European Integration Consortium - agri-migration

European Integration ConsortiumIAB, CMR, fRDB, GEP, WIFO, wiiwLabour mobility within the EU in the context ofenlargement and the functioningof the transitional arrangements- Background reports-This study has been carried out on behalf of the Employment, Social Affairs and Equal OpportunitiesDirectorate General of the European Commission (contract VC/2007/0293). The views and opinionsexpressed in this publication are those of the authors and do not necessarily represent those of theEuropean Commission.Nuremberg, 2009

Background reports1. Literature review2. Analysis of the scale, direction and structure of labour mobility3. Forecasting potential migration from the New Member States into theEU-15 - Review of Literature, Evaluation of Forecasting Methods andForecast results4. The macroeconomic consequences of labour mobility5. The impact of labour mobility on public finances and social cohesion6. Brain drain, brain gain and brain waste7. Regional effects of labour mobility

European Integration ConsortiumIAB, CMR, fRDB, GEP, WIFO, wiiwLabour mobility within the EU in the context of enlargement and the functioningof the transitional arrangementsVC/2007/0293Deliverable 1Literature reviewTito Boeri, Herbert Brücker, Anna Iara, Peter Huber, Pavel Kaczmarczyk, Richard Upwardand Hermine VidovicAbstractThis literature review is part of the study “Labour mobility within the EU in the context ofenlargement and the functioning of transitional arrangements” (VC/2007/0293). Thepurpose of this very brief review is to prove an overview on the literature relevant to thisstudy. More detailed discussions of the literature will be included in the backgroundreports to the individual deliverables. This review has two sections. In the first section weconsider the determinants and forecasts of migration following the recent EUenlargement and associated transition arrangements. In the second section we considerthe literature on the impacts of increased migration as a result of enlargement and thetransition arrangements. These impacts include the aggregate effects on wages andemployment, but also issues such as: the “brain drain”, the regional concentration ofmigrants, the effect on the welfare state and social cohesion and, finally, the impact ofremittances from migrants to their home countries.The views and opinions expressed in this publication are those of the authors and do not necessarilyrepresent those of the European Commission.

Contents1 Determinants and forecasts of migration ....................................................... 11.1 Surveys of migration intentions.............................................................. 11.2 Extrapolation studies ............................................................................ 21.3 Econometric models ............................................................................. 32 The impacts of migration............................................................................. 52.1 Impacts on wages, employment and the macro-economy.......................... 52.2 Brain drain, brain gain and brain waste................................................... 62.3 Regional impacts.................................................................................. 72.4 Public finances, the welfare state and social cohesion ............................... 82.5 Remittances ........................................................................................ 9References ...................................................................................................... 11

1 Determinants and forecasts of migrationThe substantial income gap between the incumbent EU member states and the accessioncandidates from Central and Eastern Europe has motivated numerous studies whichattempt to forecast potential migration from the new Member States (NMS) prior toenlargement. These studies have estimated the long-run stock of residents from the NMSat between 3 and 5 per cent of the population in the origin countries, while annual netmigration flows have been predicted to be between 300,000 and 400,000 persons in thefirst years following enlargement, which corresponds to 0.3-0.4 per cent of thepopulation in the countries of origin. 1These migration forecasts rely on the assumption that all Member States of the EU-15open their labour markets at the same time. However, the selective application oftransitional arrangements has affected both the scale and the direction of migration fromthe NMS. Nevertheless, at an annual net migration flow of between 200,000 and 250,000persons from the NMS-8 into the EU-15, the post-enlargement experience is not entirelyinconsistent with most of the projections, although migration flows into Ireland and theUK have greatly exceeded the forecasts.There are essentially three methods which have been used for forecasting the potentialflows of migration from the NMS. The first derives medium- and long-term migrationforecasts from surveys of migration intentions in the sending countries. The secondextrapolates the South-North migration flows in Europe during the 1960s and early1970s to future East-West migration. Finally, the third and largest part of the literaturebases migration forecasts on econometric models, which explain migration stocks andflows by economic and institutional variables. In this section we briefly outline the resultsfrom these three methods. 21.1 Surveys of migration intentionsA number of studies base forecasts of potential migration on surveys of migrationintentions in the NMS (Fassmann/Hintermann 1997; Wallace, 1998; Krieger 2003;Fassmann and Münz, 2002; see also Fouarge and Ester, 2007). Krieger (2003) is basedon the Eurobarometer Labour Mobility Survey, which covers all accession countries; theother studies are based on smaller surveys which focus only on a limited number ofcountries.Studies of migration intentions suffer from a number of problems. First, and mostimportantly, it is unclear whether or when the expressed migration intention will berealised, and if so, how long an individual will actually stay abroad. Second, the migrationintentions revealed differ a lot depending on the questionnaire and other aspects of the1 See e.g. Alvarez-Plata et al., 2003; Boeri/Brücker et al., 2001; Bruder, 2003; Hille/Straubhaar, 2001;Krieger, 2003; Layard et al., 1992; Zaiceva, 2006). Some studies have obtained however lower (Fertig,2001; Fertig/Schmidt, 2001; Dustmann et al., 2003; Pytlikova, 2007) and higher projections (Flaig 2001;Sinn et al., 2001).2 For previous literature reviews see Brücker/Siliverstovs, 2006a, 2006b; Fassmann/Münz, 2002; Hönekopp,2001; Straubhaar, 2002; Zaiceva/Zimmermann, 2007.1

survey design. Third, it is unclear whether migration intentions refer to a situationwithout legal barriers to migration or whether migration intentions reflect institutionalbarriers and are therefore a biased measure for migration under the conditions of freemovement. Many of these problems could be circumvented by panel studies which wouldallow one to show whether migration intentions are realised or not. Unfortunately, panelstudies of migration intentions do not yet exist in the NMS.However, surveys of migration intentions can provide valuable information which is notavailable from other studies. First, they deliver important insights on the human capitalcharacteristics of potential migrants (see Fouarge and Ester, 2007; Krieger, 2003, for adetailed analysis). Second, the latest Eurobarometer survey provides information on thedestination countries, which may help to analyse the spatial distribution of migrants fromthe NNS across the EU Member States.1.2 Extrapolation studiesThe extrapolation of South-North to East-West migration in Europe relies on thehypothesis that the economic and institutional conditions of “guestworker” migration inthe 1960s and early 1970s resemble migration conditions in the enlarged EU of today.Under this assumption, about 3 per cent of the population from the NMS would move tothe EU-15 within 15 years (Layard et al., 1992).The income difference measured in purchasing power parities between the EU-15 and theNMS-8 is similar to that between the members of the European Economic Community(EEC) and their neighbours in Southern Europe during the 1960s. However, there arealso important differences between the current enlargement and previous episodes. First,the present per capita GDP gap between the EU-15 and the NMS-8 at current exchangerates is substantially larger than that between the North and the South in Europe duringthe 1960s. Income differences at current exchange rates may affect migration decisionssince a part of the income obtained in host countries can be consumed in the sendingcountries. Second, labour market conditions (such as unemployment rates) in the maindestination countries in the EU-15 are generally less favourable today compared to thosein Europe during the 1960s. Third, transport and communication costs are substantiallylower today compared to the 1960s, which in turn reduces migration costs. Finally, theinstitutional and legal framework for migration was different during the guestworkerrecruitment period compared to the legal framework for the free movement of workers inthe Community of today.Thus, actual migration movements from the NMS into the EU-15 may deviate in one wayor another from the South-North migration episode in the 1960s. Nevertheless, theextrapolation exercises provide a useful hint to the magnitudes involved given that theEuropean continent faced a similar income gap in history.2

1.3 Econometric modelsThe largest part of the migration forecasts rely on econometric models, which explainmigration flows or stocks by economic and institutional variables. The key explanatoryvariables are in most models the wage and (un-)employment rates in the receiving andsending countries, the (lagged) migration stock, and a number of dummy variables,which capture institutional conditions in the destination and sending countries,particularly legal immigration barriers.Although the theoretical foundations may differ, most macro migration models areremarkably similar regarding the variables they consider and with respect to theirfunctional forms. One important difference in the literature is between stock and flowmodels, which need not however necessarily yield different estimates of the migrationpotential if properly applied. A second difference is the identifying restrictions which areimposed by different estimators. Both methodological arguments and tests of theforecasting performance suggest that standard fixed effects models outperform pooledOLS models as well as most sophisticated heterogeneous estimators.Table 1 summarises the estimation results of different studies including their data sourceand methodological foundations. The estimation results for migration stocks and flowsare expressed in per cent. This allows one to compare the findings, since the sample ofsending countries differs across the studies. 3We can distinguish studies which refer to Germany, the UK and the total EU-15 as adestination, where the latter studies are based on estimates for a panel of destinationand sending countries. The large number of studies in the literature which refer toGermany can be traced back to the fact that about 60 per cent of the immigrants fromthe NMS in the EU-15 resided in Germany before enlargement. Moreover, the Germanmigration statistics provides detailed data on migration stocks and flows by country oforigin which facilitates migration estimates compared to many other destinations in theEU-15. Many studies have therefore estimated the migration potential for Germany andthan extrapolated the estimate to the EU-15 under the counter-factual assumption thatall EU Member States will open their labour markets at the same time and that theregional distribution of migrants remains constant over time (Alvarez-Plata et al., 2003;Boeri/Brücker et al., 2001).3 Note that Table 1 is a selection of the literature. There exist numerous other studies which, by and large,resemble the findings represented in Table 1.3

Table 1 Econometric forecasts of potential migration from the NMSStudy Database Type of model Estimator Initial net inflow Long-run stockEstimates of potential immigration into Germany (extrapolations to EU-15 in parentheses)Alvarez-Plataet al. (2003)Panel of migrationstocks from 18sending countries,1967-2001Dynamic stockmodelFixed effects 0.22%(EU-15: 0.33%)2.33%(EU-15:3.82%)Boeri/Brückeret al. (2001),Brücker(2001)Panel of migrationstocks from 18sending countries,1967-1998Dynamic stockmodelFixed effects 0.22%(EU-15: 0.34%)2.53%(EU-15:3.89%)Dustmannet al. (2003)Panel of migrationflows from 18 sendingcountries, 1960-1994Static flow modelGMM withindividualeffects0.02% - 0.2% -Fertig (2001)Panel of migrationflows from 17 sendingcountries, 1960-1997Dynamic flowmodelFixed effects 0.07% -Fertig/Schmidt(2001)Panel of migrationflows from 17 sendingcountries, 1960-1997Static errorcomponentsmodelGMM 0.01% -0.06% -Flaig (2001),Sinn et al.(2001)Panel of migrationstocks from 5 sendingcountries, 1974-1997Dynamic stockmodelPooled OLS 0.64% 7.2%Estimates of potential immigration into the United KingdomDustmann etStatic flow modelal. (2003)Panel of migrationflows from 18 sendingcountries, 1960-1994GMM withindividualeffects0.004% - 0.01% -Estimates of potential immigration into the EU-15Alvarez-Plataet al. (2003)Panel of labourmigration stocks from20 sending and 15destination countries,1993-2001Dynamic stockmodelGMMsystemestimatorwithindividualeffectsEU-15:0.11% - 0.15%(labour force)EU-15:2.2% - 2.7%(labour force)Hille/Straubhaar(2001),Straubhaar(2002)Panel of migrationflows from 3 sendingand 8 destinationcountries, 1988-99Static flow model(gravity equation)Pooled OLS EU-15: 0.27% -Pytlikova(2007)Panel of gross andnet migration flowsfrom 7 NMS into 15EU/EEA countries,1990-2000Static anddynamic flowmodelFixed effectsEU/EEA-13:0.04-0.08% (net),(gross inflows:0.53-0.57)EU/EEA-13:1.5%-1.8%Zaiceva(2006)Panel of migrationflows from 3 sendingand 15 receivingcountries, 1986-1997.Static flow model(gravity equation)Fixed effects EU-15: 0.23-0.34%EU-15:3.5%-5.0%4

Altogether, the estimates of these studies are by and large consistent with the migrationdevelopment from the NMS-8 since enlargement. The annual net inflow or growth in thestock of the foreign residents from the NMS-8 can be estimated at about 200,000 p.a. in2004, and at about 250,000 p.a. in 2006 and 2007. 4 This corresponds to between 0.27and 0.34 per cent of the population in the NMS-8.The regional structure of migration across the EU, however, is very different from thatbefore the EU enlargement as a result of the selective application of the transitionalarrangements. Hence, those studies which have used the regional distribution ofmigrants before enlargement to extrapolate future estimates for a particular country tendto be less accurate. In particular, actual migration into the UK and Ireland has beenmuch larger, while actual migration to Austria and Germany has been much lower thanprojected. Actual migration inflows into the UK have been 100,000–150,000 p.a. largerthan the net flows predicted in the Dustmann et al. (2003) study for the UK (4,000-13,000). The flows to the Scandinavian countries have been at or below the predictedlevels.Future estimates of the migration potential from the NMS have to consider the thirdcountryeffects which arise because of the transitional arrangements, since migrationrestrictions in one country may affect the scale of migration in other countries. Thereexist meanwhile three annual observations since enlargement for migration from theNMS-8 into the EU-15 which can be exploited for an identification of the role of migrationrestrictions. However, counterfactual experience from migration under free movement isnot available for the NMS-8 sample. A possible way to circumvent this problem is to usethe migration data from other sending countries, which perhaps allows estimating‘normal’ regional migration patterns based on gravity-type migration equations.2 The impacts of migration2.1 Impacts on wages, employment and the macro-economyThe analysis of the macroeconomic impact of migration is typically based on generalequilibrium trade models. This type of macroeconomic modelling is very flexible andprovides a comprehensive framework which facilitates the analysis of the interactionbetween trade, migration and capital movements and their subsequent labour marketimpacts.The effect of migration on wages and unemployment depends largely on the skillcomposition of immigrants. Assuming that the low-skilled and high-skilled labour force inAustria would increase by respectively 10.5 and 2.1 per cent, Keuschnigg and Kohler(1999) estimate a 5% decrease in wages for low-skilled workers. Heijdra et al. (2002)estimate the effect of the migration from the NMS to Germany. They assume thatmigration from eastern European countries to Germany would rise from 550,000 in 20084 This estimate refers to the change in the number of foreign residents as reported by national statisticalsources in the Member States of the EU-15 and, for those countries which do not report, by the EurostatLabour Force Survey. The 2007 figures refer to summer 2007 and are very preliminary.5

to 2.5 million in 2030, with 35% of the migrant population entering the labour market.40% of the migrants are skilled and 60% unskilled. As a result, less skilled workerssuffer from reduced wages and higher unemployment, while skilled labour benefits frommigration through higher wages and lower unemployment. Brücker and Kohlhaas (2004)find that, depending on the assumptions on the qualifications of the migrant population,wages can decline by 0.5–0.6% for an immigration rate of 1% of the labour force, whilethe unemployment rate increases by 0.02–0.1 percentage points. Brücker (2007)demonstrates that if 4 % of the population from the NMS migrate into the EU-15, themain winners of migration are the migrants themselves, while blue-collar workers arenegatively affected through higher unemployment in the destination countries.However, these relatively modest negative effects of migration on wages andunemployment of the low-skilled are likely to be outweighed by positive and strongeffects of a more liberalized goods market (e.g. Brown et al. 1995, Baldwin et al. 1997).This is why most models predict that Eastern enlargement results in lower aggregateunemployment and higher wages in both the EU-15 and the NMS. Among other things, asignificant trade creation between the NMS and EU-15 is likely to affect negatively theeconomies of China, Japan, Korea and the NAFTA countries, but to have a positive impacton the economies of Russia and other members of the CIS (Ko 2006).In all models, the EU enlargement leads to a higher level of GDP. In earlier studies, thiseffect was predicted to vary between 0.1% and 0.5% in the EU-15, and between 5% and18% in the NMS. More recent studies, which take into account trade creation betweenthe old and new member countries, estimate slightly larger effects on GDP of the EU-15.Boeri and Brücker (2005) estimate a 0.5% gain in income per capita if 3% of thepopulation from the NMS migrate into the EU-15. However, these aggregate and percapita income gains may be reduced if rigidities in the labour market exist. Finally,analysing possible diversion effects due to transitional periods, Baas and Brücker (2008)conclude that the closure of labour markets in Germany has reduced the GDP effectthere, while the opening-up of the UK has led to higher GDP in that country.2.2 Brain drain, brain gain and brain wasteThe mobility of high-skilled labour has become one of the most prominent issues in therecent migration debate. The stock of high-skilled immigrants living in OECD countriesincreased from 12.4 million in 1990 to 20.4 million in 2000, whereas the total number ofworking-aged immigrants increased from 42 million to 59 million over the same period(Docquier and Marfouk (2004)). There is a general agreement that there has been asignificant outflow of the highly skilled from Central and Eastern Europe, particularly fromthe Baltic countries, Poland and countries of the Western Balkan (Inzelt 2003;Kaczmarczyk and Okólski 2005; Okólski 2006; Ribickis 2003).From the theoretical point of view, the push-pull framework is commonly used to explainmigration of the high-skilled. Factors contributing to the outflow of qualified individualsinclude: differences in wages, living standards, working conditions, chances forprofessional advancement and so on. In addition, the scale and structure of migration6

flows is largely affected by immigration policies of receiving states which often targetskilled (and young) people.The consequences of the mobility of high-skilled labour are far more controversial.Traditionally, the economic approach analyzing the consequences of the “brain drain” isbased on the assumption that a country’s competitiveness and economic growth shouldbe directly linked to the stock of human capital. Consequently, in a zero-sum game,sending countries are assumed to be losers and receiving countries winners from skilledmigration. (Grubel and Scott 1969, Berry and Soligo 1969, Bhagwati and Dellafar 1973,Bhagwati 1976, 1979).Recent theoretical and empirical contributions have emphasized more positive or “braingain” effects of high-skilled migration on sending countries. First, the outflow of highskilledlabour may have a positive impact on human capital formation by creatingincentives to invest in education and influencing propensity to acquire higher skills (e.g.Stark 2004, 2005, Mountford 1997, Beine, Docquier and Rapoport 2001). Second,remittances sent by migrants significantly reduce poverty rates in the countries of origin.Third, migrants who accumulated human and financial capital abroad contribute to higherproductivity at home upon their return (Klagge et al. 2007). Fourth, a skilled diasporacan facilitate technology transfer and bridge information gaps between potentialexporters, importers and investors at home and host countries (Docquier et al 2007).Finally, Beine, Docquier and Rapoport (2003) show that the mobility of high-skilledlabour can reduce the level of discrimination and corruption in sending countries.Another important phenomenon related to the migration of high-skilled labour isoccupational over-qualification, or the “brain waste”. In practically all OECD countriesimmigrants are more likely to be overqualified than the native born (OECD 2006). This“brain waste” is usually attributable to unobserved differences in the value of education,problems with the recognition of qualifications acquired in the home country, a lack ofhuman and social capital specific to the host country (such as language proficiency), thelocal labour market situation or, finally, various forms of discrimination. Women, recentimmigrants, and migrants from outside the OECD are typically found to have a higherprobability of being over-qualified. The problem of “brain waste” might also be relevantfor the post-enlargement migrants. While 80-90% of migrants from the NMS-8 countriesare hired for occupations that need no professional qualifications (WRS data), nationalsources suggest that, among Polish migrants in the UK, the share holding a universitydegree exceeds 25-30%. Nevertheless, relatively high rates of emigration of the skilled inPoland and other NMS countries might also be a consequence of the over-supply ofqualified individuals and inefficient use of human capital resources at home countries.2.3 Regional impactsMigrate or commute?The choice between migration or cross-border commuting has recently been examined bya number of authors (Zax, 1994, Rouwendal, 1998, Van Ommeren, Rietveld and7

Nijkamp, 2000). Such research has concluded that commuting may be of sufficientimportance that cross-border labour markets may be emerging (Overman and Puga,2000), with regional linkages in unemployment rates being equally strong across nationalborders as within countries. This has led policy makers to argue that cross-bordercommuting flows should be considered in addition to migration when considering thepotential derogation periods following the EU enlargement in 2004 (see Huber 2001,Untiedt and Alecke 2001).What affects the region of migration?A robust stylised fact from the empirical literature is that migrants are highly regionallyconcentrated. Bartel (1989) shows that close to 75% of migrants live in the 25 largestSMAS of the United States, although only 50% of the native population resides in theseregions. Similar concentrations occur in other countries (e.g. Huber (2002) for Austria;Chiswick, Lee and Miller (2002) for Australia; Edin, Fredrikson and Aslund (2001) forSweden; Hou (2005) for Canada).There are a variety of alternative explanations for such regional concentrations. Firstly,network effects, whereby migrants take advantages of positive externalities inconsumption and information arising from living in proximity to each other (Winters, deJanvry and Sadoulet 2001, Bauer, Epstein and Gang, 2002a). Secondly, herd effects,whereby informational advantages in the home country lead to the spatial clustering (seeEpstein, 2002, Bauer Epstein and Gang, 2002b). Finally, welfare magnet effects, wheremigrants gravitate to regions in which the welfare system is most generous (see Borjas,1999, Levine and Zimmermann, 1995).2.4 Public finances, the welfare state and social cohesionStudies of the US suggest that the participation of immigrants in welfare programs hasincreased in the last thirty years such that immigrants now have a higher probability ofreceiving public assistance than U.S. natives. When considering non-cash benefits, thedifference is even larger. The reasons behind this growth are still debated but appear toinclude changes in the demographic characteristics of immigrants, quicker assimilation ofthe immigrants to the welfare system (Borjas and Trejo, 1991; Borjas and Hilton, 1996),and increases in the number of refugees who have a higher welfare dependency (Borjas1995).A number of studies have examined whether such an increase might also be due to a“welfare magnet” effect, whereby immigrants are attracted by generous welfare benefits.For the US, Blank (1988), Borjas (1999), Gramlich et al. (1984), Meyer (2000),McKinnish (2005), and Gelbach (2004) find a significant impact of the generosity ofwelfare provision on the location decision of migrants.The welfare magnet literature is less developed in Europe. Hansen and Lofstrom (1999,2001, 2003), suggest that immigrants are more likely to receive both unemploymentbenefits and social assistance than natives, although this is largely due to differences in8

the characteristics of migrants vis a vis natives. Such results have been replicated instudies for Germany, where foreigners are less likely to depend on welfare than nativesonce observable characteristics are controlled for (Bird et al. 1999; Fertig and Schmidt,2001; Frick et al. 1996; Riphahn, 1998; Sinn et al., 2001) and Denmark (Riphahn andRosholm, 2001).The European studies have also found that, in contrast to evidence from the US (e.g.Borjas and Hilton, 1996), foreigners tend to assimilate out of welfare assistance. InGermany, welfare dependency declines with the duration of stay of migrant households,all other factors being equal (Fertig and Schmidt, 2001). Further, welfare usage of nonhumanitarianmigrants is well below that of humanitarian migrants and non-humanitarianmigrants tend to assimilate even more rapidly out of welfare (Hansen and Lofstrom,1999).2.5 RemittancesMigrant remittances are income earned in a host country that is sent or brought to thehome country. These include: earnings paid to migrant employees who are not residentin the host country, such as border and seasonal workers; transfers abroad by residentworkers; and cash and goods transferred by re-migrating individuals. 5The motives for remittance payments may be manifold; first, individuals may seek tosupport family consumption with their income from abroad; second, migrants may seekto accumulate savings that will be invested at home; third, remittances may protectagainst negative income shocks to the family in the home country. For bilateral flowsfrom the EU to neighbouring countries, Schiopu and Siegfried (2006) present evidence insupport of the consumption rather than the investment motive. They also show thatremittances are increasing with the skill level of the remitters. Other authors have showna role for unemployment in the home country (Dragutinovic Mitrovic and Jovicic (2006)and Schrooten (2005)). Schrooten (2005) also suggests that remittances diminish withincreasing economic development, economic openness and the development of thebanking sector in the home country.Effects of remittancesAlthough remittances serve to boost the household income of the recipients, and mayprovide a means out of poverty, remittances may have a negative impact on economicactivity in the home country if they result in the reduction of labour supply. 6 Largeremittance inflows may also result in the appreciation of the real exchange rate,comparable to the Dutch disease effect, and therefore diminish the competitiveness of5 Recently, a number of reviews of the existing findings on remittances were produced (Ghosh, 2006, Mansoorand Quillin, 2006, OECD, 2006, Page and Plaza, 2006, Ratha and Shaw, 2006). Below, we will focus on theexperience of central and eastern Europe and the western Balkans and most recent findings for otherEuropean economies.6 See Adams and Page (2003) on the poverty-alleviating effect of remittances and Chami et al. (2003) on thepossible negative effect on labour supply.9

the home country (Mansoor and Quillin, 2006). Dragutinovic Mitrovic and Jovicic (2006)also raise the prospect that countries may become dependent on remittances to financetrade deficits.The empirical evidence using macroeconomic datasets is controversial. Although Chamiet al. (2003) suggest a negative impact of remittances on economic growth, Mansoor andQuillin (2006) and Catrinescu et al. (2006) find that both the short- and the long-runimpact of remittances on economic growth appears to be small but positive.10

ReferencesAdams, Richard, and John Page, 2003, “International migration, remittances and povertyin developing countries”, World Bank Policy Research working paper no. 3179.Alvarez-Plata, P., Brücker, H. and Siliverstovs, B. (2003), Potential Migration fromCentral and Eastern Europe into the EU-15 – An Update, Report for the EuropeanCommission, DG Employment, Social Affairs and Equal Opportunities, Brussels.Baas, T. and Brücker, H. (2008) “Macroeconomic Impact of Eastern Enlargement onGermany and UK: Evidence from a CGE Model.” Applied Economics Letters,forthcoming.Baldwin, R. Francois, J. and Portes, R. (1997) “The Costs and Benefits of EasternEnlargement: The Impact on the EU and Central Europe” Economic Policy: AEuropean Forum, 12(24), pp. 125-70.Bartel Ann B (1989) Where Do New US Migrants Live? Journal of Labour Economics, Vol.7, No. 4, pp 371 – 391.Bauer, Thomas, Gil S. Epstein and Ira N. Gang (2002a) What are migration Networks,IZA Discussion Paper 200, BonnBauer, Thomas, Gil S. Epstein and Ira N. Gang (2002b) Herd effects or MigrationNetworks? The location Choice of Mexican Immigrants in the US, IZA DiscussionPaper 551Beine, M. ; F. Docquier and H. Rapoport (2001) “Brain drain and economic growth:theory and evidence” Journal of Development Economics, Volume 64, 1, pp 275-289Beine, M., F. Docquier and H. Rapoport (2003): Brain drain and growth in LDCs: winnersand losers, IZA DP no. 819.Berry, R. A. and Soligo, R. (1969), Some welfare effects of international migrationJournal of Political Economy, 77, 778-794Bhagwati, J. (1979) International Migration of Highly skilled: economics, Ethics andTaxes, Third Word Quarterly 1 (3), 17-30.Bhagwati, J. (ed.), 1976, The Brain Drain and Taxation II: Theory and empirical analysis.Amsterdam: North Holland.Bhagwati, J. N. and W. Dellafar, (1973), The Brain Drain and Income taxation WorldDevelopment, 1 94-100.Bird, E.J., H. Kayser and J.R. Frick, (1999), The Immigrant Welfare Effect, Take-up orEligibility? IZA Discussion Paper 66, IZA, Bonn.Blank, R.M. (1988), “The Effect of Welfare and Wage Levels on the Location Decisions ofFemale-Headed Households.” Journal of Urban Economics, 24(2), pp. 186-211.Boeri, T. and Brücker, H. (2005) “Why Are Europeans So Tough on Migrants?” EconomicPolicy, 44, pp. 629.Boeri, T., H. Brücker et al. (2001), The Impact of Eastern Enlargement on Employmentand Labour Markets in the EU Member States, European Commission,Berlin/Brussels/Milano.Borjas, G. (1999), “Immigration and Welfare Magnets.”, Journal of Labour Economics,17(4), pp. 607-637.Borjas, G. J. (1995), Immigration and Welfare, Research in Labor Economics, 14, pp.253-282.Borjas, G. J. and S. J. Trejo, (1991), Immigrant Participation in the Welfare System,Industrial and Labor Relations Review, 44, pp. 195-211.11

Borjas, G. J., and L. Hilton, (1996), Immigration and the Welfare State: ImmigrantParticipation in Means-Tested Entitlement Programs, Quarterly Journal ofEconomics, 111(2), pp. 575-604.Brown, Drusilla, K.; Deardorff, Alan, V.; Djankov, Simeon and Stern, R. (1995) “AnEconomic Assessment of the Integration of Czechoslovakia, Hungary and Polandinto the European Union,” American Institute for Contemporary German Studies.Brücker, H. (2001), “Die Folgen der Freizügigkeit für die Ost-West-Migration.Schlussfolgerungen aus einer Zeitreihenanalyse der Migration nach Deutschland”,Konjunkturpolitik- Applied Economics Quarterly, Supplement 52, 2001, pp. 17-54.Brücker, H. (2007) “Labor Mobility after the European Union's Eastern Enlargement: WhoWins, Who Loses?” GMF Paper Series.Brücker, H. and Kohlhaas, M. (2004) “Migration, Qualification and Labor Markets,”mimeo. Berlin: German Institute for Economic Research (DIW Berlin).Brücker, H. and Siliverstovs, B. (2006b), “On the Estimation and Forecasting ofInternational Migration: How Relevant Is Heterogeneity Across Countries?”,Empirical Economics.Brücker, H., and Siliverstovs, B. (2006a), “Estimating and forecasting Europeanmigration: methods, problems and results”, Zeitschrift für Arbeitsmarktforschung– Journal for Labour Market Research, 39(1), 35-53.Bruder, J. (2003), East-West Migration 2004-2015, Thünen Series of Applied EconomicsWorking Papers 40, University of Rostock.Catrinescu, Natalia, Miguel León-Ledesma, Matloob Piracha, and Bryce Quillin, 2006,“Remittances, institutions and economic growth”, IZA discussion paper no. 2139.Chami, Ralph, Connell Fullenkamp, and Samir Jahjah, 2003, “Are immigrant remittanceflows a source of capital for development?”, IMF working paper no. 03/189.Chiswick, Barry R., Yew Liang Lee and Paul W. Miller (2002) The determination of theGeographic Concentration among Immigrants: Application to Australia, IZA-Discussion Paper No. 462,Docquier, F. and Marfouk, A. (2004): Measuring the international mobility of skilledworkers - Release 1.0, Policy Research Working Paper n. 3382, World Bank.Docquier, F., O. Lohest, and A. Marfouk (2007): “Brain drain in developing countries”World Bank Economic Review 21: 193 - 218.Dragutinovic Mitrovic, R., and Jovicic, M. (2006), “Macroeconomic analysis of causes andeffects of remittances: a panel model of the SEE countries and a case study ofSerbia, GDN Southeast Europe working paper.Dustmann, C., Casanova, M, Fertig, M, Preston, I and Schmidt, C. (2003), The impact ofEU enlargement on migration flows, Home Office Online Report 25/03.Edin, Per-Anders, Peter Fredrikson and Olof Alsund (2001) Ethnic Enclaves and theEconomic Success of Immigrants – Evidence from a natural Experiment, CEPRDiscussion Paper No 2370Epstein, Gil S (2002) Informational Cascades and Decision to Migrate, IZA Working Paper445.Fassmann, R., and Hintermann, C. (1997), “Migrationspotential Ostmitteleuropa.Struktur und Motivation potentieller Migranten aus Polen, der Slowakei,Tschechien und Ungarn”, ISR-Scientific Report 15, Innsbruck.Fassmann, R., and Münz, R. (2002), EU-Enlargement and the future of East-Westmigration, in: IOM (ed.), New challenges for migration policy in Central andEastern Europe, Geneva, IOM.Fertig, M. (2001), The Economic Impact of EU Enlargement: Assessing the MigrationPotential, Empirical Economics, pp. 707-720.12

Fertig, M. and Schmidt, C. M., (2001), “First- and Second-Generation Migrants inGermany – What Do We Know and What Do People Think”, in Rotte R. (ed.)Migration Policy and the Economy- International Experience, mimeo.Fertig, M., and Schmidt, C. (2001), Aggregate Level Migration Studies as a Tool forForecasting Migration Streams, IZA Discussion Paper 183.Flaig, G. (2001), Die Abschätzung der Migrationspotenziale der osteuropäischen EU-Beitrittsländer. In: Konjunkturpolitik – Applied Economics Quarterly, Supplement52m pp. 55-76.Fouarge, D., and Ester, P. (2007), Mapping migration intentions in Europe: Perceivedcosts and benefits, Institute for Labour Studies, Tilburg University.Frick, J., F. Büchel and W. Voges, (1996), Sozialhilfe als Integrationshilfe für Zuwandererin Deutschland, DIW-Wochenbericht, 63 (48), 767-775.Gelbach, J.B. (2004), “Migration, the Life Cycle, and State Benefits: How Low Is theGhosh, Bimal, 2006, Migrants’ remittances and development: myths, rhetoric andrealities, Geneva: International Organization for Migration.Gramlich, E.M. and Laren D.S. (1984), "Migration and Income RedistributionResponsibilities." Journal of Human Resources, 19(4), pp. 489-511.Grubel, H G and Scott, A. (1976), “The immigration of scientists and engineers to theUnited States, 1949-61”, Journal of Political Economy, 74, 368-378.Hansen, J. and Lofstrom, M. (2001), “The dynamics of immigrant welfare and labormarket behaviour”, IZA (Institute for the Study of Labor) Discussion Paper No.360, September.Hansen, J. and Lofstrom, M. (2003), “Immigrant assimilation and welfare participation:do immigrants assimilate into or out of welfare?”, Journal of Human Resources, 38(1), Winter, pp. 74-98.Hansen, J. and M. Lofstrom, (1999), Immigration and Welfare Participation: DoImmigrants Assimilate Into or Out-of Welfare? IZA Discussion Paper 100, IZA,Bonn.Heijdra, B. Keuschnigg, C. and Kohler, W. (2002) “Eastern Enlargement of the EU: Jobs,Investment and Welfare in Present Member Countries,” CESifo GmbH, CESifoWorking Paper Series: CESifo Working Paper No. 718.Hille, H. and Straubhaar, T. (2001), “The Impact of EU-Enlargement on MigrationMovements and Economic Integration: Results of Recent Studies”, in: OECD (ed.),Migration Policies and EU-Enlargement, Paris, pp. 79-100.Hönekopp, E. (2001): “Überblick über die Ergebnisse bisher vorliegender Schätzungenzum Migrationspotential im Falle einer Arbeitskräftefreizügigkeit im Rahmen derOsterweiterung der EU”, Vortrag auf der Fachkonferenz “Mehr Konkurrenz umLohn und Arbeit? Auswirkungen der Osterweiterung der EU auf die Arbeitsmärkte– Optionen für die Politik”, Akademie für Politische Bildung Tutzing, Tutzing, 14.-16.5.Hou, Feng (2005) The Initial Destinations and Redistribution of Canada's MajorImmigrant Groups: Changes over the Past Two Decades, Manuscript, StatisticsCanadaHuber, Peter (2001) Auswirkungen der EU Erweiterung auf die Erwerbstätigkeit(Migration, Pendelwanderung), Perparity Project report, WIFO, ViennaHuber, Peter (2002) Migration and Regions – The case of Austria, in Ralph RotteMigration Policy and the Economy: International Experiences, ars et unitas,Neuried, pp157 – 178.Inzelt, A. (2003). Hungary, in: Brain-drain emigration flows for qualified scientists part 5.Flows and non-EU Europe, University of Maastricht: MERIT.13

Kaczmarczyk P., and Okólski M. (2005). Migracje specjalistow wysokiej klasy wkontekscie czlonkostwa Polski w Unii Europejskiej [Migration of the highly skilledin the context of Poland’s accession to EU], UKIE, Warsaw.Keuschnigg, C. and Kohler, W. (1999) “Eastern Enlargement to the EU: Economic Costsand Benefits for the EU Present Member States? The Case of Austria. Study forthe European Commission. Final Report”, Linz: Johannes Kepler Universität, Linz.Klagge, B. et al. (2007). High-skilled return migration and knowledge-based economicdevelopment in regional perspective. Conceptual considerations and the exampleof Poland, CMR Working Paper 19/77.Ko, Jong-Hwan (2006) “Eastern Enlargement of the EU and Its Economic Impact: A CGEApproach.” The Kyoto Economic Review, 75(1), pp. 91-113.Krieger, T. (2003), Migration trends in an enlarged EU, European Foundation for theImprovement of Working and Living Conditions, Dublin.Layard, R., Blanchard, O., Dornbusch, R. and Krugman, P. (1992): East-West Migration:The Alternatives. Boston: MIT Press.Levine, Phillip B. and David J. Zimmermann (1995) An empirical Analysis of the Welfaremagnet Debate using the NLSY. NBER Working Paper 5264Mansoor, Ali, and Bryce Quillin (ed.), 2006, Migration and remittances: eastern Europeand the former Soviet Union, Washington: The World Bank.McKinnish, T. (2005), “Importing the Poor: Welfare Magnetism and Cross-BorderMigration.”, Journal of Human Resources, 40(1), pp. 57-76.Meyer, B.D. (2000), “Do the Poor Move to Receive Higher Welfare Benefits?”, Workingpaper no. 58. Northwestern University, Joint Center for Poverty Research WorkingPaper.Mountford, A. (1997) Can a brain drain be good for growth in the source economy?Journal of Development Economics, Volume 53, 2, pp. 287-303OECD (2006). International Migration Outlook 2006. OECD:Paris.Okólski M. (2006). Costs and benefits of migration for Central European countries, CMRWorking Paper 7(65).Organisation for Economic Co-operation and Development (OECD), 2006, Internationalmigrant remittances and their role in development, in: ibid., InternationalMigration Outlook 2006, p. 139-161.Overman, Henry G; and Diego Puga, (2002) Unemployment Clusters across Europe'sRegions and Countries, Economic Policy: A European Forum, no. 34, pp. 115-43Page, John, and Sonia Plaza, 2006, “Migration remittances and development: a review ofglobal evidence”, Journal of African Economies 15, supplement 2, p. 245-336.Pytlikova, M. (2007), EU Enlargement: Migration from the New Member States, Mimeo,Aarhus School of Business, CIM and Department of Economics.Ratha, Dilip and William Shaw, 2006, Global economic prospects: economic implicationsof remittances and migration, Washington: The World Bank.Ribickis, L. (2003). Lithuania, in: Brain-drain emigration flows for qualified scientists part5. Flows and non-EU Europe, University of Maastricht: MERIT.Riphahn R. T. and Rosholm, M. (2001), “Immigrants' Time to Economic Independence:The Duration of Initial Public Transfer Reliance,”, mimeo.Rouwendal, Jan (1998) Search Theory, Spatial Labor Markets, and Commuting, Journalof Urban Economics, 43:1, 1-22Schiopu, Ioana, and Nikolaus Siegfried, 2006, “Determinants of workers’ remittances:evidence from the European neighbouring region”, ECB working paper no. 688.14

Schrooten, Mechthild, 2005, “Bringing home the money – what determines workers’remittances to transition countries?”, working paper, mimeo.Sinn, H.-W., Flaig, G., Werding, M., Münz, S., Düll, N.and Hoffmann, H. (2001), EU-Erweiterung und Arbeitskräftemigration. Wege zu einer schrittweisen Annäherungder Arbeitsmärkte, Munich: Ifo-Institut für Wirtschaftsforschung.Stark O. (2005). “The new economics of the brain drain”, World Economics, 2.Stark, O (2004) Rethinking the Brain Drain. World Development, Volume 32, 1, Pages15-22Straubhaar, T. (2002), Ost-West Migrationspotenzial: Wir groß ist es?, Jahrbücher fürNationalökonomie und Statistik, Vol. 222, pp. 20-41.Untiedt, Gerhard and Björn Alecke (2001) Pendlerpotential in den Grenzregionen an derEU-Außengrenze. Methoden, Ergebnisse und Prognosen, Preparity Project Report,Vienna, 2001.Van Ommeren, Jos; Rietveld, Piet; Nijkamp, Peter (2000) Job Mobility, ResidentialMobility and Commuting: A Theoretical Analysis Using Search Theory, Annals ofRegional Science, 34: 2, 213-32Wallace, C. (1998), “Migration Potential in Central and Eastern Europe”, IOM CooperationCentre for Europe and Central Asia, Geneva.Winters, Paul; de Janvry, Alain; Sadoulet, Elisabeth (2001) Family and CommunityNetworks in Mexico-U.S. Migration Journal of Human Resources, vol. 36, no. 1, ,pp. 159-84Zaiceva, A. (2006), “Reconciling the Estimates of Potential Migration into the EnlargedEuropean Union”, IZA Discussion Paper 2519.Zaiceva, A., and Zimmermann, K. (2007), “Migration Potential and its Labor MarketImpact after EU Enlargement”, Paper presented at the Second IZA Workshop: EUEnlargement and the Labor Market, September.Zax, Jeffrey S (1994) When Is a Move a Migration? Regional Science and UrbanEconomics, 24:3, 341-6015

European Integration ConsortiumIAB, CMR, fRDB, GEP, WIFO, wiiwLabour mobility within the EU in the context of enlargement and the functioningof the transitional arrangementsVC/2007/0293Deliverable 2Institute for Employment Research (IAB)Analysis of the scale, direction and structure of labour mobilityHerbert Brücker and Andreas DamelangAbstractThis background report is part of the study “Labour mobility within the EU in the contextof enlargement and the functioning of transitional arrangements” (VC/2007/0293). Theobjective of this report is to provide an overview on the main patterns of labour migrationin the context of the EU Eastern enlargement and on the fundamental economic forceswhich cause these patterns.The income gap between the EU-15 and the new member states from Central andEastern Europe is larger than in previous enlargement rounds. In particular, the nominalgap in per capita GNI and wage levels is high, reflecting poor capital endowments andparticularly large productivity differences in the tradable sectors. Although economicincentives to migrate are considerable at present, we also observe a fast convergence ofper capita GDP and wage levels which mitigate migration incentives over time.Particularly wages have converged very fast since the enlargement. Convergencebetween the EU-15 and the new member states is faster than convergence between theEU-15 and the candidate countries in South-Eastern Europe. Transport costs havedeclined since enlargement and depend less on geographical distance. This may be oneof the reasons for the fast shift in migration away from destinations neighbouring thenew member states toward destinations such as Ireland, the UK and Spain.The stock of foreign residents from the NMS-8 in the EU-15 has increased from 893,000persons in 2003 to about 1.91 million persons in 2007, or by 254,000 persons p.a. Thenumber of foreign residents from Bulgaria and Romania has increased from 702,000 toabout 1.86 million persons during the same period of time, or by 290,000 persons p.a.This increase in migration is associated with a shift in the regional structure of migration,i.e. away from Austria and Germany towards Ireland and the UK in case of migrants fromthe NMS-8, and towards Spain and Italy in case of migrants from Bulgaria and Romania.Migrants from the NMS are highly concentrated at the medium level of the skill spectrum,i.e. in the group with a vocational training degree. They are highly concentrated in theyoung age groups. The unemployment risk of migrant workers from the NMS is slightlyhigher than that of the native labour force in the EU-15 on average, but below that of themain other foreigner groups in the EU.The views and opinions expressed in this publication are those of the authors and do not necessarilyrepresent those of the European Commission.

Contents1 Introduction ...................................................................................................... 12 The economic incentives to migrate...................................................................... 22.1 The income gap in the enlarged EU............................................................... 22.2 Human capital investment ........................................................................... 42.3 Convergence of GDP per capita and wage levels ............................................. 62.4 Convergence of labour market conditions ...................................................... 82.5 The eroding role of distance......................................................................... 92.6 Concluding remarks.................................................................................. 113 The scale of labour mobility............................................................................... 123.1 Definitions and data restrictions ................................................................. 123.2 Immigration from the NMS-8 into the EU and EEA ........................................ 143.3 Immigration from the NMS-2 into the EU and EEA ........................................ 173.4 Immigration from the candidate countries into the EU and EEA ...................... 193.5 Main emigration trends from a sending country perspective ........................... 224 The structure of migration: Skills, age and gender ............................................... 244.1 Skill structure .......................................................................................... 254.1.1 Skill structure of immigrants from the NMS-8 .................................. 254.1.2 Skill structure of immigrants from Bulgaria and Romania .................. 274.1.3 Skill structure of immigrants from the candidate countries ................ 284.2 Does Eastern enlargement involve a brain drain? ......................................... 294.3 Is there evidence for brain waste? .............................................................. 304.4 Changing the age structure of the workforce................................................ 314.5 Gender patterns ....................................................................................... 335 Unemployment and labour market participation ................................................... 346 Conclusions..................................................................................................... 377 References ...................................................................................................... 408 Annex............................................................................................................. 41

1 IntroductionThis background report provides an overview on the key trends in labour mobility in theenlarged European Union (EU) and on the fundamental economic forces which affectthese mobility patterns in one way or another. The description of mobility patternspresented here serves as a starting point for the further analysis which is carried out inthe later sections of this study.Throughout the analysis, we distinguish four groups of countries: The first group containsthe fifteen EU member states which belonged to the Community before May 2004 (EU-15), the second group includes the eight new member states from Central and EasternEurope (NMS-8) 1 which joined at the 1 st of May, 2004. The third group consist of Bulgariaand Romania (NMS-2), which acceded in 2007, and the final group comprises the sixcandidate and potential candidate countries from South-Eastern Europe (CAND-6). 2Our analysis starts with a presentation of the main economic forces which affect labourmobility within the enlarged EU and between the EU and the candidate countries fromSouth-Eastern Europe. Migration theories state that migration decisions are driven byexpectations on income levels in the relative destinations and the social and economiccosts of migration. We therefore examine the present gap in per capita income levels andthe convergence of income levels which may affect expectations on future developments.Finally, we analyse new patterns of transport costs which arise from the emergence oflow-cost carriers in air transport. As a consequence, geographical proximity may loose itsimportant role in determining geographical migration patterns in Europe (Section 2).Section 3 presents the main migration trends in the enlarged EU and between the EU andthe candidate countries. Based on the available data from population statistics andLabour Force Survey (LFS) data we present the development of migration stocks in theenlarged EU from the NMS-8, NMS-2 and the candidate countries both from a receivingand sending country perspective. In the next step we analyse the skill, age and genderpatterns of migration from the NMS and the candidate countries (Section 4). Finally, weanalyse the labour market performance of the migrant communities from the NMS withinthe EU-15 based on standard indicators on unemployment and labour marketparticipation (Section 5).The analysis presented here – as any other analysis on migration patterns in Europe – ishampered by several shortcomings in the available data. In particular, only a minority ofthe EU member states report data on the stocks and flows of migrants by country oforigin in their population statistics. This concerns also destinations particularly relevant inthe context of the EU’s eastern enlargement such as Ireland and the UK. Large parts of1 Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovak Republic, Slovenia.2 Albania, Bosnia-Herzegovina, Croatia, the former Yugoslav Republic of Macedonia, Serbia andMontenegro, Turkey.IAB 1

our analysis rely therefore on LFS data, which may especially underreport migrants fromsmall countries. Moreover, the concept of nationality differs in the EU member states,such that the available data are not entirely comparable across countries. Finally, illegalmigration is not reported in official data by definition and reliable estimates for the scaleof illegal migrants from the NMS and the candidate countries do not exist. We thereforedo not cover illegal migration in our analysis. Nevertheless, the available data sourcesenables us to sketch a picture on the scale of labour mobility and migration in the contextof EU enlargement, as well as a picture on the skill and age structure of migration. Thereader should however consider the caveats which result from the shortcomings of theavailable data.2 The economic incentives to migrateMigration theory suggests that monetary and non-monetary arguments affect migrationdecisions (Sjaastadt, 1962; Stark, 1991). Individuals form expectations on income levelsat different destinations which are determined by the respective wage levels andemployment opportunities (Harris/Todaro, 1970). Moreover, since migration involvessunk costs, expectations on the future development of wages and employmentopportunities are relevant (Burda, 1995). If migrants are heterogeneous with regard totheir preferences or productivity, an equilibrium stock of migrants emerges eventuallywhich is determined by the differences in income levels, labour market conditions andother factors which affect the benefits and costs from migration (Brücker/Schröder,2006). We therefore describe here the current income gap within the enlarged EU andbetween the EU and the candidate countries as a starting point (Section 2.1). In the nextstep we analyse the factor endowments in the sending countries, particularly theendowments with human capital, since this may provide a first hint regarding themigration potential by skill levels (Section 2.2). On this basis we analyse the convergenceof per capita GDP and wage levels in the EU (Section 2.3) and the convergence ofemployment opportunities (Section 2.4). Finally, we discuss the implications of newpatterns of transport costs for the geographical structure of migration in the enlarged EU(Section 2.5).2.1 The income gap in the enlarged EUThe income gap between the EU-15 and the new member states from Central andEastern Europe creates substantial monetary incentives for labour mobility. Measured atpurchasing power parity standards (PPS), Eurostat (2008) estimates the GNI per capita inthe ten new member states from Central and Eastern Europe (NMS-10) at 48 per cent ofthat in the EU-15 in 2007. The GNI per capita of the eight new member states (NMS-8)which joined the EU in 2004 amounted to 53 per cent at PPS in 2007, and that ofBulgaria and Romania to about 34 per cent of that in the EU-15 at the same time. ThePPS estimate of the per capita GNI of the candidate and potential candidate countries byEurostat amounted to 38 per cent of the respective level in the EU-15, such that theIAB 2

income gap between the EU-15 and the NMS-2 resembles roughly that between the EU-15 and the candidate countries.Purchasing power parity estimates tend to understate monetary incentives for labourmobility, since migrants can consume a part of their earnings in their home countries orremit a part of the income to their families. Consequently, differences in earnings atcurrent exchange rates may affect migration decisions as well. At current exchange rates,the GNI per capita of the NMS-10 amounted to slightly more than one quarter of that inthe EU-15 in 2007. The GNI per capita at market prices of the NMS-8 is reported to be at31 per cent in 2007, and that of the NMS-2 at 17 per cent. The GNI per capita at marketprices of the CAND-6 countries amounted to 22 per cent of those in the EU-15 at thesame time (see Table 1).The wage gap is even larger. The average level of hourly gross wages and salaries in theNMS-8 was 25 per cent of that in the EU-15 in 2006, and that of the NMS-2 at about 11per cent. Note that substantial differences in wage and GNI levels across the newmember states and the candidate countries exist, ranging from a wage of 8 per cent ofthe average level in the EU-15 in Bulgaria to 57 per cent in Slovenia.Altogether, a relatively moderate GNI gap between the old and the new member statesmeasured in purchasing power parities translates in a much larger GNI gap at currentexchange rates. Low-income countries usually have a higher income in purchasing powerparities than at current exchange rates, since the productivity gap to high-incomecountries is lower in non-tradable sectors (e.g. services) compared to tradable sectors(e.g. manufacturing industries). In case of the NMS this income gap is neverthelessstrikingly high. Moreover, the high wage and GNI gap reflects rather poor endowmentswith physical capital in the new member states.IAB 3

Table 1:GNI per capita, hourly gross wages and salaries and net migration in theEU, the other EEA and the candidate countries, 2007GNI per capita at PPSGNI per capitahourly gross wagesand salariesnet migration1 2 3in EUR in % of EU-15 in EUR in % of EU-15 in EUR in % of EU-15 in 1,000 rate per 1,000Austria 31,400 114 f 32,400 112 f 15.00 103 29 3.59Belgium 29,900 108 31,500 109 17.53 120 53 5.12Denmark 31,400 114 42,500 147 24.23 166 10 1.87France 27,700 100 29,900 103 17.58 121 90 17.24Finland 29,600 107 34,000 117 15.46 106 11 0.18Germany 28,600 104 29,700 102 16.56 114 26 0.31Greece 23,800 86 20,000 69 5.71 39 40 3.62Ireland 31,000 112 36,500 126 17.55 121 69 16.93Italy 25,100 91 25,700 89 9.86 68 377 6.56Luxembourg 56,300 204 60,400 208 25.25 173 5 11.81Netherlands 33,300 121 34,800 120 17.71 122 -26 -1.59Portugal 17,600 64 14,700 51 6.72 46 26 2.48Spain 25,200 91 22,800 79 10.88 75 605 14.17Sweden 31,300 113 37,100 128 17.68 121 51 5.65United Kingdom 29,400 107 33,400 115 16.84 116 214 3.57EU-15 27,600 100 29,000 100 14.56 100 1580 4.12Cyprus 22,100 80 19,200 66 8.28 57 6 7.26Malta 18,700 68 12,800 44 7.27 50 1 2.49Czech Republic 18,700 68 f 11,500 40 f 3.71 25 35 3.40Estonia 16,700 61 10,900 38 3.51 24 0 0.12Hungary 14,800 54 9,300 32 4.16 29 21 2.11Latvia 13,900 50 8,000 28 2.92 20 -2 -1.06Lithuania 14,300 52 9,300 32 2.95 20 -5 -1.41Poland 12,900 47 7,700 27 3.34 23 -36 -0.95Slovak Republic 16,400 59 9,800 34 3.42 24 4 0.72Slovenia 22,000 80 16,300 56 8.31 57 6 3.14NMS-8 14,700 53 9,000 31 3.65 25 23 0.31Bulgaria 9,300 34 3,700 13 1.11 8 -34 -4.35Romania 9,600 35 f 5,400 19 f 1.76 12 -100 -4.61NMS-2 9,400 34 5,000 17 1.60 11 -134 -4.54NMS-10 13,200 48 7,800 27 3.03 21 -111 -1.08EU-25 25,600 93 25,900 89 12.74 88 1470 3.02EU-27 24,600 89 24,600 85 12.12 83 1477 3.03Iceland 32,000 116 46,900 162 n.a. n.a. 5 n.a.Norway 45,700 166 60,400 208 26.14 179 24 n.a.Switzerland 34,700 126 41,500 143 22.59 155 37 n.a.Albania n.a. n.a. n.a. n.a. n.a. n.a. -20 -6.43 4Bosnia-Herzegovina n.a. n.a. n.a. n.a. n.a. n.a. 8 2.05 4Croatia 13,900 50 f 8,600 30 n.a. n.a. 7 1.64Macedonia 7,300 26 f 2,700 9 f n.a. n.a. -1 -0.26Serbia-Montenegro n.a. n.a. n.a. n.a. n.a. n.a. -20 -2.45 4Turkey 10,500 38 f 6,500 22 f n.a. n.a. -3 -0.04CAND-6 10,600 38 6,500 22 n.a. n.a. -28 -0.301) Purchasing power parity standards (Eurostat estimate).2) 2006: Hourly labour cost according to Eurostat.3) 2005.4) 2005 (World Development Indicators, 2007).f) forecast.Sources: GNI and hourly labour costs: Eurostat, net migration: Eurostat, supplemented by WDI. Own calculations and presentation.2.2 Human capital investmentThe difference in the income levels between the EU-15, the new member states and thecandidate countries can be largely traced back to differences in factor endowments.Although data on physical capital stocks is scarce, it is likely that the substantial gap inGNI and wages can be largely traced back to differences in capital endowments.IAB 4

However, one important feature sets the NMS apart from traditional emigration countries:The NMS have a human capital endowment which is only slightly below that of the EU-15. In particular, school enrolment rates catch-up to average levels in the EU-15, suchthat existing differences will decline over time.Figure 1: Gross enrolment rates in secondary and tertiary education, 2006140120EU-15gross enrolment rate in per cent100806040EU-15secondarytertiaryNMS-8NMS-8NMS-2 CAND-6NMS-2CAND-62000 100 200 300 400 500population (in million persons)Source: World Bank 2007. Own calculations and presentation.Figure 1 displays the gross school enrolment rates 3 in secondary and tertiary educationfor the EU-15, the NMS-8, the NMS-2 and the CAND-6 countries, which have beencompiled by the World Bank in the World Development Indicators 2007. The gap in bothsecondary and tertiary school enrolment rates between the EU-15 and the NMS-8 is verymoderate. Note that substantial differences across individual EU-15 countries exist.However, there is a gap in the enrolment rates in tertiary education between the EU-15and the NMS-2 and the candidate countries of about 20 percentage points, which reflectsparticularly large differences in university education. However, we observe an increasingschool enrolment in all new member states, such that a convergence or even anovertaken in school enrolment is rather likely in the future. 4Compared to other countries of a similar income level the new member states possessrich endowments with human capital. This may have two consequences which arerelevant in the context of this study: The rich human capital endowment may supportfaster convergence of per capita income levels, and it may result in the emigration of a3 Note that gross school enrolment rates can exceed 100 per cent.4 The trends in school enrolment will be discussed in the report to Deliverable 7.IAB 5

elatively well-educated workforce compared to the traditional sending countries of labourmigration in Northern Africa and South-Eastern Europe.2.3 Convergence of GDP per capita and wage levelsWe find indeed strong evidence that GDP and wage levels between the old and the newmember states tend to converge. In the year 2000, the GDP per capita of the NMS-8measured in PPS amounted to 43 per cent of that in the EU-15, while it is forecasted toachieve 52 per cent in the year 2007. A similar convergence trend can be observed forBulgaria and Romania. Interestingly enough, in the candidate and potential candidatecountries we observe a slower speed of convergence compared to the new member statessince the beginning of this millennium (see Figure 2).Figure 2: Convergence of GDP per capita at PPS, 2000-2007 560%50%GDP per capita PPS in per cent of EU-1540%30%20%10%NMS-8 NMS-2 CAND-60%2000 2001 2002 2003 2004 2005 2006 2007Source: Eurostat 2008. Own calculations and presentation.A similar picture emerges regarding the convergence of the GDP per capita at currentexchange rates: The initial gap in the year 2000 declined both in case of the NMS-8 andthe NMS-2 by 10 percentage points until 2007, but only by 5 percentage points in case ofthe candidate countries during the same time span (see Figure 3).5 Values for 2007 are forecasted by Eurostat.IAB 6

Figure 3: Convergence of GDP per capita at market prices, 2000-2007 635%GDP per capita at current exchange rates in per cent of EU-1530%25%20%15%10%5%NMS-8 NMS-2 CAND-60%2000 2001 2002 2003 2004 2005 2006 2007Source: Eurostat 2008. Own calculations and presentation.We do not investigate the causes of per capita GDP convergence at this stage of ouranalysis. A number of factors may have contributed to the fast GDP convergence in thenew member states, inter alia the rich human capital endowments, the transfers of theEU in the context of the integration of the NMS into the Common Agricultural Policy (CAP)and the regional policies as well as private capital mobility and private investment.Whether migration has contributed to the convergence of GDP levels and wages will bediscussed in detail in Deliverable 4. However, it is important to note that the fastconvergence of GDP levels between the EU-15 and the NMS-8 and the NMS-2 mitigateseconomic incentives to migrate considerably over time.The impact of convergence on migration incentives is even larger if we look at thedevelopment of wages: The hourly gross wages and salaries have increased between2000 and 2006 in the NMS-8 by almost 10 percentage points, and in case of the NMS-2by 5 percentage points between 2002 and 2006. In particular, wages have jumped in theNMS-8 after enlargement in 2004. Labour mobility may have contributed to this wagehike (see Deliverable 4), but is sincerely not the only cause: Transfers into the NMS andcapital mobility may have contributed to the increasing wages as well (see Figure 4). Butthe rapid convergence since 2004 is to be interpreted carefully as it refers only to twoobservations.6 Values for 2007 are forecasted by Eurostat.IAB 7

Figure 4: Convergence of wage levels, 2000-20060.300.25hourly wages and salaries in per cent of EU-150.200.150.100.05NMS-8NMS-20.002000 2001 2002 2003 2004 2005 2006Source: Eurostat 2008. Own calculations and presentation.2.4 Convergence of labour market conditionsThe labour market conditions between the EU-15 and the new member states have alsoconverged since the trough of the transitional recession. Unemployment rates both in theNMS-8 and the NMS-2 meanwhile match the average unemployment rates in the EU-15(see Table 2). Participation rates are – due to a higher female participation in the labourforce – higher in the NMS compared to the EU-15. Altogether, unemployment risks do notcreate specific migration incentives in the NMS.However, two aspects are worthwhile to mention in this context: First, replacement ratesare in the NMS well below those in the EU-15 (OECD, 2008). This may not only createadditional migration incentives for those who are unemployed or suffer from anunemployment risk. It may also result in an underreporting of unemployment in the NMS.Second, migrants can optimise with regard to wage levels and unemployment risksacross locations. In particular, migrants from the NMS-8 cluster in countries and regionswith high wage levels and low unemployment rates in the EU-15, such that a comparisonof average unemployment and wage rates between the EU-15 and the NMS is misleading.IAB 8

Table 2:Unemployment rates in the EU, the NMS and the candidate countries,2000-20072000 2001 2002 2003 2004 2005 2006 2007Austria 3.60 3.60 4.20 4.30 4.80 5.20 4.70 4.40Belgium 6.90 6.60 7.50 8.20 8.40 8.40 8.20 7.50Denmark 4.30 4.50 4.60 5.40 5.50 4.80 3.90 3.70France 9.00 8.30 8.60 9.00 9.30 9.20 9.20 8.30Finland 9.80 9.10 9.10 9.00 8.80 8.40 7.70 6.90Germany 7.50 7.60 8.40 9.30 9.70 10.70 9.80 8.40Greece 11.20 10.70 10.30 9.70 10.50 9.80 8.90 n.a.Ireland 4.20 4.00 4.50 4.70 4.50 4.30 4.40 4.50Italy 10.10 9.10 8.60 8.40 8.00 7.70 6.80 n.a.Luxembourg 2.30 2.00 2.70 3.70 5.10 4.50 4.70 4.70Netherlands 2.80 2.20 2.80 3.70 4.60 4.70 3.90 3.20Portugal 3.90 4.00 5.00 6.30 6.70 7.60 7.70 8.00Spain 11.10 10.30 11.10 11.10 10.60 9.20 8.50 8.30Sweden 5.60 4.90 4.90 5.60 6.30 7.40 7.10 6.10United Kingdom 5.30 5.00 5.10 4.90 4.70 4.80 5.30 n.a.EU-15 7.70 7.20 7.60 7.90 8.00 8.10 7.70 7.00Cyprus 4.90 3.80 3.60 4.10 4.60 5.20 4.60 3.90Malta 6.70 7.60 7.50 7.60 7.40 7.30 7.30 6.30Czech Republic 8.70 8.00 7.30 7.80 8.30 7.90 7.10 5.30Estonia 12.80 12.40 10.30 10.00 9.70 7.90 5.90 4.90Hungary 6.40 5.70 5.80 5.90 6.10 7.20 7.50 7.20Latvia 13.70 12.90 12.20 10.50 10.40 8.90 6.80 5.90Lithuania 16.40 16.50 13.50 12.40 11.40 8.30 5.60 4.30Poland 16.10 18.20 19.90 19.60 19.00 17.70 13.80 9.60Slovak Republic 18.80 19.30 18.70 17.60 18.20 16.30 13.40 11.30Slovenia 6.70 6.20 6.30 6.70 6.30 6.50 6.00 4.70NMS-8 11.27 12.22 12.94 12.76 12.41 11.61 9.30 6.90Bulgaria 16.40 19.50 18.10 13.70 12.00 10.10 9.00 6.90Romania 7.20 6.60 8.40 7.00 8.10 7.20 7.30 n.a.NMS-2 9.66 10.04 10.98 8.77 9.13 7.97 7.75 n.a.NMS-10 10.80 11.58 12.37 11.61 11.47 10.56 8.85 n.a.Iceland n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Norway 3.40 3.60 3.90 4.50 4.40 4.60 3.50 2.60Switzerland n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Albania n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Bosnia-Herzegovina n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Croatia na na 14.70 14.10 13.60 12.60 11.10 9.10Macedonia n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Serbia-Montenegro n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Turkey 5.20 6.80 8.90 9.30 9.00 8.80 8.40 n.a.CAND-6 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.EU-25 8.60 8.40 8.70 9.00 9.00 8.90 8.20 7.20EU-27 8.60 8.50 8.90 8.90 9.00 8.90 8.20 7.10Source: Eurostat 2008. Own calculations and presentation.2.5 The eroding role of distanceTheories of the migration decision traditionally highlight the role of migration costs,particularly the costs of distance (Sjaastadt, 1962; Stark, 1991). The social and psychicIAB 9

costs of moving to an unfamiliar environment play indeed an important role and affectthe structure of migration (Brücker/Schröder, 2006). However, the role of geographicaldistance for migration costs tends to decline with the emergence of low-cost air carriers.Low-budget air transport has two important effects on migration particularly in theEuropean context: First, the role of fixed costs in transport increases, while the role ofvariable costs diminishes. As a consequence, the impact of geographical distancedecreases. Second, due to the high share of fixed costs, transport costs tend to declinewith an increasing migrant community. As a consequence, transport costs becomeendogenous: The more migrants settle in a certain location, the lower are the migrationcosts. Thus, within the European context, it becomes more and more uncertain wheremigrants settle.We have collected data on distance and different types of transport costs to illustrate thispoint. Geographical distance and the costs for road and air transport are calculated for 13sending and 15 destination countries, which gives 195 data points. The data are reportedin Annex Table 1. Road transport by car is largely determined by variable costs, i.e.gasoline, fares for ferries, and depreciation. Depreciation depends largely – albeit notonly – on the kilometres run by the vehicle. We use a standard route planning system tocalculate the costs by car, which are – as a consequence of assumptions applied here –largely linear in distance (see Figure 5).Figure 5:Transport costs by car and distance700600travelling cost by car in EUR500400300200100y = 0.1195x + 4.094R 2 = 0.763200 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000distance in kmSource: Own calculations based on the Falk-route planning system.In contrast, there is only a weak correlation between air transport costs and distance. Forthe calculation we have used the cheapest connection provided by the OPODO bookingsystem. As Figure 5 demonstrates, the costs of air transport are only weakly increasingIAB 10

with geographical distance. In particular, for the relevant range between 500 and 2,500kilometres, there is no clear correlation between air fares and distance (see Figure 6).Of course this illustrative evidence can only sketch the changing role of transport costs. Itmay, however, have very important implications for the geographical structure of labourmobility in the context of EU enlargement: While past migration patterns in the EU havebeen largely determined by geographical proximity, the emergence of low-cost carriersmakes it more and more likely that migrants choose destinations by other criteria such aslanguage, climate or labour market conditions. Moreover, network effects may becomemore important, since transport costs depend on the size of the migrant community.Thus, even if Austria or Germany open their labour markets, long-distance destinationssuch as Ireland and the UK might remain attractive destinations for migrants from theNMS in the future.Figure 6:Costs of air transport and distance800700air transport costs in EUR600500400300200100y = 0.0279x + 231.88R 2 = 0.045400 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000distance in kmSource: Own calculations based on the OPODO-booking system.2.6 Concluding remarksParticularly the nominal gap in wages between the EU-15 and the NMS as well asbetween the EU-15 and the candidate countries creates substantial migration incentivesat present. These incentives however diminish over time, since the convergence of wagesand employment conditions is fast particularly in the NMS-8. The difference in the speedof convergence between the NMS-8 and candidate countries suggests that Easternenlargement may have contributed to mitigate monetary migration incentives.IAB 11

The NMS are, relative to their income levels, well endowed with human capital. This isparticularly true for the NMS-8. Their school enrolment rates are only slightly below thoseof the EU-15 average, and well above those of the Southern EU-15 member states.School enrolment in tertiary education is substantially higher in the NMS-8 compared tothe candidate countries and other traditional sending countries of European immigration,e.g. in Northern Africa. This creates a large potential of medium and high skilled migrantsparticularly in the NMS-8.The role of geographical distance for transport costs diminishes in Europe due to theemergence of low-cost carriers in air transport. As a consequence, geographical proximityplays a less important role for the choice of migration destinations. Migrants from Centraland Eastern Europe may therefore prefer destinations even if the geographical distance islarge if other factors such as wages, employment opportunities, language, climate etc.motivate migration. Moreover, the role of network effects increases since transport costsdepend more and more on the size of the migrant community.3 The scale of labour mobilityThis section presents the main migration trends in the enlarged EU. The section startswith a brief discussion of the definitions applied in the analysis and limitations of theavailable data (Section 3.1). We then present the development of migration stocks in theenlarged EU both from the receiving (Section 3.2 - 3.4) and the sending countries(Section 3.5).3.1 Definitions and data restrictionsThroughout the analysis, we refer to the concept of citizenship in describing migrationspatterns in the context of the EU’s eastern enlargement. This excludes a part of themigrants from the new member states residing in the EU-15, e.g. ethnic Germans (socalled“Spätaussiedler”) which have migrated from the NMS into the EU-15 during the1990s. Nevertheless, the free movement of workers and the transitional arrangementsrefers to the concept of citizenship, such that we believe that a nationality-based conceptis most appropriate in the context of our analysis. It is however important to keep inmind that the definition of foreign nationals differs across destination countries in the EUdepending on legal traditions and naturalisation practices, such that figures about thestocks of foreign residents are not entirely comparable across the EU member states.Nonetheless, since migration from NMS is a recent phenomenon in most EU countries,these differences have only a minor quantitative impact. 77 Germany is the main exception here, since the number of ethnic Germans which have immigratedinto Germany has roughly the same size as the immigration of citizens from the NMS during the1990s. However, the immigration of ethnic Germans has ceased since the beginning of thisdecade.IAB 12

Moreover, our analysis is restricted to legal migration. Data on illegal migration arescarce and highly unreliable, such that we cannot cover this phenomenon empirically.Since the free movement of workers is likely to diminish incentives for illegal migrationfrom the NMS, this affects our analysis in several ways. Current immigration flows mightbe overstated if illegal migrants use the new opportunities to legalise their status ofresidency and employment in host countries. Similarly, the wage and employment effectsof immigration from the NMS may be overstated if legal activities of immigrants replaceillegal activities. Finally, migration may have a different impact on public finances if weconsider that activities in the shadow economy are replaced by activities in the firstlabour market.The figures picturing the migration trends are drawn from different data sourcesdepending on the availability of data. Priority is given to figures which are derived fromthe population statistics and provided by National Statistical Offices and Eurostat.Unfortunately, these figures are only available for about two-third of the EU-15 countries.For the remaining countries, we report the figures from the European Labour ForceSurvey (LFS), in case of UK from the UK LFS. The LFS is an EU wide household surveycollecting data about labour force participation and other socio-economic factors whichwas first implemented in 1960 by the six original EU Member States. Today, the survey –hosted by Eurostat – covers all 27 States and is a key research instrument by providingunique time series data about economic and social developments in Europe.In case of Ireland, the main destination of immigrants from the NMS in relative terms,specific data problems arise. The European LFS does not include data for Ireland for mostof the sample periods. Since 2004 we employ data from the Irish Labour Force Survey.Unfortunately, this dataset reports only aggregate figures for the NMS-8 and since 2007for the NMS-10 such that we use the contingent derived from the Personal Public ServiceNumbers (PPSN) 8 to disentangle migration from each sending country. Moreover, noinformation on the skill and age structure is available. Beyond Ireland, there are also anumber of other EU member states which do not report the entire information onimmigrants from the NMS due to low response rates. However, these countries arerelatively small such that this does not much affect the overall results.Although using three different data sources, it was not possible to obtain informationabout the stock of foreign residents for all individual sending countries. In come cases,response rates have been too small to cover all countries of origin from the NMS. As aconsequence, the aggregate figures of migration stocks from NMS-8, NMS-2, and Cand-6migrants as reported below may slightly underestimate the actual number of foreignresidents in the EU-15.8 The PPS Number is a unique reference number that helps to gain access to social welfarebenefits, public services and information in Ireland. State agencies that use PPS Numbers toidentify individuals include the Department of Social and Family Affairs, the RevenueCommissioners and the Health Services Executive (HSE) Areas.IAB 13

Some further restrictions apply to the LFS data sources in our context. First, immigrantsmay generally be under-represented in the LFS as the survey is usually carried out in thenational languages of the host countries. Second, many immigrants from the NMS areemployed as seasonal workers, e.g. in agriculture and construction, which are likely to beunderreported particularly if the LFS is undertaken off season. Third, the sample designand rotation patterns are not fully harmonised: Various schemes are used to sample theunits in the different member states. This may, in turn, lead to a long time span untilnew migration waves (households) rotate in the sample, resulting in a possible underrepresentationof migrants in the current year LFS.In contrast, migration figures in the population statistics may overstate legal migrationfrom the NMS. These statistics on the stocks of residents relies usually on registers of theforeign population, which tend to understate return migration since no incentives exist toderegister.Our analysis of the skill and age structure of immigrants from the NMS as well as on theiremployment status is based again on LFS sources. We restrict our analysis to theemployed working age population (15-64 age group) in case of skill and age structure,and to the overall working age population in case of employment status. The figures aredrawn from a special provision from the European LFS for second quarter 2006. In caseof missing information, we use the 2005 values wherever necessary.3.2 Immigration from the NMS-8 into the EU and EEAThe number of foreign residents from the NMS-8 in the EU-15 has increased from893,000 persons in the year before Eastern Enlargement (2003) to 1.91 million personsor 0.5 per cent of the population of the EU-15 by the end of 2007. This corresponds to anannual increase of 254,000 persons p.a. on average since Eastern enlargement comparedto 62,000 persons p.a. in the years from 2000 to 2003. The stock of migrants from theNMS-8 in the new member states of the EU is at about 100,000 persons small and onlyslightly increasing. In the remaining member states of the European Economic Area(Iceland, Norway, Liechtenstein) and Switzerland, the number of foreign residents fromthe NMS-8 has increased from 28,000 to approximately 61,000 persons during the 2003-2007 period (see Tables 3a/b).Since the beginning of Eastern enlargement in 2003, almost 70 per cent of theimmigrants from the NMS-8 have been absorbed by the UK and Ireland. These twocountries have replaced Austria and Germany as the main destinations for migrants fromthe NMS-8. The stock of foreign residents from the NMS-8 increased from 95,000 toabout 609,000 persons in the UK since 2000 according to the LFS data and from 44,000to about 179,000 persons in Ireland since 2004. By the end of 2007, the stock of foreignresidents from the NMS-8 achieves 4 per cent of the population in Ireland and about 1per cent of the population in the UK.IAB 14

Table 3a: Foreign residents from the NMS-8 in the EU and EEA, 2000-2007Host country 2000 2001 2002 2003 2004 2005 2006 2007in personsAustria 1 n.a. 54,797 57,537 60,255 68,933 77,264 83,978 89,940Belgium 1 9,667 12,102 14,106 16,151 19,524 25,638 32,199 42,918Denmark 1 9,101 9,447 9,805 9,807 11,635 14,282 16,527 22,146Finland 1 12,804 13,860 14,712 15,825 16,459 18,266 20,801 23,957France 3 37,832 44,946 44,857 33,858 43,138 36,237 44,181 36,971Germany 1 434,603 453,110 466,356 480,690 438,828 481,672 525,078 554,372Greece 3 13,832 12,695 14,887 16,413 15,194 19,513 18,357 20,257Ireland 4 n.a. n.a. n.a. n.a. 43,500 94,000 147,900 178,504Italy 2 40,433 40,108 41,431 54,665 66,159 77,889 91,318 117,042Luxembourg 1 n.a. n.a. 1,156 1,574 2,278 3,488 4,217 5,101 eNetherlands 1 10,063 11,152 12,147 13,048 17,814 23,155 28,344 36,317Portugal n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Spain 1 19,284 29,998 41,471 46,710 61,830 77,772 100,832 131,118 eSweden 1 23,884 22,868 21,376 21,147 23,257 26,877 33,757 42,312United Kingdom 5 94,792 105,048 93,340 122,465 120,999 219,797 357,468 609,415EU-15 706,295 755,334 833,181 892,608 949,548 1,195,850 1,504,957 1,910,370Island 1 1,865 2,232 2,462 2,547 2,644 4,251 7,803 10,782Norway 1 3,366 3,658 4,195 5,166 5,549 7,427 11,240 20,074Switzerland 1 17,598 18,733 19,997 20,308 20,909 22,060 25,711 29,786EEA-2 and CH 22,829 24,623 26,654 28,021 29,102 33,738 44,754 60,642Cyprus n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Cyprus and Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Czech Republic 1 62,095 70,581 77,947 81,484 64,546 68,300 78,428 90,258 eEstonia n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Hungary 6 4,632 4,715 3,739 5,001 3,596 6,346 7,445 8,755 eLatvia 6 n.a. n.a. 2,524 3,121 n.a. 3,755 4,119 4,526 eLithuania 6 n.a. n.a. n.a. n.a. 735 934 992 1,061 ePoland n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Slovak Republic 6 n.a. n.a. n.a. 9,372 7,698 9,057 11,017 13,429 eSlovenia 6 n.a. n.a. 418 492 203 656 711 794 eNMS-8 66,727 75,296 84,628 99,470 76,778 89,048 102,712 118,823Bulgaria n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Romania 6 n.a. 372 n.a. 372 373 365 362 359 eNMS-2 n.a. 372 n.a. 372 373 365 362 359Sources: National population statistics, Eurostat, LFS, own calculations and presentation.1) National Statistics; 2) 2000-01: Eurostat; 2002-07: National Statistics; 3) LFS annual 4) 2004-07: Irish-LFS 4th Qu. (15+);5) 2000-07: UK-LFS 2th Qu.; 6) Eurostat; e: estimatedIn contrast, Austria and Germany experienced only a modest increase in the number offoreign residents from the NMS-8 during the 2003 – 2007 period. The stock of foreignresidents from the NMS-8 has increased by about 30,000 persons in Austria. Germanyhas revised its migration statistics in 2004 such that the actual increase cannot becalculated properly. Taking the data revision into account, we can estimate the actualincrease in the number of foreign residents at 70,000 persons for the 2003 - 2007 period.Foreigners from the new member states achieve meanwhile a share of 1 per cent of thepopulation in Austria and 0.7 per cent in Germany. Other important destinations formigrants from the NMS-8 are Spain (85,000 persons), Italy (62,000 persons), Belgium(26,000 persons), The Netherlands (23,000 person) and Belgium (21,000 persons), butthe share of foreign residents from the NMS-8 in the population of these countries doesnot exceed the EU-15 average of 0.5 per cent.IAB 15

Table 3b:Foreign residents from the NMS-8 in the EU and EEA in per cent of the hostpopulation, 2000-2007Host country 2000 2001 2002 2003 2004 2005 2006 2007share of total populationAustria 1 n.a. 0.68% 0.71% 0.74% 0.84% 0.94% 1.01% 1.08%Belgium 1 0.09% 0.12% 0.14% 0.16% 0.19% 0.24% 0.31% 0.40%Denmark 1 0.17% 0.18% 0.18% 0.18% 0.22% 0.26% 0.30% 0.41%Finland 1 0.25% 0.27% 0.28% 0.30% 0.31% 0.35% 0.39% 0.45%France 3 0.06% 0.07% 0.07% 0.05% 0.07% 0.06% 0.07% 0.06%Germany 1 0.53% 0.55% 0.57% 0.58% 0.53% 0.58% 0.64% 0.67%Greece 3 0.13% 0.12% 0.14% 0.15% 0.14% 0.18% 0.16% 0.18%Ireland 4 n.a. n.a. n.a. n.a. 1.07% 2.26% 3.47% 4.09%Italy 2 0.07% 0.07% 0.07% 0.09% 0.11% 0.13% 0.15% 0.20%Luxembourg 1 n.a. n.a. 0.26% 0.35% 0.50% 0.76% 0.90% 1.06% eNetherlands 1 0.06% 0.07% 0.08% 0.08% 0.11% 0.14% 0.17% 0.22%Portugal n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Spain 1 0.05% 0.07% 0.10% 0.11% 0.14% 0.18% 0.23% 0.29% eSweden 1 0.27% 0.26% 0.24% 0.24% 0.26% 0.30% 0.37% 0.46%United Kingdom 5 0.16% 0.18% 0.16% 0.21% 0.20% 0.36% 0.59% 1.00%EU-15 0.20% 0.21% 0.23% 0.24% 0.25% 0.32% 0.40% 0.50%Island 1 0.66% 0.78% 0.86% 0.88% 0.91% 1.43% 2.57% 3.47%Norway 1 0.07% 0.08% 0.09% 0.11% 0.12% 0.16% 0.24% 0.43%Switzerland 1 0.24% 0.26% 0.27% 0.28% 0.28% 0.30% 0.34% 0.39%EEA-2 and CH 0.19% 0.20% 0.22% 0.23% 0.24% 0.27% 0.36% 0.48%Cyprus n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Cyprus and Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Czech Republic 1 0.60% 0.69% 0.76% 0.80% 0.63% 0.67% 0.76% 0.87% eEstonia n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Hungary 6 0.05% 0.05% 0.04% 0.05% 0.04% 0.06% 0.07% 0.09% eLatvia 6 n.a. n.a. 0.11% 0.13% n.a. 0.16% 0.18% 0.20% eLithuania 6 n.a. n.a. n.a. n.a. 0.02% 0.03% 0.03% 0.03% ePoland n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Slovak Republic 6 n.a. n.a. n.a. 0.17% 0.14% 0.17% 0.20% 0.25% eSlovenia 6 n.a. n.a. 0.02% 0.02% 0.01% 0.03% 0.04% 0.04% eNMS-8 0.09% 0.10% 0.12% 0.14% 0.11% 0.12% 0.14% 0.16%Bulgaria n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Romania 6 n.a. 0.00% n.a. 0.00% 0.00% 0.00% 0.00% 0.00% eNMS-2 n.a. 0.00% n.a. 0.00% 0.00% 0.00% 0.00% 0.00%Sources: National population statistics, Eurostat, LFS, own calculations and presentation.1) National Statistics; 2) 2000-01: Eurostat; 2002-07: National Statistics; 3) LFS annual 4) 2004-07: Irish-LFS 4th Qu. (15+);5) 2000-07: UK-LFS 2th Qu.; 6) Eurostat; e: estimatedThe share of Austria and Germany in the total number of foreign residents from the NMS-8 in the EU-15 has declined from almost 63 per cent in 2002 to 34 per cent in 2007,while that of Ireland and the UK has increased from 11 per cent to 41 per cent during thesame period of time. This diversion process can be inter alia explained by the selectiveapplication of the transitional arrangements for the free movement of workers. WhileIreland and the UK opened their labour markets, Austria and Germany maintained theirimmigration restrictions. Interestingly enough, other destinations which have openedtheir labour markets completely (Sweden) or partially (Denmark) have not been affectedby this diversion effect.The available data for the years 2006 and 2007 do moreover not suggest that theremoval of immigration restrictions in numerous EU member states (Finland, Greece,Italy, Portugal, Netherlands, Spain) for the second period of the transitionalIAB 16

arrangements has involved a visible increase in immigration flows from the NMS-8. Byand large, the removal of migration barriers in these ‘second-movers’ has not affectedthe scale of migration in the enlarged EU.The available evidence thus suggests that the high share of migrants from the NMS-8 inIreland and the UK cannot be explained by the selective application of transitionalarrangements for the free movement of workers alone. Other factors, such as theincreasing English language proficiency particularly among the young cohorts in the NMS,favourable labour market conditions and flexible labour market institutions, and thedeclining costs of distance, have facilitated the diversion of migration flows to thesedestinations as well.3.3 Immigration from the NMS-2 into the EU and EEAImmigration from Bulgaria and Romania – summarised as the two new member states(NMS-2) – into EU-15 countries is heavily restricted in most EU-15 countries.Nonetheless, the number of foreign residents from there has increased from 279,000persons in 2000 to 1.86 million by the end of 2007. This corresponds to an annualincrease in the number of residents of about 226,000 persons p.a. Meanwhile, the stockof foreign residents from the NMS-2 has achieved 0.49 per cent of the population in theEU-15. In the NMS-8 the stock of NMS-2 immigrants stagnates at about 77,000 persons.In the other member states of the EEA and Switzerland, immigration from the NMS-2 isat some 9,000 persons negligible (see Tables 4a/b).Immigration from Bulgaria and Romania has been facilitated by bilateral agreementsbetween Spain and Italy and the sending countries and the legalisation of immigrantsthere. Spain is the main destination for migrants from the NMS-2 at a migration stock ofabout 829,000 persons, followed by Italy with 659,000 persons. 9 By the end of 2007, theshare of NMS-2 immigrants in the population achieves 1.9 per cent in Spain and 1.1 percent in Italy. Other important destinations in the EU-15 are Germany (131,000 persons),Greece (53,000 persons), the UK (40,000 persons) and Austria (37,000 persons).9 Note that the official statistics may underreport migrants from the NMS-2 in Italy, since it doesinter alia not count people whose residence permit has expired but still stay in the country andwait for a prolongation. The Italian Caritas estimates therefore the stock of migrants from theNMS-2 in Italy at about 560,000 persons by the end of 2006.IAB 17

Table 4a: Foreign residents from the NMS-2 in the EU and EEA, 2000-2007Host country 2000 2001 2002 2003 2004 2005 2006 2007in personsAustria 1 n.a. 22,387 24,926 26,802 28,367 29,573 29,958 36,792Belgium 1 3,435 4,642 5,900 6,831 8,238 10,814 14,095 23,810Denmark 1 1,580 1,646 1,746 1,834 1,987 2,200 2,350 3,316Finland 1 786 854 873 887 909 970 1,089 1,388France 3 5,752 8,761 7,960 8,840 17,282 12,027 39,069 43,652Germany 1 124,453 126,245 131,098 133,404 112,532 112,196 112,406 131,402Greece 3 12,961 17,344 25,612 30,583 39,220 45,551 49,086 52,567Ireland 4 n.a. n.a. n.a. n.a. n.a. n.a. n.a. 24,496Italy 2 69,020 81,444 102,363 189,279 264,223 315,316 362,124 658,755Luxembourg 1 n.a. n.a. 477 498 545 700 871 1,085 eNetherlands 1 2,564 3,168 3,720 4,413 4,944 5,082 5,427 11,272Portugal n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Spain 1 43,676 97,020 190,185 277,814 410,403 508,776 649,076 828,772 eSweden 1 3,951 3,300 3,123 3,148 3,170 3,205 3,080 6,280United Kingdom 5 10,504 9,739 17,494 17,979 17,118 33,578 37,945 40,023EU-15 278,682 376,550 515,477 702,312 908,938 1,079,988 1,306,576 1,863,610Island 1 108 123 141 143 154 178 204 241Norway 1 835 893 1,049 1,205 1,313 1,427 1,520 1,543Switzerland 1 5,060 5,745 6,480 6,535 6,748 6,813 6,846 6,943EEA-2 and CH 6,003 6,761 7,670 7,883 8,215 8,418 8,570 8,727Cyprus n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Cyprus and Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Czech Republic 1 6,408 6,405 6,485 6,303 7,035 7,252 7,451 7,656 eEstonia n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Hungary 6 44,371 46,123 48,366 56,794 68,785 67,390 68,074 68,766 eLatvia 6 n.a. n.a. 26 42 n.a. 37 44 52 eLithuania 6 n.a. n.a. n.a. n.a. 33 46 107 249 ePoland n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Slovak Republic 6 n.a. n.a. n.a. 2,757 1,051 971 1,247 1,711 eSlovenia 6 n.a. n.a. 213 240 199 208 284 396 eNMS-8 50,779 52,528 55,090 66,136 77,103 75,904 77,207 78,831Bulgaria n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Romania 6 n.a. 189 n.a. 189 190 186 186 186 eNMS-2 n.a. 189 n.a. 189 190 186 186 186Sources: National population statistics, Eurostat, LFS, own calculations and presentation.1) National Statistics; 2) 2000-01: Eurostat; 2002-07: National Statistics; 3) LFS annual 4) 2004-07: Irish-LFS 4th Qu. (15+);5) 2000-07: UK-LFS 2th Qu.; 6) Eurostat; e: estimatedAgain, we observe a diversion effect: Germany has been with some 260,000 residentsthe main destination for migrants from the NMS-2 in the beginning of the 1990s, a figurewhich has declined to some 124,000 persons by the beginning of this decade. At thesame time, migration from Romania and Bulgaria to Spain and Italy has increasedsubstantially.It is worthwhile to note in this context that the figures presented here refer to legalmigration only. Incentives for illegal migration are high in case of Bulgaria and Romania,since legal immigration opportunities are limited. Anecdotal evidence suggests that actualmigration stocks from the NMS-2 in the EU-15 might be twice the official figures, butreliable evidence is missing.IAB 18

Table 4b:Foreign residents from the NMS-2 in the EU and EEA in per cent of the hostpopulation, 2000-2007Host country 2000 2001 2002 2003 2004 2005 2006 2007share of total populationAustria 1 n.a. 0.28% 0.31% 0.33% 0.35% 0.36% 0.36% 0.44%Belgium 1 0.03% 0.05% 0.06% 0.07% 0.08% 0.10% 0.13% 0.22%Denmark 1 0.03% 0.03% 0.03% 0.03% 0.04% 0.04% 0.04% 0.06%Finland 1 0.02% 0.02% 0.02% 0.02% 0.02% 0.02% 0.02% 0.03%France 3 0.01% 0.01% 0.01% 0.01% 0.03% 0.02% 0.06% 0.07%Germany 1 0.15% 0.15% 0.16% 0.16% 0.14% 0.14% 0.14% 0.16%Greece 3 0.12% 0.16% 0.23% 0.28% 0.35% 0.41% 0.44% 0.47%Ireland 4 n.a. n.a. n.a. n.a. n.a. n.a. n.a. 0.56%Italy 2 0.12% 0.14% 0.18% 0.33% 0.45% 0.54% 0.61% 1.11%Luxembourg 1 n.a. n.a. 0.11% 0.11% 0.12% 0.15% 0.19% 0.23% eNetherlands 1 0.02% 0.02% 0.02% 0.03% 0.03% 0.03% 0.03% 0.07%Portugal n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Spain 1 0.11% 0.24% 0.46% 0.66% 0.96% 1.17% 1.47% 1.85% eSweden 1 0.04% 0.04% 0.03% 0.04% 0.04% 0.04% 0.03% 0.07%United Kingdom 5 0.02% 0.02% 0.03% 0.03% 0.03% 0.06% 0.06% 0.07%EU-15 0.08% 0.10% 0.14% 0.19% 0.24% 0.29% 0.35% 0.49%Island 1 0.04% 0.04% 0.05% 0.05% 0.05% 0.06% 0.07% 0.08%Norway 1 0.02% 0.02% 0.02% 0.03% 0.03% 0.03% 0.03% 0.03%Switzerland 1 0.07% 0.08% 0.09% 0.09% 0.09% 0.09% 0.09% 0.09%EEA-2 and CH 0.05% 0.06% 0.06% 0.06% 0.07% 0.07% 0.07% 0.07%Cyprus n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Cyprus and Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Czech Republic 1 0.06% 0.06% 0.06% 0.06% 0.07% 0.07% 0.07% 0.07% eEstonia n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Hungary 6 0.43% 0.45% 0.48% 0.56% 0.68% 0.67% 0.68% 0.68% eLatvia 6 n.a. n.a. 0.00% 0.00% n.a. 0.00% 0.00% 0.00% eLithuania 6 n.a. n.a. n.a. n.a. 0.00% 0.00% 0.00% 0.01% ePoland n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Slovak Republic 6 n.a. n.a. n.a. 0.05% 0.02% 0.02% 0.02% 0.03% eSlovenia 6 n.a. n.a. 0.01% 0.01% 0.01% 0.01% 0.01% 0.02% eNMS-8 0.07% 0.07% 0.08% 0.09% 0.11% 0.10% 0.11% 0.11%Bulgaria n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Romania 6 n.a. 0.00% n.a. 0.00% 0.00% 0.00% 0.00% 0.00% eNMS-2 n.a. 0.00% n.a. 0.00% 0.00% 0.00% 0.00% 0.00%Sources: National population statistics, Eurostat, LFS, own calculations and presentation.1) National Statistics; 2) 2000-01: Eurostat; 2002-07: National Statistics; 3) LFS annual 4) 2004-07: Irish-LFS 4th Qu. (15+);5) 2000-07: UK-LFS 2th Qu.; 6) Eurostat; e: estimated3.4 Immigration from the candidate countries into the EU and EEAThe six candidate and potential candidate countries (CAND-6) from South-Eastern Europehave been one of the main sources of immigrants in Western Europe during the post-WWII period. Especially workers from Turkey and from former Yugoslavia have been the maintargets for guestworker recruitment in Austria, Germany, Switzerland and other WesternEuropean countries. In addition, migrants from Albania, one of the countries with thelowest per capita income in Europe, form an important source of immigration in Italy andGreece since the removal of emigration barriers in the beginning of the 1990s.Altogether, the stock of immigrants from the candidate countries in the EU-15 amountedIAB 19

to 4.1 million people in the EU-15 10 in 2000 and another 476,000 people residing in theother EEA countries and Switzerland at the same time. Since the EU’s Easternenlargement, the stock of migrants from this region however stagnates in the EU-15. Bythe end of 2007, the EU-15 countries reports about 4.3 million migrants from thecandidate countries (see Tables 5a/b).Table 5a:Foreign residents from the candidate countries in the EU and the EEA,2000-2007Host country 2000 2001 2002 2003 2004 2005 2006 2007in personsAustria 1 n.a. 432,149 437,481 428,386 420,237 415,857 405,949 401,885Belgium 1 66,240 56,872 54,018 53,811 52,525 53,857 54,758 66,349Denmark 1 58,086 52,841 50,319 48,146 47,304 45,494 44,872 45,065Finland 1 5,061 6,107 6,561 7,328 7,937 8,101 8,395 8,397France 3 240,328 233,120 250,124 116,420 159,829 186,629 153,974 168,246Germany 1 3,097,721 3,025,940 2,968,399 2,922,084 2,346,782 2,519,298 2,477,923 2,405,952Greece 3 181,842 209,475 252,780 288,834 338,863 343,603 337,901 376,487Ireland 4 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Italy 2 227,148 291,816 346,331 422,471 487,518 533,861 576,251 611,807Luxembourg 1 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Netherlands 1 113,851 112,596 112,195 113,584 111,725 109,321 106,411 102,798Portugal n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Spain 1 6,584 7,970 9,172 8,914 10,468 10,493 9,939 9,458 eSweden 1 48,342 42,437 36,736 33,699 32,309 30,224 27,083 27,271United Kingdom 5 61,074 83,063 89,731 96,260 81,866 77,995 106,430 102,255EU-15 4,106,277 4,554,386 4,613,847 4,539,937 4,097,363 4,334,733 4,309,886 4,325,970Island 1 609 697 740 724 699 734 813 680Norway 1 27,507 25,723 20,810 19,707 17,539 17,053 15,552 14,072Switzerland 1 447,839 452,933 455,804 452,495 445,797 436,546 423,670 413,089EEA-2 and CH 475,955 479,353 477,354 472,926 464,035 454,333 440,035 427,841Cyprus n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Cyprus and Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Czech Republic 1 8,556 7,976 8,098 7,917 9,036 9,413 10,134 10,959 eEstonia n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Hungary 6 1,916 1,965 9,628 14,310 2,962 14,459 14,913 15,391 eLatvia 6 n.a. n.a. 45 46 n.a. 79 70 72 eLithuania 6 n.a. n.a. n.a. n.a. 70 71 132 265 ePoland n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Slovak Republic 6 n.a. n.a. n.a. 2,784 1,160 1,170 1,626 2,786 eSlovenia 6 n.a. n.a. 40,424 40,553 40,306 43,371 48,130 53,577 eEU-8 10,472 9,941 58,195 65,610 53,534 68,563 75,005 83,051Bulgaria n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Romania 6 n.a. 3,027 n.a. 3,027 3,069 3,071 3,079 3,087 eEU-2 n.a. 3,027 n.a. 3,027 3,069 3,071 3,079 3,087Sources: National population statistics, Eurostat, LFS, own calculations and presentation.1) National Statistics; 2) 2000-01: Eurostat; 2002-07: National Statistics; 3) LFS annual 4) 2004-07: Irish-LFS 4th Qu. (15+);5) 2000-07: UK-LFS 2th Qu.; 6) Eurostat; e: estimated10 Unfortunately, we have no figures for Ireland, Luxembourg and Portugal.IAB 20

Table 5b:Foreign residents from the candidate countries in the EU and the EEA in percent of the host population, 2000-2007Host country 2000 2001 2002 2003 2004 2005 2006 2007share of total populationAustria 1 n.a. 5.37% 5.41% 5.27% 5.14% 5.05% 4.90% 4.83%Belgium 1 0.65% 0.55% 0.52% 0.52% 0.50% 0.51% 0.52% 0.62%Denmark 1 1.09% 0.99% 0.94% 0.89% 0.88% 0.84% 0.83% 0.83%Finland 1 0.10% 0.12% 0.13% 0.14% 0.15% 0.15% 0.16% 0.16%France 3 0.40% 0.38% 0.41% 0.19% 0.26% 0.30% 0.24% 0.26%Germany 1 3.77% 3.67% 3.60% 3.54% 2.84% 3.05% 3.01% 2.92%Greece 3 1.67% 1.91% 2.30% 2.62% 3.06% 3.09% 3.03% 3.36%Ireland 4 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Italy 2 0.40% 0.51% 0.61% 0.73% 0.84% 0.91% 0.98% 1.03%Luxembourg 1 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Netherlands 1 0.71% 0.70% 0.69% 0.70% 0.69% 0.67% 0.65% 0.63%Portugal n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Spain 1 0.02% 0.02% 0.02% 0.02% 0.02% 0.02% 0.02% 0.02% eSweden 1 0.54% 0.48% 0.41% 0.38% 0.36% 0.33% 0.30% 0.30%United Kingdom 5 0.10% 0.14% 0.15% 0.16% 0.14% 0.13% 0.18% 0.17%EU-15 1.15% 1.25% 1.26% 1.23% 1.10% 1.16% 1.15% 1.15%Island 1 0.22% 0.24% 0.26% 0.25% 0.24% 0.25% 0.27% 0.22%Norway 1 0.61% 0.57% 0.46% 0.43% 0.38% 0.37% 0.33% 0.30%Switzerland 1 6.23% 6.26% 6.26% 6.17% 6.03% 5.87% 5.66% 5.47%EEA-2 and CH 3.98% 3.99% 3.94% 3.88% 3.78% 3.68% 3.53% 3.40%Cyprus n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Cyprus and Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Czech Republic 1 0.08% 0.08% 0.08% 0.08% 0.09% 0.09% 0.10% 0.11% eEstonia n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Hungary 6 0.02% 0.02% 0.09% 0.14% 0.03% 0.14% 0.15% 0.15% eLatvia 6 n.a. n.a. 0.00% 0.00% n.a. 0.00% 0.00% 0.00% eLithuania 6 n.a. n.a. n.a. n.a. 0.00% 0.00% 0.00% 0.01% ePoland n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Slovak Republic 6 n.a. n.a. n.a. 0.05% 0.02% 0.02% 0.03% 0.05% eSlovenia 6 n.a. n.a. 2.03% 2.03% 2.02% 2.17% 2.40% 2.65% eEU-8 0.01% 0.01% 0.08% 0.09% 0.07% 0.09% 0.10% 0.11%Bulgaria n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.Romania 6 n.a. 0.01% n.a. 0.01% 0.01% 0.01% 0.01% 0.01% eEU-2 n.a. 0.01% n.a. 0.01% 0.01% 0.01% 0.01% 0.01%Sources: National population statistics, Eurostat, LFS, own calculations and presentation.1) National Statistics; 2) 2000-01: Eurostat; 2002-07: National Statistics; 3) LFS annual 4) 2004-07: Irish-LFS 4th Qu. (15+);5) 2000-07: UK-LFS 2th Qu.; 6) Eurostat; e: estimatedThe main destination for immigrants from the candidate and potential countries isGermany. In 2000, about 3.1 million or 75 per cent of the immigrants from the candidatecountries in the EU-15 resided in Germany. The German migration statistics reports 2.4million residents from the candidate countries or 56 per cent of the migrants from therein the EU-15 by the end of 2007. This decline can be largely traced back to the revision ofthe migration statistics which reduced the number of migrants from the candidatecountries by about 600,000 persons. Moreover, the repatriation of refugees from the civilwars in the former Yugoslavia and an increasing number of naturalisations following thereform of the immigration act in 2000 has contributed to this decline. Other importantdestinations for migrants from the candidate countries are Italy (612,000 persons),Austria (402,000 persons), Greece (376,000 persons) and France (168,000 persons), andamong the EEA countries Switzerland with 413,000 persons. While the number ofimmigrants from the candidate countries has declined or stagnated in most destinationIAB 21

countries, it has substantially increased in Italy (+385,000 persons) and Greece(+195,000 persons) since the beginning of this decade. This can be traced back largely tothe immigration of Albanians and some successor states of the former Yugoslavia tothese destinations.To sum up, immigrants from the candidate and potential candidate countries exceed thestock of foreign residents from the new member states at a share of 1.2 per cent of thepopulation in the EU-15 by far. However, with the notable exceptions of Italy and Greece,this stock is stagnating or declining in most destinations since the beginning of thisdecade. Tighter immigration conditions for third country nationals in most EU memberstates (Boeri/Brücker, 2005) and adverse economic conditions in main destinations suchas Germany have contributed to this development.3.5 Main emigration trends from a sending country perspectiveBy the end of 2007, the migration data from the statistics in the receiving countriesindicates that about 3.8 million emigrants from the NMS-10 resided in the EU-15. Themain sending countries are Romania (1.6 million) and Poland (1.3 million). The share ofEU-emigrants in the population of the sending countries fluctuates heavily across countrygroups and individual countries. About 2.6 per cent of the population in the NMS-8 and6.4 per cent of the population of the NMS-2 resided by the end of 2007 in the EU-15. Theemigration shares in the population vary with the per capita income level: Whileemigration shares are relatively low in the Czech Republic (1.0 per cent), Hungary (1.3per cent), and Slovenia (1.8 per cent), they are particularly high in Romania (7.2 percent), Bulgaria (4.1 per cent), Lithuania (3.8 per cent), and Poland (3.4 per cent) (seeTables 6a/b).IAB 22

Table 6a: EU-15 emigrants from the NMS-8, NMS-2 and CAND-6, 2000-2007Sending country 2000 2001 2002 2003 2004 2005 2006 2007in personsCzech Republic 42,379 52,810 58,138 71,119 62,894 71,185 90,952 104,442Estonia 18,458 20,924 22,639 26,699 26,746 30,567 32,885 36,735Hungary 84,976 94,905 98,492 94,274 91,961 102,158 105,939 132,582Latvia 21,713 19,309 22,184 24,632 24,194 32,920 42,119 42,547Lithuania 24,154 36,567 41,577 53,572 52,613 85,364 114,185 128,361Poland 476,229 531,986 545,072 576,939 606,442 757,252 992,924 1,297,647Slovak Republic 25,195 36,947 39,019 43,948 52,343 81,705 91,560 132,207Slovenia 23,814 30,697 31,218 35,672 32,355 34,698 34,395 35,848NMS-8 716,917 824,145 858,338 926,854 949,548 1,195,850 1,504,957 1,910,370Bulgaria 71,437 102,980 140,864 166,330 203,528 219,233 255,163 310,335Romania 217,669 285,075 389,045 553,508 724,697 880,738 1,072,307 1,553,276NMS-2 289,106 388,054 529,909 719,839 928,225 1,099,971 1,327,470 1,863,610Albania 412,915 434,002 514,291 581,605 670,751 717,450 743,485 805,416Bosnia-Herzegovina 227,011 323,006 323,929 330,751 313,440 314,624 310,651 319,347Croatia 249,031 316,953 329,448 334,136 324,698 326,088 322,926 316,504Macedonia 83,848 103,932 112,922 137,863 146,209 153,059 161,556 171,450Serbia-Montenegro 679,548 835,178 806,739 777,571 342,551 521,495 508,255 471,764Turkey 2,453,924 2,541,316 2,526,518 2,378,011 2,299,713 2,302,017 2,263,013 2,241,489Cand-6 4,106,277 4,554,386 4,613,847 4,539,937 4,097,363 4,334,733 4,309,886 4,325,970Sources: National population statistics, Eurostat, LFS, own calculations and presentation.2000: without Austria; 2000-2001: without Luxembourg; 2000-2003: without Ireland2004-2007: Ireland included with structure of PPSNThese figures refer to migration stocks, which hide a large number of inflows andoutflows every year. The statistics of gross migration inflows and outflows in countriessuch as Germany or the large difference between gross figures on work permits in the UKand the actual number of foreigner workers there suggests that return migration issubstantial and has increased recently. As in other migration episodes, a high share ofmigration from the new member states is temporary. The relatively short distance andfalling communication and transport costs make it likely that the share of temporarymigration is higher in case of the NMS than in other migration episodes.IAB 23

Table 6b:EU-15 emigrants from the NMS-8, NMS-2 and CAND-6 in per cent of thehome population, 2000-2007Sending country 2000 2001 2002 2003 2004 2005 2006 2007share of total populationCzech Republic 0.41% 0.52% 0.57% 0.70% 0.62% 0.70% 0.89% 1.01%Estonia 1.35% 1.53% 1.67% 1.97% 1.98% 2.27% 2.45% 2.74%Hungary 0.83% 0.93% 0.97% 0.93% 0.91% 1.01% 1.05% 1.32%Latvia 0.91% 0.82% 0.95% 1.06% 1.05% 1.43% 1.84% 1.87%Lithuania 0.69% 1.05% 1.20% 1.55% 1.53% 2.50% 3.36% 3.80%Poland 1.24% 1.39% 1.43% 1.51% 1.59% 1.98% 2.60% 3.40%Slovak Republic 0.47% 0.69% 0.73% 0.82% 0.97% 1.52% 1.70% 2.45%Slovenia 1.20% 1.54% 1.57% 1.79% 1.62% 1.73% 1.71% 1.78%EU 8 0.96% 1.10% 1.15% 1.25% 1.28% 1.61% 2.03% 2.57%Bulgaria 0.87% 1.28% 1.79% 2.13% 2.62% 2.83% 3.31% 4.05%Romania 0.97% 1.29% 1.78% 2.55% 3.34% 4.07% 4.97% 7.21%EU 2 0.94% 1.29% 1.79% 2.43% 3.15% 3.74% 4.53% 6.38%Albania 13.49% 14.12% 16.63% 18.69% 21.45% 22.83% 23.56% 25.46%Bosnia-Herzegovina 6.02% 8.50% 8.48% 8.63% 8.16% 8.19% 8.08% 8.31%Croatia 5.57% 7.14% 7.41% 7.52% 7.31% 7.34% 7.27% 7.13%Macedonia 4.14% 5.11% 5.56% 6.80% 7.19% 7.51% 7.92% 8.39%Serbia-Montenegro 6.39% 7.84% 8.60% 9.57% 4.22% 6.47% 6.30% 5.85%Turkey 3.64% 3.72% 3.65% 3.39% 3.23% 3.19% 3.18% 3.20%Cand 6 4.49% 4.93% 5.01% 4.95% 4.42% 4.63% 4.65% 4.72%Sources: National population statistics, Eurostat, LFS, own calculations and presentation.2000: without Austria; 2000-2001: without Luxembourg; 2000-2003: without Ireland2004-2007: Ireland included with structure of PPSN4 The structure of migration: Skills, age and genderThe qualification structure of migrants from the NMS is concentrated about the mean.The migrant population from the NMS has a smaller share of less skilled workers than thenative population in the EU-15, but also a smaller share of high-skilled workers comparedto the native workforce in the EU-15. However, the migrant workforce from the NMS isbetter qualified compared to the native population which stayed behind in the NMS. Ingeneral, we observe a moderate ‘brain drain’ in the sending countries, but not a largeimpact on human capital endowments in the receiving countries of the enlarged EU. Incontrast, education levels of migrants from the candidate countries are well below thoseof natives in the receiving countries of the EU-15 (Section 4.1). Not surprisingly, the ageof migrants from the NMS is well below that of natives in the receiving and the sendingcountries. In those countries which have been heavily affected by the recent immigrationepisode from the NMS the age of migrants is particularly low. Although the age ofmigrants from the NMS will grow over time, the relatively high share of temporarymigrants will result in a younger age of the migrants from the NMS compared to otherimmigrant groups with a lower share of temporary migrants (Section 4.2).IAB 24

4.1 Skill structure4.1.1 Skill structure of immigrants from the NMS-8Figure 7 and Annex Table A2 display the skill structure of the migrants from the NMS-8 inthe EU-15 by their highest level of completed education. The LFS classification ofeducation levels is based on the ISCED classification. The data reported here areaggregated to three levels: Lower secondary education (low), upper secondary education(medium), and tertiary education levels (high). Note that education degrees are notcomparable across countries. Many education degrees are therefore not acknowledged.Moreover, misclassification in the LFS is widespread if education systems differ largelybetween receiving and sending countries.Figure 7:Skill structure of immigrants from the NMS-8 in the EU-15 compared toEU-15 natives, 2006100%skill group in per cent of immigrant working age population90%80%70%60%50%40%30%20%10%0%low medium high no answerAustriaBelgiumDenmarkFinlandFranceGermanyGreeceItalyLuxembourgNetherlandsSpainSwedenUnited KingdomTotal EU 15EU-15 NativesSource: European LFS, special provision 2008. Own calculations and presentation.With a share of 61 per cent the working age population from the NMS-8 in the EU-15 isheavily concentrated in the middle of the skill spectrum. Only 17 per cent of the NMS-8migrants belong to the less qualified group, compared to 27 per cent in the nativeworkforce in the EU-15. However, the share of the high-skilled is at 22 per cent of theworking age population of the NMS-8 immigrants slightly below that of natives in the EU-15 (27 per cent).The results for the individual EU-15 member states are however quite heterogeneous:Only a very small fraction of immigrants in Austria, Sweden, and the UK belong to theless qualified group (8-12 per cent), while that fraction is substantially higher in theIAB 25

other EU-15 countries (varying between 17 per cent in Denmark to 31 per cent inItaly). 11 The share of medium skilled immigrants differs considerably. While their fractionis relatively low in Denmark and France (17-20 per cent), it is extraordinarily high in theUK, Austria, the Netherlands, and Italy. However, measurement errors bias the results tothe mean. In the UK as an example, the category of ‘unknown education’ has beenclassified as medium education during the last survey years which has biased theeducation structure of the foreign population in one way or another.Austria, Belgium, Finland, and Germany each report that approximately 25 per cent ofthe NMS-8 immigrants are highly qualified, while Sweden and Spain have values of about45 per cent. Low shares of highly qualified immigrants from the NMS-8 are found inGreece, the Netherlands, and the UK (10-20 per cent). The extremely high values forDenmark, France and Luxembourg are based on low response rates and may thereforeresult from measurement or classification errors.Altogether, the skill structure of the workforce from NMS-8 countries in the EU-15 is highcompared to other foreigner groups. In almost all EU-15 countries the share of lessskilled workers in the immigrant workforce from the NMS-8 is below that of the nativeworkforce. Belgium, Finland and Germany are notable exceptions in this respect.However, the share of high skilled workers, i.e. workers with a university degree, is inmost receiving countries well below that of the native workforce.The figures presented above refer to the skill structure of the current stock of migrants,which has been accumulated both before and after EU enlargement. As a result of thenew immigration opportunities the skill structure of migrants may have changed in thecontext of Enlargement. The LFS allows to identify the year of arrival which enables us todisentangle the skill structure of migrants which have arrived before and afterenlargement. Low response rates restrict our analysis only on the main destinations, i.e.Austria, Germany and the UK.We find indeed that the skill structure of immigrants which have arrived afterenlargement deviates from that of the earlier vintages: In Germany, which has been themain destination before enlargement, we observe that the average education level of thenew arrivals from the NMS-8 has significantly deteriorated. Particularly the share of thegroup with a low educational degree has substantially increased in Germany. In contrast,that of NMS-8 immigrants in the UK has slightly improved. In Austria, the averageeducation level of the immigrants from the NMS-8 which have arrived after Enlargementare slightly higher than that of the groups which have arrived before Enlargement, butthe differences are within the range of measurement errors. At the level of the EU-15, weobserve a slight increase in the average education level of the NMS-8 immigrants sinceenlargement, particularly the share of the less-skilled immigrants has declined (seeFigure 8).11 The figures for Luxembourg are not plausible and may suffer from low response rates.IAB 26

Figure 8:Skill structure of NMS-8 immigrant cohorts which have arrived before andafter EU enlargement in the EU-15 and selected member states, 2006skill group in per cent of immigrant/native working age population100%90%80%70%60%50%40%30%20%10%0%AT_NMS8_preAT_NMS8_postAT_nativesGE_NMS8_preLow Medium High NoAnswerGE_NMS8_postGE_nativesUK_NMS8_preUK_NMS8_postUK_nativesEU-15_NMS8_preEU-15_NMS8_postEU-15_nativesSource: European LFS, special provision 2008. Own calculations and presentation.4.1.2 Skill structure of immigrants from Bulgaria and RomaniaThe average education level of the native population in Bulgaria and Romania is belowthat of natives in the NMS-8. The skill structure of the working age population from theNMS-2 in the EU-15 reflects this lower education level of the native population in thesending countries: About 29 per cent of the immigrant population in working age fromthe NMS-2 belong to the less-educated skill group, compared to 17 per cent in theworkforce from the NMS-8 and 27 per cent in the native workforce of the EU-15. At theupper end of the skill spectrum, about 18 per cent of the NMS-2 immigrants belong tothe high-skilled group, compared to 22 per cent in the NMS-8 workforce and 27 per centin the native workforce of the EU-15.However, in the main destinations of the NMS-2 migrants, Spain and Italy, the share ofless- and high-skilled workers in the NMS-2 workforce is well below that of the nativepopulation there. Altogether, immigration from the NMS-2 has a similar impact asimmigration of the NMS-8 on the skill structure of the workforce in the main destinations:It increases the labour supply more than proportional at the medium levels of the skillspectrum, but less than proportional both at the lower and the upper end of the skillspectrum. In Greece, where immigration from the NMS-2 is important in relative terms,we observe a similar pattern (see Figure 9 and Table A2).IAB 27

Figure 9:Skill structure of immigrants from Bulgaria and Romania in the EU-15compared to EU-15 natives, 2006100%skill group in per cent of immigrant working age population90%80%70%60%50%40%30%20%10%0%low medium high no answerAustriaBelgiumFinlandFranceGermanyGreeceItalyNetherlandsPortugalSpainSwedenUnited KingdomTotal EU-15EU-15 NativesSource: European LFS, special provision 2008. Own calculations and presentation.4.1.3 Skill structure of immigrants from the candidate countriesThe qualification structure of the working age population from the six candidate andpotential candidate countries in the EU-15 displays a completely different pattern thanthat of the NMS immigrant workforce: 53 per cent belong to the less qualified group, 41to the medium qualified group and only 6 per cent to the highly qualified educationgroup. Immigration from these countries has a long history in the EU and reflects interalia the recruitment of manual workers during the 1960s and early 1970s, which leavesits traces in the skill structure of the immigrant workforce from there until today.In the main destinations of migrants from these countries, i.e. in Germany, Austria, Italyand the Netherlands, the share of the less-skilled in the working age population from thecandidate countries varies between 40 and 60 per cent, compared to 14 to 20 per cent inthe native population of the receiving countries with the exception of Italy (39 per cent).The share of the high skilled varies between 5 and 7 per cent and is thus well that of thenative population (see Figure 10 and Table A2).Altogether, the average education level of the workforce from the candidate countries iswell below that of the native labour force in the receiving countries. This is true for boththe traditional destinations such as Germany and Austria as well as new destinationssuch as Italy and Greece.IAB 28

Figure 10:Skill structure of immigrants from the candidate countries in the EU-15compared to EU-15 natives, 2006100%skill group in per cent of immigrant working age population90%80%70%60%50%40%30%20%10%0%low medium high no answerAustriaBelgiumDenmarkFinlandFranceGermanyGreeceItalyLuxembourgNetherlandsSpainSwedenUnited KingdomTotal EU 15EU-15 NativesSource: European LFS, special provision 2008. Own calculations and presentation.4.2 Does Eastern enlargement involve a brain drain?The average education level of the migrant workforce from the NMS-8 residing in the EU-15 is well above that of the native workforce staying behind. In the NMS-8, the share ofthe high-skilled segment of the workforce is at 22 per cent more than twice as high asthat of the native workforce (9 per cent), while the share of the less-skilled group is at 17per cent well below that of the native workforce in the sending countries (21 per cent). Incontrast, average education levels of the migrant workforce from the NMS-2 are notabove those of the native population: The share of the high-skilled group is about 18 percent of the migrant workforce from the NMS-2 residing in the EU-15 compared to 20 percent in the native working age population in the sending countries. Analogously, about 28per cent of the migrants from the NMS-2 belong to the less-skilled group, but only 17 percent of the native population in working age in the NMS-2. However, these figures haveto be taken with a grain of salt since survey results from the sending and receivingcountries are biased due to classification and measurement errors (see Figure 11 andTable A3).The results for the individual sending countries differ widely. The Labour Force Surveysuggests that the migrant workforce from the Czech Republic, Hungary, Bulgaria andPoland is particularly high skilled compared to the native population, while the skill levelof the migrant working age population is below the native population in case of Romaniaand Slovenia.IAB 29

Figure 11:Skill structure of immigrants in the EU-15 by country of origin comparedto NMS-8 and NMS-2 natives, 2006skill group in per cent of immigrant working age population100%90%80%70%60%50%40%30%20%low medium high no answer10%0%Czech RepublicEstoniaHungaryLatviaLithuaniaPolandSlovak RepublicSloveniaBulgaria*Notice: Figures for Serbia-Montenegro from 2005 - 2006 n.a.RomaniaAlbaniaBosniaCroatiaMazedoniaSerbia & MontenegroTurkeyNMS-8NMS-2Cand-6NMS-8 NativesNMS-2 NativesSource: European LFS, special provision 2008. Own calculations and presentation.4.3 Is there evidence for brain waste?To illustrate the issue of possible brain waste among migrants (compared to the nativepopulation), we restrict the analysis to the group of high-skilled persons in order toinvestigate whether and to what extent these highly qualified individuals work in jobsthat would generally require only medium or low qualification (see Table A4 for mediumskilled persons). We expect to see high-skilled individuals working in jobs requiring ahigh level of education; hence there should be an accumulation of individuals working asprofessionals or managers and only a minority of individuals working in fields such asagriculture, crafts or machine operating. As the dataset gives only a loose overview of theoccupational structure we refer to the occupational structure of natives in order toidentify different employment patterns. Hence, Table 7 describes the occupationalstructure of employed individuals within the EU-15 of foreigners and natives for the year2006.For our analysis we refer to data based on the International Standard Classification ofOccupations, ISCO, which enables us to distinguish the basic occupational fields in whichan individual works. We drop the ‘Armed Forces’ category due to missing values for theNMS-8, NMS-2, and Cand-6 group. There are two classes used in the table: ‘>10years’,i.e. immigrants that lived in the host country for more than ten years, and ’newlyarrived’, the group of persons that moved to the host country within the last decade.IAB 30

Table 7:Occupational structure of highly skilled employed individuals bymigration status in the EU-15, 2006immigrants from NMS-8 immigrants from NMS-2 immigrants from CAND-6> 10years newly arrived > 10years newly arrived > 10years newly arrivednativesClerks 0.7 8.9 0.9 3.4 17.6 3.3 8.1Craft and related tradeworkers15.0 12.6 15.8 32.6 13.5 20.0 3.8Elementary occupations 9.8 9.2 4.9 20.9 10.8 18.5 1.2Legislators, senior officials andmanagersPlant and machine operatorsand assemblers5.0 10.8 1.2 1.6 5.5 9.9 12.60.4 2.7 27.2 2.4 5.9 11.9 1.4Professionals 23.8 21.1 29.2 8.0 21.4 18.5 42.8Service workers and shop andmarket sales workersSkilled agricultural and fisheryworkersTechnicians and associateprofessionalsin per cent of highly skilled employed individuals aged 15-648.7 16.7 2.6 19.8 7.2 8.6 5.10.2 0.0 0.0 0.4 0.7 2.6 0.836.6 17.9 18.1 11.0 17.4 6.6 23.7Total (in persons) 41,278 91,350 15,673 109,696 57,800 33,373 42,512,500Results for immigrants can be biased due to measurement and classfication errors.-- Figures need not add up to 100 per centsince the category 'armed forces' is not reported here.Source: European LFS, special provision 2008. Own calculations and presentation.Using natives’ occupational structure as reference, which is characterised by a high shareof professionals (43 per cent), technicians (24 per cent) and legislators (13 per cent),sizeable differences between natives and foreigners become apparent. However, thesedifferences vary also between newly arrived immigrants and those who are in therespective country for more than 10 years. It is obvious that foreigners, independent oftheir origin, work more often in occupations which require only elementary skills (craftand related trade workers, elementary occupations, plant and machine operators, serviceworkers and shop and market sales workers). Moreover, the group which stays more than10 years is less represented in these occupational groups than the new arrivals.4.4 Changing the age structure of the workforceOne important feature of the recent migration wave from the new member states is thatthe immigrant population from the NMS is particularly young. Almost two-thirds (63 percent) of the working age population from the NMS-8 in the EU-15 belongs to the agegroup from 15 to 34 years, compared to 58 per cent in the immigrant workforce from theNMS-2 and 34 per cent of the native workforce in the EU-15. This can be traced back tothe fact that immigration from the NMS has started only recently. In countries likeAustria and Germany, where immigration from the NMS began already in the early1990s, the share of the 15 to 34 age group among the working age population from theNMS-8 is at 37 per cent and 49 per cent, respectively, well below that of the UK (86 percent). Due to the long migration tradition, the working age population from the candidatecountries in the EU-15 is much older than the immigrant workforce from the NMS: TheIAB 31

share of the 15 to 34 age group of the CAND-6 amounts to 46 per cent on average. Thisshare is still higher than among the native working age population, but considerablysmaller than in the workforce from the NMS (see Table 8).Table 8:Age composition of the working age population by migration status in theEU-15, 2006immigrants from NMS-8 immigrants from NMS-2 immigrants from CAND-6natives15-34 35-49 50-64 15-34 35-49 50-64 15-34 35-49 50-64 15-34 35-49 50-64in per cent of working age populationAustria 36.8 38.8 24.4 70.5 24.7 4.8 45.4 41.2 13.4 35.5 45.3 19.2Belgium 64.1 31.2 4.7 74.9 18.4 6.7 n.a. n.a. n.a. 34.6 45.1 20.3Denmark 85.0 5.7 9.4 n.a. n.a. n.a. 66.7 28.2 5.0 34.8 37.5 27.7Finland 44.0 49.5 6.4 39.8 41.9 18.3 42.3 47.8 9.9 32.6 38.1 29.3France 46.2 45.4 8.4 68.2 20.2 11.6 53.2 36.0 10.8 34.0 42.6 23.3Germany 48.7 36.5 14.9 55.5 32.9 11.6 44.3 38.2 17.5 30.3 44.9 24.8Greece 38.3 52.1 9.6 41.1 45.6 13.3 46.4 44.7 8.9 34.2 42.4 23.4Ireland n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. 43.3 35.3 21.4Italy 58.3 28.7 12.9 60.3 33.2 6.5 46.9 47.9 5.2 32.2 45.6 22.2Luxembourg 74.5 19.5 6.0 36.7 48.7 14.6 44.8 52.2 3.0 29.6 48.2 22.3Netherlands 53.8 40.6 5.6 0.0 0.0 0.0 52.9 40.9 6.2 37.5 39.6 22.8Portugal n.a. n.a. n.a. 56.7 38.9 4.3 n.a. n.a. n.a. 36.7 39.7 23.6Spain 70.9 21.6 7.5 57.3 35.1 7.6 n.a. n.a. n.a. 38.5 40.3 21.2Sweden 35.8 43.9 20.3 71.5 19.8 8.7 49.1 44.3 6.7 32.3 36.8 30.8United Kingdom 86.0 11.2 2.9 79.8 13.8 6.4 51.4 45.1 3.5 34.5 39.3 26.3Total EU 15 63.0 26.8 10.2 58.3 33.9 7.8 45.8 40.8 13.3 33.9 42.2 23.9Results for immigrants can be biased due to measurement and classfication errors.-- Figures need not add up to 100 per centsince the category 'no answer' is not reported here.Source: European LFS, special provision 2008. Own calculations and presentation.From the sending country perspective, the share of emigrants in the young cohortsincreases with the share of people which have emigrated during the last years: The shareof the 15 to 34 cohort in the migrant population is particularly high in Lithuania, Latvia,the Slovak Republic, the Czech Republic and Poland, i.e. in case of sending countries forwhich migration barriers have been recently removed. In the successor states of theformer Yugoslavia, where emigration has started already during the guestworkerrecruitment phase in the 1960s and accelerated during the civil wars in the 1990s, theaverage age of the emigrant population is high compared to the other sending countries(see Table 9).Altogether, the migrant workforce from the NMS is particularly young, which reduceslabour supply in the young cohorts substantially in the sending countries and increases itin the main destinations such as the UK and Ireland. Of course, the age of the workforcefrom the new member states will increase over time. The higher share of temporarymigration which is facilitated by the migration opportunities within the EU and thegeographical proximity may however result in a higher labour mobility among the youngcohorts of the labour force from the NMS and, hence, a lower average age of the migrantworkforce from the NMS in the EU-15 compared to other immigrant groups even in thelong-run.IAB 32

Table 9:Age composition of the working age population by migration status in thesending countries, 2006EU-15 emigrantsage group 15-34 35-49 50-64 15-34 35-49 50-64in per cent of working age populationCzech Republic 65.4 21.5 13.1 14.5 79.7 5.8Estonia 54.0 41.6 4.4 35.7 53.6 10.7Hungary 43.4 33.4 23.2 21.3 65.3 13.4Latvia 70.6 22.7 6.7 23.9 62.3 13.7Lithuania 77.6 15.2 7.2 31.2 61.0 7.9Poland 64.7 27.2 8.1 22.4 68.4 9.2Slovak Republic 69.3 23.2 7.6 16.6 78.7 4.7Slovenia 38.9 31.4 29.7 23.4 62.2 14.4Bulgaria 57.4 32.8 9.8 25.2 59.3 15.6Romania 58.4 34.4 7.2 14.2 63.9 21.9Albania 48.7 44.5 6.9 n.a. n.a. n.a.Bosnia-Herzegovina 37.6 42.6 19.8 n.a. n.a. n.a.Croatia 35.8 34.8 29.4 35.7 53.6 10.7Mazedonia 39.2 57.1 3.7 n.a. n.a. n.a.Serbia-Montenegro 44.6 31.3 24.1 n.a. n.a. n.a.Turkey 50.1 39.4 10.5 n.a. n.a. n.a.NMS-8 63.0 26.8 10.2 37.7 40.0 22.4NMS-2 58.2 34.1 7.7 36.5 41.0 22.5Cand-6 45.7 39.6 14.7 n.a. n.a. n.a.Results for immigrants can be biased due to measurement and classfication errors.--Figures need not add up to 100 per cent since the category 'no answer' is notreported here.Source: European LFS, special provision 2008. Own calculations and presentation.natives4.5 Gender patternsAnother feature of the recent immigration wave from the new member states is therelatively high participation of females in the migrant labour force. Table 10 displays theshare of females in the labour force of immigrants from the NMS-8, NMS-2, the CAND-6and natives in the EU-15. The share of females in the native labour force is below 50 percent in all EU-15 countries and particularly low in Italy and Greece. In the labour forcefrom the NMS-8 we observe a share of females of 51 per cent, which is considerablyhigher compared to the native labour force. The share of females in the NMS-2 labourforce in the EU-15 is at 47 per cent lower than that in the NMS-8 labour force, but stillhigher than that of females in the native labour force of the EU-15 (45 per cent).However, the LFS data reports for some countries implausible high shares of females inthe immigrant labour force from the NMS, such that we have to take these results with agrain of salt (see Table 10).In the labour force from the candidate countries we observe a different gender pattern:The share of females is at 34 per cent much lower than among the native and theIAB 33

immigrant labour force from the NMS, which reflects both a lower participation of femalesin the migrant population from these countries and a lower labour market participation offemales from the candidate countries residing in the EU-15.Altogether, the relatively high share of females in the immigrant labour forcedemonstrates that labour mobility from the new member states deviates from thebreadwinner model which influences migration patterns and the female labour marketparticipation in many migrant groups until today.Table 10: Share of females in labour force by migrant status in the EU-15, 2006NMS-8 NMS-2 CAND-6 nativesin per cent of labour forceAustria 46.6 56.5 35.9 45.6Belgium 63.4 53.2 n.a. 44.5Denmark n.a. n.a. 45.9 46.8Finland n.a. n.a. n.a. 48.6France 74.4 n.a. 17.2 46.8Germany 54.5 63.4 36.9 46.3Greece 55.3 58.9 30.1 39.2Italy 81.7 40.9 26.8 39.7Luxembourg 69.1 n.a. n.a. 43.3Netherlands 69.3 n.a. 34.1 45.1Portugal n.a. n.a. n.a. 46.3Spain 48.3 47.7 53.0 40.2Sweden n.a. n.a. n.a. 47.5United Kingdom 43.4 n.a. 23.9 47.0EU-15 51.3 47.4 33.5 44.7Source: European LFS, special provision 2008. Own calculations and presentation.5 Unemployment and labour market participationThis section deals with the labour market status of immigrants from the new memberstates and the candidate countries in the EU-15. We distinguish between employed,unemployed, and inactive persons in the working age population based again on theinformation provided by the European LFS. The employment share of the immigrantpopulation in working age from the NMS-8 is at 68 per cent similar to that of natives (67per cent). Interestingly enough, the LFS reports a considerably higher employment sharefor the working age population from the NMS-2 in the EU-15 (74 per cent). The share ofunemployed individuals in the working age population 12 from the NMS is at some 8.5 percent somewhat higher compared to the native population in the EU-15 (5.2 per cent).The inactivity rate is at 18 per cent (NMS-2) and 24 per cent (NMS-8) well below that of12 Note that the share of unemployed in the total working age population is not comparable to theunemployment rate, which is usually defined as the share of unemployed in the civil labourforce.IAB 34

natives in the EU-15 (28 per cent), which reflects inter alia the lower age of the migrantpopulation from the NMS. In contrast, the working age population from the candidatecountries shows a substantially higher share of inactive (36 per cent) and unemployedpersons (11 per cent) (see Table 11). The differences in the labour market performancebetween immigrants from the new member states and the candidate countries reflectboth to other demographic characteristics and differences in education levels which havebeen described above.Table 11:Employment, unemployment and inactivity by migrant status in EU-15,2006immigrants from NMS-8 immigrants from NMS-2 immigrants from CAND-6nativesemployedunemployedinactiveemployedunemployedinactiveemployedunemployedinactiveemployedunemployedinactiveAustria 68.7 6.6 24.8 60.0 5.6 34.3 61.7 7.5 30.8 70.8 3.1 26.2Belgium 52.9 3.5 43.6 53.6 19.2 27.2 27.8 12.7 59.5 61.5 5.0 33.4Denmark n.a. n.a. n.a. n.a. n.a. n.a. 60.6 5.0 34.4 77.3 3.2 19.6Finland 71.7 2.9 25.4 60.0 18.7 21.3 54.5 15.5 30.0 70.2 6.8 23.1France 61.4 6.8 31.7 27.8 34.0 38.2 43.4 9.6 47.0 64.5 5.7 29.8Germany 58.6 12.3 29.1 60.8 8.8 30.5 49.5 13.5 37.1 68.9 7.1 24.0Greece 69.9 1.6 28.4 78.4 6.6 15.0 69.0 5.2 25.8 60.6 6.0 33.4Ireland n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. 67.7 2.9 29.3Italy 57.4 3.7 38.9 73.3 8.2 18.4 62.7 8.4 28.9 58.4 4.0 37.6Luxembourg 79.1 3.5 17.4 74.2 6.1 19.7 62.8 12.3 24.9 60.9 1.9 37.2Netherlands 58.4 5.6 36.0 53.5 5.7 40.9 48.8 5.8 45.4 75.0 2.9 22.1Portugal n.a. n.a. n.a. 72.5 14.7 12.8 n.a. n.a. n.a. 68.0 5.6 26.4Spain 75.6 8.1 16.3 77.3 8.1 14.6 85.6 9.2 5.2 63.9 5.6 30.4Sweden 60.9 15.9 23.3 n.a. n.a. n.a. 46.0 12.8 41.1 74.0 6.2 19.8United Kingdom 82.3 5.4 12.4 84.1 8.2 7.7 41.1 8.3 50.6 71.7 3.9 24.4Total EU 15 67.6 8.5 23.9 73.7 8.5 17.8 53.1 11.2 35.8 66.5 5.2 28.2Results for immigrants can be biased due to measurement and classfication errors.-- Figures need not add up to 100 per centsince the category 'no answer' is not reported here.Source: European LFS, special provision 2008. Own calculations and presentation.in per cent of working age populationThe labour market performance of migrants from the new member states and thecandidate countries varies considerably across destination countries. Employment sharesof migrants from the NMS are particularly high in the UK, Luxembourg, Spain, Greece,and Italy. Note that immigration from the NMS is a recent phenomenon in thesecountries. Moreover, the UK has restricted the access to unemployment benefits formigrants from the NMS. In contrast, employment shares are particularly low in Belgium,France, Germany, and the Netherlands. Note again that particularly the country resultsmay suffer from low response rates in the LFS.Figure 12 compares the labour market performance of immigrants from the NMS-8 whichmoved before and after EU enlargement for the EU-15 and selected destinations. At theEU-15 average, the immigrant cohorts which arrived after EU enlargement arecharacterised by a higher employment and a lower inactivity ratio compared to thecohorts which arrived before enlargement. Nonetheless, the picture differs by destinationcountries. In Austria we find a high employment share among the pre-enlargementmigrants and in Germany the employment ratio is roughly the same for both groups. Incontrast, the post-enlargement cohorts outpace the employment share of theirpredecessors by far in the UK. These differences in the labour market performance mayIAB 35

eflect different migration patterns: While immigration in the UK is largely driven by theopening of the labour markets, the main channels for permanent migration from the NMSto Germany are family reunification.Figure 12:Employment, unemployment, and inactivity of NMS-8 immigrant cohortswhich have arrived before and after EU enlargement in the EU-15 andselected member states, 2006share in per cent of immigrant/native working age population100%90%80%70%60%50%40%30%20%10%0%AT_NMS8_preAT_NMS8_postAT_nativesGE_NMS8_preGE_NMS8_postemployed unemployed inactiveGE_nativesUK_NMS8_preUK_NMS8_postUK_nativesEU-15_NMS8_preEU-15_NMS8_postEU-15_nativesSource: European LFS, special provision 2008. Own calculations and presentation.Table 12 compares the employment, unemployment and inactivity shares of the workingage population of the EU-15 migrants with that of natives in the sending countries. Notethat the observable and unobservable human capital characteristics of the migrantpopulation differ from those of the native population, such that this does not provideinformation on the labour market performance of individuals in the home and the hostcountry. According to the LFS data, the inactivity rate of the migrant population of theNMS-8 and the NMS-2 is at 24 per cent and 18 per cent, respectively, well below that ofthe native population in the NMS-8 (35 per cent) and the NMS-2 (36 per cent), while theunemployment rate is slightly higher. The employment rates of the migrant populationare in most sending countries well above those of the native population. The higheractivity of the migrant population relative to the native population is not surprising, sincethe age is substantially lower and the education levels are usually higher compared to thenative population. Moreover, specific characteristics of the migrant population may play arole here. Interestingly enough, this pattern does not hold for all sending countries: Theemployment rates of migrants from the successor states of the former Yugoslavia are onaverage below those of the new member states, and that of Turkey are at 46 per centparticularly low. Although data on the labour market participation of natives in the homecountries are not available for most of these countries, these figures suggest thatemployment shares of the migrant population may be below those of natives in thesending countries.IAB 36

Table 12:Employment, unemployment and inactivity of EU-15 emigrants andnatives in the sending countries, 2006EU-15 emigrantsnativesemployedunemployedinactiveemployedunemployedinactivein per cent of working age populationCzech Republic 63.8 8.7 27.5 65.2 5.0 29.8Estonia 73.1 2.7 24.2 68.7 4.0 27.3Hungary 72.6 9.3 18.2 57.3 4.5 38.2Latvia 70.3 13.7 16.0 65.4 5.2 29.4Lithuania 75.7 9.6 14.7 63.6 3.8 32.5Poland 67.1 8.0 24.9 54.0 9.0 37.0Slovak Republic 64.4 8.5 27.1 59.2 9.3 31.5Slovenia 65.7 9.7 24.6 67.2 4.3 28.6Bulgaria 71.3 8.0 20.7 59.1 5.8 35.0Romania 74.3 8.7 17.1 59.6 4.8 35.6Albania 64.7 6.7 28.6 n.a. n.a. n.a.Bosnia-Herzegovina 60.1 10.3 29.5 n.a. n.a. n.a.Croatia 63.5 10.1 26.4 68.7 4.0 27.3Mazedonia 62.2 7.5 30.3 n.a. n.a. n.a.Serbia-Montenegro 61.0 7.0 32.1 n.a. n.a. n.a.Turkey 45.6 13.1 41.3 n.a. n.a. n.a.NMS-8 67.7 8.4 23.9 57.8 7.3 34.9NMS-2 73.6 8.5 17.8 59.5 5.0 35.5Cand-6 53.6 10.9 35.5 n.a. n.a. n.a.Results for immigrants can be biased due to measurement and classification errors.--Figures need not add up to 100 per cent since the category 'no answer' is notreported here.Source: European LFS, special provision 2008. Own calculations and presentation.Altogether, the labour market performance of the migrants from the new member statesreflects both their relatively high education level and low age compared to other foreignergroups. Employment and activity rates are above the native population in the receivingand the sending countries. However, the share of unemployed individuals in the workingage population is higher in the migrant population from the NMS compared to those ofnatives in the receiving countries.6 ConclusionsThis background report has described the main migration patterns and fundamentaleconomic conditions which may have contributed to the migration from the new memberstates and the candidate countries in the context of the EU eastern enlargement. We findthat differences particularly in nominal per capita GDP and wage levels create substantialmonetary incentives for migration, although the fast convergence of per capita GDP andIAB 37

wage levels erodes these incentives over time. The speed of wage and incomeconvergence is faster in the new member states compared to the candidate countries andhas considerably accelerated after enlargement.The removal of immigration barriers in selected EU countries is associated with asubstantial increase in migration from the NMS-8 into the EU-15 since 2004 and adiversion of migration flows away from Austria and Germany towards Ireland and the UK.An annual increase in the stock of migrants of some 254,000 persons p.a. corresponds tothe forecasts of potential migration from the NMS-8 into the EU-15 which have beenundertaken before enlargement (e.g. Alvarez-Plata et al., 2003), although the large influxof migrants to destinations such as Ireland and the UK has not been expected.Immigration from Bulgaria and Romania has substantially increased and amounts tosome 226,000 persons p.a. since the beginning of this decade, although most EUmember states have maintained their immigration restrictions vis-à-vis both countriesafter their accession in 2007. This substantial influx has been mainly facilitated bybilateral agreements between Spain and Italy and the NMS-2.The 2007 immigration data suggest that net immigration from the NMS-8 into the EU-15starts to decline, which would coincide with standard migration patterns (Brücker/Schröder, 2006), while net immigration from Bulgaria and Romania remains at the levelsof the previous years. Immigration from the candidate countries, which have been one ofthe main sources of immigration in the EU-15 during the last decades, however stagnatessince the beginning of this decade.The influx of migrants from the NMS will not much change the skill structure of theworkforce in the receiving countries of the EU-15 since they are, similar to the nativepopulation, mainly concentrated at the mean of the skill spectrum However, the shares oflow- and high-skilled workers from the NMS are slightly below those of the nativeworkforce. The average education level of migrants from the NMS-2 are slightly belowthose of the NMS-8, but compared to the education level of the native workforce in themain destinations of immigrants from the NMS-2 a similar pattern as in case ofimmigration from the NMS-8 emerges. This distinguishes the skill level of the workforcefrom the new member states in the EU-15 from that of other immigrant groups, whichare characterised by lower education levels compared to the native workforce.Comparing the skill structure of migrants with natives in the sending countries, we findthat migrants from the NMS-8 are better qualified than natives in their home countries.Particularly the share of the high-skilled group in the migrant workforce is more thantwice as high as that of the native workforce of the NMS-8. A brain drain may thereforebe an issue for some of these countries, although recent research suggests that sendingcountries can benefit from high-skilled emigration if it is temporary and involvesadditional human capital investment in the sending countries. This will be discussed indetail in Deliverable 6 of this study.Eastern enlargement is associated with an improvement of the skill structure of theworkforce from the NMS-8 if we compare migration cohorts which arrived before andIAB 38

after enlargement. An explanation of this phenomenon is not self-evident, since lowermigration barriers are often associated with a lower skill level of the migrant population(Belot/Hatton, 2008; Brücker/Defoort, 2006). A possible explanation is that theregulation of immigration from the NMS by family reunification and seasonal workpermits has reduced the skill level of the workforce which immigrated beforeenlargement.The immigrant workforce from the NMS is particularly young compared to the nativeworkforce in the receiving countries but also compared to the workforce of otherimmigrant groups. The age pattern of the migrant workforce reflects the fact thatimmigration from the NMS is in many countries a very recent phenomenon. The age ofthe immigrant workforce from the NMS will therefore increase over time. Nevertheless,geographical proximity, low transport and communication costs create together with thefree movement of workers in the EU special incentives for temporary migration, whichwill be more than proportionally utilised by the young cohorts in the labour market. It istherefore likely that the age of the workforce from the NMS will remain below that ofnatives and other immigrant groups in the long-run.The low age of the workforce from the NMS creates benefits for the public sector in thereceiving countries and costs in the sending countries. Over the life-cycle, individualscontribute in the age brackets where the migrants from the NMS are more thanproportionally represented much more to the public sector by taxes and social securitycontributions than they receive in terms of transfers. This generates a net gain for thepublic sector in the receiving countries, particularly if migration is temporary. This will bediscussed further in Deliverable 5.The patterns of labour market participation of migrants from the NMS in the EU-15 reflecttheir human capital characteristics. Inactivity rates are particularly low compared to thenative workforce as well as compared to other immigrant groups, and employment ratesare relatively high. This is not surprising given the low age of the immigrant workforceand the small share of less-skilled workers in the immigrant workforce from the NMS.However, the share of unemployed individuals is considerably higher compared to thenative workforce, reflecting problems of labour market integration of immigrants. It ishowever worthwhile to note that the unemployment risks of migrants from the NMS arelower than those of other immigrant groups. We observe moreover distinct differencesbetween destination countries, reflecting different labour market conditions andinstitutions as well as different modes of regulating the entry of immigrants from theNMS.IAB 39

7 ReferencesAlvarez-Plata, P., H. Brücker, and B. Siliverstovs (2003), Potential Migration from Centraland Eastern Europe into the EU-15 – An Update, Report for the EuropeanCommission, DG Employment, Social Affairs and Equal Opportunities, Brussels.Belot, M. and T. Hatton (2008), Immigration Selection in the OECD, Australian NationalUniversity Centre for Economic Policy Research Discussion Paper 581.Burda, M.C. (1995), Migration and the Option Value of Waiting, Economic and SocialReview, 27 (1), pp. 1-19.Boeri, T. and H. Brücker (2005), Why are Europeans so tough on migrants?, EconomicPolicy, Band 44, 621-703.Brücker, H. and C. Defoort (2006), The (Self-)Selection of International MigrantsReconsidered: Theory and New Evidence, IZA Discussion Paper 2052.Brücker, H. and P.J.H. Schröder (2006), “International Migration With HeterogeneousAgents: Theory and Evidence”, IZA-Discussion Paper 2049.Eurostat Labour Force Survey (2008), special provision, Luxembourg.Falk (2008), Online Route Planning, www.falk.de, 20 March, 2008.Harris, J.R. and M.P. Todaro (1970), “Migration, Unemployment and Development: ATwo-Sector-Analysis”, American Economic Review, Vol. 60, pp.126-142.OECD (2008), Benefits and Wages: Gross/net replacement rates, country specific filesand tax/benefit models (latest update: March 2006), online database,www.oecd.org.OPODO (2008), Online-booking system, www.opodo.com, 31 March, 2008.Sjaastad, L.A. (1962), The costs and returns of human migration, Journal of PoliticalEconomy, Vol. 70(5), pp. 80-93.Stark, O. (1991), The Migration of Labour, Oxford and Cambridge, MA: Basil Blackwell.World Bank (2007), World Development Indicators 2007, CD-Rom, Washington D.C.IAB 40

8 AnnexTable A1: Distance and transport costs, 2008AT BE DK DE FI FR GRE IE IT LX NL PT SWE SP UKdistance in kmBU 1,013 2,112 2,032 1,648 2,589 2,189 736 2,918 1,650 1,955 2,171 3,555 2,633 2,833 2,474CZE 330 915 981 355 1,500 1,050 1,980 1,713 1,300 750 890 2,790 1,350 2,320 1,226EST 1,754 2,185 1,758 1,463 20 2,530 3,289 2,950 2,825 2,185 2,088 4,224 50 3,757 2,470HRV 365 1,283 1,615 1,075 2,122 1,400 1,450 2,080 860 1,150 1,350 2,800 2,025 2,200 1,600HUN 250 1,356 1,508 886 1,814 1,490 1,426 2,154 1,208 1,192 1,407 3,113 1,870 2,554 1,667LAT 1,462 1,832 1,316 1,043 330 2,178 2,912 2,598 2,579 1,833 1,735 3,872 20 3,405 2,117LIT 1,170 1,841 1,620 1,050 630 2,132 2,616 2,600 2,240 1,840 1,744 3,884 355 3,414 2,126MAC 1,050 2,157 2,310 1,684 2,627 2,227 650 2,955 852 2,000 2,208 3,600 2,670 3,033 2,470POL 730 1,310 1,190 590 1,040 1,600 2,264 2,076 1,800 1,310 1,213 3,350 1,570 2,880 1,590ROM 1,060 2,170 2,320 1,700 2,040 2,300 1,122 2,970 1,874 2,000 2,220 3,780 2,680 3,220 2,490SVK 78 1,190 1,310 683 1,690 1,322 1,620 1,990 1,152 1,025 1,234 3,000 1,670 2,500 1,500SVN 374 1,190 1,522 1,000 2,111 1,250 1,590 1,980 720 987 1,242 2,623 1,950 2,065 1,500TK 2,040 3,145 3,300 2,671 3,200 3,214 1,500 3,950 2,675 2,980 3,200 4,600 3,660 4,020 3,455cost of road transport by car (EUR)BU 109 229 219 178 300 236 79 555 178 211 235 384 365 306 337CZE 36 99 106 39 182 113 214 425 140 81 96 300 225 250 202EST 189 236 190 158 20 273 355 558 305 236 226 456 50 405 336HRV 40 140 175 116 230 150 160 465 95 125 145 300 300 240 245HUN 27 147 163 96 216 161 154 472 131 129 152 336 280 276 250LAT 158 198 142 113 60 235 315 520 278 198 187 418 60 368 300LIT 126 199 175 113 90 230 282 520 240 198 188 419 100 368 230MAC 115 233 250 182 305 241 70 560 92 215 240 388 370 330 340POL 78 141 129 63 132 173 245 464 195 142 131 362 250 311 240ROM 114 234 250 183 240 250 121 560 202 216 240 408 370 350 270SVK 10 127 141 74 182 143 175 454 125 110 133 325 260 270 233SVN 40 129 165 108 250 135 171 453 78 106 134 283 290 223 233TK 220 340 356 289 385 350 163 665 289 320 345 495 480 435 443costs of air transport (EUR)BU 101 167 292 174 316 167 99 245 100 305 167 330 308 200 145CZE 169 106 110 200 124 195 142 104 340 620 106 198 140 140 90EST 156 119 80 125 49 159 246 127 141 247 142 500 61 250 112HRV 187 196 250 215 310 197 360 840 207 385 197 663 290 351 197HUN 221 103 315 148 325 130 224 160 140 265 147 211 290 122 129LAT 172 146 132 88 82 153 250 180 193 172 140 254 90 227 140LIT 168 110 124 100 82 165 250 168 160 168 110 183 105 220 124MAC 221 239 228 312 371 228 340 323 203 408 239 460 340 340 165POL 90 127 72 108 132 216 249 115 179 411 127 254 143 105 127ROM 165 333 188 327 290 195 125 160 180 430 311 223 293 170 145SVK n.a. 157 166 155 328 90 174 176 159 705 149 652 338 243 208SVN 176 230 253 274 357 231 330 363 310 395 171 640 368 368 343TK 296 265 250 209 493 330 208 240 330 325 204 388 250 325 134Distance refers to the distance between capitals in km.-- Travelling costs by car are computed by using the Falk-routing planer,calculated for the fastest route. Travelling costs refer to one person per car and include ferry fares.-- Air transport costs aretaken from OPODO for the cheapest carrier, booking one week before travelling.Sources: Own calculations using the Falk-route planning system and the OPODO booking system.IAB 41

Table A2:Skill composition of the working age population by migration status in theEU15, 2006immigrants from NMS-8 immigrants from NMS-2 immigrants from CAND-6nativeslow medium high low medium high low medium high low medium highshare in per cent of working age populationAustria 7.7 67.8 24.4 20.0 64.8 15.2 41.4 53.6 4.9 16.1 65.9 18.0Belgium 26.5 45.3 28.2 21.7 30.4 47.9 43.7 43.8 12.5 23.7 38.7 37.6Denmark 16.6 20.7 62.7 n.a. n.a. n.a. 62.6 31.7 5.8 20.1 47.3 32.5Finland 25.6 50.0 24.4 18.3 41.9 39.8 55.9 24.3 19.8 17.6 47.1 35.3France 23.3 17.5 59.2 29.1 40.6 30.4 75.9 17.8 6.3 24.8 45.5 29.7Germany 19.8 53.3 26.9 17.8 59.4 22.8 52.0 41.8 6.1 13.5 60.5 26.0Greece 30.5 54.0 15.5 43.2 47.2 9.5 55.1 36.9 8.0 34.7 39.3 25.9Ireland n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. 26.6 40.1 32.3Italy 30.7 60.3 9.0 27.0 66.4 6.6 59.2 35.6 5.2 39.0 45.8 15.3Luxembourg 2.7 16.6 80.7 n.a. n.a. n.a. 23.5 68.9 7.6 26.0 50.3 23.7Netherlands 18.4 60.5 18.8 25.3 27.3 47.4 45.5 41.6 9.6 25.9 43.5 30.2Portugal n.a. n.a. n.a. 12.2 80.5 7.3 n.a. n.a. n.a. 70.1 15.4 14.5Spain 19.5 33.3 47.2 31.1 45.6 23.3 5.9 53.1 41.0 44.9 21.5 33.5Sweden 11.6 43.4 45.0 26.7 44.8 28.5 39.3 48.5 7.5 14.4 55.2 29.7United Kingdom 12.4 76.0 9.5 18.1 65.7 13.7 33.8 51.7 10.7 22.9 44.9 31.5Total EU 15 16.9 60.8 21.5 28.8 53.1 18.0 52.7 40.6 6.4 27.2 45.4 27.2Results for immigrants can be biased due to measurement and classfication errors.-- Figures need not add up to 100 per centsince the category 'no answer' is not reported here.Source: European LFS, special provision 2008. Own calculations and presentation.Table A3:Skill composition of the working age population in the sendingcountries by migration status, 2006EU-15 emigrantsnativeslow medium high low medium highin per cent of working age populationCzech Republic 13.9 48.3 35.6 14.5 79.7 5.8Estonia 24.1 59.0 16.9 35.7 53.6 10.7Hungary 8.4 65.0 26.6 21.3 65.3 13.4Latvia 4.9 83.8 11.3 23.9 62.3 13.7Lithuania 19.0 62.9 15.9 31.2 61.0 7.9Poland 17.6 60.0 21.7 22.4 68.4 9.2Slovak Republic 18.2 65.9 15.8 16.6 78.7 4.7Slovenia 23.8 67.0 9.2 23.4 62.2 14.4Bulgaria 20.1 48.2 31.3 25.2 59.3 15.6Romania 30.4 55.5 14.1 14.2 63.9 21.9Albania 57.3 36.2 6.5 n.a. n.a. n.a.Bosnia-Herzegovina 43.1 48.4 8.2 n.a. n.a. n.a.Croatia 30.5 57.7 11.2 35.7 53.6 10.7Mazedonia 6.7 32.9 60.2 n.a. n.a. n.a.Serbia-Montenegro 8.8 43.3 47.9 n.a. n.a. n.a.Turkey 58.8 36.5 4.6 n.a. n.a. n.a.NMS-8 16.8 60.8 21.5 21.3 69.6 9.1NMS-2 28.4 54.0 17.5 17.0 62.7 20.3Cand-6 48.2 41.0 6.7 n.a. n.a. n.a.Results for immigrants can be biased due to measurement and classfication errors.--Figures need not add up to 100 per cent since the category 'no answer' is notreported here.Source: European LFS, special provision 2008. Own calculations and presentation.IAB 42

Table A4:Occupational structure of medium skilled employed individuals bymigration status in the EU-15, 2006immigrants from NMS-8 immigrants from NMS-2 immigrants from CAND-6> 10years newly arrived > 10years newly arrived > 10years newly arrivednativesClerks 8.6 3.2 7.4 3.0 7.3 3.1 16.4Craft and related tradeworkers18.7 18.1 27.9 29.2 30.1 37.3 14.8Elementary occupations 18.1 33.6 21.2 38.5 15.7 26.2 7.2Legislators, senior officials andmanagersPlant and machine operatorsand assemblers5.9 2.3 0.6 1.1 3.8 1.0 7.97.4 13.5 14.1 9.5 11.5 14.8 8.2Professionals 4.7 3.6 0.3 0.8 1.2 0.0 4.5Service workers and shop andmarket sales workersSkilled agricultural and fisheryworkersTechnicians and associateprofessionalsin per cent of medium skilled employed individuals aged 15-6420.3 19.4 20.6 15.3 16.6 15.4 16.80.4 1.9 5.8 1.2 0.8 0.8 2.516.0 4.6 2.1 1.5 13.0 1.5 20.9Total (in persons) 106,030 267,438 48,006 396,778 431,526 136,763 70,896,574Results for immigrants can be biased due to measurement and classfication errors.-- Figures need not add up to 100 per centsince the category 'armed forces' is not reported here.Sources: European LFS, special provision 2008. Own calculations and presentation.IAB 43

European Integration ConsortiumIAB, CMR, fRDB, GEP, WIFO, wiiwLabour mobility within the EU in the context of enlargement and the functioningof the transitional arrangementsVC/2007/0293Deliverable 3Institute for Employment Research (IAB)Forecasting potential migration from the New Member States into the EU-15:Review of Literature, Evaluation of Forecasting Methods and Forecast ResultsHerbert Brücker, Andreas Damelang and Katja WolfAbstractIn this background report we review the literature on migration forecasts, evaluatedifferent methods for forecasting migration and present a new approach to forecast themigration potential from the new member states (NMS) into the EU-15. There has been alarge literature attempting to forecast the migration from the NMS before enlargement.At a long-run migration potential of about 3 to 5 per cent of the population and an influxof between 200,000 and 300,000 persons, the mainstream of these forecasts is by andlarge consistent with the actual migration movements from the NMS-8 into the EU-15,while the migration potential from Bulgaria and Romania has been underestimated.Moreover, these studies employed explicitly or implicitly the counterfactual assumptionthat all EU-15 countries will open their labour markets at the same time, such that theywere not able to forecast the substantial changes in regional migration patterns whichtook place after EU enlargement. While this literature had to rely on coefficients fromother countries, the post-Enlargement migration enables us to exploit information onrecent migration stocks and flows for forecasts of the migration potential. However, theselective application of transitional arrangements for the free movement of workers hasdistorted bilateral migration patterns, such that the coefficients derived from bilateralmigration movements are likely to be biased. We therefore refer to the EU-15 as a singledestination which allows us to circumvent this problem. Moreover, we use information onmigration stocks and flows within the EU countries to estimate the migration potentialunder the conditions of free movement. Based on this approach, we estimate the longrunmigration potential from the NMS-8 at about 6 per cent of the population in thesending countries, and the migration potential from the NMS-2 (Bulgaria and Romania)at about 14 per cent of the population in the sending countries. The short-run net inflowof migrants from the NMS-8 is estimated to be at about 240,000 persons p.a., and thatfrom the NMS-2 at about 190,000 persons p.a. These net inflows may decline in thecourse of the financial crisis, since immigration and return migration are largelydetermined by the conditions in host countries.The views and opinions expressed in this publication are those of the authors and do not necessarilyrepresent those of the European Commission.

Contents1 Introduction ...................................................................................................... 12 A review of the literature .................................................................................... 22.1 Surveys of migration intentions.................................................................... 32.2 Extrapolation studies .................................................................................. 42.3 Forecasts based on econometric models........................................................ 53 Outline of the theoretical background and the estimation method .......................... 103.1 Theoretical background ............................................................................. 103.2 The macro migration equation ................................................................... 143.3 Identifying the impact of EU-Eastern enlargement........................................ 153.4 Other estimation issues............................................................................. 174 Data............................................................................................................... 205 Estimation results ............................................................................................ 226 A Projection of migration from the NMS-8 and the NMS-2 ..................................... 237 The impact of the financial crisis ........................................................................ 288 References ...................................................................................................... 29IAB 2

1 IntroductionThis background report briefly reviews the literature on forecasts of the migrationpotential from the new member states (NMS), discusses the forecasting methods andtheir theoretical foundations, and presents a projection for the migration potential fromthe NMS based on new estimates which considers recent migration movements.There has been a large literature attempting to forecast the migration from the NMSbefore enlargement. At a long-run migration potential of about 3 to 5 per cent of thepopulation and an influx of between 200,000 and 300,000 persons, the mainstream ofthese forecasts are by and large consistent with the actual migration movements fromthe NMS-8 into the EU-15, while the migration potential from Bulgaria and Romania hasbeen underestimated. In the course of the selective application of the transitionalarrangements, the spatial distribution of migrants across the EU-15 countries haschanged dramatically. As a consequence, forecasts for individual EU member statescarried out under the counterfactual assumption that all EU member states open theirlabour markets at the same time deviate largely from actual migration patterns whichhave emerged after EU enlargement.The data available since the EU’s Eastern enlargement enables us to apply a newapproach for projecting the migration potential. The studies carried out prior to enlargementhad to rely on data and, hence, the experience from other migration episodes,since immigration from the NMS was hampered by the iron curtain and, after thebreakdown of the Berlin wall, by immigration restrictions in the EU-15. All studiestherefore transferred elasticities estimated for other country groups to the NMS. Thisrequires that the estimated coefficients are not only constant across time, but also acrossspace. Since the migration behaviour is heterogeneous across countries, this is animportant drawback of the projections carried out before enlargement. Meanwhile, wecan use the data since enlargement for the identification of the relevant parameters forthe NMS themselves.Most migration forecasts rely explicitly or implicitly on the assumption of the irrelevanceof independent alternatives, i.e. that economic or institutional variables in third countriesdo not affect the scale of migration in another country. If this assumption is not valid, theestimated coefficients are biased. This is particularly relevant in the context of the EU’sEastern enlargement, since the selective application of transitional arrangements hascertainly affected bilateral migration patterns. We circumvent this problem by estimatingthe migration potential for the EU-15 as an aggregate. As a consequence, we cannotforecast the impact of removing immigration barriers on individual destinations such asAustria and Germany. This is in our view not possible, since the selective application ofimmigration barriers and the subsequent diversion of migration stocks and flows has noprecedent in history, such that no counterfactual evidence exists on which we can baseour estimates.We find that the projected migration potential from the NMS-8 is close to what we wouldexpect under the conditions of free movement for the other EU-15 member states.IAB 1

Altogether, the long-run migration potential from the NMS-8 is estimated to be at about5 per cent of the population, and that from the NMS-2 at about 10 per cent in case of anEU-wide introduction of a free movement. Needless to say that these forecasts rely on anumber of strong assumptions and provide no more than a hint to the actual magnitudesinvolved.The remainder of this background report is organised as follows. Section 2 briefly reviewsthe literature on forecasts of the migration potential from the NMS, which have mainlybeen carried out before EU enlargement. Section 3 outlines the theoretical and empiricalframework for the estimation of the migration potential from the NMS underconsideration of the relevant literature. Section 4 describes the data base which isemployed in the estimation of the parameters of our model and Section 5 presents theestimation results. Section 6 simulates the migration potential from the NMS-8 and theNMS-2 into the EU-15 under the status quo conditions and under free movement. Finally,Section 7 discusses how actual migration patterns may deviate from our simulationsunder the conditions of the financial crisis.2 A review of the literatureThere have been numerous studies attempting to forecast potential migration from theNMS before enlargement. Theoretical backgrounds, methodologies and data basesemployed by these studies vary widely. The overwhelming share of these studiesobtained nevertheless remarkably similar results. The mainstream of these studies hasestimated the long-run stock of residents from the NMS at between 3 and 5 per cent ofthe population in the origin countries, while annual net migration flows have beenpredicted to be between 300,000 and 400,000 persons in the first years followingenlargement, which corresponds to 0.3-0.4 per cent of the population in the countries oforigin (see e.g. Alvarez-Plata et al., 2003; Boeri/Brücker, 2001; Bruder, 2003;Hille/Straubhaar, 2001; Krieger, 2003; Layard et al., 1992; Zaiceva, 2006). Somestudies have, however, obtained lower (Fertig, 2001; Fertig and Schmidt, 2001;Dustmann et al., 2003; Pytlikova, 2007) and higher projections (Flaig, 2001; Sinn et al.,2001).These migration forecasts rely on the counterfactual assumption that all Member Statesof the EU-15 open their labour markets at the same time. The selective application oftransitional arrangements for the free movement of workers by the EU-15 countries has,however, affected both the scale and the direction of migration from the NMS. Theauthors of many studies were aware of this before enlargement:„The transitional periods can distort the regional distribution of migrants fromthe CEECs across the EU-15, that is, the diversion of migration flows awayfrom countries which restrict immigration into countries which pursue a moreliberal immigration policy.” (Alvarez-Plata et al., 2003, S. 43).However, missing historical evidence did not allow for estimating the potential diversionof migration flows triggered by the selective application of transitional arrangements inthe EU-15. Therefore, the migration forecasts carried out before enlargement cannot beIAB 2

falsified by the developments following enlargement, since the actual legal andinstitutional conditions differ from those explicitly or implicitly assumed by the migrationprojections. Nevertheless, at an annual net migration flow of between 200,000 and250,000 persons from the NMS-8 into the EU-15, the post-enlargement experience doesnot contradict the aggregate figures of most projections, although migration flows intoIreland and the UK have exceeded the forecasted figures largely.There are essentially three methods which have been used for forecasting the potentialflows of migration from the NMS. The first derives medium- and long-term migrationforecasts from surveys of migration intentions in the sending countries. The secondextrapolates the South-North migration flows in Europe during the 1960s and early1970s to future East-West migration. Finally, the third and largest part of the literaturebases migration forecasts on econometric models, which explain migration stocks andflows by economic and institutional variables.In this section we briefly outline the methodological foundations and the results whichare obtained from these three methods. For previous literature reviews see Brücker andSiliverstovs (2006a; 2006b), Fassmann and Münz (2002), Hönekopp (2001), Straubhaar(2002) and Zaiceva and Zimmermann (2007).2.1 Surveys of migration intentionsA number of studies base forecasts of potential migration on surveys of migrationintentions in the NMS (Fassmann and Hintermann, 1997; Wallace, 1998; Krieger, 2003;Fassmann and Münz, 2002; see also Fouarge and Ester, 2007). Krieger (2003) is basedon the Eurobarometer Labour Mobility Survey, which covers all accession countries; theother studies are based on smaller surveys which focus only on a limited number ofcountries.Studies of migration intentions face several methodological problems. First, and mostimportantly, it is unclear whether or when the expressed migration intention will berealised, and if so, how long an individual will actually stay abroad. As an example, onlya small fraction of the East German individuals who revealed a migration intention in theGerman Socio-Economic Panel (GSOEP) in 1991 have actually moved five years later(Schwarze, 1997), while Gordon and Molho (1995) report evidence that 90 per cent ofthe individuals who intended to move have actually moved in the UK. Therefore, thesestudies use additional questions regarding job search activities in foreign countries,employment and housing contracts etc. for the identification of serious migrationintentions for forecasts of potential migration. As a result, potential migration isestimated to be at between one-third and two-fifth of general migration intentionsrevealed in opinion polls (Fassmann/Hintermann, 1997; Krieger, 2003).Second, the migration intentions revealed in surveys differ substantially depending onthe questionnaire and other aspects of the survey design. Third, it is unclear whethermigration intentions refer to a situation without legal barriers to migration or whethermigration intentions reflect institutional barriers and are therefore a biased measure forIAB 3

migration under the conditions of free movement. Many of these problems could becircumvented by panel studies which would allow one to show whether migrationintentions are realised or not. Unfortunately, panel studies of migration intentions do notyet exist in the NMS.However, surveys of migration intentions can provide valuable information which is notavailable from other studies. First, they deliver important insights on the human capitalcharacteristics of potential migrants (see Fouarge and Ester, 2007; Krieger, 2003, for adetailed analysis). Second, the latest Eurobarometer survey provides information on thedestination countries, which may help to analyse the spatial distribution of migrants fromthe NNS across the EU Member States.According to Fouarge and Easter (2007), 7.4 per cent of the population in the NMS haverevealed a general migration intention in the 2005 wave of the Eurobarometer MobilitySurvey, compared to 2.4 per cent in the 2002 wave. It is not clear whether the differencebetween the two waves can be attributed to a higher propensity to move since thequestionnaire has changed between the two waves. It is also worthwhile noting that 5.0per cent of the EU-15 population have announced a general intention to move in the2005 Eurobarometer survey, although migration stocks from these countries number lessthan 3 per cent in the EU-15.By and large, the findings of the Eurobarometer survey are consistent with those in the1995 wave of the International Social Survey Programme (ISSP), although considerabledifferences exist in individual countries. Similarly, Fassmann and Hintermann (1997) andWallace (1998) find general migration intentions between 3 and 30 per cent of thepopulation. Following these studies, the actual migration potential derived from thegeneral migration intentions is estimated at about 3 per cent of the population in theNMS, while the findings for Bulgaria and Romania are slightly above the NMS-8 average(see Krieger, 2003; Fassmann and Hintermann, 1997; Wallace, 1998).2.2 Extrapolation studiesThe extrapolation of South-North to East-West migration in Europe relies on thehypothesis that the economic and institutional conditions of “guestworker” migration inthe 1960s and early 1970s resemble migration conditions in the enlarged EU of today.Under this assumption, about 3 per cent of the population from the NMS would move tothe EU-15 within 15 years (Layard et al., 1992). Thus, the results are very similar to theestimates of the ‘actual migration potential’ derived from surveys of migration intentions.The income difference measured in purchasing power parities between the EU-15 and theNMS-8 is indeed similar to that between the members of the then European EconomicCommunity (EEC) and their neighbours in Southern Europe during the 1960s. However,there are also important differences between the current enlargement and previousepisodes. First, the present per capita GDP gap between the EU-15 and the NMS-8 atcurrent exchange rates is substantially larger than that between the North and the Southin Europe during the 1960s. Income differences at current exchange rates may affectIAB 4

migration decisions since a part of the income obtained in host countries can beconsumed in the sending countries. Second, labour market conditions (such asunemployment rates) in the main destination countries in the EU-15 are generally lessfavourable today compared to those in Europe during the 1960s. Third, transport andcommunication costs are substantially lower today compared to the 1960s, which in turnreduces migration costs. Finally, the institutional and legal framework for migration wasdifferent during the guestworker recruitment period compared to the legal framework forthe free movement of workers in the Community of today.2.3 Forecasts based on econometric modelsThe largest part of the migration forecasts relies on econometric models, which explainmigration flows or stocks by economic and institutional variables. The key explanatoryvariables are in most models the wage and (un-)employment rates in the receiving andsending countries, the (lagged) migration stock, and a number of dummy variablescapturing institutional conditions in the destination and sending countries, particularlylegal immigration barriers.Although the theoretical foundations differ, most macro migration models are remarkablysimilar with respect to the variables they consider and regarding their functional forms(see Section 3 for a detailed discussion). 1 One important difference in the literature isbetween stock and flow models, which need, however, not necessarily yield differentestimates of the migration potential if properly applied. 2 A second difference is theidentifying restrictions which are imposed by different estimators. Both methodologicalarguments and tests of the forecasting performance suggest that standard fixed effectsmodels outperform pooled OLS models as well as most sophisticated heterogeneousestimators (Alvarez-Plata et al., 2003; Brücker and Siliverstovs, 2006a; 2006b).Table 2.1 summarises the estimation results of different studies including their datasource and methodological foundations. The estimation results for migration stocks andflows are expressed in per cent. This allows one to compare the findings, since thesample of sending countries differs across the studies. 3 We can distinguish studies whichrefer to Germany, the UK and the total EU-15 as a destination, where the latter studiesare based on estimates for a panel of destination and sending countries. The largenumber of studies in the literature which refer to Germany can be traced back to the factthat about 60 per cent of the immigrants from the NMS in the EU-15 resided in Germany1 For derivations of macro migration functions from theoretical models, see inter alia Hatton(1995), Daveri and Faini (1995), Faini and Venturini (1995) and Brücker and Schröder (2006).2 The majority of the models in the empirical literature are specified as gross- or net flow models(e.g. Hatton, 1995; Hille and Straubhaar, 2001; Pederson et al., 2004; Pytlikova, 2007). Thesemodels rely explicitly or implicitly on the assumption of a representative agent, i.e. thatindividuals do not differ with regard to their preferences or human capital characteristics relevantfor migrations. In contrast, stock models are derived from the assumption that individuals areheterogeneous, such that an equilibrium migration stock is achieved when the benefits frommigration equals its costs for the marginal individual (Brücker and Schröder, 2006).3 Note that Table 2.1 is a selection of the literature. There exist numerous other studies which, byand large, resemble the findings represented in this table.IAB 5

efore enlargement. Moreover, the German migration statistics provides detailed data onmigration stocks and flows by country of origin which facilitates migration estimatescompared to many other destinations in the EU-15. Many studies have thereforeestimated the migration potential for Germany and than extrapolated the estimate to theEU-15 under the counter-factual assumption that all EU Member States will open theirlabour markets at the same time and that the regional distribution of migrants remainsconstant over time (Alvarez-Plata et al., 2003; Boeri, Brücker et al., 2001).Among the studies for Germany, Alvarez-Plata et al. (2003), Boeri and Brücker (2001)and Brücker (2002) apply a stock model with country-specific fixed effects, while Flaig(2001) and Sinn et al. (2001) base their estimates on a stock model which is estimatedby pooled OLS. The first studies estimate the annual net inflow at 0.22 per cent of thepopulation from the NMS-8 (160,000 persons p.a.) for Germany, the latter studiesforecast the net inflow at 0.64 per cent p.a. (470,000 persons p.a.). The fixed-effectsmodels estimate the long-run migration potential at 1.7 to 1.8 million persons forGermany, and the latter studies at 5.3 million persons. Although the studies employ alsodifferent data bases, this difference can be mainly traced back to the use of fixed effectsand pooled OLS models (Brücker, 2002; Flaig, 2002). Note that regression diagnosticsrejects the pooled OLS specification and that the forecasting error of the pooled OLSmodel is about twice as high as that of the fixed effects model (see above). In case ofthe fixed effects models, an extrapolation of the estimate for Germany based on theregional distribution of migrants before enlargement provides an initial net inflow of 0.33per cent of the population in the NMS-8 p.a. (240,000 persons p.a.), and in case of thepooled OLS model a net inflow of 1.1 per cent p.a. (780,000 persons p.a.). The long-runmigration potential is estimated by the fixed effects model at 3.9 per cent of thepopulation in the NMS-8 (2.8 million persons), and in case of the pooled OLS models at12 per cent (8.8 million persons) p.a.IAB 6

Table 2.1Econometric forecasts of potential migration from the NMSStudy Database Type of model Estimator Initial net inflow Long-run stockEstimates of potential immigration into Germany (extrapolations to EU-15 in parentheses)Alvarez-Plataet al. (2003)Panel of migrationstocks from 18 sendingcountries, 1967-2001Dynamic stockmodelFixed effects 0.22%(EU-15: 0.33%)2.33%(EU-15: 3.82%)Boeri, Brückeret al. (2001),Brücker(2001)Panel of migrationstocks from 18 sendingcountries, 1967-1998Dynamic stockmodelFixed effects 0.22%(EU-15: 0.34%)2.53%(EU-15: 3.89%)Dustmannet al. (2003)Panel of migrationflows from 18 sendingcountries, 1960-1994Static flow modelGMM withindividualeffects0.02% - 0.2% -Fertig (2001)Panel of migrationflows from 17 sendingcountries, 1960-1997Dynamic flowmodelFixed effects 0.07% -Fertig andSchmidt(2001)Panel of migrationflows from 17 sendingcountries, 1960-1997Static errorcomponentsmodelGMM 0.01% -0.06% -Flaig (2001),Sinn et al.(2001)Panel of migrationstocks from 5 sendingcountries, 1974-1997Dynamic stockmodelPooled OLS 0.64% 7.2%Estimates of potential immigration into the United KingdomDustmann etal. (2003)Panel of migrationflows from 18 sendingcountries, 1960-1994Static flow modelGMM withindividualeffects0.004% - 0.01% -Alvarez-Plataet al. (2003)Panel of labourmigration stocks from20 sending and 15destination countries,1993-2001Estimates of potential immigration into the EU-15Dynamic stockmodelGMM-systemestimator withindividualeffectsEU-15:0.11% - 0.15%(labour force)EU-15:2.2% - 2.7%(labour force)Hille andStraubhaar(2001),Straubhaar(2002)Panel of migrationflows from 3 sendingand 8 destinationcountries, 1988-99Static flow model(gravity equation)Pooled OLS EU-15: 0.27% -Pytlikova(2007)Panel of gross and netmigration flows from 7NMS into 15 EU/EEAcountries, 1990-2000Static and dynamicflow modelFixed effectsEU/EEA-13:0.04-0.08%(net), (grossinflows: 0.53-0.57)EU/EEA-13:1.5%-1.8%Zaiceva (2006)Panel of migrationflows from 3 sendingand 15 receivingcountries, 1986-1997.Source: Own presentation based on the quoted studies.Static flow model(gravity equation)Fixed effects EU-15: 0.23-0.34%EU-15:3.5%-5.0%IAB 7

The estimates by Fertig (2001) and Fertig and Schmidt (2001) are substantially belowthe other forecasts: The initial net immigration rate from the NMS to Germany isestimated there at 0.01 to 0.07 per cent p.a., which corresponds to a net immigration of7,000 to 50,000 persons p.a. from the NMS. The Fertig and Schmidt (2001) study appliesan error-component model which considers country- and time-specific fixed effects, butnot any other explanatory variables such as wage differences or (un-)employment rates.As a consequence, the forecast refers to the sample average, or, more precisely, to arange of one standard deviation plus/minus the sample average. It is possible that thishas resulted in an underestimation of the migration potential from the NMS, since theincome of most sending countries in their sample is well above that of the NMS.The Dustmann et al. (2003) study estimates a flow model with GMM for Germany andthe UK, which considers also individual effects. Again, this model provides lower estimatescompared to the standard fixed effects models, although the upper range of theestimate for Germany is getting close to the estimates by Alvarez-Plata et al. (2003) andBoeri and Brücker (2001). The findings for the UK refer to flow data from the PassengerSurvey and provide, at a share of 0.004 to 0.01 per cent, a very low estimate for the UK.Note that the Dustmann et al. (2003) study - as all other studies - does not consider anypossible diversion effects which may explain the later migration surge in the UK.The gravity-type estimates for the EU-15 of Alvarez-Plata et al. (2003) and Hille andStraubhaar (2001) obtain relatively similar results. Note that the Alvarez-Plata et al.(2003) projection refers to the labour force and not to the population form the NMS,while the estimates by Hille and Straubhaar (2001) use population data. Since the labourforce is about 60 per cent of the foreign population from the NMS, the forecasted figuresare remarkably similar. Moreover, the aggregate figures from the EU-level estimates areconsistent with the extrapolations from the German estimates by Alvarez-Plata et al.(2003) and Boeri and Brücker (2001).Altogether, at the level of the EU-15, the estimates of these studies are by and largeconsistent with the migration development from the NMS-8 since enlargement: Theactual growth in the number of foreign residents numbered about 250,000 persons p.a.on average since enlargement, which corresponds to 0.34 per cent of the population inthe NMS-8. This is consistent with the projections of the Alvarez-Plata et al. (2003), Boeriand Brücker (2001), Hille and Straubhaar (2001) and Zaiceva (2005) studies, while theFlaig (2002) and Sinn et al. (2001) study provided higher, and Fertig (2001), Fertig andSchmidt (2002) and Dustmann et al. (2003) lower estimates.While the aggregate estimates of potential migration from the NMS-8 to the EU-15 are inmany studies consistent with the scale of migration after EU enlargement, the regionalstructure deviates largely from the estimates. As has been shown above, the regionalmigration patterns have dramatically changed in the course of EU enlargement. Hence,those studies which have extrapolated the regional distribution of migrants beforeenlargement tend to overstate the inflows to Austria and Germany and to understate themigration to Ireland and the UK. The same holds true for studies which base theirIAB 8

estimates for the UK on past migration flows. Actual migration inflows into the UK havebeen at about 160,000 p.a. larger than the net flows predicted in the Dustmann et al.(2003) study for the UK (4,000-13,000). Similarly, Boeri and Brücker (2001) andAlvarez-Plata et al. (2001) provided projections based on the extrapolation of theGerman forecasts which have been substantially below the actual inflows into UK andIreland after enlargement. In contrast, the flows to the Scandinavian countries have beenat or below the predicted levels.Since a counterfactual situation with a free movement of workers does not exist for theNMS, it is hardly possible to disentangle the causes for the diversion of the migrationflows from the NMS after EU enlargement empirically. Obviously, the selective applicationof the transitional arrangements is one if not the major cause of the diversion process.All studies in the literature rely, however, explicitly or implicitly on the counterfactualassumption that all EU countries will open their labour markets at the same time formigrants from the NMS. The selective application of transitional arrangements will,however, trigger additional inflows to countries which will open their labour markets andless inflows to countries which do not, as Alvarez-Plata et al. (2003) have emphasised intheir study before EU enlargement.The selective application of the transitional arrangements can, however, not explain whySweden and other Scandinavian countries received only moderate inflows from the NMS-8, while Ireland and the UK absorbed the overwhelming share. Other causes which mayhave influenced the regional allocation of migration flows from the NMS after EUenlargement are the English language, together with flexible labour market institutions.Moreover, the economic down-turn in Germany has certainly contributed to the diversiontowards more prosperous destinations. It might also be possible that the preenlargementallocation of migrants from the NMS across the EU-15 was biased by theselective application of immigration restrictions, i.e. the relatively liberal immigrationconditions in Austria and Germany compared to other destinations. Finally, the erosion ofvariable transport costs caused by low-budget air transport makes geographical migrationpatterns less stable than in previous migration episodes. As a consequence, it wasrelatively cheap for migrants from the NMS to switch from Austria and Germany toIreland and the UK and to establish new migration networks there.These arguments highlight a deeper methodological problem of forecasting migration inthe context of EU enlargement in the previous literature: All these models rely explicitlyor implicitly on the assumption of the irrelevance of independent alternatives, i.e. thatthe economic and institutional conditions in alternative destinations do not matter for thescale of migration towards a specific destination. However, the fact that main destinationssuch as Germany and Austria have maintained their immigration restrictionswhen the UK and Ireland have opened their labour markets has certainly triggeredadditional immigration flows to the latter destinations. Similarly, changing economic orsocial conditions in one destination may also affect the scale of migration in otherdestinations. The impact of third countries is particularly relevant in the context of the EUEastern enlargement, since the institutional conditions for immigration have changedIAB 9

dramatically in some destinations but not in other. This is of course hardly possible toidentify in advance, since similar evidence from previous migration episodes did not existin the EU.3 Outline of the theoretical background and the estimation method3.1 Theoretical backgroundAll econometric models in the literature attempting to forecast migration flows or stocksare macro models, which are explicitly or implicitly derived from the aggregation ofindividual decisions. Most of these models explain migration stocks or flows by wagedifferences between the destination and the sending country, labour market variableswhich should capture employment opportunities in the respective locations, and by a setof institutional and distance variables which should approximate migration costs and legalor administrative barriers to migration. These specifications of the migration functionhave a long tradition in the literature, which can be summarised under the umbrella ofthe ‘human capital approach’ (Sjaastad, 1962).4 The choice of economic variables isprimarily based on the classical theoretical contributions by Ravenstein (1889), Hicks(1932), Sjaastad (1962), Todaro (1969) and Harris and Todaro (1970). The firstcontributions suggest that the net present value of the difference in wages and othersources of income between the host and the source countries could be regarded as aprimary determinant of the migration decision, while the latter two papers introduce therole of labour markets in the decision-making process.More specifically, the standard model in the literature is derived from the followingassumptions: The utility of individuals is inter alia determined by expectations of incomelevels in the respective locations. Utility is concave in the income differential. Explicitly orimplicitly, most models of the migration decision assume that other arguments enter theutility function as well. In particular, non-monetary factors such as the disutility fromleaving a familiar social and cultural environment and the role of family ties (Mincer,1978) are considered.5 Depending on the assumptions on the utility function, thefunctional form of the macro-migration function is specified both in semi-log form (e.g.Hatton, 1995) and in double-log form (e.g. Faini/Venturini, 1995).Expectations on income levels are conditioned by employment opportunities, such thatthe expected income levels depend on the probability of employment in the respectivelocation (Todaro, 1969; Harris/Todaro, 1970). Moreover, since employment opportunitiesof migrants in host countries are below those from natives, some models predict that thecoefficient for the employment rate in the host country is larger than that in the sourcecountry (Hatton, 1995).4 More precisley, migration is understood as an “investment in the productive use ofhuman resources” (Sjaastad, 1962) by these migration theories.5 See Faini and Venturini (1995) for a model which considers non-pecuniary argumentsexplicitly in the utility function.IAB 10

If capital markets are not perfect, liquidity constraints affect migration decisions.Consequently, for a given income difference between the host and the source country,the income level in the source country has a positive impact on migration (seeFaini/Venturini, 1995, for a formal exposition, and Hatton and Williamson, 2002, as wellas Pederson et al., 2004, for contrasting evidence).Migration networks alleviate the costs of adapting to an unfamiliar environment, suchthat the costs from migration are expected to decline with the stock of migrants alreadyexisting in the destination country (Massey et al., 1984; Massey/Espana, 1987). Distanceserves as a proxy for pecuniary and non-pecuniary migration costs, which are expectedto increase with the spatial distance between the source and the destination country (seeSchwarz, 1962, for a detailed discussion of the role of distance). The time trend isincluded as a proxy for the variation in the costs of migration, which are expected to fallover time in the course of decreasing transport and communication costs.Among the institutional variables, most models consider dummy variables for conditionswhich facilitate immigration (e.g. bilateral guestworker recruitment agreements, freemovement of workers within the EU) or hinder emigration (e.g. the iron curtain in theformer COMECON countries). Moreover, some models include certain variables for pushand pull factors in sending and receiving countries such as dummy variables fordictatorship, the Freedom House political and civil right indices etc.Risk and uncertaintySome models in the literature explicitly consider the risk aversion of individuals. E.g.Banerjee and Kanbur (1986) have developed a model which assumes that individuals arerisk-averse. They consider in a specification of a regional migration function the variancein expected income levels. The higher the variance, the lower the migration rate ifindividuals are risk adverse. This is, however, seldom applied in the context ofinternational migration since time-series data on the distribution of income do not existin most countries.The impact of uncertainty on the migration decision under the assumption of fixedmigration costs has been analysed theoretically in an option-value framework by Burda(1995). The model treats migration as an irreversible investment, such that the optionvalue of waiting is increasing in the uncertainty about the net returns to migration.Hatton (1995) derives an error-correction specification of the migration function fromthese assumptions, without changing the contents of the long-run migration function ofthe standard migration models. Moreover, Hatton (1995) assumes that individuals arerisk-averse, but that uncertainty focuses on employment opportunities. As aconsequence, the model expects that the coefficient for the employment variable in thedestination country is larger than that for the income variables.IAB 11

Limitations of the human capital approachAltogether, the human capital theories of migration expect that the wage differencebetween the host and the home country and the employment rate in the host countryhave a positive impact on migration, while the employment rate in the host country has anegative impact. At a given wage difference between the host and the home country, theincome level in the home country is expected to have a positive impact, since it relaxesliquidity constraints. Finally, the scale of migration is expected to decline with geographicaldistance, since this variable approximates fixed and variables migration costs.There exist numerous microeconomic models of the migration decision in the literaturewhich go far beyond these considerations. Inter alia, these models analyse the role ofportfolio diversification of families in the absence of perfect capital markets (Stark, 1991)and the role of relative deprivation (Stark, 1984). However, few of these theoreticalcontributions have developed macro migration functions which can be applied empirically.Moreover, the estimation of more complex macro models is hindered by datalimitations, e.g. time series information on the income distribution in the receiving andsending countries is rarely available for longer time spans.Migration flows versus migration stocksThus, although the microeconomic migration literature is richer than the standard macromigration model suggests, a consensus has evolved in the literature to explain migrationby income variables, labour market variables such as (un-)employment rates, distance,and institutional variables. One important difference in the specification of macromigration functions in the literature refers to the choice of the dependent variable. Whilethe larger part of the literature employs net or gross migration flows on the left-handside of the macro migration equation (e.g. Faini/Venturini, 1995; Hatton, 1995;Hille/Straubhaar, 2001; Pederson et al., 2004; Pytlikova, 2007), a minority of the studieschooses the migration stock (i.e. the number of residents) as the dependent variable(e.g. Boeri/Brücker, 2001; Brücker/Schröder, 2006; Flaig, 2001; Sinn et al., 2001).The difference between these two specifications can be traced back to the underlyingassumptions regarding the aggregation of individual migration decisions: The flow modelis implicitly based on the assumption of a representative agent, i.e. that the behaviour ofindividuals is homogeneous. In contrast, the stock model is based on the assumption thatindividuals differ with regard to their preferences or human capital characteristics, whichin turn determine the benefits and costs of migration. The specific form of the macromigration function depends then on the assumptions which are made regarding thedistribution of preferences or human capital characteristics. For a formal derivation of astock model which considers heterogeneous preferences see Brücker/Schröder (2006).The specification of the migration function in flow or stock form has importantimplications for migration forecasts: In case of the flow-specification of the migrationfunction, net migration flows do not cease before (expected) income levels between thehost and the source country have converged to a certain threshold level which capturesIAB 12

the uniform level of migration costs. In contrast, in case of the stock model, netmigration ceases when the benefits from migration equals the costs for the marginalmigrant. Consequently, net migration flows converge to zero when the migration stockapproaches its equilibrium level.This might be an explanation for the phenomenon that in case of the Southernenlargement of the EU, where still substantial income differences between the incumbentand the new member states from Southern Europe existed, net migration flows havestagnated or even declined after the application of the rules for the free movement ofworkers. Note that most receiving countries have built up substantial migration stocksfrom the later EU members from Southern Europe already in the 1960s and 1970s.Permanent vs. temporary migration modelsThe overwhelming share of migration is temporary, i.e. migrants return to their homecountry before the end of their lifetime. Moreover, many individuals have severalmigration episodes during their lifetime, a phenomenon called ‘replicated migration’ inthe literature. Nevertheless, most macro migration models in the literature treatmigration as permanent. There exist, however, many theoretical models in the literaturewhich consider temporary migration (e.g. Djajic/Milbourne, 1986). There, the length ofan individual migration episode is explained by expected earnings in the home and thedestination countries and the costs of staying in a foreign country which are determinedby individual preferences. Brücker and Schröder (2006) have derived the consequencesof temporary migration for migration stocks and flows in temporary migration frameworkwith heterogeneous agents. The length of individual migration episodes varies dependingon the expected difference in earnings and individual preferences. The equilibrium stockof migrants increases with the difference in earnings for a given distribution of individualpreferences. Analogously, the gross emigration and return migration rates are increasingin the earnings difference, while the net migration ceases to zero if the equilibrium isachieved and the rates of population growth in the foreign and the home country areequal. Thus, the stock model is consistent with a temporary migration framework.Bilateral versus multi-country modelsTheoretically, most models in the literature are two-country or two-region models, i.e.migrants decide whether to migrate into a foreign country (region) or to stay at home.This simplifies the modelling effort tremendously. However, migration decisions are infact optimising decisions across space, i.e. (potential) migrants compare the netdifference in utility between all possible locations including the home country or region.In empirical applications this is usually ignored, i.e. it is explicitly or implicitly assumedthat migration between two countries or regions is driven by the differences in expectedincome and other factors between these two regions, but not by the immigrationconditions in third countries. Technically, this assumption is called ‘Irrelevance ofIndependent Alternatives’ (IIA). All gravity models and other macro migration modelsIAB 13

applied in the empirical literature rely on this assumption. If this assumption is not valid,the estimated coefficients are biased. While the IIA assumption might be not toodemanding if we consider migration towards destinations which differ largely in theircharacteristics and/or geographical distance, it is particularly dangerous in the context ofmigration in the enlarged EU, since many destinations are similar with respect to theirincome levels, culture and other factors. In this case it is likely that institutional factorssuch as the selective application of the transitional arrangements for the free movementof workers have an impact on migration movements, such that conditions in thirdcountries matter. However, explaining bilateral migration movements by the entire set ofpossible migration alternatives is not a viable estimation strategy, such that simpler toolshave to be applied to circumvent the problem (see below).3.2 The macro migration equationFollowing the overwhelming majority, we apply here a parsimonious specification of themacro migration function in our econometric model. The theoretical approach follows thetemporary migration framework with heterogeneous agents originally developed byBrücker and Schröder (2006). Individuals have the choice to stay at home or to move fora certain period of their life time (or their entire life) to another country. They choose thelength of the stay in the foreign country such that they maximise utility over their lifetime. The utility of individuals depends on their income in the respective locations, butalso on non-monetary factors such as social relations, cultural links etc. At a givendifference in the net present value of earnings, the time spend abroad depends on theweight individuals assign to monetary earnings and to the non-pecuniary factors relevantfor their utility in the respective locations (see Djajic and Milbourne, 1986; Dustmannand Kirchkamp, 2002; for similar models). Under the assumption that these preferencesare not uniform across individuals, an equilibrium relationship between migration stocksand the difference in income levels between the host and the home country emerges. Atthis equilibrium, the gross emigration rate and the gross return migration rate are equal,such that net migration ceases (Brücker and Schröder, 2006).More specifically, the long-run macro migration function is specified in the following form:⎛ w ⎞mst = a + a ln + a ln ( e ) + a ln ( e ) + ε⎝ ⎠*ftfit 0 1 ⎜ ⎟ 2 ft 3 it fitwit, (1)*where mst fit denotes the long-run or equilibrium share of migrants residing indestination f in the population from sending country i, w ft and w it the wage rate in thedestination and the sending country, and e ft and e jt the employment rate in therespective countries and ε fit the disturbance term. The subscript f denotes thedestination, i the index of sending countries and t the time index.The variables of the model are derived from the standard human capital model, i.e. theutility is determined by expectations on income levels, which are in turn conditioned byemployment opportunities. Individuals are risk averse, but uncertainty focuses onIAB 14

employment opportunities. Hence, it is expected that the coefficient for the employmentrate in the receiving country is larger than the coefficient for the employment rate in thehome country (Hatton, 1995).The dynamics of the model are specified here in form of a simple partial adjustmentmechanism, i.e. as⎛ w ⎞mst = b + b ln + b ln ( e ) + b ln ( e ) + b mst + ν⎝ ⎠ftfit 0 1 ⎜ ⎟ 2 ft 3 it 4 fi, t−1fitwit, (2)where the coefficient b 4 < 1 captures the dynamic adjustment of the model. Therestriction that b 4 < 1 is needed for the dynamic stability of the model. Note that thisdoes not rule out that networks of previous migrants alleviate migration costs andfacilitate further migration. In contrast, we follow here the literature that migrationnetworks or migration chains reduce migration costs (Bauer et al., 2002a; 2002b;Massey et al., 1984; Massey and Espana, 1987). However, since the preference foramenities in the home country tends to increase for the marginal individual the higherthe share of the population is that already lives abroad, the declining costs for migrationresulting from networks are eventually offset by the low preferences to move abroad ofthe remaining population.Of course, the specific functional form of the model depends on the underlyingassumptions regarding the utility function. The model may thus be specified both indouble-log or semi-log form (see e.g. Hatton, 1995, for a discussion). 63.3 Identifying the impact of EU-Eastern enlargementSo far we have ignored all institutional restrictions and applied the traditional irrelevanceof independent alternatives (IIA) assumption. Since institutional conditions in alternativedestinations have turned out to be quite relevant as the diversion of migration flowsaway from Germany and Austria towards the UK and Ireland has demonstrated, weemploy here another approach than in the previous literature. Instead of estimating themodel in equation (2) for bilateral country pairs, we estimate the migration from anumber of destinations into the entire EU-15 assuming that the choice to move into theEU-15 is independent from other possible destinations. Since the overwhelming share ofthe migrants from the NMS and the other countries included in the sample moves to theEU-15, ignoring other destinations does not seem to be too restrictive. By treating the EUas a single destination country, we circumvent the IIA problem and should obtainconsistent estimates of the parameters as long as other alternative destinations outsidethe EU do not affect the scale of migration into the EU-15 and as long the EU-15countries are relatively homogeneous in their characteristics such that a change in the6 The semi-log form employed here provides a better forecasting performance than a double-logspecification. See Brücker/Siliverstovs (2006).IAB 15

egional structure of migration within the EU does not largely affect overall migration intothe EU-15.Although income levels and employment opportunities across the individual EU countriesare relatively homogeneous, there still exist some differences which might be hidden ifwe average all variables of the model across the destination countries in the EU-15. Wehave therefore weighted all earnings and employment variables by the share of therespective country in the migrants from a specific sending country in the EU-15 in orderto capture the relevant values for the explanatory variables. We expect that thisincreases the explanatory power of the model. 7The second problem is the identification of the impact of the remaining immigrationrestrictions. Since a free movement counterfactual does not exist for the NMS, wedecided to include in our sample three groups of sending countries: The member statesfrom the EU-15, for which the free movement of workers was granted for the entire or apart of the sample period, the NMS-8, for which the transitional arrangements applysince 2004, and the NMS-2, for which no transitional arrangements apply during thesample period, but bilateral agreements which have facilitated migration. We assumethat immigration restrictions affect both the absolute terms and the slope parameters ofthe model.In general form we can then write the migration function under consideration of theimmigration restrictions as∑ ∑ ∑ ∑∑ ∑∑ (3)mst = α x + β x + γ z + η z x + λ z x + δ mst + νfit j jft k kit n nfit nj nfit jft nk nfit kit fi, t − 1 fitj k n n j n kwhere z denotes a dummy variable which captures an institutional regime which affectsthe migration opportunities and costs, x an explanatory variable such as the wage andthe employment rate which affects migration incentives, α, β, γ, η, λ and δ coefficients, jan index for variables which capture economic conditions in destination f, k an index forvariables which capture economic conditions in sending country i, n an index for aninstitutional regime which affects migration between destination country f and sendingcountry i, and ν the error term.Thus, different institutional regimes can affect migration in our model via the absoluteterms and via the slope parameters for the economic variables considered by the model.Under the assumption that the slope parameters are uniform across countries for a giveninstitutional regime, we can use the estimated parameters of the model to identify how achange in the institutional regime affects migration. As an example, if the NMS respondsimilarly as other EU member states under free movement to the explanatory variablessuch as the income differential and the employment rate, we can use the estimated7 The migration shares are of course endogenous which may bias the results. We have thereforeused both the average values of the variables in the EU-15 and lagged values of the explanatoryvariables as instruments which did not change the results significantly.IAB 16

parameters of the free movement dummy and the interaction dummies of the freemovement with the explanatory variables for identifying the impact of a switch of aninstitutional regime which is characterised by transitional arrangements to freemovement. However, it is worthwhile noting that countries might be heterogeneous, i.e.that the migration behaviour of the NMS may differ in one way or another from that ofthe EU-15 member states. The assumption of homogenous slope parameters is, however,needed for the identification of the effects of different institutional conditions.In the specification of the model we consider the following institutional regimes:• transitional arrangements for the NMS-8 between 2004 and 2007 and for theNMS-2 in 2007;• bilateral (guestworker) agreements between individual EU-15 and the NMS-2which were in place since the end of the 1990s;• restricted immigration, which holds for third countries such as Turkey, Moroccoand Tunisia as well as for the NMS before the transitional arrangements or thebilateral agreements were in place;• emigration restrictions which were in place for citizens from most NMS under theso-called ‘iron curtain’.For each regime we created a dummy variable, which was included as a level variableand as an interaction variable with all other explanatory variables of the model.3.4 Other estimation issuesThe error term is specified here as one-way error component model, i.e. asν it = µ i + ε it , (4)where µ i is a country specific fixed effect which captures all time-invariant variables suchas geographical distance, language, and cultural proximity migration decisions, and ε it iswhite noise.The specification of the error term has important implications for the scale of migrationforecasts. In the literature, most macro migration models are either estimated by pooledordinary least squares (pooled OLS) or with a fixed effects estimator. While the firstapproach assumes that the intercept term (constant) is uniform for all countries, thelatter approach allows the intercept term to differ across countries. However, bothestimators impose the restriction of uniform slope parameters. In the case of Germany,The pooled OLS estimator has provided much larger migration forecasts (Flaig, 2001;Sinn et al., 2001) compared to the fixed effects estimator (Alvarez-Plata et al., 2003;Boeri/Brücker, 2001; Brücker, 2001).IAB 17

The intercept term captures all time-invariant country specific effects such asgeographical factors, culture, historical links, language etc. Most of these variables areunobservable and can therefore not be explicitly considered in pooled OLS models.Imposing the restriction of a uniform intercept term can therefore produce inconsistentand potentially misleading estimates of the parameter values unless the constants areidentical across countries (Baltagi, 1995). In the context of migration regressions, theregression diagnostics unambiguously rejects the pooled OLS specification whencompared to the fixed effects model (see e.g. Alecke et al., 2001; Alvarez-Plata et al.,2003; Brücker, 2001; Fertig, 2001; Pytlikova, 2007).There remain, however, two arguments why pooled OLS models are still used forforecasts from the NMS: Firstly, since most forecasts of potential migration from the NMSdo not include data from the NMS in the data base, the estimated fixed effects cannot beused for forecasting migration from the NMS. The NMS are usually not included in thecountry sample since migration from there has been hindered by the Iron Curtain andlater by the immigration restrictions in the EU-15. A widely applied procedure tocircumvent this problem is to explain the fixed effects in an auxiliary regression by timeinvariantfactors (e.g. language, geographical location etc.) (Fertig, 2001; see alsoAlvarez-Plata et al., 2001; Boeri/Brücker, 2001; Zaiceva, 2006). This allows a consistentestimation of the slope parameters, even if the fixed effects are not entirely explained.Note that about 90 per cent in the variance of the fixed effects has been explained inauxiliary regressions (see Alvarez-Plata et al., 2003).Secondly, it is sometimes argued that the within transformation by the fixed effectsestimator reduces the variance of the data such that the coefficients cannot be properlyidentified (Flaig, 2001). However, note that the variance of income levels andunemployment rates is pretty large over time if we consider that many models base theirestimates on data bases which cover between three and four decades. Not surprisingly, adetailed analysis of the forecasting performance of different estimators finds that the outof-sampleforecasting error of pooled OLS models is about twice as high as that of fixedeffects models (Alvarez-Plata et al., 2003; Brücker/Siliverstovs, 2006a; 2006b).Another alternative would be to specify the individual term in form of a random effectsmodel. This is rarely applied in the macro migration literature, since it is rather unlikelythat individual effects follow a random distribution in cross-country regressions. Indeed,the standard Hausman-test clearly rejects the random effects specification in the contextof macro migration models. The forecasting error of the random effects model ismoreover clearly larger than that of fixed effects models (Brücker/Siliverstovs, 2006a;2006b).The different migration behaviours across countries can affect not only the interceptterm, but also the slope parameters of the model. The homogeneity assumption ofstandard panel models can therefore yield inconsistent and biased estimates of theparameters (Pesaran and Smith, 1995).Several alternatives to the restriction of uniform slope parameters can be considered.The regressions can be estimated individually, after which the means of the estimatedIAB 18

coefficients can be calculated. This ‘Mean Group’ estimator produces consistent results ifthe group dimension of the panel tends to infinity (Pesaran and Smith, 1995) – which is,however, not the case in the samples at hand for migration forecasts. Another alternativeis the ‘Pooled Mean Group’ estimator, which constrains the long-term coefficients to bethe same but allows for heterogeneous short-run coefficients. This estimator is anintermediate case – it imposes fewer restrictions on the adjustment process, but thesame restrictions on the long-term coefficients as standard panel models. In case of acointegration relationship between the variables, similar assumptions on the convergenceof the estimated parameters to their true values as in the individual case apply (Pesaranet al., 1997).Although the theoretical arguments against the homogeneity assumption of pooledestimators are appealing, there exists evidence in many other empirical contexts that theforecasting performance of traditional panel estimators such as the fixed effectsestimator is superior relative to estimators with heterogeneous slope parameters (Baltagiet al., 2002; Baltagi et al., 2000; Baltagi and Griffin, 1997). The reason for this finding isthat individual regressions can yield highly unstable results if data sets have a limitedtime-dimension.Brücker and Siliverstovs (2006a, 2006b) and Helmert (2007) have tested in themigration context inter alia the forecasting performance of the Pooled Mean Group(PMG), the Mean Group (MG) estimator and individual OLS (IOLS) regressions for eachcountry. They find that the forecasting error of the PMG and the MG is much higher thanthat of homogeneous estimators such as the fixed effects and the pooled OLS estimators,which confirms the findings by Baltagi et al. (2002). However, the forecastingperformance of the individual OLS regressions depends on the forecasting criteria: Theforecasting performance is poorer compared to the homogeneous panel estimators if weuse the Root Mean Squared Error as the criterion, which measures the mean forecastingerror from all countries in absolute terms. However, if we apply the Root Mean SquaredPercentage Error, which measures the mean percentage error of the forecasts for allcountries, the individual OLS estimator outperforms the homogeneous panel estimators.Thus, the panel estimators are more appropriate if we are interested in the number ofmigrants which move from all NMS into the EU-15 (or an individual EU-15 country), whilethe individual OLS regressions are more useful if we are interested in the percentage ofmigrants which leave each individual NMS country (Helmert, 2007).Thus, most empirical models are based on a dynamic specification of the migrationfunction and employ country-specific fixed effects. It is well known that these types ofmodels may cause a simultaneous equation bias if the time dimension of the data athand is limited (Nickell, 1981; Kiviet 1995). Although the simultaneous equation biasdisappears with the time dimension of the panel, it can still be relevant for the size of apanel with between 15 and 35 observations over time as is usually employed in themigration literature (Judson and Owen, 1999).Arellano and Bond (1991), Arellano and Bover (1995) and Blundell and Bond (1998) havedeveloped Generalized Methods of Moments (GMM) estimators, which difference the dataand use either first-differences or first-differences and lags of the level variables asIAB 19

instruments. This allows a consistent estimation of dynamic models with fixed effects.However, the gains from an unbiased estimation by GMM might be offset by losses inefficiency if the group dimension of the data set is limited (Baltagi et. al., 2000). Brückerand Siliverstovs (2006a; 2006b) show that the forecasting performance of the GMMestimators is poor compared to standard fixed effects and other panel models.To sum up, against the background of the experience in the literature, we employ astandard fixed effects estimation approach here.4 DataOur sample consists of 28 sending countries during the period 1982 to 2007: The ‘old’ EUmember states with the exception of Luxembourg (14), the NMS-8, the NMS-2 (Bulgariaand Romania), the (former) Yugoslavia, Morocco, Tunisia, and Turkey. This sample thuscovers - with the exception of the Commonwealth of Independent States (CIS) countries- the entire European continent and some main sending countries at the Europeanperiphery. Moreover, the EU-15 is the main destination for migrants from these countriessuch that the assumption of the irrelevance of independent alternatives is not toodemanding. For this reason we have excluded the CIS countries from the sample, sinceethnic disentangling plays an important role there. Other destinations such as Russia aretherefore important alternatives to the EU-15 in case of the CIS. Altogether, our samplecovers more than 80 per cent of the immigrants residing in the EU-15. Due to datalimitations, the sample is not balanced. Note that we can include only those sendingcountries for which (almost) the entire EU-15 report migration stocks.The data on migration stocks are derived from the statistics of the EU-15 destinationcountries. Whenever possible, we have used the national population statistics, and theEurostat Labour Force Survey in the remaining cases. However, in order to avoidstructural breaks we rely only on one data source for a given destination. These datahave then been aggregated to calculate the number of migrants in the EU-15. Sincenational data sources and nationality concepts differ across the EU, some measurementerror is unavoidable.As an approximation for average earnings we have used the GDP per capita. Weemployed in our regressions both the GDP per capita at purchasing power parities and atcurrent exchange rates. Since the forecasting performance of the income variable atcurrent exchange rates has turned out to be better as the income measured atpurchasing power parities, we decided to use the GDP per capita at current exchangerates in the regressions presented here. Note that particularly in the case of temporarymigration the GDP at current exchange rates affects migration decisions, since a part ofthe income is consumed in home countries. Moreover, the measurement error for theGDP per capita at current exchange rates is likely to be smaller compared to thepurchasing power parity estimates. The GDP per capita at current exchange rates hasbeen taken from the World Development Indicators (World Bank, 2008), while the GDPper capita at purchasing power parity has been derived from the series provided byIAB 20

Angus Maddison and the University of Groningen, which has been extrapolated from theWold Bank series. For the calculation of the employment rates we used the standardisedunemployment rates (ILO norm) provided by Eurostat which have been complementedby national statistical sources in some cases. The population figures have been takenfrom Eurostat. The destination country variables (i.e. the EU-15 variables) have beencalculated by weighting the variables across the destinations with the immigrant sharesas outlined above.The institutional variables are defined as follows: TRANS it is a dummy variable which hasa value of 1 if the transitional arrangements for the free movement of workers betweenthe EU-15 and the NMS-8 are in place and of zero otherwise; GUEST it is a dummyvariable which has a values of 1 if migration from Bulgaria and Romania is facilitated bybilateral guestworker agreements and of zero otherwise; 8 RESTRICT it is a dummyvariable which has a value of 1 if the country does not participate in the free movementof the EU and the EEA and if immigration is not facilitated either by transitionalarrangements for the free movement or by guestworker agreements; IRON it is a dummyvariable which has a value of 1 if emigration is effectively hindered by the iron curtainand of zero otherwise.Several aspects are important to notice in this context. The institutional variablesconsidered here are of course only rough approximations of the institutional conditions inthe EU-15. As an example, we are not able to capture changes in the application of thetransitional arrangements during the 2004-2007 period in individual EU member states,i.e. countries which have decided to open their labour markets during the sample period.This would require including a dummy variable and the respective interaction dummyvariables for each year since 2004, which would in turn make any identificationimpossible. A similar argument applies for changes of immigration policies of the EU-15vis-à-vis Bulgaria and Romania during the phase which we characterise here asinfluenced by bilateral migration agreements. However, in our view these changes in theimmigration policies during the 2004-2007 period did not affect migration flows from theNMS-8 and the NMS-2 much, such that our identification strategy captures the mainchanges in the immigration regimes of the EU-15 during the sample period. A moredetailed consideration of the institutional regimes would require estimating the model asa panel of destination and sending countries, which would in turn run into the difficultiesof employing the irrelevance of independent alternatives assumption. This would yieldextremely biased results if migration in one EU-15 country is affected in one way oranother by the immigration policies of other EU-15 countries, which is certainly the casein the context of the EU’s Eastern enlargement in our view.8 This holds for Bulgaria and Romania in the years from 1998 until the end of the sample period.The traditional source countries of guestworker recruitment in the EU such as Spain, Portugal,and Turkey have not been subject of those agreements during the sample period. We did notinclude a transitional arrangement dummy for the one observation in 2007, since (i) theimmigration conditions did not change in the EU-15 for Bulgaria and Romania after 2007 with theexception of Sweden and Finland which are no main destinations for the NMS-2, and (ii) one yearis not sufficient to identify this variable properly,IAB 21

A detailed description of the data set and the descriptive statistics is available from theauthors upon request.5 Estimation resultsThe estimation results are displayed in Table 5.1. We have estimated four specificationsof the model here. First, we estimated a simple fixed effects model which considers theincome difference between the EU-15 and the sending country and the immigrationrestrictions – including the interaction terms between the immigration restrictions andthe income differential – only. Second, we employed a fixed effects model whichconsiders in addition the employment rates in the EU-15 and the sending countries. Ascan be seen in the regression diagnostics, the explanatory power of the second model ishigher and the forecasting error substantially lower. The forecasting error has beencalculated for the ten NMS for the years 2001 to 2007 by using the root mean squarederror (RMSE) and the root mean squared percentage error (RMSPE) as forecastingcriteria. Third, we estimated this model also with Feasible GLS and cross-sectionalweights allowing for heteroscedasticity in the disturbances. Testing this model againstthe second specification suggests that heteroscedasticity is present. Moreover, thepredictive power of the model is higher compared to the second model. Finally, weestimated the same model allowing furthermore for serial correlation in the error termssince our specification tests suggest that the disturbances are indeed serially correlated.The forecasting error declines however only marginally in this specification compared tothe third one. The last model is our preferred specification which we use for thecalculation of the forecasts. 99 The specification tests are available from the authors upon request.IAB 22

Table 5.1Estimation resultsModel (1) Model (2) Model (3) Model (4)coefficient t -statistics coefficient t -statistics coefficient t -statistics coefficient t -statisticsln (mst i,t-1 ) 0.963 *** 48.92 0.957 *** 44.12 0.958 *** 50.7 0.960 *** 51.12ln (y eu,t-1 /y i,t-1 ) 0.002 * 1.64 0.002 * 1.71 0.003 *** 2.63 0.003 *** 2.67ln (e eu,t-1 ) 0.014 1.17 0.012 1.19 0.011 1.17ln (e i,t-1 ) -0.004 -0.56 -0.005 -0.84 -0.005 -0.83TRANS it ×ln (y eu,t-1 /y i,t-1 ) -0.002 * -1.65 0.001 0.28 0.001 0.46 0.001 0.47TRANS it ×ln (e eu,t-1 ) 0.088 ** 2.17 0.082 ** 2.51 0.082 ** 2.51TRANS it ×ln (e i,t-1 ) -0.020 -1.21 -0.016 -1.26 -0.016 -1.25GUEST it ×ln (y eu,t-1 /y i,t-1 ) -0.002 -1.58 -0.007 * -1.66 -0.004 -0.83 -0.004 -0.79GUEST it ×ln (e eu,t-1 ) -0.142 -1.47 -0.037 -0.35 -0.037 -0.32GUEST it ×ln (e i,t-1 ) 0.013 0.72 -0.002 -0.1 -0.003 -0.11RESTR it ×ln (y eu,t-1 /y i,t-1 ) -0.004 *** -3.15 -0.003 ** -2.4 -0.003 *** -2.95 -0.003 *** -3.00RESTR it ×ln (e eu,t-1 ) 0.011 0.88 0.007 0.63 0.007 0.65RESTR it ×ln (e i,t-1 ) 0.002 0.22 0.003 0.56 0.003 0.46IRON it ×ln (y eu,t-1 /y i,t-1 ) -0.001 -0.8 -0.006 * -1.94 -0.002 -0.68 -0.002 -0.63IRON it ×ln (e eu,t-1 ) -0.048 -0.36 -0.038 -0.33 -0.039 -0.33IRON it ×ln (e i,t-1 ) -3.582 -0.46 -0.471 -0.07 -0.341 -0.05TRANS it 0.001 1.12 0.003 * 1.85 0.002 * 1.95 0.002 * 1.89GUEST it 0.004 * 1.85 0.004 ** 2.01 0.002 1.24 0.002 1.13RESTR it 0.001 1.38 0.002 * 1.71 0.001 ** 1.53 0.001 ** 1.5IRON it -0.001 -0.45 -0.003 * -1.5 -0.002 * -0.97 -0.002 * -0.92WAR it 0.007 *** 5.02 0.008 *** 5.03 0.006 *** 3.2 0.006 *** 3.09obs. 552 552 529 529Wald ∠(51) 2 statistics 64,491 *** 65,651 *** 107,016 *** 109,977 ***R 2 0.87 0.88 - -RMSE (2001-07) 0.000185 0.000121 0.000025 0.000023RMSPE (2001-07) 0.480 0.146 0.060 0.057The dependent variable is ln (mst it ). -- ***, **, * denote the significance at the 1-, 5- and 10 per cent level, respectively.-- All models includecountry dummy variables.-- Model (1) and model (2) are estimated by LSDV. Model (3) is estimated by weighted Feasible GLS using theaverage GDP per capita in the sending country as a weight.-- Model (4) is estimated by weighted Feasible GLS allowing for panelspecificfirst-order autocorrelation.The qualitative results confirm largely our theoretical expectations. The income differencebetween the EU-15 and the sending countries has in all four specifications the expectedpositive sign and appears significant. The employment rate in the EU-15 has theexpected positive sign, while the employment rates in the sending countries have theexpected negative signs, although both variables do not appear as significant.The interaction dummy variables can only be interpreted together with the signs and thesize of the level dummy variables. As a consequence, the impact of the income gap aswell as the impact of the employment variables are either reduced or increase with therespective dummy variables. As expected, the civil wars in the former Yugoslavia haveexerted a strong positive impact on migration from the affected countries into the EU-15.6 A Projection of migration from the NMS-8 and the NMS-2The coefficients of model (4) in Table 5.1 are used for the simulation of future migrationmovements from the NMS into the EU-15. More specifically, we have calculated twoscenarios:IAB 23

• The first scenario assumes that the status quo regarding the institutionalconditions continues. This means that (i) the transitional arrangements for theNMS-8 are employed in the same way as during the 2004-2007 period, and (ii)the immigration conditions for Bulgaria and Romania remain the same as underthe bilateral agreements which are in place since the end of the 1990s.• The second scenario assumes that rules of the free movement of workers isintroduced in the entire EU, such that the values of all dummy variables andinteraction terms which capture the remaining immigration restrictions for theNMS are assumed to be zero.The results of the scenarios are displayed in Table 6.1 and Table 6.2. As a rule of athumb, our projections indicate that the present stock of migrants residing in the EU-15stood in 2007 at one half of the potential which will be realised by the year 2020 if thecurrent migration conditions prevail and at about two-fifth if the free movement isintroduced in the entire EU-15. During the same period of time, the net growth of theforeign population from the NMS-8 and the NMS-2 in the EU-15 will have declined fromabout 430,000 persons p.a. to 200,000 persons p.a. under the current institutionalconditions and from 515,000 persons p.a. to 235,000 persons p.a. under free movement.More specifically, the model predicts that the stock of migrants from the NMS-8 couldincrease from about 1.9 million in 2007 to 3.8 million in 2020 if the present restrictionsare maintained, while it could increase to 4.4 million under free movement in the sameperiod of time. This corresponds to 5.2 per cent of population of the sending countries(1.0 per cent of the population of the EU-15) under the current immigration conditionsand to 6.1 per cent of the population of the sending countries (1.13 per cent of thepopulation of the EU-15) under free movement. Thus, our scenario predicts thatremoving the immigration restrictions in important destinations such as Germany andAustria would trigger an additional migration of about 600,000 persons in the long-run ifmigrants from the NMS-8 behave in the same way as other migrants from the EU-15.However, the model does not make any predication on the allocation of migrants acrossdifferent destinations in the EU-15.Concerning migrants from Bulgaria and Romania, their stock could increase from about1.8 million persons in 2007 to 3.9 million in 2020 under the present immigrationrestrictions, while it could increase to 4.0 million when the free movement is introduced.This corresponds to 13.4 per cent of the population of the sending countries (1.0 per centof the population of the EU-15) under the current institutional conditions, and to almost14 per cent of the population of the sending countries (1.1 per cent of the population ofthe EU-15) when the free movement is introduced. Note again that the free movementscenario is derived from the assumption that migrants from Bulgaria and Romaniabehave in the same way as other EU-15 migrants. Given that income levels in Bulgariaand Romania deviate substantially from the sample mean, the forecasts for these twocountries are less reliable than those for the NMS. Thus, actual migration figures underIAB 24

free movement may deviate from the scenario presented here and the actual differencebetween the restricted and the free movement scenario might be larger.Table 6.1 Projection of migration stocks, 2007-2020 10CZ EE HU LT LV PL SK SI BG RO NMS-8 NMS-2 NMS-102006 79,094 32,020 106,618 102,455 40,826 1,039,283 109,336 30,265 246,187 1,045,873 1,539,898 1,292,060 2,831,9582007 105,918 33,998 119,465 111,631 46,554 1,280,756 120,728 32,347 272,521 1,550,240 1,851,395 1,822,761 3,674,1572008 119,002 36,861 136,072 127,552 55,159 1,437,604 146,399 30,036 293,502 1,722,887 2,088,685 2,016,389 4,105,0742009 130,731 39,479 151,031 142,414 63,136 1,583,665 170,450 27,713 313,881 1,889,149 2,308,619 2,203,030 4,511,6492010 141,177 41,866 164,429 156,270 70,515 1,719,462 192,956 25,378 333,679 2,049,272 2,512,052 2,382,951 4,895,0032011 150,408 44,033 176,346 169,166 77,326 1,845,495 213,991 23,035 352,913 2,203,493 2,699,799 2,556,406 5,256,2062012 158,487 45,992 186,861 181,150 83,597 1,962,242 233,626 20,686 371,602 2,352,038 2,872,641 2,723,640 5,596,2812013 165,477 47,753 196,049 192,267 89,355 2,070,160 251,928 18,333 389,764 2,495,124 3,031,321 2,884,888 5,916,2092014 171,436 49,326 203,981 202,558 94,625 2,169,687 268,962 15,979 407,415 2,632,962 3,176,555 3,040,377 6,216,9312015 176,422 50,722 210,725 212,065 99,434 2,261,240 284,789 13,625 424,571 2,765,753 3,309,021 3,190,324 6,499,3452016 180,487 51,950 216,346 220,826 103,803 2,345,219 299,468 11,273 441,248 2,893,689 3,429,372 3,334,937 6,764,3082017 183,683 53,018 220,907 228,879 107,756 2,422,005 313,055 8,925 457,460 3,016,957 3,538,228 3,474,417 7,012,6452018 186,059 53,936 224,466 236,259 111,315 2,491,965 325,604 6,583 473,224 3,135,735 3,636,186 3,608,959 7,245,1452019 187,662 54,711 227,079 243,000 114,499 2,555,446 337,167 4,248 488,552 3,250,194 3,723,812 3,738,746 7,462,5582020 188,536 55,352 228,802 249,135 117,328 2,612,781 347,793 1,921 503,459 3,360,499 3,801,648 3,863,958 7,665,6052006 79,094 32,020 106,618 102,455 40,826 1,039,283 109,336 30,265 246,187 1,045,873 1,539,898 1,292,060 2,831,9582007 105,918 33,998 119,465 111,631 46,554 1,280,756 120,728 32,347 272,521 1,550,240 1,851,395 1,822,761 3,674,1572008 135,413 39,185 153,674 129,543 56,627 1,437,886 143,097 36,031 310,851 1,747,009 2,131,456 2,057,860 4,189,3162009 163,082 44,064 185,751 146,413 66,077 1,585,241 164,112 39,484 346,668 1,933,606 2,394,224 2,280,274 4,674,4982010 189,012 48,649 215,797 162,287 74,934 1,723,282 183,837 42,718 380,092 2,110,484 2,640,516 2,490,576 5,131,0922011 213,284 52,954 243,907 177,214 83,225 1,852,451 202,332 45,743 411,240 2,278,080 2,871,110 2,689,320 5,560,4302012 235,978 56,993 270,174 191,236 90,977 1,973,172 219,656 48,569 440,221 2,436,809 3,086,755 2,877,030 5,963,7852013 257,170 60,777 294,687 204,397 98,216 2,085,848 235,864 51,205 467,139 2,587,071 3,288,163 3,054,210 6,342,3742014 276,930 64,318 317,530 216,736 104,967 2,190,867 251,010 53,661 492,097 2,729,246 3,476,019 3,221,343 6,697,3622015 295,329 67,629 338,783 228,294 111,253 2,288,598 265,144 55,945 515,191 2,863,703 3,650,974 3,378,894 7,029,8682016 312,432 70,721 358,524 239,106 117,096 2,379,396 278,314 58,066 536,511 2,990,789 3,813,655 3,527,300 7,340,9552017 328,303 73,603 376,827 249,208 122,519 2,463,599 290,569 60,031 556,147 3,110,842 3,964,659 3,666,989 7,631,6482018 343,002 76,286 393,762 258,635 127,542 2,541,530 301,951 61,849 574,183 3,224,181 4,104,557 3,798,364 7,902,9212019 356,586 78,780 409,396 267,419 132,185 2,613,500 312,504 63,526 590,699 3,331,115 4,233,896 3,921,814 8,155,7102020 369,111 81,094 423,796 275,591 136,466 2,679,804 322,269 65,070 605,772 3,431,938 4,353,200 4,037,710 8,390,910Own Projection. See text for assumptions.forecast under status quo conditions (nationals residing in the EU-15 in persons)forecast under free movements of workers (nationals residing in the EU-15 in persons)The annual net immigration or, more precisely, the net growth of the number of foreignresidents from the NMS-8 will decline from about 237,000 persons at the beginning ofthe projection period to 78,000 in 2020 under the transitional arrangements. In case ofintroducing the free movement, this figure will increase to about 280,000 persons p.a. atthe beginning of the projection period. The net increase of the foreign residents from theNMS-2 is estimated to be about 194,000 persons at the beginning of the projectionperiod and at 125,000 persons at the end under the current immigration restrictions. Anintroduction of the free movement will increase this figure to 235,000 persons p.a. at thebeginning of the projection period. Compared to the average net inflows during the firstthree years under the transitional arrangements our model predicts that the net inflowswill slightly decline, which can be already observed in 2008 e.g. in the UK.10 The start values of the migration stocks deviate slightly from those provided in Deliverable 2since the data sources on which the estimates are based differ for consistency reasons in somecountries from those presented in Deliverable 2.IAB 25

Table 6.2 Projection of the net growth of migration stocks, 2008-2020CZ EE HU LT LV PL SK SI BG RO NMS-8 NMS-2 NMS-102007 26,824 1,978 12,846 9,175 5,727 241,474 11,392 2,081 26,334 504,367 311,498 530,701 842,1992008 13,084 2,863 16,607 15,921 8,605 156,848 25,671 -2,310 20,981 172,647 237,289 193,627 430,9172009 11,729 2,619 14,959 14,863 7,977 146,061 24,050 -2,324 20,379 166,262 219,935 186,641 406,5762010 10,446 2,387 13,397 13,855 7,379 135,797 22,506 -2,335 19,798 160,123 203,433 179,921 383,3542011 9,230 2,167 11,917 12,896 6,811 126,033 21,035 -2,343 19,234 154,221 187,747 173,455 361,2022012 8,079 1,959 10,515 11,984 6,271 116,747 19,635 -2,349 18,689 148,545 172,841 167,234 340,0752013 6,990 1,761 9,188 11,116 5,758 107,918 18,302 -2,353 18,162 143,087 158,681 161,248 319,9292014 5,960 1,573 7,932 10,291 5,271 99,527 17,034 -2,354 17,651 137,838 145,233 155,489 300,7222015 4,985 1,396 6,744 9,507 4,808 91,553 15,827 -2,354 17,156 132,791 132,466 149,947 282,4132016 4,065 1,228 5,621 8,761 4,369 83,979 14,679 -2,352 16,677 127,937 120,350 144,613 264,9642017 3,196 1,068 4,560 8,053 3,953 76,787 13,587 -2,348 16,213 123,268 108,857 139,481 248,3382018 2,376 918 3,559 7,380 3,558 69,959 12,549 -2,342 15,764 118,778 97,957 134,541 232,4992019 1,603 775 2,614 6,741 3,184 63,481 11,563 -2,335 15,328 114,459 87,626 129,787 217,4132020 874 641 1,723 6,135 2,829 57,336 10,626 -2,327 14,906 110,305 77,837 125,212 203,0482007 26,824 1,978 12,846 9,175 5,727 241,474 11,392 2,081 26,334 504,367 311,498 530,701 842,1992008 29,495 5,187 34,210 17,912 10,073 157,130 22,369 3,684 38,329 196,769 280,060 235,098 515,1582009 27,669 4,879 32,077 16,869 9,450 147,354 21,015 3,454 35,817 186,597 262,768 222,414 485,1822010 25,930 4,585 300,455 15,875 8,857 138,041 19,725 3,234 33,425 176,879 516,701 210,303 727,0042011 24,272 4,305 28,110 14,926 8,291 129,169 18,495 3,025 31,148 167,596 230,595 198,743 429,3382012 22,694 4,038 26,267 14,022 7,752 120,721 17,324 2,826 28,981 158,729 215,645 187,710 403,3542013 21,191 3,784 24,513 13,161 7,239 112,676 16,208 2,636 26,919 150,262 201,408 177,180 378,5892014 19,761 3,542 22,843 12,340 6,751 105,019 15,146 2,456 24,958 142,176 187,855 167,134 354,9892015 18,399 3,311 21,253 11,557 6,286 97,732 14,134 2,284 23,093 134,456 174,956 157,550 332,5052016 17,103 3,091 19,741 10,812 5,844 90,798 13,171 2,121 21,321 127,087 162,681 148,407 311,0882017 15,871 2,882 18,303 10,103 5,423 84,203 12,254 1,965 19,636 120,053 151,004 139,689 290,6922018 14,699 2,683 16,935 9,427 5,023 77,932 11,382 1,818 18,036 113,340 139,898 131,375 271,2742019 13,584 2,494 15,635 8,784 4,642 71,970 10,553 1,677 16,516 106,934 129,339 123,450 252,7892020 12,526 2,314 14,399 8,172 4,281 66,304 9,765 1,544 15,073 100,823 119,304 115,895 235,199Own Projection. See text for assumptions.forecast under status quo conditions (nationals residing in the EU-15 in persons)forecast under free movements of workers (nationals residing in the EU-15 in persons)The scenarios presented in Table 6.1 and Table 6.2 refer to our point estimates.However, actual migration stocks and flows may deviate substantially from the pointestimates. The forecast intervals which we have derived by bootstrapping are prettylarge: In Poland, the lower bound of the 95-per cent interval stands at about two millionpersons, while the upper bound predicts about 3.2 million persons in 2020 (Figure 6.1).Similarly, in Romania the lower forecasting bound amounts to about 3 million persons,while the upper bound estimates the migration potential in 2020 at about 3.7 millionpersons (Figure 6.2). Overall, we expect that the migration potential from the NMS canbe about one-third above or below the point forecast of the migration stock in 2020.IAB 26

www.netzerotools.comRemote Control - Description of CommandsCommands and Queries 4[:SENSe]:SWEep1 Configures memory recording of sampled values (REALtime).[:SENSe]:SWEep2 Configures memory recording of averaged values (AVERage).The settings for memory recording of averaged and sampled values share the sameconfiguration space and all of the settings must be set when changing from averaged tosampled recording or vice versa. The [SENSe:]SWEep1|2[:STATe] OFF commands resetthe settings to default values except for triggers.[SENSe:]SWEep1|2:TIME | MAXDescriptionThis command specifies maximum memory record length in seconds. This record lengthincludes pretrigger. If the synchronization is on, then the max. recording duration forSWEep2 is directly dependent on exact number of averaging intervals that will berecorded, which is calculated as specified record length / nominal averaging interval.Parameter(s) | MAX*RST stateMAXExample(s)SWE1:TIME 1.0SWE1:TIME? Response: 1.0Invalidates-Invalidated by-[SENSe:]SWEep1|2:TIME:MAX?DescriptionReturns maximum recording time in seconds according to current memory recordingsettings (total amount of available memory, set of variables to record, sampling factorand instrument's sampling frequency).Response*RST stateMaximum recording time according to *RST settings for memory recording.Example(s)SWE1:TIME:MAX?Invalidates-Invalidated by-www.netzerotools.com4-47

insufficient to identify the parameters of the model properly. Second, the free movementscenario assumes that the slope parameters for the explanatory variables such as theincome difference and the employment rates are the same under free movement for theEU-15 sending countries and the NMS. This need, however, not to be the case. Third,particularly the migration data used for the estimates are subject to measurement errorwhich may bias the results in one way or another. Finally, the projections presented hereare based on estimates of long-run equilibrium relationships between the migrationstocks and the explanatory variables and the speed of adjustment to these long-runrelationships. The estimates do therefore not capture short-term fluctuations in thebusiness cycle appropriately, such that short-term migration movements may deviatesubstantially. This is particularly relevant in the context of the current financial crisis (seebelow).Thus, the projections presented here provide no more than a clue to the possibledevelopment of future migration movements from the NMS and should therefore beinterpreted with great care.7 The impact of the financial crisisThe current financial crisis may reduce the short-term migration substantially comparedto the projections presented in Table 6.1 and Table 6.2. It is an open question at presentwhether the NMS or the EU-15 will be more than proportionally affected by the financialcrisis. According to the recent forecasts, important sending countries such as Poland andRomania are less affected by the decline in GDP growth than the EU-15 countries, whileothers such as Hungary and the Baltic countries are more than proportionally affected.Nevertheless, since Poland and Romania alone account for about 80 per cent of themigrant population, these developments would reduce the short-term migrationpotential.More importantly, it is worthwhile noting that employment opportunities in the receivingand the sending countries do not affect the scale of migration in a symmetric way.Migration is largely driven by the opportunity to work, which in turn depends on theopportunity to find employment in the receiving countries. If employment opportunitiesin the receiving countries tend to decline, net immigration contracts irrespective ofmigration conditions in the sending countries. In the two main destinations of migrantsfrom the NMS in the EU-15 in absolute terms, the United Kingdom and Spain,unemployment has already started to increase substantially in the course of the currentfinancial crisis. Moreover, the prospects are bleak for 2009 according the forecasts of thenational governments and the European Commission. As a consequence, immigrationfrom the NMS will decline in these destinations, while return migration will tend toincrease. Net migration figures might thus decline or even become negative in the courseof the crisis, although the exact impact is uncertain at the present stage. Altogether,labour mobility between the EU-15 and the NMS will act as a buffer for natives in thereceiving countries in the current crisis, while it might further increase unemployment inthe sending countries if return migration becomes large.IAB 28

8 ReferencesAlecke, B., P. Huber, and G. Untiedt (2001), What a difference a constant makes: howpredictable are international migration flows?, in: OECD (ed.): Migration Policiesand EU Enlargement. The Case of Central and Eastern Europe, Paris, pp. 63-78.Alvarez-Plata, P., H. Brücker, and B. Siliverstovs (2003), Potential Migration from Centraland Eastern Europe into the EU-15 – An Update, Report for the EuropeanCommission, DG Employment, Social Affairs and Equal Opportunities, Brussels.Arellano, M., S. Bond, (1991), Some Tests of Specification for Panel Data: Monte CarloEvidence and an Application to Employment Equations, Review of EconomicStudies, 58, pp. 277-97.Arellano, M., O. Bover (1995), Another look at the instrumental-variable estimation oferror-components models, Review of Economic Studies, 58, pp. 277-97.Baltagi, B.H., G. Bresson, and A. Pirotte (2002), Comparison of Forecast forHomogeneous, Heterogeneous and Shrinkage Estimators. Some EmpiricalEvidence for US Electricity and Natural-Gas Consumption, Economics Letters, 76,pp. 375-82.Baltagi, B.H., J.M. Griffin, and W. Xiong (2000), To Pool of Not to Pool: Homogeneousversus Heterogeneous Estimators Applied to Cigarette Demand, Review ofEconomics and Statistics, 82(1), pp. 117-126.Baltagi, B.H., J.M. Griffin (1997), Pooled Estimators vs. Their HeterogeneousCounterparts in the Context of Dynamic Demand for Gasoline, Journal ofEconometrics, 77, pp. 303-27.Baltagi, B.H. (1995), Econometric Analysis of Panel Data. Chichester: John Wiley andSons.Banerjee, A., J.J. Dolado, D.F. Henry, and G.W. Smith (1986), Exploring EquilibriumRelationships in Econometrics through Static Models: Some Monte Carlo Evidence,Oxford Bulletin of Economics and Statistics, 48(3), pp. 253-77.Banerjee, B., S.M. Kanbur (1986), On the estimation and specification of macro ruralurbanmigration functions, Oxford Bulletin of Economic and Statistics, 43, pp. 7-29.Barro, R., X. Sala-i-Martin (1991), Convergence across States and Regions, BrookingsPapers on Economic Activity, 1, pp. 107-182.Barro, R., X. Sala-i-Martin (1995), Economic Growth, New York et al.: McGraw-Hill.Bauer, T., K.F. Zimmermann (1999), Assessment of Possible Migration Pressure and ItsLabour Market Impact Following EU Enlargement to Central and Eastern Europe,IZA Research Report Nr. 3, IZA, Bonn.Blundell, R.W., S.R. Bond (1998), Initial conditions and moment restrictions in dynamicpanel data models, Journal of Econometrics, 87, pp. 115-143.Boeri, T., H. Brücker (2005), Why are Europeans so tough on migrants?, EconomicPolicy, 44, pp. 1-76.IAB 29

Boeri, T., H. Brücker et al. (2001), The Impact of Eastern Enlargement on Employmentand Labour Markets in the EU Member States, EuropeanBerlin/Brussels/Milano.Commission,Brücker, H. (2001), Die Folgen der Freizügigkeit für die Ost-West-Migration.Schlussfolgerungen aus einer Zeitreihenanalyse der Migration nach Deutschland,Konjunkturpolitik- Applied Economics Quarterly, Supplement 52, pp. 17-54.Brücker, H., P.J.H. Schröder (2006), International Migration With Heterogeneous Agents:Theory and Evidence, IZA-Discussion Paper, IZA, Bonn.Brücker, H., B. Siliverstovs (2006a), Estimating and forecasting European migration:methods, problems and results, Zeitschrift für Arbeitsmarktforschung – Journal forLabour Market Research, 39(1), 35-53.Brücker, H., B. Siliverstovs (2006b), On the Estimation and Forecasting of InternationalMigration: How Relevant Is Heterogeneity Across Countries?, Empirical Economics,31(3), pp. 735-754.Bruder, J. (2003), East-West Migration 2004-2015, Thünen Series of Applied EconomicsWorking Papers 40, University of Rostock.Burda, M.C. (1995), Migration and the Option Value of Waiting, Economic and SocialReview, 27(1), pp. 1-19.Daveri, F., R. Faini (1999), Where Do Migrants Go?, Oxford Economic Papers, 51, pp.595-622.Decressin, J., A. Fátas (1995), Regional Labour Market Dynamics in Europe, EuropeanEconomic Review, 39(9), pp. 1627-57.Djajic, S., R. Milbourne (1988), A general equilibrium model of guestworker migration,Journal of International Economics, 25, pp. 335-351.Dustmann, C., M. Casanova, M. Fertig, I. Preston, and C.M. Schmidt (2003), The impactof EU enlargement on migration flows, Home Office Online Report 25/03,http://www.homeoffice.gov.uk/rds/pdfs2/rdsolr2503.pdf.Engle, R.F., C. Granger (1987), Co-integration and error correction: interpretation,estimation, and testing, Econometrica, 66, pp. 251-76.Faini, R., A. Venturini (1995), Migration and Growth: The Experience of Southern Europe,CEPR Discussion Paper No. 964.Fachin, S. (2007), Long-run trends in internal migrations in Italy: a study in panelcointegration with dependent units, Journal of Applied Econometrics, 22(2), pp.401-428.Fassmann, R., C. Hintermann, (1997), Migrationspotential Ostmitteleuropa: Struktur undMotivation potentieller Migranten aus Polen, der Slowakei, Tschechien undUngarn, ISR-Scientific Report 15, Innsbruck.Fassmann, R., R. Münz (2002), EU-Enlargement and the future of east-west migration,in: IOM (ed.), New challenges for migration policy in Central and Eastern Europe,Geneva, IOM.IAB 30

Fertig, M. (2001), The Economic Impact of EU Enlargement: Assessing the MigrationPotential, Empirical Economics, 26(4), pp. 707-720.Fertig, M., C.M. Schmidt (2001), Aggregate Level Migration Studies as a Tool forForecasting Migration Streams, IZA Discussion Paper 183, IZA, Bonn.Flaig, G. (2001), Die Abschätzung der Migrationspotenziale der osteuropäischen EU-Beitrittsländer, Konjunkturpolitik – Applied Economics Quarterely, Supplement52m, pp. 55-76.Fouarge, D., P. Ester (2007), Mapping migration intentions in Europe: Perceived costsand benefits, Institute for Labour Studies, Tilburg University.Gordon, I., I. Molho (1995), Duration dependence in migration behaviour: cumulativeinertia versus stochastic change, Environment and Planning, A 27, pp. 1961-1975.Granger, C. W. J., P. Newbold (1974), Spurious regression in econometrics, Journal ofEconometrics, 2(2), pp. 111-20.Hamilton, J. (1994) Time series analysis, Princeton, NJ: Princeton University Press.Harris, J.R., M.P. Todaro (1970), “Migration, Unemployment and Development: A Two-Sector-Analysis”, American Economic Review, 60, pp.126-142.Hatton, T.J. (1995), A Model of U.K. Emigration, 1870-1913, Review of Economics andStatistics, 77, pp. 407-415.Hatton, T.J., J.G. Williamson (2002), What Fundamentals Drive World Migration?, NBERWorking Paper 9159.Helmert, T. (2007), Schätzung von Migrationspotenzialen: Analyse der Prognose-qualitätverschiedener Schätzverfahren für Paneldaten, Diplom-Thesis, University ofBamberg, Department of Economics.Hicks, J.R. (1932), The Theory of Wages, London: McMillan.Hille, H. and T. Straubhaar (2001), “The Impact of EU-Enlargement on MigrationMovements and Economic Integration: Results of Recent Studies”, in: OECD (ed.),Migration Policies and EU-Enlargement, Paris, pp. 79-100.Hönekopp, E. (2001), Überblick über die Ergebnisse bisher vorliegender Schätzungenzum Migrationspotential im Falle einer Arbeitskräftefreizügigkeit im Rahmen derOsterweiterung der EU, Vortrag auf der Fachkonferenz „Mehr Konkurrenz um Lohnund Arbeit? Auswirkungen der Osterweiterung der EU auf die Arbeitsmärkte –Optionen für die Politik“, Akademie für Politische Bildung Tutzing, Tutzing, 14.-16.5.2001.Im, K.S., M.H. Pesaran, and Y. Shin (2003), Unit root testing in heterogeneous panels,Journal of Econometrics, 115(1), pp. 53-74.Judson, R.A., A.K. Owen (1999): Estimating dynamic panel data models: a guide formacroeconomists, Economic Letters, 65, pp. 9-15.Kiviet, J. (1995), „On Bias, Inconsistency and Efficiency of Various Estimators in DynamicPanel Data Models”, Journal of Econometrics, 68, pp. 53-78.IAB 31

Krieger, T. (2003), Migration trends in an enlarged EU, European Foundation for theImprovement of Working and Living Conditions, Dublin.Layard, R., O. Blanchard, R. Dornbusch, and P. Krugman (1992): East-West Migration:The Alternatives. Boston: MIT Press.Maddison, A. (1995), Monitoring the World Economy 1820-1992, Paris: OECD.Massey, D.S., F.G. Espana (1987), The Social Process of International Migration, Science,237, pp. 733-738.Massey, D.S., L. Goldring, and J. Durand (1984), Continuities in Transnational Migration:An Analysis of Nineteen Mexican Communities, American Journal of Sociology,99(6), 1492-1533.Mayda, A.M. (2005), International Migration: A Panel Data Analysis of Economic andNon-Economic Determinants, IZA Discussion Paper 1590.Mincer, J. (1978), Family migration decisions, Journal of Politcal Economy, 86, pp. 749-73.Nickell, S. (1981), Biases in dynamic models with fixed effects, Econometrica, 49, pp.1417-1426.Patterson, K. (2000), An Introduction to Applied Econometrics: A Time Series Approach,New York: Palgrave.Pedersen, P., M. Pytlikova, and N. Smith (2004), Selection or network effects? Migrationflows into 27 OECD countries, IZA Discussion Paper 1104, IZA, Bonn.Pesaran, M.H., Y. Shin, and R.P. Smith (1997), Pooled Mean Group Estimation ofDynamic Heterogeneous Panels, mimeo, Trinity College, Dublin (revised:November 1998).Pesaran, M.H., R.P. Smith (1995), Estimating long-run relationships from dynamicheterogeneous panels, Journal of Econometrics, 68, pp. 79-113.Portes, J., S. French (2005), The Impact of Free Movement of Workers from Central andEastern Europe on the UK Labour Market: Early Evidence, Department for Workand Pensions Working Paper 18.Pytlikova, M. (2007), EU Enlargement: Migration from the New Member States, Mimeo,Aarhus School of Business, CIM and Department of Economics.Ravenstein, E. (1889), The laws of migration, Journal of the Statistic Society, 59, pp.214-301.Schwarze, J. (1996), Beeinflusst das Lohngefälle zwischen Ost- und Westdeutschland dasMigrationsverhalten der Ostdeutschen?, Allgemeines Statistisches Archiv, 80, pp.50-68.Sinn, H.-W., G. Flaig, M. Werding, S. Münz, N. Düll and H. Hoffmann (2001), EU-Erweiterung und Arbeitskräftemigration. Wege zu einer schrittweisen Annäherungder Arbeitsmärkte, Munich: Ifo-Institut für Wirtschaftsforschung.Sjaastad, L.A. (1962), The costs and returns of human migration, Journal of PoliticalEconomy, 70(5), pp. 80-93.IAB 32

Stark, O. (1984), Rural-to-Urban Migration in Less Developed Countries: A RelativeDeprivation Approach, Economic Development and Cultural Change, 32(3), pp.475-486.Stark, O. (1991), The Migration of Labour, Oxford and Cambridge, MA: Basil Blackwell.Stock, J.H. (1987), Asymptotic Properties of Least Squares Estimators of CointegratingVectors, Econometrica, 55, pp. 1035-1056.Straubhaar, T. (2002), Ost-West Migrationspotenzial: Wir groß ist es?, Jahrbücher fürNationalökonomie und Statistik, 222, pp. 20-41.Todaro, M. (1969), A model of labour migration and urban unemployment in lessdeveloped countries, American Economic Review, 59(1), pp. 139-48.University of Groningen (2007), Groningen Growth and Development Centre, PurchasingPower Parity Estimates, Online.Wallace, C. (1998), Migration Potetnial in Central and Eastern Europe, IOM CooperationCentre for Europe and Central Asia, Geneva.Walterskirchen, E., R. Dietz (1998), Auswirkungen der EU-Erweiterung auf denösterreichischen Arbeitsmarkt, Studie des österreichischen Instituts fürWirtschaftsforschung, Wien.Zaiceva, A., K.F. Zimmermann (2007), Migration Potential and its Labor Market Impactafter EU Enlargement, Paper presented at the Second IZA Workshop: EUEnlargement and the Labor Market, September.Zaiceva, A. (2006), Reconciling the Estimates of Potential Migration into the EnlargedEuropean Union, IZA Discussion Paper 2519, IZA, Bonn.Zaiceva, A. (2004), Implications of EU accession for international migration: anassessment of potential migration pressure, CES Ifo Working Paper 1184.IAB 33

European Integration ConsortiumIAB, CMR, fRDB, GEP, WIFO, wiiwLabour mobility within the EU in the context of enlargement and the functioningof the transitional arrangementsVC/2007/0293Deliverable 4Institute for Employment Research (IAB)The macroeconomic consequences of labour mobilityTimo Baas, Herbert Brücker, Andreas Hauptmann and Elke J. JahnAbstractThis deliverable examines the impact of the EU Eastern enlargement on wages,unemployment and other macroeconomic variables. For this purpose we employ twogeneral equilibrium models which both analyse the economic consequences of labourmobility in the context of the EU Eastern enlargement in a setting of imperfect labourmarkets. The first model is based on a nested production function, which enables us toexamine the migration effects for the different cells of the labour market. The secondmodel is based on a CGE-framework, which allows us to consider the links betweenlabour migration, trade and international capital mobility. Moreover, it enables us toexamine the sectoral implications of labour mobility in detail. Both models assume thatcapital stocks adjust to labour supply shocks at least in the long-run.We analyse the impact of Eastern enlargement during the years from 2004 to 2007 andcompare it to the situation where no enlargement took place. We find remarkably similarresults in both simulation models. The EU Eastern enlargement has only a moderateimpact on labour markets. Especially in the long-run, labour mobility is neutral for wagesin both the sending and the receiving countries and has only a negligible impact on theunemployment rate. Nevertheless our simulations suggest that increased labour mobilityyields an aggregate gain in terms of GDP in the enlarged EU.Furthermore we examine the potential effects of introducing free movement in theenlarged EU. Based on our projections we contrast a prolongation of the migrationrestrictions until the end of the transitional periods with a scenario where we allow forfree movement already at the beginning of 2009. Although the impact on the entire EU israther small, single receiving countries are affected differently. This is becauseintroducing free movement also changes the regional distribution of migration flows.The views and opinions expressed in this publication are those of the authors and do not necessarilyrepresent those of the European Commission.IAB

Contents1 Introduction..................................................................................................... 12 A review of the literature ................................................................................... 23 Theoretical considerations.................................................................................. 44 Migration scenarios ........................................................................................... 64.1.1 Transitional arrangements vs. no EU Eastern enlargement .................. 64.1.2 Free movement vs. prolongation of transitional arrangements ............. 94.1.3 Accounting for differences between migrants' jobs and skills ............. 115 Assessing the Labour Market Effects: A Wage Curve Approach .............................. 125.1 Theoretical background ............................................................................ 125.2 Data...................................................................................................... 145.3 Estimation results ................................................................................... 165.3.1 Adjustment of capital stocks ......................................................... 165.3.2 Estimates of the wage curve ......................................................... 175.3.3 Estimates of the elasticities of substitution...................................... 195.4 Simulation results ................................................................................... 215.4.1 The impact of Eastern enlargement on the UK and Germany,2004-2007 ................................................................................. 225.4.2 The impact of Eastern enlargement on the EU-25, 2004-2007 ........... 235.4.3 The impact of migration from Bulgaria and Romania, 2004-2007................................................................................................ 285.4.4 The impact of transitional arrangements and the freemovement of workers from the NMS-8, 2008–2011 ......................... 315.4.5 The impact of transitional arrangements and the freemovement of workers from Bulgaria and Romania, 2008-2014 .......... 325.5 Conclusions ............................................................................................ 346 The macroeconomic consequences of labour mobility: The impact ofmigration, trade and capital mobility in a multisectoral CGE model......................... 356.1 Outline of the model ................................................................................ 356.2 Data and calibration of the model .............................................................. 366.3 Simulation results ................................................................................... 376.3.1 Germany.................................................................................... 406.3.2 UK ............................................................................................ 426.3.3 Hungary .................................................................................... 456.3.4 Poland ....................................................................................... 496.3.5 Slovenia .................................................................................... 526.3.6 Slovakia..................................................................................... 556.4 Conclusions ............................................................................................ 587 References..................................................................................................... 598 Appendix ....................................................................................................... 628.1 Appendix A............................................................................................. 628.2 Appendix B............................................................................................. 64

8.3 Appendix C............................................................................................. 70TablesTable 1: Migration stock for the NMS-8, 2003-2007 scenario..................................... 7Table 2: Migration stock for the NMS-2, 2003-2007 scenario..................................... 8Table 3: Migration stock forecasts for the NMS-8 (2007-2011) ................................ 10Table 4: Migration stock forecasts for the NMS-2 (2007-2014) ................................ 11Table 5: Adjustment of the capital-labour ratio in EU countries................................ 17Table 6: Estimate of the dynamic wage curve model.............................................. 19Table 7: Estimates of the inverse elasticity of substitution ...................................... 20Table 8: Estimates of the inverse elasticity of substitution: a literature review ........... 21Table 9: The impact of Eastern enlargement on the UK and Germany, 2004-2007...... 23Table 10: The macroeconomic impact of migration from the NMS-8, 2004-2007 .......... 24Table 11: The impact of migration from the NMS-8 on wages, 2004-2007 .................. 26Table 12: The impact of migration from the NMS-8 on unemployment, 2004-2007 ...... 27Table 13: The macroeconomic impact of migration from the NMS-2, 2004-2007 .......... 28Table 14: The impact of migration from the NMS-2 on wages, 2004-2007 .................. 30Table 15: The impact of migration from the NMS-2 on unemployment, 2004-2007 ...... 31Table 16: Short-run effects of transitional arrangements and the free movement ofworkers from the NMS-8, 2008-2011 ...................................................... 32Table 17: Short-run effects of transitional arrangements and the free movement ofworkers from Bulgaria and Romania, 2008-2014....................................... 33Table 18: Simulation Results, Key Macroeconomic Figures, NMS-8............................. 39Table 19: Simulation results Germany, key macroeconomic figures............................ 41Table 20: Simulation results Germany, sectoral impact ............................................ 42Table 21: Simulation results UK, key macroeconomic figures .................................... 44Table 22: Simulation results UK, sectoral impact ..................................................... 45

Table 23: Simulation results Hungary, key macroeconomic figures ............................ 46Table 24: Simulation results Hungary, sectoral impact ............................................. 48Table 25: Simulation results Poland, key macroeconomic figures ............................... 49Table 26: Simulation results Poland, sectoral impact................................................ 51Table 27: Simulation results Slovenia, key macroeconomic figures ............................ 52Table 28: Simulation results Slovenia, sectoral impact ............................................. 54Table 29: Simulation results Slovakia, key macroeconomic figures............................. 56Table 30: Simulation results Slovakia, sectoral impact ............................................. 57Table A1: The short-run effects of transitional arrangements and the free movement ofworkers from the NMS-8 on the structure of wages and unemployment, 2008-2011 .................................................................................................. 62Table A2: The short-run effects of transitional arrangements and the free movement ofworkers from the NMS-2 on the structure of wages and unemployment, 2008-2014 .................................................................................................. 63

1 IntroductionThis deliverable examines the impact of labour mobility on wages, (un-)employment,GDP and other macroeconomic variables in the context of the EU Eastern enlargement.Our analysis addresses both the destination and the sending country perspective. Wedistinguish two main labour supply shocks here: the migration from the NMS-8 and fromthe NMS-2 into the EU-15. The first group covers the eight Central and Eastern Europeancountries 1 which joined the EU in May 2004; the second group Bulgaria and Romaniawhich joined the EU in January 2007. The candidate countries, which may accede duringthe next decade, are not considered at this stage of the study, since Easternenlargement has only modestly affected migration from there, if at all.The study is based on two macroeconomic models which address different aspects of themacroeconomic implications of migration. The first model employs a general equilibriumframework for analysing the effects of migration in a setting with imperfect labourmarkets. The model uses a nested production function which groups the labour force byeducation, work experience, and national origin. This enables us to examine the wageand employment effects of migration for the different segments of the labour market.This model can be applied for both the analysis of the short-run and the long-run effectsof labour mobility.The second model analyses the labour market effects of labour mobility also on basis of amodel with imperfect labour markets. In contrast to the first model, the impact ofmigration on different industries is modelled within a computable general equilibrium(CGE) framework. This enables us to assess not only the sectoral impact of migration,but also the links between labour mobility and international trade and capital mobility. Inthis second model we focus on the analysis of the UK, Germany, Poland, Hungary,Slovenia, and Slovakia. The rather broad range of countries allows us, however, tocapture the different ways by which the sending and receiving countries in the enlargedEU are affected by labour mobility.The analysis of the impact of immigration on the destination and sending countries in theenlarged EU is carried out here for both models in two steps. In the first step, we analysethe impact of the actual migration movements which took place under the currentinstitutional and legal conditions during the years from 2004 to 2007 and contrast thiswith a counterfactual scenario of no EU enlargement. In the second step, based on ourprojections, we contrast a prolongation of the migration restrictions until the end of thetransitional periods with a scenario where we allow for free movement already at thebeginning of 2008. The purpose of these scenarios is to grasp the main changes inimmigration policies which have been carried out in the context of the EU Easternenlargement.1 Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovak Republic, and Slovenia.IAB 1

The remainder of this deliverable is organised as follows. First, we review the relevantliterature. Second, we discuss the theoretical considerations which form the basis of thelater analysis. Third, we describe the immigration scenarios employed in the simulations.Fourth, we present the first simulation model, the estimates of the relevant parametersand the simulation results. Fifth, we describe the CGE model which analyses the linksbetween migration, sectoral change, international trade and international capitalmovements and present the simulation results based on this model. The final sectiondraws conclusion on the macroeconomic and structural effects of migration in the contextof the EU’s Eastern enlargement.2 A review of the literatureThe impact of migration on wages and employment in the context of the EU’s Easternenlargement has been addressed meanwhile by numerous studies. We can distinguishthree strands in the literature: The first strand of literature is based on econometricestimates using the regional variance of the migration share for the identification of thewage and employment effects of immigration. The second approach uses the variance ofthe migration share across the education and experience cells of the labour market at thenational level for identification. Finally, the third approach uses CGE or othermacroeconomic models for the simulation of the labour market effects.The spatial correlation approach has been widely applied in the US and Europeanliterature during the 1990s for an evaluation of the labour market effects of immigration.Both the wage and employment effects of migration are small and seem to cluster aboutzero (see Borjas, 2003; Friedberg and Hunt, 1995, for a discussion). Recent metaanalysesof this literature indicate that an increase in the labour force by 1 per centreduces native wages by less than 0.1 per cent and increases the unemployment risk ofnatives by less than 0.1 percentage point (DeLonghi et al., 2005; 2006). A recent studyfor the UK based on this approach finds that immigration from the NMS has a smallpositive impact on wages and a small negative impact on unemployment of natives(Lemos and Portes, 2008), supporting earlier findings by Dustmann et al. (2005) for theUK. Both effects are, however, insignificant.The spatial correlation approach may yield spurious results if migrants are not randomlydistributed across locations. Large parts of this literature therefore rely either on naturalexperiments or use instrumental variable or difference-in-difference estimators inaddressing this endogeneity problem (see e.g. Dustmann and Glitz, 2005, for adiscussion). It remains nevertheless controversial whether the wage and employmenteffects of immigration can be properly identified by the spatial correlation approach.Another part of the empirical literature therefore uses the variance of migrants acrosseducation and experience cells in the labour market at the national level foridentification. In his seminal study, Borjas (2003) finds for the US that a 1 per centincrease of the labour force through immigration reduces native wages substantially byabout 0.3 to 0.4 per cent. Similar results are obtained by Aydemir and Borjas (2006). Incontrast, Ottaviano and Peri (2006) reconcile the findings of the spatial correlationstudies. They estimate that the impact of immigration on native wages is almost neutral,while foreign workers tend to lose substantially. Similar results have been obtainedIAB 2

ecently for the UK by Manacorda et al. (2006) and for Germany by Brücker and Jahn(2008), D’Amuri et al. (2008), and Felbermayr et al. (2008). All these studies find thatan increase of the foreign labour force by 1 per cent reduces native wages by less than0.1 per cent and increases native unemployment risks by less than 0.1 percentagepoints.The third strand of the literature addresses the macroeconomic impact of migration onbasis of general equilibrium models. This type of macroeconomic modelling is veryflexible and provides a comprehensive framework which facilitates the analysis of theinteraction between trade, migration and capital movements and their subsequent labourmarket impacts. A number of these studies have addressed the labour market effects ofimmigration in the context of the Eastern enlargement. The main focus of this literatureis on the changing skill composition of the labour force through immigration. Assumingthat the low-skilled and high-skilled labour force in Austria would increase by 10.5 and2.1 per cent, respectively, Keuschnigg and Kohler (1999) estimate a 5 per cent decreasein wages for low-skilled workers. Heijdra et al. (2002) estimate the effect of migrationfrom the NMS to Germany. They assume that migration from Eastern European countriesto Germany would rise from 550,000 in 2008 to 2.5 million in 2030, with 35 per cent ofthe migrant population entering the labour market. 40 per cent of the migrants areassumed to be skilled and 60 per cent unskilled. As a result, less skilled workers sufferfrom reduced wages and higher unemployment, while skilled labour benefits frommigration through higher wages and lower unemployment. Brücker and Kohlhaas (2004)find that, depending on the assumptions on the qualifications of the migrant population,wages can decline by 0.5–0.6 per cent for an immigration rate of 1 per cent of the labourforce, while the unemployment rate increases by 0.02–0.1 percentage points. In anotherstudy, Brücker (2007) demonstrates that if 4 per cent of the population from the NMSmigrate into the EU-15, the main winners of migration are the migrants themselves,while blue-collar workers are negatively affected through higher unemployment in thedestination countries.Altogether, this literature finds wage and employment effects of immigration which aresomewhat larger than those found by the econometric literature. However, the stillrelatively modest negative effects of immigration on wages and unemployment ofparticularly low-skilled workers are outweighed by positive and strong effects resultingfrom the integration of the NMS into the goods markets of the EU (e.g. Brown et al.,1995; Baldwin et al., 1997). Consequently, most models predict that Easternenlargement results in lower aggregate unemployment and higher wages in both the EU-15 and the NMS.Not surprisingly, all CGE models predict that enlargement increases the GDP in thereceiving countries and the total EU. In earlier studies, this effect was predicted to varybetween 0.1 per cent and 0.5 per cent in the EU-15, and between 5 per cent and 18 percent in the NMS. More recent studies, which take into account trade creation between theold and new member countries, estimate slightly larger effects on GDP of the EU-15.Boeri and Brücker (2005) estimate a 0.5 per cent gain in the income per capita if 3 percent of the population from the NMS migrate into the EU-15. However, these aggregateand per capita income gains may be reduced if rigidities in the labour market exist.Finally, analysing possible diversion effects due to transitional periods, Baas and BrückerIAB 3

(2008) conclude that the closure of labour markets in Germany has reduced the GDPeffect, while the opening-up of the UK has resulted in a higher GDP.Most studies addressing the macroeconomic effects of migration in the context of the EUEastern enlargement employ a CGE framework. A notable exception is the recent studyby Barrett et al. (2007). This study uses a large new Keynesian macroeconometric modelto describe the absorption of a labour supply shock triggered by the EU Easternenlargement. In contrast to the general equilibrium framework, these types ofmacroeconomic models are less rigorously founded on theoretical models but cover ahuge variety of economic relations. Interestingly enough, the differences between theBarrett et al. (2007) study and the results reported from the CGE literature are quitesmall. 23 Theoretical considerationsFrom a global perspective, international migration increases the productive use of humanresources and hence, global output. Many simulation models suggest that the gains fromopening labour markets to international migration can easily dwarf potential gains from afurther liberalization of international goods and capital markets (Hamilton and Whalley,1984). This has been also demonstrated for labour migration within the Europeancontinent (Boeri and Brücker, 2005).But international migration does not only create winners. The standard textbook model ofmigration predicts that international labour mobility generates aggregate gains fornatives in the receiving countries, while natives left behind in the sending countries tendto lose (e.g. Wong, 1995, Ch. 14). Moreover, production factors in receiving countrieswhich are net complements to migrant labour tend to win, while those which are netsubstitutes tend to lose. More specifically, labour is expected to lose at the destination.The converse applies to the sending countries.One key assumption of the textbook model of migration is that labour markets clear.Relaxing this assumption yields different results (Boeri and Brücker, 2005; Levine,1999). In case of rigid labour markets and unemployment, migrants can replace nativeworkers in recipient countries. Hence, unemployment can increase, which mayfurthermore trigger higher welfare expenditures for both natives and migrants. As aconsequence, natives in the receiving countries may lose, while those in the sendingcountries may gain. Considering labour market rigidities is particularly relevant in thecontext of this study, since many EU countries still suffer from high and persistingunemployment rates. The concern that migration from the new member states mayincrease unemployment is therefore one of the main arguments for the application oftransitional arrangements for the free movement of workers. Indeed we find in our2 Barell et al. (2007) find that immigration of 1 per cent of the population leads to a 1.1 per centincrease in GDP while Baas and Brücker (2008) report a 1 per cent increase in GDP.IAB 4

simulations rising unemployment and shrinking wages in the short-run, which are causedby wage rigidities.However, labour migration may have very different effects in the different cells of thelabour market. It may create additional labour demand for certain types of labour andmay reduce it for others. Depending on the wage flexibility in the different segments ofthe labour markets, it may therefore either increase or reduce aggregate unemployment.Moreover, depending on the elasticities of substitution between native and foreignlabour, labour immigration may increase wages and employment opportunities of nativesin the host countries, even if aggregate wages decline and the aggregate unemploymentis increasing (see e.g. Ottaviano and Peri, 2006, for US evidence).An important issue for an assessment of the migration impacts is the adjustment of othermarkets in the economy. The standard migration model is based on the assumption thatcapital stocks are fixed, which is hardly realistic if we consider that investors exploitprofit opportunities. Indeed, it is one of the few empirically supported facts in economicsthat the capital-output ratio and, hence, the productivity adjusted capital intensity ofproduction remains constant over time (Kaldor, 1961). This implies that capital stocksadjust in one way or another to labour supply shocks, which in turn implies that theaggregate impact of migration on wages is mitigated when capital adjusts in the longrun.We thus consider the adjustment of capital stocks here and examine empiricallywhether and to what extent capital stocks adjust even in the short-run.International links via goods and capital markets can further reduce the impact of labourmobility on wages and unemployment in the receiving and sending countries. Thestandard models of trade theory suggest that the impact of labour mobility on factorprices and employment opportunities is mitigated if migration, trade and capitalmovements are substitutes (see Venables, 1999, for a discussion). Under the extremeassumption that international demand on the goods markets is perfectly elastic,international migration has no impact on wages and employment opportunities. Althoughthis is empirically not very likely, trade and capital movements may contribute to reducethe migration impacts.Against this background the two types of models employed here may deliver slightlydifferent results: The first model analyses the domestic adjustment of economies mainlyvia the labour market. It considers the elasticities of substitution and complementaritiesin the different cells of the labour market in detail. Adjustments in other markets are onlyconsidered as long as they affect the capital-output ratio. Considering the capital-outputratio enables us, however, to capture the adjustment of capital stocks via domestic orinternational investment, which may be the most important channel of adjustment. Thesecond type of model goes beyond this in considering the adjustment of the sectoralstructure of the economy via international trade and shifts in the structure of demandand production. We therefore expect that the short-term migration impact on both thereceiving and the sending countries will be smaller in the second type of model.IAB 5

4 Migration scenariosThe analysis of the impact of migration on the destination and sending countries in theenlarged EU follows two questions. (i) What is the impact of Eastern enlargement duringthe years from 2004 to 2007 compared to a scenario where no enlargement took place?(ii) What are the potential implications if free movement is introduced in the entire EU atthe beginning of 2008 compared to a scenario where the present immigration restrictionsunder the transitional arrangements for the free movement of worker continue? Thepurpose of these scenarios is to grasp the main changes in immigration policies whichhave been carried out in the context of the EU Eastern enlargement.4.1.1 Transitional arrangements vs. no EU Eastern enlargementFirst we analyse the impact of the migration which took place since EU enlargement from2004 to 2007. As has been outlined in Deliverable 2, the EU Eastern enlargementinvolved a distinct increase in migration from the NMS-8 and a diversion of migrationflows away from Austria and Germany towards Ireland and the UK. In ourcounterfactual scenario we assume that the pre-enlargement conditions for migrationbetween the NMS on the one hand and the EU-15 on the other hand prevail. Thisscenario does not assume that no labour mobility takes place, but that both the overallscale and the regional distribution of immigration flows stay at their pre-enlargementlevels. We thus base the immigration from 2004 to 2007 on an extrapolation of theaverage immigration during the 1999-2003 period in this counterfactual scenario. Thisscenario is contrasted by the EU Eastern enlargement scenario. In the EU Easternenlargement scenario we have calculated the actual increase in the migration stocksbetween 2004 and 2007. 3 The difference between these two scenarios is treated here asthe “EU enlargement effect”, i.e. the migration effect which has been caused by the EU’sEastern enlargement. Table 1 displays the scenarios for the EU-15 and the individualreceiving countries from the NMS-8. 4 The foreign population from the NMS-8 in the EU-15 has increased from 874,000 in 2003 to 1.9 million persons in 2007 or by one millionpersons. According to our counterfactual scenario, the increase would have been a mere199,000 persons without enlargement, yielding a migration effect of 837,000 personswhich can be attributed to the EU’s Eastern enlargement.3 We have, in case of missing information in some countries, estimated the 2007 figures, whichyield slightly higher results than the actual figures presented in Deliverable 2.4 Note that due to missing information Portugal is excluded throughout the simulations.IAB 6

Table 1:Migration stock for the NMS-8, 2003-2007 scenarioForeign residents from NMS-8in personsForeign residents from NMS-8in per cent of populationBenchmarkCounterfactualscenarioEnlargementscenarioEnlargementeffectBenchmarkCounterfactualscenarioEnlargementscenario2003 2007 2007 2003 2007 2007EnlargementeffectAT 60255 64596 89940 25344 0.75 0.81 1.12 0.32BE 16151 23242 42918 19676 0.16 0.22 0.41 0.19DE 427958 492123 554372 62249 0.52 0.60 0.68 0.08DK 9807 11220 22146 10926 0.18 0.21 0.41 0.20ES 46710 82863 131118 48255 0.11 0.20 0.31 0.12FI 15825 19154 23957 4803 0.30 0.37 0.46 0.09FR 33858 29690 36971 7281 0.06 0.05 0.06 0.01GR 16413 21582 20257 -1325 0.16 0.20 0.19 -0.01IE 34246 60657 178504 117847 0.86 1.52 4.47 2.95IT 54665 74909 117042 42133 0.10 0.13 0.20 0.07LU 1574 2568 5101 2533 0.36 0.58 1.15 0.57NL 13048 16861 36317 19456 0.08 0.11 0.23 0.12SE 21147 19301 42312 23011 0.24 0.22 0.47 0.26UK 122465 154198 609415 455217 0.21 0.27 1.05 0.78CZ 71019 95954 104442 8488 0.70 0.94 1.03 0.08EE 26070 33922 36735 2813 1.93 2.51 2.72 0.21HU 87680 88285 132582 44297 0.88 0.88 1.33 0.44LT 53557 88922 128361 39439 1.55 2.58 3.73 1.14LV 23863 32559 42547 9987 1.02 1.40 1.83 0.43PL 532942 632111 1297647 665536 1.42 1.68 3.45 1.77SI 35051 40958 35848 -5110 1.76 2.05 1.80 -0.26SK 43938 60252 132207 71955 0.82 1.12 2.45 1.34EU-15 1) 874122 1072964 1910370 837406 0.24 0.29 0.52 0.23NMS-8 874122 1072964 1910370 837406 1.21 1.48 2.64 1.161) Without Portugal.Notes: The stock of foreign residents in 2003 is used as a benchmark. The counterfactual scenario assumes that immigration flowscontinue at their pre-enlargement levels, while the enlargment scenario refers to the actual figures observed in 2007. Therefore thedifference of the enlargement- and the counterfactual scenario is treated as the "enlargement effect".Sources: Own calculations and estimates based on the figures from national population statistics and the European LFS.Immigration from Bulgaria and Romania has already accelerated before enlargement as aconsequence of the immigration policies in Spain and Italy. The foreign population fromBulgaria and Romania in the EU-15 has grown between the end of 2003 and 2007 from694,000 to 1.9 million persons or by 1.2 million persons (see Table 2). We can notattribute this increase to the EU’s Eastern enlargement since the NMS-2 joined the EU-15at January 1 st , 2007. Therefore, we use a zero immigration scenario as a counterfactualto the actual increase from the population from Bulgaria and Romania in our lateranalysis. This measures, however, the impact of relaxed immigration conditions in theEU-15 for these two countries and not the EU Eastern enlargement effect.IAB 7

Table 2:Migration stock for the NMS-2, 2003-2007 scenarioForeign residents from NMS-2in personsForeign residents from NMS-2in per cent of populationBenchmarkEnlargementscenario Difference BenchmarkEnlargementscenarioDifference2003 2007 2003-2007 2003 2007 2003-2007AT 26802 36792 9990 0.34 0.46 0.12BE 6831 23810 16979 0.07 0.23 0.16DE 107850 131402 23552 0.13 0.16 0.03DK 1834 3316 1482 0.03 0.06 0.03ES 277814 828772 550958 0.67 1.98 1.32FI 887 1388 501 0.02 0.03 0.01FR 8840 43652 34812 0.02 0.07 0.06GR 30583 52567 21984 0.29 0.50 0.21IE 17526 24496 6970 0.44 0.61 0.17IT 189279 658755 469476 0.33 1.15 0.82LU 498 1085 587 0.11 0.25 0.13NL 4413 11272 6859 0.03 0.07 0.04SE 3148 6280 3132 0.04 0.07 0.03UK 17979 40023 22044 0.03 0.07 0.04BG 159243 310335 151092 2.04 3.97 1.93RO 535041 1553276 1018234 2.47 7.16 4.70EU-15 1) 694284 1863610 1169326 0.19 0.51 0.32NMS-2 694284 1863610 1169326 2.35 6.32 3.961) Without Portugal.Notes: The stock of foreign residents in 2003 is used as a benchmark. The enlargment scenario refersto the actual figures observed in 2007. The simulation is based on the net migration flows observed forthe period 2003 to 2007.Sources: Own calculations and estimates based on the figures from national population statistics andthe European LFS.The immigration influx varies widely across the EU-15 countries. The net inflow ofresidents from the NMS-8 which has been caused by EU enlargement amounts to 3 percent of the population in Ireland, 0.8 per cent in the UK and 0.6 per cent in Luxembourgcompared to 0.2 per cent at the EU-15 level according to our scenario. The net inflow ofresidents from the NMS-2 in the 2004-2007 period amounts to 1.3 per cent of thepopulation in Spain, 0.8 per cent of the population in Italy and 0.2 per cent of thepopulation in Greece, compared to 0.3 per cent at the EU-15 level.Among the NMS-8, an outflow of about 1.8 per cent of the population in Poland has beencaused by the EU Eastern enlargement according to our scenarios during the 2004 to2007 period, compared to 1.2 per cent for all NMS-8 countries. During the same periodof time, the net outflow amounted 4.7 per cent of the population in Romania and 1.9 percent of the population in Bulgaria.IAB 8

4.1.2 Free movement vs. prolongation of transitional arrangementsIn the next step we analyse the potential impact of removing the remaining immigrationrestrictions which are in place under the transitional arrangements. In case of theNMS-8, the remaining EU-15 countries which have still immigration restrictions in placehave to decide whether to maintain these restrictions or to introduce the free movementin 2009. Particularly relevant is this decision in case of Austria and Germany, since thesetwo countries are still important destinations for migrants from the NMS. In case ofBulgaria and Romania, most EU member states have to decide whether to prolong theimmigration restrictions which are still in place vis-à-vis the NMS-2 in the second phaseof the transitional arrangements beginning with January 1 st , 2009.For the assessment of the macroeconomic effects of transitional periods we employ twopolicy scenarios. Both policy scenarios rely on the migration forecasts carried out inDeliverable 11. The status quo scenario is based on the assumption that the migrationrestrictions which are applied at present will be maintained until the end of thetransitional period. Germany and Austria thus employ the same set of immigrationrestrictions for workers from the NMS-8 until the end of the transitional periods, whilethe UK, Ireland, and Sweden continue to grant workers from the NMS-8 free access totheir labour markets. Analogously, the EU member states maintain their immigrationrestrictions which are currently in place vis-à-vis Bulgaria and Romania. Consequently,we assume that the overall scale of immigration from the NMS-8 and the NMS-2 followsthe status quo scenario outlined in Deliverable 11, and that the regional distribution ofthe inflows of migrants across the EU-15 destination countries remains constant duringthis period. The free movement scenario is again based on the projections carried out inDeliverable 11. Note that the free movement scenario relies on the assumption that theelasticity of migration with respect to the income difference and labour market variablesis similar in the NMS compared to other sending countries in the EU-15. Nevertheless,the free movement scenario expects that immigration from the NMS-8 and the NMS-2will further accelerate if the free movement is introduced compared to its level under thetransitional arrangement. The difference between the free movement scenario and thestatus quo scenario illustrates the migration effect caused by the introduction of freemovement in 2009.Introducing the free movement will affect not only the overall scale of migration in theenlarged EU, but also the regional distribution of migrants across destination countries.Due to missing historical evidence, we can hardly forecast the future distribution ofmigrants from the NMS across the EU-15. Therefore, we have to base our free movementscenario an assumptions here. We assume that the regional migration pattern before2004 reflect the free choice of migrants such that future migration under the freemovement will display a similar regional pattern. As a consequence some countries (e.g.Germany and Austria) receive more migrants while others (e.g. UK and Ireland) attractless. This counterfactual policy scenario is of course based on the heroic assumption ofconstant behaviour of migrants and ignores that network effects etc. established since2004 will certainly affect future migration flows. The reversion in the geographicalstructure of migration flows to the pre-enlargement structure can thus be considered asthe most extreme assumption. The actual regional migration pattern is likely to beIAB 9

etween the present regional distribution and the regional distribution of migration flowsbefore EU enlargement.Table 3 displays the scenarios for the EU-15 and the individual NMS-8 countries betweenthe end of 2007 and 2011. As briefly mentioned above, the introduction of free movementincreases the overall stock of migrants by 86,000 persons. The diversion ofmigration flows is illustrated by the increase of 0.3 and 0.2 per cent of population inAustria and Germany and the decrease by 0.9 and 0.2 per cent of population in Irelandand the UK.Table 3: Migration stock forecasts for the NMS-8 (2007-2011)Foreign residents from NMS-8 in personsForeign residents from NMS-8 in per cent of populationStatus Quo Free movement Free movementStatus Quo Free movementBenchmark scenario scenario effectBenchmark scenario scenario2007 2011 2011 2007 2011 2011Free movementeffectAT 83978 106452 127768 21316 1.03 1.30 1.56 0.26BE 42918 65669 64071 -1598 0.40 0.62 0.60 -0.02DE 554372 661819 847899 186080 0.68 0.81 1.04 0.23DK 22146 32634 33198 564 0.41 0.60 0.61 0.01ES 100832 151287 150856 -431 0.23 0.34 0.34 0.00FI 23957 30869 36395 5526 0.45 0.59 0.69 0.10FR 36971 39617 57038 17421 0.06 0.07 0.09 0.03GR 20257 23525 31055 7530 0.19 0.22 0.29 0.07IE 178504 301117 263438 -37680 4.10 6.91 6.04 -0.86IT 107251 151947 161436 9489 0.18 0.26 0.27 0.02LU 5101 8099 7583 -516 1.10 1.74 1.63 -0.11NL 36317 56095 54160 -1935 0.22 0.35 0.33 -0.01SE 42312 60301 63650 3348 0.46 0.66 0.70 0.04UK 609415 1023305 899896 -123410 1.02 1.71 1.50 -0.21CZ 102198 146687 177213 30526 0.99 1.42 1.72 0.30EE 36444 46480 50816 4336 2.72 3.48 3.80 0.32HU 128345 185227 218068 32841 1.30 1.87 2.20 0.33LT 124885 182420 186470 4049 3.69 5.39 5.51 0.12LV 41996 72768 75726 2957 1.84 3.19 3.32 0.13PL 1270620 1835359 1840739 5380 3.41 4.92 4.94 0.01SI 35701 26389 37326 10936 1.77 1.31 1.85 0.54SK 124142 217405 212084 -5321 2.30 4.03 3.93 -0.10EU-15 1) 1864331 2712735 2798441 85705 0.50 0.72 0.75 0.02NMS-8 1864331 2712735 2798441 85705 2.59 3.77 3.89 0.121) Without Portugal.Notes: The stock of foreign residents in 2007 is used as a benchmark. The status quo scenario refers to migration projections assuming that thetransitional arrangements are prolonged, while the free movement scenario refers to projections which assume that free movement is introducedin the entire EU. Therefore the difference of the status quo and the free movement scenario is treated as the "free movement effect".Sources: Own calculations and estimates.With regard to Bulgaria and Romania, the introduction of free movement increases thestock of migrants in the EU-15 by 104,000 persons between the end of 2007 and 2014(compare Table 4).IAB 10

Table 4: Migration stock forecasts for the NMS-2 (2007-2014)Foreign residents from NMS-2 in personsForeign residents from NMS-2 in per cent of populationBenchmarkStatus QuoscenarioFree movementscenarioFree movementeffectBenchmarkStatus Quoscenario2007 2014 2014 2007 2014 2014Free movement Free movementscenario effectAT 29958 37345 71051 33706 0.37 0.46 0.87 0.41BE 23810 48735 42224 -6511 0.22 0.46 0.40 -0.06DE 131402 165977 310020 144043 0.16 0.20 0.38 0.18DK 3316 5492 6849 1358 0.06 0.10 0.13 0.03ES 649076 1322727 1155400 -167328 1.45 2.96 2.59 -0.37FI 1388 2123 2998 874 0.03 0.04 0.06 0.02FR 43652 94756 73368 -21388 0.07 0.16 0.12 -0.04GR 52567 84840 110236 25396 0.49 0.79 1.03 0.24IE 24496 34728 54964 20236 0.56 0.80 1.26 0.46IT 415893 748562 814316 65754 0.71 1.27 1.38 0.11LU 1085 1946 2129 183 0.23 0.42 0.46 0.04NL 11272 21341 21283 -58 0.07 0.13 0.13 0.00SE 6280 10878 12614 1737 0.07 0.12 0.14 0.02UK 40023 72384 78106 5722 0.07 0.12 0.13 0.01BG 273506 408399 460295 51896 3.56 5.32 6.00 0.68RO 1160713 2243435 2295262 51827 5.39 10.41 10.65 0.24EU-15 1) 1434218 2651834 2755557 103723 0.38 0.71 0.73 0.03NMS-2 1434218 2651834 2755557 103723 4.91 9.07 9.43 0.351) Sources: Without Own Portugal. calculations and estimates based on the figures from national population statistics and the European LFS.Notes: The stock of foreign residents in 2007 is used as a benchmark. The status quo scenario refers to migration projections assuming that thetransitional arrangements are prolonged, while the free movement scenario refers to projections which assume that free movement is introducedin the entire EU. Therefore the difference of the status quo and the free movement scenario is treated as the "free movement effect".Sources: Own calculations and estimates.Throughout our simulations, we have used the actual activity and employment rates ofthe immigrant population derived from the European Labour Force Survey (Eurostat,2008) for the calculation of the labour supply shocks. Moreover, we used the skill andage composition of the immigrant workforce for the analysis of the labour market effectsfrom the same data source. However, since migrants from the NMS are employed inoccupations which do not correspond to their educational attainment, we madeadjustments for the ‘brain waste’ in the receiving countries.4.1.3 Accounting for differences between migrants' jobs and skillsFor an empirically meaningful assessment of the migration impact, we have to makeassumptions on the skill structure of the labour supply shock. As has been outlined inBackground Report 2, the skill level of migrants from the NMS is higher than that ofnatives who stay behind in the sending countries, even if we control for cohort effects(see Background Report 2). We apply here the assumption that there is no selection withrespect to unobservable abilities relative to the native population in the home countries,such that migrants from the NMS would be employed in their home countries similar tonatives with the same skill levels and work experience.In the receiving countries, the occupational structure of employment suggests thatmigrants from the NMS are employed below their educational levels: a large share ofmigrants is employed in occupations which need only elementary skills irrespective oftheir educational attainment. As a consequence, the wage level of migrants from theNMS in the UK is well below that of natives in the receiving countries with similareducation and work experience (see Background Report 6 in this report, and Barret andDuffy, 2008 for evidence from Ireland). Moreover, the returns to education do notIAB 11

increase significantly with the time spend in the receiving countries, although it is tooearly to ultimately assess the labour market assimilation of migrants from the NMS(Background Report 6). Overall, migrants from the NMS compete to a large extent in theless-skilled segments of the labour market with natives and other foreigners in the EU-15, although their educational attainment is relatively high.Using the skill level of migrants from the NMS as reported in the Labour Force Surveywould therefore bias our simulations of the migration impact. In order to avoid this, wehave classified migrants according to their occupational breakdown, which has beenrelated to the skill level of the workforce. As a result, we find much higher shares ofmigrants from the NMS in the group with low education, and much lower shares in thegroup with high education. This revised breakdown provides in our view a much betterapproximation of the skill structure of the labour supply shock from the NMS than theskill breakdown reported by the Labour Force Survey.5 Assessing the Labour Market Effects: A Wage Curve ApproachThe first model we apply here for the assessment of labour mobility on wages,employment and other macroeconomic variables is based on a framework whichconsiders imperfect labour markets and unemployment. In contrast to the overwhelmingshare of the literature which addresses the wage and employment effects of labourmigration separately, we analyse the wage and employment effects of migrationsimultaneously in a general equilibrium framework. We apply an aggregate wage curveapproach for this purpose, which relies on the empirical observation that wages respondto changes in the unemployment rate, albeit imperfectly. This allows us to considerinstitutional and other labour market rigidities, which are particularly relevant in theEuropean context.The empirical framework is based on a nested production function grouping the labourforce by education, experience, and national origin. The elasticities of the wage curveand of the production function are estimated. Moreover, we consider the adjustment ofcapital stocks and estimate the speed of adjustment empirically.The analysis in this section is organised as follows: First, we outline the theoreticalbackground (Section 5.1). Second, we describe the databases which are employed forthe empirical analysis (Section 5.2). Third, we present the estimation strategy and theestimation results for the adjustment of capital stocks, the elasticities of the wage curveand of the production function (Section 5.3). Fourth, we simulate the employment andwage impact of migration on the receiving and sending countries in the enlarged EU(Section 5.4). Finally, Section 5.5 concludes.5.1 Theoretical backgroundThe labour market is modelled here in form of an aggregate wage curve. The wage curveis based on the empirical observation that wages decline when the employment rateIAB 12

increases. This enables us to capture the employment and the wage effects of migrationsimultaneously in a joint framework (Boeri and Brücker, 2005; Brücker and Jahn, 2008;Levine, 1999).The wage curve can be based on different theoretical foundations (see Blanchflower andOswald, 1994; Layard et al., 1991, for a discussion). In our context, two modellingtraditions are particularly important. First, the wage curve can be derived frombargaining models (see e.g. Layard and Nickell, 1986; Lindbeck, 1993), which assumethat trade unions are concerned about both their employed and unemployed members.Consider the case where wages are fixed in a bilateral bargaining monopoly betweentrade unions and employer federations. Once wages are fixed, firms hire workers untilthe marginal product of labour equals the wage rate. Both parties that participate in thewage bargain are aware of this. Higher unemployment means that more union membersare without work and that employed members who are dismissed will have a lowerprobability of finding new employment. Consequently, the negotiated wage is lower whenunemployment is higher and vice versa.Second, in a completely non-unionised environment, a wage curve can be explained byefficiency-wage considerations (Shapiro and Stiglitz, 1984), where the productivity ofworkers is linked to the wage level. Unemployment works here as disciplining devicesince it determines the difficulties in finding a new job. As a result, firms will reduce theremuneration of workers if the unemployment rate is increasing since they can achievethe same level of productivity at a lower wage if unemployment is higher.Both approaches have in common that they replace the conventional labour supply curveby a wage fixing function and that they rely on standard assumptions about labourdemand (Blanchflower and Oswald, 1995; Layard and Nickell, 1986). Bargaining andefficiency wage models may play different roles in different countries depending on theirlabour market institutions. Therefore we do not derive the wage curve from a specificwage bargaining or efficiency wage model here. We simply assume that a wage-fixingmechanism exists which responds to the unemployment rate, albeit imperfectly. Oncewages are fixed, profit-maximising firms hire workers until the marginal product of labourequals the wage rate.The production-side of the economy is modelled in form of a nested production function,which groups the labour force by education, experience, and national origin (see Borjas,2003; Card and Lemieux, 2001; Ottaviano and Peri, 2006, for a similar approach).However, data limitations restrict the number of cells in the labour market. Wedistinguish three education groups, three experience groups, and native and foreignworkers. We assume that the production function is characterised by a constant elasticityof substitution (CES) between the individual factors.The production function determines the marginal product of labour. Since firms are freein their hiring decisions, it follows that profit-maximising firms hire workers until thewage rate equals the marginal product of labour. At the same time, the elasticity of thewage curve determines the relation between wages and the unemployment rate, andhence, both the wage and employment response to an exogenous labour supply shock.IAB 13

This allows deriving the wage and employment response of the economy to theimmigration of labour simultaneously.The details of the model are described in Brücker and Jahn, 2008.5.2 DataAn EU-wide data set which provides detailed information on wages, employment, andlabour supply for larger time-series does not exist. Our empirical approach thereforefollows the strategy to exploit both the existing data sources for the EU-15 and the newmember states and empirical estimates on the elasticities of the production function forcountries where more detailed data sets exist. For the EU-15 we use information fromthe European Community Household Panel (ECHP); for selected NMS, we use wage andlabour force data which have been collected in the framework of the EU-KLEMS project.The ECHP is a household survey which provides individual information on wages, theemployment status, human capital characteristics such as education and workexperience, and national origin. This information is used to estimate the elasticities of theaggregate production functions in the EU-15. Due to missing wage information we had toskip Sweden and Luxembourg from the panel. We use the unweighted average for theparameters in the remaining EU-15 for these two countries.The data set has, however, a number of limitations: First, since it relies on surveyinformation, particularly the measurement of wages is inaccurate. Measurement errorcan result in an attenuation bias, i.e. an underestimation of the inverse of the relevantelasticities. Second, the response rates for the immigrant community are low. This forcesus to base our analysis on relatively broad categories. Still, the information suffers frominsufficient information particularly in the foreigner cells. Third, the time dimension of thepanel is limited. At a maximum we have eight observations over time, for a number ofcountries we have only six observations. However, since the elasticities of the productionfunction are identified by fixed effects regressions, the time dimension is crucial for aproper identification.Compared to the literature, studies in individual countries suffer from data limitations aswell, but less than ours. The time dimension of the data sets in the US studies (e.g.Borjas, 2003; Ottaviano and Peri, 2006) is also limited to eight observations. But therewe have decennial information from the population censuses and not annual information.The variance of the data is therefore higher in the US data bases. The existing Germanstudies (Brücker and Jahn, 2008; D'Amuri et al., 2008; Felbermayr et al., 2008) arebased on administrative data or household surveys with a longer time dimension andaccurate wage information. The British study by Manacorda et al. (2006) is based onlabour force survey data with similar measurement problems as our dataset but it has alarger time dimension and more observations than the ECHP.Altogether, it is likely that our estimates of the elasticities of the production functionsuffer from an attenuation bias which can be traced back to the limitation of the data setemployed. Nevertheless, the ECHP is the only available data source which provides therelevant information for most EU member states, i.e. information on wage levels,IAB 14

employment status, and human capital characteristics of the workforce which we needfor the identification of the elasticities of the production function. We therefore use theelasticities from the literature for a sensitivity analysis in our simulations.The ECHP information is supplemented by information from other data sources. The wagecurves are estimated on basis of aggregate wage and unemployment data provided bythe Eurostat Labour Force Survey. This data series enables us to cover more informationover time than the ECHP data.The adjustment of capital stocks to labour supply shocks is measured on basis ofinternationally comparable capital stock data provided by the OECD. Finally, thesimulations are based on the structure of employment and unemployment by natives andforeigners provided by the Eurostat Labour Force Survey (LFS) data. We use these datafor the simulations since the picture on the employment structure is more accurate in theLFS data compared to the ECHP.There exists no complete data set for the new member states which provides informationon wages and employment status by education and experience. We therefore use for aselection of countries – Czech Republic, Hungary, Poland, and Slovakia – data on wagesand employment provided by Vienna Institute for International Economic Comparisons(wiiw), which have been collected in the context of the EU KLEMS project. This data setprovides information on wages, employment and unemployment by three education andthree age groups. The data set contains no information by nationality. We thus focus inthis country group on the impact of emigration on wages and employment, but do notconsider the impact of immigration into these countries. Note that the foreigner share israther small in these countries, such that the ignoring the different effects of emigrationon native and foreign workers does not bias our results seriously.For those NMS countries which are not covered by the data set, we use the unweightedaverage of the estimated parameters for the country sample described above. Thestructure of wages and employment by education and experience groups is alsoextrapolated from the average structure of the countries on which we have information.However, we use the available information on GDP and average wages for thesecountries for the evaluation of the wage and employment effects. Regarding thestructure of wages across the different cells of the labour market we use again theunweighted average from the NMS countries for which information on the wage structureis available.Time series on capital stocks are not available for the NMS. We therefore assume thatcapital stocks adjust in this country group to labour supply shocks at the average speedwhich we observe in the EU-15.Altogether, data on wages and employment status in the different cells of the labourmarket is available only for a subsample of the EU-15 and NMS countries covered by ouranalysis. Moreover, the survey information used is subject to measurement error, whichmay in turn result in an attenuation bias. Although we may overestimate the elasticity ofsubstitution across the different cells of the labour market, the available time-series mayallow us to identify both the average elasticity of the wage curves properly as well as theIAB 15

adjustment of capital stocks to labour supply shocks. Thus, while our analysis may bedistorted in individual cells of the labour market, this kind of analysis may provide areasonable picture of the overall trends in the economies involved. Moreover, as arobustness check, we use elasticities estimated by other studies for a sensitivity analysis.5.3 Estimation resultsThe simulation of the model requires three sets of parameters: Estimates of theadjustment of capital stocks to labour supply shocks, estimates of the elasticity of thewage curve and estimates of the elasticity of substitution between the factors ofproduction.5.3.1 Adjustment of capital stocksFollowing Ottaviano and Peri (2006), we estimate the adjustment of the capital-labourratio asln κ t = β 0 + β 1 ln κ t-1 + β 2 ln κ t-2 + β 3 TREND t + γ ∆ ln L t + ε t , (1)where κ t is the capital-labour ratio, TREND t a deterministic time trend, L t the labourforce, ε t the error term, and ∆ the difference operator, β and γ coefficients and t the timeindex. The numbers of lags of the dependent variable which are included have beenchosen by significance level of the respective lag.Thus, equation (1) is a dynamic model, where the short-run impact of the labour supplyshocks, ∆lnL t , is captured by the estimate of the parameter γ. Other factors which mayaffect the capital-labour ratio are captured by the deterministic time trend. Theinterpretation of the coefficient γ is straightforward: a coefficient of -1 implies that thecapital-labour ratio declines by 1 per cent if the labour force grows by 1 per cent, whichcorresponds to the case where the capital stock is fixed. The size of the coefficients onthe lagged capital output ratio determines the speed of adjustment to capital-labour ratiobefore the labour supply shock.Note that the unit-root tests indicate that the capital-labour ratio is stationary, while thelabour force is a non-stationary I(1) variable. This can be interpreted as support for thetheoretical assumption that labour supply shocks have a short-run but not a long-runimpact on the capital-labour ratio.Since the labour force might be endogenous, we have estimated equation (1) both withOLS and Two-Stage-Least Squares. In the later regression we have used the first andsecond lag of the change of the labour force as instruments.IAB 16

Table 5:Adjustment of the capital-labour ratio in EU countriesOLS-RegressionsIV-Regressions∆ ln L adj. R 2 adj. R 2t ln k t-1 ln k t-2 ∆ ln L t ln k t-1 ln k t-2coeff. t-stat. coeff. t-stat. coeff. t-stat. coeff. t-stat. coeff. t-stat. coeff. t-stat.AT -0.13 -0.37 0.87 *** 29.34 - - 0.998 -0.17 -0.49 0.82 *** 20.91 - - 0.998BE -0.58 *** -3.11 0.99 *** 20.78 - - 0.999 -0.57 *** -2.93 0.99 *** 18.86 - - 0.999DK -0.64 *** -2.83 0.96 *** 13.76 - - 0.998 -0.69 *** -2.95 0.97 *** 13.74 - - 0.998FIN -0.90 *** -11.73 0.96 *** 39.40 - - 0.998 -0.90 *** -11.73 0.96 *** 39.40 - - 0.997FR -0.40 *** -3.60 0.95 *** 73.47 - - 1.000 -0.41 *** -3.83 0.93 *** 63.76 - - 0.998DE -0.80 *** -10.72 0.93 *** 33.04 - - 0.999 -0.83 *** -10.62 0.96 *** 26.58 - - 0.999GR -0.80 *** -3.96 1.45 *** 9.67 -0.53 *** -4.50 0.998 -0.81 *** -3.92 1.46 *** 9.51 -0.53 *** -4.41 0.997IE -0.84 *** -6.37 0.90 *** 7.39 -0.21 * -1.83 0.997 -0.84 *** -6.31 0.91 *** 7.36 -0.21 * -1.87 0.997IT -0.72 *** -7.15 0.89 *** 35.02 - - 1.000 -0.69 *** -5.85 0.89 *** 23.17 - - 0.992NL -0.61 ** -2.71 0.78 *** 10.94 - - 0.960 -0.64 ** -2.69 0.74 *** 7.96 - - 0.942PT -0.86 ** -2.75 0.86 *** 10.14 - - 0.991 -0.86 ** -2.75 0.86 *** 10.14 - - 0.987SP -0.72 *** -7.38 0.91 *** 39.24 - - 0.999 -0.71 *** -7.53 0.88 *** 35.64 - - 0.999SWE -0.49 *** -4.38 1.29 *** 10.22 -0.33 ** -2.58 0.999 -0.49 *** -4.27 1.28 *** 10.07 -0.33 ** -2.54 0.999UK -0.80 *** -8.80 0.86 *** 16.26 - - 1.000 -0.80 *** -8.45 0.86 *** 14.84 - - 0.995EU-14 -0.66 *** -15.47 0.95 *** 123.1 - - 0.999 -0.65 *** -14.77 0.95 *** 129.85 - - 0.997Dependent variable is ln κ t .-- ***,**,* denote the significance at the 1-, 5-, and 10-per cent level, respectively.-- Thecountry regressions cover 38 observations (1 lag) or 37 observations (2 lags).-- Each regressions includes a constantand a deterministic time trend.-- The IV-regressions use the first and the second lag of the log change of the labourforce as instruments.-- The panel regressions is estimated with GLS allowing for heteroscedastic disturbances.The regression results are displayed in Table 5. We find that the estimated coefficient γvaries in most countries between -0.6 and -0.9, indicating that capital stocks adjustalready in the first period. We have two outliers – Austria and France – where theestimated coefficients for the parameter γ are very small and suggest that the capitalstocks adjust already in the first period largely to labour supply shocks. The coefficientsfor the lagged dependent variable (or the sum of the lagged dependent variable) vary inmost regressions between 0.7 and 0.95, indicating that between 5 and 30 per cent of aninitial shock on the capital-labour ratio disappears within one year.Altogether we find strong evidence that capital stocks adjust to labour supply shocks andthat these adjustment processes are rather fast in most countries, although the resultsdiffer for the individual countries.In our simulations we apply the estimated coefficients for the individual EU-15 countries.For the NMS, where long time series for capital stocks do not exist, we apply the panelestimate of the coefficients for the EU-15 as parameters in our simulations. This yields anestimate of -0.65 for the parameter γ, and one of 0.95 for the lagged capital-labourratio.5.3.2 Estimates of the wage curveThe wage curve is usually estimated either at the regional or at the sectoral level(Blanchflower and Oswald, 1994; 2005). However, there also exist a number ofestimates at the national level (see Card and David, 1995, for a detailed discussion andGuichard and Laffargue, 2000, for a recent contribution). Since we want to identify themacro impact of the adjustment of wages to the unemployment rate we follow here thenational level approach (see Brücker and Jahn, 2008, for a detailed discussion).IAB 17

More specifically, we estimateln w t = β 0 + β 1 ln w t-1 + β 2 ln w t-2 + β 3 TREND t + η ln u t + ε t , (2)where w t is the wage rate, TREND t a deterministic time trend, u t the unemployment rate,ε t the error term, β and η coefficients and t as before the time index. The numbers oflags of the dependent variable which are included have been chosen by significance levelof the respective lag. Following the literature (Blanchflower and Oswald, 1994; 2005) weestimate equation (2) with two-stage-least-squares using the first and the second lag ofthe unemployment rate as instruments.Our findings are displayed in Table 6. Our estimates vary country by country and are notcompletely robust with regard to the lag specification of the model. We therefore decidedto use the more robust panel estimate for the EU-15 for our simulations. This yields along-run elasticity of -0.13 for the aggregate wage curve. This is slightly higher than theelasticity of -0.1 which is found in large parts of the regional level literature(Blanchflower and Oswald, 1995; 2005; Longhi et al., 2005). However, this is notsurprising in our view, since we estimate here the macro response of wages to changesin the unemployment rate rather than the regional wage response to changes in theregional unemployment rate. In case of centralised wage bargaining it is however ratherlikely that the macro response exceeds the regional response.For the new member states we have only short time series between 10 and 15 years,making it difficult to identify the wage curve. We find a rather large elasticity of -0.26 forthe NMS. However, this may be influenced by the transitional recession and not robustdue to structural breaks. We therefore employed the wage curve which we found in thepanel estimate for the EU-15 in our simulations also for the NMS.IAB 18

Table 6:Estimate of the dynamic wage curve modeldependent variable:ln (w it )ln (w i,t-1 ) ln (w i,t-2 )ln (u i,t-1 )regressiondiagnosticsshort-runlong-runcountry coeff. se coeff. se coeff. se coeff. adj. R 2 obs.Austria 0.80 *** 0.07 - -0.013 0.01 -0.063 0.998 36Belgium 0.83 *** 0.12 -0.49 *** 0.13 -0.017 0.01 -0.025 0.999 35Denmark 0.51 ** 0.19 0.07 0.18 -0.002 0.07 -0.004 0.995 36Finland 0.76 *** 0.11 - -0.022 ** 0.01 -0.092 0.993 36France 1.75 *** 0.09 -0.80 0.10 -0.014 * 0.04 -0.320 0.997 35Germany 0.19 0.13 - -0.067 *** 0.01 -0.083 0.997 15Greece 0.83 *** 0.08 -0.033 0.02 -0.197 0.927 36Ireland 0.77 *** 0.13 -0.008 0.02 -0.034 0.999 34Italy 0.54 *** 0.16 - -0.052 * 0.03 -0.113 0.999 36Luxembourg 0.35 *** 0.17 - -0.064 ** 0.02 -0.099 0.958 36Netherlands 1.56 *** 0.11 -0.62 *** 0.11 -0.015 0.02 -0.221 0.999 35Portugal 0.51 *** 0.16 - -0.063 *** 0.02 -0.129 0.999 11Spain 0.87 *** 0.07 -0.033 ** 0.01 -0.243 0.991 36Sweden 0.79 *** 0.10 - -0.050 *** 0.01 -0.238 0.998 36United Kingdom 0.67 *** 0.10 - -0.028 ** 0.01 -0.086 0.994 36EU-15 0.93 *** 0.01 -0.009 *** 0.00 -0.130 0.988 498NMS-10 0.69 *** 0.06 -0.083 *** 0.02 -0.269 1.00 99***, **, * denote the significance at the 1 per cent, 5 per cent and 10 per cent level, respectively.-- In each regressio theunemployment rate ist instrumented with the first and the second lag of the unemployment rate.-- All regressionsinclude a deterministic time trend and a squared deterministic time trend.-- We report White-heteoscedastic consistentstandard errors in the fixed effects regressions.-- The F-test rejects the Null hypothesis of no country specfic fixedeffects.5.3.3 Estimates of the elasticities of substitutionThe simulation of the model presented above requires the estimation of the elasticities ofsubstitution between labour of different education groups, of different experience groupsand natives and foreigners. We have estimated these elasticities step by step on basis ofthe ECHP data. In case of the NMS as emigration countries with a rather small foreignershare we have not estimated the elasticity of substitution between natives andforeigners.The results are displayed in Table 7. In most countries the elasticities of substitutionhave the expected signs. It is worthwhile noting that our findings confirm the suggestionby Ottaviano and Peri (2006) that natives and foreigners are imperfect substitutes in thelabour market. However, the coefficients which we have estimated are quite small,indicating a high elasticity of substitution between natives and foreigners.IAB 19

Table 7:Estimates of the inverse elasticity of substitutionBetween Between Betweeneducation experience natives andgroups groups immigrantsEU-15 AT 0.04 0.02 0.08BE 0.08 0.02 0.07DE 0.00 0.05 0.11DK 0.02 -0.11 0.02ES 0.50 0.00 0.00FI 0.05 0.03 0.07FR 0.02 0.16 0.00GR 0.24 0.04 0.12IE 0.21 0.04 0.01IT 0.03 0.07 0.08LU 0.12 0.04 0.04NL 0.12 0.04 0.04SE 0.12 0.04 0.04UK 0.23 0.05 0.06NMS-8 CZ 0.04 0.13 --EE 0.05 0.06 --HU 0.08 0.10 --LT 0.05 0.06 --LV 0.05 0.06 --PL 0.06 0.01 --SI 0.05 0.06 --SK 0.03 0.02 --NMS-2 BG 0.05 0.06 --RO 0.05 0.06 --Sources: Own estimates based on ECHP data and the EUKLEMS data.We have therefore compared our findings with those of the literature (see Table 8). Theinverse elasticities found in other studies are on average larger than those obtained byus, particularly those in the US studies. However, many results in the European studieslook relatively similar to our findings. As a robustness check, we have employed thelargest elasticities found in the literature in a sensitivity analysis. Employing theseelasticities does not change our findings qualitatively and quantitatively to a large extentwith one exception: The size of the elasticity of substitution between native and foreignlabour can change results in an important way. Beyond this caveat, our simulations arerather robust. The sensitivity analysis is available from the authors upon request.IAB 20

Table 8:Estimates of the inverse elasticity of substitution: a literature reviewBetween Between Betweeneducation groups age or experience groups natives and immigrantsCountry Min Max Min Max Min MaxAydemir and Borjas (2007) CA 0.05 0.42 0.03 0.14 -0.02 -Card and Lemieux (2006) CA 0.13 0.28 0.16 0.17 - -Bruecker and Jahn (2008) DE 0.15 0.31 0.03 0.06 0.01 0.13D'Amuri, Ottaviano and Peri (2008) DE - - - - 0.05 0.06Felbermayr, Geis and Kohler (2008) DE 0.22 0.24 -0.02 0.06 0.06 0.12Fitzenberger, Garloff and Kohn (2004) DE 0.11 0.09 0.09 0.03 - -Amuedo-Dorantes and de la Rica (2008) ES 0.65 0.69 0.22 0.34 - -Aydemir and Borjas (2007) MEX 0.36 2.02 0.23 0.29 - -Card and Lemieux (2001) UK 0.34 0.42 0.23 0.26 - -Manacorda, Manning and Wadsworth (2006) UK 0.09 0.17 0.10 0.11 0.15 0.36Aydemir and Borjas (2007) US 0.27 0.33 0.12 0.32 -0.01 -Borjas (2003) US 0.29 - 0.74 0.76 - -Card and Lemieux (2006) US 0.33 0.48 0.20 0.27 - -Ottaviano and Perri (2006) US 0.38 0.54 0.16 0.30 0.07 0.27Bruecker and Jahn (2008, unpublished) UK 0.23 - 0.05 0.06 0.04 0.09Sources: Own presentation based on the studies quoted above.5.4 Simulation resultsWe simulate first the impact of the EU Eastern enlargement on migration between theNMS-8 and the EU-15 during the 2004 to 2007 period, and then the impact of migrationbetween the NMS-2 and the EU-15 during the same period of time. In each scenario wedistinguish between the short-run and the long-run effects of migration. In the short-runscenario we assume that the capital-labour ratio adjusts as estimated by equation (1), inthe long-run scenario we assume that the capital stock adjusts completely to theincreasing labour supply.In all scenarios we have calculated the following effects:• First, the impact of migration on aggregate GDP, on GDP per capita and the totalfactor income per native. The first variable captures the overall effect on outputand the second one the output effect per capita. Both indicators should not bemisunderstood as welfare indicators. They do in particular not capture whethernatives in the receiving countries lose or gain. The third indicator comprises thetotal factor income of the native population based on the assumptions thatmigrants do not bring capital and that natives own the entire capital stock of theeconomy. Under these strong assumptions, this is an indication for the change intotal earnings of the native population.• Second, we have calculated the aggregate effects on the labour market. Thiscovers the wage rate and the aggregate unemployment rate.• Third, we have analysed the wage and unemployment effects in detail fordifferent groups in the labour market, distinguishing between high-, medium- andlow- skilled workers.IAB 21

5.4.1 The impact of Eastern enlargement on the UK and Germany, 2004-2007Based on the detailed estimation of the parameters, including the elasticities of thewage-setting curves for different education and experience groups in the labour market,we have first simulated the impact of Eastern enlargement on the UK and Germany.According to our scenarios, Eastern enlargement involves an increase in the labour forcethrough immigration from the NMS-8 of about 1.3 per cent in the UK, but only of 0.1 percent in Germany. The immigration from the NMS-2 is negligible in both countries.Our simulation results indicate that the immigration from the NMS will decrease the GDPper capita in the UK by about 0.34 per cent in the short-run while the long-run effect isalmost neutral. The short-run decrease can be attributed to the fact that migrants do notbring capital. However, the factor income of the native population, i.e. the income ofnative labour and capital, will increase by 0.31 per cent in the long-run and only slightlydecline by 0.06 per cent in the short-term. Wages, however, decline in the short-run byabout 0.29 per cent and unemployment increases by about 0.26 percentage points in theshort-run. In the long-run, when capital stocks have adjusted, the wage impact is zerowhile the unemployment rate is slightly increasing by 0.18 percentage points. The resultsfor Germany display a similar picture, but are much smaller due to the lowerimmigration.We find that the effects are very balanced across the different groups of the labour forcein the UK and Germany, with the notable exception of workers with no vocationaltraining. In the UK, these workers are much more affected by declining wages in theshort-term (-0.67) compared to workers with vocational training (-0.23), a high school(-0.27) or a university degree (-0.26). In the long-run, these effects diminish (Table 9).Similarly, the unemployment rate of workers with no vocational training tends toincrease more than that of other workers. In the long-run, the unemployment rateremains by and large unchanged for all groups in the labour market in the UK, except forworkers with no vocational training. It is also important to note that the native workforcetends to win from migration slightly in the long-run both in terms of higher wages andlower unemployment risks, while the foreign workforce loses substantially (Table 9).It is worthwhile to note that the ceteris paribus condition applies for these results, i.e.that other currents may affect wages and the unemployment rate in one direction oranother. In fact, unemployment has increased in the UK slightly by about 0.5 percentagepoints from 2004 to 2007 which is in the range of normal fluctuations which we observesince the beginning of this decade and before the financial crisis began. We thusconclude that our findings are by and large consistent with actual developments.However, the unemployment rate of the foreign workforce has increased by less than 0.5percentage points during the simulation period, i.e. by much less than our simulationresults suggest. Again, the findings presented here do not predict the actualdevelopment of the unemployment rate or wage growth for certain groups in the labourmarket, but the potential impact of migration under the assumption that anything else isequal and that the values of the parameters of our structural model remain constant.IAB 22

Table 9: The impact of Eastern enlargement on the UK and Germany, 2004-2007NMS-8NMS2Germany United Kingdom Germany United KingdomShort-run Long-run Short-run Long-run Short-run Long-run Short-run Long-runChanges in per cent (unemployment rate: changes in percentage points)Macro figuresChange of labour force 0.10 0.10 1.28 1.28 0.04 0.04 0.07 0.07GDP 0.01 0.07 0.44 0.81 0.00 0.03 0.03 0.05GDP per capita -0.07 -0.01 -0.34 0.03 -0.02 0.00 -0.01 0.01Factor income per native -0.03 0.03 -0.06 0.31 -0.01 0.01 0.00 0.02Unemployment 0.04 0.02 0.26 0.18 0.02 0.01 0.01 0.01Wages -0.03 0.00 -0.29 0.00 -0.01 0.00 -0.01 0.00Wages by educationAll -0.03 0.00 -0.29 0.00 -0.01 0.00 -0.01 0.00No vocational -0.07 -0.04 -0.67 -0.38 -0.03 -0.02 -0.04 -0.02Vocational -0.03 0.00 -0.23 0.06 -0.01 0.00 -0.01 0.00High school -0.03 0.00 -0.27 0.02 -0.01 0.00 -0.01 0.00University -0.03 0.00 -0.26 0.05 -0.01 0.00 -0.01 0.00Native wages by educationAll natives -0.02 0.00 -0.24 0.05 -0.01 0.00 -0.01 0.00No vocational -0.04 -0.01 -0.52 -0.23 -0.01 0.00 -0.03 -0.01Vocational -0.02 0.01 -0.20 0.09 -0.01 0.00 -0.01 0.00High school -0.03 0.00 -0.21 0.08 -0.01 0.00 -0.01 0.00University -0.02 0.01 -0.20 0.10 -0.01 0.00 -0.01 0.01Non-native wages by educationAll non-natives -0.07 -0.04 -0.89 -0.60 -0.03 -0.02 -0.05 -0.03No vocational -0.12 -0.09 -4.45 -4.17 -0.04 -0.03 -0.25 -0.23Vocational -0.04 -0.02 -0.85 -0.56 -0.02 -0.01 -0.05 -0.03High school -0.07 -0.04 -0.75 -0.47 -0.02 -0.01 -0.04 -0.03University -0.07 -0.03 -0.62 -0.31 -0.03 -0.02 -0.03 -0.02Unemployment by educationAll 0.04 0.02 0.26 0.18 0.02 0.01 0.01 0.01No vocational 0.10 0.06 1.02 0.92 0.04 0.02 0.06 0.05Vocational 0.03 0.00 0.15 0.06 0.01 0.00 0.01 0.00High school 0.03 0.01 0.14 0.04 0.01 0.00 0.01 0.00University 0.01 0.00 0.04 0.02 0.01 0.00 0.00 0.00Native unemployment by educationAll natives 0.02 0.00 0.07 -0.01 0.01 0.00 0.00 0.00No vocational 0.05 0.02 0.18 0.08 0.02 0.01 0.01 0.00Vocational 0.02 -0.01 0.06 -0.02 0.01 0.00 0.00 0.00High school 0.01 0.00 0.07 -0.02 0.00 0.00 0.00 0.00University 0.01 0.00 0.01 -0.01 0.00 0.00 0.00 0.00Non-native unemployment by educationAll non-natives 0.11 0.07 1.69 1.59 0.04 0.03 0.11 0.10No vocational 0.17 0.13 5.58 5.47 0.07 0.05 0.63 0.62Vocational 0.07 0.03 1.30 1.18 0.03 0.01 0.09 0.08High school 0.09 0.06 0.40 0.22 0.03 0.02 0.02 0.01University 0.05 0.03 0.06 0.02 0.02 0.01 0.00 0.00Source: Own estimates and simulation, see text.5.4.2 The impact of Eastern enlargement on the EU-25, 2004-2007Table 10 presents the impact of migration from the NMS-8 to the EU-15 caused byEastern enlargement on GDP during the 2004-2007 period. We find that immigrationfrom the NMS-8 increases the GDP of the enlarged EU in the short-run by about 0.11 perIAB 23

cent and in the long-run, after the adjustment of capital stocks, by about 0.20 per cent.While the GDP in the EU-15 increases by about 0.26 per cent it falls in the NMS-8 byabout 1.10 per cent in the long-run. This is not surprising since the first group ofcountries receives additional labour and, after the adjustment of capital stocks, additionalcapital. The reverse holds for the sending countries.The impact of migration on the GDP per capita is largely influenced by two factors: First,since immigrants do not bring physical capital by assumption, the capital endowment percapita falls in the receiving and increases in the sending countries in the short-term. Inthe long-term, when capital stocks adjust to changes in the labour supply, this effectdisappears. Second, the rate of participation in the labour market is higher among themigrant population compared to the population average in the receiving countries. As aconsequence, the GDP per capita tends to rise in the receiving countries. Our simulationsdemonstrate that the GDP per capita tends to increase in the sending countries in theshort-term, while it remains largely constant in the receiving countries.Table 10: The macroeconomic impact of migration from the NMS-8, 2004-2007Change oflabourGDPGDP per capitaFactor incomeper native Unemployment WagesShort-run Long-run Short-run Long-run Short-run Long-run Short-run Long-run Short-run Long-runChanges in per cent (unemployment rate: changes in percentage points)AT 0.42 0.31 0.34 0.00 0.02 0.12 0.15 0.02 0.02 -0.02 0.00BE 0.22 0.11 0.17 -0.08 -0.02 0.01 0.07 0.07 0.05 -0.04 0.00DE 0.10 0.04 0.10 -0.03 0.02 -0.01 0.04 0.03 0.01 -0.03 0.00DK 0.23 0.13 0.20 -0.08 -0.01 0.00 0.07 0.02 0.00 -0.05 0.00ES 0.19 0.03 0.11 -0.08 -0.01 -0.04 0.04 0.05 0.02 -0.04 0.00FI 0.09 0.03 0.08 -0.06 -0.01 -0.02 0.04 0.03 0.01 -0.03 0.00FR 0.01 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00GR -0.01 0.00 -0.01 0.01 0.00 0.00 -0.01 0.00 0.00 0.00 0.00IE 4.87 0.80 2.93 -2.07 -0.02 -0.77 1.31 0.87 0.37 -1.61 0.00IT 0.11 0.04 0.08 -0.03 0.01 0.00 0.04 0.02 0.01 -0.03 0.00LU 1.00 0.81 1.13 0.23 0.55 0.34 0.65 0.12 0.05 -0.25 0.00NL 0.14 0.09 0.12 -0.03 -0.01 0.02 0.04 0.02 0.01 -0.02 0.00SE 0.38 0.25 0.33 -0.01 0.07 0.05 0.12 0.05 0.03 -0.06 0.00UK 1.28 0.50 0.89 -0.28 0.10 -0.05 0.34 0.21 0.11 -0.29 0.00CZ -0.08 -0.07 -0.11 0.01 -0.03 0.01 -0.03 -0.02 0.00 0.03 0.00EE -0.21 -0.09 -0.19 0.12 0.02 0.12 0.02 -0.04 0.00 0.06 0.00HU -0.44 -0.34 -0.49 0.10 -0.04 0.10 -0.04 -0.04 0.00 0.11 0.00LT -1.14 -0.55 -1.15 0.61 -0.01 0.61 -0.01 -0.32 -0.01 0.31 0.00LV -0.43 -0.26 -0.46 0.17 -0.03 0.17 -0.03 -0.09 0.00 0.12 0.00PL -1.77 -0.88 -1.94 0.90 -0.18 0.90 -0.18 -0.59 0.03 0.43 0.00SI 0.26 0.15 0.21 -0.10 -0.05 -0.10 -0.05 0.02 0.00 -0.04 0.00SK -1.34 -0.53 -1.51 0.82 -0.18 0.82 -0.18 -0.55 0.00 0.43 0.00EU-15 1) 0.36 0.13 0.26 -0.09 0.03 -0.02 0.10 0.06 0.02 -0.09 0.00NMS-8 -1.16 -0.52 -1.10 0.65 0.05 0.65 0.05 -0.42 -0.02 0.25 0.00Total 0.11 0.11 0.20 0.11 0.20 0.16 0.25 -0.03 0.00 -0.07 0.001) Without Portugal.Source: Own estimates and simulation, see text.More importantly, the total gross factor income of natives in the receiving countries isincreasing in the long-run. Several factors contribute to this fact. First, natives in thesending countries tend to benefit from migration if they differ in their factor endowments(human capital, physical capital) from the migrant population. However, if theunemployment rate is increasing, the effects on the aggregate income of natives areambiguous. When capital adjusts in the longer term, adverse shocks on employment areIAB 24

mitigated and total factor income increases with a larger capital stock. The converseholds for the sending countries.It is important to note in this context that our calculation of the gross factor income pernative is based on the assumption that the capital stock of the economy is owned by thenative population. This is a strong assumption since we may have an inflow of foreigncapital and savings by the migrant population. In the first case some of the additionalincome may flow abroad and in the second case to the migrant population. Nevertheless,since it is likely that most of the investment is undertaken by natives, this approximationdoes not distort the picture largely.Under the assumptions of our simulations, the total factor income of the nativepopulation increases by 1.3 per cent in Ireland and by 0.3 per cent in the UK in the longrun.In the short-run, the factor income of the native population declines slightly in theUK and, reflecting the labour supply shock of 5 per cent, by 0.8 per cent in Ireland. Withthe exception of Luxembourg, the impact on the other receiving countries is negligible.Depending on the scale of the emigration shock in the NMS-8, the total factor income ofthe native population declines in the long-run when capital stocks have adjusted.In the short-run, the unemployment in the receiving countries increases by 0.06percentage points, while it remains stable after the adjustment of capital stocks. In thecountries mainly affected, our simulations suggest that the unemployment rate mayincrease by 0.2 percentage points in the UK and 0.9 percentage points in Ireland in theshort-run. In the long-run, the unemployment rate increases by 0.1 percentage points inthe UK and 0.4 percentage points in Ireland.In contrast to these results, we do not find any visible increase in the unemployment ratein Ireland in the course of the EU’s Eastern enlargement despite the substantial influx ofmigrants there. This may be traced back to a faster adjustment of the capital stock thanassumed by our model or by other adjustment mechanisms not considered by our modelsuch as international trade.We find that the unemployment rate is declining in the sending countries as aconsequence of the outflow of labour. The same holds true for the entire EU sincemigrants tend to move out of countries or regions with an unemployment rate at orabove the average level of the enlarged EU and move to countries having unemploymentrates below the EU average.In our model, migration affects aggregate wages only in the short-run, since aggregatefactor proportions remain unchanged in the long-run due to the adjustment of capitalstocks. At the average of the EU-15, wages decline slightly by 0.1 per cent, but increasein the sending countries by 0.3 per cent in the short-run. Again, Ireland is at a wagedecrease of 1.6 per cent the most affected country, while the wage decreases are at 0.3per cent in the UK and Luxembourg and only limited in the other affected countries. Incontrast, depending on the outflow, wages increase by 0.4 per cent in Poland andSlovakia in the short-run, such that migration contributed slightly to the wageconvergence there. Nevertheless, the wage impact is rather moderate and cannot be feltin most receiving and sending countries.IAB 25

Migration affects the different groups in the labour market in different ways. We havetherefore analysed how the different groups are affected in terms of their wages andunemployment risks. Table 11 displays the wage effects by skill group. We find that lowandmedium skilled workers are slightly more affected by declining wages in the EU-15 (-0.10 and -0.09 per cent) compared to high-skilled workers (-0.07 per cent) in the shortrun.In the long-run, we find that migration from the NMS-8 reduces wages of the lowandmedium-skilled by only 0.01 per cent, and increases wages of high-skilled by 0.02per cent. This pattern reflects the high concentration of migrant workers from the NMS atthe low and medium skill spectrum and that migrants from the NMS are employed wellbelow their reported skill levels.Table 11: The impact of migration from the NMS-8 on wages, 2004-2007AllLow-skilledMedium-skilledHigh-skilledShort-run Long-run Short-run Long-run Short-run Long-run Short-run Long-runChanges in per centAT -0.02 0.00 -0.02 0.00 -0.02 0.00 -0.02 0.00BE -0.04 0.00 -0.03 0.01 -0.03 0.00 -0.05 -0.01DE -0.03 0.00 -0.03 0.00 -0.03 0.00 -0.03 0.00DK -0.05 0.00 -0.05 0.00 -0.05 0.00 -0.05 0.00ES -0.04 0.00 -0.03 0.01 -0.14 -0.09 -0.01 0.04FI -0.03 0.00 -0.03 0.00 -0.03 0.00 -0.03 0.00FR 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00GR 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00IE -1.61 0.00 -1.72 -0.19 -1.84 -0.23 -1.34 0.30IT -0.03 0.00 -0.03 0.00 -0.03 0.00 -0.03 0.00LU -0.25 0.00 -0.13 0.12 -0.14 0.11 -0.63 -0.38NL -0.02 0.00 -0.02 0.00 -0.02 0.00 -0.03 0.00SE -0.06 0.00 -0.05 0.01 -0.05 0.01 -0.08 -0.02UK -0.29 0.00 -0.35 -0.07 -0.35 -0.06 -0.19 0.11CZ 0.03 0.00 0.03 0.00 0.02 0.00 0.03 0.01EE 0.06 0.00 0.07 0.01 0.06 0.00 0.06 0.00HU 0.11 0.00 0.09 -0.01 0.10 -0.01 0.12 0.01LT 0.31 0.00 0.32 0.02 0.30 -0.01 0.33 0.01LV 0.12 0.00 0.11 0.00 0.11 -0.01 0.13 0.01PL 0.43 0.00 0.41 0.01 0.39 -0.03 0.51 0.06SI -0.04 0.00 -0.06 -0.02 -0.04 0.00 -0.03 0.01SK 0.43 0.00 0.36 -0.02 0.41 -0.02 0.49 0.05EU-15 1) -0.09 0.00 -0.10 -0.01 -0.09 -0.01 -0.07 0.02NMS-8 0.25 0.00 0.23 0.00 0.23 -0.02 0.30 0.03Total -0.07 0.00 -0.09 -0.01 -0.08 -0.01 -0.06 0.031) Without Portugal.Source: Own estimates and simulation, see text.In the NMS-8, high-skilled natives benefit more from emigration (+0.30 per cent) thanless- and medium-skilled workers (+0.23 per cent each) in the short-run. In the longrun,wages of the high-skilled increase by 0.03 per cent, while the wages of the mediumskilleddecline by 0.02 per cent. This can be traced back to the fact that the laboursupply in the medium range of the skill spectrum is substantially larger in the NMS-8compared to the EU-15, such that the composition of the migrant workforce changesIAB 26

labour endowments in the receiving and the sending countries in different ways (Table11).Finally, Table 12 displays the effects of migration from the NMS-8 on the unemploymentrisks of different groups in the labour market. Immigration from the NMS-8 increases theunemployment rate of less-skilled workers in the EU-15 by 0.07 percentage points, ofmedium-skilled workers by 0.06 percentage points, and of high-skilled workers by 0.02percentage points. In the long-run, the impact of immigration on employment is largelyneutral. A measurable impact is only found in Ireland. Note that it is rather likely that alarger part of the increasing unemployment risk is absorbed by the migrant populationand not by natives.Table 12: The impact of migration from the NMS-8 on unemployment, 2004-2007AllLow-skilledMedium-skilledHigh-skilledShort-run Long-run Short-run Long-run Short-run Long-run Short-run Long-runChanges in percentage pointsAT 0.02 0.02 0.03 0.02 0.01 0.00 0.09 0.09BE 0.07 0.05 0.09 0.06 0.08 0.06 0.03 0.02DE 0.03 0.01 0.04 0.01 0.02 0.00 0.03 0.02DK 0.02 0.00 0.03 0.00 0.01 0.00 0.03 0.01ES 0.05 0.02 0.04 0.00 0.16 0.12 0.01 -0.02FI 0.03 0.01 0.04 0.00 0.04 0.01 0.02 0.01FR 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00GR 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00IE 0.87 0.37 1.32 0.57 0.86 0.43 0.31 -0.01IT 0.02 0.01 0.02 0.01 0.02 0.01 0.01 0.00LU 0.12 0.05 0.04 -0.04 0.04 -0.02 0.47 0.40NL 0.02 0.01 0.03 0.02 0.01 0.01 0.01 0.00SE 0.05 0.03 0.08 0.05 0.04 0.02 0.07 0.05UK 0.21 0.11 0.29 0.14 0.25 0.16 0.04 -0.02CZ -0.02 0.00 -0.11 -0.07 -0.01 0.00 0.00 0.00EE -0.04 0.00 -0.08 -0.01 -0.05 0.00 -0.02 0.00HU -0.04 0.00 -0.10 0.00 -0.04 0.00 -0.01 0.00LT -0.32 -0.01 -0.61 -0.11 -0.33 0.01 -0.15 0.00LV -0.09 0.00 -0.14 -0.02 -0.09 0.00 -0.05 0.00PL -0.59 0.03 -1.12 -0.23 -0.61 0.06 -0.26 0.00SI 0.02 0.00 0.05 0.02 0.02 0.00 0.01 0.00SK -0.55 0.00 -1.55 -0.21 -0.52 0.00 -0.28 -0.12EU-15 1) 0.06 0.02 0.07 0.03 0.06 0.03 0.02 0.00NMS-8 -0.42 -0.02 -0.81 -0.21 -0.41 0.00 -0.19 -0.03Total -0.03 0.00 -0.01 -0.01 -0.07 0.01 0.00 -0.011) Without Portugal.Source: Own estimates and simulation, see text.In the NMS-8, the unemployment rate is declining in the short-term for the less-skilled (-0.81 percentage points), compared to -0.41 percentage points for the medium skilledand -0.19 percentage points for the high-skilled. In the long-run, the unemployment-riskis declining by -0.21 percentage points for the less-skilled, while the effects for themedium- and high-skilled are rather negligible (Table 12).IAB 27

5.4.3 The impact of migration from Bulgaria and Romania, 2004-2007While we have analysed in the previous section the impact of migration flows which havebeen caused by the EU’s Eastern enlargement during the period 2004 to 2007, weanalyse here the impact of migration from the NMS-2 during the same period comparedto a zero migration scenario. We cannot contrast the Eastern enlargement migrationflows with a no EU enlargement counterfactual here, since the NMS-2 joined the EU notbefore 2007.Table 13 displays the aggregate effects on GDP and factor income. The immigration fromthe NMS-2 of about 0.50 per cent of the labour force of the EU-15 increases the GDP ofthe EU-15 by 0.13 per cent in the short-run and 0.30 per cent in the long-run, while itreduces it in the NMS-2 by 2.91 per cent in the short-run and by 4.07 per cent in thelong-run. The GDP per capita in the EU-15 falls by 0.19 per cent in the short-run and by0.02 per cent in the long-run. The decrease in the short-run reflects the fact that theimmigration from the NMS-2 reduces the capital stock per capita in the short-run, whichis only partially compensated by higher labor market participation. Finally, the totalfactor income of the native population in the EU-15 is slightly reduced in the short-run,but it increases in the long-run. It is worth noting that the total factor income of nativesin the main receiving countries, Spain and Italy, increase by 0.46 and 0.43 per cent,respectively, in the long-run (Table 13).Table 13: The macroeconomic impact of migration from the NMS-2, 2004-2007Change oflabour forceGDPGDP per capitaFactor incomeper native Unemployment WagesShort-run Long-run Short-run Long-run Short-run Long-run Short-run Long-run Short-run Long-runChanges in per cent (unemployment rate: changes in percentage points)AT 0.13 0.09 0.10 -0.04 -0.03 0.03 0.04 0.01 0.01 -0.01 0.00BE 0.22 0.09 0.15 -0.07 -0.01 0.00 0.06 0.07 0.05 -0.04 0.00DE 0.04 0.02 0.04 -0.01 0.01 0.00 0.02 0.01 0.00 -0.01 0.00DK 0.03 0.02 0.03 -0.01 0.01 0.00 0.01 0.01 0.01 -0.01 0.00ES 2.29 0.42 1.33 -0.88 0.01 -0.44 0.46 0.65 0.24 -0.50 0.00FI 0.01 0.00 0.01 -0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00FR 0.06 0.03 0.05 -0.03 -0.01 0.00 0.02 0.01 0.01 -0.01 0.00GR 0.31 0.08 0.22 -0.13 0.01 -0.03 0.11 0.07 0.01 -0.08 0.00IE 0.33 0.09 0.24 -0.08 0.06 -0.04 0.11 0.06 0.02 -0.11 0.00IT 1.27 0.42 0.90 -0.39 0.08 -0.05 0.43 0.26 0.09 -0.32 0.00LU 0.15 0.10 0.15 -0.03 0.02 0.04 0.08 0.01 0.00 -0.04 0.00NL 0.04 0.03 0.04 -0.02 0.00 0.00 0.01 0.01 0.00 -0.01 0.00SE 0.05 0.02 0.03 -0.02 -0.01 0.00 0.01 0.01 0.01 -0.01 0.00UK 0.07 0.05 0.07 0.01 0.03 0.01 0.02 0.01 0.00 -0.01 0.00BG -1.93 -0.98 -1.98 0.97 -0.05 0.97 -0.05 -0.60 -0.08 0.50 0.00RO -4.70 -3.60 -4.83 1.15 -0.14 1.15 -0.14 -0.61 -0.16 0.84 0.00EU-15 1) 0.50 0.13 0.30 -0.19 -0.02 -0.05 0.13 0.13 0.05 -0.10 0.00NMS-2 -3.97 -2.91 -4.07 1.10 -0.12 1.10 -0.12 -0.57 -0.10 0.76 0.00Total 0.18 0.11 0.28 0.11 0.28 0.25 0.41 0.08 0.04 -0.10 0.001) Without Portugal.Source: Own estimates and simulation, see text.While the impact of immigration from the NMS-2 on unemployment in the EU-15 isalmost neutral in the long-run, it increases by 0.13 percentage points in the short-run.According to our simulations, the unemployment rate would have increased by 0.65percentage points in Spain and 0.26 percentage points in Italy in the short-run.IAB 28

However, we observe a distinct decline of the unemployment rate in Spain during theperiod of observation. There may be several explanations for this puzzle: Capital stocksmay have adjusted faster than projected, or the elasticity of the wage curve may belarger than according to our estimates.Wages decline in our model in the receiving countries by about 0.10 per cent in theshort-run. This is relatively moderate. In the two mainly affected receiving countries,Spain and Italy, wages decline by about 0.50 per cent (Spain) and 0.32 per cent (Italy)in the short-run. In the two sending countries, wages increase by 0.50 per cent(Bulgaria) and 0.84 per cent (Romania) in the short-run, while the long-run effects ofemigration on wages are neutral (Table 13).At the level of the EU-15, the short-run impact of immigration from the NMS-2 on thestructure of wages is – at between -0.05 and -0.15 per cent for the different skill groups– rather moderate. However, we observe distinct differences in the main destinationcountries: The wages for the less skilled (-0.02 per cent) and the medium skilled(-0.93 per cent) decrease in Spain in the long-run, while those of the high skilled tend torise (+0.46 per cent). In contrast, the effects on the structure of wages are ratherneutral in Italy in the long-run. In the sending countries, the wages tend to increase forthe high-skilled by 0.15 per cent in the long-run, while they decline for the medium andthe less skilled moderately. In the short-run, we observe again the largest wage increasefor high skilled workers (Table 14).IAB 29

Table 14: The impact of migration from the NMS-2 on wages, 2004-2007AllLow-skilledMedium-skilledHigh-skilledShort-run Long-run Short-run Long-run Short-run Long-run Short-run Long-runChanges in per centAT -0.01 0.00 -0.01 0.00 -0.01 0.00 -0.01 0.00BE -0.04 0.00 -0.04 0.00 -0.03 0.01 -0.05 -0.01DE -0.01 0.00 -0.01 0.00 -0.01 0.00 -0.01 0.00DK -0.01 0.00 -0.01 0.00 -0.01 0.00 -0.01 0.00ES -0.50 0.00 -0.48 -0.02 -1.42 -0.93 -0.09 0.46FI 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00FR -0.01 0.00 -0.01 0.00 -0.01 0.00 -0.01 0.00GR -0.08 0.00 -0.11 -0.02 -0.09 -0.01 -0.05 0.04IE -0.11 0.00 -0.11 0.00 -0.12 -0.01 -0.11 0.01IT -0.32 0.00 -0.31 0.00 -0.33 -0.01 -0.30 0.02LU -0.04 0.00 -0.02 0.02 -0.05 -0.01 -0.05 -0.01NL -0.01 0.00 -0.01 0.00 -0.01 0.00 -0.01 0.00SE -0.01 0.00 -0.01 0.00 -0.01 0.00 0.00 0.00UK -0.01 0.00 -0.01 0.00 -0.01 0.00 -0.01 0.00BG 0.50 0.00 0.49 0.02 0.46 -0.05 0.56 0.05RO 0.84 0.00 0.80 -0.04 0.77 -0.06 1.06 0.21EU-15 1) -0.10 0.00 -0.15 0.00 -0.12 -0.04 -0.05 0.05NMS-2 0.76 0.00 0.76 -0.03 0.71 -0.06 0.88 0.15Total -0.10 0.00 -0.14 0.00 -0.11 -0.04 -0.05 0.051) Without Portugal.Source: Own estimates and simulation, see text.The unemployment rate in the receiving countries increases for the less skilled by 0.20percentage points, for the medium skilled by 0.14 percentage points in the short-run andonly slightly by 0.03 percentage points for the high-skilled. In the long-run, theunemployment rate is declining for the high-skilled, but slightly increasing for the lowandmedium-skilled. Particularly affected are again medium skilled workers in Spain. Inthe sending countries, we observe that less-skilled and high-skilled workers benefitparticularly from falling unemployment rates in the long-run, while the medium skilledbenefit less than proportional (Table 15).IAB 30

Table 15: The impact of migration from the NMS-2 on unemployment, 2004-2007AllLow-skilledMedium-skilledHigh-skilledShort-run Long-run Short-run Long-run Short-run Long-run Short-run Long-runChanges in percentage pointsAT 0.01 0.01 0.02 0.01 0.00 0.00 0.04 0.04BE 0.07 0.05 0.15 0.11 0.03 0.01 0.05 0.03DE 0.01 0.00 0.02 0.00 0.01 0.00 0.01 0.01DK 0.01 0.01 0.00 0.00 0.00 0.00 0.02 0.02ES 0.65 0.24 0.64 0.20 1.40 0.99 0.06 -0.28FI 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00FR 0.01 0.01 0.01 0.00 0.02 0.02 0.01 0.01GR 0.07 0.01 0.07 0.02 0.07 0.00 0.03 -0.01IE 0.06 0.02 0.08 0.03 0.06 0.02 0.03 0.01IT 0.26 0.09 0.26 0.06 0.30 0.15 0.14 0.01LU 0.01 0.00 0.01 -0.01 0.02 0.01 0.02 0.01NL 0.01 0.00 0.01 0.00 0.00 0.00 0.01 0.01SE 0.01 0.01 0.03 0.02 0.01 0.01 0.00 0.00UK 0.01 0.00 0.01 0.00 0.01 0.01 0.00 0.00BG -0.60 -0.08 -1.12 -0.23 -0.49 0.00 -0.38 -0.11RO -0.61 -0.16 -0.62 -0.26 -0.66 -0.12 -0.54 -0.31EU-15 1) 0.13 0.05 0.20 0.06 0.14 0.09 0.03 -0.03NMS-2 -0.57 -0.10 -0.66 -0.20 -0.59 -0.07 -0.41 -0.15Total 0.08 0.04 0.15 0.05 0.07 0.07 0.01 -0.041) Without Portugal.Source: Own estimates and simulation, see text.5.4.4 The impact of transitional arrangements and the free movement ofworkers from the NMS-8, 2008–2011In this section we address the impact of a prolongation of the transitional arrangementsfor the free movement of workers from the NMS-8 as well as the implications ofintroducing the free movement for them. We evaluate the impacts during the 2008-2011period, i.e. until the date the transitional arrangements will finally expire. Note thatintroducing the free movement would trigger not only an increase of aggregate migrationbut also a reversal in the geographical distribution of the migration flows.Table 16 displays the macroeconomic effects of the prolongation of the transitionalarrangements and the introduction of the free movement. The difference between thesescenarios is interpreted as the effect of introducing the free movement in all remainingcountries in 2009. As a consequence of the redirection of migration flows away from theUK and Ireland we find that the GDP declines by 0.11 per cent in the UK and by 0.17 percent in Ireland, while the GDP increases in Germany by 0.11 per cent and by 0.24 percent in Austria in the free movement case compared to a prolongation of the transitionalarrangements. However, since both countries have to open their labour markets anywayin 2011, the effects are modest. The unemployment rate rises by 0.08 percentage pointsIAB 31

in Germany and 0.02 percentage points in Austria, while wages tend to decline (-0.08per cent in Germany and -0.02 per cent in Austria). 5Table 16: Short-run effects of transitional arrangements and the free movement ofworkers from the NMS-8, 2008-2011Change oflabour forceGDPGDP percapitaFactor incomeper native Unemployment WagesChanges in per cent (unemployment rate: changes in percentage points)AT 0.33 0.24 -0.01 0.09 0.02 -0.02BE -0.02 -0.01 0.01 0.00 -0.01 0.00DE 0.28 0.11 -0.09 -0.02 0.08 -0.08DK 0.01 0.01 0.00 0.00 0.00 0.00ES 0.00 0.00 0.00 0.00 0.00 0.00FI 0.11 0.03 -0.08 -0.02 0.03 -0.03FR 0.03 0.02 -0.01 0.00 0.01 0.00GR 0.07 0.02 -0.05 -0.01 0.02 -0.02IE -1.26 -0.17 0.57 0.24 -0.23 0.44IT 0.02 0.01 -0.01 0.00 0.00 -0.01LU -0.18 -0.15 -0.04 -0.06 -0.02 0.05NL -0.01 -0.01 0.00 0.00 0.00 0.00SE 0.05 0.03 0.00 0.01 0.01 -0.01UK -0.30 -0.11 0.08 0.02 -0.05 0.07CZ -0.30 -0.27 0.02 0.02 -0.06 0.08EE -0.32 -0.15 0.18 0.18 -0.07 0.09HU -0.33 -0.25 0.08 0.08 -0.03 0.08LT -0.12 -0.06 0.07 0.07 -0.03 0.03LV -0.13 -0.08 0.05 0.05 -0.03 0.04PL -0.01 -0.01 0.01 0.01 -0.01 0.00SI -0.54 -0.31 0.23 0.23 -0.05 0.09SK 0.10 0.04 -0.06 -0.06 0.04 -0.03EU-15 1) 0.02 0.02 -0.01 0.00 0.01 0.00NMS-8 -0.12 -0.12 0.00 0.00 -0.01 0.04Total 0.00 0.01 0.01 0.02 0.01 0.001) Without Portugal.Source: Own estimates and simulation, see text.5.4.5 The impact of transitional arrangements and the free movement ofworkers from Bulgaria and Romania, 2008-2014The selective application of immigration restrictions vis-à-vis workers from Bulgaria andRomania by the EU-15 countries has affected – similar to the NMS-8 – both the overallscale and the geographical distribution of migration flows from the NMS-2. ParticularlySpain and Italy experienced an immigration surge, while inflows to Germany and Austria5 For the effects on the structure on wages and unemployment see Table A1 in Appendix A.IAB 32

declined. Introducing the free movement of workers for Bulgaria and Romania willtherefore again both increase the number of immigrants and change the geographicaldistribution of immigration flows. The regional structure will change to a smaller extentcompared to the NMS-8.Our macroeconomic simulations reflect this picture. In Germany, the GDP will increase ifthe free movement is introduced, while the GDP per capita falls, wages tend to decline,and the unemployment rate tends to rise in the short-run (Table 17). This is offset in thelong-run due to the adjustment of capital stocks. Then GDP increases further, while thewage and unemployment effects diminish. The same picture can be drawn for Italy: GDPincreases there by 0.06 per cent, wages shrink by 0.04 per cent, and the unemploymentrises there by 0.03 percentage points as a consequence of further immigration. For Spainwe obtain a slightly different picture: The scale of migration under the transitionalarrangements and under the free movement is almost the same in the EU-15; however,the share of Spain in the overall inflows will decline if free movement is introducedaccording to our scenarios.Table 17: Short-run effects of transitional arrangements and the free movement ofworkers from Bulgaria and Romania, 2008-2014Change oflabour forceGDPGDP percapitaFactor incomeper native Unemployment WagesChanges in per cent (unemployment rate: changes in percentage points)AT 0.41 0.28 -0.12 0.11 0.03 -0.02BE -0.08 -0.03 0.02 0.00 -0.02 0.01DE 0.22 0.09 -0.06 -0.01 0.06 -0.06DK 0.03 0.02 0.00 0.00 0.01 -0.01ES -0.59 -0.12 0.21 0.10 -0.16 0.12FI 0.01 0.01 -0.01 0.00 0.00 0.00FR -0.04 -0.02 0.02 0.00 -0.01 0.00GR 0.35 0.08 -0.15 -0.03 0.07 -0.09IE 0.79 0.25 -0.16 -0.06 0.12 -0.24IT 0.17 0.06 -0.05 0.00 0.03 -0.04LU 0.04 0.03 -0.01 0.01 0.00 -0.01NL 0.00 0.00 0.00 0.00 0.00 0.00SE 0.02 0.01 -0.01 0.00 0.01 0.00UK 0.02 0.01 0.00 0.00 0.00 0.00BG -0.68 -0.36 0.33 0.33 -0.21 0.17RO -0.24 -0.19 0.05 0.05 -0.03 0.04EU-15 1) 0.03 0.03 0.00 0.01 0.00 -0.01NMS-2 -0.36 -0.24 0.13 0.13 -0.08 0.07Total 0.01 0.03 0.03 0.03 -0.01 -0.011) Without Portugal.Source: Own estimates and simulation, see text.Altogether, the enlarged EU is a winner of the free movement of workers within the EU.The joint GDP rises by 0.03 per cent and income of natives rises by 0.03 per cent relativeIAB 33

to a scenario where the present immigration restrictions under the transitionalarrangements are prolonged during the 2008–2014 period. 65.5 ConclusionsIn this section we applied a general equilibrium framework for the analysis of the impactof migration in the enlarged EU on wages, employment, and some macroeconomicaggregates. We modelled wage rigidities in form of a wage curve, assuming that wagesrespond imperfectly to an increase in the unemployment rate. We find an averageelasticity of the wage curve of -0.13, which is slightly higher than that found by theaverage of regional level studies. In our view, the higher elasticity reflects the impact ofcentralised wage setting, resulting in a higher elasticity of the wage curve if it ismeasured at the national level. Another important figure driving the results of our studyis the finding that capital stocks adjust to an increasing labour supply, although theseadjustments may take time. The speed of adjustment has been estimated and isconsidered in our simulations.The simulation of the impact of migration from the NMS-8 and the NMS-2 provides anumber of interesting insights. First, we observe that the additional migration from theNMS-8 caused by the EU’s Eastern enlargement during the 2004-2007 period hasincreased the aggregate GDP of the enlarged EU by about 0.11 per cent in the short-runand 0.20 per cent in the long-run, while the migration from the NMS-2 has increased theGDP of the enlarged EU by 0.11 in the short-run and by 0.28 per cent in the long-runduring the same period of time. Second, we observe that the total factor income ofnatives in the receiving countries tends to increase in the long-run, while it declines onlyslightly in the short-run. This can be traced back to the fact that complementary factorincomes tend to increase in case of migration. Third, we find that the unemployment isslightly increasing in the receiving countries in the short-run, while it is falling in thesending countries. The long-run effects of migration on the aggregate unemploymentrate are by and large neutral. Fourth, wages decline slightly in the receiving countriesand increase in the sending countries in the short-run, while the long-run impact ofmigration on wages is neutral. Fifth, we find that low- and medium skilled workers areslightly more affected by declining wages in the EU-15 compared to high-skilled workersin the short-term. This pattern reflects that migrants from the NMS are heavilyconcentrated at the low and medium ranges of the skill spectrum if we adjust for theiremployment structure.An important caveat is crucial to highlight here. In Ireland and Spain, which are thecountries mainly affected by immigration from the NMS-8 and the NMS-2, respectively,our simulations yield relatively large effects particular with respect to unemployment andwages. However, the labour supply shocks in both countries have not resulted in visiblechanges of the unemployment rates there. It is thus likely that we tend to overstate the6 For the effects on the structure on wages and unemployment see Table A2 in Appendix A.IAB 34

migration effects on these countries. There might be three explanations for this puzzle:First, capital stocks may adjust faster than predicted by our estimates. Second, the wageresponse might be larger than is expected by our estimates of the wage curve. As anexample, Bentolila et al. (2007) argue that immigration itself has changed the bargainingposition of workers, such that responsiveness of wages has increased through higherimmigration. Thus, wages may decline even further, while the unemployment effects aresmaller compared to our simulations. Third, there may be other adjustment mechanismswhich are not considered by our model but mitigate the effects of migration on wagesand unemployment such as sectoral change and international trade. The latter aspect isaddressed by the model presented in the next section.6 The macroeconomic consequences of labour mobility: The impact ofmigration, trade and capital mobility in a multisectoral CGE modelIn this section we examine the effects of labour mobility in the context of EU Easternenlargement on two destination economies, the UK and Germany and the sendingeconomies Poland, Hungary, Slovakia, and Slovenia. The study is based on a computablegeneral equilibrium (CGE) model comprising 16 commodities, 16 domestic industries andreflecting trade of intermediary and final goods as well as the movement of capital.CGE models have been widely applied for the analysis of the impact of the EU integrationprocess. Integration in this sense is typically modelled as a reduction in transactioncosts, especially the cost of trade, of capital movement, and of migration betweencountries. The strength of this kind of numerical CGE models lies in the illustration of thecomplex interactions underlying these processes. With this CGE model we are thereforeable to examine interactions between trade, capital movements and migration and toanalyse the impact of migration at the sectoral level.The analysis in this section proceeds in four steps. In Section 6.1 we briefly outline theunderlying theoretical model. Section 6.2 describes the calibration of the model and thedata used. In Section 6.3 we present the simulation results for the different policyscenarios and the counterfactual scenario. This allows us to consider the impact ofmigration in the specific context under the transitional arrangements (2004-2007) andbased on our migration projections the effects of free movement (2008-2011 for theNMS-8 and 2008-2014 for the NMS-2). We describe the scenarios first and present thenthe results country by country. In Section 6.4 we summarise the sectoral results anddiscuss their impact on the economy again country by country. Section 6.5 concludes.6.1 Outline of the modelThe CGE model employed here can be classified as a standard comparative static modelbased on the IFPRI 7 framework. The IFPRI type models follow the neoclassic-structuralistmodelling tradition first presented in Dervis et al. (1982). The equations of the model are7 IFPRI (2002) provide a standard CGE model, easy to enhance. Most modern CGE models arebased on this framework, due to the excellent report procedures included in the model code.IAB 35

derived from microeconomic assumptions about the behaviour of price taking agents.Consumers maximize utility subject to their budget constraints. Producers choose inputsso as to minimize production costs. Production technologies are characterised by a CESor Leontief function whereby resources are limited and distributed by market forces.The model consists of n = 16 commodities, m = 16 domestic industries, and h = 2 typesof households, migrants and natives. In total there are 2 agricultural industries, 4manufacturing industries and 10 service industries. Each commodity corresponds to anindustry. The consideration of two types of households allows considering the differentconsumption behaviour of natives and migrants. The empirical basis of the model isformed by the current input-output matrices from Eurostat which enable us to considerthe recent developments in the interconnection between trade, factor movements andproduction.In order to capture the effects of the European integration process, we enhanced the twocountry framework of the IFPRI model to a three country framework which reflects onecountry and two regions, the EU and the rest of the world (see Baas and Brücker, 2008).The economies of Germany and the UK are linked to the EU and to the rest of the worldvia trade in goods and services, capital flows and the migration of labour. Transactioncosts within the EU are lower; therefore we consider the different trade pattern emergingfrom EU integration and distinguish between Intra- and Extra-EU trade.Governmental consumption is restricted to tax income and borrowing, which hasimplications for other economic agents.An important feature of the model is the reflection of labour market imperfections by awage curve which is novel in the CGE literature on the effects of EU Eastern enlargement(compare Baas and Brücker, 2008). The consideration of labour market rigidities throughthe specification of a wage curve postulates a negative relationship between the realwage rate and the unemployment rate (Blanchflower and Oswald, 1994, see also the firstsection of this deliverable). Hence, migration leads to lower wages and higherunemployment in the destination country, while unemployment is reduced and wagesrise in the sending country. Nevertheless, we model a short-run scenario reflectingimperfect adjustment of the capital-output ratio, which should fully adjust in the longrun.The adjustment parameters in the model are therefore estimated.The technical features of the model are described in detail in Appendix B.6.2 Data and calibration of the modelThe numerical specification of the CGE model is undertaken by using the Eurostat supplyand demand matrices. The matrices are compiled according to the European Systems ofAccounts ESA 95 which provide common classifications and a harmonised methodologyalong the convention in harmonising national gross domestic products within theEuropean Union. The transmission of input-output tables is compulsory since the end of2002. This concerns annual supply and demand matrices and five-yearly symmetricinput-output-matrices. Nevertheless, data quality and the transmission of matrices differalong member states. Some supply and demand matrices are not symmetric while otherIAB 36

matrices suffer missing or hidden values. The application of CGE-modelling on base ofthese matrices is therefore restricted.The supply and demand matrices provide detailed information on the economic system.The demand table provides inter alia information on intermediate consumption, theapplication of factors of production, taxation and subsidies at the activity level andconsumption of households, the government, and external trade. The supply matricesshow inter alia the production of marketed output, the import of goods and services, andsales taxes. The demand and supply matrix is combined to a symmetric input-outputmatrix with industries and activities. Since the classification of goods relies on CPA 8systematic, goods and activities use the same nomenclature which facilitates thecalibration of the model.Beside the data obtained from Eurostat matrices, additional data is needed to reflectinter alia the level of labour market restrictions, the welfare system, and trade issues.Hence, a social accounting matrix (SAM) is compiled as an extended symmetric inputoutputtable. Whenever possible we used Eurostat data to build the SAM matrix of acountry.After the specification of the SAM, the theoretical model parameters are calibrated to realvalues. Thus, in a first step, the model is solved using the SAM variables as variables ofthe model. This provides us with information about the parameters of the model. In asecond step the model is solved using the calibrated parameters. The solution of thesecond run is compared with the SAM data. If the model matches this data, the baseyear model is calibrated and can be used for simulation. In Appendix B we provide thekey equations of the theoretical model, while Appendix C presents a figure of a typicalSAM.6.3 Simulation resultsThe following six subsections present the country-specific macroeconomic effects of theEU Eastern enlargement on Germany, Hungary, Poland, the UK, Slovakia, and Slovenia.The simulations presented here consider the impact of migration on GDP, thegovernment, trade, capital movements, and the structure of the economy by sectors. Asoutlined in the introduction, the effects of migration are captured by two policyscenarios: The first scenario describes the effects of Eastern enlargement under thetransitional arrangements whereas the second scenario describes a situation of freemovement beginning in 2009. The first scenario covers a time period from 2004 to 2007,the second a period from 2008 to 2011 (2008-2014 for the NMS-2 countries).The selection of countries which are considered here is particularly relevant. The UK isthe country which has been in absolute terms mainly affected by migration in theaftermath of enlargement, since it has almost completely removed the barriers forworker mobility vis-à-vis the new member states. In contrast, Germany still heavilyrestricts migration from the NMS, but has been in absolute terms the main destination8 Statistical Classification of Products by Activity in the European Economic Community.IAB 37

for migration from there before enlargement. The four sending countries differ withrespect to their size and the amount of migrants working abroad. Therefore thesecountries are affected by the EU Eastern enlargement very differently. According to ourestimates, about 1.3 million migrants from Poland will reside in the EU-15 by 2007 in theEastern enlargement scenario, while only 630,000 Polish migrants would live there in thecase without enlargement. The difference accounts for almost two per cent of the Polishworkforce. While Slovakia experiences a similar effect of EU enlargement, theneighbouring country Slovenia is much less affected by emigration, as well as themedium sized Hungary.The results in Table 18 reflect these differences in migration after enlargement. Ingeneral, the sending countries experience a reduction in GDP and unemployment whilewages increase. Per capita GDP is therefore higher after the enlargement. Otherwise, thereceiving countries’ GDP and unemployment rates are higher and wages are lower withEU-enlargement, but GDP per capita declines.In the second policy scenario we see a partial reversion of the effects of migrationdiversion after EU Enlargement. On the one hand the main destination country after2004, the UK, gains fewer migrants with free movement. Therefore the GDP declines,wages rise and unemployment is reduced. On the other hand, Germany experiences arise in migration with free movement. That’s why the GDP increases, while wages declineand unemployment rises. Germany regains the role as a mayor destination country forNMS-8 migrants in this scenario. However, since there are only two years left of thepossibility to apply transitional periods, effects are small.IAB 38

Table 18:Simulation Results, Key Macroeconomic Figures, NMS-8EnlargementeffectsGermany UK Hungary Poland Slovenia SlovakiaFreemovementeffectEnlargementeffectsFreemovementeffectEnlargementeffectsFreemovementeffectEnlargementeffectsFreemovementeffectEnlargementeffectsFreemovementeffectEnlargementeffectsFreemovementeffect2004-2007 2008-2011 2004-2007 2008-2011 2004-2007 2008-2011 2004-2007 2008-2011 2004-2007 2008-2011 2004-2007 2008-2011changes in percentGDP 0.06 0.17 0.86 -0.20 -0.23 -0.19 -0.92 -0.01 0.17 -0.38 -0.44 0.03GDP per capita -0.02 -0.06 -0.03 0.01 0.18 0.15 0.81 0.02 -0.06 0.16 0.81 -0.08Exports intra EU 0.12 0.33 1.24 -0.29 -0.21 -0.17 -1.25 -0.01 0.20 -0.45 -0.26 0.01Exports extra EU 0.12 0.32 1.09 -0.26 -0.21 -0.17 -1.24 -0.01 0.20 -0.45 -0.27 0.01Imports intra EU 0.05 0.12 0.81 -0.19 -0.25 -0.20 -0.80 0.00 0.16 -0.35 -0.54 0.04Imports extra EU 0.05 0.13 0.89 -0.21 -0.24 -0.20 -0.81 0.00 0.16 -0.36 -0.54 0.04Wages -0.02 -0.06 -0.34 0.08 0.12 0.10 0.32 0.01 -0.05 0.13 0.34 -0.03changes in percentage pointsUnemployment rate 0.02 0.06 0.13 -0.03 -0.08 -0.07 -0.48 -0.01 0.03 -0.07 -0.45 0.04Notes: The simulation results indicate the difference between the status-quo scenario and the counterfactual scenario of no enlargement.Sources: Own estimates.IAB 39

In both scenarios our results predict moderate effects of migration on wages andunemployment. The migration effect is mitigated in case of a partial adjustment of thecapital stock and a redistribution of factors among sectors. Therefore, we observe anincrease in labour but also an increase in capital in the destination countries. In thesending countries, capital is correspondingly reduced. The sectoral factor mobilityassures, as a second effect, that the new factor endowments are distributed to theirmost productive use.Migration also affects trade patterns. In all countries except Poland, migration improvesthe trade balance. In Germany, we observe only a small migration effect of 0.05 per centon imports, but a strong 0.12 per cent increase in exports (see Table 18). Interestingly,in most countries trade with EU countries (Intra-EU trade) and trade with third countries(Extra-EU trade) reacts similar. Only in the UK, Intra-EU trade reacts more strongly thanExtra-EU trade.In the reminder of this section we take a closer look at country specific effects. Theseeffects are driven by the production structure of the economy, the openness of theeconomy and the migration shock.6.3.1 GermanyThe migration structure in the aftermath of EU Eastern enlargement changes migrationpatterns heavily. Germany as the former main receiving country is therefore no longerthe main destination of migrants after the enlargement. Indeed, we estimate an increasein migration by 62,000 compared to pre-enlargement figures. This is only a moderateincrease which shows the strict application of transitional agreements. Hence,macroeconomic effects in Germany are small. This migration pattern is reversed if weassume free movement from 2009 on.If we assume that migrants are employed as their already migrated counterparts, thelabour supply shock increases the labour force in the enlargement scenario by 42,000.This figure considers an employment rate of NMS-migrants in Germany of 64 per cent,which is only slightly higher than the corresponding employment rate of natives.As the simulation results show, migration from the NMS-8 countries has only a smallimpact on the German economy (see Table 19). In the enlargement scenario, theincrease in GDP is small at 0.06 per cent, while the free movement scenario addsanother 0.17 per cent. The impact of migration from Bulgaria and Romania (NMS-2), isalmost neglible. The GDP rises by 0.01 per cent in the enlargement scenario. However, inthe free movement scenario we observe a 0.14 per cent increase in GDP after all.As discussed in Chapter 5, we use a wage curve for modelling the labour market. Hence,a labour supply shock leads by assumption to lower wages and higher unemployment. Inthe enlargement scenario, wages are shrinking by about 0.02 per cent. Therefore, asIAB 40

expected, EU enlargement has not affected the key macroeconomic variables of Germanyvery much. This is due to the small labour supply shock.If Germany abstains from applying the transitional arrangements in 2009, migrationenhances the labour force in the free movement scenario by an additional 0.28 per centand increases the GDP by 0.17 per cent (see Table 19). The additional migration leads toa rise in the unemployment rate of 0.06 percentage points and a reduction of wages byabout -0.06 per cent. However, effects on GDP per capita are modest at -0.06 per centsince the labour market participation rate of migrants is higher than that of natives.Table 19:Simulation results Germany, key macroeconomic figuresEnlargement effectFree movement effectNMS-8 NMS-2 NMS-8 NMS-2Base year 2004-2007 2004-2007 2008-2011 2008-2014Changes in per centGDP 2211200 0.06 0.01 0.17 0.14GDP per capita 26791 -0.02 0.00 -0.06 -0.04Private consumption 1239350 0.03 0.00 0.08 0.08Investment 377050 0.04 0.01 0.10 0.09Government consumption 453240 0.04 0.01 0.11 0.10Taxes 231490 0.06 0.01 0.15 0.13Exports intra EU 514790 0.12 0.02 0.33 0.26Exports extra EU 311461 0.12 0.02 0.32 0.25Imports intra EU -405720 0.05 0.01 0.12 0.11Imports extra EU -278971 0.05 0.01 0.13 0.12Wages 29 -0.02 0.00 -0.06 -0.04Capital 841910 0.02 0.00 0.05 0.06Labour force 42551 0.10 0.02 0.28 0.22Changes in percentage pointsUnemployment rate 9 0.02 0.00 0.06 0.04Source: Own estimates and simulation, see text.As we see in Table 20, migration influences the sectoral structure and the trade patternof the economy. However, sectoral adjustments are small. Only the manufacturing sectorproducing tradable goods is affected by the labour supply shock in the free movementscenario above the average production increase (an increase of 0.3 per cent in case offree movement), while all other sectors enhance their production only slightly (0.2 percent in total with free movement).IAB 41

Table 20:Simulation results Germany, sectoral impactEnlargement effectFree movement effectBase Year NMS-8 NMS-2 NMS-8 NMS-2Changes in per centAgriculture, hunting and forestry 47730 0.10 0.00 0.10 0.10Fishing 420 0.00 0.00 0.10 0.10Mining and quarrying 12590 0.00 0.00 0.20 0.20Manufacturing 1357440 0.10 0.00 0.30 0.20Electricity, gas and water supply 91220 0.00 0.00 0.10 0.10Construction 189440 0.10 0.00 0.10 0.10Wholesale and retail trade 1) 343810 0.00 0.00 0.20 0.20Hotels and restaurants 62070 0.10 0.00 0.10 0.10Transport, storage andcommunication261690 0.00 0.00 0.10 0.10Financial intermediation 221390 0.10 0.00 0.10 0.10Real estate, renting and businessactivitiesPublic administration and defence;compulsory social security676450 0.10 0.00 0.10 0.10175940 0.10 0.00 0.10 0.10Education 114210 0.10 0.00 0.10 0.10Health and social work 204850 0.10 0.00 0.10 0.10Other community, social andpersonal service activities153330 0.10 0.00 0.10 0.10Activities of households 6620 0.10 0.00 0.10 0.10Total 3919200 0.00 0.00 0.20 0.201) Includes also the repair of motor vehicles, motorcycles, and personal and household goods.Source: Own estimates and simulation, see text.6.3.2 UKIn the aftermath of EU-enlargement the UK opted for the free movement of workers fromNMS-countries. The only obligation for migrants is to register, yet access to welfare isrestricted. Migration therefore increases heavily by 455,000, while the labour forceincreases by 340,000. This strong increase in the labour force is initially driven by themigration shock itself, but also from the high employment rate of migrants of 75 perIAB 42

cent. Interestingly, the employment rate of NMS-2 migrants is even higher with 84 percent. Both figures are even larger than the employment rate of natives and essentiallylarger than the employment rate of NMS-migrants in Germany. Accordingly, themacroeconomic effects of migration are strong.In Table 21 we see the development of key macroeconomic figures in the enlargementand free movement scenario. As we can see, macroeconomic effects are driven by thelarge immigration from NMS-8 countries. Therefore the GDP in the enlargement scenarioincreases by 0.86 per cent. The GDP per capita shrinks with 0.03 per cent only modestly.The high participation rate of NMS-8 workers compensates to some extend their lowcapital endowment.The impact of migration on trade is similar to Germany. Migration enhances exports andimports, but the effect on exports is stronger. However, for the UK the differencebetween imports (0.81 Intra EU / 0.89 Extra EU) and exports (1.24 Intra EU / 1.09 ExtraEU) are relatively smaller and more differentiated among destinations than in Germany.Therefore, the trade balance with the rest of the world improves only modestly, while thetrade balance with other EU countries improves strongly.In all models, a wage curve drives the labour market effects. Given the size of the shock,we find a relatively small rise in unemployment (0.13 percentage points, EUenlargement)and a small reduction in wages (0.34 per cent, EU-enlargement).In the free movement scenario, we predict a decrease in migration. The labour force isreduced by 0.3 per cent compared to a situation where some EU-countries like Germanyand Austria stay closed. This leads to a partial reversion of the effects of migrationobserved with transitional periods. The GDP is shrinking and the rise in GDP per capita isalmost negligible, while exports and imports are lower. We also see that theimprovement of the trade balance is partly reversed, if all countries adapt freemovement. Consequently, wages rise by 0.08 per cent, while unemployment is reducedby 0.03 percentage points.IAB 43

Table 21:Simulation results UK, key macroeconomic figuresEnlargement effectFree movement effectNMS-8 NMS-2 NMS-8 NMS-2Base year 2004-2007 2004-2007 2008-2011 2008-2014Changes in per centGDP 1147947 0.86 0.02 -0.20 0.02GDP per capita 19313 -0.03 0.00 0.01 0.00Private consumption 727827 0.76 0.02 -0.18 0.01Investment 179922 0.73 0.02 -0.17 0.01Government consumption 259197 0.90 0.02 -0.21 0.02Taxes 140934 0.85 0.02 -0.20 0.02Exports intra EU 142337 1.24 0.02 -0.29 0.02Exports extra EU 126816 1.09 0.02 -0.26 0.02Imports intra EU -162886 0.81 0.02 -0.19 0.01Imports extra EU -125266 0.89 0.02 -0.21 0.02Wages 22 -0.34 -0.01 0.08 0.00Capital 391375 0.34 0.01 -0.08 0.01Labour force 29652 1.28 0.02 -0.30 0.02Changes in percentage pointsUnemployment rate 5 0.13 0.00 -0.03 0.00Source: Own estimates and simulation, see text.If we look at the results for production, we see an overall increase. However, somesectors like Manufacturing, Education and Health, and Social Work enhance theirproduction above average. This can be traced back to two facts: On the one hand thereis a direct increase in labour supply in these sectors by migrants; on the other hand,native workers shift sectors if they can be more productive there. This second indirecteffect can outpace the direct migration effect as is the case in the manufacturing sector.Altogether, production is rising strongly in the enlargement scenario by 0.8 per cent, dueto a sharp rise in the labour force. In the free movement scenario, where no country optsfor transitional periods, we see an overall lower production of 0.2 per cent. The sectorswhich gained most from direct or indirect migration effects lose more, that’s why we seea slight reversion of the migration-driven sectoral distribution of additional production.IAB 44

Table 22:Simulation results UK, sectoral impactEnlargement effectFree movement effectBase Year NMS-8 NMS-2 NMS-8 NMS-2Changes in per centAgriculture, hunting and forestry 21935 0.70 0.00 -0.20 0.10Fishing 1801 0.50 0.00 -0.10 0.10Mining and quarrying 32508 0.40 0.00 -0.10 0.10Manufacturing 401402 1.10 0.00 -0.30 0.10Electricity, gas and water supply 49691 0.70 0.10 -0.10 0.10Construction 158998 0.70 0.00 -0.10 0.10Wholesale and retail trade 1) 233390 0.90 0.00 -0.30 0.10Hotels and restaurants 65163 0.80 0.10 -0.20 0.10Transport, storage and communication 168203 0.80 0.00 -0.20 0.10Financial intermediation 153374 0.60 0.10 -0.20 0.10Real estate, renting and businessactivitiesPublic administration and defence;compulsory social security362583 0.60 0.00 -0.20 0.10107425 0.90 0.00 -0.20 0.10Education 82117 0.90 0.10 -0.20 0.10Health and social work 130207 0.90 0.10 -0.20 0.10Other community, social and personalservice activities94650 0.70 0.10 -0.20 0.10Activities of households 4957 0.90 0.00 -0.30 0.10Total 2068403 0.80 0.10 -0.20 0.101) Includes also the repair of motor vehicles, motorcycles, and personal and household goods.Source: Own estimates and simulation, see text.6.3.3 HungaryIn the aftermath of EU-enlargement Hungary reports an unemployment rate at about 6per cent. The compensation of employees in Hungary was only 36.5 per cent of EU-25average, but above the NMS-8 figure of 29.7 per cent. Migration therefore affected theHungarian economy below the average of NMS-8 countries. We estimate a migrationIAB 45

effect of EU-enlargement of 44.000 emigrants which is 0.41 per cent of the Hungarianlabour force.Table 23:Simulation results Hungary, key macroeconomic figuresEnlargementeffectNMS-2Free movementeffectBase year 2004-2007 2008-2011Changes in per centGDP 18575041 -0.23 -0.19GDP per capita 1831431 0.18 0.15Private consumption 10354737 -0.25 -0.20Investment 4533796 -0.24 -0.20Government consumption 4812376 -0.28 -0.23Taxes 2636108 -0.24 -0.20Exports intra EU 7918169 -0.21 -0.17Exports extra EU 2923220 -0.21 -0.17Imports intra EU -6686784 -0.25 -0.20Imports extra EU -5280472 -0.24 -0.20Wages 2195 0.12 0.10Capital 7403523 -0.15 -0.12Labour force 4265 -0.41 -0.34Changes in percentage pointsUnemployment rate 9 -0.08 -0.07Source: Own estimates and simulation, see text.The reduction in the labour force is reducing production and therefore GDP. Nevertheless,since the population declines, per capita GDP is rising. In Table 23 the GDP is reduced atabout 0.23 per cent in the EU enlargement scenario.The assumption of a partial adjustment of the capital stock leads to a decline in capitalendowment; consequently investment is reduced in the simulation model by 0.24 percent. Nevertheless, the trade balance is slightly improving. Exports and Imports aremoving closely among the same rate as GDP is shrinking, but the decline of exports isweaker (0.21 per cent) than the decline of imports (0.25 per cent).If the application of transitional periods would be dropped in 2009 by the remainingcountries, migration from Hungary would increase. This would strengthen the effectsalready seen in the enlargement scenario. The GDP is declining, while GDP per capita isimproving (0.15 per cent). The trade balance is again slightly improving, while exportsdecline less (0.17 per cent) than imports (0.20 per cent). In both scenarios Intra-EU andExtra-EU imports and export react roughly similar.IAB 46

As we see in Table 24, the reduced labour force does not lead to a strong redistributionof production among sectors. Production is shrinking in the enlargement and the freetrade scenario by 0.2 per cent and almost all sectors are reducing their production at thisamount. Hence, there is no big difference between tradable goods and non-tradablegoods.IAB 47

Table 24:Simulation results Hungary, sectoral impactEnlargementeffectFree movementeffectBase year 2004-2007 2008-2011Changes in per centAgriculture, hunting and forestry 1650517 -0.20 -0.20Fishing 23954 -0.30 -0.20Mining and quarrying 106092 -0.30 -0.20Manufacturing 14914072 -0.20 -0.20Electricity, gas and water supply 1394533 -0.20 -0.10Construction 2057310 -0.20 -0.10Wholesale and retail trade 1) 3752648 -0.20 -0.20Hotels and restaurants 688831 -0.30 -0.20Transport, storage and communication 2364454 -0.30 -0.20Financial intermediation 1155924 -0.20 -0.20Real estate, renting and businessactivitiesPublic administration and defence;compulsory social security4593807 -0.20 -0.202050603 -0.30 -0.20Education 1238609 -0.30 -0.20Health and social work 1326516 -0.30 -0.20Other community, social and personalservice activities1234931 -0.30 -0.20Activities of households 2)Total 38552800 -0.20 -0.201) Includes also repair of motor vehicles, motorcycles, and personal and household goods.2) Blank fields indicate missing values in the I/O-tables.Source: Own estimates and simulation, see text.IAB 48

6.3.4 PolandIn the aftermath of EU-enlargement unemployment in Poland was high at 19.6 per cent.Additionally, the compensation of employees was at 28.8 per cent of the EU-25 and thusbelow the average of NMS-8 countries (29.7 per cent). Emigration from poland thereforewas strong; 666,000 migrants left Poland in the aftermath of the EU-enlargementNevertheless, the Polish participation rate was low with 51 per cent, which reduces theimpact of the migration shock on the Polish economy. The reduction of labour force iswith 1.71 per cent below the population shock, but still strong. Hence, we can expectlarge macroeconomic effects.Table 25:Simulation results Poland, key macroeconomic figuresEnlargementeffectNMS-2Free movementeffectBase year 2004-2007 2008-2011Changes in per centGDP 843156 -0.92 -0.01GDP per capita 22061 0.81 0.02Private consumption 546077 -0.75 0.00Investment 158028 -0.78 0.00Government consumption 165567 -0.88 0.00Taxes 108194 -0.84 0.00Exports intra EU 185441 -1.25 -0.01Exports extra EU 83540 -1.24 -0.01Imports intra EU -179284 -0.80 0.00Imports extra EU -116214 -0.81 0.00Wages 24 0.32 0.01Capital 412916 -0.64 0.00Labour force 16946 -1.71 -0.02Changes in percentage pointsUnemployment rate 20 -0.48 -0.01Source: Own estimates and simulation, see text.Emigration from Poland leads to a strong decrease in GDP (see Table 25). In theenlargement scenario, migration reduces GDP by 0.92 per cent. As we see in bothscenarios, trade is strongly affected by the labour supply shock. Intra-EU and Extra-EUexports are declining by roughly 1.25 per cent and imports decline by 0.8 per cent.Consequently, the trade balance is worsening. The strong decline in trade indicates aIAB 49

edistribution of production among sectors (see Table 26). We can see, that tradablesectors like manufacturing reduce their production by 1 per cent, while service sectorslike hotel and restaurant reduce their production by 0.8 per cent, only. However, mostother service sectors reduce their production like the average of all sectors by 0.9 percent. Nevertheless, the labour supply shock enhances wages by 0.32 per cent andstrongly reduces unemployment by 0.48 percentage points.The effects of the free movement scenario are negligible in all categories due to thediminutive decrease of labour supply with free movement of workers to all EU-countries.IAB 50

Table 26:Simulation results Poland, sectoral impactEnlargementeffectFree movementeffectBase year 2004-2007 2008-2011Changes in per centAgriculture, hunting and forestry 78123 -0.80 0.00Fishing 476 -0.90 0.00Mining and quarrying 26835 -1.00 -0.10Manufacturing 493498 -1.00 0.00Electricity, gas and water supply 68749 -0.80 0.00Construction 115113 -0.80 0.00Wholesale and retail trade 1) 260694 -0.90 0.00Hotels and restaurants 19457 -0.80 0.00Transport, storage and communication 128485 -0.90 0.00Financial intermediation 55051 -0.80 0.00Real estate, renting and businessactivitiesPublic administration and defence;compulsory social security192624 -0.80 0.0063339 -0.90 0.00Education 44994 -0.90 0.00Health and social work 46915 -0.90 0.00Other community, social and personalservice activities51825 -0.90 0.00Activities of households 5275 -0.70 0.00Total 1651452 -0.90 0.001) Includes also the repair of motor vehicles, motorcycles, and personal and household goods.Source: Own estimates and simulation, see text.IAB 51

6.3.5 SloveniaIn the aftermath of EU-enlargement, unemployment in Slovenia was comparatively lowat 6.7 per cent. The compensation of employees was with 57.7 per cent of the EU-25average well ahead of the NMS-8 figure of 29.7 per cent. Migration thus affected theSlovenian economy only slightly. We estimated a migration effect of EU-enlargementwhich is even lower than the counterfactual assumption of no enlargement. However,numbers are small, 5100 emigrants stay after EU-enlargement in Slovenia and do notmove into the EU-15 countries.Table 27:Simulation results Slovenia, key macroeconomic figuresEnlargementeffectNMS-2Free movementeffectBase year 2004-2007 2008-2011Changes in per centGDP 5813540 0.17 -0.38GDP per capita 2914007 -0.06 0.16Private consumption 3332074 0.14 -0.31Investment 1433367 0.15 -0.33Government consumption 1213919 0.17 -0.37Taxes 864309 0.16 -0.36Exports intra EU 1746315 0.20 -0.45Exports extra EU 1223014 0.20 -0.45Imports intra EU -2284272 0.16 -0.35Imports extra EU -850877 0.16 -0.36Wages 3363 -0.05 0.13Capital 1936348 0.10 -0.20Labour force 959 0.23 -0.54Changes in percentage pointsUnemployment rate 7 0.03 -0.07Source: Own estimates and simulation, see text.Slovenia is an exception in the countries analysed in this chapter. The EU enlargementhas lead to a lower migration than we would predict without enlargement. Therefore, theGDP and the unemployment rate are higher while GDP per capita and wages are lowerwith enlargement. This effect is only reversed if all countries allow free movement ofworkers from the NMS-8. In the free movement scenario the labour force in Slovenia isreduced by 0.54 per cent. Therefore, the usual pattern of sending countries is reached,the GDP declines by 0.38 per cent and GDP per capita rises by 0.16 per cent. Intra-EUand Extra-EU exports and imports react very similar in this scenario. While exports areIAB 52

educed by 0.45 per cent more strongly than imports (0.35 per cent), the trade balanceis slightly worsening.The sectoral structure of Slovenia shows a shock which enhances production in all sectorsequally (see Table 28). Thus, we see no big divergence in tradable and non-tradablegoods in the enlargement scenario. The stronger reduction of exports in the freemovement scenario follows a reduction of manufacturing production by 0.4 per cent,which is above the average of 0.3 per cent.IAB 53

Table 28:Simulation results Slovenia, sectoral impactEnlargementeffectFree movementeffectBase year 2004-2007 2008-2011Changes in per centAgriculture, hunting and forestry 294424 0.10 -0.20Fishing 2226 0.20 -0.30Mining and quarrying 50559 0.20 -0.30Manufacturing 4247767 0.20 -0.40Electricity, gas and water supply 307089 0.20 -0.30Construction 1065401 0.20 -0.30Wholesale and retail trade 1) 1134677 0.20 -0.30Hotels and restaurants 243865 0.20 -0.30Transport, storage and communication 868086 0.20 -0.30Financial intermediation 372874 0.20 -0.30Real estate, renting and business activities 1305820 0.20 -0.30Public administration and defence;compulsory social security524485 0.20 -0.30Education 376102 0.20 -0.30Health and social work 400073 0.20 -0.30Other community, social and personalservice activities325362 0.20 -0.30Activities of households 1336 0.20 -0.30Total 11520146 0.20 -0.301) Includes also the repair of motor vehicles, motorcycles, and personal and household goods.Source: Own estimates and simulation, see text.IAB 54

6.3.6 SlovakiaSlovakia is a small country which is heavily affected by migration. In the aftermath ofEU-enlargement, unemployment in Slovakia was comparatively high at 17.6 per cent andtherefore higher than in all other NMS-8 countries except Poland. The compensation ofemployees was with 23.2 per cent of the EU-25 average lower than in all other NMS-8countries. Migration thus affected the Slovakian economy heavily. We estimated amigration effect of EU-enlargement of 72,000 emigrants, which is high compared to thesmall size of Slovakia.The GDP in Slovakia is reduced by 0.44 per cent due to enlargement. Interestingly,exports are reacting half as much to the migration shock than imports (see Table 29).This indicates strong differences in the reduction of production among sectors.Furthermore, Intra-EU and Extra-EU exports (0.26 / 0.27 per cent) and imports (0.54per cent) are reacting very similarly.The opening up of labour markets in the remaining EU-15 countries does not lead tostrong effects in Slovakia. Surprisingly, migration is slightly higher with transitionalperiods than with free movement. Therefore, we see a small increase in GDP and lowerGDP per capita due to a lower reduction in labour supply with free movement. However,these effects are extremely small.IAB 55

Table 29:Simulation results Slovakia, key macroeconomic figuresEnlargementeffectNMS-2Free movementeffectBase year 2004-2007 2008-2011Changes in per centGDP 1357312 -0.44 0.03GDP per capita 252328 0.81 -0.08Private consumption 762032 -0.63 0.05Investment 356776 -0.61 0.04Government consumption 275032 -0.72 0.05Taxes 138065 -0.54 0.04Exports intra EU 859842 -0.26 0.01Exports extra EU 153022 -0.27 0.01Imports intra EU -796449 -0.54 0.04Imports extra EU -252944 -0.54 0.04Wages 231 0.34 -0.03Capital 719469 -0.46 0.03Labour force 2624 -1.23 0.10Changes in percentage pointsUnemployment rate 18 -0.45 0.04Source: Own estimates and simulation, see text.The production in Slovakia is reduced by 0.6 per cent in the enlargement scenario (seeTable 30). Nevertheless, manufacturing is only reduced below average (0.4 per cent),while the non-tradable sectors reduce production more heavily (0.6 to 0.7 per cent).Hence, the sectoral structure is heavily affected by the migration shock.IAB 56

Table 30:Simulation results Slovakia, sectoral impactEnlargementeffectFree movementeffectBase year 2004-2007 2008-2011Changes in per centAgriculture, hunting and forestry 111422 -0.60 0.00Fishing 595 -0.60 0.10Mining and quarrying 13084 -0.70 0.00Manufacturing 1176469 -0.40 0.00Electricity, gas and water supply 246955 -0.60 0.00Construction 219925 -0.70 0.10Wholesale and retail trade 1) 342576 -0.50 0.10Hotels and restaurants 35122 -0.70 0.10Transport, storage and communication 252829 -0.60 0.00Financial intermediation 79830 -0.60 0.10Real estate, renting and businessactivitiesPublic administration and defence;compulsory social security274403 -0.60 0.00129796 -0.70 0.00Education 53273 -0.70 0.00Health and social work 64502 -0.70 0.00Other community, social and personalservice activities72893 -0.70 0.00Activities of householdsTotal 3073675 -0.60 0.101) Includes also repair of motor vehicles, motorcycles, and personal and household goods.2) Blank fields indicate missing values in the I/O-tables.Source: Own estimates and simulation, see text.IAB 57

6.4 ConclusionsIn this section we addressed the trade and sectoral effects of labour mobility within aCGE-model. As our results show, countries are reacting very differently to the laboursupply shock: In Germany, exports are affected nearly twice as much from the migrationshock than imports, while in the UK these differences are much smaller. This reflects thedifferent trade structure and the different degree of openness of the two countries.However, differences occur also among sending countries: In Hungary and Slovenia weobserve only a slight departure from the pre-shock sectoral production structure, while inSlovakia we observe a strong sectoral redistribution of factors.Nevertheless, our results predict moderate effects of migration on wages and unemploymenton the aggregate level. In brief, our results can be summarised as follows:First, migration effects are mitigated in case of a partial adjustment of the capital stock.Second, we observe strong trade effects which mitigate the migration shock and fosterthe redistribution of factors among tradable goods and non-tradable goods in somecountries. Third, a redistribution of factors leads in Slovakia to differences in thedistribution of production among sectors between simulated and initial values.IAB 58

7 ReferencesAydemir, Abdurrahman and George J. Borjas. 2006. “A Comparative Analysis of theLabor Market Impact of International Migration: Canada, Mexico, and the United States.”National Bureau of Economic Research, Inc, NBER Working Papers: 12327.Baas, Timo and Herbert Brücker. 2008. "Macroeconomic impact of Eastern Enlargementon Germany and UK: evidence from a CGE model." Applied Economics Letters,forthcomming.Baldwin, Richard E., Joseph F. Francois and Richard Portes. 1997. “The Costs andBenefits of Eastern Enlargement: The Impact on the EU and Central Europe.” EconomicPolicy: A European Forum, 12:24, pp. 125-70.Barrell, Ray, John FitzGerald and Rebecca Riley. 2007. “EU Enlargement and Migration:Assessing the Macroeconomic Impacts.” National Institute for Economic and SocialResearch Discussion paper no. 292, London.Barrett, A. and D. Duffy. 2008. “Are Ireland’s Immigrants Integrating into its LabourMarket?”, International Migration Review, 42(3), forthcoming.Bentolila, Samuel, Juan Dolado, and Juan Jimeno. 2007. "Does Immigration Affect thePhillips Curve? Some Evidence for Spain." Kiel Institute for the World Economy: Kiel.Blanchflower, David G. and Andrew J. Oswald. 1994. The wage curve: Cambridge andLondon:MIT Press.Blanchflower, David G. and Andrew J. Oswald. 1995. "An Introduction to the WageCurve." Journal of Economic Perspectives, 9:3, pp. 153-67.Blanchflower, David G. and Andrew J. Oswald. 2005. "The Wage Curve Reloaded."National Bureau of Economic Research, Inc, NBER Working Papers: 11338.Boeri, Tito and Herbert Brücker. 2005. "Why are Europeans so tough on migrants?"Economic Policy, 20:44, pp. 629-703.Borjas, George J. 2003. "The Labor Demand Curve is Downward Sloping: Reexaminingthe Impact of Immigration on the Labor Market." Quarterly Journal of Economics, 118:4,pp. 1335-74.Brown, Drusilla K., Alan V. Deardorff, Simeon D. Djankov and Robert M. Stern. 1995. “AnEconomic Assessment of the Integration of Czechoslovakia, Hungary and Poland in theEuropean Union.” Working Paper from the American Institute for Contemporary GermanStudies.Brücker, Herbert. (2007). “Labor Mobility after the European Union's Eastern Enlargement:Who Wins, Who Loses?” GMF Paper Series.IAB 59

Brücker, Herbert and Elke J. Jahn. 2008. "Migration and the Wage Curve: A StructuralApproach to Measure the Wage and Employment Effects of Migration." Institute for theStudy of Labor (IZA).Brücker, Herbert and Michael Kohlhaas. 2004. “Dynamische Effekte der Migration undZuwanderungsbedarf in Teilarbeitsmärkten.“ Zuwanderungsbeirat der DeutschenBundesregierung.Card, David. 1995. "The Wage Curve: A Review." Journal of Economic Literature, 33:2,pp. 785-99.Card, David and Thomas Lemieux. 2001. "Can Falling Supply Explain the Rising Return toCollege For Younger Men? A Cohort_Based Analysis." Quarterly Journal of Economics,116:2, pp. 705-46.D'Amuri, Francesco, Gianmarco I. P. Ottaviano, and Giovanni Peri. 2008. "The LaborMarket Impact of Immigration in Western Germany in the 1990's." National Bureau ofEconomic Research, Inc.Dervis, Kemal and Sherman Robinson. 1982. "A General Equilibrium Analysis of theCauses of a Foreign Exchange Crisis: The Case of Turkey." Weltwirtschaftliches Archiv,118:2, pp. 259-80.Dustmann, Christian, Francesca Fabbri and Ian Preston. 2005. “The Impact ofImmigration on the British Labour Market.” Economic Journal, 115, pp. F324-F341.Dustmann, Christian and Albrecht Glitz. 2005. “Immigration, Jobs, and Wages: Theory,Evidence and Opinion.” mimeo, CEPR and CREAM.Felbermayr, G., W. Geis, and W. Kohler. 2008. "Absorbing German Immigration: Wagesand Unemployment." mimeo. Carls-Universität: Tübingen.Friedberg, Rachel M. and Jennifer Hunt. 1995. “The Impact of Immigrants on HostCountry Wages, Employment and Growth.” Journal of Economic Perspectives, 9:2, pp.23-44.Guichard, Stephanie and Jean-Pierre Laffargue. 2000. "The Wage Curve: The Lessons ofan Estimation over a Panel of Countries." CEPII research center, Working Papers.Hamilton, Bob and John Whalley. 1984. "Efficiency and Distributional Implications ofGlobal Restrictions on Labour Mobility." Journal of Development Economics, 14:1/2, pp.61.Heijdra, Ben J., Christian Keuschnigg and Wilhelm Kohler. 2002. “Eastern Enlargement ofthe EU: Jobs, Investment and Welfare in Present Member Countries.” CESifo GmbH,CESifo Working Paper Series: CESifo Working Paper No. 718.IAB 60

Kaldor, Nicholas. 1961. "Capital Accumulation and Economic Growth," in The Theory ofCapital. F.A. Lutz and D.C. Hague eds. New York: St. Martins.Keuschnigg, Christian and Wilhelm Kohler. 1999. “Eastern Enlargement to the EU:Economic Costs and Benefits for the EU Present Member States? The Case of Austria.Study for the European Commission. Final Report.” Johannes Kepler Universität: Linz.Layard, Richard and Stephen Nickell. 1986. "Unemployment in Britain." Economica,53:210, pp. S121-S69.Layard, Richard, Stephen Nickell, and Richard Jackman. 1991. Unemployment:Macroeconomic performance and the labour market: Oxford; New York; Toronto andMelbourne:Oxford University Press.Lemos, S. and J. Portes. 2008. “New Labour? The Impact of Migration from the NewMember States from Central and Eastern European Countries on the UK Labour Market.”University of Leicester WP 08/29, Leicester, UK.Levine, Paul. 1999. "The welfare economics of immigration control." Journal of PopulationEconomics, 12:1, pp. 23.Lindbeck, Assar. 1993. Unemployment and Macroeconomics. Cambridge, MA: MIT Press.Longhi, Simonetta, Peter Nijkamp, and Jacques Poot. 2005. "A Meta-Analytic Assessmentof the Effect of Immigration on Wages." Journal of Economic Surveys, 19:3, pp. 451-77.Longhi, Simonetta, Peter Nijkamp and Jaques Poot. 2006. “The Impact of Immigration onEmployment of Natives in Regional Labor Markets: A Meta-Analysis.” IZA DiscussionPaper 2044, IZA: Bonn.Manacorda, M., Alan Manning, and Jonathan Wadsworth. 2006. "The Impact ofImmigration on the Structure of Male Wages: Theory and Evidence from Britain." Centrefor Economic Performance, LSE, CEP Discussion Papers.Ottaviano, Gianmarco I. P. and Giovanni Peri. 2006. "Rethinking the Effects ofImmigration on Wages." National Bureau of Economic Research, Inc, NBER WorkingPapers: 12497.Shapiro, Carl and Joseph E. Stiglitz. 1984. "Equilibrium Unemployment as a WorkerDiscipline Device." American Economic Review, 74:3, pp. 433.Venables, Anthony J. 1999. "Regional Integration Agreements: A Force for Convergenceor Divergence?" The World Bank, Policy Research Working Paper Series: 2260.Wong, Kar-yiu. 1995. International trade in goods and factor mobility: Cambridge andLondon: MIT Press.IAB 61

8 Appendix8.1 Appendix ATable A1: The short-run effects of transitional arrangements and the free movementof workers from the NMS-8 on the structure of wages and unemployment, 2008-2011WagesUnemploymentAllLowskilledHighskilledAllLowskilledMediumskilledMediumskilledHighskilledChanges in per cent (unemployment rate: changes in percentage points)AT -0.02 -0.02 -0.02 -0.02 0.02 0.03 0.01 0.07BE 0.00 0.00 0.00 0.00 -0.01 -0.01 -0.01 0.00DE -0.08 -0.08 -0.08 -0.08 0.08 0.12 0.07 0.08DK 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00ES 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00FI -0.03 -0.03 -0.03 -0.04 0.03 0.04 0.04 0.02FR 0.00 0.00 0.00 0.00 0.01 0.01 0.01 0.00GR -0.02 -0.02 -0.03 -0.01 0.02 0.01 0.02 0.01IE 0.44 0.47 0.50 0.38 -0.23 -0.36 -0.23 -0.09IT -0.01 -0.01 -0.01 -0.01 0.00 0.01 0.00 0.00LU 0.05 0.03 0.03 0.12 -0.02 -0.01 -0.01 -0.09NL 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00SE -0.01 -0.01 -0.01 -0.01 0.01 0.01 0.01 0.01UK 0.07 0.08 0.08 0.05 -0.05 -0.07 -0.06 -0.01CZ 0.08 0.08 0.07 0.11 -0.06 -0.37 -0.04 -0.01EE 0.09 0.10 0.09 0.09 -0.07 -0.12 -0.07 -0.03HU 0.08 0.07 0.08 0.09 -0.03 -0.08 -0.03 -0.01LT 0.03 0.03 0.03 0.04 -0.03 -0.07 -0.03 -0.02LV 0.04 0.04 0.04 0.04 -0.03 -0.05 -0.03 -0.01PL 0.00 0.00 0.00 0.00 -0.01 -0.01 -0.01 0.00SI 0.09 0.13 0.09 0.08 -0.05 -0.10 -0.04 -0.02SK -0.03 -0.03 -0.03 -0.04 0.04 0.12 0.04 0.02EU-15 1) 0.00 0.01 -0.01 -0.01 0.01 0.01 0.02 0.02NMS-8 0.04 0.04 0.03 0.05 -0.01 -0.04 -0.01 0.00Total 0.00 0.01 -0.01 0.00 0.01 0.00 0.01 0.021) Without Portugal.Source: Own estimates and simulation, see text.IAB 62

Table A2: The short-run effects of transitional arrangements and the free movementof workers from the NMS-2 on the structure of wages and unemployment, 2008-2014WagesUnemploymentAllLowskilledHighskilledAllLowskilledMediumskilledMediumskilledHighskilledChanges in per cent (unemployment rate: changes in percentage points)AT -0.02 -0.02 -0.01 -0.03 0.03 0.05 0.00 0.14BE 0.01 0.01 0.01 0.02 -0.02 -0.05 -0.01 -0.02DE -0.06 -0.06 -0.06 -0.06 0.06 0.09 0.05 0.05DK -0.01 -0.01 -0.01 -0.01 0.01 0.00 0.00 0.02ES 0.12 0.11 0.36 0.01 -0.16 -0.15 -0.34 -0.01FI 0.00 0.00 0.00 -0.01 0.00 0.01 0.00 0.01FR 0.00 0.00 0.01 0.00 -0.01 -0.01 -0.01 -0.01GR -0.09 -0.12 -0.10 -0.05 0.07 0.08 0.08 0.04IE -0.24 -0.23 -0.25 -0.23 0.12 0.18 0.13 0.06IT -0.04 -0.04 -0.04 -0.03 0.03 0.03 0.04 0.02LU -0.01 -0.01 -0.01 -0.01 0.00 0.00 0.01 0.01NL 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00SE 0.00 -0.01 0.00 0.00 0.01 0.01 0.00 0.00UK 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00BG 0.17 0.16 0.15 0.19 -0.21 -0.39 -0.16 -0.13RO 0.04 0.04 0.04 0.05 -0.03 -0.03 -0.03 -0.03EU-15 1) -0.01 0.00 -0.01 -0.02 0.00 -0.01 0.00 0.01NMS-2 0.07 0.05 0.06 0.10 -0.08 -0.13 -0.07 -0.08Total -0.01 0.00 -0.01 -0.02 -0.01 -0.02 0.00 0.011) Without Portugal.Source: Own estimates and simulation, see text.IAB 63

8.2 Appendix BIn this Appendix the key equations of the theoretical CGE model are explained. TheAppendix is divided in six sections which describe the modeling of production,consumption, trade, the income of households, the government and the equilibriumconditions.The domestic productionThe production in the model is organized by activities. These activities use labour andcapital on the one hand and intermediaries on the other hand to produce final output.The production function is therefore nested. The upper nest describes the combination ofvalue added and intermediaries, while the lower nest describes the production of valueadded by the combination of labour and capital. If there are different kinds of labour orcapital, the combination of each type to composite labour or composite capital is done inthe lowest nest.The production of value added is described by a CES production function. The factors ofQF,production are imperfect substitutes while the variable f acomposite labour reflecting lower nests.can be either labour or(B.1)QVAα⎛=α ∑δ⎝vava va −ρaa ⎜ faQFf,af∈F⎞⎟⎠1vaρawheref∈ F factorfis element of the set of factorsQVAavalue added in quantity unitsQF f , afactor demand by activities avaa aefficiency parameter of the CES value added functionvaδ f , ashare parameter of factor f in activity avaρ aexponent of the CES value added functionFactor demand is derived according to the profit maximation hypothesis. Every factor isused up to the quantity where it’s marginal return equates marginal costs.IAB 64

(B.2)⎛⎞WF WFDIST PVA( 1 tva ) QVA ∑ QF QF⎝⎠vavava −ρava −ρa−1f f , a= −a ⎜ δf , a f , a ⎟ δf , a f , af∈F'−1wheretva aPVAaWF fvalue added tax for activity aprice of value addedprice of factor fWFDIST f , adistribution factor for wages of factor f in activity . (exogen)The upper nest of the production function combines intermediaries and value added.According to different production structures along activities, different productionfunctions,CES or Leontief have to be used in this nest.The intermediary goods demanded by each activities are in turn produced by differentactivities. Therefore demand of intermediaries is demand to a composite good producedby different activities. The share of each product in this set is determined according tocost minimization and therefore relative prices.Extra and Intra-EU TradeWhether a final product is consumed domestically or exported into another EU-country oroutside the EU is determined by profit maximization. Therefore a CET function is used toallocate production to domestic use or Intra-EU and Extra-EU exports.(B.3)( ( ) ) 1ρc t ρc 1t δctt t tQX = α δ QE −δQDc c c c c cwithQX cQDcQE cquantity of the production of good cquantity of production sold domesticallyquantity of production exported ctα cdisplacement parameter of the CET functiontδ cshare parameter of the CET functionIAB 65

tρ cexponent of the CET functionImports are treated similar to exports. The quantity of imports is determined by a CETtypeArmington function. Additionally, imports and domestic products are only imperfectsubstitutes. This reduces the impact of world prices on domestic prices and consumption.(B.4)( ( ) ) 1−ρc q −ρc 1q ρcqq q qQQ = α δ QM + −δQDc c c c c cwithQQcQMcdomestic supplyquantity of importsqα cshift parameter of the Armington functionqδ cshare parameter of the Armington functionqρ cexponent of the Armington funktionThe income of nongovernmental institutionsNongovernmental Institutions receive wages and capital income from their factorendowments. Additionally they receive transfers from the state or other domestic orforeign nongovernmental institutions. These earnings are spent for consumption,savings, taxes, or transfers.(B.5)∑∑YI = YIF + TRII + transfr CPI + transfr EXR + transf EXREUi i ,' i i , gov i , row euf∈ F i'= INSDNG'withYI iIncome of Institution iYIF i , fIncome of institution i from factor fTRII ii ,'transfers from institution i to instirtution i'shif i , ftffshare of income from factor f by domestic nongovernmental institutionsdirect tax on factor fIAB 66

transfr i , govtransfers from government to institution itransfr irow ,transf eutransfers from ROW-countries to institution itransfers from EU-countries to institution iCPIconsumer price indexIncome from labour is divided in earned income and unemployment benefits. Therelationship between unemployment and wages is specified by a wage curve. Thereforelabour market rigidities in different countries can be considered.The domestic consumptionThe domestic demand is divided into household consumption and investment demand ofenterprises. Since investment demand is equal to household savings, it is discussed inthe equilibrium section.The consumption of households is a function of disposable income and is derived from aStone-Geary demand function:(B.6)⎛PQ QH = PQ γ + β EH −∑PQ γ −∑∑PXACγ⎝m m, , , ⎜m hc c h c c h c h h c c', h a, c' a, c',hc'∈C a∈Ac'∈C⎞⎟⎠withQH ch ,consumption of good c by household hmγ ch ,consumption of home produced good c by household hmβ ch ,household h marginal share of consumption expenditure for good cThe household maximizes a Stone-Geary utility function with regard to her budgetconstraints.The governmentThe State in the model is financed by taxes, customs duties, credit, and transfers byother institutions. The income of the state is spent for consumption of goods,investment, transfers, and savings.IAB 67

(B.7)∑ ∑ ∑YG = TINS YI + tf YF + tva PVA QVAi i f f a a ai∈INSDNG f∈F a∈A∑ ∑ ∑+ ta PA QA + tm pwm QM EXR + te pwe QE EXRa a a c c c c c ca∈A c∈CM c∈CE∑∑+ tq PQ QQ YIF + trnsf EXRc c c gov gov,rowc∈C f∈FwithYGstate incomeThe equilibrium conditionsThe model is closed by solving five equilibrium conditions, market clearing on factor andgoods markets, an even balance of payments, a balanced budget of the state sector andsaving equal investment. The goods markets are in equilibrium if supply equals demand,while the governmental sector is in equilibrium if income equals spending. Thereforegovernmental savings have to be flexible. The third equilibrium condition is savinginvestmentequilibrium, where savings have to equal investment.Forth, the factor market reach equilibrium if supply of a factor meets its demand. Thesupply of factors is exogenous.(B.8)∑a∈AQFf , a= QFSwithQFS fquantity of factor fIn the labour markets, supply of labour is additionally restricted by labour marketrigidities. Therefore a wage-curve describes the unemployment rate at a specific wagelevel.The equilibrium of the balance of payments is solved separately for Intra-EU and Extra-EU trade, reflecting quasi fixed exchange rates in the EU.(B.9)∑ ∑ ∑ ∑pmrw QMRW + trnsfr = perw QERW + trnsfr + FSAV + FSAVc c rdw, f c c i,rdw eu rdwc∈CM f ∈F c∈CE i∈INSDwithFSAV rdw Foreign savings (Extra-EU) in foreign currency unitsIAB 68

FSAV eu Foreign savings (Intra-EU) in foreign currency unitsIAB 69

8.3 Appendix CFigure C 1: Social accounting matrixExpendituresReceipts Activities Commodities Factors Housholds Enterprises Government Savings/InvestmentRest of WorldTotalActivities MarketedoutputsHomeconsumedoutputsActivityincomeCommoditiesIntermediateInputsTransaktionskostenPrivateconsumptionGovernmentconsumptionInvestment Exports DemandFactors Value-added Factor incomefrom RoWFactor incomeHousholds Factor income tohousholdsTransfers tohousholdsSurplus tohousholdsTransfers tohousholdsTransfers tohousholdsHousholdsincomeEnterprises Factor income toenterprisesTransfers toenterprisesTransfers toenterprisesEnterpriseincomeGovernmentProducer taxes,value added taxVAT,Factor income togovernment,Income taxesSurplus togovernment,direct taxTransfers toGovernmentGovernmentincomeSavings/InvestmentSavings ofHousholdsSavings ofenterprisesGovernmentsavingsRoW savingsSavingsRest of World Imports Surplus toRoWTransfers toRoWForeignexchangeoutflowTotal Activity SupplyexpendituresFactorexpendituresHousholdexpendituresEnterpriseexpendituresGovernmentexpendituresInvestmentForeignexchange inflowSource: IFPRI (2002)IAB 70

European Integration ConsortiumIAB, CMR, fRDB, GEP, WIFO, wiiwLabour mobility within the EU in the context of enlargement and the functioningof the transitional arrangementsVC/2007/0293Deliverable 5Fondazione Rodolfo DeBenedetti (fRDB)The impact of labour mobility on public finances and social cohesionTito Boeri and Paola MontiAbstractMost Europeans fear that immigrants are a fiscal burden via their disproportionateparticipation in the generous social welfare programmes of the EU. However, to datethere has not been any systematic attempt to evaluate the net contribution of migrantsto welfare systems. The purpose of this background note is to contribute to filling thisgap.We first draw on data from the EU-SILC surveys to understand whether migrants to EUcountries tend to be disproportionately represented in the population of beneficiaries ofsocial transfers. Descriptive statistics point to marked differences – with respect tonatives – in migrants’ access to contributory and non-contributory benefits; migrantsappear to be under-represented among the recipients of contributory benefits, while theopposite is true for non-contributory allowances, such as social assistance and housingbenefits. This contributes to explain the widespread perception of welfare abuse.However, once we control for confounding factors which are likely to correlate withmigrant status and influence the likelihood of receiving non-contributory benefits, (i.e.,relevant individual and household characteristics), we find that migrant status has little –if any – impact on the likelihood of being a recipient of social welfare benefits. Next wecarry out a (static) analysis of the net fiscal position of natives’ and migrants’ householdwith respect to the government; adding up how much they contribute to the state budgetvia payroll and income taxes and how much they draw from it in terms of access to avariety of welfare programmes. Our analysis suggests that migrants pay lower taxes thannatives, and yet a significant portion of migrants are net-contributors to the state budget.Finally, we analyse the determinants of public opinion perceptions about immigrants inthe EU countries; in particular, we analyse whether negative perceptions about migrationare stronger in countries with a more generous redistributive system, or adopting specificmigration policies. This analysis sheds light on the optimal interaction of social andmigration policies that could cope with the concerns of public opinion with respect tomigration from both New Member States and non-EU countries.The views and opinions expressed in this publication are those of the authors and do not necessarilyrepresent those of the European Commission.

Contents1 The impact on public finances.............................................................................. 11.1 Introduction .............................................................................................. 12 Migrants representation in the welfare state .......................................................... 22.1 Residual Welfare Dependency of Migrants...................................................... 32.1.1 Contributory benefits...................................................................... 42.1.2 Non-contributory allowances............................................................ 62.2 Residual Welfare Dependency of Migrants...................................................... 92.2.1 Contributory benefits.................................................................... 102.2.2 Non-contributory allowances.......................................................... 133 The net fiscal position of migrants...................................................................... 163.1.1 Taxes ......................................................................................... 173.1.2 Net Balance with respect to the state budget ................................... 183.1.3 Residual net dependency .............................................................. 214 Perceptions ..................................................................................................... 224.1 Crime ..................................................................................................... 224.2 Fiscal Contribution.................................................................................... 265 Conclusions and policy implications .................................................................... 316 References ...................................................................................................... 337 Annex............................................................................................................. 36

1 The impact on public finances1.1 IntroductionThe participation of immigrants in welfare programmes has been widely investigated byempirical research in Europe and the US. Several studies document that the immigrantpopulation is over-represented among the pool of welfare recipients, notably socialassistance recipients, although for some European countries – but not all - thosedifferences can be explained by differences in socio-economic characteristics (Barrett andMcCarthy 2008, Boeri 2006, Hansen and Lofstrom 1999) As these programmes are noncontributory,a relatively high take-up rate may generate pressures on the fiscal budget,although the overall fiscal impact of immigration is still uncertain (Auerbach andOreopoulos 1999). However, social assistance is often the main source of income forthose migrants who are not entitled to contributory benefits (e.g. unemploymentbenefits) since they have not accumulated enough work experience in the host country tobe eligible for social insurance.In this note, we provide a first overview of the available evidence as to whether migrantsto EU member states are more than proportionally represented among the recipients ofwelfare benefits. We employ the recently released survey on European Union Statistics onIncome and Living Conditions, EU-SILC 2004-2006. Regrettably, the EU-SILC databaseprovides an extremely rough division of immigrants with respect to their citizenship orcountry of birth, distinguishing only between immigrants from EU-25 countries and fromthird countries. This implies that it is not possible to use this otherwise rich data sourceto conduct analysis that focus on immigrants from the New Member States (NMS); stillwe argue that such an aggregate analysis can be informative about the phenomenon ofinterest given the remarkable similarities of migrants from NMS and other Europeanmigrants, especially when considering their human capital characteristics, as depicted inDeliverable 2. Furthermore, we will focus our attention particularly on those countrieswhose European migrants come prevalently from the NMS, like Greece (53% of other EU-25-citizen living there come from NMS), Finland (47%), Iceland (46%), Ireland (44%) orAustria (36%). 1The rather detailed coverage of the take-up of welfare programmes offered by the EU-SILC enables us to evaluate the number of non-citizens benefiting from a certain kind ofgovernment transfer as well as the amount received, in order to assess the extent of theirparticipation to social assistance schemes. We then make a comparison with the nativepopulation in order to consider whether migrants are more or less dependent on statetransfers and evaluate their relative pressure on the public budget. In the next step, wetest the presence of migrant residual dependency, that is to say, whether controlling forpersonal characteristics the migrant status directly affects the probability of receiving a1 See Table A1 in the appendix.fRDB 1

particular kind of transfer. We employ a standard probit regression model for this purpose(Section 2).Throughout the analysis we always disentangle contributory from non-contributorytransfers: the first group includes individual benefits designed to cover against the risksof unemployment, longevity (interacted with labour market risk), sickness, disability aswell as survival to the death of the main breadwinner (survivor’s pension); the secondgroup is household-related and it includes housing and family allowances, as well astransfers targeted specifically to groups with a higher risk of social exclusion. 2Furthermore we will distinguish between migrants coming from other countries within theEU-25 or outside of this union; 3 in the case of households we keep an eye also on mixedcouples, where only the household head or his/her spouse is a native citizen.We then move on to estimate the contribution of migrants to the state budget throughtheir taxes and the mandatory social security contributions paid by both employees andemployers. After estimating this quantity, we subtract from it the amount of transfersreceived by the household as a whole, in order to compute the net fiscal position ofmigrants relative to native citizens. A descriptive analysis of the data is, once more,followed by a more in-depth, multivariate analysis of the residual impact of the status ofmigrant on the household’s net position with respect to the state budget (Section 3).Next, we will focus on public opinion perceptions of migrants, drawing on a number ofvery specific questions raised in the European Social Survey (ESS) in 2002. In particular,we will compare across various European countries the opinions about migration and itseffect on crime and the state budget. As a further step, we will examine the link betweenthese perceptions and the individual-specific characteristic of the respondents, in order toidentify the profile of the average citizen concerned with migration related issues. Finally,for different EU countries we will compare the average stance on migration matters withthe generosity of their welfare state as well as crime rates (Section 4).Finally, Section 5 offers a summary of our findings and provides some policyrecommendations.2 Migrants representation in the welfare stateThe EU-SILC database enables us to distinguish migrants from the EU-25 countries andfrom third countries. As a consequence, we will always share migrants into these twodifferent groups. However, when looking at migrant households in stead of individuals, a2 These categories correspond rather closely to the definitions of the EU-SILC; groups at risk of “socialexclusion may be identified (among others) as destitute people, migrants, refugees, drug addicts,alcoholics, victims of criminal violence”. For more details, see the SILC User Database Variable Description(epunet.essex.ac.uk/EU-SILC_UDB.pdf).3 For Estonia, Germany, Latvia and Slovenia even this rough classification is unfeasible, as this distinction isnot provided by the dataset.fRDB 2

third category will be added: mixed households (i.e. households involving at least anative and a migrant).For purposes of cross-country comparability, we consider the difference in the share ofwelfare recipients between migrants and native population in the host country, i.e.:[1]RMM i−iRNNWhere R M represents welfare recipients among the i group of migrants (where i could bewhether a migrant from the EU-25 or a migrant from outside the EU-25 4 ), M i the totalnumber of migrants belonging to the specific group i, R N native welfare recipients, and Nnatives. Thus, a positive number indicates an over-representation of migrants in welfareschemes, since the percentage of recipient in the migrant population is higher than theshare of welfare recipients in the native population. A negative number points instead toan under-representation of migrants.Secondly, we will also compare the nominal amount of transfer received, on average, bynatives and by migrants in the three years going from 2004 to 2006, as explained inequation [2]:[2]BMM i−iBNNWhere B M represents the total amount of welfare benefits received by i-migrants, B N thetotal amount of welfare benefits received by natives. The difference between these twoquantities will tell us how much more or less the average household can rely on,depending on its migrant status.These two measures enable us to look both at the numbers (how many migrants benefitfrom the welfare with respect to natives) and the quantities (how much they receive, onaverage, with respect to natives).2.1 Residual Welfare Dependency of MigrantsThe EU-SILC survey offers a detailed source of information on participation to welfareprograms. This allows us to evaluate both the number of migrants benefiting fromspecific government transfers as well as the amount received, and make a comparisonwith the native population. In order to carry out such estimates, we distinguish between4 As previously explained, categories will become three when looking at migrant households (in stead ofindividuals): migrant households from the EU25 countries, migrant households from non-EU25 countries,and mixed households (at least one migrant and one native).fRDB 3

two main categories of state transfers: contributory benefits and non-contributoryallowances.2.1.1 Contributory benefitsWe begin with the analysis of contributory benefits. Since the main condition ofentitlement for contributory benefits is that the claimant must have paid sufficientpersonal insurance contributions, the unit of analysis for this category of benefits is theindividual, independently of the existence of other household members. Our preliminaryresults are displayed in Table 1a. As it can be seen, apart from a few countries – notablyDenmark, Finland, Lithuania and Slovakia – migrants are under-represented amongrecipients of contributory benefits in most countries.The results are similar when, rather than considering take-up rates, we consider theaverage difference in benefits received by migrants with respect to natives (Table 1b). Asalready observed, such a difference is, in most cases, negative. In particular, this isalways true for EU-15 countries, while evidence is more mixed for the NMS.fRDB 4

Table 1a. Contributory Benefits: migrant under-representationCountryEU-25 immigrantsExtra EU-25 immigrantsAll immigrantsEU-15 Austria -0.10 [5.67]*** -0.14 [12.55]***Belgium -0.02 [2.37]** -0.13 [9.10]***Denmark 0.04 [1.91]* 0.05 [3.77]***Finland -0.03 [1.28] 0.08 [4.69]***France -0.01 [0.44] -0.09 [8.69]***Germany + -0.08 [5.86]***Greece -0.19 [7.50]*** -0.25 [22.71]***Ireland -0.14 [11.54]*** -0.25 [13.62]***Italy -0.17 [7.96]*** -0.19 [24.76]***Luxembourg -0.18 [34.54]*** -0.24 [18.95]***Netherlands -0.06 [1.63] -0.17 [3.65]***Portugal -0.12 [3.24]*** -0.28 [15.24]***Spain -0.07 [2.00]** -0.22 [14.38]***Sweden -0.08 [5.04]*** -0.17 [10.51]***United Kingdom -0.01 [0.81] -0.24 [23.39]***New MemberStatesOther CountriesCyprus -0.05 [3.92]*** -0.24 [19.39]***Czech Republic 0.05 [1.05] -0.37 [9.78]***Estonia + 0.06 [8.91]***Hungary -0.25 [6.35]*** -0.34 [5.71]***Latvia + 0.11 [13.43]***Lithuania 0.06 [0.91] 0.08 [3.01]***Poland -0.03 [0.38] -0.19 [3.78]***Slovakia 0.18 [3.68]*** -0.06 [0.65]Slovenia ++ 0.10 [15.40]***Iceland -0.09 [3.27]*** -0.04 [7.65]***Norway -0.07 [4.10]*** -0.13 [7.64]***Notes: averages over the available years; t statistics in brackets, ***,** and * denote significance at 1, 5 and 10 percent respectively; +the EU-SILC does not distinguish between EU-25 and extra-EU25; ++ migrants identified by country of birth; the EU-SILC does notdistinguish between EU-25 and extra EU-25 migrants.Source: own calculations on data from EU-SILC 2004-2006.fRDB 5

Table 1b. Difference in average benefits receivedCountryEU-25 immigrantsExtra EU-25 immigrantsAll immigrantsEU-15 Austria -2,152 [197.29]*** -3,288 [522.39]***Belgium -520 [105.21]*** -1,833 [279.64]***Denmark -195 [10.09]*** -1,182 [91.48]***Finland -1,424 [63.97]*** -1,919 [117.02]***France -1,040 [278.06]*** -2,274 [720.17]***Germany + -1,675 [679.30]***Greece -163 [19.94]*** -1,844 [524.54]***Ireland -1,426 [173.19]*** -1,922 [165.71]***Italy -1,967 [245.00]*** -3,254 [1317.72]***Luxembourg -4,901 [230.47]*** -6,074 [118.46]***Netherlands -1,831 [65.18]*** -3,723 [123.12]***Portugal -548 [54.89]*** -1,469 [352.86]***Spain -304 [31.49]*** -1,865 [457.92]***Sweden -1,197 [158.50]*** -2,214 [292.27]***United Kingdom -402 [85.88]*** -2,636 [1026.91]***New MemberStatesOther CountriesCyprus -86 [7.19]*** -1,592 [123.65]***Czech Republic 37 [8.83]*** -877 [285.47]***Estonia + 92 [89.95]***Hungary -588 [128.04]*** -884 [123.39]***Latvia + 141 [199.44]***Lithuania 39 [6.30]*** 315 [121.18]***Poland 350 [50.43]*** -628 [150.41]***Slovakia 347 [60.44]*** -40 [4.28]***Slovenia ++ 434 [89.41]***Iceland -2,455 [33.53]*** -1,366 [74.14]***Norway -402 [85.88]*** -2,636 [1026.91]***Notes: figures are in euros, averages over the available years; t statistics in brackets, ***,** and * denote significance at 1, 5 and 10percent respectively; + the EU-SILC does not distinguish between EU-25 and extra-EU25; ++ migrants identified by country of birth; theEU-SILC does not distinguish between EU-25 and extra EU-25 migrants.Source: own calculations on data from EU-SILC 2004-2006.Summarising, whatever measure we take, take-up rates or difference in average socialspending, when we look at unconditional distributions (not controlling for individuals’characteristics), migrants appear to be under-represented among the recipients ofcontributory transfers.This result is hardly surprising as contributory benefits typically require a minimumvesting period for eligibility. Many migrants may have not yet accumulated sufficientlylong contributions to be eligible for these transfers. Furthermore, migrants are muchyounger than natives who are therefore more represented in the population ofpensioners.2.1.2 Non-contributory allowancesWe now move to non-contributory allowances. The unit of analysis here is not theindividual, but the household. This is partly due to the way EU-SILC provide variables onfamily allowances, partly because most non-contributory benefits – funded from generalfRDB 6

taxation – are not related to insurance contributions but to specific household needs orcircumstances (support careers, educational allowances, child benefits, etc.) andtherefore they have to be considered on a household basis.When we focus on non-contributory allowances, the picture changes. A significantdifference between non-EU-25 migrants and migrants from the EU-25 emerges. In mostcountries, EU-25 migrants are under-represented also among recipients of noncontributorybenefits. Interestingly, among them there are several countries where theshare of NMS migrants in the population of EU-25 migrants is relatively large (Greece,Austria, Ireland, Italy). Unfortunately, limited data availability – the EU-SILC data do notdisentangle migrants from the NMS from migrants from the EU-15 countries – does notallow for further investigations focusing only on migrants from the New Member States.A different picture emerges for migrant households from outside Europe. In most EU-15countries those migrants’ households are indeed over-represented as welfare recipientsand seem therefore to be more dependent on social assistance than the average nativehousehold (Table 2a). In those countries, not only non-EU-25 migrants are more likely toreceive non-contributory allowances, but the average subsidy is generally higher than fornatives (Table 2b). The relative size of allowances transferred to the households is shownin Table 2b. This suggests that migrants from outside Europe receive, on average, morethan natives almost everywhere in EU-15 and by a significant margin in the Nordics.In NMS non-EU-25 migrant households are generally equally represented amongrecipients of non-contributory transfers than natives. The average subsidies are,however, lower than those of natives.There are also a few countries where data do not allow distinguishing EU-25 migrantsfrom non-EU migrants (this is the case of Germany, Estonia, Latvia and Slovenia).Finally, mixed households (involving at least a native and a migrant) are overrepresentedboth in terms of take-up rates and average social spending per household inall countries (except Greece and Cyprus, see Table 2a).fRDB 7

Table 2a. Non-contributory Allowances: non-EU migrant households over-representationCountryEU-25 immigranthouseholdsExtra EU-25 immigranthouseholdsImmigranthouseholdsMixed householdsEU-15 Austria -0.04 [1.23] 0.13 [6.60]*** 0.23 [8.84]***Belgium -0.02 [0.85] 0.10 [3.59]*** 0.25 [10.40]***Denmark 0.05 [1.06] 0.28 [10.85]*** 0.15 [6.44]***Finland -0.08 [1.69]* 0.36 [10.22]*** 0.22 [7.49]***France 0.02 [0.78] 0.43 [25.04]*** 0.20 [8.74]***Germany + 0.10 [3.16]*** 0.15 [6.31]***Greece -0.07 [1.85]* -0.08 [5.03]*** -0.04 [1.53]Ireland -0.10 [4.36]*** -0.02 [0.73] 0.00 [0.17]Italy -0.24 [6.27]*** 0.05 [4.28]*** 0.13 [6.06]***Luxembourg 0.17 [18.28]*** 0.37 [15.30]*** 0.14 [6.97]***Netherlands -0.04 [0.43] 0.41 [2.62]*** 0.14 [2.91]***Portugal -0.27 [3.80]*** -0.05 [1.30] 0.25 [4.91]***Spain -0.03 [1.02] 0.04 [2.60]*** 0.05 [2.06]**Sweden -0.11 [3.51]*** 0.19 [5.63]*** 0.18 [7.25]***United Kingdom 0.00 [0.10] 0.00 [0.13] 0.10 [4.12]***New MemberStatesOther CountriesCyprus -0.38 [15.29]*** -0.45 [14.13]*** -0.08 [3.44]***Czech Republic -0.25 [3.57]*** -0.07 [1.11] 0.18 [2.89]***Estonia + -0.09 [7.04]*** 0.13 [4.60]***Hungary -0.11 [1.29] -0.13 [1.21] 0.23 [2.72]***Latvia + -0.09 [5.97]*** 0.03 [1.27]Lithuania -0.24 [1.48] -0.14 [1.86]* -0.01 [0.14]Poland -0.21 [1.51] -0.12 [0.85] 0.17 [2.18]**Slovakia -0.08 [0.62] -0.01 [0.04] -0.08 [0.89]Slovenia ++ 0.09 [5.90]*** 0.04 [1.88]*Iceland -0.01 [0.14] 0.09 [4.67]*** 0.02 [1.63]Norway -0.03 [0.84] 0.24 [6.93]*** 0.15 [5.82]***Notes: averages over the available years; t statistics in brackets, ***,** and * denote significance at 1, 5 and 10 percent respectively; +the EU-SILC does not distinguish between EU-25 and extra-EU25; ++ migrants identified by country of birth; the EU-SILC does notdistinguish between EU-25 and extra EU-25 migrants.Source: own calculations on data from EU-SILC 2004-2006.fRDB 8

Table 2b. Difference in average allowances receivedCountryEU-25 immigranthouseholdsExtra EU25 immigranthouseholdsImmigranthouseholdsMixed householdsEU-15 Austria -493 [60.49]*** 707 [153.67]*** 1,736 [355.62]***Belgium -282 [82.14]*** 1,736 [354.54]*** 727 [208.75]***Denmark 76 [8.75]*** 1,555 [280.31]*** 559 [112.11]***Finland 361 [18.11]*** 4,209 [294.16]*** 2,052 [196.74]***France -198 [85.01]*** 2,918 [1517.96]*** 1,735 [923.60]***Germany + 1,057 [383.68]*** 1,649 [885.74]***Greece 27 [9.88]*** -99 [91.07]*** -91 [62.22]***Ireland -95 [7.97]*** 3,326 [231.05]*** 596 [57.42]***Italy -304 [94.91]*** 131 [150.16]*** 366 [213.12]***Luxembourg 1,481 [97.95]*** 3,124 [67.44]*** 1,020 [38.62]***Netherlands 1,494 [56.97]*** 4,357 [123.07]*** 43 [4.09]***Portugal -185 [33.35]*** 32 [15.81]*** 337 [156.19]***Spain -92 [26.28]*** -35 [24.12]*** 311 [151.65]***Sweden -393 [51.65]*** 3,501 [440.42]*** 1,982 [360.35]***United Kingdom 1,197 [166.36]*** 885 [257.33]*** 979 [265.98]***New MemberStatesOther CountriesCyprus -945 [68.66]*** -837 [49.33]*** -106 [9.46]***Czech Republic -312 [76.39]*** -166 [53.73]*** 240 [84.33]***Estonia + -98 [88.44]*** 41 [23.61]***Hungary -47 [8.59]*** -68 [9.28]*** 409 [89.12]***Latvia + -72 [93.20]*** 20 [23.67]***Lithuania -145 [14.23]*** -85 [16.70]*** 18 [7.10]***Poland -123 [38.24]*** -57 [19.97]*** 80 [46.27]***Slovakia 290 [42.29]*** -146 [6.38]*** 126 [35.20]***Slovenia ++ -4 [0.79] 148 [27.82]***Iceland -654 [8.81]*** -480 [13.68]*** 739 [45.86]***Norway -260 [18.62]*** 7,011 [480.58]*** 1,954 [168.16]***Notes: figures are in euros, averages over the available years; t statistics in brackets, ***,** and * denote significance at 1, 5 and 10percent respectively; + the EU-SILC does not distinguish between EU-25 and extra-EU25; ++ migrants identified by country of birth; theEU-SILC does not distinguish between EU-25 and extra EU-25 migrants.Source: own calculations on data from EU-SILC 2004-2006.Thus, migrants in EU-15 – especially from outside EU-25 – would seem to have mainlyaccess to non-contributory schemes of social assistance, probably because socialassistance and housing benefits target the poorest fraction of the population regardless oftheir contribution history. On the contrary, migrants in the NMS, whatever origin theyhave, are under-represented among recipients of both contributory and non-contributorybenefits,2.2 Residual Welfare Dependency of MigrantsThe over-representation of migrants among recipients of non-contributory transfers maybe due to the fact that migrants have more children and lower educational attainmentsthan natives or to a higher dependency from transfers also compared to natives with thesame observable characteristics (the so-called residual welfare dependency). In order todisentangle the two effects, we resort to a multivariate analysis framework, runningfRDB 9

country-specific probit models of the probability of receiving social assistance, controllingfor some observable characteristics, and including among the regressors a dichotomousvariable describing migrant status. A positive and statistically significant coefficient forthis variable is informative as to the presence of welfare dependency among migrants.2.2.1 Contributory benefitsIn our first set of regressions, the dependent variable is a dummy equal to 1 when theindividual receives any type of contributory benefits and 0 otherwise. We control forindividual- and community-level characteristics, 5 and we add separate dummies formigrants coming from the EU-25 and those coming from non-EU countries.As expected, we find that having a low personal income before transfers increasessignificantly the probability of receiving some kind of benefit. Unsurprisingly, singles withchildren are also more likely to receive transfers, while house-owners, and persons withhigher education, are less likely to receive a contributory transfer. Other things beingequal, men are more dependent on welfare than women. 6As far as the migrant dummy is concerned, Table 3 confirms our preliminary results 7 :with a very few exceptions migrants are equally likely or less likely than natives toreceive contributory transfers. This is the case of Germany and Estonia, two countrieswhose data do not allow separating EU-25 from non-EU-25 migrants. Another countrywhere this happens is Denmark.Germany – in spite of data limitation – is a very interesting case, because our findingsare at odds with those of earlier literature on the subject. As an example, Barrett (2008)states that, although unadjusted data show higher use of welfare by immigrants, inGermany this difference can be explained by controlling for observable characteristics.5 We control for sex, age and age square, education, income, number of children, size of the household,whether the house is of property, the density of the neighbouring area, dummies for different regions(NUTS2) and years.6 See Table A2 in the appendix.7 For ease of exposition, we reported in the Table only the change in the estimated probability induced by ashift of the migrant dummy variables from 0 to 1; for the whole regression results, see Table A2 in theappendix.fRDB 10

Table 3. Change in the probability of receiving contributory benefits due to migrant statusMigrant dummiesCountryEU-25Extra EU-25All countriesObs.EU-15 Austria -0.082 [3.21]*** -0.011 [0.68] 41,843Belgium -0.052 [4.03]*** -0.200 [12.39]*** 40,460Denmark 0.010 [0.31] 0.074 [3.81]*** 48,740Finland -0.110 [3.01]*** 0.020 [0.76] 90,745France -0.063 [3.72]*** -0.109 [7.29]*** 76,103Germany + 0.048 [2.37]** 75,937Greece -0.046 [1.39] -0.081 [4.84]*** 51,344Ireland -0.125 [8.44]*** -0.180 [8.13]*** 46,340Italy -0.107 [3.53]*** -0.007 [0.52] 192,440Luxembourg -0.040 [4.15]*** -0.103 [5.61]*** 30,476Netherlands 0.004 [0.08] -0.128 [1.83]* 17,750Portugal -0.123 [2.63]*** -0.116 [4.02]*** 43,240Spain -0.032 [1.81]* -0.096 [5.60]*** 119,170Sweden -0.180 [7.81]*** -0.245 [11.65]*** 47,573United Kingdom 0.004 [0.15] -0.141 [7.98]*** 58,626New MemberStatesOther CountriesCyprus -0.031 [2.20]** -0.137 [6.75]*** 26,751Czech Republic 0.044 [0.72] -0.275 [4.21]*** 32,112Estonia + 0.049 [4.23]*** 41,102Hungary -0.210 [3.57]*** -0.402 [5.35]*** 46,059Latvia + -0.034 [2.69]*** 24,893Lithuania -0.157 [2.35]** 0.042 [0.99] 30,049Poland -0.180 [2.03]** -0.229 [3.90]*** 110,235Slovakia 0.122 [2.08]** -0.227 [3.35]*** 38,388Slovenia ++ 0.009 [1.00] 74,347Iceland -0.023 [0.63] -0.025 [2.64]*** 26,488Norway -0.038 [1.45] -0.201 [6.79]*** 47,259Notes: z statistics in brackets, ***,** and * denote significance at 1, 5 and 10 percent respectively; + the EU-SILC does not distinguishbetween EU-25 and extra EU-25; ++ migrants identified by country of birth; the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants.Source: own elaborations on data from EU-SILC 2004-2006.Pushing our analysis a step further, we run a separate regression on the probability ofreceiving every single type of benefit that we have detailed information about, notablyunemployment, old-age, survivors’, sickness and disability benefits. The results aresummarised in Figure 1.fRDB 11

Figure 1. Change in probability of receiving contributory benefits due to migrant statusEU25 migrants0.020-0.02-0.04-0.06-0.08-0.1-0.12-0.14AT FI FR GR IE IT UKbenefits unempl old-age survivors sicknerss disabilityExtra EU25 migrants0.150.10.050-0.05AT FI FR GR IE IT UK-0.1-0.15-0.2benefits unempl old-age survivors sicknerss disabilityNotes: only significant coefficients of migrant dummies displayed; the first column reports coefficients from Table 3. Forexpositional ease, we report in the figure the change in the estimated probability induced by a shift of the migrant dummyvariable from 0 to 1; when the estimated effect lacks statistical significance at conventional confidence level, we don’t report it;thus, Figure 1 succinctly provides information of the size and significance of the estimated effect for the variable of interest.* The EU-SILC does not distinguish between EU-25 and extra EU-25 migrants.** Migrants identified by country of birth; the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants.Source: own elaborations (or calculations) on EU-SILC data 2004-2006Even when considering separately each single contributory social programme, we obtainthe same result: migrants do not display a significant welfare dependency, especiallythose who come from other EU countries.fRDB 12

The only important exception is represented by unemployment benefits in Denmark(which are largely non-contributory), Finland and Germany and by sickness benefits inDenmark, Estonia and Slovakia, for which we observe residual dependency of non-EUmigrants.2.2.2 Non-contributory allowancesWe carry out the same analysis for non-contributory allowances.In this case both the household size and the number of children increase significantly theprobability of receipt. Also an older head of the household and a low level of educationincrease the probability of receiving the transfer. 8Table 4. Change in the probability of receiving non-contributory benefits due to migrant statusMigrant household dummiesCountryEU-25 Extra EU-25 All countriesMixedObs.EU-15 Austria -0.023 [0.53] -0.073 [3.33]*** 0.002 [0.07] 17,470Belgium -0.046 [2.37]** 0.097 [2.90]*** 0.037 [2.10]** 17,744Denmark 0.005 [0.06] 0.067 [1.42] 0.060 [2.61]*** 21,054Finland -0.141 [2.00]** 0.162 [2.62]*** -0.005 [0.14] 37,252France 0.034 [1.35] 0.295 [10.13]*** 0.130 [6.41]*** 32,679Germany + 0.179 [3.73]*** 0.032 [1.29] 30,168Greece -0.055 [1.66]* -0.059 [3.84]*** 0.009 [0.42] 19,620Ireland -0.168 [5.21]*** -0.038 [0.80] 0.069 [2.96]*** 18,797Italy -0.154 [1.68]* -0.017 [1.19] 0.058 [2.45]** 75,098Luxembourg 0.053 [2.62]*** 0.090 [1.49] 0.025 [0.96] 12,661Netherlands 0.061 [0.56] 0.421 [2.86]*** -0.013 [0.26] 9,234Portugal -0.177 [1.50] -0.205 [6.15]*** 0.123 [2.82]*** 15,208Spain -0.054 [3.76]*** -0.018 [2.17]** 0.004 [0.29] 44,184Sweden -0.184 [3.90]*** 0.035 [0.70] 0.059 [2.38]** 20,326United Kingdom -0.060 [0.95] -0.229 [9.64]*** -0.014 [0.59] 23,329New MemberStatesCyprus -0.391 [11.63]*** -0.506 [11.00]*** -0.115 [4.74]*** 9,191Czech Republic -0.261 [6.74]*** -0.222 [4.34]*** -0.014 [0.25] 13,005Estonia + -0.068 [3.68]*** 0.046 [2.04]** 13,991Hungary -0.123 [1.55] -0.258 [2.39]** 0.248 [3.03]*** 15,576Latvia + -0.024 [1.16] -0.010 [0.45] 7,699Lithuania +++ -0.173 [2.21]** 0.054 [0.89] 9,123Poland 0.009 [0.06] -0.171 [1.93]* 0.016 [0.30] 32,536Slovakia -0.022 [0.21] 0.291 [1.49] -0.049 [0.78] 11,856Slovenia ++ 0.006 [0.27] 0.083 [5.55]*** 19,612IcelandOther Countries-0.232 [2.80]*** -0.047 [1.11] -0.020 [0.58] 9,919Norway -0.150 [3.84]*** 0.106 [1.81]* 0.101 [3.93]*** 20,164Notes: z statistics in brackets, ***,** and * denote significance at 1, 5 and 10 percent respectively; + the EU-SILC does not distinguishbetween EU-25 and extra EU-25; ++ migrants identified by country of birth; the EU-SILC does not distinguish between EU-25 and extraEU-25 migrants; +++ migrant households from EU-25 countries excluded from the estimation because of their limited number.Source: own elaborations on data from EU-SILC 2004-2006.8 See Table A3 in the appendix.fRDB 13

Our multivariate analysis supports the view that EU-25 migrants are less welfaredependent then natives. In fact, even controlling for observable characteristics, theyseem to be equally or less likely than natives to receive non-contributory benefits.Moreover, regression results support what we previously observed for countries whose EUmigrants come prevalently from the New Member States. They all show a negative(Finland, Greece, Ireland) or non-significant (Austria) correlation between the (EU-25)migrant status and the probability of receiving transfers.Importantly, our estimates also suggest that being a migrant household from outsideEurope does not explain benefit receipt in a large number of countries. Put it anotherway, the over-representation of non-EU migrants in the pool of welfare recipients is inseveral countries explained by observable characteristics making them more eligiblerather than by a residual dependency effect.There are, however, important exceptions in countries with a rather generous welfareprogramme in place: in Belgium, Finland, France, Germany, the Netherlands and Norwaysome residual dependency of non-EU migrants is observed. Those findings are quiteconsistent with previous studies on Nordic countries, typically characterized by generouswelfare systems. Sweden is, however, an important exception. Although not showingevidence of migrants’ welfare dependency in the present work, earlier literature onSweden usually found that differences in welfare participation in that country cannot beexplained by observable socio-economic characteristics (Hansen, 1999). As alreadymentioned, Germany is another case, where our results are different from those of theearlier literature (Barrett, 2008).Mixed households are, in most countries, over-represented even after controlling for theirobservable characteristics. A number of explanations can be possibly provided for thisresult, such as better access to information on welfare programmes, assortative matingand household formation influenced by the welfare access opportunities.Also in this case, we run separate probit regressions for housing allowances, familyrelatedtransfers and subsidies targeting specific marginal groups. The results aresummarised in Figure 2.fRDB 14

Figure 2. Change in the probability of receiving non-contributory benefits due to migrant statusEU25 migrant households0.10.050-0.05AT FI FR GR IE IT UK-0.1-0.15-0.2-0.25Allowances family social exclusion housingExtra EU25 migrant households0.40.30.20.10-0.1AT FI FR GR IE IT UK-0.2-0.3allowences family social exclusion housingfRDB 15

Mixed households0.140.120.10.080.060.040.020-0.02AT DE FI FR GR IE IT UK-0.04-0.06Allowances family social exclusion housingNotes: For expositional ease, we report in the figure the change in the estimated probability induced by a shift of the migrantdummy variable from 0 to 1; when the estimated effect lacks statistical significance at conventional confidence level, we don’treport it; thus, Figure 1 succinctly provides information of the size and significance of the estimated effect for the variable ofinterest. The first column reports coefficients from Table 4;* The EU-SILC does not distinguish between EU-25 and extra EU-25 migrants.** Migrants identified by country of birth; the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants.Source: own elaborations (or calculations) on EU-SILC data 2004-2006Our analysis suggests that residual welfare dependency concerns only non-EU migrantsand mixed households. Housing benefits in France are disproportionately targeted to non-EU migrants, perhaps because of their segregation in villes nouvelles and peripheralareas, where massive public housing schemes have been implemented. In NordicCountries there is also residual welfare dependency of non-EU migrants, notably in familyallowances and housing benefits.3 The net fiscal position of migrantsIn this section we evaluate the net fiscal position of migrant, non-migrant and mixedhouseholds in a static sense, that is, we consider only the difference between the currentcontributions and taxes paid by each household member (and her/his employer) and thecurrent amount of transfers received by the state in terms of social programmes. Noconsideration is made of the lifetime contributions and benefits paid/received by thedifferent households.fRDB 16

As the EU-SILC did not report gross-wages and taxes for Greece, Italy, Latvia andPortugal, 9 these countries had to be dropped from our analysis. Moreover, the EU-SILCdoes not provide information on employers’ social security contributions; 10 thus, weimputed these contributions by applying the rules as detailed in the OECD publication“Taxing Wages” (editions 2003/2004 to 2005/2006). The latter provides a routine foreach country belonging to the OECD that can be used to calculate the average employers’social security contributions, conditional on gross-wages. This means that we also had todrop non-OECD EU countries, such as Cyprus, Estonia, Lithuania and Slovenia.3.1.1 TaxesTable 5 suggests that in the EU-15 migrants, on average, contribute less to tax revenuesand social security contributions than natives. This result is hardly surprising as taxes aretypically progressive and social security contribution proportional to earnings andmigrants are generally concentrated at the low end of the income (and earning)distribution.Mixed household are, once more, an important exception. They pay, on average, moretaxes than native citizens.9 This information is available for Spain only for 2006.10 The EU-SILC committee decided that this information must be provided from MS only from 2007 onwards.fRDB 17

Table 5. Difference in average taxes paid: migrant households lower participation to the statebudgetEquation [2]GroupCountryEU-25 migranthouseholdExtra EU-25migrant householdMixed householdEU-15 Austria -3159 [73.98]*** -2402 [99.68]*** 2550 [99.66]***Belgium -3918 [50.29]*** 18983 [156.01]*** -2805 [40.06]***Denmark -9549 [177.99]*** -7264 [212.63]*** 6456 [210.32]***Finland -7176 [62.49]*** -10834 [131.58]*** 1227 [20.44]***France -2304 [149.65]*** -8911 [700.86]*** 3249 [261.45]***Germany+ -5166 [290.28]*** 3031 [252.00]***Ireland -429 [9.51]*** -4312 [79.25]*** 4795 [122.23]***Luxembourg 1155 [19.70]*** -12669 [70.53]*** 5521 [53.93]***Netherlands -4102 [40.65]*** -5483 [40.25]*** 3295 [81.77]***Spain -1526 [43.67]*** -194 [13.46]*** 1766 [85.85]***Sweden -7012 [135.72]*** -15041 [278.89]*** 2790 [74.75]***United Kingdom -5405 [170.05]*** 1828 [120.29]*** 4539 [279.01]***New MemberStatesOther CountriesCzech Republic 451 [14.27]*** 1096 [45.88]*** 1184 [53.94]***Hungary -8 [0.22] -1105 [22.78]*** 1558 [51.17]***Poland -1081 [35.13]*** -123 [4.50]*** 978 [59.32]***Slovakia -2427 [59.96]*** -969 [7.17]*** 594 [28.01]***Iceland -8723 [28.16]*** 12938 [88.41]*** 7479 [111.17]***Norway 422 [7.18]*** -13592 [221.59]*** 10178 [208.30]***Notes: z statistics in brackets, * significant at 10 per cent, ** significant at 5 per cent, *** significantat 1 per cent; + the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants.Source: EU-SILC 2004-2006.3.1.2 Net Balance with respect to the state budgetWe now turn our attention to the net balance between, on the one hand, taxes andcontributions paid and, on the other hand, state transfers received. We consider first netcontributors to the state budget. Table 6a shows the difference between the share of netcontributors among migrants and the share of net contributors in the native population.fRDB 18

Table 6a. Relative share of net contributory: migrant households over-representationGroupCountryEU-25 migranthouseholdEquation [1]Extra-EU25migranthouseholdMixed householdsEU-15 Austria 0.05 [1.53] 0.19 [9.36]*** 0.22 [8.70]***Belgium -0.02 [0.89] -0.15 [5.46]*** 0.12 [5.00]***Denmark -0.26 [6.36]*** -0.14 [5.99]*** 0.10 [4.83]***Finland -0.11 [2.50]** -0.39 [11.87]*** 0.11 [4.17]***France 0.00 [0.11] -0.17 [9.80]*** 0.11 [4.75]***Germany + -0.11 [3.60]***0.11 [4.66]***Ireland 0.10 [3.74]*** 0.10 [3.24]*** 0.16 [5.56]***Luxembourg 0.23 [25.36]*** -0.03 [1.32] 0.16 [7.69]***Netherlands -0.05 [0.51] 0.22 [1.51] 0.23 [5.18]***Spain 0.06 [0.94] 0.34 [12.06]*** 0.21 [4.50]***Sweden -0.07 [2.24]** -0.26 [8.70]*** 0.05 [2.14]**United Kingdom -0.10 [2.25]** 0.16 [7.65]*** 0.13 [5.01]***New MemberStatesCzech Republic 0.19 [2.49]** 0.28 [4.15]*** 0.20 [3.07]***Hungary 0.27 [3.28]*** 0.40 [3.83]*** 0.26 [3.04]***Poland -0.18 [1.18] 0.10 [0.64] 0.13 [5.72]***Slovakia -0.36 [2.93]*** -0.11 [0.31] -0.04 [0.46]Other Countries Iceland 0.17 [3.73]*** 0.16 [9.58]*** 0.05 [4.17]***Source: EU-SILC 2004-2006.Norway 0.05 [1.84]* -0.32 [10.02]*** 0.13 [5.72]***Notes: the difference between the share of net contributors among migrants and the share of netcontributors in the native population, as in Equation [1]; t statistics in brackets, * significant at 10 percent, ** significant at 5 per cent, *** significant at 1 per cent; averages over the available years; + theEU-SILC does not distinguish between EU-25 and extra EU-25 migrants.Unexpectedly, it would seem that in a number of countries migrant households are atleast as equally represented in the group of net-contributors to the state budget.Remarkable examples are Ireland for both EU-25 and non-EU-25 migrants, Austria, Spainand the UK in case of non-EU migrants.Next we consider the net balance of migrant and non-migrant households. Table 6bdisplays the difference between the average balance between taxes and transfers fornative and migrant households.fRDB 19

Table 6b. Difference in average position with the government: migrant households lowerparticipation to the state budgetEquation [2]GroupCountryEU-25 migranthouseholdExtra-EU-25migrant householdMixed householdsEU-15 Austria 1100 [19.31]*** 2815 [87.59]*** 6265 [183.60]***Belgium -2559 [32.15]*** 20327 [163.48]*** -2734 [38.22]***Denmark -9060 [125.55]*** -9752 [212.23]*** 7177 [173.81]***Finland -7830 [55.83]*** -13641 [135.65]*** 2386 [32.55]***France -107 [5.22]*** -8013 [471.83]*** 5134 [309.33]***Germany+ -3159 [128.66]*** 4356 [262.47]***Ireland 3350 [56.06]*** -3703 [51.43]*** 6232 [120.05]***Luxembourg 9022 [105.91]*** -4708 [18.04]*** 10767 [72.40]***Netherlands -2850 [20.90]*** -3739 [20.31]*** 5812 [106.75]***Spain -852 [16.98]*** 4193 [202.88]*** 4601 [155.69]***Sweden -4434 [73.15]*** -14875 [235.11]*** 2587 [59.09]***United Kingdom -6842 [167.47]*** 5987 [306.53]*** 7028 [336.09]***New MemberStatesCzech Republic 2082 [52.22]*** 3144 [104.37]*** 1525 [55.08]***Hungary 888 [20.32]*** 769 [13.13]*** 2308 [62.83]***Poland -1682 [40.52]*** 1632 [44.24]*** 1446 [65.04]***Slovakia -3082 [57.50]*** -463 [2.58]*** 741 [26.39]***Other Countries Iceland -3395 [8.33]*** 15545 [80.76]*** 7466 [84.37]***Norway 4610 [58.79]*** -13660 [166.84]*** 12561 [192.59]***Notes: z statistics in brackets, * significant at 10 per cent, ** significant at 5 per cent, *** significant at1 per cent; + the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants.Source: own elaborations on data from EU-SILC 2004-2006.While we find once more a net contribution of non-EU migrants in the UK, Spain (wherethere are many first-generation migrants) and Austria, the migrants to the Nordiccountries contribute significantly less than what they receive from the state budget(roughly between 10 and 15,000 Euros per year).The fact that migrants receive more than what they pay is consistent with theprogressiveness of taxes in the EU and the distributive goals of European welfare states.Significantly, migrants are over-represented in the population of net contributors to thestate budget, but those receiving more than what they contribute apparently receivesignificantly more than what they pay into the system.fRDB 20

3.1.3 Residual net dependencyAs in the case of benefit receipt, it is important to consider if a negative net fiscal positionof migrants survives to a control of their personal characteristics. Thus, we runregressions in which the dependent variable is the household net-position with respect tothe government. The full results of our estimations are shown in the appendix (Table A4),while Table 7 reports the coefficients of the dummy variables associated to EU and thirdcountrymigrants or mixed households.Table 7. Incidence of migrant status on the net-position with the governmentGroup Country Migrant dummiesObsEU-25 migrant Extra EU-25 migrant Mixed householdhouseholdhouseholdAll 2462.53 [6.35]*** 2014.28 [2.71]*** 3569.92 [9.23]*** 335868EU-15 Austria 1066.6 [0.80] 1914.76 [3.97]*** 3582.3 [4.77]*** 17475Belgium 3809.16 [2.49]** 53182.25 [1.89]* -756.34 [0.96] 10823Germany+ 746.16 [0.89] 546.81 [0.55] 30173Denmark -461.78 [0.54] -4312.91 [6.30]*** 794.38 [1.03] 21096Spain 688.76 [0.31] 2147.91 [3.80]*** 1767.21 [1.44] 12146Finland 16821.53 [1.43] -4790.31 [5.86]*** -959.6 [1.67]* 37267France 419.56 [0.68] -242.97 [0.45] 2080.39 [3.04]*** 32687Ireland 1852.13 [1.71]* -4162.06 [3.35]*** -1547.57 [1.66]* 18815Luxembourg 3353.49 [4.80]*** 3727.97 [2.42]** 2798.16 [2.30]** 12663Sweden -837.71 [0.93] -1180.55 [0.86] -1124.11 [1.36] 20360United Kingdom 921.52 [0.63] 3942.71 [2.81]*** 2850.88 [2.53]** 20030New MemberStatesCzech Republic 766.27 [0.52] 243.87 [0.37] 326.59 [0.38] 12247Hungary -3003.7 [3.45]*** 918.12 [1.36] -356.84 [0.55] 15579Poland 1871.31 [2.23]** -932.38 [1.14] 4077.46 [1.86]* 32536Slovakia 426.66 [0.36] 116.23 [0.17] 222.71 [0.66] 11875Other Countries Iceland 126.61 [0.09] -2806.24 [2.31]** -2093.21 [1.83]* 9919Norway 2936.72 [2.28]** -1210.19 [1.16] 1340.34 [1.88]* 20177Notes: t statistics in brackets, * significant at 10 per cent, ** significant at 5 per cent, significant at 1per cent;+ theEU-SILC does not distinguish between EU-25 and extra EU-25 migrants.Source: own elaborations on data from EU-SILC 2004-2006.Table 7 suggests that in most countries the documented net negative position ofmigrants with respect to the state budget is explained by their personal characteristics.In other words, there is no residual net dependency once we control for education,number of children and the other relevant covariates affecting the probability of receivingsocial transfers, as discussed in Section 2. This result is particularly clear for EU migrants,since the dummy coefficients are almost all non significant and even positive.If we look at the non-EU migrants, however, there is still a minority of countries(Denmark, Finland, Ireland and Iceland) where migrants are residually dependent,notably they receive more than what they contribute even when account is made for theirpersonal characteristics.fRDB 21

Finally, mixed households tend to maintain a non-negative position with respect to thestate budget, despite some exceptions (Finland, Ireland and Iceland).4 PerceptionsAfter having looked at the fiscal position of migrant households, we now want to see howEuropean citizens perceive migration and its impact on the nation wellbeing. Theiropinions will be assessed by drawing on results from the European Social Surveys (ESS),a public opinion survey carried out in many EU countries, 11 which includes a number ofquestions about immigration. We will focus on the perceptions about the influence ofmigration on crime problems, as well as opinions on the impact of foreign guests on thefiscal balance of the recipient country. Although the ESS does not have a longitudinaldesign, access to micro data from the survey enables us to control for individualcharacteristics, as well as cyclical factors potentially affecting individual perceptions.4.1 CrimeThe typical profile of a criminal is a young male with low education and experiencingfinancial difficulties, mainly associated with low incomes or long-term unemployment(Freeman, 1991; Levitt, 1998; Grogger, 1998). At the same time, the first settlement ofnew migrants are usually carried out by young males, and those with low education andincome constraints tend to receive a higher attention from the general public, probablydue to the stark differences from the native population. Public opinion may combinethese two phenomena of marginalisation and mentally associate migration withcriminality even when it is faulty (see Buonanno et al., 2008).Figure 3 shows the opinions of EU citizens with respect to the contribution offered bymigration to crime rates. There is some cross-country variation, but almost 70% of therespondents believe that crime problems are made worse by migrants. 12 The average EUcitizen does associate migration with higher crime rates.11 Austria, Belgium, Czech Republic , Germany, Denmark, Spain, Finland, France, United Kingdom, Greece,Hungary, Ireland, Italy, Luxemburg, Netherlands, Norway, Poland, Portugal, Sweden, Slovenia.12 The respondents were asked: “Are [country’s] crime problems made worse or better by people coming tolive here from other countries?” and were allowed to give a grade from 0 (being: Crime problems made worst)to 10 (being: Crime problems made better).fRDB 22

Figure 3. “Are national crime problems made worse or better by people coming to live herefrom other countries?” % of responses0.700.600.500.400.300.200.100.00worse a little worse neutral a little better betterEU IT AT DE FI UK GR IENotes: own elaboration on data from ESS-2002; EU-average showed in the first column;Answers regrouped as follows: 0-2, worse; 3-4, little worse; 6-7, little better; 8-10, better.We are interested in unfolding which social and economic factors might influence theseperceptions. In order to attain this, we run an OLS regression of the probability of statingthat migrants contribute to crime rates controlling for observables characteristics ofrespondents, as well as self-reported involvement in humanitarian organisations orfriendship with migrants. 13 Cross-country variation is taken into account by includingcountry dummies in the regressors. The results are displayed in Table 8.13 We control for sex, age and age square, education and labour status of respondent, also adding a dummyfor migrants, one for high and one for medium total household income. Furthermore we include dummiesfor declared domicile description (big city, suburbs or outskirts of a big city, town or small village, countryvillage), a dichotomy variable equal to one if the respondent is friend or works with a migrant, and anotherone for having taken part to a humanitarian organisation. We also introduce a dummy for political views(left of right) and variable controlling for the declared time spent watching the TV, listening to the radio andreading newspapers (from 0: no time at all, to 7: more than 3 hours a day). Finally, we control for thefeeling of safety when walking alone at home in the dark (1: very safe; 4: very unsafe) and for thosewhose household members were victim of burglary/assault in the last 5 years.fRDB 23

Table 8. Migrants and crime problems, incidence of personal characteristicsInfluence of immigration on crime(1) (2) (3) (4)MaleAgeAge square-0.09 -0.18 -0.17 -0.17[4.66]*** [8.26]*** [8.30]*** [8.38]***-0.02 -0.03 -0.03 -0.03[8.50]*** [9.01]*** [9.66]*** [9.73]***0.0001 0.0001 0.0002 0.0001[5.57]*** [7.26]*** [7.59]*** [7.58]***0.09 0.08 0.06 0.05Secondary EducationTertiary EducationUnemployedMigrantMedium IncomeHigh Incomecitysuburbstowncountry villageKnow an immigrantHumanitarianorganisationsUnsafe walking aloneVictim of crimeLeftRightTVRadioNewspaperConstant[2.86]*** [2.15]** [2.00]** [1.56]0.55 0.43 0.43 0.39[15.12]*** [10.90]*** [11.56]*** [10.44]***-0.14 -0.13 -0.11 -0.10[3.09]*** [2.61]*** [2.41]** [2.25]**0.71 0.60 0.66 0.65[12.48]*** [9.74]*** [11.47]*** [11.41]***0.08 0.06 0.05 0.05[2.81]*** [2.07]** [1.87]* [1.77]*0.12 0.09 0.07 0.06[4.01]*** [2.76]*** [2.18]** [1.87]*0.27 0.31 0.30 0.30[5.69]*** [6.05]*** [6.07]*** [6.20]***0.15 0.15 0.15 0.17[3.09]*** [2.99]*** [3.18]*** [3.46]***0.11 0.13 0.13 0.14[2.52]** [2.67]*** [2.82]*** [3.05]***0.05 0.04 0.04 0.05[1.20] [0.92] [1.00] [1.20]0.32 0.32 0.31[13.34]*** [14.25]*** [13.89]***0.26 0.25 0.24[9.05]*** [8.86]*** [8.60]***-0.20 -0.19 -0.19[12.92]*** [13.35]*** [13.23]***-0.13 -0.13 -0.12[4.88]*** [5.15]*** [5.02]***0.21 0.21[8.36]*** [8.38]***-0.11 -0.11[4.58]*** [4.67]***-0.02[3.70]***-0.01[2.05]**0.02[2.58]***3.30 3.86 3.50 3.60[35.88]*** [36.41]*** [35.71]*** [35.69]***Country Dummies yes yes yes yesObservations 40291 35408 39882 39683R-squared 0.01 0.01 0.01 0.01Notes: robust t statistics in brackets, * significant at 10%, ** significant at 5%, *** significant at1%; dependent variable on a scale from 0, worse, to 10, better, depending on answer to thequestion: “Are [country]’s crime problems made worse or better by people coming to live herefrom other countries?”. Reference person: female living in farm or countryside with lower thansecondary education. Country of reference: IcelandSource: ESS-2002.fRDB 24

Other things being equal, young males with lower education and living in rural areas aremore concerned about the association between crime rates and migration. Furthermore,the higher education and family income, the less migrants are perceived as a threat tothe security. The opposite holds for people identifying themselves as belonging to theright of the political spectrum or reporting feelings unsafe while walking back home aloneor who have been victim of an assault.Tertiary education has broadly the same effect on opinion, as well as being a migrant.Moreover, more frequent personal contacts with migrants – through work, volunteering,friendship or simply by living in a highly populated area – are associated with lessconcerns about the contribution of migrants to crime rates.Finally, it seems that media (TV or radio) exposure increases negative perceptions ofmigrants. Next, we consider the relationship between perceptions and measured crimerates, across countries both unconditionally and controlling for the characteristics ofrespondents (in which case we report country dummies).Figure 4. Crime and migration perceptions.21.5IE.15SEIEFRIT1.5SISEFIDKFRITESGBPTPL.1SIDKFILUBENLDEGBES PTHUPL0-.5NOBECH DE NLLUHUCZ.05NOCHCZ.5 1 1.5 2 2.5Total Incarcerated (every 100 persons)Crime problems made better(Percentage of respondents)-1.5 1 1.5 2 2.5Total Incarcerated (every 100 persons)Regression country dummiesNotes: % of people answering strictly less than 5 to the ESS-2002 question about migration and crime.Regression dummies from column (1) Table 8; The choice of the specification does not affect the regression results..Source: share of incarcerated taken from the “United Nations Survey on Crime Trends and the Operations of Criminal JusticeSystems (7th and 8th)”, year 2002fRDB 25

As pointed out by Figure 4, there is not a clear relationship between perceptions andmeasured crime rates. If anything, the relationship is mildly negative.4.2 Fiscal ContributionIn order to evaluate public opinion perceptions about the migrants’ contribution to thestate budget, we rely on the ESS, whose 2002 wave contained a very specific question onthe topic: “Most people who come to live here work and pay taxes. They also use healthand welfare services. On balance, do you think people who come here take out more thanthey put in or put in more than they take out?.”Let’s first take a look at the rough data, as presented in Figure 5.Figure 5. “Do you think migrants take out more than they put in or put in more than they takeout?” % of responses0.450.40.350.30.250.20.150.10.050take out more take out a little more balance put in a little more put in moreEU IT AT DE FI UK GR IENotes: Aggregation of data from ESS-2002. EU-average showed in the first columnAnswers from 0 to 10 regrouped as follows: 0-2, take out more; 3-4, take out a little more; 5, balance; 6-7, put in a little more; 8-10, put in more.Even if, on average, European citizens believe that migrants balance out the resourcesthey receive with the taxes they pay, the distribution of answers is quite skewed to theleft. In other words, when comparing those respondents who take a stance, the majoritybelieves that migrants are a burden rather than an asset for public finances. Indeed,more than one European out of four (28%) believes that migrants balance out theiraccount with the government, but 22 and 23% respectively believes that they take outmore or a little more than what they contribute to. However there are relevantdifferences across countries: while 43% of Slovenians believe that foreigners depend onthe welfare state, almost one Italian out of five believes that they contribute more thanwhat they receive.fRDB 26

Once more, we consider conditional perceptions, controlling for individualcharacteristics. 14Table 9. Migrants and welfare state, incidence of personal characteristicsDo migrants put in more than what they take out?(1) (2) (3) (4)MaleAgeAge squareSecondary EducationTertiary EducationUnemployedMigrantMedium IncomeHigh Incomecitysuburbstowncountry villageKnow an immigrantHumanitarianorganisationsLeftRightTVRadioNewspaperConstant0.00 0.00 0.00 0.00[1.86]* [1.39] [1.58] [1.17]-0.02 -0.02 -0.02 -0.02[5.13]*** [5.45]*** [6.00]*** [5.85]***0.00 0.00 0.00 0.00[3.24]*** [4.66]*** [4.97]*** [4.74]***0.01 0.01 0.01 0.01[4.97]*** [3.59]*** [3.89]*** [3.46]***0.05 0.04 0.04 0.03[16.62]*** [12.55]*** [12.91]*** [11.79]***-0.22 -0.21 -0.18 -0.18[4.32]*** [3.99]*** [3.69]*** [3.51]***0.07 0.06 0.07 0.07[22.04]*** [17.48]*** [20.93]*** [21.07]***0.01 0.00 0.01 0.01[3.53]*** [2.30]** [2.57]** [2.53]**0.01 0.01 0.01 0.01[4.08]*** [2.46]** [2.46]** [2.21]**0.02 0.01 0.01 0.01[4.91]*** [2.44]** [2.70]*** [2.63]***0.01 -0.02 0.00 0.00[2.06]** [0.38] [0.03] [0.02]0.01 0.00 0.00 0.00[2.17]** [0.35] [0.73] [0.81]-0.05 -0.11 -0.09 -0.09[1.12] [2.10]** [1.93]* [1.86]*0.03 0.03 0.03[16.52]*** [17.88]*** [17.64]***0.03 0.03 0.02[11.78]*** [12.09]*** [11.58]***0.02 0.02[9.55]*** [9.44]***-0.09 -0.10[3.55]*** [3.67]***-0.02[3.66]***-0.01[2.44]**0.00[3.57]***0.19 0.20 0.17 0.18[42.05]*** [39.97]*** [39.26]*** [39.02]***Country Dummies yes yes yes yesObservations 39138 34966 39138 38943R-squared 0.01 0.01 0.01 0.01Notes: robust t statistics in brackets, * significant at 10%, ** significant at 5%, *** significant at 1%;dependent variable on a scale from 0, worse, to 10, better, depending on answer to the question:“Taxes and services: immigrants take out more than they put in or less?”; Reference person: femaleliving in farm or countryside with lower than secondary education; Country of reference: Iceland.Source: ESS-2002.14 We use the same set of control variables as in Table 4, but those variables related to feeling of safety andcriminal victim. See note above.fRDB 27

The typical profile of persons concerned about the net fiscal position of migrants is similarto that of those concerned about crime rates: it is mainly young people, unemployed,living in rural areas, with low levels of literacy and income sources, right-winged when itcomes to politics, with little or no contact with the world of migration, who fear about thefiscal burden of migrants. Media once more strengthen these beliefs whilst people readingnewspapers give a higher importance to the economic contributions of foreigners.Finally, we relate the perceptions of EU citizens to the level of generosity of welfaresystems and the characteristics of social policies in different European countries, in orderto assess whether negative perceptions about migration are stronger in countries with amore generous redistributive system or adopting specific policies.Since the non-contributory part of public transfers turned out to be the one wheremigrants were over-represented, we focus our attention on the share of non-contributoryallowances over the total government expenditure for social benefits. Figure 6 shows aninverse relationship between this quantity and the percentage of people stating thatmigrants are net contributors to public finances. This is in line with our findings: as longas the share of non-contributory benefits is relatively low as compared to the contributoryones, people appreciate more the fiscal contribution of migrants. On one side, when mostof the transfers are given on a non-contributory basis, in which migrants are overrepresented,the average citizen is more concerned about his/her welfare dependency.fRDB 28

Figure 6. Migrants perceived dependency and non-contributory transfers.51.5PT.4PT1IT.5ES.3.2.1ITPLESCHBESINLATCZGRFR GB.05 .1 .15 .2Non Contributory Allowances (% Social Expenditure)DESENOFIHULUDKIE0-.5-1PLCHBESI.05 .1 .15 .2Non Contributory Allowances (% Social Expenditure)NLCZATGRFRDESEGBNOFIHULUDKIEMigrants give more(Percentage of respondents)Regression country dummiesNotes: percentage of people answering strictly less than 5 to the ESS-2002 question about migration and welfare contribution.Regression dummies from column (1) in Table 9, the choice of regression doesn’t have a qualitative influence, given the veryhigh correlation of dummies from different columns.Share of non contributory benefits calculated as: Social transfers in kind / (Social benefits + social transfers in kind).Source: Eurostat, statistics on Social Protection Expenditure 2002.In order to be more rigorous we control for individual specific characteristics, and plottingthe dummies from the first regression of Table 9 on the vertical axis, we find that anegative correlation still holds even conditionally on the respondent’s personal features,as displayed in the right-panel of Figure 6.As expected, a similar relation holds even when we consider the share of GDP spend forfamily allowances, housing or social exclusion, as displayed in Figure 7.fRDB 29

Figure 7. Migrants perceived dependency and non-contributory transfers1.5PT1.5PT1IT1IT.50-.5-1ESPLNLCHCZGBGRSIBEFRHU IESE ATNOFIDELUDK.50-.5-1ESAT NOLUFIBE CHCZ DENLHU IESEGRDKFRGB1 2 3 4Family/Children (% GDP)0 .5 1 1.5Housing (% GDP)1.5PT1.5PT1IT1IT.50-.5-1ESPLAT SI NO SEFR LU FIBE CHGBHU IE CZDEGRDKNL.50-.5-1ESPLCHCZSINLBEGRHU IEAT NO SEFIFR LUGBDEDK0 .5 1 1.5Social exclusion n.e.c. (% GDP)1 2 3 4 5 6Non Contributory Allowances (% GDP)Notes: Regression dummies from column (1) in Table 9; The choice of regression doesn’t have a qualitative influence, giventhe very high correlation of dummies from different columns.Share of non contributory benefits calculated as: Social transfers in kind / (Social benefits + social transfers in kind)Source: Eurostat, statistics on Social Protection Expenditure 2002Furthermore it is interesting to notice a slightly positive relation between the perceptionsof European citizens about migrants’ contribution to the state budget and the average netfiscal position of migrants, as calculated using the EU-SILC 2005, as well as thegovernment net lending, as provided by the European Commission for the ExcessiveDeficit Procedure.fRDB 30

Figure 8. Migrants perceived dependency and net fiscal position1.51.5PT11IT.5.5ES0-.5PLFIDKHUNOFRNLCZIEDESE ATLUGBBE0-.5HUCZSI SE ATPLFRNLDKBEGBIEDELUFINO-1-1GR0 5000 10000 15000 20000Migrants' net fiscal position (average 2005)-10 -5 0 5 10Net Lending (+) or borrowing (-) as % GDPNotes: Regression dummies from column (1) in Table 9; the choice of regression doesn’t have a qualitative influence, given thevery high correlation of dummies from different columns.Source: Eurostat, statistics on Social Protection Expenditure 2002;Net fiscal position calculated as country-average of the variable constructed in section 3, EU-SILC 2005;Eurostat, Government Deficit and Debt, Excessive Deficit Procedure 2002.5 Conclusions and policy implicationsThere is a widespread perception in Europe that migrants are a burden for public finance.This view is deeply rooted in the countries with a more generous redistributive systemand is stronger among poorer and less educated individuals. We document in this studythat migrants are indeed over-represented among beneficiaries of non-contributorytransfers, while they are under-represented among recipients of contributory schemes.However, EU-25 migrants are under-represented also among recipients of noncontributorybenefits. Interestingly, among them there are several countries where theshare of NMS migrants in the population of EU-25 migrants is relatively large (Greece,Austria, Ireland, Italy). Furthermore, especially in Nordic countries, there is someevidence of “residual dependency” of migrants, thereby they receive transfers more thannatives when control is made of their educational attainments and family characteristics.We also try for the first time to estimate the net fiscal position of migrants vis-à-vis thestate budget. Our estimates depend on a number of assumptions and caveats that arefRDB 31

detailed in the report. They suggest that the net fiscal position of migrants is not differentthan the one of natives. They pay less, but also receive less than natives. It should bestressed that our calculations are static, that is compare current contributions and taxesand current transfers rather than analysing them over the lifetime. Thus the young age ofmigrants may contribute to explain our results.The main policy implications of our findings is that countries should look carefully at thedesign of their social welfare systems in order to minimise moral hazard and preventmigrants from falling into unemployment and poverty traps. Adopting a more Beveridgianwelfare system is not always an option as some schemes (e.g., social assistance) canonly be funded out of general Government revenues. But much can be done to reducelong-term dependency from such transfers, as suggested by ongoing policy experimentsalong the route of activation policies (Boeri, 2005).fRDB 32

6 ReferencesAuerbach, A.J. and P. Oreopoulos (1999), “Analyzing the Fiscal Impact of U.S.Immigration”, American Economic Review, 89(2), pp. 176-180.Barrett, Alan and Yvonne McCarthy (2007), “Immigrants in a Booming Economy:Analysing their Earnings and Welfare Dependence”, Labour, 21(4-5), pp. 789-808.Barrett, Alan and Yvonne McCarthy (2008), "Immigrants and Welfare Programmes:Exploring the Interactions between Immigrant Characteristics, Immigrant WelfareDependence and Welfare Policy," IZA Discussion Papers 3494, IZA, Bonn.Bird, E.J., H. Kayser and J.R. Frick (1999), “The Immigrant Welfare Effect, Take-up orEligibility?”, IZA Discussion Paper 66, IZA, Bonn.Blank, R.M. (1988), “The Effect of Welfare and Wage Levels on the Location Decisions ofFemale-Headed Households”, Journal of Urban Economics, 24(2), pp. 186-211.Blau, F. D. (1984), “The use of transfer Payments by Immigrants”, Industrial and LaborRelations Review, 37, pp. 222-39.Boeri, T. (2005), “An Activating Social Security System”, De Economist, 153, n.4.Boeri, T. (2006), “Migration Policy and the Welfare State”, paper presented to theConference “Reinventing the Welfare State”, Tilburg.Boeri, T., G. Hanson and B. McCormick (eds.) (2002), Immigration Policy and the WelfareSystem, Oxford University Press, Oxford.Borjas, G.J. (1999), “Immigration and Welfare Magnets”, Journal of Labour Economics,17(4), pp. 607-637.Borjas, G.J. (1995), "Immigration and Welfare, 1970-1990”, Research in Labor Economics,14, pp. 253-282.Borjas, G.J. and S. J. Trejo (1991), “Immigrant Participation in the Welfare System”,Industrial and Labor Relations Review, 44, pp. 195-211.Borjas, G.J. (1996), “The Earnings of Mexican Immigrants in the United States”, Journalof Development Economics, 51, pp. 69-98.Borjas, G.J., and L. Hilton (1996), “Immigration and the Welfare State: ImmigrantParticipation in Means-Tested Entitlement Programs”, Quarterly Journal of Economics,111(2), pp. 575-604.Brueckner, J. K. (2000), “Welfare Reform and the Race to the Bottom: Theory andEvidence”, Southern Economic Journal, 66(3), pp. 505-525.fRDB 33

Buonanno, P., M. Bianchi and P. Pinotti (2008). "Do Immigrants Cause Crime?" ParisSchool of Economics Working Paper 05/2008Fertig, M. and Schmidt, C. M., (2001), “First- and Second-Generation Migrants inGermany – What Do We Know and What Do People Think”, in Rotte R. (ed.) MigrationPolicy and the Economy- International Experience, mimeo.Fix, M.E., J.S. Passel and W. Zimmermann (1996), "Facts about immigrants' use ofwelfare", The Urban Institute.Freeman, R.B. (1991). “Crime and the employment of disadvantaged youths”, NBERWorking Paper No. 3875.Frick, J., F. Büchel and W. Voges (1996), „Sozialhilfe als Integrationshilfe für Zuwandererin Deutschland“, DIW-Wochenbericht, 63(48), pp. 767-775.Gelbach, J.B. (2004), “Migration, the Life Cycle, and State Benefits: How Low Is theBottom?”, Journal of Political Economy, 112(5), pp. 1091-1130.Gramlich, E.M. and D.S. Laren (1984), "Migration and Income RedistributionResponsibilities", Journal of Human Resources, 19(4), pp. 489-511.Grogger, J. (1998), “Market wages and youth crime”, Journal of Labor Economics, 16(4),pp. 756–791.Hansen, J. and M. Lofstrom, (1999), “Immigration and Welfare Participation: DoImmigrants Assimilate Into or Out-of Welfare?”, IZA Discussion Paper 100, IZA, Bonn.Hansen, J. and M. Lofstrom, (2001), “The dynamics of immigrant welfare and labormarket behaviour”, IZA Discussion Paper 360, IZA, Bonn.Hansen, J. and M. Lofstrom, (2003), “Immigrant assimilation and welfare participation:do immigrants assimilate into or out of welfare?”, Journal of Human Resources, 38(1),pp. 74-98.Levitt, S.D. (1996), “The effect of prison population size on crime rates: Evidence fromprison overcrowding litigation”, Quarterly Journal of Economics, 111(2), pp. 319–351.Levitt, S.D. (1998), “Juvenile crime and punishment”, Journal of Political Economy,106(6), pp. 1156–1185.McKinnish, T. (2005), “Importing the Poor: Welfare Magnetism and Cross-BorderMigration”, Journal of Human Resources, 40(1), pp. 57-76.Meyer, B.D. (2000), “Do the Poor Move to Receive HigherWelfare Benefits?”, Workingpaper no. 58. Northwestern University, Joint Center for Poverty Research Working Paper.fRDB 34

Pederson, P. (2000), “Immigration in a High Unemployment Economy: The Recent DanishExperience”, IZA Discussion Paper 165, IZA, Bonn.Riphahn, R.T. and M. Rosholm, (2001), “Immigrants' Time to Economic Independence:The Duration of Initial Public Transfer Reliance”, mimeo.Riphahn, R. T., (1998), „Immigrant Participation in the German Welfare Program”,Finanzarchiv, 55, pp. 163-185.Sinn, H.-W., G. Flaig, M. Werding, S. Munz, N. Düll and H. Hofmann, (2001), „EU-Erweiterung und Arbeitskräftemigration, Wege zu einer schrittweisen Annäherung derArbeitsmärkte“, ifo-Institut für Wirtschaftsforschung, Munich.Walker, J.R. (1994), “Migration Among Low-Income Households: Helping the WitchDoctors Reach Consensus,” discussion paper 94-1031, Institute for Research on Poverty,University of Wisconsin-Madison.fRDB 35

7 AnnexTable A1a. Share of migrants from New Member States over migrant population inEuropean CountriesCountryShare of immigrants fromNMS-10 over total EU-25immigrantsShare of immigrantsfrom NMS-12 over totalimmigrantsAustria 35.90 14.15Belgium 4.00 4.17Bulgaria 52.32 6.78Croatia No information No informationCyprus 4.41 8.56Czech 79.71 41.12Denmark 13.03 5.17Estonia 69.11 3.10Finalnd 46.77 26.74France 3.37 2.37Germany 19.35 8.98Greece 52.86 14.55Hungary 41.82 59.36Iceland 45.47 28.20Ireland 43.52 No informationItaly 33.26 16.18Latvia 82.73 5.68Lithuania 76.74 13.17Luxembourg 1.37 1.46Malta No information No informationNetherlands 7.84 3.84Norway 7.90 5.06Poland 11.97 3.96Portugal 2.74 6.58Romania 34.59 9.64Slovakia 88.88 58.98Slovenia** 18.68 2.80Spain 9.80 20.21Sweden 10.19 6.17Switzerland 2.33 1.91Turkey No information No informationUnited Kingdom 28.06 13.43Source: Own elaborations on National Population Statistics andEUROSTAT LFS.fRDB 36

Table A1b. Descriptive Statistics - CitizensCountry Natives MigrantsTotalEU-25Extra EU-25Austria 38424 680 1970 41074Belgium 36917 2022 1138 40077Cyprus 23425 1497 1314 26236Czech Republic 28861 146 173 29180Germany 72964 147774441Denmark 45688 652 1470 47810Estonia 35056 533240388Spain 26421 149 907 27477Finalnd 87679 488 790 88957France 70644 1632 2458 74734Greece 48676 318 1857 50851Hungary 44931 158 69 45158Ireland 43211 1539 700 45450Iceland 18174 240 7564 25978Italy 185964 497 4156 190617Lithuania 29151 54 275 29480Luxembourg 17211 11264 1415 29890Latvia 19762 488324645Netherlands 17172 142 92 17406Norway 44747 902 829 46478Poland 107955 36 92 108083Portugal 41958 181 689 42828Sweden 46174 959 1018 48151Slovenia 66732 614972881Slovakia 37507 88 34 37629United Kingdom 54749 621 2216 57586Total 1290153 24265 49067 1363485fRDB 37

Table A1c. Descriptive Statistics – HouseholdsCountry Natives Migrants Mixed TotalEU-25Extra EU-25Austria 15791 183 617 606 17197Belgium 15715 704 336 752 17507Cyprus 7922 348 200 561 9031Czech Republic 11868 43 56 85 12052Germany 28638 266693 29597Denmark 19493 109 357 781 20740Estonia 11585 1593676 13854Spain 11408 52 307 173 11940Finalnd 35828 114 202 472 36616France 29888 518 826 884 32116Greece 18418 90 593 340 19441Hungary 15192 37 23 60 15312Ireland 17446 371 228 504 18549Iceland 5940 60 689 3052 9741Italy 72185 141 1660 596 74582Lithuania 8825 7 27 106 8965Luxembourg 6711 4344 361 1018 12434Latvia 5559 1214846 7619Netherlands 8971 25 10 163 9169Norway 18880 242 206 618 19946Poland 32255 11 11 54 32331Portugal 14706 46 163 157 15072Sweden 18991 223 217 685 20116Slovenia 16512 12751493 19280Slovakia 11579 20 2 69 11670United Kingdom 21680 117 550 591 22938Total 481986 7805 11989 16035 517815fRDB 38

Table A2. Change in the probability of receiving contributory-benefits:individual probit regression(1) (2) (3) (4) (5) ( 6) (7) (8) (9) (10) (11) (12) (13) (14)All AT BE CY CZ DE + DK EE + ES FI FR GR HU IE- 0.065 - 0.082 - 0.052 - 0.031 0.0 44 0.010 - 0.032 - 0.110 - 0.063 - 0.046 - 0.210 - 0.125EU25 Migrant[15.55]*** [3.21]*** [4.03]*** [2.20]** [0.72] [0.31] [1.81]* [3.01]*** [3.72]*** [1.39] [3.57]*** [8.44]***Extra EU25 - 0.048 - 0.011 - 0.200 - 0.137 - 0.275 0.048 0.074 0.049 - 0.096 0.020 - 0.109 - 0.081 - 0.402 - 0.180Migrant [16.54]*** [0.68] [12.39]*** [6.75]*** [4.21]*** [2.37]** [3.81]*** [4.23]*** [5.60]*** [0.76] [7.29]*** [4.84]*** [5.35]*** [8.13]***Male0.094 0.201 0.197 0.021 - 0.087 0.130 - 0.002 - 0.062 0.174 0.014 0.110 0.159 0.003 0.12 2[88.43]*** [27.65]*** [28.32]*** [3.20]*** [9.91]*** [22.10]*** [0.26] [7.30]*** [42.13]*** [2.57]** [20.49]*** [24.64]*** [0.42] [18.14]***Age- 0.005 - 0.003 0.020 - 0.029 - 0.030 - 0.018 0.010 - 0.030 - 0.002 0.012 - 0.002 - 0.003 0.007 0.004[26.21]*** [2.44]** [19.27]*** [24.33]*** [15.18]*** [15.54]*** [7.38]*** [16.76]*** [3.67]*** [10.64]*** [1.87]* [3.22]*** [4.10]*** [3.73]***Age^20.000 0.000 0.000 0.000 0.001 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000[119.37]*** [14.95]*** [7.44]* ** [35.63]*** [24.98]*** [33.25]*** [4.20]*** [28.99]*** [23.44]*** [6.68]*** [17.47]*** [20.45]*** [9.78]*** [8.42]***Secondary 0.013 0.018 0.004 - 0.025 0.052 0.036 - 0.017 - 0.011 - 0.018 0.027 0.045 0.020 - 0.023 - 0.037Education [10.13]*** [2.28]** [0. 59] [3.14]*** [4.08]*** [4.24]*** [1.99]** [0.94] [3.50]*** [3.86]*** [7.58]*** [2.85]*** [2.80]*** [4.92]***Tertiary - 0.002 - 0.006 0.012 - 0.056 0.025 - 0.009 - 0.025 - 0.027 0.000 - 0.017 0.056 0.061 - 0.037 - 0.034Education [1.33] [0.55] [1.40] [6.07]*** [1.29] [0.92] [2.53]** [1.94]* [0.01] [2.12]** [7.04]*** [6.32]*** [2.92]*** [3.82]***High income (before- 0.267 - 0.261 - 0.348 - 0.146 - 0.223 - 0.249 - 0.419 - 0.083 - 0.234 - 0.428 - 0.354 - 0.261 - 0.300 - 0.218transfers) [179.04]*** [26.37]*** [33.35]*** [2.56]** [16.91]*** [21.61]*** [42.51]*** [7.06]*** [36.23]*** [53.14]*** [48.99]*** [31.25]*** [26.31]*** [21.55]***0.129 0.239 0.291 - 0.003 0.199 0.245 0.150 0.155 0.076 0.162 0.105 - 0.012 0.165 0.107Low income (b.t)[87.60]*** [26.61]*** [31.69]*** [0 .07] [17.76]*** [25.68]*** [14.15]*** [13.93]*** [13.45]*** [19.41]*** [15.11]*** [1.45] [14.87]*** [11.50]***House Owner - 0.043 - 0.064 - 0.108 - 0.038 - 0.011 - 0.095 - 0.045 0.003 - 0.019 - 0.028 - 0.037 0.012 - 0.043 - 0.104[28.88]*** [7.50]*** [13.64]*** [2 .76]*** [1.00] [15.27]*** [5.19]*** [0.12] [2.54]** [3.75]*** [5.90]*** [1.46] [2.91]*** [10.68]***Single0.057 0.151 0.020 0.005 - 0.113 0.041 - 0.192 0.001 0.082 0.075 - 0.047 0.181 - 0.055 0.197[18.20]*** [7.33]*** [1.10] [0.24] [4.52]*** [2.34]** [7.9 6]*** [0.05] [7.91]*** [4.30]*** [3.09]*** [9.85]*** [2.48]** [10.15]***0.115 0.142 0.080 0.068 0.035 0.114 - 0.040 0.049 0.198 0.106 - 0.081 0.334 - 0.019 - 0.013Single with child[37.10]*** [6.68]*** [5.18]*** [2.83]*** [1.54] [8.17]*** [1.89]* [2.58]** * [11.73]*** [6.86]*** [5.78]*** [12.29]*** [0.97] [0.71]1 child- 0.072 - 0.037 - 0.048 - 0.041 - 0.112 - 0.006 - 0.078 - 0.099 - 0.023 - 0.128 - 0.065 - 0.043 - 0.046 - 0.050[30.46]*** [2.24]** [3.01]*** [2.40]** [5.47]*** [0.34] [4.40]*** [7.02]*** [2.50]** [9.5 2]*** [5.40]*** [3.24]*** [2.57]** [3.68]***2 children- 0.071 - 0.057 - 0.074 - 0.041 - 0.078 - 0.061 - 0.094 - 0.078 - 0.020 - 0.083 - 0.078 - 0.024 - 0.121 - 0.091[35.02]*** [3.79]*** [4.36]*** [3.53]*** [4.15]*** [5.49]*** [6.05]*** [5.85]*** [2.61]*** [6.88]** * [6.86]*** [1.94]* [8.95]*** [7.96]***3 children- 0.066 - 0.106 - 0.036 0.141 0.051 - 0.113 - 0.071 0.244 - 0.176 - 0.074 - 0.215[2.77]*** [1.09] [0.28] [0.72] [0.40] [1.50] [0.83] [2.11]** [1.32] [0.50] [3.65]***4+ children- 0.093 - 0.065 0. 433 - 0.263 0.065 - 0.174 0.041 0.290 - 0.147[4.72]*** [0.90] [3.78]*** [3.63]*** [0.44] [1.76]* [0.33] [3.09]*** [2.44]**2 household - 0.067 - 0.072 - 0.169 - 0.097 - 0.151 - 0.110 - 0.196 - 0.016 - 0.104 - 0.021 - 0.158 - 0.021 - 0.076 - 0.012members [19.8 6]*** [3.52]*** [9.46]*** [3.97]*** [5.42]*** [5.94]*** [7.80]*** [0.64] [11.51]*** [1.10] [10.20]*** [1.23] [2.86]*** [0.63]3 hh - 0.112 - 0.102 - 0.202 - 0.132 - 0.206 - 0.195 - 0.311 - 0.093 - 0.135 - 0.051 - 0.226 - 0.032 - 0.152 - 0.054members [29.35]*** [4.3 8]*** [10.77]*** [5.51]*** [6.65]*** [8.66]*** [10.28]*** [3.39]*** [15.03]*** [2.16]** [13.38]*** [1.69]* [5.05]*** [2.35]**4 hh - 0.130 - 0.137 - 0.293 - 0.157 - 0.287 - 0.239 - 0.371 - 0.081 - 0.175 - 0.062 - 0.296 - 0.031 - 0.158 - 0.021members [28.23]*** [4.4 0]*** [10.53]*** [4.76]*** [7.30]*** [7.91]*** [9.12]*** [2.66]*** [15.67]*** [2.05]** [12.64]*** [1.36] [4.47]*** [0.73]5 hh - 0.120 - 0.138 - 0.243 - 0.113 - 0.244 - 0.218 - 0.351 - 0.120 - 0.126 - 0.120 - 0.243 - 0.017 - 0.176 - 0.014members [23.35]*** [4.54]** * [9.20]*** [3.93]*** [5.33]*** [4.71]*** [6.35]*** [3.46]*** [10.25]*** [1.78]* [9.46]*** [0.68] [4.55]*** [0.46]6 hh - 0.165 - 0.135 - 0.286 - 0.160 - 0.346 - 0.276 - 0.427 - 0.105 - 0.182 - 0.114 - 0.311 - 0.046 - 0.254 - 0.058members [34.60]*** [4.16]*** [12.4 3]*** [5.58]*** [8.74]*** [9.55]*** [10.69]*** [3.07]*** [17.23]*** [3.41]*** [14.51]*** [1.85]* [6.99]*** [1.83]*7 hh - 0.191 - 0.177 - 0.300 - 0.176 - 0.348 - 0.301 - 0.434 - 0.164 - 0.198 - 0.170 - 0.318 - 0.054 - 0.256 - 0.051members [39.99]*** [5.38]*** [13.1 3]*** [5.90]*** [8.73]*** [10.68]*** [11.08]*** [4.82]*** [18.41]*** [4.53]*** [14.97]*** [2.17]** [6.92]*** [1.58]Densly - 0.021 - 0.017 0.007 - 0.016 0.018 - 0.044 - 0.058 - 0.074 - 0.023 - 0.027 - 0.014 - 0.020 - 0.005 0.000populated area [15.05]*** [1.86]* [ 1.04] [1.58] [1.41] [6.80]*** [6.69]*** [7.84]*** [4.80]*** [2.89]*** [2.21]** [1.21] [0.38] [0.00]Notes: z statistics in brackets; * significant at 10percent; ** significant at 5percent; *** significant at 1percent; + the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants; ++ migrants identified by country of birth; the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants. Low income variable defined as equivalizedincome lower than 60 percent of median income; High income variable defined as equivailzed income greater than 4/3 of median income.fRDB 39

Table A2 (Continued). Change in the probability of receiving contributory-benefits:individual probit regressionEU25 Migrant(15) (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) (26) (27) (28)IS IT L L LV + NO P PT SE SK UK SI ++ N UK- 0.023 - 0.107 - 0.157 T - 0.040 U- 0.038 - 0.180 L - 0.123 - 0 .180 0.122 0.003 0.004 L 0.004[0.63] [3.53]*** [2.35]** [4.15]*** [1.45] [2.03]** [2.63]*** [7.81]*** [2.08]** [0.08] [0.08] [0.15]- 0.025 - 0.007 0.042 - 0.103 - 0.034 - 0.201 - 0.229 - 0.116 - 0.245 - 0.227 - 0.161 0.009 - 0.128 - 0.141Extra EU25Migrant [2.64]*** [0.52] [0.99] [5.61]*** [2.69]*** [6.79]*** [3.90]*** [4.02]*** [11.65]*** [3.35]*** [8.21]*** [1.00] [1.83]* [7.98]***Mal- 0.015 0.159 - 0.048 0.188 - 0.057 0.033 0.056 0.100 - 0.003 - 0.045 0.106 - 0.037 0.134 0.107e [2.08]** [41.10]*** [4.74]* ** [19.18]*** [5.65]*** [5.23]*** [12.51]*** [14.80]*** [0.49] [6.28]*** [14.38]*** [7.09]*** [10.90]*** [15.86]***- 0.016 - 0.024 - 0.029 - 0.007 - 0.036 0.014 - 0.009 - 0.006 0.019 - 0.010 - 0.034 - 0.008 - 0.026 - 0.034Age[13.55]*** [31.60]*** [13.72]*** [3.4 6]*** [16.59]*** [12.32]*** [7.44]*** [4.91]*** [19.21]*** [6.28]*** [19.06]*** [6.00]*** [12.47]*** [20.90]***Age^20.000 0.000 0.001 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.001[24.27]*** [51.51]*** [25.47]*** [11.30]*** [26.89] *** [1.60] [24.74]*** [18.56]*** [5.79]*** [19.41]*** [30.68]*** [21.59]*** [21.07]*** [33.38]***Secondary - 0.015 - 0.048 - 0.046 - 0.016 - 0.001 - 0.027 0.037 - 0.031 0.077 0.092 - 0.010 0.043 0.048 - 0.008Educatio [2.09]** [11.41]*** [3.33]*** [1.72]* [0.1 0] [3.10]*** [6.70]*** [2.78]*** [10.17]*** [9.24]*** [1.21] [6.90]*** [3.77]*** [1.08]nTertiary - 0.039 - 0.061 - 0.005 - 0.035 - 0.034 - 0.097 0.025 0.065 0.030 0.071 - 0.002 0.084 0.059 0.008Educatio [3.21]*** [9.01]*** [0.27] [2.69]*** [1.93]* [9.75]*** [2.76]*** [4.79]*** [3.37]*** [5.04]*** [0.24] [7.71]*** [3.75]*** [0.90]High n- 0.177 - 0.196 - 0.140 - 0.191 - 0.010 - 0.384 - 0.310 - 0.148 - 0.316 - 0.231 - 0.127 - 0.216 - 0.301 - 0.122income (befortransfers) e[18.49]*** [34.62]*** [8.43]*** [14.10]*** [0.67] [43.58]*** [38.46]*** [13.79]*** [40.12]*** [21.19]*** [12.80]*** [28.17]*** [18.49]*** [13.34]***Low income (b.t) [22.04]*** [4.46]*** [20.88]*** [17.62]*** [15.35]*** [16.22]*** [23.82]*** [27.87]*** [25.16]*** [19.45]*** [34.36]*** [11.27]*** [1.99]** [39.42]***0.206 0.024 0.312 0.228 0.209 0.146 0.178 0.247 0.201 0.204 0.306 0.081 0.032 0.321- 0.088 - 0.001 0.090 0.008 - 0.050 0.029 - 0.020 - 0.025 0.021 0.005 - 0.100 - 0.001 - 0.045 - 0.093House Owner [7.41]*** [0.21] [2.55]** [0.65] [3.81]*** [2.77]* ** [1.53] [2.91]*** [3.14]*** [0.42] [11.27]*** [0.07] [3.70]*** [11.53]***Single[0.96] [5.21]*** [0.88] [3.71]*** [3.92]*** [0.60] [4.53]*** [6.66]*** [16.99]** * [0.80] [4.45]*** [8.18]*** [0.61] [5.75]***0.021 0.056 - 0.027 0.104 - 0.104 - 0.011 0.066 0.147 - 0.313 - 0.019 0.102 - 0.190 0.037 0.1250.147 0.098 0.232 0.300 0.102 0.067 0.234 0.206 - 0.022 0.147 0.043 0.056 0.025 0.044Single with child [6.85]*** [8.02]*** [10.39]*** [9.80]*** [4.27]*** [3.84]*** [17.55]*** [8.68]*** [1.44] [6.66]*** [1.95 ]* [3.14]*** [0.55] [2.14]**1 child- 0.022 - 0.102 - 0.061 - 0.078 - 0.085 - 0.104 - 0.047 - 0.072 - 0.076 - 0.056 - 0.073 - 0.050 - 0.064 - 0.066[1.68]* [12.35]*** [2.38]** [4.67]*** [3.63]*** [7.80]*** [4.36]*** [5.19]*** [5.64]*** [3.41]*** [3.54]*** [4.43]*** [3.03]*** [3.22]***2 children- 0.042 - 0.089 - 0.046 - 0.054 - 0.027 - 0.096 - 0.057 - 0.074 - 0.046 - 0.052 - 0.099 - 0.041 - 0.080 - 0.091[3.44]*** [10.29]*** [3.01]*** [3.28]*** [1.59] [7.96]*** [8.37]*** [5.85]*** [3.92]*** [4.79]*** [6.81]*** [4.88]*** [1.78] * [6.69]***3 children0.049 0.035 0.067 - 0.118 0.022 0.045 - 0.141 0.065 - 0.105 0.074[0.50] [0.20] [0.43] [1.09] [0.21] [0.62] [1.25] [0.39] [0.68] [0.44]- 0.065 - 0.156 0.304 0.006 - 0.300 0.064 - 0.228 - 0.178 0.0324+children [0 .55] [1.93]* [2.31]** [0.05] [2.58]*** [1.07] [3.12]*** [2.25]** [0.22]2 household - 0.088 - 0.040 - 0.038 - 0.051 - 0.070 - 0.113 - 0.039 - 0.037 - 0.239 - 0.072 - 0.008 - 0.153 - 0.060 0.003member [3.86]*** [3.46]*** [0.95] [1.92]* [2.21]** [5.66]*** [2.0 2]** [1.59] [12.39]*** [2.46]** [0.34] [4.84]*** [1.05] [0.13]ssssss3member hh4member hh5member hh6member hh7member hh- 0.121 - 0.064 - 0.056 - 0.049 - 0.102 - 0.157 - 0.050 - 0.070 - 0.384 - 0.133 - 0.065 - 0.243 - 0.105 - 0.054[5.16]*** [4.88]*** [1.24] [1.63] [2.77]*** [6.54]*** [2.34]** [2.89]*** [16.4 4]*** [4.24]*** [2.21]** [7.61]*** [1.88]* [1.89]*- 0.153 - 0.082 - 0.084 - 0.110 - 0.122 - 0.165 - 0.065 - 0.085 - 0.468 - 0.162 - 0.031 - 0.284 - 0.134 - 0.022[5.13]*** [5.28]*** [1.59] [3.07]*** [2.75]*** [5.46]*** [2.86]*** [3.03]*** [15.29]*** [4 .73]*** [0.77] [8.41]*** [2.37]** [0.57]- 0.121 - 0.095 - 0.057 - 0.084 - 0.083 - 0.156 - 0.095 - 0.080 - 0.435 - 0.157 - 0.077 - 0.303 - 0.114 - 0.081[4.26]*** [5.37]*** [0.99] [2.19]** [1.61] [4.05]*** [4.01]*** [2.68]*** [13.03]*** [4.57]*** [1.47] [9.53]*** [1.68]* [1.57]- 0.165 - 0.101 - 0.083 - 0.119 - 0.196 - 0.224 - 0.066 - 0.123 - 0.526 - 0.207 - 0.051 - 0.369 - 0.028 - 0.045[5.75]*** [5.82]*** [1.51] [3.53]*** [4.26]*** [6.97]*** [2.81]*** [4.29]*** [17.81]*** [5.99]*** [1.22] [11.13]*** [0.27] [1.11]- 0.172 - 0.128 - 0.126 - 0.124 - 0.183 - 0.253 - 0.091 - 0.145 - 0.550 - 0.247 - 0.074 - 0.420 - 0.004 - 0.073[6.29]*** [7.25]*** [2.36]** [3.62]*** [3.87]*** [7.79]*** [3.87]*** [5.17]*** [21.17]*** [7.02]*** [1.75]* [12.83]*** [0.02] [1.83]*- 0.017 - 0.038 0.034 - 0.040 0.010 0.029 - 0.021 - 0.036 0.026Denslypopulated area [2.42]** [9.42]*** [3.48]*** [4.68]*** [1.43] [3.70]*** [2.15]** [4.29]*** [3.07]***Notes: z statistics in brackets; * significant at 10percent; ** significant at 5percent; *** significant at 1percent; + the EU-SILC does not distinguish between EU-25 and extra EU-25migrants; ++ migrants identified by country of birth; the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants. Low income variable defined as equivalizedincome lower than 60 percent of median income; High income variable defined as equivailzed income greater than 4/3 of median income.fRDB 40

Table A3. Change in the probability of receiving non-contributory allowances:household probit regression(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)All AT BE CY CZ DE DK EE ES FI FR G R HU IEEU25 Migrant - 0.062 - 0.023 - 0.046 - 0.391 - 0.261 0.005 - 0.054 - 0.141 0.034 - 0.055 - 0.123 - 0.168House [5.07]*** [0.53] [2.37]** [11.63]*** [6.74]*** [0.06] [3.76]*** [2.00]** [1.35] [1.66]* [1.55] [5.21]***Extra EU25 - 0.041 - 0.073 0.097 - 0.506 - 0.222 0.179 0.067 - 0 .068 - 0.018 0.162 0.295 - 0.059 - 0.258 - 0.038Migrant House [5.22]*** [3.33]*** [2.90]*** [11.00]*** [4.34]*** [3.73]*** [1.42] [3.68]*** [2.17]** [2.62]*** [10.13]*** [3.84]*** [2.39]** [0.80]Mixed 0.050 0.002 0.037 - 0.115 - 0.014 0.032 0.060 0.046 0.004 - 0 .005 0.130 0.009 0.248 0.069Household [6.76]*** [0.07] [2.10]** [4.74]*** [0.25] [1.29] [2.61]*** [2.04]** [0.29] [0.14] [6.41]*** [0.42] [3.03]*** [2.96]***Male0.000 0.004 - 0.018 - 0.052 0.003 - 0.004 - 0.100 0.024 - 0.004 - 0.035 - 0.021 - 0.052 - 0.013 - 0.017[0. 13] [0.25] [1.49] [2.11]** [0.16] [0.25] [6.92]*** [1.58] [0.87] [2.96]*** [2.11]** [6.04]*** [0.85] [1.34]Age0.001 0.027 0.039 - 0.007 0.029 0.042 - 0.021 0.015 0.002 - 0.001 0.003 - 0.010 0.004 - 0.024[2.58]*** [9.45]*** [9.66]*** [2.36]** [8.45]*** [14 .09]*** [7.96]*** [4.53]*** [2.97]*** [0.30] [1.86]* [8.53]*** [1.35] [12.88]***Age^20.000 0.000 - 0.001 0.000 0.000 - 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000[23.89]*** [13.15]*** [11.44]*** [0.08] [11.95]*** [18.90]*** [3.79]*** [8.17]** * [7.11]*** [7.86]*** [10.86]*** [8.48]*** [7.29]*** [13.73]***- 0.057 0.037 0.016 0.027 - 0.081 - 0.063 0.003 0.051 0.006 - 0.023 - 0.033 - 0.043 - 0.031 0.017SecondaryEducation [21.58]*** [2.60]*** [1.58] [1.91]* [3.41]*** [3.71]*** [0.22] [2.70]*** [1.60 ] [2.13]** [3.99]*** [6.10]*** [2.35]** [1.48]Tertiary - 0.046 0.103 0.031 0.026 - 0.126 - 0.031 0.065 0.079 0.017 0.017 - 0.034 - 0.022 - 0.033 - 0.020Education [14.95]*** [6.22]*** [2.89]*** [1.58] [5.22]*** [1.79]* [4.58]*** [3.56]*** [4.62]*** [1.46] [3. 29]*** [2.66]*** [1.92]* [1.71]*- 0.160 - 0.190 - 0.048 - 0.122 - 0.218 - 0.256 - 0.178 - 0.088 - 0.018 - 0.190 - 0.207 - 0.027 - 0.149 - 0.148High income(beforetransfers) [64.38]*** [19.64]*** [5.52]*** [8.36]*** [19.46]*** [29.44]*** [16.03]*** [7.00]*** [5.36] *** [20.03]*** [24.11]*** [3.71]*** [11.30]*** [12.58]***0.049 0.000 0.009 0.054 0.077 0.036 0.157 0.115 0.018 0.239 0.115 0.052 0.073 0.156Low income (b.t)[15.78]*** [0.02] [0.72] [2.96]*** [4.60]*** [2.88]*** [10.85]*** [6.35]*** [4.43]*** [20.29]* ** [10.98]*** [6.25]*** [4.66]*** [11.48]***- 0.092 - 0.024 0.025 0.105 - 0.044 - 0.015 - 0.265 - 0.002 0.001 - 0.241 - 0.222 - 0.026 - 0.006 0.020House Owner [33.07]*** [2.19]** [2.44]** [4.82]*** [3.35]*** [1.78]* [22.35]*** [0.07] [0.18] [20.43]*** [27.17]** * [2.92]*** [0.27] [1.47]Single- 0.227 - 0.335 - 0.262 - 0.376 - 0.250 - 0.047 0.041 - 0.395 - 0.037 - 0.066 - 0.212 - 0.082 - 0.400 - 0.453[40.10]*** [11.50]*** [12.14]*** [8.19]*** [8.65]*** [1.37] [0.30] [16.21]*** [6.04]*** [1.72]* [9.27]*** [6.02]*** [13.23] *** [14.13]***0.226 0.402 0.367 0.026 0.314 0.593 0.430 0.285 0.030 0.440 0.186 - 0.012 0.313 0.121Single with child[25.51]*** [7.44]*** [8.26]*** [0.37] [5.91]*** [9.52]*** [3.90]*** [6.94]*** [2.87]*** [9.58]*** [6.81]*** [0.47] [5.88]*** [2.62]***1 child0.281 0.545 0.635 0.276 0.352 0.441 0.342 0.453 0.055 0.370 0.346 0.014 0.470 0.287[44.65]*** [16.18]*** [15.50]*** [9.21]*** [10.37]*** [8.59]*** [4.00]*** [16.32]*** [7.34]*** [14.11]*** [16.56]*** [0.99] [14.01]*** [17.98]***2 children0.2 89 0.649 0.542 0.289 0.371 0.404 0.353 0.549 0.058 0.316 0.394 0.022 0.512 0.283[46.00]*** [14.72]*** [12.29]*** [11.73]*** [11.68]*** [8.07]*** [4.28]*** [14.17]*** [7.23]*** [12.74]*** [16.08]*** [1.35] [16.70]*** [16.18]***3 children0.457 0 .241[5.34]*** [1.27]4+ children0.5072 household [5.53]***- 0.053 - 0.082 - 0.014 - 0.155 - 0.052 0.238 0.268 - 0.121 0.013 0.166 - 0.127 - 0.051 - 0.281 - 0.206members [8.09]*** [2.3 3]** [0.52] [3.17]*** [1.36] [6.58]*** [1.97]** [4.49]*** [2.12]** [4.24]*** [5.28]*** [3.13]*** [7.75]*** [7.11]***- 0.023 - 0.092 0.014 - 0.225 - 0.082 0.531 0.462 - 0.118 0.010 0.325 - 0.137 - 0.053 - 0.288 - 0.4063 hh members[2.89]*** [2.40]** [0.43] [3.8 2]*** [1.95]* [12.57]*** [3.82]*** [4.03]*** [1.36] [7.25]*** [4.59]*** [3.10]*** [7.57]*** [10.01]***4 hh members- 0.006 - 0.014 0.070 - 0.147 0.016 0.515 0.587 - 0.077 0.029 0.432 - 0.127 - 0.025 - 0.331 - 0.345[0.58] [0.25] [1.58] [2.10]** [0.27] [8.02]** * [2.69]*** [2.11]** [3.51]*** [6.46]*** [3.05]*** [1.05] [6.14]*** [6.50]***0.041 0.046 0.226 - 0.128 0.135 0.575 - 0.085 0.011 0.491 - 0.014 0.150 - 0.180 - 0.3745 hh members [2.86]*** [0.71] [2.58]*** [1.40] [1.53] [5.67]*** [1.67]* [0.84] [8.14]*** [0.22] [4.09]*** [2.68]*** [5.80]***0.070 - 0.072 0.122 - 0.097 - 0.022 0.677 0.537 - 0.097 0.036 0.525 - 0.123 - 0.058 - 0.319 - 0.5546 hh members [5.83]*** [1.26] [2.36]** [1.28] [0.35] [12.90]*** [3.80]*** [2.32]** [3.37]*** [9.42]*** [2.66]*** [2.42]** [ 6.16]*** [9.02]***0.155 0.120 0.425 0.052 0.060 0.680 0.516 0.032 0.078 0.548 0.078 - 0.003 - 0.202 - 0.4537 hh members [11.92]*** [1.79]* [6.89]*** [0.74] [0.86] [14.88]*** [5.00]*** [0.69] [6.05]*** [11.75]*** [1.45] [0.12] [3.49]*** [7.07]***Notes: z statistics in brackets; * significant at 10percent; ** significant at 5percent; *** significant at 1percent; + the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants; ++ migrants identified by country of birth; the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants. Low income variable defined as equivalizedincome lower than 60 percent of median income; High income variable defined as equivailzed income greater than 4/3 of median income.fRDB 41

Table A3 (Continued). Change in the probability of receiving non-contributory allowances:household probit regression(15) (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) (26) (27) (28)IS IT LT LU LV + NO PL PT S E SK UK SI ++ NL UKEU25 Migrant - 0.232 - 0.154 0.053 - 0.150 0.009 - 0.177 - 0.184 - 0.022 - 0.069 0.061 - 0.060House [2.80]*** [1.68]* [2.62]*** [3.84]*** [0.06] [1.50] [3.90]*** [0.21] [0.75] [0.56] [0.95]Extra EU25 - 0.047 - 0.017 - 0.173 0.090 - 0.024 0.106 - 0.171 - 0.205 0.035 0.291 - 0.245 0.006 0.421 - 0.229Migrant House [1.11] [1.19] [2.21]** [1.49] [1.16] [1.81]* [1.93]* [6.15]*** [0.70] [1.49] [9.61]*** [0.27] [2.86]*** [9.64]***Mixed - 0.020 0.058 0.054 0.025 - 0.01 0 0.101 0.016 0.123 0.059 - 0.049 - 0.020 0.083 - 0.013 - 0.014Household [0.58] [2.45]** [0.89] [0.96] [0.45] [3.93]*** [0.30] [2.82]*** [2.38]** [0.78] [0.78] [5.55]*** [0.26] [0.59]MaleAgeAge^2- 0.033 0.087 - 0.021 - 0.048 - 0.007 - 0.088 - 0.019 0.004 - 0.138 0.034 - 0.002 - 0. 057 - 0.059 0.000[1.63] [14.34]*** [1.06] [1.78]* [0.37] [5.36]*** [2.12]** [0.24] [10.25]*** [1.66]* [0.11] [4.08]*** [3.29]*** [0.02]0.014 - 0.007 - 0.013 0.048 - 0.011 0.007 0.007 0.022 - 0.016 0.008 0.028 0.027 - 0.008 0.029[4.15]*** [6.95]*** [4 .47]*** [8.96]*** [3.47]*** [2.17]** [4.99]*** [7.72]*** [7.81]*** [2.14]** [14.49]*** [9.95]*** [2.43]** [16.54]***0.000 0.000 0.000 - 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000[8.94]*** [4.61]*** [1.20] [10.61]*** [0.31] [7.28]*** [12.44]*** [12.86]*** [0.57] [7.36]*** [19.98]*** [14.19]*** [1.27] [22.42]***Secondary - 0.011 - 0.070 0.004 - 0.038 0.012 - 0.056 - 0.061 - 0.012 - 0.085 - 0.128 - 0.067 - 0.060 - 0.111 - 0.070Education [0.70] [14.48]*** [0.21] [2.08]** [0.64] [3.29]* ** [7.67]*** [0.72] [6.30]*** [5.24]*** [6.64]*** [4.66]*** [6.39]*** [7.50]***Tertiary 0.012 - 0.096 0.059 - 0.023 0.006 - 0.002 - 0.176 - 0.098 - 0.022 - 0.075 - 0.068 - 0.152 - 0.061 - 0.063Education [0.60] [13.29]*** [2.41]** [1.07] [0.25] [0.09] [18.51]*** [6.24]*** [1.52] [2.74]*** [6.26]*** [9.25]*** [3.26]*** [6.26]***High income(before- 0.230 - 0.128 - 0.090 - 0.137 - 0.053 - 0.195 - 0.196 - 0.051 - 0.203 - 0.142 - 0.131 - 0.091 - 0.254 - 0.134transfers) [15.82]*** [24.62]*** [5.64]*** [7.51]*** [2.86]*** [20.4 2]*** [29.92]*** [4.31]*** [19.94]*** [10.63]*** [12.47]*** [8.45]*** [16.28]*** [13.93]***Low income 0.085 - 0.025 0.097 0.094 0.032 0.223 0.018 - 0.043 0.171 0.039 0.190 0.012 0.135 0.188(b.t) [4.48]*** [3.95]*** [5.14]*** [3.31]*** [1.55] [15.46]*** [2.22]** [3.02]*** [12.53]*** [2.40]** [14.18]*** [0.90] [6.29]*** [15.33]***House OwnerSingle0.032 - 0.052 - 0.034 0.034 - 0.010 - 0.086 - 0.060 0.111 - 0.064 - 0.071 - 0.297 - 0.206 - 0.144 - 0.296[1.39] [8.23]*** [0.78] [1.55] [0.49] [4.36]*** [4.48]*** [8.45]*** [5.74]*** [3.74]*** [26.61]*** [8.53]*** [9.65]*** [28.71]***- 0.385 - 0.370 - 0.128 - 0.317 - 0.311 - 0.269 - 0.128 - 0.367 0.016 - 0.163 - 0.417 - 0.232 0.044 - 0.416[7.02]*** [34.29]*** [3.50]*** [6.09]*** [8.56]*** [5.47]*** [8.46]*** [13.99]*** [0.23] [3.89]*** [13.41]*** [6.71]*** [0.09] [13.91]***0.105 - 0.007 0.060 0.522 0.228 0.381 0.152 0.041 0.277 0.409 0.218 0.190 0.516 0.193Single with child [1.93]* [0.38] [1.38] [8.19]*** [4.19]*** [6.72]*** [6.51]*** [0.96] [4.26]*** [5.70]*** [4.10]*** [ 4.55]*** [1.90]* [3.71]***1 child0.027 0.133 - 0.011 0.449 0.390 0.543 0.151 0.328 0.450 0.460 0.499 0.374 0.521 0.498[0.92] [10.89]*** [0.38] [7.52]*** [10.20]*** [10.49]*** [10.04]*** [12.31]*** [8.74]*** [13.90]*** [15.26]*** [17.08]*** [12.40]*** [15.26]***0.002 0.193 - 0.018 0.456 0.385 0.463 0.139 0.282 0.420 0.433 0.463 0.413 0.4822 children[0.06] [13.42]*** [0.81] [8.92]*** [9.81]*** [10.90]*** [11.79]*** [8.94]*** [9.19]*** [15.69]*** [16.10]*** [21.15]*** [19.05]***3 children0.025 0.1544+ children[0.11] [1.06]0.351[1.75]*2 household - 0.198 - 0.135 - 0.049 0.037 - 0.178 0.108 - 0.013 - 0.183 0.216 0.155 - 0.274 0.029 0.256 - 0.281members [3.51]*** [9 .04]*** [1.21] [0.60] [4.38]*** [2.10]** [0.68] [4.55]*** [3.18]*** [3.28]*** [7.67]*** [0.70] [0.51] [8.17]***- 0.213 - 0.150 - 0.062 0.093 - 0.182 0.216 - 0.013 - 0.222 0.488 0.119 - 0.290 0.061 0.288 - 0.2943 hh members [3.25]*** [9.86]*** [1.29] [1.30] [3 .76]*** [3.45]*** [0.57] [5.88]*** [6.84]*** [2.25]** [7.65]*** [1.35] [0.61] [8.12]***4 hh members- 0.196 - 0.161 0.111 0.235 - 0.113 0.350 0.050 - 0.228 0.557 0.283 - 0.380 0.221 0.420 - 0.393[2.44]** [8.00]*** [1.71]* [2.33]** [1.74]* [3.67]*** [1.95]* [4.99]*** [4.82]*** [4.85]*** [6.21]*** [4.65]*** [1.03] [6.48]***5 hh members- 0.149 - 0.131 0.190 0.229 0.006 0.345 0.168 - 0.197 0.503 0.287 - 0.280 0.273 0.461 - 0.267[1.79]* [5.66]*** [1.94]* [1.90]* [0.07] [2.92]*** [4.80]*** [4.24]*** [4.89]*** [4. 47]*** [3.66]*** [5.32]*** [1.21] [3.51]***6 hh members- 0.139 - 0.199 0.106 0.253 - 0.041 0.500 0.064 - 0.210 0.576 0.319 - 0.336 0.246 - 0.347[1.57] [10.54]*** [1.46] [2.35]** [0.58] [5.82]*** [2.22]** [4.66]*** [7.06]*** [5.46]*** [6.07]*** [4.84]*** [6.59]***7 hh members- 0.093 - 0.180 0.328 0.479 0.127 0.564 0.173 - 0.173 0.566 0.477 - 0.304 0.436 - 0.311[1.01] [8.94]*** [4.00]*** [4.77]*** [1.64] [7.54]*** [5.51]*** [3.58]*** [9.33]*** [9.12]*** [5.22]*** [9.27]*** [5.59]***Notes: z statistics in brackets; * significant at 10percent; ** significant at 5percent; *** significant at 1percent; + the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants; ++ migrants identified by country of birth; the EU-SILC does not distinguish between EU-25 and extra EU-25 migrants. Low income variable defined as equivalizedincome lower than 60 percent of median income; High income variable defined as equivailzed income greater than 4/3 of median income.fRDB 42

Table A4. Net fiscal position of Households: incidence of individual characteristics(1) (2) (3) (4) (5) (6) (7) (8) (9)AT BE CZ DE + DK ES FI FR HUEU25 Migrant House 1066.60 3809.16 766.27 -461.78 688.76 16821.53 419.56 -3003.70[0.80] [2.49]** [0.52] [0.54] [0.31] [1.43] [0.68] [3.45]***Extra EU25 Migrant House 1914.76 53182.25 243.87 746.16 -4312.91 2147.91 -4790.31 -242.97 918.12[3.97]*** [1.89]* [0.37] [0.89] [6.30]*** [3.80]*** [5.86]*** [0.45] [1.36]Mixed Household 3582.30 -756.34 326.59 546.81 794.38 1767.21 -959.60 2080.39 -356.84[4.77]*** [0.96] [0.38] [0.55] [1.03] [1.44] [1.67]* [3.04]*** [0.55]Male -745.13 3165.56 -381.46 -427.15 1065.55 89.70 358.94 -615.70 195.58[2.17]** [2.94]*** [2.97]*** [1.48] [4.55]*** [0.33] [1.55] [2.15]** [1.24]Age 426.53 1790.33 116.78 580.75 -68.60 359.06 462.74 613.59 148.52[10.16]*** [6.02]*** [6.60]*** [14.47]*** [2.10]** [8.40]*** [14.14]*** [15.80]*** [7.50]***Age^2 -6.19 -18.93 -1.82 -8.61 -0.88 -4.74 -6.24 -7.80 -2.28[15.31]*** [6.04]*** [10.33]*** [21.87]*** [2.68]*** [11.64]*** [19.36]*** [20.98]*** [12.28]***Secondary Education -1287.93 1088.56 -412.74 -1562.70 119.01 -946.06 442.78 -1958.79 -569.21[4.89]*** [0.87] [3.20]*** [5.65]*** [0.64] [3.34]*** [2.10]** [8.09]*** [6.25]***Tertiary Education 777.64 8448.16 1883.44 -764.67 3907.77 942.27 5310.38 3915.59 3164.43[1.97]** [4.41]*** [8.13]*** [2.47]** [14.36]*** [2.75]*** [20.76]*** [9.79]*** [8.55]***High income (before transfers) 21429.97 25663.52 6294.90 18918.46 21819.80 12044.18 23459.03 24047.17 7230.92[70.55]*** [17.84]*** [39.44]*** [72.49]*** [61.39]*** [43.39]*** [66.49]*** [84.91]*** [51.10]***Low income (b.t) -16871.70 -18974.00 -5216.36 -18478.10 -20490.10 -8108.95 -17454.40 -17868.50 -2874.97[59.70]*** [15.54]*** [43.12]*** [65.90]*** [93.04]*** [25.96]*** [100.75]*** [63.15]*** [25.92]***House Owner 1189.53 508.27 -101.36 1222.47 2861.41 41.94 1894.42 1625.59 349.25[4.18]*** [0.86] [0.86] [6.02]*** [14.73]*** [0.12] [6.55]*** [7.04]*** [1.79]*Single -831.65 -4001.83 312.12 1668.59 1367.93 2643.87 -1004.11 901.20 249.78[1.04] [1.36] [0.98] [2.48]** [1.77]* [4.70]*** [0.87] [1.50] [0.80]Single with child -3492.02 -932.98 -895.28 -2891.15 -1215.39 -162.38 -4741.12 -639.56 -1783.64[3.94]*** [0.46] [3.08]*** [4.84]*** [1.62] [0.28] [5.68]*** [1.09] [5.55]***1 child 1996.15 -1141.92 1025.19 1101.04 4111.16 3504.05 7504.46 2905.47 288.96[2.96]*** [1.09] [3.84]*** [1.80]* [5.54]*** [7.79]*** [7.35]*** [5.71]*** [1.09]2 children 1977.94 -2895.28 1555.92 377.51 5196.56 2227.72 6306.18 3055.54 1209.93[2.45]** [1.89]* [3.97]*** [0.63] [5.60]*** [2.74]*** [9.10]*** [3.53]*** [1.28]3 children 865.78 2145.81 -9692.76 -21033.40 6649.06 -1546.21 -5062.40 303.88[0.25] [0.30] [2.69]*** [1.81]* [4.59]*** [0.20] [0.91] [0.56]4 children 3039.90 -9866.17 -462.15 -1642.48 7031.31 4877.22 3642.18 -3414.65[1.25] [2.60]*** [0.51] [0.89] [2.23]** [3.06]*** [1.26] [4.22]***5+ children -10383.20 -21365.70 12003.84 17077.49 7366.83 -10458.90[8.96]*** [2.39]** [3.76]*** [6.34]*** [2.25]** [5.74]***2 household members -1792.42 5363.29 -615.04 -268.71 2684.52 2069.37 -1936.33 -1383.66 -978.90[2.17]** [1.41] [1.90]* [0.40] [3.52]*** [3.88]*** [1.74]* [2.35]** [3.21]***3 hh members -955.30 8441.37 -469.09 179.79 8479.67 999.47 -668.95 505.45 -1209.26[0.94] [2.14]** [1.19] [0.20] [6.82]*** [1.71]* [0.59] [0.59] [2.71]***4 hh members 1240.19 13095.37 593.26 2680.28 6987.47 2589.35 2354.97 3255.88 -501.64[0.85] [3.68]*** [1.12] [2.08]** [4.69]*** [3.78]*** [1.03] [3.07]*** [0.91]5 hh members 3303.67 17458.01 545.31 7616.60 22409.49 2525.62 918.30 1911.02 5.82[2.09]** [3.58]*** [0.79] [2.11]** [2.49]** [2.55]** [0.54] [1.25] [0.01]6 hh members 4716.70 14582.57 1453.42 7713.94 12487.33 177.44 7326.38 9225.74 1362.33[2.80]*** [3.94]*** [2.22]** [5.40]*** [6.98]*** [0.09] [3.01]*** [6.55]*** [1.85]*7 hh members 11772.61 29546.97 3905.20 14390.23 15402.60 1544.96 13307.66 11970.72 4081.44[5.86]*** [7.00]*** [5.40]*** [8.50]*** [4.65]*** [0.34] [5.06]*** [6.83]*** [5.66]***Densly populated area 47.15 -69.64 -132.38 632.97 1463.59 165.91 1459.06 -189.37 163.41[0.15] [0.11] [0.76] [2.98]*** [6.21]*** [0.63] [4.61]*** [0.74] [0.87]Thinly populated area -988.71 -762.42 -178.06 -526.54 -1177.08 -155.36 -767.59 -329.98 -16.81[3.56]*** [0.86] [0.98] [2.01]** [5.84]*** [0.57] [3.71]*** [1.13] [0.09]Year dummies Yes Yes Yes Yes Yes Yes Yes Yes YesCountry dummiesRegional dummies Yes Yes Yes Yes Yes Yes Yes Yes YesConstant 3128.14 -34368.20 902.05 4562.59 14414.69 -6086.82 1023.32 -4283.63 -429.42[2.32]** [5.34]*** [1.68]* [2.83]*** [13.84]*** [3.98]*** [0.69] [2.71]*** [0.60]Observations 17475 10823 12247 30173 21096 12146 37267 32687 15579R-squared 0.66 0.15 0.59 0.65 0.67 0.60 0.59 0.60 0.43Notes: z statistics in brackets; * significant at 10percent; ** significant at 5percent; *** significant at 1percent; + the EU-SILC does notdistinguish between EU-25 and extra EU-25 migrants; ++ migrants identified by country of birth; the EU-SILC does not distinguishbetween EU-25 and extra EU-25 migrants. Low income variable defined as equivalized income lower than 60 percent of medianincome; High income variable defined as equivailzed income greater than 4/3 of median income.43

Table A4 (Continued). Net fiscal position of Households: incidence of individual characteristics(10) (11) (12) (13) (14) (15) (16) (17) (18)IE IS LU NO PL SE SK UK AllEU25 Migrant House 1852.13 126.61 3353.49 2936.72 1871.31 -837.71 426.66 921.52 2462.53[1.71]* [0.09] [4.80]*** [2.28]** [2.23]** [0.93] [0.36] [0.63] [6.35]***Extra EU25 Migrant House -4162.06 -2806.24 3727.97 -1210.19 -932.38 -1180.55 116.23 3942.71 2014.28[3.35]*** [2.31]** [2.42]** [1.16] [1.14] [0.86] [0.17] [2.81]*** [2.71]***Mixed Household -1547.57 -2093.21 2798.16 1340.34 4077.46 -1124.11 222.71 2850.88 3569.92[1.66]* [1.83]* [2.30]** [1.88]* [1.86]* [1.36] [0.66] [2.53]** [9.23]***Male -143.06 1701.65 2448.04 770.03 -113.75 372.32 -71.86 35.10 -101.47[0.32] [3.29]*** [2.49]** [2.21]** [1.65]* [1.42] [0.45] [0.09] [0.84]Age 484.19 1307.16 1353.97 353.94 175.77 261.45 58.11 718.99 693.33[9.84]*** [13.37]*** [13.09]*** [7.52]*** [21.10]*** [7.09]*** [3.66]*** [12.16]*** [41.58]***Age^2 -5.79 -15.51 -16.42 -5.28 -2.27 -3.40 -1.06 -9.02 -8.49[13.40]*** [16.44]*** [16.38]*** [11.02]*** [29.31]*** [9.26]*** [6.28]*** [15.16]*** [51.23]***Secondary Education -305.16 815.32 -2445.96 299.13 -670.43 1127.40 -342.73 -2091.79 -1885.72[0.84] [2.03]** [3.92]*** [0.98] [16.53]*** [4.27]*** [2.30]** [5.83]*** [17.52]***Tertiary Education 8070.79 11294.12 4732.94 5463.95 1376.27 7431.96 -151.64 2734.85 2536.51[12.89]*** [14.57]*** [4.73]*** [12.51]*** [12.81]*** [18.58]*** [0.98] [4.28]*** [15.79]***High income (before transfers) 20620.69 24157.40 28332.72 25895.01 4476.75 26916.31 4559.44 26001.06 18698.59[49.60]*** [33.28]*** [38.60]*** [69.56]*** [82.56]*** [75.47]*** [43.24]*** [52.86]*** [139.48]***Low income (b.t) -10427.00 -18556.70 -19421.60 -24443.40 -2311.43 -20819.50 -3646.77 -13990.30 -16717.30[27.97]*** [32.24]*** [27.97]*** [85.52]*** [44.17]*** [89.82]*** [37.41]*** [40.24]*** [149.03]***House Owner 2107.57 2394.72 -437.21 858.08 443.48 2272.22 58.99 3274.36 1098.12[5.36]*** [4.44]*** [0.77] [2.69]*** [4.38]*** [8.95]*** [0.40] [11.38]*** [11.47]***Single 931.19 -480.87 3300.16 1105.02 405.16 -6772.40 1710.34 -177.57 -874.83[0.93] [0.31] [1.98]** [1.33] [3.38]*** [10.35]*** [4.75]*** [0.11] [3.14]***Single with child -3465.21 -3888.20 -384.51 -3324.35 -1100.98 -5098.06 1045.21 -6815.40 -4333.44[3.89]*** [2.77]*** [0.21] [3.87]*** [9.29]*** [8.22]*** [0.76] [4.57]*** [16.63]***1 child 4846.24 3782.54 2986.75 5566.11 761.68 4997.44 497.57 1401.33 3629.59[3.72]*** [3.98]*** [2.15]** [6.88]*** [6.72]*** [9.56]*** [1.69]* [1.23] [15.39]***2 children 2350.07 123.51 4655.82 5986.65 619.53 5490.77 729.34 7119.42 5566.93[2.63]*** [0.11] [2.70]*** [6.07]*** [5.91]*** [6.06]*** [2.62]*** [3.89]*** [15.02]***3 children 2141.43 -2024.79 4809.72 3456.86 171.76 655.18 -40.33 2023.64 4322.82[0.95] [0.37] [0.71] [0.44] [0.20] [0.22] [0.03] [0.23] [2.36]**4 children -5751.79 -3228.78 1729.83 4493.14 789.76 10389.84 122.46 39301.51 11575.61[2.16]** [1.51] [0.67] [1.73]* [3.22]*** [4.27]*** [0.56] [1.53] [3.50]***5+ children 1657.09 4595.18 112.64 -19369.20 -915.33 6595.98[0.38] [2.11]** [0.15] [3.41]*** [1.46] [2.17]**2 household members -3510.82 1536.10 -557.34 402.57 -52.74 -4937.78 -173.23 -2567.78 -3051.43[3.69]*** [1.05] [0.36] [0.48] [0.41] [7.63]*** [0.49] [1.62] [11.10]***3 hh members -5340.54 5388.57 -1934.53 5950.91 -130.54 -1711.15 282.56 -860.50 -3537.55[4.33]*** [3.01]*** [0.93] [4.78]*** [0.79] [1.78]* [0.43] [0.42] [9.66]***4 hh members -3643.98 7075.66 2440.85 6082.13 478.48 -296.96 712.11 -228.75 -2590.56[2.00]** [3.23]*** [0.89] [3.90]*** [2.25]** [0.25] [1.30] [0.07] [5.23]***5 hh members -3061.91 12416.99 4617.81 8036.88 253.22 972.28 625.71 -582.24 -4250.07[1.41] [4.87]*** [1.46] [3.24]*** [1.19] [0.49] [0.81] [0.20] [6.64]***6 hh members -2528.67 13546.82 2610.26 14817.11 500.75 8534.98 1778.09 8.48 -2057.36[1.20] [5.52]*** [0.81] [7.87]*** [2.14]** [5.21]*** [3.49]*** [0.00] [3.72]***7 hh members -881.89 19277.20 9027.58 20627.51 969.20 13431.85 3062.27 268.97 -2392.46[0.38] [7.32]*** [2.50]** [8.84]*** [4.02]*** [6.11]*** [5.89]*** [0.08] [3.76]***Densly populated area 93.19 -144.45 1869.85 38.96 384.01 355.09 -444.08 -354.51[0.16] [0.23] [5.81]*** [0.44] [0.73] [3.57]*** [1.15] [3.33]***Thinly populated area -1598.93 -1143.77 -1112.45 31.44 -292.57 -2751.59 -28.67 2706.81 -449.94[3.80]*** [2.61]*** [1.48] [0.10] [3.38]*** [7.46]*** [0.36] [1.15] [3.52]***Year dummies Yes Yes Yes Yes Yes Yes Yes Yes YesCountry dummiesYesRegional dummies Yes Yes Yes Yes Yes Yes Yes YesConstant -11096.00 -16292.70 -26115.70 4811.71 -2702.02 12339.69 111.20 -9550.80 -2174.66[6.32]*** [6.96]*** [9.07]*** [3.64]*** [9.99]*** [12.16]*** [0.22] [4.26]*** [4.46]***Observations 18815 9919 12663 20177 32536 20360 11875 20030 335868R-squared 0.45 0.59 0.62 0.63 0.56 0.65 0.48 0.41 0.45Notes: z statistics in brackets; * significant at 10percent; ** significant at 5percent; *** significant at 1percent; + the EU-SILCdoes not distinguish between EU-25 and extra EU-25 migrants; ++ migrants identified by country of birth; the EU-SILC does notdistinguish between EU-25 and extra EU-25 migrants. Low income variable defined as equivalized income lower than 60percent of median income; High income variable defined as equivailzed income greater than 4/3 of median income.fRDB 44

European Integration ConsortiumIAB, CMR, fRDB, GEP, WIFO, wiiwLabour mobility within the EU in the context of enlargement and the functioningof the transitional arrangementsVC/2007/0293Deliverable 6CMR, Centre of Migration Research, University of WarsawAgnieszka Fihel†, Paweł Kaczmarczyk†, Nina Wolfeil††, Anna śylicz†Brain drain, brain gain and brain wasteAbstractDespite a rapidly growing scholarly interest in skilled migration generally, there is as yetonly limited evidence on the extent and effects of the outflow of skilled workers from theNew Member States (NMS), the so-called “brain drain”. Economic theory predicts that abrain drain can have positive or negative impacts on the sending country, and so anyassessment of the actual effect remains but an adverse impact of skilled migration uponthe sending country cannot be ruled out a priori. The assessment of its effects becomesan empirical issue. Drawing robust conclusions from the empirical evidence is difficult,partly because of severe data limitations, but it is important because the lack of evidenceis matched by a widespread popular perception that skilled migration represents asignificant economic problem for NMS.The aim of this report is to provide an assessment of the scale and impact of highlyskilled migration from the NMS. We draw mainly on Labour Force Surveys from each ofthe EU27 countries in 2006. The statistical analysis confirms that migrants from the NMSare positively selected with respect to education. This education differential is not simplythat the result of differences in the age structure. However, claims about the size of theoutflow of skilled workers may have been overemphasized. With regard to the impacts ofhighly skilled migration, we refer to both static effects (drain effect) and dynamic effects(brain effect). We show that the drain effect is rather limited and, at least in case ofPoland, the most important sending country, recent mobility is to be understood in termsof brain overflow resulting from an oversupply of highly skilled labour. The brain effect,however, also appears to be limited, mainly due to the relatively low rates of return tohuman capital observed in main destination countries.† Centre of Migration Research, University of Warsaw†† Department of Geography and Regional Research, University of ViennaThe views and opinions expressed in this publication are those of the authors and do not necessarilyrepresent those of the European Commission.

Contents1 Introduction ...................................................................................................... 12 Methodological Issues......................................................................................... 23 Theoretical background....................................................................................... 44 Scope of the phenomena – overview of highly skilled migration from NMS ................ 75 Contextual issues............................................................................................. 126 High-skilled mobility and its impact on sending countries...................................... 186.1 Poland – recent migration and mobility of the highly skilled........................... 196.2 Selectivity of the recent outflow from Poland ............................................... 246.3 Drain effect or brain effect? ....................................................................... 296.3.1 Drain effect ................................................................................. 296.3.2 Brain effect ................................................................................. 317 Case studies.................................................................................................... 407.1 Mobility of health care professionals ........................................................... 407.2 Mobility of students .................................................................................. 438 Conclusions..................................................................................................... 489 References ...................................................................................................... 51

1 IntroductionThere is yet only limited evidence on the extent and effects of the outflow of skilledworkers from the New Member States (NMS), despite a rapidly growing scholarly interesttowards skilled migration. Nevertheless, there is widespread agreement that there hasbeen a significant outflow of highly skilled workers from the area (Balaz et al., 2004), anda perception that the outflow has had a negative impact on the average human capitalendowment of the domestic workforce, and a resulting detrimental effect on economicgrowth (Radu 2003; Straubhaar and Wolburg 1999; Wolburg 1996; Wolburg and Wolter1997; Salt and Findlay, 1989; Salt 1997, 2001).A major obstacle to the analysis of migration from the NMS countries is the lack, or thepoor reliability, of statistical data. 1 This report represents an important step towards amore solid understanding of the scale of skilled migration from NMS, and of its impact onsending countries. The impacts of recent highly skilled mobility are difficult to estimate.This is not only due to data limitations, but also to the complexity of the phenomenon. Inthis report we attempt to assess both positive and negative features of the phenomenonby looking at both static and dynamic aspects.We start by assessing the scale of highly skilled mobility in the post-2004 period, with ananalysis of the selection of migrants from the population as a whole. We argue that anintegrated approach to the analysis of skilled migration is needed, and we suggest thatharmonised Labour Force Surveys can be used as a consistent data source in bothmigrant-sending and migrant-receiving EU countries.The analysis then concentrates on exploring two main issues. First, does the structure ofthe recent outflow merely reflect the composition of the sending countries’ populations(in terms of age and education)? Is the scale of the drain of highly skilled thereforeexaggerated for this reason? Second, how have the institutional changes related to EUenlargement(particularly the introduction of Transitional Arrangements) influenced post-2004 migratory patterns? How have these changes influenced the patterns of highlyskilled mobility?It is important to note that we do not provide an account of the overall economic impactof migration for NMS, as several relevant aspects — such as those defined by Grubel andScott (1966) — exceed the scope of this report. For example, we will neither deal withthe ensuing flow of remittances, nor with their possible impact on trade patterns, onforeign direct investment flows or on technological spillovers. Still, we argue that thisresearch represents an important step forward in a context where there is a very limitedunderstanding, and this may also lay the ground for further research on the economicimpact of migration from NMS.1 Poland represents an isolated exception. Official sources of information, such as register and census data,tend to greatly underestimate the scale of migration, and especially the scale of migration by highly skilledindividuals. This can lead to underestimates of the scale of the brain drain (Fihel et al., 2006).CMR 1

The structure of the report is as follows. Section 1 defines the scope of the analysis andSection 2 provides a description of the main methodological challenges that have to beaddressed. Section 3 provides a brief overview of the relevant theoretical and empiricalliterature on the impact of skilled migration upon sending countries. Section 4summarises the patterns of skilled migration from the NMS, using both existinginternational datasets and also Labour Force Surveys conducted in origin and destinationcountries. Section 5 considers the background of the observed migration processes.Section 6 provides, firstly, an in-depth analysis of the highly skilled mobility andselectivity issues presented for Poland, and, secondly, offers an assessment of the impactof recent mobility on source countries. Section 7 includes two case studies relevant to thestudy and dealing with mobility of medical professionals, and students. Section 8concludes.2 Methodological IssuesThe aim of this report is to provide an integrated assessment of the scale of highly skilledmigration from the NMS and its impacts on source countries. As a first step, this requiresreliable data on the size of the outflow of skilled workers, and then to use these data toprovide an assessment of whether the consequences of this outflow are beneficial ordetrimental for NMS.It is important to note that the heading “NMS” hides significant differences with respectto the aggregate size of current migration flows. 2 The limited size of pre- and postaccessionmigration flows from the Czech Republic, Hungary and Slovenia 3 means thatconcerns about the negative effects of the outflow are very limited. We therefore focusmore on those countries that are experiencing larger outflows. In particular Poland hasexperienced large outflows and has migration data of relatively high quality andreliability.A simple comparison of the skill composition of current migration flows with thecorresponding composition of the resident population represents an important first step— and even a challenging one given existing data limitations, but it cannot support anyconclusion about the detrimental or beneficial character of current migration patterns. 4Even when one observes that skilled workers are over-represented among migrants,further evidence is needed before one can conclude that migration is reducing the humancapital endowment of the country of origin. For example, a positive selection of migrants2 See Deliverable 2 for information on aggregate flows from NMS.3 For example, the proportion of Hungarian R&D personnel working abroad for more than 6 months wasestimated at 2% in 2001, but immigration led the net balance of R&D personnel in Hungary to be close to zero(Inzelt, 2003).4 Note that the higher propensity of younger individuals to migrate means that any direct comparison of theskill composition of the migrant and resident population is going to be influenced by the different age structureof the two groups. When average educational achievement is negatively related to age — as it is the case ofthe NMS (see Section 3), then it is informative to estimate cohort-specific measures of the skill composition ofthe migrant and of the resident population, in order to gain a better understanding of the extent to whichmigration is selective with respect to education.CMR 2

with respect to education may lead to an substantial increase in educational investmentsdetermined by the migration prospect itself, and this is precisely the critical factor thatsimple descriptive statistics on skill composition fail to reveal. 5 This factor can be at leastpartly captured by some data on the evolution of tertiary and secondary enrolment rates(see Section 4).With respect to the definitional issues, there is no international system of recoding skilledmigration, as there is no accordance on what the term “skilled” should mean (Lowell andFindlay, 2002; Salt, 1997). The term “skilled” is usually interpreted in the literature interms of the formal level of education and qualifications, which is relatively easy tomeasure (e.g. in years of formal education), contrary to other possible definitions of skills(Csedö, 2008). It is common practice to identify skilled individuals with highly educatedworkers, as it is hard to gather reliable information on the extent of “on-the-job”experience, 6 not to mention the difficulty in measuring something as fuzzy as innateability, although these two components are admittedly important factors in determiningthe skill level of a worker. In this research we will adhere to a definition of skilled workerbased on years of formal education, and — as Dumont and Lemaître (2005) and Docquierand Marfouk (2004) — we move beyond a strict dichotomy between skilled and unskilledworkers, attempting to provide statistics also on the migration of workers that hold asecondary-degree or vocational training. 7 However, in accordance with the wide typologyof skilled migrants presented by Salt (1997), we will extend the analysis to incorporatepost-secondary students, a group which, although not yet formally qualified, forms partof the phenomenon of (potential) skill flows. The general analysis of individuals withtertiary education or gaining tertiary education will be supplemented by a case study ofmedical health care workers.We will use the most commonly used meaning of the term brain drain, i.e. it will beunderstood as selective migration of well educated people (typically from less- towardsmore-developed countries. The term “brain drain” is sometimes used with regard to theimpacts of highly skilled mobility, i.e. in such cases when emigration of tertiary educatedpersons for permanent or long stays abroad reaches significant levels, visible in theeconomy, and is not offset by welfare gains or feedback effects from remittances,technology transfer, investments, or trade. In this case the negative effects of the flow onthe economy of the sending country dominate. On the other hand, a “brain gain” occurs ifthe sending country experiences net benefits (for example in terms of welfare) from theemigration of the skilled. A positive effect may dominate as the possibility of working5 By similar arguments, one could argue that if a positive selection of migrants with respect to education ismatched by high domestic unemployment rates for qualified people, then this signals the existence of a likelyoverinvestment in education that could not be sustainable anyway, so that an eventual later decline ineducational investments should not be attributed to current migration patterns.6 OECD (2002) proposes a definition of highly skilled workers that includes workers that completed tertiaryeducation, and workers that did not complete tertiary education but are employed in occupation where such aqualification is usually required. This definition that captures the idea of skill acquisition through “on-the-job”experience, is data-demanding and thus hard to implement.7 Whenever possible, we will attempt to distinguish different types of mobility (such as short-term migration orrepeated migration spells), as these could have different economic impacts on sending countries.Distinguishing between these different types, however, is often not possible with the available data.CMR 3

abroad for higher wages may create an incentive to pursue education, which in turn mayraise domestic educational levels and stimulate economic growth (Stark, 2004).With regard to consequences of the outflow of skilled workers for the sending country, weconsider both static and dynamic aspects captured by the brain effect/drain effectdichotomy as understood by Beine et al. (2001), as well as other labour market effectssuch as the brain overflow (see theoretical discussion and definitions of the possibleeffects of the brain drain in Section 2). As far as migrants themselves are concerned, animportant dimension of the analysis of the effects of the mobility of high skilled labourarises from considering the extent to which the migrants’ qualifications are “adequately”employed in the receiving country. When highly skilled workers migrate into forms ofemployment not requiring the application of their skills and experience, brain waste mayoccur (Salt, 1997). We argue that this perspective is particularly useful in the context ofthe project.An attempt to assess the impacts of the outflow of skilled workers from the NMS posesseveral analytical challenges. The first one relates to the statistical assessment of thephenomenon. Official sources of information, although in principle consistent with adefinition based on years of formal education, offer a frail ground to analyze migrationfrom the NMS. Population registers fail to record the migrants who left the country butdid not modify their residence status, whereas population censuses most often provide noinformation about the individuals who left the country after the previous round of thecensus. 8 We argue that harmonised LFS can represent a way to overcome current datalimitations, and to derive sound estimates about the skill composition of migration flowsfrom NMS, and to implement the integrated approach that we deem as necessary toachieve a proper balance of the impact of the impact of migration upon the human capitalendowment of these countries.The next problem is the difficulty of defining an adequate counterfactual against whichwe can assess the impact of migration upon human capital formation This counterfactualshould be informative about the human capital endowment that the NMS would beexperiencing in a no-migration (or in a pre-accession) scenario. A combination ofdifferent data sources shall help to achieve a better approximation of a reliablecounterfactual.3 Theoretical backgroundThe economic literature on the consequences of skilled migration for sending countries isusually divided into two distinct parts. The first dates back to the 1970s, and produced atheoretical consensus that regarded the impact of skilled emigration as detrimental. Thesecond, from the 1990s, reversed the earlier theoretical consensus, and attempted to8 A further problem with data sources is that national statistics differ across countries with respect to theadopted definition of a skilled worker; Poland, for instance, defines a highly skilled individual as a universitygraduate who has also acquired at least a M.A. degree.CMR 4

support its prediction of a possible brain gain with econometric analysis on newlycollected international migration data. In fact this distinction is somewhat artificial. Theearlier literature already contained elements that were later developed by the so callednew economics of brain drain, and some of the more recent studies have provided sometheoretical and empirical results that are actually closer to the pessimistic conclusions ofthe earlier literature.Although a paper by Grubel and Scott (1966) had emphasized that positive feedbackeffects — in terms of remittances and technology acquisition — had the potential to offsetthe losses caused by the migration of skilled workers, the emphasis of the early literaturewas on the losses rather than the gains of the brain drain. In addition, the fiscal costs ofproviding public education to the migrants, and the existence of intra-generationalpositive educational externalities implied that the brain drain had detrimental welfareeffects on non-migrants (e.g. Bhagwati and Hamada, 1974). 9A central innovation of the new economics of the brain drain is to model migration as aprobabilistic event, i.e. the outcome of a lottery where the would-be migrant has apositive probability p of actually migrating, where p

positive feedback effects suggested by Grubel and Scott (1966). In this report we willfollow this particular line of reasoning on the effects of highly skilled mobility.Early attempts to gather internationally comparable data on skill-specific migration rates(Barro and Lee, 1993; Carrington and Detragiache, 1998) provided a very limitedcoverage of NMS. More recently they have been extended by Dumont and Lemaître(2005) and Docquier and Marfouk (2004), later adjusted by Beine et al. (2007) to correctfor skill acquisition abroad. This has offered the opportunity to assess the empiricalvalidity of the theoretical predictions of the new economics of the brain drain, and theresults are broadly consistent with the idea that there is a possibility for the beneficialbrain drain to occur (Beine et al., 2001, 2008; Docquier et al., 2008). Individualcountry-studies, such as McKenzie and Rapoport (2008) on Mexico, have shown that thereverse can occur, with a reduction in educational attainment in the areas characterizedby higher emigration rates. Such a finding may be explained by the fact that migrantsexperience very low (or even zero) returns to human capital in the destination countries.High-wage countries need not be (at least for the migrants) high-return to human capitalcountries. Recent theoretical contributions (Egger and Felbermayr, 2007; Brücker et al.,2007) have shown that the optimistic conclusions of the newer literature crucially hingeon this assumption, and the scope for a beneficial brain drain can be substantiallyreduced.This latter point has a critical methodological implication for the study. The analysis ofthe impact of migration upon migrant sending countries cannot be separated from ananalysis of the labour market performance of migrants in destination countries. Theoccurrence of so-called brain waste (e.g. Mattoo et al., 2005), a situation where migrantsare employed in occupations for which they are over-qualified, influences the impact ofmigration itself upon human capital formation in migrant sending countries. This need foran integrated approach to the analysis of migration breaks down the usual separationbetween analysis focused either on migrant-destination countries or on migrant-sendingcountries. The research on the economic effects of migration upon the countries of originthus needs to draw insights also from one of the main components of the literature upondestination countries, namely the analysis of the labour market assimilation of themigrants (e.g. Chiswick, 1978).A slightly different category of effects, albeit impossible to disentangle form the previousones, is brain overflow. This effect occurs when there is (intentional or unintentional)oversupply of educated professionals in the sending country. In such a case, migration ofthe highly skilled occurs at low or zero alternative costs, and reduces the labour marketsupply-demand inequality in the sending country. Additionally, when a brain overflowoccurs, both the drain and the brain effects are limited. The drain effect is unimportantbecause of the probable high domestic unemployment rate for skilled workers. The braineffect is unimportant because domestic labour market conditions do not adequatelyreward skill formation.CMR 6

4 Scope of the phenomena – overview of highly skilled migration fromNMSOne goal of this research project is to move beyond current data sources and employmicro data drawn from the Labour Force Surveys (LFS). We begin by comparingdescriptive statistics from the LFS with those from the most common international datasources, that is Dumont and Lemaître (2005), Docquier and Marfouk (2004) and Beine etal. (2007). These datasets gather information about foreigners and foreign-bornindividuals from censuses, or administrative registers, from OECD member countries, andthen compare these data with information on the resident population in migrant sendingcountries to break down aggregate migration figures by skill composition, and to deriveestimates of skill-specific emigration rates. 12The term emigration rate needs to beinterpreted with caution, as what Dumont and Lemaître (2005) and Docquier andMarfouk (2004) actually define as such is the ratio of the population born in – or holdingthe citizenship of - a given country and currently residing in OECD countries over thepopulation residing in the home country. 13Thus, this measure — both in its aggregateand in its skill-specific versions — captures migration flows that have been accumulatingover a long period of time, and it thus conveys relatively limited information on therecent pattern of migration. 14With these caveats in mind, Table 1 reports the estimates of the emigration rates, brokendown by educational level, from the NMS for the year 2000, 15while Table 2 displaysinformation about the skill composition — also referred to as the selection rate — ofcurrent migrants to OECD countries from NMS.12 We refer to Dumont and Lemaître (2005) and Docquier and Marfouk (2004) for more details, and for anunderstanding of the differences between these two datasets.13 Neither Dumont and Lemaître (2005) nor Docquier and Marfouk (2005) are able to distinguish betweenmigrants who completed their formal education at home and those who studied abroad (while Beine et al.,2007 estimate a model that is meant to adjust for this), and this further limits the possibility to use these datato make any inference about the extent – and the character – of brain drain from NMS.14 For instance, according to Dumont and Lemaître (2005), over 1.2 million Polish were living abroad in 2000,and an estimated 328,000 had completed tertiary education; leaving aside concerns about data quality, it hasto be stressed that this reflects the whole Polish post-war migration history.15 The adjustment introduced by Beine et al. (2007) to correct for education acquired abroad does notsignificantly change the estimated emigration rate for highly educated migrants from NMS.CMR 7

Table 1: Emigration rates from NMS, year 2000Level of qualificationCountry Low Medium High TotalBulgaria 9.1 6.3 6.6 7.6Czech Republic 4.2 1.9 10.4 3.7Estonia 4.3 4.9 11.5 6.0Latvia 1.8 2.6 8.8 3.5Lithuania 6.2 3.6 8.6 5.6Hungary 2.7 3.8 13.2 4.4Poland 3.4 2.8 14.1 4.4Romania 4.6 2.0 11.8 3.7Slovenia 7.1 4.3 11.5 6.7Slovakia 10.1 9.1 16.7 10.4Notes: Docquier and Marfouk (2005) count as migrants all foreign-bornindividuals aged 25 and above who live in an OECD member country; a highlevel of qualificationcorresponds to at least to tertiary education, a medium levelto secondary education.Source: Docquier and Marfouk (2005)The emigration rate is highest amongst highly skilled workers in all countries exceptBulgaria. Table 2 shows estimates of the proportion of migrants in each skill category.Docquier and Marfouk (2005) estimate that the highly-educated between 16% and 51%of all migrants. Dumont and Lemaître (2005) estimate rather lower shares, between 15%and 37%.Table 2: Selection rates of emigrants from NMS, year 2000Selection rateCountry Low Medium High HighBulgaria 52.8 30.8 16.4 14.5Czech Republic 39.4 27.6 33.1 24.0Estonia 32.0 27.9 40.1 32.0Latvia 22.3 26.4 51.2 37.4Lithuania 42.7 28.2 29.1 22.1Hungary 31.7 29.2 39.1 28.7Poland 30.0 30.4 39.5 25.7Romania 34.5 34.2 31.3 26.3Slovenia 47.1 26.8 26.1 17.5Slovakia 37.9 42.2 20.0 13.8Source: Docquier and Marfouk (2005); Dumont and Lamaître (2005)As a complimentary data source we use the data from the EUROSTAT LFS conducted inthe year 2006 in all the EU27 countries. We rely on the LFS conducted in the NMS toestimate the skill structure of the resident native population, to use it as a benchmarkagainst which the corresponding composition of the migrant population from eachcountry can be compared (see Table 3). For seven out of ten countries — the exceptionsbeing Estonia, Lithuania and Slovenia — the share of highly skilled among migrants isCMR 8

higher than the corresponding share among the resident population. With respect tomedium skilled workers, the picture is more mixed, as for half of the countries mediumskilled workers are overrepresented among the resident population, while for the otherhalf the reverse occurs.Table 3: Skill composition of native population and of emigrants from NMS to EU15countries, year 2006Resident population, nativesMigrant populationMigrant population, age adjustedCountry low medium high low medium high low medium highBulgaria 31.3 50.8 17.9 24.0 48.5 30.2 24.2 45.9 29.1Czech Republic 16.7 72.1 11.2 14.8 48.8 36.4 19.3 48.3 34.9Estonia 22.7 49.8 27.5 35.8 49.4 14.8 26.8 43.1 19.2Hungary 27.6 57.4 15.0 9.0 38.7 35.4 9.3 66.2 23.9Lithuania 21.3 56.1 22.6 25.9 38.7 35.4 22.1 39.3 39.8Latvia 25.4 56.8 17.8 - - - - - -Poland 21.3 64.1 14.7 26.1 48.2 25.7 21.5 47.6 26.8Romania 33.0 57.5 9.6 33.2 53.3 13.5 33.2 53.1 13.6Slovenia 23.4 58.8 17.8 34.4 59.2 6.4 33.2 60.0 6.6Slovak Republic 19.2 69.1 11.7 18.2 62.6 19.2 17.3 63.3 21.3Note: the age adjusted selection rates are computed applying the age distribution of the resident populationto migrants’ age-specific skill composition.Source: Own Calculations based on Eurostat Labour Force Survey.However, these comparisons are influenced by the possibly different age structure of thetwo groups. To assess the actual relevance of this confounding factor, the last threecolumns of Table 3 report the skill composition of the migrant population, computed as aweighted average of the skill composition within each one of three age brackets, 16 withweights given by the age structure of the resident population. Such an adjustment doesnot produce a major impact on the estimated skill composition, and the direction of theinduced change in the share of highly skilled people is not constant across countries.The absence of major changes is probably due to the fact that, as Table 4 shows, theshare of the resident population in the younger age brackets is not necessarily lower thanthe corresponding share in the migrant population. The former is actually higher inEstonia, Hungary and Slovenia. In Slovenia the elderly are also over-represented amongmigrants.16 The three age brackets are 15 to 34, 35 to 49 and 50 to 64 respectively.CMR 9

Table 4: Age structure of resident workforce and emigrant population from NMS, year2006Resident population, natives Migrant populationCountry 15-34 35-50 50-64 15-34 35-50 50-64Bulgaria 28.6 43.3 28.1 40.1 49.6 10.3Czech Republic 30.1 40.1 29.8 34.8 47.5 17.7Estonia 35.4 40.0 24.6 29.8 63.3 6.9Hungary 30.0 40.0 30.0 22.3 54.0 23.7Lithuania 32.6 43.2 24.2 45.6 45.9 8.5Latvia 33.0 41.4 25.6 53.8 41.1 5.1Poland 33.6 39.7 26.6 43.5 46.0 10.5Romania 33.6 41.3 25.0 42.0 50.7 7.3Slovenia 29.5 43.2 27.4 25.9 40.4 33.6Slovak Republic 34.5 40.8 24.8 54.2 38.9 7.0Source: authors’ elaboration on EUROSTAT, Labor force surveysIt is important to recall that LFS data, like the data used by Dumont and Lemaître (2005)and Docquier and Marfouk (2004), are stock data. So, for countries with a long-standingmigration history, this data source does not necessarily provide an accurate picture of thecharacteristics of the recent migration process. Keeping in mind these limitations, acomparison of Table 3 with Table 2 shows that — with the exception of Bulgaria — thehare of highly skilled workers among migrants is lower than the corresponding figuresfrom Dumont and Lemaître (2005) and Docquier and Marfouk (2004). Although such acomparison is only suggestive, given the differences across the datasets used, it isnevertheless possible that the claims about the size of skilled migration from the NMSmight have been overstated.Recently published OECD data does not cover the post-accession period. However it ispossible to use this data to explore the duration of the migrant stock in each destinationcountry. Table 5 presents the stock of immigrants born in NMS staying in the OECDcountries around 2000 by education level and duration of stay.CMR 10

Table 5: Stock of immigrants born in NMS in the OECD countries, by education level andduration of stay, around 2000Country ofHighly-skilledresidenceTotal Total More than 10 years 5 to 10 years Up to 5 years UnknownAustralia 116,988 26,616 20,948 2,858 2,369 441Austria 137,151 17,365 12,487 4,878--Belgium 35,866 8,097 4,136 1,495 2,464 2Canada 359,725 138,980 92,370 24,855 19,835 1,920Czech Republic 112,337 6,528 6,528 -- -Denmark 19,068 4,809 3,228 662919 -Finland 2,211 n.a. n.a. n.a.n.a. n.a.France 159,333 39,269 20,589 6,495 6,400 5,785Germany 1,546,414 269,998 254,080 15,9180 0Greece 81,863 11,610 1,180 1,942 4,347 4,141Hungary 70,846 8,750 6,542 2,208- -Ireland 13,281 3,528 147 210 1,896 1,275Italy 149,430 19,315 3,509 3,433 5,854 6,519Luxembourg 2,225 637 149 92373 23Netherlands 5,012 n.a. n.a. n.a.n.a. n.a.New Zealand 5,301 1,527 747 309432 39Norway 13,170 3,632 2,606 582444 -Spain 98,260 16,200 4,520 1,880 9,800 -Sweden 78,985 24,730 17,555 2,190 3,895 1,090Switzerland 58,247 28,235 5,012 1,349 6,957 14,917United Kingdom 51,008 n.a. n.a. n.a.n.a. n.a.United States 849,339 272,959 177,089 46,493 49,377 -OECD - Total 3,966,060 902,785 633,422 117,849 115,362 36,152Note: NMS include Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania,Slovakia, Slovenia,Source: OECD database, based on national population censuses and LFSsFrom Table 5 it follows that, on average, around 70% of highly skilled migrants from theNMS had been abroad for more than 10 years. Thus, as suggested before, the problemwith the stock data is that migration rates derived from them is a cumulative effect oflong-lasting migration process and should not be used (directly) to analyze recentmigratory phenomena.CMR 11

Table 6: Stock of immigrants born in Poland in the OECD countries, by education leveland duration of stay, around 2000Country ofHighly-skilledresidenceTotal Total More than 10 years 5 to 10 years Up to 5 years UnknownAustralia 56,292 13,706 11,313 1,500 663 230Austria 31,642 5,254 3,370 1,884 n.a. n.a.Belgium 18,880 3,153 1,779 542 831 1Canada 177,525 61,455 45,360 12,635 2,935 525Czech Republic 15,519 n.a. n.a. n.a. n.a. n.a.Denmark 10,247 2,738 2,185 315 238 -France 103,829 18,130 10,922 2,225 2,121 2,862Germany 1,021,656 168,777 155,647 13,130 n.a. n.a.Greece 14,547 2,413 393 615 476 929Ireland 1,956 624 66 42 351 165Italy 31,413 5,423 1,314 1,087 1,355 1,667Luxembourg 931 253 51 37 161 4Netherlands 5,012 n.a. n.a. n.a. n.a. n.a.New Zealand 1,851 483 297 84 96 6Norway 6,578 1,986 1,601 233 152 -Spain 15,600 3,440 1,580 760 1,100 -Sweden 36,530 11,120 8,775 845 1,445 55Switzerland 10,814 6,182 1,840 404 1,401 2,537United Kingdom 39,618 n.a. n.a. n.a. n.a. n.a.United States 452,005 119,465 80,008 23,935 15,522 -OECD - Total 2,052,445 424,602 326,501 60,273 28,847 8,981Source: OECD database, based on national population censuses and LFSsTable 6 repeats the analysis for Polish migrants only. It is worth noting that the providednumbers seem to be quite reliable. In 2000, over 425,000 highly-skilled people born inPoland were registered abroad. However, of these, over 327,000 had stayed abroad forlonger than 10 years, compared to only 60,000 between five and 10 years and only29,000 for less than five years. Interestingly, there are significant differences observedbetween destination countries. In Germany, for example, more than 90% of all Polishhighly skilled migrants had lived there for more than 10 years. But in the ‘new’destinations (Spain, Italy, Ireland) the share of long-term migrants was less than 25%.Last but not least, according to this OECD data, Poles constituted around 50% of allhighly skilled migrants originating from the NMS.5 Contextual issuesIn this section we provide the context for the recent migration of high-skilled workers,including basic data on the demographic and educational structures in the NMS and theposition of high-skilled workers in labour market in the origin countries. From Deliverable2 it is clear that the populations of the NMS are relatively young compared to thepopulations in the EU15 countries. This is reflected, for example, in the share of personsaged 15-34, which is significantly higher in the NMS than the EU25 average. This isparticularly the case in three countries: Poland, Romania and Slovakia. It is important tonote that those aged 15-34 are, on average, the most geographically mobile and alsoCMR 12

most likely to be actively participating in the learning process. Thus, analysis ofdemographic data reveals, firstly, a relatively high migratory potential, but also,secondly, may suggest that the structure of educational attainment can differ betweenthe NMS and the EU15 countries.Table 7 shows the percentage of the population aged 25 to 64 having completed at leastupper secondary education. According to the data the NMS can be described as havinghigh levels of human capital relative to EU15 countries. This is particularly the case forthe NMS8 countries, and especially in the Czech Republic, Poland and the Baltic States.Bulgaria and Romania (NMS2) have lower levels of human capital, but still have higherlevels than the average in the EU15.Table 7: Percentage of the population aged 25 to 64 having completed at least uppersecondary education, 2000-2007Country 2000 2001 2002 2003 2004 2005 2006 2007EU15 61.5 61.5 62.4 63.9 65.2 66.2 66.7 67.5NMS10 79.4 79.7 80.7 82.1 83.1 84.2 : :Bulgaria 67.5 71 71.6 71.2 71.7 72.5 75.5 77.4Czech Republic 86.1 86.3 87.9 88.5 89.1 89.9 90.3 90.5Estonia 86.1 87.1 87.6 88.5 88.9 89.1 88.5 89.1Latvia 83.2 79.6 82.2 83.2 84.6 84.5 84.5 85.0Lithuania 84.2 84.2 84.9 86.1 86.6 87.6 88.3 88.9Hungary 69.4 70.0 71.4 74.1 75.3 76.4 78.1 79.2Poland 79.9 80.2 80.9 82.3 83.6 84.8 85.8 86.3Romania 69.3 70.6 71.1 70.5 71.5 73.1 74.2 75.0Slovenia 75.3 75.8 77.0 78.1 79.7 80.3 81.6 81.8Slovakia 83.8 85.1 86.0 86.7 87.0 87.9 88.8 89.1Source: Own elaboration based on EUROSTAT dataTable 8 refers to students at the tertiary level, and shows the trends in the number ofstudents between 2000 and 2006. The data reveal that, leading up to accession, most ofthe NMS experienced a significant (and in some cases enormous) increase in the numbersof students. In the EU25 as a whole the average number of students in 2006 was around16% higher than in 2000. In the case of the NMS the ratio was much higher (with theexception of Bulgaria). The largest increases in number of students were in Romania(85% increase), Lithuania (63%), Slovakia (46%), and Poland (36%). Such an enormousincrease raises the question of the quality of the tertiary education and obviously, this isan issue that needs closer attention in future analyses.CMR 13

Table 8: Trends in the number of students (ISCED 5-6) 17 , EU25 and NMS, 2000=100Country 2000 2001 2002 2003 2004 2005 2006EU25 100.00 103.49 107.38 111.05 113.89 115.43 116.42Bulgaria 100.00 94.53 87.41 88.21 87.45 91.04 93.19Czech Republic 100.00 102.48 112.14 113.13 125.70 132.56 132.99Estonia 100.00 107.84 113.06 118.66 122.57 126.49 127.43Latvia 100.00 112.72 121.16 130.37 140.02 143.31 143.75Lithuania 100.00 111.48 122.07 137.49 149.88 160.30 163.17Hungary 100.00 107.62 115.40 127.16 137.48 141.97 142.85Poland 100.00 112.37 120.68 125.56 129.42 134.09 135.84Romania 100.00 117.81 128.63 142.27 151.50 163.23 184.49Slovenia 100.00 109.19 118.38 121.12 124.58 133.89 136.99Slovakia 100.00 105.89 111.99 116.34 121.19 133.48 145.62Source: Own elaboration based on EUROSTAT dataTable 9 shows that, in more than half of the NMS, the share of students in the populationaged 15-29 is higher than the EU25 average. The share of students is particularly high inEstonia, Lithuania, Latvia and Poland. The highest increases were noted directly beforethe EU-enlargement.17 The International Standard Classification of Education (ISCED) is commonly used in order to compare oneducation data between countries. In the context of this report four ISCED levels are of special importance:level 3 - (upper) secondary education, level 4 - post-secondary non-tertiary education, level 5 - first stage oftertiary education, level 6 - second stage of tertiary education.CMR 14

Table 9: Percentage of students (ISCED level 5 and 6, tertiary and post-tertiary) amongthe population aged 18-24, EU countriesCountry 1998 1999 2000 2001 2002 2003 2004 2005New member statesBulgaria 22.4 23.1 23.1 21.6 21.2 21.9 21.9 23.3Cyprus 13.6 12.3 13.8 15.8 19.5 20.9 18.0Czech Republic 13.7 14.9 17.1 17.4 19.0 19.6 21.9 23.2Estonia 25.3 28.2 28.7 28.9 28.5 28.4 28.5 28.7Hungary 15.5 17.6 19.2 20.0 22.0 24.3 26.3 27.5Latvia 22.6 25.1 24.1 26.1 27.9 29.0 29.9 29.5Lithuania 23.6 25.7 28.2 30.1 31.1 33.0 34.1 35.2Malta 11.3 11.8 13.1 13.3 15.0 13.9 16.4Poland 19.4 22.1 24.5 27.2 28.9 30.0 30.9 32.5Romania 10.8 12.2 13.9 16.6 19.5 20.5 21.5 22.6Slovak Republic 17.0 17.8 18.6 18.9 18.8 20.3Slovenia 25.0 27.9 29.2 31.2 33.8 34.9 35.7 38.1EU15Austria 17.6 18.0 17.0 17.5 16.7 17.3 18.0 18.4Belgium 32.9 32.8 32.6 32.9 33.1 33.5 33.7Denmark 16.5 17.1 17.8 18.4 19.1 19.7 20.2 20.7Finland 28.1 28.9 29.6 29.9 29.8 31.2 31.5 31.8France 29.3 29.5 29.7 29.7 29.3 29.5 29.7 29.9Germany 14.4 14.6 14.8 15.2 15.9 16.6 17.5 17.7Greece 29.9 31.8 34.7 33.3 37.3 40.3 44.4 44.5Ireland 24.6 24.7 25.7 26.2 27.0 27.1 27.9 27.2Italy 23.1 22.3 22.1 22.8 23.9 25.6 26.8 27.5LuxembourgNetherlands 23.9 24.8 25.8 26.2 26.5 26.7 27.4 28.3Portugal 21.3 22.3 23.2 24.9 24.6 25.1 25.3 25.0Spain 27.9 29.0 30.4 30.4 30.3 30.3 30.3 29.7Sweden 17.6 20.4 21.1 21.5 22.6 23.6 24.2 24.0United Kingdom 21.3 22.5 22.3 22.7 23.4 22.9 22.3 22.3Source: Eurostat, own elaborationOf course, the fact that there has been an increase in educational investments does notnecessarily mean that the labour market situation is improving. 18 In at least threecountries among the NMS (Bulgaria, Poland and Slovakia) there were severe labourmarket disequilibria observed in the pre-enlargement period. However, in most casessince 2003 the unemployment rate has fallen, and as of 2007 only in Poland and Slovakiawere unemployment rates higher than the EU15 average. These developments are, atleast to some extent, likely to be linked to post-accession labour mobility.18 In contrary, in many CEE countries, education at tertiary level tends to be perceived as an ‘escape fromunemployment’ and thus may reflect negative changes on the domestic labour market.CMR 15

Table 10: Unemployment rates in the NMS10, percent, for persons with upper secondaryand post-secondary non-tertiary education – levels 3-4 and with tertiaryeducation – levels 5-6 (ISCED 1997), 2000-2007 (2 nd quarter)Levels 3-4 Levels 5-62000 2002 2004 2005 2006 2007 2000 2002 2004 2005 2006 2007EU15 7.9 7.4 8.1 7.9 7.4 6.5 4.9 4.6 5.1 4.8 4.5 3.9NMS10 14.1 15.5 14.9 14.1 11.3 8.2 5.1 4.9 5.5 5.0 4.1 3.2Bulgaria 15.8 17.7 11.3 9.1 7.7 5.6 6.7 8.2 5.8 4.2 3.8 2.7Czech Rep. 7.9 6.4 7.5 7.1 6.3 4.7 3.0 1.8 2.1 2.1 2.5 1.3Estonia 14.8 10.3 10.7 10.1 6.2 5.3 5.0 4.7 6.0 3.2 4.1 3.2Latvia 14.9 13.0 10.6 8.9 6.0 5.4 7.4 6.6 3.6 3.9 2.7 3.1Lithuania 20.3 14.6 12.8 9.7 6.5 4.6 9.4 6.8 6.7 3.8 2.4 2.1Hungary 6.5 5.1 5.4 6.9 6.6 6.3 1.4 1.8 2.2 2.5 2.6 2.5Poland 17.1 21.2 20.4 19.4 15.2 10.5 5.4 6.6 7.3 6.8 5.5 3.8Romania 9.5 10.0 8.4 8.2 7.7 7.1 3.6 4.1 3.1 3.3 3.1 2.8Slovenia 7.0 6.1 6.1 6.0 6.5 4.7 2.2 2.5 2.8 3.1 3.0 2.8Slovakia 18.4 17.8 17.0 14.4 12.1 9.4 5.2 3.9 5.9 5.2 3.0 4.2Source: Own elaboration based on EUROSTAT dataThe left hand panel of Table 10 shows that in most cases the unemployment rate forthose with upper secondary and non-tertiary education (ISCED 3-4) fell between 2000and 2004 towards the EU15 average. Poland and Slovakia were exceptions, withsignificantly higher unemployment rates. The right hand panel of Table 10 showsequivalent unemployment rates for those with tertiary education. In the post-accessionperiod Poland was the only country among the NMS with relatively high unemploymentrate of the well educated, which suggests that there may be problems in absorbing thelarge numbers of highly educated workers. Once again, this suggests that Poland shouldbe characterised as a country experiencing brain overflow” rather than “brain drain”.CMR 16

Table 11: Unemployment rates in the NMS10, percent, for persons aged 15-24 with uppersecondary and post-secondary non-tertiary education – levels 3-4 (ISCED1997), 2000-2007 (2 nd quarter)2000 2001 2002 2003 2004 2005 2006 2007EU15 14.1 12.0 12.5 13.1 13.6 13.6 12.9 11.9NMS10 27.3 30.6 31.4 31.9 31.2 29.8 23.2 16.3Bulgaria 30.4 33.3 31.0 23.0 19.7 17.7 16.3 9.8Czech Rep. 14.1 13.2 13.0 13.9 16.7 15.3 13.8 8.1Estonia 17.3 21.8 . 23.4 18.5 21.8 . .Latvia 17.9 19.2 21.1 14.5 18.4 13.4 11.6 6.7Lithuania 26.2 30.5 18.5 26.9 23.0 20.3 . 7.2Hungary 11.0 9.4 10.0 10.5 12.0 16.6 13.7 14.2Poland 35.7 39.9 42.2 42.9 40.6 39.0 29.9 21.3Romania 22.0 21.0 25.0 22.8 24.0 21.7 19.9 19.8Slovenia 14.5 13.4 12.4 13.8 13.2 12.7 14.5 6.4Slovakia 35.0 36.6 35.6 30.6 28.6 23.5 21.3 13.5Source: Own elaboration based on EUROSTAT dataTable 10 referred to the whole working-age population. As noted, however, the mostdynamic changes with regard to education were in the youngest age groups. Table 11shows that the labour market situation of young people in the NMS was relatively poor,particularly in the pre-accession period. This is true not only for poorly-skilled individuals,but also to those persons with upper secondary and post-secondary (non-tertiary)education, particularly in case of Poland and Romania.Table 12: Unemployment rates in the NMS10, percent, for persons aged 15-39 withtertiary education - levels 5-6 (ISCED 1997), 2000-2007 (2nd quarter)*2000 2001 2002 2003 2004 2005 2006 2007EU15 6.1 5.3 6.0 6.1 6.3 6.3 5.7 5.1Bulgaria 7.9 9.5 10.3 8.6 7.4 5.2 4.5 2.8Czech Rep. 3.8 4.0 2.3 3.2 2.3 2.9 3.0 2.0Estonia . 8.8 . . 8.3 . . .Latvia 7.6 . 7.7 4.5 . 4.1 . 3.5Lithuania 11 8.4 8.3 5.6 7.2 4.7 . 2.4Hungary 2.0 1.4 2.5 1.8 3.1 3.6 3.9 4.1Poland 7.7 9.0 9.4 9.7 10.1 9.6 7.9 6.1Romania 5.6 5.5 5.3 4.4 4.7 5.8 5.6 3.9Slovenia 3.2 2.8 3.9 4.9 4.0 4.6 4.9 4.6Slovakia 7.3 8.0 5.2 5.9 8.4 6.3 4.4 5.8* if availableSource: Own elaboration based on EUROSTAT dataTable 12 refers to unemployment rates of persons aged 15-39 who completed educationat the tertiary level. It can be seen that in at least two cases (Poland and Slovakia)highly-educated young people still face serious problems on the domestic labour market.The most difficult situation is in Poland, where the unemployment rate among those whoCMR 17

achieved tertiary education was much higher than the EU15 average. To conclude, it isnecessary to consider the following factors that may have a profound effect on the extentand the consequences of high-skilled migration from the NMS:- The NMS populations are generally younger than populations of the ‘old’ Europe,particularly in the cases of Poland, Romania and Slovakia. We would expect thesecountries to have high migratory potential.- The NMS populations are relatively well educated (in some cases much better thanthe EU15 countries). This situation is, to some extent, the consequence ofcommunist past (which may raise the question on the quality of educationobtained) but mainly the outcome of the ‘educational breakthrough’ as observedin the 1990s and 2000s (particularly in Romania, Lithuania, Slovakia, and Poland).Consequently, a key point in the analysis of migration is the demographic andeducational structure of sending populations. The data presented in this section shows amarked increase in enrolment in tertiary and post-tertiary education in most NMSbetween 1998 and 2005, a time when the corresponding figures in EU15 countriesremained roughly constant, or at least fell short of matching the NMS increase. Thissuggests that the recent substantial increase in the supply of highly skilled workers couldmore than offset any drain of skilled workers from NMS, even if one cannot attribute theobserved increase in enrolment rates to the migration pattern from these countries.Nevertheless, even though the causal relationship might be weak, it is true that thefigures provided in this section contribute to further reduce possible concerns about adetrimental effect of skilled migration from these countries. 19In many cases the labour market position of young people in the NMS who obtainedtertiary education is actually more favourable than in the EU15, although there areimportant exceptions, such as Poland. This casts some doubts on the migratory potentialof this group. On the other hand — and this is of great importance in the context of thisreport — the data suggest that, in the case of Poland, the outflow of well-educatedindividuals can be seen as “brain overflow” rather than a “brain drain”. This question willbe the subject of more in-depth analysis in the next section which analyses the impactsof high-skilled mobility on sending countries.6 High-skilled mobility and its impact on sending countriesThe main aim of this section is to provide an in-depth analysis of high-skilled mobilityand its consequences. The scope of the analysis depends partly data availability, and as aresult most attention will be paid to Poland, a country which sends the largest numbersof migrants abroad, but also offers migration data of relatively high reliability. Thedeparture point will be an overview of recent trends in international mobility and stylized19 As suggested by the analysis of skill specific unemployment rates for young workers in NMS provided above.CMR 18

facts on the migration of the highly skilled. Against this background a selectivity analysiswill be provided for both the pre- and post-accession periods. We focus in particular onthe effects of the EU enlargement and the introduction of the Transitional Arrangements,and on the structural patterns of mobility from the NMS and the consequences of theoutflow on the sending countries.6.1 Poland – recent migration and mobility of the highly skilledPoland is the most important migrant sending countries among the NMS, with significantmigration flows recorded since the early 1970s. In the 1980s the number of long-termmigrants amounted to between 1.1 and 1.3 million people, about 3% of the totalpopulation. In addition, more than 1 million people spent more than three but less than12 months abroad (Kaczmarczyk and Okólski, 2002). National census data from 1988indicated that around 900,000 permanent citizens of Poland (approximately 2% of thetotal population) resided abroad on a temporary basis. Most of the data available suggestthat, in the very first phase of transition, the international mobility of Poles declined.Data from the LFS data indicate a significant decline in the scale of migration between1994 and 1998 (from over 200,000 to 150,000 people staying abroad every quarter).However, since the late 1990s, migration from Poland has been on the rise again. The2002 National Census indicated that around 790,000 Polish citizens (1.8% of the totalpopulation) were staying abroad. Generally, prior to the EU enlargement, Poland was oneof the most important European migrant sending countries, with significant numbers ofits citizens employed in Germany (with seasonal migration playing an important role —around 250,000 people a year in the early 2000s), the United States of America andsouthern European countries (Italy, Spain).The recent estimates provided by the Polish Central Statistical Office constitute the mostreliable data set made available thus far (see Table 13). 2020 For more details see the Polish Country Study in this project.CMR 19

Table 13: Polish citizens staying abroad for longer than 2 months by destination country,estimates (000s)Destination May 2002 End of 2004 End of 2006 End of 2007Total 786 1000 1950 2270European Union 451 750 1550 1860Austria 11 15 34 39Belgium 14 13 28 31France 21 30 49 55Germany 294 385 450 490Ireland 2 15 120 200Italy 39 59 85 87Netherlands 10 23 55 98Spain 14 26 44 80Sweden 6 11 25 27United Kingdom 24 150 580 690Source: Central Statistical Office (2008).As is shown in Table 13, the stock of migrants from Poland more than doubled since EUenlargement. Over 80% of Polish migrants in 2007 were residents of other EU countriescompared to 57% in 2002, while the most important destination country became theUnited Kingdom, with 30% of the total. Germany — the most favourable destinationcountry for Polish migrants in the pre-accession phase — received ‘only’ 22% of theoutflow. Notable increases were also observed in Ireland, the Netherlands and Sweden.The massive post-accession migration of Poles is confirmed by data obtained from majordestination countries, particularly from the UK, which became the most attractivedestination country for Polish migrants after May 2004. According to the InternationalPassenger Survey in 2006 the number of visitors from Poland was 4.8 times higher thanit was in 2003, exceeding 1.6 million. 21 From Worker Registration Scheme data, over500,000 Poles registered with the system up until September 2007. The inflow wasparticularly high in 2005 and 2006, and only began diminishing in 2007. Beginning in thefiscal year 2003/2004 Polish citizens appeared among the top ten countries of origin ofincoming migrants that were allocated a National Insurance number. The total number ofNational Insurance Numbers allocated to Poles between 2003 and 2007 amounted toaround 470,000. Poles thus constitute the most important migrant group, accountable forover 30 percent of the total inflow of foreigners to the insurance system.The data presented above is also supported by the UK LFS data, which indicates thatbetween early 2006 and early 2007 the number of Poles residing in the UK increasedfrom 209,000 to 406,000 (Kaczmarczyk and Okólski, 2008).With regard to the composition of the migrating population, post-accession migrationfrom Poland can be expressed both in terms of continuity and change. The mostimportant aspect of continuity is the predominance of labour migration. According to the21 Note that this data refers not to migration per se but rather depicts the scale of and trends in mobility,including tourism.CMR 20

LFS around 80% of migrants take up employment while staying abroad. The prevalenceof short-term mobility also remains more or less stable. In the first half of 2000, asignificant proportion of all temporary migrants (over 60%) stayed abroad for less thantwelve months. However, a long-term mobility trend also began to emerge after EUenlargement. For example, the proportion of short-term migrants in the total number ofmigrants decreased from 63% in 2005 to 54% in the second quarter of 2007 (Kępińska,2007). This suggests that Polish migrants are prolonging their stays abroad.One of the most prominent changes in the structure of Poles’ post-accession mobilityrefers to destination countries (see Table 14). However, according to available data,recent migration from Poland is not best understood in terms of a particular concentrationin selected countries (i.e., mostly in the UK and Ireland) but rather as a gradual ‘spillingover’. In fact, Polish citizens are targeting almost all EU/EEA countries and have becomeincreasingly active contributors to their labour markets. The widening range ofdestination countries for Poles is not the only element changing. In general, recent Polishmigration is more regular than irregular (that is, more frequently legal than clandestine),more of a long-term duration than circular, and more individualistic than related(subordinated) to household strategies (Kaczmarczyk and Okólski, 2008).Traditionally, a considerable part of Polish migration was ascribed to the mobility of thehighly skilled. However, this thesis seems to be rather questionable with reference toalmost the whole post-war period. With the exception of an episode of (partially) forcedand politically motivated migration of persons of Jewish descent (1968-1971), when over13,000 mostly highly educated persons left Poland, the share of persons with tertiaryeducation among all migrants did not differ significantly from that of the totalpopulation. 22 The situation changed in the late 1970s and 1980s. The overrepresentationof the highly skilled is particularly true in the case of the massive outflow in the 1980s.Calculations based on the policy register’s data show that of almost 700,000 emigrantswho left Poland between April 1st, 1981 and December 6th, 1988, 15% had a higherdegree and 31% had secondary education. If we consider that for the whole populationthe share of university graduates was approximately 7%, the data show that there was aconsiderable overrepresentation of the highly educated amongst emigrants (Sakson,2002). According to estimates of Okólski (1997), the scale of the emigration of high-classspecialists in the 1980s was so large that the number of emigrants in this category eachyear (15,000) constituted approximately a quarter of Polish university graduates of allhigher education institutions.As follows from various data sources, the situation has changed much duringtransformation. According to the official data, since 1990 the share of individuals with thelowest level of education amongst migrants has been increasing, while the share ofindividuals with the highest level of educational attainment has been decreasing. At thethreshold of transformation in 1988, 37% of migrants aged 15 or above had an22 In case of emigrating Poles of Jewish descent this share was over eight times higher than in the totalpopulation.CMR 21

elementary or lower than elementary education, compared to 9% of migrants who had ahigher degree. In 2003, there were 55% in the former group, and 4% in the latter. Theseobservations were proved by the majority of studies conducted both in Poland and in thereceiving countries. 23Table 14: Permanent residents of Poland (aged 15 and above) living abroad for morethan one year (as of May 15, 2002), of which those with at least universitydiploma, by country of destination (actual residence) and year of departureYear of departureTotalCountry of residenceTotal Germany Italy UK other U15 U.S. Canada Otherx 39.0 4.2 2.4 10.1 21.8 4.2 18.3of which those with university diploma 14.0 20.6 3.1 6.0 12.9 26.8 7.1 23.51988 and before 15.6 21.8 2.1 3.2 12.4 24.3 13.2 22.91989-1991 11.8 26.2 2.0 2.5 10.8 28.4 10.7 19.41992-1994 13.4 17.7 3.1 4.2 13.8 32.0 7.9 21.21995-1997 13.4 19.2 3.7 6.4 13.7 29.4 4.9 22.61998-2001 15.2 19.4 3.8 9.8 13.6 25.8 3.2 24.2Source: Kaczmarczyk and Okólski, 2005.According to the Polish census of 2002, among 576,000 permanent residents aged 15 ormore years who at the census date lived abroad for at least 12 months, 24 0.7% held adoctor’s degree, 10.1% a university diploma and 3.2% other tertiary education diploma.Respective shares for the general population were 0.3%, 7.4% and 2.7%. Altogether theeducation of migrants was much better than actual residents (14.0% vs. 10.4%). As canbe seen in Table 14, the share of highly educated migrants was the highest among thosewho left Poland before the onset of transition (15.6%), became rather low among thosewho emigrated in 1989-1991 (11.8%), and rose among those leaving in the followingyears.For obvious reasons the population census cannot serve as a source of information on themost recent emigration from Poland. Another source of information about emigrationfrom Poland, namely the population register, reflects only a very small part of the totaloutflow. 25 It reveals that during the 1990s the percentage of highly skilled personsamong emigrants was very low, approximately 2% (Figure 1). Since 2004 this sharestarted to increase very rapidly, to reach in 2005 8% for men and 11% for women. No23 CMR research in the years 1994-1999 indicated that the claim about the brain drain can be upheld only inrelation to big urban centers. More importantly in quantitative terms, migration from the peripheral regionswas dominated by individuals with no more than secondary educational attainment, of poor human capital,taking up employment almost exclusively in the secondary sectors of labour markets in the host countries.Similar results were provided by studies conducted both in Poland and in the receiving countries. Each of thesestudies supported the observation that a greater propensity to migrate was typical for people with low culturalcompetencies and no knowledge of foreign languages who encountered problems with finding their feet in thenew post-communist reality, particularly on the labour market. These people were almost fully dependent onthe employment offer addressed to unskilled workers, willing to start work any time and for any period of time(usually on an extremely short-term basis). Exceptions to the case — such as Ireland or the Scandinaviancountries — only confirmed the general rule (Kaczmarczyk and Okólski, 2005).24 That was 1.8% of the total number of permanent residents of Poland aged 15 or more years.25 Mainly due to definition of migrant applied. According to the official data emigrant from Poland is a personwho left with an intention to settle abroad and de-listed her-/himself from the place of permanent residence inPoland.CMR 22

data for subsequent years is available as since 2006 the information about educationlevel of emigrants ceased to be collected in the population register system.Figure 1: The share of emigrants with post-secondary level of education on all registeredemigrants by sex, in percent, 1994-2005Source: population register, after Okólski (1997-2001), Okólski and Kępińska (2002), Kępińska(2003-2007), Recent trends in international migration – OECD Sopemi report for Poland, variousyears.The re-emergence of the highly skilled outflow from Poland and the increase in its scalesince the EU enlargement is also reflected by the Labour Force Survey, which remains themost comprehensive data source on the educational structure of Polish emigrants.According to the CMR Migrants’ Database based on the Polish LFS, 26 the pre-accessionoutflow from Poland was dominated by people with secondary vocational and vocationaleducation (61% of migrants, Table 16). After 2004 the share of University graduatesincreased significantly: from 15% to 20%, which in comparison to 14% of Universitygraduates in the overall population of Poland (in 2004) is the sign of high selectivity ofmigration with respect to education (see Section 5.2). In particular, this is the caseamong female migrants, out of whom 27% were highly-skilled persons.However, as we already argued in Section 1, this picture may be misleading without anassessment of the structure of the Polish population. In the last twenty years, Polandexperienced a true educational breakthrough (see Section 4). Between 1970 and 2001,the share of university graduates among the Polish population increased from 2% to12%. At the end of the 1990s, the number of students was 2.6 times higher than in1990. Nowadays in Poland there are over 1.8 million students, and data from the CentralStatistical Office shows that in the early 2000s the gross enrolment ratio (the rate of allstudying to the whole population) in the age group 19-24 was over 30% (see section 4),which means that as far as the universality of higher education is concerned, Poland has26 See below for details on the construction of the dataset.CMR 23

almost reached the standards of the EU15. If we take into consideration that a higherpropensity to migrate is typically a feature of relatively young persons (aged 18 to 35),the recent increase in the highly skilled migration is a natural phenomenon and reflectschanges in the demographic and educational structure of sending population and migrantgroup.Table 16: The education structure of Polish pre- and post-accession migrants by sex, inper centPre-accessionPost-accessionLevel of education Total Men Women Total Men WomenUniversity degree 14.7 12.0 18.3 19.8 15.6 27.0Secondary 14.0 7.1 23.1 14.2 8.8 23.8Secondaryvocational 26.1 26.0 26.3 28.1 29.8 25.1Vocational 34.8 45.4 20.9 30.9 39.2 16.2Primary 9.9 9.3 10.9 7.0 6.6 7.8Unfinished 0.4 0.2 0.5 0.0 0.0 0.0Total 100.0 100.0 100.0 100.0 100.0 100.0Notes: Pre-accession migrants - aged 15 and over who have been abroad for at least 2months in the period 1999-2003; post-accession migrants - in the period may 2004 -December 2006; University degree - including bachelor, master and Ph.D. degree.Source: CMR Migrants’ Database, based on the Polish LFS.6.2 Selectivity of the recent outflow from PolandPoland’s accession to the EU was expected to affect the migration patterns of the Polishpopulation in many ways. Below we will present an account of the scale and diversity ofthose influences by comparing migrant characteristics of the immediate pre- and postaccessionperiod.The analysis will be based on two migrant databases extracted from Polish LFS. Due tolimited number of migrant cases in the samples a dedicated data set was created. Thisdata set consists of two sub-sets. The first one includes all residents of Poland aged 15or above who, at the time of the survey, resided in a foreign country for longer than twomonths (hereafter referred to as “temporary migrants”). The second one includes thosetemporary migrants whose stay abroad did not exceed one year (hereafter, “short-timetemporary migrants”). 27All migrants in the databases were divided into two groups according to the time of theirdeparture from Poland: those who left between the 1 st quarter of 1999 and the 1 st quarter27 The number of migrants in the first database was 6,693. In the second database 3,700, which allows us toprovide various structural breakdowns at both the country and regional level.CMR 24

of 2004 (“pre-accession migrants”) and between the 1 stquarter of 2006 (“post-accession migrants”). 28quarter of 2005 and the 4 thTable 17 shows that the accession seemed to have a significant impact on the geographyof outflow. The top 3 countries of the pre-accession period — those that accounted foralmost two-thirds of the total outflow — lost their predominance in the post-accessionperiod, replaced by three countries whose importance before May 1 st , 2004 was moderateor very low. The former three countries consisted of destinations known for extensive andwell-developed (and in the case of a few destination countries, long-lasting) networks ofPolish migrants, whereas the latter three happened to be the only EU countries which onMay 1 st 2004 did not introduce restrictions to the access of Polish migrants to their labourmarkets. It is worth noting that the shift in the geography of outflows was more markedin the population of short-term than long-term migrants.Table 17: All and short-term temporary migrants from Poland sorted by major groups ofdestination countries before and after EU accessionGroupAll migrants Of which short-term migrantsof countries before after before afterCountries granting Polish citizens afree access to labour market afterMay 1 st , 2004 * 12.1 42.2 10.3 46.5(of which the United Kingdom) (9.7) (31.3) (8.2) (34.4)Top-3 countries of the preaccessionperiod ** 62.9 36.1 63.7 34.8(of which Germany) (31.9) (18.8) (38.2) (20.4)Countries whose share in the totaloutflow was at least 3 percent inany period ***12.7 11.0 13.9 9.5Other countries 12.3 10.7 12.1 9.2Notes: * United Kingdom, Ireland, Sweden,Belgium, the NetherlandsSource: Kaczmarczyk and Okólski 2008.** Germany, USA, Italy, ***France, Spain,The data strongly support a hypothesis of a shift from predominantly network-driven topredominantly labour demand-driven migration. This hypothesis can also be supportedby the analysis of the distribution of migrants, sorted by their region of residence prior tomigration, in the period before May 1 st 2004, with that which occurs in the period afterMay 1 st 2004. It can be concluded that the post-accession migrants were more evenlydistributed across regions than were the pre-accession migrants. Temporary migrationbecame more readily accessible to people across Poland, which seems consistent with thehypothesis that stresses the role of demand as an impetus for outflow. Additionally, asshown in the Polish country study (within this research project) recent changes in the28 For methodological reasons, migrants recorded between the 2nd quarter of 2004 and the 4th quarter of2004 were not included in the databases.CMR 25

anking of destination countries is to be linked to institutional changes (particularly, theintroduction of Transitional Arrangements).One of the most striking tendencies within migrant selectivity was a change with respectto education levels. A predominant part of Poland’s population aged 15+ comprises (as ofmid-2004) persons with educations below secondary levels, where only 12% haveuniversity diplomas (or their equivalent). Before the accession no selectivity effect wasobserved among people with post-secondary education, while those with vocationaleducation, being by far the largest group among migrants, exhibited a moderate positiveselectivity. After the accession, the selectivity index value (SI) 29 remained almostunchanged in the latter group and became much higher in the former group. Generally,post-accession Polish migrants are definitely positively selected with respect to education.Table 18: Migrant selectivity indexes (SI) for post-secondary and vocational educationbefore and after EU accession (all migrants), by selected countries ofdestinationEducational level/ country of destination Before accession After accessionAll countriesPost-secondary 0.02 0.42Vocational 0.34 0.30United KingdomPost-secondary 1.09 1.13Vocational 0.07 0.11GermanyPost-secondary -0.29 -0.52Vocational 0.51 0.57Source: Kaczmarczyk and Okólski 2008.Three categories of educational attainment encountered a pretty similar loss, namelyaround 4%. Those were: tertiary (university diploma or equivalent 30 ), other postsecondaryand secondary completed 31 , and vocational. 32 In the group with educationlevels lower than vocational the loss was merely 1%. There were, however, considerabledifferences between males and females. Males with post-secondary (other than tertiary)and secondary education suffered the largest loss (5.8%), followed by those withM29 V = iPV = i−The migrant selectivity index is illustrated by the following formula:S I M PV = i=P V = iPwhere: SI V=i – index for category i of variable V; M V=i and P V=i – number of migrants and number of people inthe general population, respectively, falling into category (or value) i of variable V, and M and P – overallnumber of migrants and people in the general population, respectively. The selectivity of outflow takes place ifthe index assumes a non-zero value for any category (value) of a given variable. A positive SI value meansthat migrants falling into a specific category (variable) of a given variable are relatively more numerous thanpeople in the general population with the same characteristics, whereas a negative SI value (but equal to orhigher than -1) means the opposite. The higher the positive value or the lower the negative value of SI, thestronger the selectivity.30 (Usually) at least 16 years of schooling.31 Usually at least 12 years of schooling.32 Usually at least 10 to 11 years of schooling.CMR 26

vocational education (5.4%), tertiary (5.0%) and lower (1.4%). Amongst females, thelargest loss was noted among those with tertiary education (3.3%), whereas women withpost-secondary and secondary education lost 3.1%, with vocational –2.4% and withlower –0.6%.From Table 18 it follows that distinctive differences were noted with regard to the mostimportant destination countries. This is clearly supported by following figures showing SIfor tertiary and vocational education, and for the UK and Germany (by Polish regions).Figure 2: Migrant selectivity indexes (SI) for tertiary educationFigure 3: Migrant selectivity indexes (SI) for vocational educationSource: Kaczmarczyk and Okólski 2008.From the above it follows that the selectivity of migrants in various categories ofeducation levels was diversified according to the target country (and also category ofsettlement – see section 5.3). Generally, the United Kingdom strongly ‘attracted’ theCMR 27

highly educated and appeared largely neutral with regard to the poorly educatedwhereas, in a striking contrast, Germany ‘repelled’ the highly educated and moderately(positively) ‘attracted’ people with low education levels. Those two destination-specifictendencies — although visible in the pre-accession period — appeared to be reinforcedafter accession. They can be interpreted in many ways. According to the model presentedby McKenzie and Rapoport (2008) structural changes in post-accession mobility can beattributed to migrant networks. The model predicts negative self-selection of Polishmigrants to Germany due to relatively extensive and long-lasting migrant networks inthis country and positive self-selection in case of those countries where networks areweak or non-existent (e.g. UK or Ireland). On the other hand, the change in patterns ofmobility of the highly skilled in the post-accession period is to be linked to institutionalchanges since 2004, particularly to the opening of different forms of labour markets.Figure 4: Share of university graduates among Polish migrants in the post-accessionperiod, by type of restrictions imposed on the labour market access10090888070605040302010031 302511211511422114 1311 92812Czech Rep.IrelandUKSwedenGreeceFranceItalyNetherlandsSpainNorwayDenmarkBelgiumAustriaGermanyCanadaUSSwitzerlandno restrictions short-term restrictions long-term restrictions outside EEASource: Fihel and Kaczmarczyk, 2008.The above shows that, in general, those countries which did not introduce restrictions onmobility are gaining the “best” migrants (in terms of skills). On the other hand, countrieswhich did impose short- or long-term restrictions are attractive predominantly to personswith relatively poorer education. However, significant differences within all groups maysuggest that this pattern is to be attributed predominantly to the structure of demand ineach labour market rather than to institutional arrangements in the post-accessionperiod.CMR 28

6.3 Drain effect or brain effect?Regardless of the methodological issues and the uncertainty as to the real scale of thephenomenon, most of the data clearly indicates that there is a positive selection ofemigrants from Poland and other NMS with regard to education. The next step is toassess the impact of post-accession mobility on the sending countries. We follow the lineof reasoning of Beine et al. (2001) and analyse the consequences of the highly skilledmobility in a static (drain effect) and dynamic (brain effect) framework.6.3.1 Drain effectA massive outflow of migrants — as has been observed in some of the NMS — may havea significant impact on the labour market in sending countries. Consequences of outmigrationinclude an eventual decline in unemployment (so-called export ofunemployment), labour shortages (due to the outflow of workforce) and a correspondingpressure on wages.A back-on-the-envelope analysis of the labour market data seems to support thesehypotheses. In case of Poland, between the 2 nd quarter of 2004 and the 1 st quarter of2007 the number of unemployed individuals decreased from 3.1 million to 1.5 million andthe unemployment rate fell below 10 per cent, compared with as much as 20% in 2002(Kaczmarczyk and Okólski, 2008). A similar situation was observed in other NMS.Furthermore, the number of vacancies is rising rapidly. Almost 13% of Polish companiesreported hiring difficulties in the second quarter of 2007, compared to only 1.8%reporting such difficulties in 2005. The shortage of workers became particularly severe inconstruction and in manufacturing, and this situation is, again, a common feature of mostimportant migrant sending countries of the region (World Bank, 2007).However, even if there is a gradual improvement in the labour market of most sendingcountries, this can be attributed to out-migration only to a limited extent. Rather, asshown in the Polish country study, changes on the NMS labour markets can be attributedto a complex set of factors. Migration plays an important but not decisive role. Theimpact of mobility on labour markets in the region is largely exaggerated.Moreover, the most severe labour shortages are observed in construction, manufacturingand agriculture and as noted by Grabowska-Lusinska and śylicz (2008), these arepredominantly manual jobs. But at the national level, the drain effect is hardly visibleexcept in some specific cases such as medical professionals. This conclusion referspredominantly to Poland. In case of other countries, particularly the Baltic States, theoutflow may have far larger impact but statistical evidence is still missing.Kaczmarczyk and Okólski (2008) therefore suggested that recent outflow from Poland,and to some extent also from other countries of the region, should be regarded as a brainoverflow rather than a brain drain. The proposed “crowding-out” hypothesis can besummarized as follows. Due to long-lasting historical processes, the number of people inPoland, their spatial distribution and their human capital characteristics do not match theCMR 29

needs of a modern economy. Past migration from Poland, even in massive numbers, didnot have a significant impact on the population and economy, mostly due to positivenatural increase in the 1980s and 1990s. Recent mobility, for the first time in the modernhistory of Poland, may seriously influence labour market mechanisms, particularly if itincludes people living in villages or tiny towns with still visible remnants of thesubsistence sector. A large part of workforce in these areas can be seen as redundant ineconomic terms (both because of its excessive size and skill mismatch) and thereforeout-migration may be analyzed in terms of an overflow and not drain. 33 In this context itis useful to compare the data on migration selectivity with regard to education aspresented above with other data on migration structure. A change worth noting thatoccurred immediately after the 2004 EU enlargement was a decline in the proportion ofresidents of the rural settlements within the migrating population, as well as a rise in thenumber of residents of the urban areas. A general tendency both in the pre- and postaccessionperiod (as shown in Table 19) was an overrepresentation of migrantsoriginating from rural areas (relative to the respective resident population) and, to alesser extent, from medium and small towns. However, at the country level, thedifferences were rather moderate.Table 19: All and short-term temporary migrants from Poland by type of residence(category of settlement) prior to migration, before and after EU accessionCategoryof settlementResidentpopulationAll migrants Of which short term(mid-2004) before after before afteraccession accession accession accessionTown, 100,000 ormore inhabitants29.1 21.0 23.3 20.1 24.0Town, up to100,000 inhabitants 32.3 35.8 32.3 35.5 35.7Village 38.6 40.5 38.6 44.4 40.3Source: Kaczmarczyk and Okólski, 2008.Changes in selectivity are more clearly visible when comparing persons with vocationaland post-secondary education originating from settlements of different type (Table 20).33 See also the Polish country study within this projectCMR 30

Table 20: Migrant selectivity indexes (SI) for post-secondary and vocational educationafter EU accession (all migrants), by categories of settlement (migrants’ placesof residence prior to migration)Category of settlement Post-secondary VocationalTown, 100,000 or more inhabitants 0.27 0.18Town, up to 100,000 inhabitants 0.55 0.18Village 1.10 0.46All settlements 0.42 0.30Source: Kaczmarczyk and Okólski, 2008.The selectivity analysis indicates that indeed the accession and particularly the openingof the British labour market to Polish migrant workers did not only attract more Poles tothe United Kingdom, but above all it made migration worthwhile for many more highlyeducated individuals (in particular males) originating from villages or medium and smalltowns. In general, a significantly stronger propensity to migrate can be observed amongpeople originating from economically backward regions, characterized by a highproportion of the population living in medium-sized or small towns and in villages, arelatively large semi-subsistence sector, and very limited employment opportunities. Dueto recent migration these regions lost many young and highly educated persons. Anincreasing number among those migrants were newcomers to the labour market, peoplewho had just completed their formal education. To assess the impact of recent migrationfrom Poland on human capital formation and the situation in the labour market it isnecessary to consider the structure of opportunities. Having in mind structural features ofrecent migrants and characteristics of their domestic regional and local labour markets,this kind of migration can be easily described in terms of brain overflow (outflow of anexcessive supply of labour) and might be seen as a relief (rather than a threat) for thePolish labour market.Nevertheless, a few remarks need to be made. First, the long-term impact of recentoutflow is unknown. It may be true that even if the brain drain effect is not visible in theshort term, the migration of highly skilled may have detrimental effects in the long-run.Second, the impact of the outflow on the attractiveness of Poland and other countries ofthe region for foreign investors is hard to estimate. However, one has to note that cheapand relatively skilled labour constituted one of the most important competitiveadvantages of the NMS economies. Thus, we cannot exclude the possibility that highlyskilled mobility will influence the scale of future FDI inflows and their structure.6.3.2 Brain effectOne of the critical assumptions of the theoretical model presented by Beine et al. (2001)is that human capital (acquired through education) is not only transferable but also isrewarded a higher return abroad. This assumption implies, in turn, that migration maypositively influence the motivation to gain higher education and thus turn brain drain intobrain gain. It is therefore important to analyse the position of migrants in receivinglabour markets and examine to what extent skills of current migrants are employed in anCMR 31

efficient way and whether there is a wage premium for skills which could induce thosewho stayed in sending country to acquire more human capital.The UK Worker Registration Scheme (WRS) data may serve as the basic source ofinformation (Accession Monitoring Report 2008). If we assume that the number andstructure of applications to the WRS can be treated as an accurate measure of grossinflows, the WRS data allow one to build quite a precise picture of contemporary labourmigration to the UK. The data reveal that migrant workers from the NMS tend toconcentrate in only five sectors, among them administration, business and management(39%), hospitality and catering (19%), agriculture (10%), and manufacturing (7%) playthe most prominent role. (cf. Fig. 5).Figure 5: Top five sectors in which registered EU-8 workers are employed, May 2004 -March 2008Source: Own elaboration based on Home Office WRS dataThe high share of NMS migrants in “Administration, Business and Management” mightsuggest that these migrants achieve a relatively successful position in the UK labourmarket. However, this picture is largely misleading. It turns out that jobs in this sectorare mainly simple jobs which do not demand high skills. It is therefore more useful toexamine data on occupations rather than sectors. 34 Among the top occupations, suchposts as process operative (over 212,000 applicants, 27% of all recorded), warehouseoperative (63,590, 9%), packer (46,515, 6%), kitchen and catering assistant (44,810,6%), cleaner, domestic staff (42,120, 5%) or farm worker (32,515, 4%) dominate. Noneof these occupations could be described as demanding high level of skills or education.Only minor changes were recorded since May 2004.As a next step we look at the wage level of different groups of migrants in the UK labourmarket in order to assess the impact of education acquired on the labour marketperformance and throughout test the hypothesis of the existence of a brain waste.CMR 32

Methodological issuesThe basis for comparison of the return to education of Polish (and other NMS) workers will be the UKLabour Force Survey (LFS) and Poland’s Badanie Aktywności Ekonomicznej Ludności (BAEL, or PolishLFS). Both surveys are conducted quarterly. In the case of the UK LFS, data analysed contains all of thequarters of 2000-2007. The Polish LFS data used for reference comprises two quarters: the secondquarter of 2002 (i.e., two years prior to accession), and the second quarter of 2006 (i.e., two years postaccession).From the overall UK LFS survey, only the records of Polish, other NMS8 – later referred to as NMS7,NMS2 and EU15 (other than UK) migrants have been used (In general, the immigrants from NMS8 otherthan Poland could not be treated separately due to the size of the sample population). Out of these, lessthan 2/5 could be used due to the fact that the question on net earnings of the persons interviewed isasked only twice during the five interview waves. Contrary to Drinkwater et al. (2008), we have decidedto analyse both the first wave and fifth wave responses: LFS is a household survey, and it was often thecase that the immigrants interviewed within the same household in the first and fifth wave were differentpersons, and the repetitions of interviewed persons were not very common. Thus we decided to take intoconsideration two waves which allowed us to obtain larger samples.The Polish and NMS7 migrants were divided into three categories. First, those who arrived in the UK forthe first time for working purposes prior to the EU enlargement. These are referred to as pre-accessionmigrants. Second, those who arrived in 2005 or later, who are referred to as post-accession migrants.Third, those who arrived in 2004, but who were interviewed later, could not have their arrival dateaccurately determined and have been excluded from further analysis. This division reflects the fact thatPolish and NMS7 citizens were not granted free access to the UK labour market prior to the 2004enlargement. The legal status of EU15 nationals on the UK market has not changed over the period2000-2007 dates, so this group may be considered as homogenous. Also, even after their EU accession in2007 the Romanians and Bulgarians have not been granted full access to the British labour market, sothis group of labour migrants may also be considered homogenous.The most straightforward way of measuring the level of skill acquired by an individual is his/her highestacquired level of education. This method can not be used for immigrants (especially Polish) featured inthe UK LFS data, due to the fact that the immigrant qualification levels cannot be mapped easily on theBritish scale included in the LFS, resulting in a great proportion of responses (both from the higher andlower end of the scale) falling into the “other” category. Therefore, following Clark and Drinkwater(2008), migrants’ level of education is measured by the age left full-time education. Note that thisvariable can be easily transformed into a more intuitive one such as number of years of education.Subsequently, years of education were considered an educational proxy also for Poles on the Polish labourmarket, although, of course, an appropriate rank variable has more explanatory power in this case.Several illustrative characteristics of the sample populations are depicted in Table 21. Theresults for the Polish LFS and the UK LFS are not fully comparable, especially as regardsthe income level. However, it is worth noting that the average income in Poland hasincreased between 2002 and 2006 (the two reference periods considered, as a proxy ofthe labour market situation in the pre-accession and post-accession periods). On theother hand, the average income level of both pre-accession and post-accession Polishmigrants in the UK has steadily decreased in the post-accession period (for migrants fromother NMS8 countries this was not the case). EU15 citizens earn, on average, more than34 This is particularly true in case of administration, business and management whereby the problem is thatworkers in this sector work predominantly for recruitment agencies so could be employed in a variety ofoccupations.CMR 33

other EU migrants in the UK, although their average schooling level is lower. This may bean indication of a deepening brain waste.Table 21: Selected descriptive statistics, LFS samplesPL 02PL 06 PL 04 inUK UKNMS8 04 EU15 in UKin UKNMS2 inUKAverage (log) income 5.45 5.55 5.32 5.29 5.14 5.17 5.47 5.41Average years ofschooling (after 7) 13.6 14.1 14.1 13.2 12.9 11.5 12.4 13.5Average length ofemployment (months) 118 117 63 8 45 9 66 36Fraction of females 48% 48% 64% 42% 71% 46% 54% 65%Average age 39.0 39.5 38.3 28.0 34.3 28.7 37.3 34.3Source: own evaluation based on UK LFS (2000-2007) and Polish LFS (2nd quarters 2002 and 2006)The average level of schooling (measured by the years of education) of post-accessionmigrants in comparison to the pre-accession migrants has declined slightly (while, on theother hand, the level of schooling has increased in Poland in recent years). This couldmean that the brain waste effect is somewhat mitigated. The average age of a postaccessionmigrant is a decade lower than the age of a pre-accession migrant (however,one should note that the group of pre-accession migrants comprises persons who mayhave been living in the UK for decades).To measure the rate of return to human capital we will assess the wage level of differentgroups of migrants on the UK labour market. In order to do so, the data on net weeklypay of full-time workers were broken down by age groups and skill level. A variablereferring to age left full-time education was used as proxy of the highest level of formaleducation achieved. Tables 22-25 include data on net pay for three immigrant groups(NMS7 citizens, Poles, EU14 citizens) and for UK-born workers. In all cases the wage dataare expressed both in nominal terms and relative to the average pay in the respectivegroup.CMR 34

Table 22: Net weekly pay of full-time workers from NMS7 in the UK (nominal and relativeto the average)Pre-accession migrantsAge groupsPost-accession migrantsAge groupsAge left fulltimeeducation15-20 21-29 31-45 45+ Total 15-20 21-29 31-45 45+Total- 227.00 150.00 187.50 182.43 - 162.50 - 216.14 204.22Less than 15- 82.3 54.4 67.9 66.1 - 76.7 - 102.0 96.3- 240.20 217.50 279.30 252.33 171.40 195.77 176.78 182.00 184.1416 to 17- 87.0 78.8 101.2 91.4 80.9 92.4 83.4 85.9 86.9138.33 232.37 250.26 257.06 238.05 180.57 210.06 198.25 213.56 207.2318 to 2050.1 84.2 90.7 93.1 86.3 85.2 99.1 93.5 100.8 97.8226.77 430.46 423.11 350.63 - 217.05 251.89 323.73 255.18More than 2182.2 156.0 153.3 127.1 - 102.4 118.8 152.7 120.4163.00 219.00 471.00 - 284.33 - 251.00 - 195.00 223.00Students59.1 79.4 170.7 - 103.0 - 118.4 - 92.0 105.2144.50 230.98 321.33 315.72 275.96 176.75 209.27 206.59 226.37 211.97Total52.4 83.7 116.4 114.4 100.0 83.4 98.7 97.5 106.8 100.0Source: own elaboration based on the LFS dataTable 22 shows that well educated migrants from the NMS7 acquired around 20% higherpay than the average (post-2004), however, the difference was lower in case of postaccessionmigrants than in case of those who were employed in the UK prior to theEU-enlargement. This effect is clearly visible while analyzing data on nominal weekly pay(255 versus 350 GBP), and particularly in case of persons aged 31-45 (252 versus 423GBP).Table 23: Net weekly pay of full-time workers from Poland in the UK (nominal andrelative to the average)Pre-accession migrantsPost-accession migrantsAge left full-timeeducationLess than 1516 to 1718 to 20More than 21StudentsTotalAge groupsSource: own elaboration based on the LFS dataAge groups15-20 21-29 31-45 45+ Total 15-20 21-29 31-45 45+ Total- 231.00 174.00 181.50 192.00 - 266.75 176.00 219.50 226.00- 73.8 55.6 58.0 61.4 - 117.4 77.5 96.6 99.5120.00 200.00 242.33 257.17 243.12 145.67 190.50 226.10 195.08 197.2438.4 63.9 77.5 82.2 77.7 64.1 83.9 99.5 85.9 86.862.50 234.35 279.94 261.55 250.89 207.22 202.81 220.47 236.42 217.6520.0 74.9 89.5 83.6 80.2 91.2 89.3 97.1 104.1 95.8-274.83 394.57 393.38 354.54 - 223.97 306.04 255.99 244.6787.9 126.1 125.7 113.3 - 98.6 134.7 112.7 107.7120.00 - - - 120.00 518.00 - - - 518.0038.4 - - - 38.4 228.1 - - - 228.191.25 260.45 352.77 334.35 312.83 212.95 212.36 249.97 240.42 227.1429.2 83.3 112.8 106.9 100.0 93.8 93.5 110.1 105.8 100.0In the case of Poland, the average weekly pay of persons with the highest level ofeducation was significantly lower in the post-accession period – in nominal terms thedifference equalled (on average) over 100 GBP, in relative terms over 5 percentagepoints. Contrary to the NMS7, migrants relatively higher wages were noted in case of welleducated persons aged 31-45. Generally, the wage level of Polish workers in the UK wasslightly lower than for citizens of other NMS.CMR 35

Table 24: Net weekly pay of full-time workers from the EU14 in the UK (nominal andrelative to the average)EU15 immigrantsAge groupsAge left full-time education15-20 21-29 31-45 45+ Total172.13 242.40 249.44 256.05 250.82Less than 1555.4 78.1 80.3 82.5 80.8165.62 275.65 314.53 303.23 294.9516 to 1753.3 88.8 101.3 97.6 95.0176.14 253.78 360.63 372.94 324.3618 to 2056.7 81.7 116.1 120.1 104.5348.29 523.33 519.99 464.66More than 21112.2 168.5 167.5 149.6224.00 220.28 385.00 342.67 240.07Students72.1 70.9 124.0 110.3 77.3114.40 270.01 359.05 306.48 310.53Total36.8 87.0 115.6 98.7 100.0Source: own elaboration based on the LFS dataTable 25: Net weekly pay of full-time native workers in the UK (nominal and relative tothe average), 2002 and 2006Age left full-timeeducationLess than 1516 to 1718 to 20More than 21StudentsTotal15-20 21-29 31-45 45+ Total 15-20 21-29 31-45 45+ Total150.00 246.14 256.29 254.61 253.00 143.08 281.14 303.28 294.49 293.4847.9 78.6 81.9 81.3 80.8 40.4 79.3 85.5 83.1 82.8158.53 245.59 297.77 310.96 283.04 166.50 269.44 330.91 335.33 314.9050.6 78.4 95.1 99.3 90.4 47.0 76.0 93.3 94.6 88.8166.72 257.02 356.45 369.97 316.42 187.90 272.45 392.81 414.18 354.2153.3 82.1 113.9 118.2 101.1 53.0 76.8 110.8 116.8 99.9325.44 510.08 474.45 438.32 361.92 530.41 550.13 480.74103.9 162.9 151.5 140.0 102.1 149.6 155.2 135.6154.00 188.14 - - 176.76 187.30 226.13 300.00 - 209.5849.2 60.1 56.5 52.8 63.8 84.6 59.1159.91 271.48 343.04 321.44 313.07 171.26 304.20 383.30 367.79 354.5451.1 86.7 109.6 102.7 100.0 48.3 85.8 108.1 103.7 100.0Source: own elaboration based on the LFS data2002, 2nd quarter 2006, 2nd quarterAge groupsAge groupsTables 24 and 25 show that there is a completely different pattern in the case of EU14migrants and native workers (UK-born). With regard to migrants from the EU14, theaverage weekly pay was almost 50% higher in case of highly skilled than it was for anaverage worker. In case of native workers this difference was not that high (36%) butstill position of the well educated on the labour market was quite favourable (in bothperiods under consideration).All the above suggests that recent migrants from NMS cannot secure a wage level whichwould be relevant to their skill level. On the other hand, return to education on the UKCMR 36

labour market was the highest in case of the EU14 workers. As a point of referenceinformation on the weekly pay of workers in Poland has been provided (Table 26). 35Table 26: Net weekly pay of full time workers in Poland (nominal, in PLN and relative tothe average), 2002 and 2006Age left full-timeeducationLess than 1516 to 1718 to 20More than 21StudentsTotal15-20 21-29 31-45 45+ Total 15-20 21-29 31-45 45+ Total600.20 795.22 866.45 843.86 842.10 640.00 943.96 935.94 905.09 920.4455.1 73.0 79.5 77.4 77.3 51.4 75.8 75.2 72.7 73.9763.75 762.32 923.94 936.40 908.71 890.91 900.13 1003.52 1105.78 1034.8870.1 69.9 84.8 85.9 83.4 71.6 72.3 80.6 88.8 83.1650.18 897.94 1040.95 1082.99 1012.36 709.41 959.06 1168.31 1165.85 1118.9559.6 82.4 95.5 99.4 92.9 57.0 77.0 93.8 93.6 89.91132.86 1369.78 1460.46 1343.36 1197.53 1583.04 1806.93 1551.79103.9 125.7 134.0 123.2 96.2 127.1 145.1 124.6606.15 943.28 1390.65 1390.00 1050.47 651.75 960.99 1498.24 1598.33 1072.7655.6 86.5 127.6 127.5 96.4 52.3 77.2 120.3 128.4 86.2640.33 958.55 1131.88 1154.09 1090.02 705.00 1041.35 1301.39 1346.87 1245.0558.7 87.9 103.8 105.9 100.0 56.6 83.6 104.5 108.2 100.0Source: own elaboration based on the LFS data2002, 2nd quarter 2006, 2nd quarterAge groupsAge groupsFrom Tables 26 and 23 it follows that while significant increases in nominal wages wereobserved for skilled workers in Poland, the situation on the UK labour market wascompletely different. The same holds true in the case of relative values. In the case ofpeople with the highest level of education employed in Poland the average weekly payrelative to the average pay increased from 123 to 125% (between 2002 and 2006), whilein the UK labour market a decrease has been noted. Although the “education premium” inPoland not that high, it is still significantly higher than in the UK.These observations are supported by Marcinkowska et al. (2008). Their analysis, basedon the Polish LFS data as well as other data on earnings (the October Earnings Survey),shows that people with tertiary education constitute the only group with a serious wagepremium for skills (see Table 27). Additionally, education is one of the main factorsexplaining variance in earnings on the Polish labour market. Its importance (measured byTheil coefficient) increased from 12% in 1996 to around 22% in 2004. Thus, we couldconclude that incentives to invest in higher levels of skill is the result of the situation inthe Polish labour market itself, rather than from any returns to education fromemployment abroad.35 One has to note that this data are hardly comparable with the data presented above. This is not due todifferent currencies (relative values can be used), but mainly due to selectivity patterns among Polishmigrants. As clearly stated above, Polish migrants — particularly in the post-accession period — do notconstitute a random sample of the total population. In contrast, in the case of migration to the UK a clearpositive selection of migrants is visible. Additionally, specific migration strategies (e.g. short term or circularmobility) may also significantly influence the wage level. Thus, such a comparison may be biased due to selfselectionproblems.CMR 37

Table 27: Average net earnings in selected group of workers according to Polish LFS, inPLN, 2000-2006, by education2000 2001 2002 2003 2004 2005 2006Total (in thous.) 1,023.96 1,103.72 1,132.21 1,132.94 1,162.57 1,213.05 1,298.72tertiary 145.81 148.07 145.37 141.79 140.41 139.49 140.04secondary technical orpost-secondary 98.95 99.28 98.09 97.99 97.20 97.17 95.79secondary 97.51 98.73 99.53 95.05 94.13 91.81 91.76vocational 87.07 84.62 85.14 85.24 85.00 84.11 82.93primary or lower 76.70 75.57 74.79 75.51 74.79 73.72 71.88Source: Marcinkowska et al., 2008Average earnings = 100A similar conclusion with regard to the position of NMS migrants in the UK labour marketcan be drawn from the analysis provided by Clark and Drinkwater (2008), who showedthat, according to the UK LFS data, the rate of return to human capital is far lower formigrants coming from the NMS than it is for natives or migrants from the EU15 countries.The authors of the UK country study conducted within this project derive similarconclusions: the returns to education for the NMS migrants in the UK are relativelysmaller than the returns experienced by the natives; this effect is most visible for postaccessionmigrants, whose returns are four times lower than returns to natives. However,the return to education increases with the duration of a migrant’s stay. Apart from aneconometric analysis, the authors also conduct a comparative analysis of theoccupational structure of the natives and migrants, by educational attainment. Thisexercise proves that, in general, NMS migrants are employed in less-skilled (and lowerpaying) occupations than their equally educated native counterparts. This effect is moststriking for post-accession migrants, among whom a remarkable 36% of highly educatedare employed in elementary occupations, compared to 1% of similarly educated natives.All this evidence serves as a clear indication that a “brain waste” effect occurs for highlyskilled NMS migrants.However, the analysis of the effects of the performance of migrants abroad and the scopeof the brain waste is incomplete without a comparison to the sending-country situation.Even if highly-skilled workers from the NMS are employed in low skill occupations, theymay still more efficiently employed than at home, which may mitigate the brain wasteeffect. Fihel et al. (2008) prove this is not so in the case of Poland (for whom the brainwaste effects are the highest).Comparing Polish (and other NMS) migrants in the UK LFS and the Polish LFS, Fihel et al.(2008) employ a Mincerian framework to assess the returns to education. For thedifferent migrant groups considered, the return to skills is the highest amongst EU15 andNMS8 (except Polish) nationals in the UK. A similar, but slightly lower, level of return toeducation may be found among the pre-accession migrants from Poland. On the otherhand, post-accession migrants from Poland and migrants from Bulgaria and Romania(NMS2) have dramatically lower returns to education: each additional year of educationon average brings about an increase of net earnings 2.5 smaller than for the othergroups.CMR 38

A comparison of the coefficients on the education variables for Polish migrants and theresident population of Poland, for the pre-accession and post-accession periods suggestthat the returns to skills in Poland have not changed much between 2002 and 2006. Onthe other hand, the returns to education for pre-accession and post-accession migrantsfrom Poland have changed dramatically. Even bearing in mind the possible differencesarising from the fact that the UK LFS and the Polish LFS are not perfectly comparable,one may assume that while the pre-accession migrants in the UK experienced a higherreturn to skills in the UK than they could have in Poland, in the case of the post-accessionmigrants the situation is quite the opposite. Therefore, recent flows suggest more ofbrain waste characteristics.Fihel et al. (2008) also look for other indications of a brain waste. Evidence of the factthat recent Polish migrants’ skills are not put to the best use in the UK is also derivedfrom the dramatic decline in the explanatory power of the econometric model specified(for those migrants), when compared to older migrants or migrants originating fromother countries. The fact that all the variables included in the model, such as education,experience and individual demographic characteristics, do not explain the variance ofindividual earnings as well as in the case of other migrants is a clear indication that thesevariables on the whole have less impact on the income level of an individual. This meansthat the jobs undertaken by recent Polish migrants in the UK have little to do with theirtrue skills.Thus, a combined analysis of the UK and Polish LFS suggests that there is, indeed, abrain waste observed among Polish skilled migrants in the UK, which has increased inscale after Poland’s EU accession. In other words, recent NMS migrants’ skills are not putto their best use. This is particularly true in case of Polish workers. Additionally, as statedabove, people with tertiary education constitute the only group on the Polish labourmarket which acquire a skill wage premium. Such a situation may have seriousconsequences both in terms of human capital formation in Poland (decline in propensityto acquire higher skills for those who are planning going abroad) as well as regardingfuture of labour mobility. One can imagine that those whose skills are not being usedefficiently are more prone to go back if the return to education is higher on the Polishlabour market. Thus, the recent position of migrants from NMS in the UK (or in the EU15more generally) labour market may significantly influence future dynamics of migration(in a negative way) and scale of returns (in a positive way). However, with regard to theformer point, the strategies of migrants should be considered as well. For short-termmigrants, their position in the labour market in destination country is perhaps not asimportant compared to those who intend to to settle there. 36 Additionally, one has toremember that statistical evidence shows significant differences between migrants fromselected NMS. This refers both to their structural characteristics as well as to position onthe labour market in destination countries, including the wage premium for skills. In thecase of Poland a large part of recent mobility is to be explained in terms of brainCMR 39

overflow. However, in the case of the Baltic States, migration is far more significant inrelative terms (migrants as a share of sending population or workforce) and thus itsimpacts on labour market phenomena (unemployment, wages, shortages) may begreater.Last but not least, the fact that the recent outflow from Poland, which may be considereda selective mobility of the well-educated, has the characteristics of a brain overflowsignifies also that the two main effects of the outflow of skilled workers — namely, thebrain drain effect and the brain gain effect — are less visible and very hard to trace. Thebrain gain effect is hard to assess primarily because the time span available for analysisis too short. The number of years that have passed since the NMS accession is smallerthan or equal to the number of years necessary to obtain higher education. Therefore,although a trend of the growth of the popularity of education may be observed, it isimpossible to determine whether this is partly a result of accession. On the other hand,the brain drain effect is hard to assess in general, due to the different scales ofqualification mismatches in specific sectors and regions in Poland and the fact that thelevels of educational attainment are still growing for reasons other than migration, whichhave a far more significant (positive) impact on the number of highly skilled.7 Case studies7.1 Mobility of health care professionalsOne of the most controversial issues in the world-wide debate is the migration of medicalprofessionals. This phenomenon is above all a consequence of the permanent demand forthis type of migrant in highly developed countries, mainly due to unfavourabledemographic trends as well as fluctuations in labour markets. Additionally, this fieldrepresents a typical example of intangible services where the human capital flow cannotbe easily substituted with mobility of goods and services. In effect, potential immigrantsmay expect highly beneficial financial and social conditions, integration support and, in atleast several receiving countries, simplified immigration procedures. There are thereforestrong pull factors to encourage migration among medical professionals from the NMS.Data on the mobility of medical professionals from the NMS is rather limited, but most ofthe data sources do not indicate dramatically high level of migration. According to theOECD data bases only two EU10 countries were noted among these with relatively highexpatriation rates among doctors and nurses. In the case of Hungary respective numberswere as high as 2,538 doctors (expatriation rate: 7.2%) and 2,117 nurses (2.4%). In the36 Additionally, one may argue that absolute wages are crucial for short-term migrants while long-termmigrants and those who decide to settle are oriented towards higher relative wages. In other words, if weCMR 40

case of Poland they equalled 5,821 doctors (5.8%) and 9,153 nurses (4.6%). Note that insome countries expatriation rates among doctors were over 10%, and a few cases higherthan 50% (International Migration Outlook, 2007).The major shortcoming of the above presented data is that they do not allow one toassess the scale of recent outflows. These data refer to the stock of migrants which is anoutcome of cumulative inflows in last decades. Apart from the OECD data, which is basedon censuses and registers (partially also on the LFS data), information about the mobilityof medical professionals is rather anecdotal. One typical example is the outcome of astudy on migratory potential among heath care professionals completed shortly beforethe EU enlargement. This study showed that a large proportion of medical professionalsat least planned to go abroad. In the case of Hungary, 25% of all health careprofessionals declared a definite plan and another 48% an intention to leave. In theCzech Republic results were quite similar, and only slightly smaller in case of Poland.Research conducted among Estonian health care professionals gave the result of 5.4%respondents (which is about 700-800 individuals) who had definite plan to work abroad,17.9% who developed such plans and 32.3% who had vague plan. Only 44.4% ofrespondents did not take the migration into account. These results suggest a ratherdramatic picture of mass students’ and professionals’ outflow. However, the samesurveys reveal the temporary character of intended emigration. Of those Estonianmedical professionals who want to work abroad, only 6.5% want to leave the homecountrypermanently, with 44.5% intending to live abroad for two years and 22% for onlya few months. The percentages of physicians and nurses who want to emigratepermanently (in those who want to emigrate at all) were as high as 25%for Poland, 11%for the Czech Republic, 7% for Hungary and 5% for Lithuania (Andres, Kallaste, Priinits,2004, Aidis, Krupickaitè, Blinstrubaitè, 2005).Of course, it is hard to assess to what extent individuals’ migration intentions wereactually realised in the post-accession period. Recently published data from the UKGeneral Medical Council shows that scale of the phenomenon is relatively low, althoughdoes seem to be increasing over time (Pollard et al., 2008).apply the relative deprivation approach (Stark, 1991) the reference group for the former is sendingcommunity while for the latter it is receiving society.CMR 41

Figure 6: Number of doctors born in NMS registered to the General Medical Council,2005-2007Source: Pollard et al. 2008According to the General Medical Council data, between 2005 and 2007 an increase ofover 25% were recorded with regards to registered doctors born in NMS and working inthe UK. Three countries — Poland, Hungary and the Czech Republic — are responsible formost of the inflow in the post-accession period.In the case of Poland, some indication of the scale of potential migration of medicalprofessionals is provided by the issuing of certificates confirming qualifications andprofessional experience required by employers in Western European states. The numberof issued certificates (6,724 as of the end of December 2007) amounted to 5.7% of thetotal number of medical doctors in Poland. In the case of dentists, certificates were issuedto 1,924 persons (6.3% of the total). For semi-skilled medical staff, around 9,300certificates were issued to nurses and midwives, which amounts to 0.3% of thisprofessional group in Poland.It follows that migration of the so-called ‘white personnel’ is a noticeable phenomenon.However its scale is not so large as to pose a threat to the healthcare system in theshort-term. This threat is not that significant because the Polish educational systemproduces medical professionals at a rate still higher than their potential outflow to otherstates. In fact, to some extent migration of medical specialists may be viewed as anotherexample of overflow rather than a drain of workers. This may be particularly true in thecase of young professionals trapped in organizational structures with limited chances forpromotion. Nonetheless, the outflow of medical doctors may be painful in the case ofcertain specializations.CMR 42

Table 28: Certificates issued to Polish medical professionals – specialties with thehighest number of certificates issued and the highest share in total number ofactive specialists (May 2004 – June 2006)SpecialtyNo. of economicallyactive doctorsNo. of certificatesissuedShare of certificates in thetotal no. of specialistsSpecialties with the highest number of certificates issuedAnaesthesiology 3,984 625 15.6Surgery 5,395 334 6.1Orthopedics 2,261 168 7.4Internal diseases 11,792 163 1.38Radiology 1,993 154 7.7Specialties with the highest relation of certificates issued to the number of active specialistsAnaesthesiology 3,984 625 15.6Plastic surgery 142 21 14.7Chest surgery 218 28 12.8Radiology 1,993 154 7.7Orthopedics 2,261 168 7.4Total 81,346 3,074 3.7Source: Kaczmarczyk and Okólski, 2005; Kaczmarczyk, 2008; Ministry of Health.Table 28 shows that this especially refers to anaesthesiology (here the percentage ofpotential migrants amounted to almost sixteen percent), chest surgery (12.8%), plasticsurgery (14.7%), as well as radiologists (7.7%). The outflow problem has a considerableimpact on specialties of the most difficult position in terms of income on the Polish labourmarket (anaesthesiologists, radiologists) or of high demand on foreign labour markets(plastic surgeons). Moreover, a temporary or permanent imbalance on local and regionallabour markets is likely to happen.7.2 Mobility of studentsAs shown in previous parts of the study, if we consider recent migration from the NMStowards those countries that opened their labour markets already in 2004 (e.g. UK,Ireland) it is the young and well-educated who migrate. At the same time, theory andevidence also indicate that there is a strong connection between student mobility andsubsequent labour mobility. The last few years have brought about a change in policiestowards highly skilled migration. Many industrialised countries introduced targetedpolicies in order to attract foreign talent. Amongst the bundle of measures to recruithighly skilled migrants are in most cases also measures targeted on the retention offoreign graduates. While these policies are mainly designed for third country nationals,European governments also strengthened their activities to gain international graduatesfrom European countries (Mechtenberg 2005; Bologna Process Working Group 2007,Universities UK 2008b). However, policies towards foreign graduates from the NMSremained ambivalent. While certain countries (UK, Ireland, Sweden) opened their labourmarkets already in 2004 and treat NMS graduates as other EU citizens, other countries(Austria, Germany) apply transitional regulations also to NMS graduates (BMAS, 2006,2007). On the 1 st of November 2007 Germany introduced a new regulation and facilitatedCMR 43

the labour market entrance of NMS graduates. Although they still need a work permit thepriority check is abolished. NMS graduates need to show a work offer and the local labouragency in charge will issue the document. Before the 1 st of November 2007, the labouragency checked whether there were any other Germans, EU-citizens or persons holding apermanent work permit before they would issue the work permit for the NMS-graduate(BMAS 2007).Statistical evidence about retention rates is available from overseas immigrationcountries because their immigration authorities collect data on the change from one visacategory to the other. Such data shows great variation across disciplines, sendingcountries and levels of education involved. For the US, in a long-term perspective it isestimated that around 58% of the former PhD-students are retained (Suter and Jandl,2006). In Europe, data on retention rates is best available for non EU nationals becauseEU citizens do not have to apply for a work permit. In Sweden, data on work permitsshow that the proportion of guest students who applied for a work permit between 2000and 2005 varied significantly by nationality (32% Iran, 6% USA) (Suter and Jandl,2006). In the UK there is a source of information on the retention of graduates with EUcitizenships since they are included in a survey on the destination of university leaverssix months after graduation. The numbers for the last years indicate that the retention ofEU graduates in the UK is rising. While in 2000/01 19.3% of all respondents took up workin the UK, in 2004/05 the number rose to 26.6% (Suter and Jandl 2006). In 2006/07,18% of the non-UK EU students indicated that they wanted to take up a full-time job inthe UK six month after graduation (Department for Innovation Universities and Skills2008).In 2005 the EU25 countries hosted over 1.1 million international students (UNESCO,2007). A considerable part of this is a result intra-European mobility. In 2004 2.2% ofthe total European student population (401,000 students) were enrolled at a university inanother European country for at least one year (Eurydice, 2007). These numbers excludedata on mobility in European programmes so we should add another 144,000 mobileErasmus-students in the academic year 2004/05. By 2006/07 the number of Europeanstudents participating in Erasmus increased to 159,000. In fact, Erasmus can be regardedas a motor of European student mobility and the increase in short term mobility is mainlydue to European programmes. In contrast, European degree mobility increased onlymoderately (Teichler, 2007). France, Germany, the UK and the United States attracttogether more than 50% of all worldwide mobile students (OECD, 2007). Studentmobility in Europe follows very specific patterns. Former colonial powers (UK, Portugal,France, Belgium and Spain) still attract huge numbers of students from their formerterritories, while Austria and Germany import students from CEE. The Nordic countriesalso show a special relation to transition countries since they host a comparatively largenumber of students from the Baltic States (cf. Kuptsch, 2006).CMR 44

Table 29: Foreign students from EU-8+2 countries in selected target countries academicyear 2006/07source country Bulgaria Czech Rep. Estonia Hungary Lithuania Latvia Poland Romania Slovakia Slovenia totalGermany 12,170 2,132 724 2,434 1,667 886 14,493 4,156 1,569 524 40,755United Kingdom 710 1,150 535 1,040 1,485 880 6,770 740 890 285 14,485target countryFrance* 2,615 772** 3,188 4,675 11,250Austria 1,309 528 40 1,199 77 48 1,467 707 1,301 567 7,243Sweden 317 349 393 709 112 142 2,781 907 20 46 5,776Netherlands 500 350 100 400 150 100 1,250 300 150 3,300Finland*** 32 325 135 265 208 70 500 47 97 52 1,731Ireland 117 152 97 41 80 44 539 66 39 11 1,186* data contains only foreign students at Universities, ** 2006, *** data on international exchange studentsSource: own calculation based on target countries dataTable 29 gives the number of NMS students enrolled in the eight target countries in2006/07. Germany is the most important target country for student migration from CEE.Over 40,000 students from the region were enrolled at German higher educationinstitutions. The second and third most important countries (UK and France) onlyrecruited approximately 15,000 and 11,250 students in this year. The data derives fromsources in the target countries. However, for the sake of more recent data (academicyear 2006/07) this brings about the disadvantage of a low comparability since the targetcountries use different concepts when producing data on student mobility. Wheneverpossible, data on international students were used (i.e. inwards mobile students) and inthe remaining cases data on foreign students (i.e. students with foreign citizenship).Table 30: Number of EU10 mobile students abroad in 2005outbound mobilitytotal abroadmost important target countriesratio (%)Poland 31,455DE (15,893); FR (3,217); USA (2,988); UK(2,183), AT (1,357), others (5,817) 1.5Bulgaria 26,272DE (12,913); USA (3,806); FR (2,903); AT(1,696); TR (1,111); others (3,843) 10.7Romania 21,672DE (4,520); FR (4,320); USA (3,360); HU(3,171); IT (1,521); others (4,780) 2.7Slovakia 18,747CZ (10,119); HU (2,341); DE (1,707); AT(1,515); USA (636); others (2,429) 9.9HungaryDE (2,881); AT (1,344); USA (976); FR (601); UK7,777 (584); others (1,391) 1.6DE (2,439); USA (942); FR (654); UK (606);Czech Rep. 7,057Austria (500); others (1,913) 1.9Lithuania 6,514DE (1,729); RF (1,376); USA (663); PL (558); LV(538); others (1,650) 3.6Estonia 3,580RF (1,057); DE (776); FI (599); USA (296); UK(187); others (665) 5.5Latvia 3,483DE (919); RF (884); USA (426); UK (271); EE(174); others (809) 2.8Slovenia 2,735DE (623); AT (619); USA (320); UK (317); IT(305); others (551)2.3Source: UNESCO Global Education Digest, 2007.Table 29 can be complimented with information originated from sending countries(UNESCO education statistics). Table 30 gives an overview about the most importantsending countries in quantitative terms, their main destinations and the rate of outboundCMR 45

mobility in comparison to all students enrolled in tertiary education. In quantitative termsPoland is the main sending country from the region. In 2005 31,455 Polish studentsstudied outside the borders of their home country. This is followed by Bulgaria (26,272students abroad), Romania (21,672), Slovakia (18,747) and Hungary (7,777). For eightof the 10 countries under consideration Germany is the most important target country.If we take the outbound mobility rate into consideration, we get an impression which ofthe sending countries has the most mobile students: The ranking is headed by Bulgarianstudents. 10.7% of all Bulgarians study outside of Bulgaria. On the second and thirdposition follow Slovakia (9.9%) and Estonia (5.5%). Interestingly, Polish students – whorepresent in quantitative terms the most important sending country – are the leastmobile. Only 1.5% of all Polish students are enrolled abroad.If we look at the changing patterns in mobility from the NMS between 2004 and 2005 thegeneral trend towards new target countries is already observable. In 2004 the UK wasthe fifth most important target country for Czech students. One year later it was alreadyon the fourth position. These findings are supported by the Eurostudent 2008 report.26% of the surveyed Czech students spent studies abroad in the UK, while only 20%went to Germany. In the Estonian case in 2005 the UK is included in the list of the fivemost important target countries for the first time. The number of Hungarian studentsenrolled in the UK increased between 2004 and 2005 from 371 to 584 students. Between2004 and 2005 the share of the UK as fourth most attractive country for students fromLatvia increased from 4.9% to 8%. In the case of Poland, the UK is for the first timeincluded in the top five of target countries for Polish students abroad (UNESCO, 2006,2007).Table 30 summarizes the observed trends in the intra-European mobility of students.Table 30: Enrolment trends with regard to NMS citizensAT DE FI FR GB IE NL SEPLgeneral trendexception fromthe general trendBG RO EE, SL RO,CZ CZ, RO,BG,Sl,SKSource: own elaborationAmong the target countries in Western Europe we may differentiate three groups:(1) One group of countries (AT, DE, FR) traditionally attracted many NMS students butfaces declining enrolments. In Austria enrolments from NMS decreased between 2003and 2004 but are slowly recovering. An exception is the number of Bulgarian studentsthat continues to decrease. A probable reason for the decrease between 2003 and 2004 isthe anticipated different tuition regulation which might have convinced potentialCMR 46

candidates to postpone their enrolment in Austria. The decreasing Bulgarian enrolmentmight be explained with a decreasing young population in Bulgaria and hence anincreasing supply of state funded university capacities back home. In quantitative termsGermany is the second largest target country for international students in Europe and themost important target country for students from the NMS. In the winter term 2006/07the Federal Statistical Office counted 246,369 students with foreign citizenship(Statistisches Bundesamt, 2007). Foreign students represented 12.4% of all studentsenrolled in Germany and educational foreigners represented 9.5% of all students inwinter term 2006/07. China is the most important sending country for educationalforeigners to German universities. But remarkably, two of the NMS countries follow in theranking. Bulgaria sent 11,816 educational foreigners to Germany in winter term 2006/07.Almost the same number of educational foreigners (11,651) came from Poland. Althoughthe numbers have been growing considerably over the last decade, recently we observe adecrease for most of NMS with the exception of Romania. The development becomeseven more visible if we take the numbers for newly enrolled educational foreigners fromNMS into consideration. Bulgaria and Poland rank on the second and third position of allinwards mobile students in Germany but, since the winter term 2005/06 numbers aredeclining. A possible explanation for the decreasing numbers of NMS students might bedemographic changes in the source countries in combination with entrance criteria fortertiary education. However, tuition fees are not responsible for the slow-down in recentyears because they were only introduced in summer term 2007 in some federal states inGermany. It has been argued that tuition fees will have an influence on the enrolment ofinternational students in Germany (CESifo 2007; DAAD, 2005). France also experienced adecline in the traditional strong enrolments of Polish and Bulgarian students. In France,we observe a general decline in overall and foreign enrolments which might explain thisdecrease as well. Romanian enrolment, however, is increasing which might be due tolinguistic ties.(2) A second group of countries is characterized by increasing enrolment from the NMS(UK, IE). The most interesting case is probably the changing pattern of student mobilitytowards the UK since 2004. The United Kingdom has a long tradition as a target countryfor mobile students. General arguments that explain the attractiveness of Great Britainfor education migrants are the perception of a high quality of education, English asinstruction language and comparatively short degrees (HEPI, 2008). From all EU15countries it has the highest intake of mobile students: over 2.3 million students areenrolled in higher education in the academic year 2006/07 (HESA 2008a). Both the totalenrolment and foreign enrolment are rising. The total number of students increasedbetween 2005/06 and 2006/07 by 1.1% (HESA, 2008a). Between 2004/05 and 2005/06the enrolment of foreign domicile students in the UK increased from 13.9% to 14.1%. In2005/06 there were 106,000 foreign domicile students from the EU enrolled in the UKand 224,000 international students (non-EU foreign domicile students) (UUK 2007). Noneof the NMS is a main supplier of students to the UK. However, this pattern seems likely tochange. Between 2005/06 and 2006/07 the numbers of Polish students increased by56%. Both the numbers of Latvian and Lithuanian students also grew considerably(HESA, 2008a). The main reason for this is the changing tuition fee policy. The UKCMR 47

charges a seven times higher fee for international students than for EU or home students.From the day of accession on NMS-students had to pay the home fee and studies inBritain became affordable. The situation even improved with the new tuition schemeintroduced in 2006/07. Before this, students had to pay the fees in advance but now,they are only charged upon graduation if they earn a certain amount (Aston, 2004).(3) A third group of countries (NL, SE, FI) attracts a smaller share of CEE students andshows diverse patterns which might be due to data restrictions. In the Netherlands, a feecharging country, the number of especially Polish students had been increasing butrecently numbers are decreasing. In Sweden we observe a downward trend for somecountries (PL, LT, HU, LV, EE) and a slight upwards trend for others (CZ, BG,RO, SK,SL).Finland only offers data on credit mobility but even in credit mobility there is a generaldownwards trend since 2006 with the exception of Estonia and Slovenia.At this stage it is not possible to assess how much influence labour market policy towardsNMS citizens had on their decision to study in the UK. But one may suppose someinfluence combined with a strong impact of tuition policies. Whereas there is little scopeof the possible destination countries to influence the slowing demand of study abroad inthe source countries due to demographic trends, destination countries will be in futureneed to develop targeted policies to those who are willing to go abroad. Recentdevelopments in the UK (new retention policies, report of retention rates, projections offuture demand, strong marketing, innovative tuition policies) gives reason to believe thatthe UK is in a good position to attract the high-skilled. In contrast, former market leaders(DE, AT) will probably see a further decline in numbers of NMS-students if they do notchange their strategies.8 ConclusionsThe economic literature argues that there may be positive and negative impacts from theoutflow of skilled workers. The theory suggests that there may be complex linkagesbetween the mobility of the highly-skilled and socio-economic processes in sending andreceiving countries. In particular, the analysis of the impact of the outflow on the sendingcountries (including impact on the human capital formation) cannot be separated from anassessment of the labour market performance of migrants in destination countries.As shown in Sections 3 and 4, descriptive statistics on the skill composition of migrantscannot provide unambiguous arguments for or against the brain drain (the selectiveoutflow of the highly skilled). To assess the scale and consequences of migration of thistype it is necessary to control for additional effects such as the evolution in enrolmentrates, labour market performance in sending countries, selectivity of migration withrespect to age and so on. Unfortunately, official data sources offer a relatively weak basisfor an analysis of the highly skilled mobility and its consequences. This is due both toincompleteness of migration data and other methodological issues (e.g. definitions). Weargue that reference to harmonised LFS data gives an opportunity to overcome well-CMR 48

known migration data limitations (as shown among others in the case of Poland) andtherefore we made extensive use of this particular data set.Our conclusions are as follows:(1) The scale of the brain drain from the NMS in the post-accession period has beenexaggerated. The apparent positive selection of migrants from the well-educatedis mostly due to demographic developments (in particular the age structure ofsending populations) and changes with regard to educational attainment.(2) The study reveals significant changes in the selectivity of migration from Polandwhich can be related to the migration policy of EU15 countries, and particularlythe introduction of Transitional Arrangements. As a consequence of these changesthere has been a significant shift in migratory trajectories: the UK seems to be thewinner and attracts relatively well skilled migrants while Germany remainsattractive mostly for poorly educated and relatively older individuals.(3) Analysis of the impacts of highly skilled mobility was presented in the frameworkproposed by Beine et al. (2001). The results strongly suggest that the outflow ofhighly-skilled from Poland should be interpreted in terms of the crowding-out (orbrain overflow) hypothesis rather than the brain drain hypothesis. This conclusion,however, does not necessarily hold true for all of the NMS.(4) The analysis of returns to human capital, based on the UK example, clearlysuggests that recent migrants from the NMS do not work in jobs which match theirskills or competencies. This suggests that “brain waste” may be a serious issue.This, in turn, implies that the incentive to invest in more human capital in thesource countries may be reduced, so that any “brain gain” effect is weakened.(5) In theory, the EU enlargement presented an opportunity for an extension of thebrain gain effect. The introduction of the Transitional Arrangements and attemptsto induce selective inflow of the highly skilled (as it is in case of Germany) was notsuccessful so far and therefore it would be difficult to assess its effects. On theother hand, in those countries which decided to open their labour markets formigrants from the NMS, the rate of return to education is very low. Consequently,the overall effect of the outflow of highly skilled is extremely difficult to estimate.(6) Information on the mobility of health care professionals from the NMS is ratherlimited, but most of the data do not indicate high levels of outflow. Nevertheless,imbalances in local and regional labour markets are already clearly visible andseriously influence public debates.(7) Recent years have witnessed an increase in the scale of students’ mobility fromthe NMS. However, the relative scale of the phenomenon remains very low. Forexample, while Poland is the main sending country of students in the region theoutbound mobility ratio is still less than 2%.CMR 49

(8) A comparison of the two most important students’ receiving countries (Germanyand the UK) shows that recent patterns of mobility are complex. A possibleexplanation for the decreasing numbers of NMS students in Germany might bedemographic changes in the source countries, in combination with entrancecriteria for tertiary education (tuition fee policy). On the other hand, changingtuition fee policies after the EU enlargement are possibly the best explanation forthe high growth rates of NMS citizens studying at universities in the UK.(9) According to the UK evidence, retention rates of students are relatively low(around 20%). This suggests that countries of origin can greatly benefit from therecent wave of student mobility, mostly via a positive impact on the human capitalformation after return.CMR 50

9 ReferencesAccession Monitoring Report. 2004-2008. London: Home Office: Department for Workand Pensions.Aidis, Ruta, Krupickaitë, Dovilë, and Lina Blinstrubaitë. 2005. “The loss of intellectualpotential: migration tendencies amongst university students in Lithuania”, Geografija2005, T 41, No. 2, pp. 33-40.Andres, Vork, Kallaste, Epp, and Marit Priinits. 2004. “Migration intentions of health careprofessionals: the case of Estonia”. In: Pop Daniel (ed.) New patterns of labour migrationin CEE. AMM Publishing House.Aston, Libby. 2004. Projecting Demand for UK Higher Education from the AccessionCountries, Higher Education Policy Institute, 82 available at:http://www.hepi.ac.uk/pubdetail.asp?ID=122&DOC=Reports, last access 2008-06-15.Baláz, Vladimír, Williams, Allan, and Kollár, Daniel. 2004. "Temporary versus permanentyouth. Brain drain: economic implications", in International Migration 42 (4), pp. 3-32.Barro, Robert, and Lee, Jong-Wha. 1993. “International Data on Educational Attainment”,in Journal of Monetary Economics, Vol. 32, pp. 363-94.Beine, Michel, Docquier, Frédéric, and Rapoport, Hillel. 2001. “Brain drain and economicgrowth: theory and evidence” in Journal of Development Economics, Vol. 64, No. 1 pp.275-89.Beine, Michel, Docquier, Frédéric, and Rapoport, Hillel. 2007. “Measuring InternationalSkilled Migration: A New Database Controlling for Age of Entry”, in The World BankEconomic Review, Vol. 21, No. 2, pp. 249–54.Beine, Michel, Docquier, Frédéric, and Rapoport, Hillel. 2008. “Brain drain and humancapital formation in developing countries: winners and losers” forthcoming in TheEconomic Journal.Bhagwati, Jagdish, and Hamada, Koichi. 1974. “The brain drain, international integrationof markets for professionals and unemployment”, in Journal of Development Economics,Vol. 1, No. 1, pp. 19-42.BMAS. 2007. Press Release (2007-09-19) Zuwanderung gestalten. Zugang ausländischerHochschulabsolventen wird erleichtert, available at:http://www.bmas.de/coremedia/generator/19492/2007__09__15__zuwanderung__gestalten.html last access 2008-06-14.Bologna Process Working Group. 2007. Key Issues for the European Higher EducationArea – Social Dimension and Mobility. Report from the Bologna Process Working Group onSocial Dimension and Data on Mobility of Staff and Students in Participating Countries,80 p., May 2007, available at:CMR 51

Brücker, Herbert, Defoort Cécili, and Grün, Carola. 2007. “Does Migration Trigger HumanCapital Investment?”, paper presented at the First Conference on “Transnationality ofMigrants, Riga, 26-27 October.Bundesministerium für Bildung, Wissenschaft und Kultur. Various years. Statistiken zumösterreichischen Universitätswesen, Statistisches Taschenbuch, available at:http://archiv.bmbwk.gv.at/universitaeten/stats/stat_tb.xml, last access 2008-06-15.Bundesministerium für Wissenschaft und Forschung. 2008. Studierende aus den neuenMitgliedsstaaten an Universitäten nach Schulformen, bm.wf Abt. I/9, communication viae-mail 2008-06-26.Carrington, William, and Detragiache, Enrica. 1998. “How big is the brain drain?”, IMFWorking paper WP/98/102.CESifo. 2007. Tuition Fees in Europe 2007/2008, http://www.cesifogroup.de/pls/guestci/download/CESifo%20DICE%20Report%202007/CESifo%20DICE%20Report%204/2007/dicereport407-db5.pdf, last access 2008-06-15.Chiswick, Barry. 1978. “The Effect of Americanization on the Earnings of Foreign-bornMen”, in The Journal of Political Economy, Vol. 86, No. 5, pp. 897-921.Clark, Ken, and Drinkwater, Stephen. 2008. ‘The Labour Market Performance of RecentMigrants’. Oxford Review of Economic Policy (Autumn) (in press).Csedo, Krisztina. 2008. “Negotiating Skills in the Global City: Hungarian and RomanianProfessionals and Graduates in London”. In: Journal of Ethnic and Migration Studies, Vol.34, No. 5., pp. 803-823.DAAD. 2005. Protokoll der AG 1: Studiengebühren: Chancen und Risiken für dasAuslands- und Ausländerstudium, 2005-11-17, 2 p., available at:Department for Innovation, Universities & Skills. 2008. European Union StudentsStudying in EnglishDocquier, Frédéric, and Marfouk, Abdeslam. 2004. “Measuring the international mobilityof skilled workers - Release 1.0,” Policy Research Working Paper n. 3382, World Bank(August 2004).Docquier, Frédéric, Faye, Ousmane, and Pestieau, Pierre. 2008. “Is migration a goodsubstitute for education subsidies?” forthcoming in Journal of Development Economics.Drinkwater, Stephen, Eade, John, and Garapich, Michał. 2008. Poles apart? EUEnlargement and the Labour Market Outcomes of Immigrants in the UK, mimeoDumont, Jean-Christophe, and Lemaître, Georges. 2005. “Counting immigrants andexpatriates in OECD countries: a new perspective”, Mimeo: OECD.CMR 52

Egger, Hartmut, and Felbermayr, Gabriel. 2007. “Endogenous Skill Formation and theSource Country Effects of International Labour Market Integration”, CESifo Working PaperNo. 2018.Eurydice, Eurostat, European Commission. 2007. Key data on higher education in Europe,available at: http://www.eurydice.org/ressources/Eurydice/pdf/0_integral/088EN.pdf lastaccess 2008-06-14.Fihel, Agnieszka, and Kaczmarczyk, Paweł. 2008. “Migration – a threat or a chance?Recent migration of Poles and its impact on the Polish labour market”. In: Burrel, K., ed.Polish Migration to the UK in the 'New' European Union: After 2004, London: Ashgate(forthcoming).Fihel, Agnieszka, Kaczmarczyk, Paweł and Okólski Marek. (2006). “Labour Mobility in theEnlarged European Union: International Migration from the EU8 countries”, CMR WorkingPaper No. 14/72, Centre of Migration Research, Warsaw.Fihel, Agnieszka, Kaczmarczyk, Paweł, and śylicz Anna. 2008. Brain drain or brain gain?Post-accession mobility of the highly skilled from Poland and its consequences, CMRWorking Papers (forthcoming).Grabowska-Lusinska, Izabela, and Anna śylicz. 2008. Czy polska gospodarka potrzebujecudzoziemców? [Does the Polish economy need foreign labour?]. Warsaw: CMR UW.Grubel, Herbert, and Scott, Anthony. 1966. “The international flow of human capital”, inAmerican Economic Review, Vol. 56, pp. 268-74.HEPI Higher Education Policy Institute. 2008. The Bologna Process and UK’s InternationalStudent Market, http://www.hepi.ac.uk/downloads/36Bolognaprocessfull.pdf, last access2008-06-15.HESA. 2008a. available at: http://www.hesa.ac.uk/.Higher Education Authority. 2007. Domiciliary of origin of full-time students 2006/07,HEA Statistics Section, available at:Higher Education Authority. 2008. Statistics Section, Domiciliary of origin of full-timestudents 1996/97-2006/07, communication via e-mail 2008-06-27.http://skylla.wz-berlin.de/fulltext/wzbmit_artikel/2005/f-12236.pdf last access 2008-06-14.http://www.daad.de/de/download/hochschulen/veranstaltungen/aaa_2005/ag1_protokoll.pdf, last access 2008-06-15.http://www.dcsf.gov.uk/londonbologna/uploads/documents/Socialdimensionandmobilityreport.pdf, last access 2008-06-14.CMR 53

http://www.hea.ie/files/files/file/statistics/2007/FTDomiciliaryofOrigin07.xls last access2008-06-27.http://www.international.ac.uk/resources/International%20Focus%20Issue%2016.01.08%20final.pdf last access 2008-06-14.http://www.uis.unesco.org/template/pdf/ged/2007/EN_web2.pdf last access 2008-06-14.Inzelt, Annamaria. 2003. “Hungary”. In: Brain-drain emigration flows for qualifiedscientists part 5. Flows and non-EU Europe, Maastricht: University of Maastricht,MERIT.Lowell and Findlay, 2002.Kaczmarczyk, Paweł, and Okólski Marek. 2002. From net emigration to net immigration.Socio-economic aspects of international population movements in Poland, in: Rotte, R.and P. Stein (eds.). Migration Policy and the Economy: International Experiences,München: ars et unitas.Kaczmarczyk, Paweł, and Okólski Marek. 2005. Migracje specjalistow wysokiej klasy wkontekscie czlonkostwa Polski w Unii Europejskiej [Migration of the highly skilled in thecontext of Poland’s accession to EU], UKIE, WarsawKaczmarczyk, Paweł, and Okólski Marek. 2008. “Demographic and economic impacts ofmigration on Poland”, in Oxford Review of Economic Policy, Autumn (in press).Kaczmarczyk, Paweł. 2008. Brains on the move? Recent migration of the highly skilledfrom Poland and its consequences, unpublished manuscript.Kępińska, Ewa and M. Okólski, Marek. 2002. „Recent trends in International migration.The 2002 SOPEMI report for Poland”, CMR Working Papers, 48.Kępińska, Ewa. 2006. “Recent trends in International migration. The 2005 SOPEMI reportfor Poland”, CMR Working Papers, 60.Kępińska, Ewa. 2007. “Recent trends in International migration. The 2007 SOPEMI reportfor Poland”, CMR Working Papers, 29/87.Kuptsch, Christiane. 2006. Students and talent flow – the case of Europe: From castle toharbour?, p.33-60 in: Kuptsch, C./Fong, P.E. (eds.): Competing for global talent,International Institute for Labour Studies, Geneva 2006.Lowell, B. Lindsay and Allan Findlay. 2002. Migration of Highly Skilled Persons fromDeveloping Countries: Impact and Policy Responses, A Project Report for theInternational Labor Office and the United Kingdom's Department for InternationalDevelopment.Marcinkowska, Izabela, Ruzik, Anna, Strawiński, Paweł, and Walewski, Mateusz. 2008.Badanie struktury i zmian rozkładu wynagrodzeń w Polsce w latach 2000-2006.Warszawa: MPiPS.CMR 54

Mattoo, Aaditya, Ozden, Caglar, and Neagu Cristina. 2005. “Brain Waste? EducatedImmigrants in the U.S. Labour Market”, World Bank Policy Research Working Paper No.3581.McKenzie, David, and Rapoport, Hillel. 2008. “Self-selection patterns in Mexico-U.S.migration: The role of migration networks”, Journal of Development Economics,forthcoming.Mechtenberg, Lydia. 2005. Wettbewerb um kluge Köpfe. Die Länder Europas im Bologna-Prozess, in: WZB-Mitteilungen, Heft 107, March 2005, 4 p, available at:Mountford, Andrew. 1997, “Can a brain drain be good for growth in the sourceeconomy?”, in Journal of Development Economics, Vol. 53, No. 2, pp. 287-303.OECD. 2002. International Mobility of the Highly Skilled, Paris: OECD.OECD. 2005. Trends in International Migration: SOPEMI 2004, Paris: OECDOECD. 2006. International Migration Outlook 2006. OECD: Paris.OECD. 2007. Education at a Glance OECD Indicators, 41 p., available at:http://www.oecd.org/dataoecd/36/4/40701218.pdf last access 2008-06-14.Okólski Marek. 1997. Statystyka imigracji w Polsce. Warunki poprawności, ocena stanuobecnego, propozycje nowych rozwiązań, ISS Working Papers – Seria: Prace Migracyjne(currently CMR Working Papers), No. 2.Okólski Marek. 2006. “Costs and benefits of migration for Central European countries”,CMR Working Paper 7(65).Pollard, Naomi, Latorre, Maria, and Jaideep Sriskanradajah. 2008. Floodgates orturnstiles? Post-EU enlargement migration flows to (and from) the UK. London: IPPR.Radu, Dragos. 2003. “Human capital content and selectivity of Romanian migration”,Working Paper No. 365, Luxembourg Income Study.Sakson, Barbara. 2002. Wpływ „niewidzialnych” migracji zagranicznych latosiemdziesiątych na struktury demograficzne Polski. Warszawa: Szkoła GłównaHandlowa.Salt, John, and Findlay, Andrew. 1989. “International migration of highly-skilledmanpower: theoretical and developmental issues”, in R. Appleyard (ed.), “The Impact ofInternational Migration on Developing Countries”. Paris: Organisation for Economic Cooperationand Development.Salt, John. 1997. “International movements of the highly skilled”. OECD OccasionalPapers 3.Salt, John. 2001. “Current Trends in International Migration in Europe”. CDMG, 33.CMR 55

Stark, Oded, and Wang, Yong. 2002. “Inducing Human Capital Formation: Migration as aSubstitute for Subsidies”, in Journal of Public Economics, Vol. 86, No. 1, pp. 29-46.Stark, Oded, Helmenstein Christian, and Prskawetz, Anna. 1997. “A brain gain with abrain drain“, in Economics Letters, Vol. 55, pp. 227-34.Stark, Oded. 1991. Migration of Labour. Oxford: Basil Blackwell.Stark, Oded. 2004. “Rethinking the Brain Drain. World Development”, Volume 32, 1, pp.15-22Statistisches Bundesamt [Federal Statistical Office]. 2007. Bildung und Kultur.Studierende an Hochschulen, Wintersemester 2006/2007, in: Fachserie 11, Reihe 4.1,491 p. Statistisches Bundesamt Wiesbaden 2007.Straubhaar Thomas, and Wolburg, Martin. 1999. “Brain Drain and Brain Gain in Europe:an evaluation of the East-European Migration to Germany”. Jahrbücher fürNationalökonomie und Statistik, vol 218, N 5-6.Suter, Brigitte, and Jandl, Michael. 2006. Comparative Study on Policies towards ForeignGraduates – Study on Admission and Retention Policies towards Foreign Students inIndustrialised Countries, ICMPD Vienna, October 2006, available at:http://www.imiscoe.org/news/newsletters/documents/ComparatovestudyICMPD.pdf, lastaccess: 2008-06-01Teichler, Ulrich. 2007. The Changing Role of Student Mobility, 10 p., available at:http://portal.unesco.org/education/en/files/53904/11858759925Changing_role_student_mobility.pdf/Changing_role_student_mobility.pdf last access 2008-06-14.UNESCO. 2006. Global Education Digest 2006. Comparing Education Statistics Across theWorld, available at: http://www.uis.unesco.org/TEMPLATE/pdf/ged/2006/GED2006.pdflast access 2008-06-14.UNESCO. 2007. Global Education Digest 2007. Comparing Education Statistics Across theWorld, available at:Universities UK International Unit. 2008: International Focus 16/01/08, Making it easierfor international students to stay, available at:Wolburg, Martin and Achim Wolter. 1997. “Mobility within and into European Union:Some Stylized Facts”. Discussion Paper of the Europa-Kolleg Hamburg.Wolburg, Martin. 1996. “On Brain Drain, Brain Exchange, Division of Labour andeconomic growth within Common Market”. University of the federal Armed forces,Hamburg, Discussion Papers in Economic Policy, 66.World Bank. 2007. From Shortage of Jobs to Shortage of Skills. Quarterly EconomicReport Part II: Special Topic. Washington: World Bank.CMR 56

European Integration ConsortiumIAB, CMR, fRDB, GEP, WIFO, wiiwLabour mobility within the EU in the context of enlargement and the functioningof the transitional arrangementsVC/2007/0293Deliverable 7Austrian Institute of Economic Research (WIFO)Regional effects of labour mobilityPeter Huber and Klaus NowotnyAbstractThis deliverable analyses the regional impact and distribution of migration and crossbordercommuting in the EU27 using European Labour Force data. Furthermore a casestudy of migration and commuting potentials in one of the border regions, which can bedeemed to be most affected from these flows (the border region of the new memberstates to Austria) is conducted by using the first two waves of the LAMO householdsurvey conducted in the CENTROPE region in 2004-2005 and 2006-2007.With respect to the regional structure of migration in the EU we find the largest localclusters of migrants in the EU 15 in the Île de France as well as Inner and Outer Londonand a markedly different settlement structure of migrants relative to natives: 23.9% ofall migrants would have to change their region of residence in order to achieve a uniformdistribution of migrants across EU-15 NUTS 2 regions. Migrants from the NMS-8 show alower degree of concentration than those from Bulgaria and Romania or the candidatecountries, while they are more regionally concentrated than migrants from othercountries. The biggest local clusters of NMS migrants can be observed in the Londonareas and Vienna. Looking at individual sending countries, Polish migrants show thelowest tendency to cluster regionally among migrants from the NMS. Furthermore, lowskilled migrants with primary education are much more spatially concentrated thanmigrants with secondary or tertiary education, which confirms earlier findings.The concentration of migrants did not differ substantially between migration cohorts:those who moved during the last 10 years are about as concentrated as those whomigrated earlier. However, the target regions of more recent migration waves areconsiderably different from those of earlier cohorts. This applies in particular to migrantsfrom the NMS-8, where the different institutional regimes since accession have shiftedthe target country structure, which also affects the regional patterns of migration.Although the geographical concentration increased for more recent cohorts of migrantsfrom the NMS-8, the correlation of local concentrations across time is rather low and eveninsignificant for some CEE countries. However, a regression analysis shows that—even

after controlling for geographic and economic characteristics of the regions—ethnicnetworks do play a significant role in explaining the location choice of migrantsCross-border commuting in the EU 27 in general is limited to individual border regionsand has a relatively low magnitude when considering the overall European labour market.In the two years observed cross-border commuters accounted for only 0.5% of totalemployment in the EU. Cross-border commuting is of relevance in a small number ofborder regions, only. These are mostly located at the German-French and French-Belgianborders, on the Austro-German border, at the Czech-Slovak border, in the Balticcountries and in Western Hungary as well as the German-Polish border and potentiallysouthern Sweden. These borders are mostly characterised by strong linguistic, historic orinstitutional ties. In these border regions usually slightly more than 1% of the employedcommute across borders and in individual cases cross-border commuting may surpassthe 5% mark. For most border regions outside these "hot spots” out-commuting is below0.5% of the employed.There are also some differences in the importance of cross-border commuting betweenthe EU 15 and NMS 12. In particular, NMS 12 countries receive much fewer cross-bordercommuters than EU 15 countries as a percentage of the employed in the country of work.In addition outbound cross-border commuting from the NMS 12 is strongly orientedtowards the EU 15 countries rather than non-EU countries. By contrast, outbound crossbordercommuting in the EU 15 is more strongly oriented to non-EU countries rather thanto the NMS 12.Cross-border commuters - in contrast to internal commuters in the EU 27 - are also notnecessarily better qualified than non-commuters and are drawn more thanproportionately from manufacturing workers, males and the age group of the 20 to 29year olds. These characteristics apply even more strongly to cross-border commutersfrom the NMS 12 than to commuters from the EU 15. While these results are largelyconsistent with the findings of earlier case studies in the literature, they also suggest thatcross border commuters – in contrast to migrants – are not as strongly positively selectedon educational criteria, but stem primarily from the intermediate qualification level.Finally, - while our results in this respect are subject to a rather unsatisfactory datasituation, our findings also imply that after controlling for other influences on crossbordercommuting - flows from the NMS 12 to the EU 15 are not significantly smallerthan those among the EU 15 countries, while flows from the EU 15 to the NMS 12 aresignificantly lower than those among the EU 15. The primary difference in the factorsdetermining cross-border migration in the NMS 12 and the EU 15 seems to be a closerassociation of cross-border commuting with the industrial specialisation in the NMS 12than the EU 15.In our case study of the CENTROPE region we analyse migration and commutingpotentials in the border regions of the new member states to Austria using two waves ofa household survey conducted in the Austrian-Czech-Slovak-Hungarian border region in2004-2005 and 2006-2007. 10.9% of the interviewed in the CENTROPE regions of theCzech Republic, Hungary or Slovakia expressed the wish to migrate to one of the EU 15in the future (and thus belonged to the general migration potential). 3.8% of thepopulation in the region were willing to migrate and had taken first steps to prepare forcross border migration or commuting. They belonged to the expected migration potential.1.3% of the population applied for a work permit and or already had a job offer abroad(real migration potential) in 2006-2007. An additional 5.6% of the population in theregion under consideration expressed the wish to commute to the EU 15 in the future

(and belonged to the general commuting potential). 1.4% of the population in the regionwere willing to commute had taken first preparatory step. 0.8% of the population hadapplied for a work permit and/or already had a job abroad (real commuting potential) in2006-2007.Relative to the first wave of interviews in 2004-2006 this represents a decrease in themigration potential of between 1.5 percentage points (general migration potentials) and0.1 percentage points (real migration potential) of the population. Commuting potentialsdeclined more strongly for the general and expected commuting potentials, while the realcommuting potential increased slightly. A comparison with the Austrian subregionssuggests that the general migration potential in Austria is as high as in the NMS-regions.Analysing the determinants and structure of potential commuters and migrants suggeststhat, the presence of kids or a spouse in the household is a more serious impediment forthe willingness to migrate than for the willingness to commute; gender differences in thewillingness to commute are larger than for the willingness to migrate (although womenare both significantly less willing to commute and to migrate), and the willingness tomigrate reduces much more strongly with age than does the willingness to commute.Also both those willing to commute as well as those willing to migrate aredisproportionately often drawn from the two extremes of the educational distribution, andare thus often either highly or less educated. When, however, including education in amultivariate regression analysis we find that education has no significant effect on boththe willingness to migrate and to commute. This implies that potential migrants as wellas potential commuters in the region considered are neither positively nor negativelyselected.The willingness to commute also decreases much more rapidly with distance to thenearest potential workplace than the willingness to migrate while the latter is positivelyinfluenced by English and other foreign language knowledge. The willingness to commuteis, however, more strongly associated with German language knowledge. In addition, thewillingness to migrate is also more strongly influenced by the presence of networks andprevious experience of working abroad than the willingness to commute.Analysing the changes in the preferences associated with the willingness to migrate andcommute, our data suggests that the proportion of those willing to migrate to Germanyand Austria is about 40% and thus substantially lower than in previous studies. On theother hand, the share of potential migrants preferring the United Kingdom is substantiallyhigher than in the earlier literature. Those who prefer Austria do so mainly because of itsgeographical proximity and its high wage level. All other motives, such as language skills,resident family members, relatives or friends, education or training opportunities as wellas the relative easiness of obtaining a residence or work permit speak for the UK.The views and opinions expressed in this publication are those of the authors and do not necessarilyrepresent those of the European Commission.

Contents1 Introduction ...................................................................................................... 12 Regional Concentration of Migrants in Europe ........................................................ 52.1 Introduction .............................................................................................. 52.2 Why do migrants concentrate in specific areas?.............................................. 52.2.1 Migration networks......................................................................... 62.2.2 Herd behaviour .............................................................................. 92.3 Measuring the regional concentration of migrants........................................... 92.4 The regional concentration of migrants in Europe ......................................... 112.4.1 Regional concentration: facts and figures ........................................ 112.4.2 Does regional concentration differ by individual characteristics?......... 132.4.3 Ethnic migration clusters............................................................... 162.4.4 An analysis by receiving country .................................................... 212.4.5 Is there evidence of network or herd migration?............................... 222.5 Summary ................................................................................................ 283 Cross-border Commuting in the EU27................................................................. 313.1 Introduction ............................................................................................ 313.2 Data and Extent of Cross-Border Commuting in the EU 27............................. 323.2.1 Data Issues ................................................................................. 323.2.2 Some stylized facts on the extent of commuting in the EU 27from the sending region perspective ............................................... 343.2.3 Stylized facts from the receiving region and place to placeperspective ................................................................................. 393.3 The Structure of Commuting Flows ............................................................. 413.3.1 Comparing cross-border commuters, non commuters andinternal commuters ...................................................................... 413.3.2 Differences between NMS 12 and EU 15 Flows ................................. 423.4 The Determinants of Out-Commuting Flows................................................. 443.5 Conclusions ............................................................................................. 484 The CENTROPE Region: Economic Background..................................................... 504.1 Economic development of CENTROPE .......................................................... 534.1.1 GDP and GDP per capita................................................................ 534.1.2 Specialisation and sectoral structure............................................... 554.1.3 Education, R&D and high technology resources ................................ 574.2 Cross border flows.................................................................................... 584.2.1 Cross–border enterprise co-operation ............................................. 584.2.2 Cross–border labour market mobility .............................................. 594.3 Labour Market Development of CENTROPE................................................... 594.3.1 The structure of employment and unemployment rates in theNUTS 2 regions of CENTROPE ........................................................ 594.3.2. Development of unemployment and employment rates ..................... 65

5 Cross-Border Migration and Commuting Potentials in the CENTROPE Region............ 685.1 Migration and commuting potentials ........................................................... 705.2 Mobility, migration and commuting potentials towards Austria ....................... 756 Determinants and Structure of Potential Migration and Commuting in theCENTROPE region............................................................................................. 776.1 Theoretical aspects................................................................................... 776.2 Explanatory Variables ............................................................................... 836.2.1 Individual level variables to capture income differentials ................... 846.2.2 Individual level variables affecting migrations costs.......................... 886.3 Estimation results..................................................................................... 917 Motives, expectations and preferences of potential migrants and commuters inthe CENTROPE Region ...................................................................................... 997.1 Mobility motives....................................................................................... 997.2 Choice of country and region of work .........................................................1027.2.1 Country preferences ....................................................................1027.2.2 Regional preferences ...................................................................1057.3 Length and timing of migration .................................................................1075.4 Expectations concerning type of work ........................................................1118 Conclusions....................................................................................................1159 References .....................................................................................................120

1 IntroductionSpatial labour mobility (i.e. commuting and migration) by definition always involves achange in place of work. Thus the analysis of cross-border labour mobility also alwaysinvolves a regional component. This applies both to the choice of the region of residenceas well as to the method in by which spatial labour mobility is achieved. These regionalissues have also received increasing attention in the recent literature. For instance withrespect to the choice of method of mobility recent literature (e.g. Zax, 1994, Rowendal,1998, Van Ommeren, Rietveld and Nijkamp, 2000) has repeatedly stressed that spatiallabour mobility can be achieved either by migration or by commuting. As pointed out bythis literature these two processes are closely related: a person residing in a region andreceiving an job offer in another region can either choose to work in this other regionwithout changing place of residence, in which case she will become a commuter, or shemay choose to move both her place of residence as well as place of work. In this case shewill become a migrant. Furthermore, studies (see Renkow and Hoover, 2000, Clark andWhithers, 1999, Rowendahl, 1999, Van Ommeren, Rietveld and Nijkamp, 1999, Elliason,Lindgren and Westerlund, 2003) which consider this relationship empirically and primarilyconcentrate on migration and commuting choices within a specified metropolitan area,find a strong relationship between migration and commuting.Similarly, with respect to migration, the high regional concentration of migrants in certaingeographic locations is one of the most robust stylised facts repeatedly stressed in theeconomic literature on migration. In her seminal paper Bartel (1989) shows that in theUS close to 75% of the migrants live in the 25 largest SMAS of the United States,although only 50% of the native population resides in these regions. Similar stylised factshave been found to apply in the few studies that have conducted similar analyses forcountries other than the US (e.g. Huber, 2002 for Austria, Chiswick, Lee and Miller, 2002for Australia, Edin, Fredrikson and Aslund, 2001 for Sweden, Hou, 2005 for Canada, andBlom 1999, for the Oslo metropolitan area).Furthermore Bartel (1989) and the related literature show a number of further stylisedfacts. In particular this literature suggests that:1. Education plays an important role in the location choice of migrants, with the degreeof geographic concentration decreasing with increasing education of migrants.2. There are important differences in the geographic concentration of migrant groupsdepending on their ethnicity, where in general the concentration decreases withincreasing duration of residence and increasing average educational attainment butincreases with the degree to which the tastes of the migrant group underconsideration differ from those of the native population.3. There are also important differences as to the geographic distribution of migrantsaccording to ethnicity, with certain migrant groups locating in different parts of thecountry. In particular with respect to migrant groups from nearby sending countriesWIFO 1

proximity to the home country seems to play an important role in determining theregion of residence.4. Migrants are also more mobile than natives within the country of residence, where ingeneral migrants tend to move out of their ethnic enclaves as their stay in the hostcountry prolongs (see also Rephann and Vencatasawmy, 2000, Sündeln, 2007).By contrast the literature on commuting suggests that commuting flows may differ frommigration flows in a number of important ways. For instance commuting flows are muchmore dependent on distance between sending and receiving regions than migration flows.Since this is also to be expected from cross-border commuting flows, this implies aregionally asymmetric impact of cross-border commuting, which, - in contrast tomigration flows, which as shown in the last chapter are often concentrated in the urbancentres of a country – on account of their high distance dependence should be expectedto be concentrated in border regions.There may, however, also be more subtle differences between migration and cross-bordercommuting flows. In this respect for instance White (1986) as well as Rouwendahl (1999)show that commuting within a country is strongly focused on males. This can beexplained by the higher alternative costs of travelling time for women, which arise onaccount of their role in childcare and household production, as well as a higher share ofpart time workers among women, which leads to higher commuting costs per work hour.Furthermore commuters within a country may differ from non-commuters with respect toage and education. Rouwendahl (1999) find that the willingness for mobility decreaseswith age, and Van Ommeren (1999), Hazans (2003) as well as Rouwendahl (1999) allfind that higher educated workers are more likely to be commuters within a country thanless educated workers.One could expect that some of these "stylized facts” carry over to cross-bordercommuters while others may differ. Indeed some of the recent case studies (see: Buch etal, 2008 and Gottholmseder and Theurl 2006, 2007) confirm that cross bordercommuters are mostly male but also suggest that cross-border commuters differ fromcommuters within a country both with respect to education and age structure. Comparingcross-border commuters from Vorarlberg to Switzerland to internal commuters and noncommutersin the same region Gottholmseder and Theurl 2006 find that on account of alow commuting share among the under 25 year olds there is no clear evidence that crossbordercommuters are younger than non commuters, and according to the regressionresults in Gottholmseder and Theurl (2007) neither age nor education is a significantdeterminant of cross-border commuting. This thus suggests that with respect to age andeducation the process of selection of cross-border commuters differs from that of internalcommuters.In the context of European integration these regional issues of labour mobility arebecoming increasingly relevant both from an analytical as well as from a policyperspective. This is because with respect to cross-border commuting and migration andtheir interrelationship recent results suggest that cross-border labour markets may beWIFO 2

emerging in the European Union. For instance Overmann and Puga (2000) find thatregional linkages in unemployment rates are equally strong across national borders aswithin countries. Furthermore, issues of the interrelationship between cross-bordercommuting and migration have also received heightened attention by policy makers inthe context of the debate on enlargement of the EU by the 10 new member states whichjoined the EU in 2004. In this debate in particular Austrian and German policy makersrepeatedly argued that due to the vicinity to of major centres to the external border ofthe EU, cross-border commuting flows may be an additional impact on migration andshould be considered in a debate on potential derogation periods (see Huber 2001,Untiedt and Alecke 2001).Furthermore, the regional concentration of migrants raises issues as to whether migrationhas a differential impact on different regional economies (see Card and Lewis, 2005 for arecent contribution), whether the potential formation of enclaves has a negative orpositive effect on the probability of integration of foreigners (see Borjas, 1995, Betrand,Luttmer and Mullainathan, 2000, Cutler and Glaeser, 1997, Chiswick and Miller, 2002,Cardak and McDonald, 2004 for contributions analysing the implications of enclaves onthe economic and social success of migrants) and what policy activities could help thoseregional labour markets most strongly affected by migration and commuting in thedouble task of integrating the new arrivals and adjusting to the increase in labour supply(see Edin, Fredrikson and Aslund, 2001a for a contribution to this literature focusing on aEuropean country).In this report on deliverable 7 to the study “Labour Mobility within the EU in the contextof enlargement and the functioning of the transitional arrangements” our primary aim isto describe the regional concentration of migrants in Europe and to analyse both theregional and the educational structure of cross-border commuters, with particularemphasis on labour mobility from the new member states and candidate countries. Weaddress this issue by first focusing on a descriptive analysis of European Labour Forcesurvey data and second conducting a case study of the Austrian-Czech-Hungarian-Slovakborder regions in which special attention is paid to potential commuting and migrationchoices.With respect to the analysis of data from the European Labour Force Suvey in the firsttwo chapter of this deliverable we assess the regional impact of migration andcommuting at the level of NUTS2 regions. This part of the study thus gives answer to thequestions of where foreign born citizen (resp. citizen of foreign nationality) in the EU liveand how the distribution of foreign born individuals (resp. citizen of foreign nationality)evolved over the last decades. To this end in chapter 2 of this report we use LFS data onthe nationality and country of birth to estimate the number of foreign born (resp foreignnationals) residing in the EU NUTS 2 regions, for those regions where the data is reliable(i.e. representative) and analyse the regional settlement structure of foreign born(foreign nationals) by country group (i.e. other EU15 countries, New EU member States,other countries) and by educational characteristics. Finally we also follow the literature byWIFO 3

estimating a single equation model to determine which factors determine the locationchoices of migrants.In chapter 3, by contrast, we use data from the European Labour Force Survey to analysethe extent of cross-border commuting in the EU. In contrast to previous literature (see forexample: Buch et al, 2008, Gottholmseder and Theurl , 2006, 2007, van der Velde,Jansen and van Houtum, 2005, Greve and Rydbjerg, 2003a, 2003b, Bernotat andSnickars, 2002) which mostly focused on case studies for individual regions, we focus onthe complete EU 27. Given the paucity of empirical results on the extent and structure ofcross-border commuting for the EU 27, our aims are primarily descriptive: In particularwe first of all want to know how many people can be assumed to commute across bordersin the EU 27 currently and how their demographic structure differs from that of bothcommuters within a country and persons, who both live and work within the same region(i.e. non-commuters). Second of all we want to know in which regions and countries ofthe EU 27 cross-border commuting currently plays an important role and thirdly – withrespect to labour mobility from the 12 new member states (NMS 12) to the 15 oldmember states (EU 15), - we want to know how both the structure and extent of currentcross-border commuting from the NMS12 (which are still influenced by the transitionalperiods applied in a number of EU countries) differs from cross-border commuting flowsin the unregulated regime of the EU 15.In chapters 4 to 7 of this study we then conduct a case study of the CENTROPE region(see: Palme and Feldkircher, 2006, Huber and Mayerhofer, 2006) as an example of aregion that may be particularly strongly affected by commuting and migration. Thisregion encompasses Czech, Slovak and Hungarian borders and the metropolitan areas ofVienna and Bratislava as well as the Eastern regions of Austria. It is probably one of mostaffected by migration and commuting and thus provides a unique area for studying theeffects of labour mobility associated with enlargement. This case study evolves aroundthree questions:• First, we analyse the development of cross-border commuting and migrationpotentials in the border region surrounding Austria,• Second, we analyze how the structure of those willing to migrate or commute differsfrom stayers and whether it has changed over time.• Third, we want to assess whether cross border migrants, commuters and stayersdiffer with respect to their motives for becoming mobile or not.Thus, after a short description of the economic background of the CENTROPE region inchapter 4, chapter 5 presents a description of the data and the cross-border commutingand migration potentials in the region. Chapter 6 investigates the determinants ofmobility as well as the choice between migration and commuting and Chapter 7 presentsdata about the motives of prospective migrants and commuters. Finally, chapter 8 drawssome conclusions for the study.WIFO 4

2 Regional Concentration of Migrants in Europe2.1 IntroductionA common characteristic of migration movements is that migrants tend to cluster inspecific regions of the host countries. Some of these concentrations can be explained bythe fact that a region serves as a “port of entry” (both literally as well as figuratively) orby favourable labour market conditions. But looking closer, it can be observed that manymigrant clusters consist predominantly of individuals with the same ethnic background.This indicates that immigrants tend to settle where other migrants from the same sourcecountry have gone, leading to regional ethnic concentrations of migrants.This concentration not only has important implications for regional housing and labourmarkets, but also increases the risk of emerging parallel societies and affects localgovernments: depending on social security regulations, large clusters of migrants canlead to an increased burden for local (and national) welfare institutions if they areassociated with a higher welfare participation among its members.Against this background in this chapter we analyse the concentration of migrants in theEU-15 using recent data from the European Labour Force Survey to shed light onquestions such as: where can we observe the largest concentrations of migrants inEurope? Do more recent migration waves differ in their concentration from previousmigrants? Does the concentration in specific regions differ by age cohorts? Are moreeducated migrant groups less concentrated in specific areas while low-skilled workers relyon networks? Does the concentration of migrants differ by country of origin? Whichtrends in regional preferences of migrants can be observed in the EU-15, and is thereevidence of network or herd migration?2.2 Why do migrants concentrate in specific areas?Several hypotheses have been developed in the economic literature on the topic ofmigrant’s locational choice within the receiving country. Apart from some regions being“natural hubs” for migrants—e.g., cities which act as “ports of entry” because ofinfrastructure endowments (like sea- or airports) or administrative institutions (likecentral immigration offices)—, regional concentrations can e.g. also arise in high-wageareas with favourable labour market conditions. 1This can however not fully account for the observation that migrants tend to settle whereother migrants from the same country of origin migrated before, resulting in a geographicconcentration of migrants with similar ethnicity in specific locations. Since a seminal1 This applies both to labour market conditions in the “official” labour market as well as in the informal sector(Amuendo-Dorantes and de la Rica 2005).WIFO 5

study on ethnic migrant concentration in the U.S. by Bartel (1989), several hypotheseshave been developed to explain this phenomenon.2.2.1 Migration networksOne of the most frequently cited theories is that clustering allows the formation ofmigrant networks which produce externalities for members of the same ethnic groupbecause the costs of migration decreases with the number of previous migrants. Thisleads to “self-perpetuating” migration (Massy et al. 1993; Carrington, Detragiache, andVishwanath 1996) from a specific source country. Above reducing migration costs,networks can also provide help with the settlement process or decrease the perceivedalienation in the host country (Bauer, Epstein and Gang 2000).Furthermore, networks can provide their members with ethnic goods like food, clothing,social organisations, religious services, media (like radio, newspapers, etc.) or marriagemarkets (Chiswick and Miller, 2005). The provision of ethnic goods can be expected toincrease with the stock of migrants with similar ethnic background, creating incentivesfor other immigrants to settle in these regions where they can enjoy a larger supply ofethnic goods. The concentration of migrants will be more pronounced, the higher theshare of ethnic goods in the migrants’ consumption basket and the more dissimilar thesource and target cultures. If there are economies of scale in the production of ethnicgoods (as can be expected, e.g., for religious services or media), the lower price of ethnicgoods in regions with large ethnic concentrations reduces the costs of living there, whichis an incentive for immigrants to move into this region even if they could earn a higherwage somewhere else (Chiswick and Miller, 2005). 2Regions with large networks can also be attractive because they increase the labourmarket prospects of new migrants: they can benefit from a better availability ofinformation in the network which increases labour market opportunities (Gross andSchmitt 2003). E.g., Edin, Fredriksson and Åslund (2001) found a statistically significantpositive effect of ethnic concentration on migrant earnings. Other studies, however,showed that clustering negatively influences the economic success of migrants (Bartel1989, p. 388). One explanation for this is that migrant concentration is negativelycorrelated with foreign language fluency (Lazear 1999), which is in turn a prerequisite forintegrating into the host countries labour market. 32 While networks will thus have a positive impact on imports (e.g., of ethnic goods) from the source to the hostcountry, they can have a positive influence on exports to the source country of migrants as well(Bandyopadhyay, Coughlin and Wall 2008).3 Migration also affects the local labour market for natives in the host country, Whether migration is welfareenhancingor welfare-decreasing depends, however, on the relative skill composition of migrants comparedto natives, which determines the substitutability of native vs. Migrant labour (Bartel 1989, Borjas 1999).WIFO 6

However, if employers with migration background prefer to employ other migrants (ofsimilar ethnic origin) instead of natives, a separate migrant labour market can emerge. 4Gross and Schmitt (2003) show that a small and homogeneous market for migrant labourcan even sustain a higher wage than the larger anonymous “general” labour market. Forsuch a migrant labour market to be sustainable, the ethnic community must neither betoo small nor too large.Network size is thus crucial not only for new arrivals, but also for previous migrantsalready living in the region (Heitmueller 2006). As the concentration of migrantsincreases, there can, however, also be negative effects on previous migrants’ utility:continuing migration reduces the income differentials between sending and receivingcountries and the wages of the previous migrant cohorts. A similar effect will arise ifhousing prices increase following an influx of migrants into a region. This negative effectof decreasing wages and/or increasing housing prices will at some point dominate thepositive network externality effect, leading to a decline in the attractiveness of a formerlypopular migrant cluster (Portnov 1999). There will thus be an optimal size of the regionalnetwork beyond which every new migrant decreases the utility of previous migrantsalready living in this region. 5Local networks can, however, still grow beyond this optimal size (from the point of viewof migrants already living in the region), if the region still provides the maximum utilitycompared to all other available regions, even if new migrants take into account that theirutility will decrease with every other migrant that follows (Bauer, Epstein and Gang2002). Even if migrants already living in the region could theoretically provide no morepositive network effects (e.g., by withholding information or refusing to help with job orresidence search) it has been shown by Heitmueller (2006) that, in the absence ofcoordination and a collective sanctioning mechanism, there is an incentive to increase thenetwork beyond the optimum. This can occur because the utility gain arising from anincreased personal network (e.g., family and friends) exceeds the utility decline from thepotential wage loss arising from one additional migrant.Networks can also affect the selection of migrants by skills. Miranda (2007) analysed theeffect of migrant networks on educational achievement of family and friends in the sourcecountries. He concluded that migrant networks have a positive effect on the education offamily members at home if remittances are used to enhance the educational achievementof family members (provided the qualifications are portable). However, as networksfacilitate family chain migration, they can also generate incentives to drop out ofeducation at an early stage, especially if education is non-portable. Thus, remittances canlead to a negative selection of migrants, inducing the migration of low-skilled followers.4 E.g., self-employed migrants tend to prefer hiring other migrants with the same ethnic background(Andersson and Wadensjö 2007). The same result can arise if natives are reluctant to work in firms led bymigrants.5 If prospective emigrants take this into consideration when deciding where to migrate to, an inversely U-shaped effect of network size on the probability of moving to a specific region can arise (Bauer, Epstein andGang 2002).WIFO 7

This is consistent with the proposition often found in the literature that the “pioneers”(i.e., those among the first wave of migrants) are “likely to be the most able” (Lazear1999, p. 118), e.g. because they can expect the highest returns from migration and willthus find it easier to cover migration costs. Low-skilled followers on the other hand willfind it worthwhile to wait until the network has grown and migration costs have fallen.Furthermore, as Stark (1994) has shown, under asymmetric information—i.e., whenemployers have no information on the true skill (or effort) of applicants—low-skill (orlow-effort) workers might choose to relocate to regions where a considerable stock ofhigh-skilled migrants settled before. By doing so, they can mingle with these high-skilled(or high-effort) migrants to obscure their skill signals to employers. E.g., if employersobserve that migrants with a specific ethnic background have good skills and/or showhigh work efforts, they might be predisposed to hire other workers with the same ethnicbackground. If employers cannot observe skill or effort beforehand, low-skilled followerscan use this predisposition to earn higher wages, leading to a negative selection ofmigrants by skills. Eventually, if the proportion of low-skilled migrants becomes too large,this predisposition might cease or reverse, and followers are no longer able to exploitemployers’ asymmetric information. High-skilled followers will then no longer find itprofitable to move to this region, and high-skilled migrants already living there willconsider relocating to other areas where their skills are not obscured by low-skillmigrants. This is consistent with the observation that high-skilled migrants are moredispersed and less concentrated in specific areas, while concentration was found to behighest among low-skilled workers (Bartel 1989).A negative selection can also occur because networks attract migrants with poor nativelanguage skills: if the network is large enough, knowledge of the host language might nolonger be necessary, because all transactions can be carried out within the network. Thisdecreases the opportunity costs of not learning the host language. Large ethnicconcentrations can thus act as “language traps”, sustaining the migrants’ poor languageabilities (Bauer, Epstein and Gang 2005) which can have a negative effect on earnings.This separation between migrants and natives tends to increase with the cultural distancebetween the immigration group and the native population (Blom 1999). Furthermore,networks can aggravate negative selection if they facilitate access to welfare provisions.Bertrand, Luttmer and Mullainathan (2000) have shown that a larger network increasesthe probability of welfare participation for individuals from high welfare language groups,and that social networks strongly influence welfare participation. Regions with highconcentrations of migrants can thus face an increased burden in terms of social securityprovisions as well as a higher demand for public goods (Bartel 1989, p. 390). This appliesespecially to regions with more generous welfare provisions. 66 Furthermore, Lazear (1999) shows that government transfers can reduce the incentives to assimilate, thuscounteracting integration efforts by governments.WIFO 8

2.2.2 Herd behaviourHerd behaviour can constitute another explanation for the clustering of migrants inspecific regions. Herd behaviour can occur if there is imperfect information as to whichamong alternative target locations provides the highest utility. If a potential migrantobserves only the outcome of previous migrants’ destination choices, but not the “signal”that determined their choice, she might discount her private information aboutalternative target regions and follow the flow of previous migrants (Epstein 2002; Bauer,Epstein and Gang 2005) in the belief that they must have had information which is notavailable to her. E.g., an individual might migrate to a specific city simply because shehas observed other migrants from her country doing so, even though she would havemigrated to another region based on her private information.Herd behaviour can lead to inefficiencies if previous migrants also discounted theirprivate information in favour of the belief that those who went there before them hadinformation they do not have, while they could have gained a higher utility by followingtheir private information. Herd behaviour and network effects are—although conceptuallydifferent—not mutually exclusive: both effects can exist simultaneously and determinethe location decisions of migrants. The presence of network externalities in this contextcan even increase the probability that herd behaviour will be observed (Epstein 2002).Herd behaviour, on the other hand, can lead to a steady inflow of new migrants even ifthe negative wage or housing price effects already dominates the network externalityeffect in the target country.2.3 Measuring the regional concentration of migrantsFollowing Bartel (1989) we measure the regional concentration of migrants by thecoefficient of geographic association. Individuals are considered migrants if they wereborn in a country different from their current country of residence. Suppose there are Igroups of migrants living in the geographical area under investigation, which consists ofR regions. These groups of migrants can, e.g., be defined by country of origin, skill level,age, years since arrival in their current country of residence or other characteristics.Definem as the number of migrants from a specific group j ∈ I living in a specificrjregion r ∈ R, n as the number of natively born individuals living in region r andras the total number of migrants from all groups I in this region with i,j ∈ I . Thecoefficient of geographic association can then be defined as:IR⎛⎞⎜ mrjnr+i 1 rijmax∑ m= ⎟G = ∑−,0R R Ir=1 ⎜∑ mr 1 rjr 1( nr m= ∑ += ∑ i=1 ri ) ⎟⎝⎠The index is constructed by taking the sum over the (positive) differences between thepercentage of migrants from group j living in region r and the percentage of the totalpopulation living in r . The indexG can be interpreted as the proportion of this specificj∑Ii=1mriWIFO 9

group that would need to change its region of residence—together with a similar numberof members of the rest of the population—in order to achieve an allocation of migrants jover regions which follows the distribution of the total population without changing theshare of the region’s population in total population (i.e., leaving the total number ofindividuals residing in each region unchanged).In the case of G = 0 , there is no difference between the geographic distribution of groupjj and the total population, and the members of this group are not regionallyconcentrated. In the case of total segregation—i.e., if all members of j live in only oneregion r where no natives or members of other groups reside—the index takes on thevalue G = 1− m j rj∑ ( n + mr ir)R∑ , which corresponds to one minus the proportion of groupIj in total population. E.g., if group j constitutes 10 % of total population, 90% of itsmembers (together with a similar number of natives and members of other migrantgroups) would have to change their region of residence in order to achieve a uniformdistribution of group j across regions without changing the general distribution of thetotal population. Because of this appealing interpretation the coefficient of geographicassociation has been used in a variety of studies on the regional concentration ofmigrants. 7 The coefficient of geographic association is, however, prone to the “modifiableareal unit problem” (MAUP): results vary with the geographical unit of analysis. Generallyspeaking, the higher the level of aggregation, the smaller the coefficient. Therefore,comparisons across countries are of limited usefulness if the regional units are not similarin characteristics (such as population, size etc.). This also applies to the present analysis:although the “nomenclature des unités territoriales statistiques” (NUTS) ensures at leastsome comparability, the characteristics of the regions in the EU are far too heterogeneousto allow a direct comparison of migrant concentrations across countries. 8 It is thereforenot possible to compare, e.g., the concentration of migrants in Germany to theconcentration of migrants in Ireland because German and Irish regions are not directlycomparable. This also makes comparisons between the concentration of migrants inEurope to that found in other studies—e.g. of the U.S.A. or Australia—difficult.Gjcan,however, be compared among different groups of migrants (e.g., by nationality or skill7 An alternative to the coefficient of geographic concentration is the index of dissimilarity (Duncan and Duncan1955) which is defined as:1 mD = × −2mR rj rj ∑r = 1 R R∑r = 1 rj ∑r = 1The dissimilarity index gives the percentage of members of group j which would have to change residencein order to achieve a geographical distribution similar to that of the natively born population. Because theindex of geographic concentration is more widely used in the literature (see, e.g., Bartel 1989, Chiswick,Lee and Miller 2002), we do not consider the dissimilarity index.8 As an example, the NUTS 2 regions in the EU-15 differ substantially by population, from more than 11 Mio.persons living in Île de France to Åland’s (Finland) 26,800 inhabitants. They are also heterogeneous withrespect to size: Övre Norrland (Sweden), for example, has an area of more than 165,000 km² (and about 3inhabitants per km²), while Bruxelles-Capitale is smaller by a factor of more than 1,000 (161 km², withabout 6,200 inhabitants per km². Source: Eurostat). The coefficients for different groups of target regionsare therefore not generally comparable.nnrWIFO 10

level) within the EU to assess whether they show a higher or lower level of geographicalconcentration.2.4 The regional concentration of migrants in EuropeTo analyse the regional concentration of migrants in Europe, the coefficient ofgeographical association is calculated for all NUTS 2 regions in the EU-15 using the mostrecent European Labour Force Survey (LFS) data available (2007). The LFS is a regularquestionnaire surveyed among a representative sample of households in all countries ofthe EU-27. We define as migrants all individuals who were not born in the member statethey reside in, while all those who still live in their country of birth are considered“natives”. 92.4.1 Regional concentration: facts and figuresAt the NUTS 2 level of aggregation, the coefficient of geographical association in the EU-15 is Gj=0.239: 10 23,9% of all migrants in Europe would have to change their place ofresidence (together with a similar number of “natives”, i.e., those who still live in thecountry they were born in) in order to achieve an even distribution of migrants acrossEurope without changing the relative population across regions (see table 2.1).The region with the largest share of migrants in Europe is the Île de France regionincluding the French capital Paris: 5.7% of all migrants in Europe live in this region.Large proportions of migrants can also be found in Outer (3.2%) and Inner London(2.8%) as well as Cataluña and the Comunidad de Madrid in Spain (both 2.7%). Île deFrance is also the European region with the largest difference between the percentage ofmigrants and the percentage of the total population in Europe (the “localconcentration”) 11 , closely followed by Inner and Outer London. The Darmstadt region9 EU-15 citizens can either be natives or migrants, depending on whether they still reside in their country ofbirth or in another EU-15 member state. Individuals who provided no information on this question wereclassified as natives.10 As mentioned in section 2.3, the coefficient of geographical association cannot be directly compared acrossstudies because of the MAUP. However, to get a sense of the size of the coefficient, the value found here isabout the same in magnitude as the one found by Chiswick, Lee and Miller (2002) for Australia( G =0.221), but considerably smaller than the values reported by Bartel (1989) in her comparison amongj29 U.S. SMSAs ( G =0.308 tojG =0.525).j11 The „local concentration“ is defined formally as:gIm n + ∑∑ ∑ ∑rjr= −rj R R Imi = 1( + )m n mr r i= = =1 1 1rj r riRwhere, of course, Gj=∑ max ( grj,0).rir = 1WIFO 11

(including the city of Frankfurt am Main) and the Comunidad de Madrid can also be foundamong the regions with the highest local concentration (see figure 2.1). 12Figure 2.1: Local concentration of migrants in the EU-15Source: European Labour Force Survey 2007, Eurostat, WIFO.The data also show that more recent migration waves do not differ substantially in theirconcentration from previous migrants: the coefficient of geographical association forthose who moved to their current country of residence more than 10 years ago isGj=0.296, while it is only slightly smaller for those who migrated during the last 9 years( Gj=0.288). However, the target regions of more recent migration waves areconsiderably different from those of earlier cohorts. 13While the largest localconcentration of migrants who moved more than 10 years ago can be found in the Île deFrance, Inner and Outer London, Darmstadt and Provence-Alpes-Côte d’Azur regions, themore recent cohorts concentrated in Spain (Cataluña, Comunidad de Madrid and12 The largest relative concentration, defined as the ratio of the proportion of migrants living in this regionamong all migrants in the EU-15 to the proportion of all non-migrants living in this region among all nonmigrantsin the EU-15 can be found, however, in Inner London: the percentage of migrants living there is3.8 times larger than the share of native population. Luxembourg (relative ratio of migrants to natives 3.3),the Brussels region (3.1), Outer London (2.8) and Vienna (2.7) also show a high degree of relativeconcentration.13 It cannot be ruled out that the more recent migration waves also include repeat migrants, i.e. individualswho migrated more than 10 years ago and repeated again in the meantime. The “true” target regions of allprevious migrants can therefore not be observed.WIFO 12

Comunidad Valenciana) 14 . The correlation between the local concentration of earlier andmore recent cohorts is rather low ( r xy=0.159) and significant only at the 5 percent level.This indicates that there has been a shift in regional preferences of migrants.2.4.2 Does regional concentration differ by individual characteristics?The coefficient of geographical association differs noticeably with age: older migrants aremore concentrated than younger migrants. E.g., the age group of 50-59 years has acoefficient ofGj=0.309, while it is onlyGj=0.251 for individuals in their twenties andeven lower for those between 30 and 39 ( G =0.234). With respect to preferred targetregions, the age groups differ only slightly. Inner and Outer London, Cataluña, theComunidad de Madrid and Île de France comprise the 5 regions with the highest localconcentration for younger migrants, while the region Provence-Alpes-Côte d’Azur as wellas the German agglomerations Darmstadt, Düsseldorf and Stuttgart are also popular withmigrants age 50 and older (see table 2.1).Taking into account time since migration, we can roughly compare cohorts of similar age.It can be seen that the concentration of younger migrants from more recent waves isslightly higher than that of earlier waves (table 2.1): e.g., for migrants in their twentieswho migrated during the last 10 years, the coefficient of geographical association is0.309, while it is only 0.259 for those who migrated more than 10 years ago or earlierand who are now in their thirties. The same can be observed, albeit less pronounced, forthose in the 30-39 age group who migrated during the last decade ( G =0.305 vs.Gj=0.283 for those who migrated more than 10 years ago and who are now in theirforties). Interestingly, the opposite is true for the cohorts of 40-49 and 50-59, who arenow less concentrated among regions than earlier groups of working-age migrants of thesame age (see table 2.1).jjThese changes in concentration were accompanied by changes in target regions. German(industrial) regions like Darmstadt, Stuttgart, Detmold (with cities of Bielefeld andPaderborn) or the Ruhr Area regions of Arnsberg (Dortmund, Bochum etc.) andDüsseldorf were especially popular with earlier migration waves alongside the Londonareas, Île de France, South Holland (Rotterdam, The Hague etc.) and Provence-Alpes-Côte d’Azur. However, while more recent migrants also concentrated in the (Inner andOuter) London areas, the Spanish regions of Cataluña, Cumunidad de Madrid, ComunidadValenciana and Andalucía were amongst the regions with the highest local concentrationwhile at the same time there are only minor differences across age groups. It can thus be14 These are also the regions with the largest absolute inflows of migrants during the last decade. Outer andInner London and the Île de France also experienced large concentrations of recent migrants, as haveSouthern and Eastern Ireland as well as other Spanish regions (Andalucía, Región de Murcia and Canarias).WIFO 13

concluded that the regional concentration does not vary as strongly with age as it doeswith timing of migration. 15Table 2.1: Coefficients of geographical association and local concentration in the EU-15General, by age and educationMigrant groupTime since migrationGeneral Total 0.239 FR10 UKI1 UKI2≤ 10 years 0.288 ES51 ES30 ES52> 10 years 0.296 FR10 UKI1 UKI2Age 15-19 Total 0.241 ES30 ES51 UKI1≤ 10 years 0.318 ES30 ES51 UKI1> 10 years 0.314 UKI1 DE71 -Age 20-29 Total 0.251 UKI1 UKI2 ES51≤ 10 years 0.309 ES51 UKI2 ES30> 10 years 0.317 FR10 UKI1 DEA5Age 30-39 Total 0.234 UKI1 ES52 ES30≤ 10 years 0.305 ES51 ES30 ES52> 10 years 0.259 FR10 UKI2 UKI1Age 40-49 Total 0.243 FR10 UKI1 UKI2≤ 10 years 0.319 ES30 ES51 ES52> 10 years 0.283 FR10 UKI2 UKI1Age 50-59 Total 0.309 FR82 DE71 DEA1≤ 10 years 0.333 ES30 ES52 ES51> 10 years 0.353 FR10 UKI2 FR82Age 60 + Total 0.371 FR10 FR82 DE11≤ 10 years 0.386 ES52 ES61 -> 10 years 0.388 FR10 FR82 DE11Primary education Total 0.377 FR10 ES51 FR82≤ 10 years 0.409 ES51 ES30 ES52> 10 years 0.432 FR10 FR82 FR71Secondary education Total 0.243 UKI2 UKI1 FR10≤ 10 years 0.288 ES30 UKI2 ES51> 10 years 0.316 DE71 DE11 FR10Tertiary education Total 0.279 FR10 UKI1 UKI2≤ 10 years 0.359 UKI1 ES30 FR10> 10 years 0.292 FR10 UKI2 UKI1GjTop 3 regions byg (NUTS2)rjNotes: Empty cells indicate that the number of migrants with the respective characteristic in thisregion is below reliability limits according to the EU LFS publishing guidelines. See the appendix fora list of NUTS 2 codes used.Source: European Labour Force Survey 2007, Eurostat, WIFO.Education does also play an important role in explaining the regional concentration ofmigrants. Broken down by educational levels, table 2.1 reveals that low skilled migrants15 These concentrations of earlier migrants in German regions can partly be explained by the German “guestworker” scheme implemented until the 1970s, where workers were actively recruited abroad by Germangovernment offices. Many of these guest workers did not move back to their home countries, but stayed inGermany. Family reunion laws, network and herd migration will have led to an ongoing in-migration intothese regions even after the 1970s. The high local concentrations in Spanish regions of more recentWIFO 14

with primary education are much more concentrated ( G =0.377) than migrants withsecondary ( G =0.243) or tertiary education ( G =0.279). 16 This accords with the findingsjof Bartel (1989) for the U.S.A. and suggests that migrant networks are more importantfor low-skilled workers: they need to rely on networks to find jobs, while medium- andhigh-skilled workers, who are more likely to know the host country’s native language,tend to disperse more. The concentration of high-skilled migrants who moved during thelast 10 years ( G =0.359) has, however, increased (compared to those who migrated 10years ago or earlier,jprimary and secondary education (fromrespectively).jGj=0.292), while it has decreased for more recent migrants withGj=0.432 to 0.409 and fromjGj=0.316 to 0.288,Comparing the largest local clusters shows that more recent migrant cohorts tend toprefer Spanish regions. This observation is most pronounced, but not limited to, lowskilledmigrants: while those with primary education who migrated more than 10 yearsago mainly concentrated in French regions (with 13.7% living in the Île de France), lowskilledworkers who migrated during the last decade can mainly be found in Spain,especially in Cataluña (12.6%). Nevertheless, the correlation between the localconcentrations of earlier and more recent low-skill migrants is statistically significant atthe 5 percent level, albeit rather small ( r xy=0.137). Migrants with secondary education,on the other hand, tended to migrate to Germany more than 10 years ago: apart fromthe Île de France, Darmstadt, Stuttgart, Arnsberg, Düsseldorf and Cologne are among theregions with the highest concentrations of medium-skilled earlier migrants. More recentcohorts with secondary education, however, also prefer Spain (e.g., the Comunidad deMadrid, Cataluña or Comunidad Valenciana), while not a single German region can befound among the 10 regions with the highest local concentration of migrants in this skillgroup. The local concentrations of the migration waves are also not significantlycorrelated, pointing to a substantial change in regional preferences of medium-skilledmigrants.Finally, many Spanish regions can also be found among the most highly concentratedregions for recent high-skill migrants while Stockholm, Darmstadt, South and NorthHolland (with the Dutch capital Amsterdam), the Provence-Alpes-Côte d’Azur or Brusselsregions (which were important for earlier high-skill migrants) are no longer among thepreferred destinations for this skill group. Nevertheless, there is a high and significantcorrelation between the local concentrations of earlier and more recent migrant cohortswith tertiary education ( r xy=0.598).migration waves can be attributed to large in-migration from Morocco, Ecuador and Columbia during thelast 10 years, which may be due to language similarities.16 Primary education: ISCED levels 0 or 1; secondary education: ISCED 2-4; tertiary education: ISCED 5 and6.WIFO 15

2.4.3 Ethnic migration clustersTo investigate the possibility of network or herd migration, we also analyse the regionalconcentration of migrants by ethnicity. We define ethnicity using the migrants’ country oforigin, although this could be considered a very simplified definition by socialanthropology standards as ethnicity must not necessarily coincide with nationalboundaries. The migrants’ country of origin can either be deduced from the country ofbirth or from the nationality of the migrants. Unfortunately, while the latter is a morevague definition (migrants might attain their host country’s nationality, in which casetheir true country of origin is no longer observed), the former is not available formigrants in all EU-15 countries. E.g., Germany does not ask for the country of birth in itsLabour Force Survey questionnaire, only for nationality. Unfortunately, neither thenationality nor the country of birth is available for migrants in Ireland.Using country of birth, the coefficient of geographical association can therefore only becalculated for 13 EU countries excluding Germany and Ireland. For migrants from theeight new CEE member states which joined the EU in 2004 (NMS-8), a concentrationcoefficient ofGj=0.451 can be observed, which is smaller than the coefficient forBulgaria and Romania ( G = 0.542, see table 2.2). While the latter are mostlyjconcentrated in Spain and Italy, the former show a high degree of concentration in(Outer and Inner) London and eastern Austria (Vienna and Lower Austria). However,other U.K. regions (e.g., East Anglia, Leicestershire, Rutland and Northamptonshire, WestYorkshire, Gloucestershire, Wiltshire and Bristol/Bath Area) as well as Southern Swedenand Stockholm are also among the most highly concentrated regions in the EU-15 (seefigure 2.2).Migrants from the candidate countries Turkey, Macedonia and Croatia show a degree ofconcentration which is about equal in size to that of migrants from Romania and Bulgaria(0.556). Migrants from other countries, which make up for 65.9% of all migrants inEurope according to this definition, are more evenly distributed across regions. Thecoefficient associated with these migrants ( G =0.284) is even lower than the one formigrants from EU-15 countries (0.307). 17It could be expected that using citizenship to define country of origin will put some ofthese results into perspective, especially since a large proportion of all migrants inEurope (28.3%) live in Germany, for which data on country of birth is not available. Theresults do, however, not change considerably. The coefficient of geographical associationis only slightly different from that measured using country of birth to define ethnicity( Gj=0.458). Migrants from the NMS-8 are still most concentrated in Outer and InnerLondon as well as Vienna (see figure 2.3). The distribution of migrants from Bulgaria andRomania is also largely unchanged, although the coefficient of geographical associationj17 It should, however, be noted that this even distribution is in past also due to the ethnic heterogeneity of themigrants subsumed under the category of other countries.WIFO 16

increases slightly from 0.542 to 0.605. Changing the definition of ethnicity from countryof birth to nationality thus increases the coefficients slightly, but leaves most resultsunchanged.Table 2.2: Coefficients of geographical association and local concentration in the EU-15by country groupsMigrant groupTime since migrationEU-15 Total 0.307 FR10 UKI1 LU00≤ 10 years 0.370 UKI1 ES52 LU00> 10 years 0.307 FR10 FR82 LU00NMS-8 Total 0.451 UKI2 AT13 UKI1≤ 10 years 0.578 UKI2 UKI1 UKH1> 10 years 0.453 AT13 AT12 SE22Bulgaria and Romania Total 0.542 ES52 ES30 ES61≤ 10 years 0.592 ES52 ES30 ES61> 10 years 0.551 ITE4 ES30 AT13Candidate countries Total 0.556 AT13 NL33 AT12≤ 10 years 0.538 AT13 ITD3 UKI1> 10 years 0.581 AT13 NL33 AT12Other Total 0.284 FR10 UKI1 UKI2≤ 10 years 0.334 ES51 ES30 UKI2> 10 years 0.316 FR10 UKI1 UKI2Migrant groupTime since migrationEU-15 Total 0.301 FR10 UKI2 UKI1≤ 10 years 0.287 UKI1 UKI2 FR10> 10 years 0.314 FR10 UKI2 FR82NMS-8 Total 0.458 UKI2 UKI1 AT13≤ 10 years 0.537 UKI2 UKI1 UKH1> 10 years 0.498 AT13 ES30 DE11Bulgaria and Romania Total 0.605 ES52 ES30 ES61≤ 10 years 0.626 ES52 ES30 ES61> 10 years 0.656 ES30 ITE4 ES52Candidate countries Total 0.591 DE30 DEA1 DE71≤ 10 years 0.530 AT13 - -> 10 years 0.635 DE30 DEA1 DE71Other Total 0.320 ES51 FR10 ES30≤ 10 years 0.362 ES51 ES30 ES52> 10 years 0.323 FR10 UKI1 ITC4GjGjCountry of birthNationalityTop 3 regions byTop 3 regions bygrjgrj(NUTS2)(NUTS2)Notes: Country of birth: Germany and Ireland not included. Nationality: Ireland not included. Emptycells indicate that the number of migrants with the respective characteristic in this region is belowreliability limits according to the EU LFS publishing guidelines. See the appendix for a list of NUTS2 codes used.Source: European Labour Force Survey 2007, Eurostat, WIFO.WIFO 17

Figure 2.2: Local concentration of NMS-8 migrants in the EU-15Based on country of birthSource: European Labour Force Survey 2007, Eurostat, WIFO. Germany and Ireland not included.Large differences can, however, be observed for migrants from candidate and othercountries. Although the concentration coefficient for migrants from Turkey, Croatia andMacedonia increases by only 0.035 points, it is clear from table 2.2 that migrants fromthese countries are mostly concentrated in Germany, especially in Berlin, Düsseldorf andDarmstadt, Stuttgart and Oberbayern. The regional distribution of migrants from othercountries, which make up for only 29.2% of all migrants in the EU-15 (excluding Ireland)according to this definition, does also change significantly if citizenship is used to defineethnicity.Looking at the individual countries of origin in more detail (see table 2.3), it can be seenthat the regional concentration of migrants differs substantially between migrants withdifferent backgrounds and that there are some substantial differences depending on thedefinition of ethnicity (by country of birth—and thus excluding Germany—or bynationality). The largest concentration of migrants from NMS-8 countries can be foundamong Estonians: Using country of birth as the proxy for ethnic background, thecoefficient of geographical association isGj=0.877, which is slightly lower than ifnationality is used as a proxy ( G =0.897). The largest local concentration can be foundjin the geographically close region of Southern Finland, where more than 40% of allEstonian migrants in the EU-15 reside. Relatively high concentrations of migrants canalso be found for the other Baltic states, Latvia (0.765) and Lithuania (0.747).WIFO 18

Figure 2.3: Local concentration of MNS-8 migrants in the EU-15Based on nationalitySource: European Labour Force Survey 2007, Eurostat, WIFO. Ireland not included.Migrants from Slovenia also show a high degree of concentration ( G =0.834-0.878,depending on definition). Migrants from Slovakia are substantially more concentratedthan Czech or Hungarian migrants, although for the latter this depends on the definitionof ethnicity. The lowest concentration of all NMS-8 countries can be found for Polishmigrants ( G =0.443-0.477) which are mostly concentrated in the U.K. (especially in thejLondon areas, East Anglia and West Yorkshire) and in Vienna.By country of birth and nationality Polish workers abroad are thus less concentrated thanmigrants from the EU candidate countries Turkey and Croatia. Using nationality to defineethnicity, the largest concentrations of migrants from these countries can be foundamong German regions (Düsseldorf, Berlin and Darmstadt for Turkish, Stuttgart forCroatian migrants) as well as Vienna (which experienced a large inflow of Croatianmigrants during and following the Yugoslav wars). Migrants from the third candidatecountry, Macedonia, cluster mainly in Italy (Veneto).jWIFO 19

Table 2.3: Coefficients of geographical association in the EU-15 by country of originG jCountry of Origin Country of birth NationalityPoland 0.443 0.477Czech Republic 0.619 0.656Hungary 0.637 0.741Slovakia 0.755 0.767Estonia 0.877 0.897Lithuania 0.747 0.756Latvia 0.765 0.837Slovenia 0.834 0.878Bulgaria 0.581 0.624Romania 0.562 0.637Turkey 0.614 0.623Croatia 0.689 0.719Macedonia 0.722 0.777Morocco 0.500 0.568Algeria 0.734 0.734Ecuador 0.769 0.814India 0.611 0.686Albania 0.762 0.798Pakistan 0.688 0.711Tunisia 0.640 0.711China 0.425 0.463Columbia 0.722 0.811Notes: Country of birth: Germany and Ireland notincluded. Nationality: Ireland not included.Source: European Labour Force Survey 2007, Eurostat,WIFO.Other large migrant groups show varying degrees of concentration. Migrants fromMorocco ( G =0.500-0.568) and Algeria (0.734), the (based on country of birth) twojlargest migrant groups in Europe, are mainly concentrated in France or Spain, which canbe explained by former colonial ties. This holds, in part, also for Tunisian migrants. Theyall are, however, less concentrated than migrants from Ecuador 18 , who moved almostexclusively to Spanish regions, especially the Comunidad de Madrid, Cataluña andValencia ( G =0.769-0.814). A similar pattern can be observed for migrants fromjColombia. For both Ecuadorian as well as Colombian migrants, language can be expected18 Ecuadorians are a sizeable group among non-EU migrants in the EU15. In Spain they account for over600.000 migrants.WIFO 20

to be the main driving force for migrating to Spain. Former colonial ties also play a rolefor Indian and Pakistani migrants, who mostly cluster in U.K. regions. Large localconcentrations of migrants from Albania can be found in the nearby Greek region of Attikias well as in Italy (Tuscany and Lombardy).Table 2.4: Coefficients of geographical association within EU-15 countries by sendingcountry groupsG jby sending countriesReceiving country EU-15 NMS-12 Candidates OtherAustria 0.202 0.231 0.257 0.266Belgium 0.286 0.342 0.350 0.324Germany 0.115 0.251 0.307 0.233Denmark 0.122 0.185 0.249 0.114Spain 0.205 0.257 - 0.223Finland 0.082 0.334 - 0.247France 0.204 0.311 0.308 0.284Greece 0.180 0.264 0.412 0.159Italy 0.099 0.250 0.352 0.163Netherlands 0.150 0.129 0.160 0.205Portugal 0.143 0.539 - 0.347Sweden 0.119 0.220 0.248 0.120United Kingdom 0.231 0.263 0.624 0.353Notes: Calculations for Germany based on nationality, country of birth otherwise.Ireland and Luxembourg not included. Empty cells indicate that the number ofmigrants from the respective country group is below reliability limits according to theEU LFS publishing guidelines for all regions with positive local concentrations.Source: European Labour Force Survey 2007, Eurostat, WIFO.2.4.4 An analysis by receiving countryThe preceding results are, however, based on data for all EU-15 regions and could thusbe influenced by differences in national migration regimes. To account for this possibility,table 2.4 reports the coefficient of geographic concentration for different sending countrygroups within all EU-15 countries for which data can be considered representative andwhich are composed of more than one region. Controlling for national differences in thisway reduces the measure of geographic concentration somewhat. While table 2.2suggests that between 30% and 60% of the foreign born would have to change region ofresidence to achieve a distribution equivalent to that of natives (depending on theirregion of origin), on a within-country basis this figure reduces to 20% to 40% ofmigrants which would have to change residence within a country. This reduction,however, was to be expected, since now a much smaller number of regions is consideredin calculation.WIFO 21

It must be borne in mind, however, that comparisons across countries are not possible intable 2.4 due to the MAUP. Comparisons between different sending country groups in thesame country are, however, possible. These suggest that migrants from the candidatecountries are the most concentrated in almost all EU-15 countries. The coefficients ofgeographic concentration of other migrant groups are higher only in Austria, France andthe Netherlands. By contrast, migrants from other EU-15 countries are the leastconcentrated in all countries except for Greece the Netherlands and Denmark. Finallymigrants from both the 12 new member states and other countries typically havecoefficients of geographic concentration between these two extremes.2.4.5 Is there evidence of network or herd migration?The local concentrations of migrants with similar ethnic background can also becompared across migration waves: if migrants who moved during the last 10 yearsconcentrated in the same regions as those who migrated more than 10 years ago, thiscould be seen as evidence for either network or herd migration. Whether this is the casewill be tested by computing the coefficient of correlation between the local concentrationof these different migrant cohorts. As mentioned above, there is a measurement error inthe concentration of migrants across EU-15 NUTS 2 regions if ethnicity is defined bynationality, because some migrants will have acquired the citizenship of the country theylive in. This is especially true for those who migrated more than 10 years ago. 19 Thus,comparisons across migration waves are more meaningful if ethnicity is defined bycountry of birth, but in this case, German regions could not be included in thiscomparison. Therefore, comparisons across migrant waves where nationality was used todefine ethnicity are also reported, although they should be interpreted with caution asthey are probably biased.As table 2.4 shows, the coefficients of concentration change little between cohorts formost ethnic groups. For migrants from some NMS-8 countries, the coefficient hasdecreased: especially more recent migrants from the Czech Republic, Slovakia, Lithuaniaand Latvia are considerably less concentrated than their predecessors, at least ifnationality is used to define the country of origin. But generally, an increase ingeographical concentration between migrant cohorts can be observed: e.g., more recentPolish emigrants are more concentrated than their predecessors ( G =0.586, vs. 0.449using country of birth). The increase in regional concentration is however less impressiveif nationality defines ethnicity.j19 If migrants would acquire citizenship randomly with the same rate across regions (i.e., if each period aconstant (across regions and across ethnicities) percentage of migrants would acquire citizenship of thecountry of residence), the relative concentration would remain unchanged. In this case, nationality couldalso be used to derive ethnicity. However, because of differences in citizenship laws across Europe, theconcentration coefficients cannot be expected to remain unbiased.WIFO 22

Table 2.5: Cohort differences in regional concentration by countries and correlationbetween local concentrationsCountry of birthNationalityCountry of origin > 10 years ≤ 10 years Correlation > 10 years ≤ 10 years CorrelationGjPoland 0.449 0.586 0.205*** 0.529 0.555 0.102Czech Republic 0.696 0.704 0.221*** 0.820 0.712 0.121*Hungary 0.669 0.758 0.399*** 0.806 0.830 0.322***Slovakia 0.902 0.767 0.464*** 0.940 0.790 0.123*Estonia 0.936 0.903 0.822*** 0.972 0.906 0.723***Lithuania 0.864 0.779 0.427*** 0.945 0.776 0.138**Latvia 0.885 0.779 0.105 0.980 0.852 -0.016Slovenia 0.842 0.958 0.674*** 0.903 0.966 0.101NMS-8 0.453 0.578 0.256*** 0.490 0.537 0.159**Bulgaria 0.634 0.626 0.552*** 0.772 0.645 0.567***Romania 0.575 0.614 0.372*** 0.709 0.664 0.571***Turkey 0.638 0.599 0.767*** 0.656 0.583 0.832***Croatia 0.696 0.810 0.699*** 0.740 0.800 0.620***Macedonia 0.739 0.767 0.735*** 0.828 0.811 0.843***Morocco 0.523 0.541 0.450*** 0.574 0.609 0.676***Algeria 0.764 0.659 0.889*** 0.784 0.704 0.921***Ecuador 0.759 0.775 0.824*** 0.854 0.851 0.844***India 0.633 0.640 0.917*** 0.732 0.707 0.846***Albania 0.783 0.745 0.818*** 0.829 0.786 0.819***Pakistan 0.731 0.669 0.889*** 0.745 0.721 0.839***Tunisia 0.675 0.658 0.920*** 0.746 0.739 0.835***China 0.470 0.457 0.632*** 0.624 0.465 0.544***Columbia 0.637 0.782 0.669*** 0.825 0.828 0.553***rxyGjrxyNotes: *** significant at 1%, ** significant at 5%, * significant at 10% level. See the appendix for a list ofNUTS 2 codes used.Source: European Labour Force Survey 2007, Eurostat, WIFO.In either case, the target countries of recent Polish immigration waves have shiftedconsiderably: while earlier emigrants concentrated mostly in Vienna and the Comunidadde Madrid, later migrant waves have shifted to U.K. regions (especially the London areas,East Anglia and West Yorkshire). This holds irrespective of the definition of ethnicity.Accordingly, the correlation of local concentration values between migration waves israther low and insignificant if computed using nationality. This is also the case for mostother NMS-8 countries, suggesting that recent migration patterns did not follow networkmotives strongly. Only among Estonians (and, albeit less strongly, in Hungarians) a highcorrelation can be observed which indicates a strong tendency towards network migration(especially to Southern Finland). Considering migrants from the NMS-8 as a single group,WIFO 23

the correlation between the local concentrations of different migration cohorts is ratherlow.Table 2.6: Cohort differences and correlation between local concentrations within EU-15countriesEU-15NMS-12Receiving country > 10 years ≤ 10 years Correlation > 10 years ≤ 10 years CorrelationG jAustria 0.173 0.249 0.960*** 0.221 0.318 0.949***Belgium 0.289 0.317 0.624** 0.251 0.518 0.885***Germany 0.122 0.178 0.325** - - -Denmark 0.117 0.195 0.816* 0.289 - -Spain 0.196 0.263 0.852*** 0.577 0.241 0.612***Finland 0.096 0.316 -0.123 0.375 0.281 0.982***France 0.212 0.274 0.486** 0.298 0.417 0.874***Greece 0.187 0.249 0.265 0.261 0.283 0.922***Italy 0.106 0.160 -0.045 0.279 0.246 0.796***Netherlands 0.149 0.156 0.930*** 0.090 0.212 0.818***Portugal 0.179 - - - - -Sweden 0.141 0.146 0.056 0.219 0.235 0.866***UK 0.201 0.354 0.737*** 0.316 0.263 0.884***r xyG jr xyCandidate countriesOther countriesReceiving country > 10 years ≤ 10 years Correlation > 10 years ≤ 10 years CorrelationG jAustria 0.259 0.249 0.986*** 0.253 0.293 0.985***Belgium 0.330 0.405 0.886*** 0.341 0.329 0.986***Germany 0.308 - - 0.284 0.216 0.657***Denmark 0.236 0.292 0.977*** 0.197 0.054 0.808*Spain - - - 0.207 0.240 0.723***Finland - - - 0.283 0.201 0.990***France 0.315 - - 0.286 0.299 0.924***Greece 0.430 - - 0.178 0.157 0.892***Italy 0.315 0.441 0.878*** 0.140 0.233 0.777***Netherlands 0.175 - - 0.211 0.181 0.963***Portugal - - - 0.291 0.454 0.987***Sweden 0.263 0.236 0.899*** 0.136 0.114 0.897***UK 0.737 0.596 0.756*** 0.369 0.333 0.988***rxyG jrxyNotes: *** significant at 1%, ** significant at 5%, * significant at 10% level. Calculations for Germanybased on nationality, country of birth otherwise. Ireland and Luxembourg not included. Empty cellsindicate that the number of migrants from the respective country group is below reliability limitsaccording to the EU LFS publishing guidelines for all regions with positive local concentrations.Source: European Labour Force Survey 2007, Eurostat, WIFO.For the two countries which joined the EU in 2007, there is still a strong correlationbetween the local concentration of earlier and more recent migration waves, as is alsothe case for the three candidate countries, especially Turkey and Macedonia. HighlyWIFO 24

significant and positive correlations can also be observed for migrants from othercountries with a high number of migrants in the EU-15 like Algeria (with the largestcluster in Île de France), Ecuador (Comunidad de Madrid), India (Outer London), Tunisia(Île de France), Albania (Attiki) or Pakistan.These results suggests a substantial change of migration patterns for migrants from theNMS. However, part of these differences in the settlement structure of migrant cohortsare due to institutional changes in migration regimes in the EU-15 during the last decade,in particular with respect to enlargement. Performing the same analysis as in table 2.5for migrants within a country (see table 2.6) corroborates this finding. While there are nounambiguous patterns in the concentration of different cohorts within countries (exceptfor the case of migrants from the EU-15, where the more recent cohort is unambiguouslymore concentrated), within-country correlation coefficients are much larger. This isespecially true for migrants from the 12 new member states, the candidate countries andother countries. This thus suggests that, after the choice of country is fixed, networkeffects and/or herd effects are still important for the decision where to settle within acountry. The only exception to this are migrants from the EU-15. Here coefficients ofcorrelation are mostly rather low or even negative.Thus, even though the preferences of migrants over target regions have apparentlychanged between cohorts, it can not be concluded that network and/or herd effects donot affect the locational choice of migrants in the EU-15. These changes are rather theresult of changes in immigration laws during the last years, especially due to EUenlargement and thus rather reflect national migration regimes than the absence ofnetwork migration.In order to measure the importance of networks on migration decisions, a conditionallogit regression 20 is estimated which measures the influence of network size on theprobability of migration to a specific region after controlling for other factors affectinglocational choice. The empirical estimation is based on a theoretical model where amigrant can choose among all available locations. The observed move of an individual toa specific region can then be interpreted as a move to the location which providesmaximum utility (see Bartel, 1989 for details).As explanatory variables in the regression, the population and area of the target regionare used alongside the unemployment rate and the average income per employedperson. The local network size is defined using the proportion of migrants from the samecountry of origin living in this region for more than 10 years. To allow for a decreasingmarginal utility of networks, the squared value of this variable also enters the regression.20 See also Bartel (1989), Bauer, Epstein and Gang (2000, 2002, 2005) or Jaeger (2007).WIFO 25

Table 2.7: Conditional logit regression of locational choiceCountry of birthBulgaria andBulgaria andSource country All NMS-8 Romania All NMS-8 RomaniaModel (1) (2) (3) (4) (5) (6)Population (2006, in 100,000) 1.250*** 1.298*** 1.199*** 1.268*** 1.324*** 1.202***(0.000) (0.001) (0.001) (0.000) (0.001) (0.001)Area (in 1,000 km²) 1.001*** 1.014*** 1.021*** 1.001*** 1.023*** 1.024***(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)Unemployment rate (2006, in %) 0.968*** 0.975*** 0.981*** 0.964*** 0.989*** 0.999***(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)Ø yearly income per employee 1.021*** 1.017*** 1.028*** 1.022*** 1.006*** 1.019***(2004, in 1,000 €) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)Capital region 0.723*** 0.613*** 0.738*** 0.720*** 1.093*** 0.912***(0.001) (0.002) (0.003) (0.001) (0.004) (0.004)Distance between region and country of 0.502*** 0.029*** 0.731***origin (in 1,000 km) (0.000) (0.000) (0.014)Squared Distance between region and 1.017*** 1.368*** 0.863***country of origin (in 1,000 km) (0.000) (0.004) (0.004)Proportion of migrants from same country 1.287*** 1.388*** 1.690*** 1.267*** 1.332*** 1.759***living in region (in %) (0.000) (0.001) (0.003) (0.000) (0.001) (0.003)Squared proportion of migrants from 0.996*** 0.995*** 0.971*** 0.996*** 0.994*** 0.967***same country living in region(in %) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)Pseudo-R² 0.206 0.201 0.317 0.215 0.227 0.321Bulgaria andBulgaria andSource country All NMS-8 Romania All NMS-8 RomaniaModel (7) (8) (9) (10) (11) (12)Population (2006, in 100,000) 1.296*** 1.357*** 1.237*** 1.314*** 1.376*** 1.239***(0.000) (0.001) (0.001) (0.000) (0.001) (0.001)Area (in 1,000 km²) 1.000*** 1.013*** 1.021*** 1.001*** 1.024*** 1.023***(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)Unemployment rate (2006, in %) 0.977*** 0.976*** 0.972*** 0.970*** 0.987*** 0.987***(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)Ø yearly income per employee 1.020*** 1.018*** 1.041*** 1.021*** 1.004*** 1.034***(2004, in 1,000 €) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)Capital region 0.712*** 0.633*** 0.626*** 0.740*** 1.254*** 0.760***(0.001) (0.002) (0.003) (0.001) (0.005) (0.004)Distance between region and country of 0.383*** 0.018*** 0.866***origin (in 1,000 km) (0.000) (0.000) (0.016)Squared Distance between region and 1.030*** 1.456*** 0.864***country of origin (in 1,000 km) (0.000) (0.004) (0.003)Proportion of migrants from same country 1.186*** 1.207*** 1.362*** 1.160*** 1.183*** 1.377***living in region (in %) (0.000) (0.001) (0.001) (0.000) (0.000) (0.001)Squared proportion of migrants from 0.998*** 0.997*** 0.990*** 0.998*** 0.997*** 0.989***same country living in region(in %) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)Pseudo-R² 0.195 0.194 0.327 0.211 0.227 0.330Notes: Odds ratios reported. Standard errors in parentheses. *** significant at 1%, ** significant at 5%, * significant at 10%level. Dependent variable: locational choice by NUTS 2 regions. Ireland, overseas territories and exclaves excluded in bothmodels. Germany excluded in model based on country of birth. Coefficients of country dummies not reported.Source: European Labour Force Survey 2007, Eurostat, WIFO.NationalityTo control for national differences in immigration laws as well as other country-fixedeffects, dummies for the receiving countries are included as is a dummy variable forWIFO 26

those regions which comprise national capitals. To proxy for the costs of migration (or thecosts of visiting relatives at home), the distance of migration 21 and its squared value arealso included in a second specification. Table 2.7 shows the results of the conditional logitregression. The coefficients shown can be interpreted as odds ratios, with values largerthan 1 indicating an increase in the odds while values smaller than 1 indicate a decreasein the odds of an event occurring (i.e., choosing a specific region) when the independentvariable changes by one.As the table shows, there are only minor differences between the regression based oncountry of birth (excluding Germany and Ireland) and nationality (excluding Ireland).Differences between a general estimation including migrants from all countries andrestricted estimations including only migrants from the NMS-8 or Bulgaria and Romaniaare also to a large extent similar. Both population and area have a positive effect on thechoice of the preferred target region: the larger the region, the higher the probability thatit will be chosen as preferred destination of migration. Higher unemployment isassociated with a lower probability of choosing this region: an increase in theunemployment rate of 1 percentage point decreases the odds of migrating to this regionbetween 2.3% in model (7) and 3.2% in model (1). The attractiveness of a regionincreases with the average annual income per employee: the odds of choosing a regionincrease by about 2% if average yearly income is € 1,000 higher, all else equal.Surprisingly, capital regions are significantly less attractive, ceteris paribus, than otherregions, even after controlling for country specific fixed effects: Generally, the odds of amigrant choosing a capital city region are 27.3% (or 19.6%, respectively, if nationality isused to define country of origin) lower than otherwise. This thus suggests that migrantsprefer larger urban areas (as illustrated by the positive effects of region size), but notnecessarily capitals. The distance between the region of residence in the EU-15 and thecountry of origin also significantly affects locational choice. There are, however,differences between regions of origin. Generally, the probability of choosing a regionceteris paribus decreases with the distance from the home country. This negative effecthowever becomes smaller as the distance increases, as is shown by the odds ratio > 1 ofsquared distance in models (4), (5), (10) and (11). For migrants from Bulgaria andRomania, however, the effect is increasingly negative. The effect of distance is thus notunambiguous when looking at different groups of origin countries.Even after controlling for these demographic, geographic and economic factors, networksshow a significant effect on locational choice: the higher the share of migrants from thesame country of origin already living in the region, the higher the probability that morerecent migrants choose the same region. This supports the hypothesis that network andherd effects positively influence the migration of subsequent cohorts. However, adecreasing effect of network size on the probability of choosing a particular region can beobserved. Generally, for migrants from all countries the effect of ethnic networks21 The distance is measured as the crow flies between the capital of the migrants’ home countries and thelargest city within their region of residence in the EU-15.WIFO 27

ecomes negative at a network size of 29.7% (model 1, 40.1% according to model 7): ifmore than 29.7% of all migrants from a given country of origin already live in this region,it will become less attractive for following migrants with the same ethnic background.This indicates that there is an "optimum size” for migrant networks. If the network growsbeyond this size, there will be negative effects for both "newcomers" as well asestablished migrants, e.g. because of decreasing wage levels or increased housing andrental prices (Portnov, 1999), and migrants will no longer want to migrate to this region.For migrants from the NMS-8 (models 2 and 8), the optimal network size is also about30%, which is considerably larger than for migrants from Bulgaria and Romania, asshown by models (3) (9.0%) and (9) (15.4%). 222.5 SummaryThis chapter analysed the regional (NUTS2 level) concentration of migrants in the EU-15using recent data from the European Labour Force Survey. We observe the largest localclusters of migrants in the Île de France as well as Inner and Outer London andsubstantial differences in the settlement structure of foreigners and natives. Generally,23.9% of all migrants would have to change their region of residence (together with asimilar number of natives) in order to achieve a uniform distribution of migrants acrossEurope (and leave the relative population of the regions unchanged). This concentrationdoes not differ substantially between different cohorts: those who migrated during thelast 10 years are about as concentrated as those who migrated earlier. However, thetarget regions of more recent migration waves are considerably different from those ofearlier waves, indicating a shift in regional preferences of migrants.Comparing the regional concentration by individual characteristics shows that low skilledmigrants with primary education are much more concentrated than migrants withsecondary or tertiary education. This points to the importance of networks for low-skilledgroups, confirming results found, e.g., by Bartel (1989). However, high-skilled migrantswho moved during the last 10 years are today more concentrated than those who movedmore than 10 years ago, while the opposite is true for the medium- and low-skilled. Ourdata show a substantial change in target regions for all skill groups across time: morerecent migrant cohorts tend to prefer Spanish regions. This change is most pronounced,but not limited to, the low-skilled, while for highly skilled migrants a significantcorrelation between the local concentrations of recent and earlier migrant cohorts can stillbe observed. This indicates that regional preferences did not change as much for highskillmigrants as they did for the medium- and low-skilled.Analysing the concentration by country of origin, a smaller geographical association canbe observed for migrants from the 8 new CEE member states than, e.g., for migrantsfrom Romania and Bulgaria or the candidate countries. Migrants from the NMS-8 are thus22 The optimal network size computed from the other models is: model (4) 30.1%, model (5) 25.5%, model(6) 8.3%, model (10) 41.2%, model (11) 28.6%, model (12) 14.6%.WIFO 28

less concentrated in specific regions than Bulgarians or Romanians. The biggest localclusters of NMS migrants can be observed in the London areas and Vienna. Looking at theindividual countries, the largest concentration of migrants from the MNS-8 can beregistered among Estonians and Slovenians, while Polish migrants show the lowesttendency to cluster.Comparing the regional concentration across cohorts, an increase can be observed forrecent migrants from the NMS-8. At the same time, the correlation of local concentrationsacross cohorts is rather low and even insignificant for some CEE countries. The resultsthus suggest that some of the recent migration movements from the NMS-8 to the EU-15are different from the regional migration trends observed before. The opposite holds truefor migrants from Bulgaria and Romania, the candidate or other countries: regions whichhad a high local concentration of earlier migrants from one of these countries, alsotended to have a high local concentration of more recent migrants with the same ethnicbackground.However, these low or insignificant correlations for NMS-8 countries must not beinterpreted as disproving network or herd migration in Europe. They rather show thatinstitutional changes in the course of EU accession have led to a severe redistribution ofmigrant flows from the NMS-8. Institutional settings more favourable for migrationdecreased the costs of relocating into these regions, and by that also the opportunitycosts of not moving to where networks are. Initial movements “against the current” intonew regions without (or with only small) networks could, after some time, have creatednew (or led to the accumulation of larger) networks, attracting further in-migration anddrawing migrants away from “traditional” network hubs. This is supported by theobservation that there is still a significant correspondence in target regions across cohortsfor nationals of those countries for which the institutional setting has not changed, likeTurkey, Croatia, Macedonia or other countries. 23 This is confirmed by a conditional logitregression, which shows that networks do have a significantly positive effect on locationalchoice even after controlling for economic and geographic characteristics of the regions.This effect is, however, decreasing, which suggests an “optimal” network size formigrants.The results found in this chapter thus point on the one hand to significant changes in thestructure of migration from the NMS-8 to the EU-15 during the last years which can beattributed to changes in the institutional environment, especially EU accession. It canthus be expected that another shift in regional concentration of migrants from thesecountries will be observed after the end of the transitional period, because thestandardisation of freedom of movement regulations in the EU-15 after the end of thetransitional period contributes to a more even distribution of migrants from the NMS-8 tothe EU-15. On the other hand, the results point to significant network and herd effects in23 This applies, in principle, also to Bulgaria and Romania. Although the institutional situation has changed forthese countries with EU accession in 2007, they had only had one post-accession year at the time of theinterview. It can thus be expected that this had only a minor impact on regional concentrations of allBulgarian and Romanian migrants who moved during the last 10 years.WIFO 29

migration. These could increase the size of those networks which accumulated during thelast 10 years, which would pertain especially to Spanish regions and lead to a permanentchange in regional preferences of migrants. We thus observe two opposing effects, oneleading to a more even distribution of migrants across the EU-15, the other to increasedconcentration. How pronounced these effects will be and which effect will dominate is,however, ambiguous.WIFO 30

3 Cross-border Commuting in the EU273.1 IntroductionAs stressed in the last chapter one channel through which cross-border labour mobilitymay asymmetrically affect regional labour supply is through the spatial concentration ofmigration. Another one, however, is cross-border commuting. While the migrationchannel has been much analysed in recent literature, cross-border commuting has beenlargely ignored. To the best of our knowledge there is no single study to date whichanalyses the extent, structure and motivations for cross-border commuting from aEuropean perspective. By contrast the literature on commuting has either focused oncommuting choices within a country (see e.g. White, 1986, Hazans, 2003, Rouwendahl,1999, Van Ommeren, 1999) or on individual case studies of cross-border commuting orspecific border regions (see for example: Buch et al, 2008, Gottholmseder and Theurl ,2006, 2007, van der Velde, Jansen and van Houtum, 2005, Greve and Rydbjerg, 2003a,2003b, Bernotat and Snickars, 2002).This research suggests that commuting flows may differ from migration flows in anumber of important ways. For instance much of this literature indicates that commutingflows are much more dependent on distance between sending and receiving regions thanmigration flows. Since this is also to be expected from cross-border commuting flows,this implies a regionally asymmetric impact of cross-border commuting, which, - incontrast to migration flows, which as shown in the last chapter are often concentrated inthe urban centres of a country – on account of their high distance dependence may beexpected to be concentrated in border regions.There may, however, also be more subtle differences between migration and cross-bordercommuting flows. In this respect for instance White (1986) as well as Rouwendahl (1999)show that commuting within a country is strongly focused on males. This can beexplained by the higher alternative costs of travelling time for women, which arise onaccount of their role in childcare and household production, as well as a higher share ofpart time workers among women, which leads to higher commuting costs per work hour.Furthermore commuters within a country may differ from non-commuters with respect toage and education. Rouwendahl (1999) find that the willingness for mobility decreaseswith age, and Van Ommeren (1999), Hazans (2003) as well as Rouwendahl (1999) allfind that higher educated workers are more likely to be commuters within a country thanless educated workers.One could expect that some of these "stylized facts” carry over to cross-bordercommuters while others may differ. Indeed some of the recent case studies (see: Buch etal, 2008 and Gottholmseder and Theurl 2006, 2007) confirm that cross bordercommuters are mostly male but also suggest that cross-border commuters may differfrom commuters within a country both with respect to education and age structure.Comparing cross-border commuters from Vorarlberg to Switzerland to internalcommuters and non-commuters in the same region Gottholmseder and Theurl (2006)WIFO 31

find that on account of a low commuting share among the under 25 year olds there is noclear evidence that cross-border commuters are younger than non commuters, andaccording to the regression results in Gottholmseder and Theurl (2007) neither age noreducation is a significant determinant of cross-border commuting. This thus suggests thatwith respect to age and education the process of selection of cross-border commutersdiffers from that of internal commuters.In this chapter - before analysing the specific case of the so called CENTROPE region, -we are also interested in cross border commuting. In contrast to previous literature wehowever focus on the complete EU 27. Given the paucity of empirical results on theextent and structure of cross-border commuting for the EU 27, our aims are primarilydescriptive: In particular we first of all want to know how many people can be assumedto commute across borders in the EU 27 currently and how their demographic structurediffers from that of both commuters within a country and persons, who both live andwork within the same region (i.e. non-commuters). Second of all we want to know inwhich regions and countries of the EU 27 cross-border commuting currently plays animportant role and thirdly – with respect to labour mobility from the 12 new memberstates (NMS 12) to the 15 old member states (EU 15), - we want to know how both thestructure and extent of current cross-border commuting from the NMS12 (which are stillinfluenced by the transitional periods applied in a number of EU countries) differs fromcross-border commuting flows in the unregulated regime of the EU 15.3.2 Data and Extent of Cross-Border Commuting in the EU 273.2.1 Data IssuesThe data we use to address these issues are taken from the annual results of theEuropean Labour Force Survey for the years 2005 and 2006, which is a regularquestionnaire presented to a representative sample of households in all countries of theEU 27 (see: http://circa.europa.eu/irc/dsis/employment/info/data/eu_lfs/index.htm for apresentation of the questionnaire and its methodology). In this questionnaire personsthat were employed in paid employment for at least one hour in the week preceding theinterview are asked both for their place of residence as well as on their place of work.Furthermore, respondents are also interviewed on a number of demographic andworkplace characteristics (such as occupation and branch of employment, age, gender,highest completed education and others). Thus from this questionnaire it is possible tocalculate estimates of both the extent and structure of commuting within the EU 27.Furthermore this data can be analysed from the perspective of the sending regions (byanalysing outgoing commuter flows from a particular region), the receiving region (byanalysing incoming commuters to a particular region) or from a place to placeperspective (by analysing commuting between a particular pair of regions).While this data is thus well suited for our purposes, its analysis is subject to a number ofcaveats. The first of these arises with respect to the number of countries analysed. Not allof the national questionnaires in the EU 27 pose the question concerning the place ofWIFO 32

work. This applies to Greece, Portugal and Cyprus. Thus we have to exclude thesecountries from our analysis. Furthermore, the disaggregated data in the questionnaire forSlovenia grossly disaccords with the data provided in official data sources from Eurostat,so that in order to avoid data uncertainties we also exclude Slovenia from our sample ofcountries.Another caveat applies to missing data and non response problems. In our data in 20050.5% of the employed in the European did not respond to the question on place of work,in 2006 non response was at 0.1%. While these figures seem small, the rate of nonresponsein 2005 is of about the same magnitude as the extent of cross bordercommuting (see below). Given that respondents are more likely to be able to answer thequestion concerning the place of work when working in the same region as the region ofresidence; this may imply that commuting and in particular cross-border commuting maybe underestimated. In order to allow the reader an evaluation of this potentialunderreporting we thus also report the share of non respondents. Furthermore for ItalianData the share of non respondents exceeds 5% in each of the years considered and in theNetherlands for 2005 it exceeds 9%. Thus we also exclude Italian data from our analysisand use only data from 2006 when considering the structure of commuting flows, sincethis seems to be the most reliable.In addition the regional grid on which this data can be analysed from a receiving regionor place to place perspective differs from that on which an analysis from the sendingregion perspective can be conducted. Depending on the year analysed, between 40% and50% of the cross-border commuters provide information only on the country to whichthey commute but not on the NUTS 2-region. Thus analysis from a receiving region andplace to place perspective cannot be conducted on the same regional level as the sendingregion perspective (which is NUTS 2-level), but only on a national level.Furthermore in a number of cases the size of cross border commuting flows (in particularwhen disaggregating by region, age, industry or occupation) is well below the confidencebounds provided by Eurostat. Thus to avoid misinterpretation of the data, we follow therules of reporting suggested by Eurostat (seehttp://circa.europa.eu/irc/dsis/employment/info/data/eu_lfs/index.htm) by listing allfigures where high standard errors of the estimates may be expected in italics andsuppressing all numbers where commuting levels are below the lower confidence boundssuggested by EUROSTAT.Finally, the possibility of comparison of data across geographical entities (both nationaland regional level) is rather limited. Our data is available on the NUTS 2-level; thisimplies that commuting can only be measured if the workplace of a commuter is locatedin another NUTS 2-region than the place of residence. Thus any commuting within thesame NUTS 2-region is not registered. Since as shown in the last chapter NUTS 2 regionsdiffer vastly in terms of size and commuting is highly distance dependent, this impliesWIFO 33

that the extent of commuting measured in this study cannot easily be compared betweencountries and/or regions of a different size. 243.2.2 Some stylized facts on the extent of commuting in the EU 27 from thesending region perspectiveGiven these caveats we define as cross-border commuters all persons, who work inanother country than they live in. We compare these cross – border commuters to thosepersons, who live in the same NUTS 2-region as they work in (which are referred to asnon-commuters), as well as to the group of persons, who work in a different NUTS 2-region than they live in within the same country (internal commuters). 25Given thesedefinitions, the sum of the employed in these three groups is defined as the number ofemployed at place of residence while the sum of the non-commuters plus incomingcommuters (both from cross-border as well as internal commuting) gives the employedat the workplace. We use this number of employed at place of residence to normalizecommuting flows in table 3.1. This table in conjunction with table 3.2 and Figure 3.1shows a number of stylised facts concerning the extent of cross-border commuting in theEU.First according to LFS Data cross border commuting is a rather rare event in theEuropean Union. In the two years considered only around 0.6% of the employedcommuted across borders. This seems small relative to the approximately 7.4% of thepopulation that commuted across NUTS 2-regions within their respective countries. Thisalso seems to apply on a regional (NUTS 2) level. Among the 218 NUTS 2-regionsincluded in our sample the share of outward cross-border commuting in total employmentat the place of residence is higher than 5% only in 9 regions, which are the three Slovakregions, the French region of Alsace and Lorraine, the Provinces of Luxemburg andLimburg in Belgium and Freiburg in Germany as well as Vorarlberg in Austria. In a further31 regions this share is between 1% and 5% of the employed at the place of residence.For the vast majority of NUTS 2-regions (152), however, less than 0.5% of the residentemployed commutes across borders (see Figure 3.1)In addition, the extent commuting seems to have been relatively stable over the timeperiod considered. Although the absolute number of both cross-border and internalcommuters is slightly higher in 2006 than in 2005 the increase is of approximately thesame magnitude as the decrease in the number of non-respondents to this question inthe 2006 Labour Force Survey.24 This could be avoided if data on travelling time to the workplace were collected for all persons interviewed.Unfortunately this is not the case in the European Labour Force Survey.25 It should be noted that this definition does not impose any restrictions on the frequency of the commute.Thus we cannot separately identify daily, weekly and monthly commuters. This has the implication that fora small number of flows we find relatively distant commuting even to non-European destinations. Since theextent of these flows is rather small, we include them in our analysis.WIFO 34

Table 3.1: The Extent of Outbound Cross-border commuting by year and Country Group2005 2006 2005 2006Absolute (1,000)In % of employed at place ofresidenceCross border CommutersEU Total 1,036.1 1,170.2 0.6 0.6- of this EU 15* 723.1 792.8 0.5 0.6- of this NMS 10** 268.6 330.1 1.0 1.1- of this Bulgaria, Romania 44.4 47.3 0.4 0.4Internal CommutersEU Total 13,031.8 13,634.7 7.4 7.6- of this EU 15* 12,287.2 12,845.0 9.0 9.3- of this NMS 10** 655.6 692.6 2.3 2.4- of this Bulgaria, Romania 89.0 97.1 0.7 0.8Non RespondentsEU Total 862.1 124.5 0.5 0.1- of this EU 15* 860.9 121.8 0.6 0.1- of this NMS 10** 0.8 1.9 0.0 0.0- of this Bulgaria, Romania 0.4 0.8 0.0 0.0Notes: * excluding Greece, Portugal and Italy, **excluding Cyprus and SloveniaSource: Eurostat-LFS, WIFO-calculationsFurthermore, both internal and cross-border commuting is highly dependent on countries’geography. High rates of outbound cross-border commuting occur primarily in borderregions or in regions close to the border, while high rates of outbound internalcommuting are found primarily in the vicinity of large urban agglomerations of London,Berlin, Vienna, Prague and Stockholm, as well as in the smaller NUTS 2-regions of theBenelux countries 26 (see Figure 3.1). Thus small countries (such as Belgium, Austria andthe Baltic Countries), where most regions are located close to the border in general havehigher shares of outbound cross-border commuting than large countries (such as Spain).The major "hot spots” of cross-border commuting in the European Union seem to belocated at the German-French and French-Belgian borders, on the Austro-German border,at the Czech-Slovak border, in the Baltic countries and in Western Hungary as well as theGerman-Polish border and potentially southern Sweden. This suggests that cross bordercommuting in the EU occurs primarily between countries which either share a commonlanguage (e.g. France Belgium and Switzerland or Austria, Germany and Switzerland) orhave been a single country until very recently (such as in the case of the Czech Republicand Slovakia) or where special institutional arrangements influence the possibility of26 As already noted the smallness of these regions complicates a direct comparison of commuting shares withlarger regions. Since NUTS 2-regions differ vastly in terms of size and commuting is highly distancedependent, this implies that the extent of commuting measured is automatically higher in small regions.WIFO 35

cross-border commuting (as in the Austro-Hungarian case). 27 By contrast most otherborder regions are characterised by rather low cross-border commuting rates. In theregions outside of these "hot spots” the share of out commuting cross-border commutersis lower than 0.5% of the resident workforce even when considering only border regions(see: Figure 3.1).Aside from size and geography, however, also other factors seem to be important for theextent of cross border commuting, since there is a large variation in outbound commutingamong countries of similar size. In particular in general outbound commuting tends to behigher in regions with lower GDP per capita levels and lower unemployment rates (seesection 3.4 for further details) and there seems to be a core-periphery pattern in bothcross-border and internal commuting. Regions which may be considered to be locatedmore in the centre of the EU such as for example the regions of Austria, Belgium,Germany, Netherlands and others in general tend to have higher internal and crossbordercommuting rates, while regions which are located more in the periphery (e.g.Spain, Bulgaria, Romania) tend to have low commuting rates. 28Finally, we also find that, the share of cross-border commuters is somewhat higher in theNMS 10 than in the EU 15. We would have expected that the opposite is the case onaccount of both the shorter time span the NMS 10 have integrated in the EU and becausefor important receiving countries there are still institutional barriers for cross-bordercommuters. These high cross-border commuting rates in the NMS 10 may, however, bedistorted by both the high share of cross-border commuters from Slovakia, which resultsfrom a high share of commuters from Slovakia to the Czech Republic with which it formeda single country until 2002, and a large number of small countries among the NMS 10,which distorts cross border commuting flows upwards (see Figure 3.1). By contrast, -accordance with expectations - cross-border commuting shares in Bulgaria and Romaniaare lower than in the EU 15.27 Cross-border commuting at the Austro-Hungarian border was already substantially liberalised in the 1990s,by providing a special commuting quota to Hungarian commuters.28 These comparisons are, however, also influenced by relative region size, which is larger for the moreperipheral regions and leads to a downward bias for commuting flows in these regions.WIFO 36

PortugalGibraltarP ortugalÍslandGibraltarIrelandÍslandEspañaIr elandFærørneIsle of ManEspañaChannel IslandsUnited KingdomFæ rørneIsle of ManAndor raChannel IslandsFranceUnited KingdomAndor raFr anceNederlandBelgique-BelgiëLuxembourg (Gr and-Duché)MonacoDanmar kLiechtensteinSchweiz/Suisse/SvizzeraNederlandBelgique-BelgiëLuxem bour g (Grand-Duché)MonacoNor geDeutschlandNor geDanm arkLiechtensteinSchweiz/Suisse/S vizzeraDeutschlandSan MarinoItaliaSverigeSan MarinoItaliaCeska RepublikaÖsterr eichSverigeSlovenijaMaltaCeska RepublikaÖsterr eichSlovenijaMaltaHrvatskaPolskaBosna i HercegovinaHr vatskaSlovenska RepublikaMagyarorszagPolskaMontenegroSlovenska RepublikaMagyar orszagBosna i HercegovinaMontenegroSuomi / FinlandLietuvaS er biaEestiShqiper iaSuomi / FinlandLietuvaSerbiaE estiShqiper iaLatvijaMakedonijaLatvijaElladaMakedonijaElladaBelar us'RomaniaBelarus'Rom aniaBulgariaBulgariaMoldovaMoldovaUkraineUkraineRossijaRossijaKypros / KibrisTurkiyeKypros / KibrisTurkiyeGr uzijaFigure 3.1: The Extent of Outbound Cross-border commuting in EU 27 NUTS 2-regions(2006)Internal CommuteNUTS2 Europa 2006= 2.50= 5.00= 10.00= 20.00= 50.00Cross-border CommutingNUTS2 Europa 2006= 0.20= 0.50= 1.00= 5.00= 30.00Notes: Figure shows commuting rates in % of the employed at the place of residence. Top panel =internal commuters, bottom panel =external commutersSource: Eurostat LFS.WIFO 37

Table 3.2: The Extent of Outbound Cross-border commuting by EU Countries (2006)Absolute (thousands)Share in percentInternalCommutersCross-BorderCommutersNonRespondentsInternalCommutersCross-BorderCommutersNonRespondentsAbsolute ( thousands)PercentAustria 397.9 39.7 0.9 10.1 1.0 0.0Belgium 828.3 95.0 0.2 19.4 2.2 0.0Germany 3,846.5 173.2 56.1 10.3 0.5 0.2Denmark 0.0 5.5 27.0 0.0 0.2 1.0Spain 382.7 55.6 0.0 1.9 0.3 0.0Finnland 66.9 3.0 0.0 2.7 - 0.0France 1,468.9 279.0 19.9 5.9 1.1 0.1Ireland 264.8 0.0 8.8 13.1 0.0 0.4Luxemburg 0.0 1.7 0.0 0.0 0.9 0.0Netherlands 1,056.2 32.4 4.4 12.9 0.4 0.1Sweden 195.7 38.3 3.1 4.4 0.9 0.1U.K. 4,337.0 69.4 1.5 15.4 0.2 0.0Czech Republic 230.7 25.1 0.1 4.8 0.5 0.0Estonia 0.0 10.7 0.0 0.0 1.7 0.0Hungary 147.5 24.9 0.0 3.8 0.6 0.0Lituania 0.0 26.2 0.0 0.0 1.7 0.0Latvia 0.0 14.3 0.1 0.0 1.3 0.0Malta 0.0 - 0.1 0.0 - 0.0Poland 216.3 71.6 0.2 1.5 0.5 0.0Slovakia 98.1 156.8 1.4 4.3 6.8 0.1Bulgaria 39.2 10.3 0.8 1.3 0.3 0.0Romania 57.9 36.9 0.0 0.6 0.4 0.0Notes: Figures in bold italics=unreliable data due to few observations, - = no data reported on account of thelow number of observationsSource: EUROSTAT-LFS, own calculationsTable 3.3: The Extent of Inward Cross-border Commuting from the EU 27 by Countrygroups and yearAbsolute (thousands)Share in percentIn percent of employment atworkplace2005 2006 2005 2006 2005 2006EU 15 652.2 722.7 62.9 61.8 0.5 0.6NMS 12 99.4 115.7 9.6 9.9 0.3 0.3Other countries 284.5 331.7 27.5 28.4 n.a. n.a.Source: EUROSTAT-LFS, own calculations, n.a.=not availableWIFO 38

3.2.3 Stylized facts from the receiving region and place to place perspectiveWhen considering commuting flows from the receiving region perspective (see Table 3.3and Figure 3.1) the total share of incoming cross-border commuters from the EU 27 inthe total number of employed at the workplace is at about 0,5% or lower and may thusalso be considered small. Among the individual EU 15 receiving countries apart from theobvious outlier of Luxemburg (where over a third of the employed in that region arecommuters from other countries) only Belgium, Ireland, Austria and the Netherlandsreceive a share of cross-border commuters from other EU 27 countries in excess of 1% ofthe employed at the workplace. For the NMS 10 and NMS 2 inward cross-bordercommuting flows are of an even lower relevance. Among these countries the share ofinward cross-border commuters in total employment exceeds the 1% mark only in theCzech Republic (on account of the large number of commuters from Slovakia), and the0.5% mark in Hungary. For all other NMS 12 countries the size of inward commutingflows from other EU 27 countries may be considered negligible.Figure 3.2: The Extent of Inward Cross-border Commuting from the EU 27 by EU 27Country (in % of employment at the workplace)40.035.036.430.025.020.015.010.05.00.01.5 1.8 1.70.0 0.5 0.7 0.1 0.1 0.4 0.1 0.5 1.6 0.0 0.1 0.1 1.1 0.0 0.0 0.2 0.1 0.3AT BE BG CZ DE DK EE ES FI FR HU IE LT LU LV MT NL PL RO SE SK UKSource: Eurostat LFS, own calculationsFurthermore when considering place to place cross border commuting flows (see table3.3) a clear differentiation between the EU 15 and the NMS 12 emerges. Most of theoutbound cross-border commuting from the EU 15 countries is either with other EU 15countries (these flows account for more than 90% of outbound cross-border commutingin Belgium, Luxemburg and the Netherlands) or with other non EU 27 countries (theseflows are particularly important for Denmark, and the UK where more than 50% of theoutgoing commuting flows go to non-EU countries). By contrast commuting patterns fromWIFO 39

the NMS 12 (with the exception of Slovakia where more than 60% of commuting flowsare with other EU 12 countries and Malta, where commuting flows are to low to representreliable estimates) are much more strongly focused on the EU 15. In all of the NMS 12countries (with the two mentioned exceptions) more than 70% of all cross bordercommuting flows go to EU 15 countries.Table 3.4: Place to Place Cross-border Commuting by Country Groups and YearSending RegionReceiving Region EU 15 NMS 12 Other countries TotalYear 2006Absolute (thousands)EU 15 456.3 7.0 259.8 723.1NMS 12 195.9 92.5 24.7 313.0Share in percentEU 15 63.1 1.0 35.9 100.0NMS 12 62.6 29.5 7.9 100.0Year 2005Absolute (thousands)EU 15 479.7 10.6 302.5 792.8NMS 12 234.9 104.2 28.0 367.1Share in percentEU 15 60.5 1.3 38.2 100.0NMS 12 64.0 28.4 7.6 100.0Source: EUROSTAT-LFS, own calculationsThus while there seem to be few differences between commuting flows from the NMS 12and the EU 15 when considering the sending regions only, there are some significantdifferences from a receiving country perspective as well as from the place to placeperspective. In particular when considering inbound cross-border commuting we find thatas a percentage of the employed in the country of work, NMS 12 countries receive muchfewer cross border commuters than EU 15 countries. In addition when consideringoutbound cross-border commuting we find that cross border commuting from the NMS 12is strongly oriented towards the EU 15 countries rather than non-EU or other EU12countries. This can be explained by the fact that most non-EU countries close enough tothe NMS 12 to be destinations for cross-border commuting have substantially lowerincome levels than the NMS 12. By contrast outbound cross-border commuting in theEU 15 is more strongly oriented to non-EU countries. Again this can be explained bydifferences in income levels.WIFO 40

3.3 The Structure of Commuting Flows3.3.1 Comparing cross-border commuters, non commuters and internalcommutersThe descriptive evidence provided so far thus suggests that the extent of cross-bordercommuting in the EU in general is limited to individual border regions and has a relativelylimited magnitude when considering the overall European labour market. In the twoyears observed cross-border commuters accounted for only 0.5% of total employment inthe EU. In particular cross-border commuting is of relevance only in a limited number ofborder regions, which are mostly characterised by strong linguistic, historic orinstitutional ties, only.Aside from the extent of commuting, we are, however, also interested in thedemographic, and occupational structure of cross-border commuters. Thus table 3.5presents some descriptive evidence on the demographic structure of cross-bordercommuters in comparison both to non-commuters and internal commuters. As can beseen from this table cross-border commuters differ most significantly from noncommutersby the high share of male cross-border commuters, a disproportionately highshare of persons in the age group between 20 and 29 and a stronger focus onintermediate (secondary level) educated workers. Furthermore, cross border commutersoften work in less skilled occupations such as elementary occupations, plant and machineoperators or as crafts and related trade workers as well as in the construction andmanufacturing sector. Thus in contrast to non-commuters, cross border commuters aredisproportionately often medium skilled male manufacturing and construction workers,which also work in medium to less qualified manufacturing jobs.This qualification profile also carries over to the comparison with internal commuters.Relative to cross-border commuters, internal commuters are clearly more highlyqualified. 36% of the internal commuters but only 26% of the cross-border commuters inthe EU have completed a tertiary education. In addition internal commuters are muchmore strongly concentrated in service sector employment (internal commuters 70%,cross-border commuter 53%) and typically work as legislators, professionals ortechnicians (internal commuters 50%, cross-border commuters 42%). This suggests thatwhile internal -commuters are a clearly positively selected group among the employed,cross-border commuters, by contrast, seem to be primarily selected from medium skilland manufacturing workers. Furthermore, selectivity by age and gender also seems to bestronger when considering cross-border commuters than internal commuters. The shareof males among cross-border commuters is substantially higher than among internalcommuters and cross-border commuters are also much more strongly focused on the agegroup of the 20 to 29 year olds.These results are in accordance with some of the findings of previous studies. Forinstance Buch et al (2008) finds that cross border workers in the German – Danish borderregions are disproportionately drawn from among manufacturing workers and mostlystem form the age group of the over 25 year olds and Gottholmseder and Theurl (2005WIFO 41

and 2006), focusing on cross-border commuters from Vorarlberg to Switzerland, find thatthe (relative) majority of them are male, 25-to 35 years old, medium skilledmanufacturing workers, that are neither positively nor negatively selected on educationalgrounds.Table 3.4: Commuting flows in the EU27 by Demographic and Job Characteristics (in % oftotal flows, 2006)NoncommutersInternalcommutersCross-bordercommutersNo responseEmploymentat place ofresidenceFemale 46.1 36.5 28.3 - 45.2Male 53.9 63.5 71.7 67.0 54.8Aged 20 to 29 19.4 21.6 27.50 23.60 19.6Aged 30 to 39 26.0 27.8 26.80 32.00 26.1Aged 40 to 49 26.8 26.5 26.50 22.40 26.8Aged 50 to 59 19.9 18.1 15.30 11.40 19.7Aged 60 or more 5.0 3.4 2.40 6.50 4.9Not Available 0.2 0.3 - - 0.2Primary 4.5 2.4 2.50 3.70 4.4Secondary 68.7 61.3 71.30 62.30 68.1Tertiary 26.6 36.0 25.90 33.80 27.3Legislators senior officials and managers 7.9 13.1 6.6 9.2 8.3Professionals 13.8 18.1 12.9 20.5 14.2Technicians and associate professionals 15.6 18.5 13.8 14.5 15.8Clerks 10.7 11.5 6.8 - 10.7Service workers and shop and market sale 14.0 9.8 11.4 11.3 13.6Skilled agricultural and fishery workers 4.9 1.1 1.4 - 4.6Craft and related trades workers 13.6 11.8 23.8 17.5 13.6Plant and machine operators and assemble 8.7 7.6 12.9 - 8.7Elementary occupations 9.9 6.2 9.3 7.8 9.7Armed forces 0.5 1.3 0.9 - 0.5No answer 0.4 1.1 0.2 - 0.4Aggriculture 6.2 1.4 3.5 - 5.8Construction 7.6 8.3 17.3 14.4 7.7Manufacturing 19.3 19.7 25.6 18.4 19.4Market Services 36.2 41.3 35.7 36.8 36.6Non-Market Services 30.5 28.7 17.0 23.2 30.3Non-Response 0.2 0.7 0.9 - 0.3Notes: - = no data reported on account of the low number of observations , column sums for individualcharacteristics are 100%Source: EUROSTAT-LFS, own calculations3.3.2 Differences between NMS 12 and EU 15 FlowsThese distinct differences in the demographic, educational and occupationalcharacteristics of cross-border commuters both relative to internal-commuters andstayers, can be expected to arise from a number of factors that may be considered to bespecific either to the receiving region such as the industrial structure and thus thestructure of labour demand in border regions (which may be more strongly focused onWIFO 42

manufacturing activities than that of urban centres, which are the basin for attraction forinternal commuting flows), or to the particular sending and receiving region pairconsidered (such as differences in the returns to education in sending and receivingregions). While a detailed analysis of the structure of cross-border commuters relative tointernal commuters on a place to place basis is beyond the scope of this study, due to thementioned data problems, we were interested to what degree the structure of crossbordercommuting flows differs between different EU 15 countries and between EU 15and NMS 12 countries, since this is a particularly interesting case, on account of the factthat cross-border commuting with the most relevant receiving countries is still regulatedfor flows between the NMS 12 and the EU 15, while it is unregulated within the EU 15countries.Table 3.5: Commuting flows in the EU27 by Demographic and Job Characteristics andReceiving Region and Place to Place criteria (in % of total flows, 2006)Recieving RegionPlace to Place FlowsEU 15 EU 12 Other From EU 15to EU 15From EU 12to EU 15OtherFemale 29.7 24.2 26.7 32.2 31.7 28.6Male 70.3 75.8 73.3 67.8 68.3 71.4Aged 15 to 19 1.7 - - - - -Aged 20 to 29 29.3 37.6 20.0 20.8 42.3 24.1Aged 30 to 39 26.0 25.1 29.0 31.5 27.1 27.9Aged 40 to 49 26.8 21.8 27.4 28.4 19.6 27.3Aged 50 to 59 14.4 12.5 18.3 15.6 8.1 16.6Aged 60 or more 1.7 - - 2.4 0.8 2.6Not Available 2.3 - 3.2 - - -Primary 71.7 85.2 65.7 62.5 87.0 69.6Secondary 25.6 12.9 31.0 30.9 11.9 27.9Tertiary7.2 - - 8.8 - 7.7Legislators senior officials and managers 11.6 7.2 17.9 17.1 - 13.7Professionals 12.2 12.5 17.7 15.2 6.2 16.7Technicians and associate professionals 7.2 2.2 7.7 10.0 - 5.9Clerks 12.5 4.8 11.2 9.8 18.0 8.5Service workers and shop and market sale 1.7 - - - 5.7 -Skilled agricultural and fishery workers 22.5 35.6 22.5 15.8 28.8 23.4Craft and related trades workers 12.7 21.8 10.2 15.8 9.6 12.9Plant and machine operators and assemble 11.6 11.2 - 5.4 23.6 8.5Elementary occupations - - - - - -Armed forcesNo answer 4.4 - 2.4 - 11.4 2.4Construction 16.2 32.6 14.6 9.3 28.1 17.7Manufacturing 23.0 42.9 25.1 30.0 14.3 31.2Market Services 38.8 18.2 35.0 38.1 28.7 30.2Non-Market Services 17.0 - 21.0 19.9 17.3 17.5in %Notes:- = no data reported on account of the low number of observationsSource: EUROSTAT-LFS, own calculationsTo perform this analysis in a first step we separated migration flows by receiving countryinto migration flows received by the EU 15 countries, those received by NMS 12 countriesWIFO 43

and those received by other countries outside the EU 29 (see left hand side panel of table2.6). Here we hypothesized that (given the persisting differences in industrial structurebetween the EU 15 and the NMS 12, 30 and the fact that most of the cross-bordercommuting flows to the NMS 12 mostly come from other NMS 12 countries and not fromthe EU 15) cross-border commuting should be particularly relevant for the group ofyoung, male manufacturing workers with intermediate education levels in the NMS 12.As can be seen from the results (reported in the left hand side panel of table 3.6) thishypothesis is confirmed. The share of cross-border commuters with completed secondaryeducation, working in manufacturing or in elementary occupations or employed as plantand machine operators or crafts, is particularly high among those workers that commute(from one of the EU 27 countries) to one of the NMS 12 countries. At the same time,however, the cross-border commuters that commute to the EU 15 are also more stronglyconcentrated in these education, occupation and industry groups, than either noncommutersor internal commuters. A clearly better than average occupational,educational and industrial structure can be found among those who commute from anEU 27 country to a non-EU country.In a second step we also divided commuting flows by place to place categories. Here wefocused on all cross-border commuting flows from an EU 15 country to another EU 15country, cross-border flows from an NMS 12 to an EU 15 country and all other commutingflows (see right hand side panel of table 3.6). 31 We find somewhat more pronounceddifferences between flows between different EU 15 countries and flows from the NMS 12to EU 15 countries. In particular cross-border commuters within the EU 15 tend to besubstantially better educated than commuters from the NMS 12 to the EU 15. A largershare of them also works in market services and in occupations such as legislators,professionals and technicians. Indeed when comparing the structure of cross-bordercommuters within the EU 15 (in table 3.6) to that of noncommuters (in table 3.6) we findthat aside from the focus on males and younger workers, cross-border commuters withinthe EU 15 do not differ very strongly from non-commuters. By contrast commuters fromthe NMS 12 to the EU 15 are much more strongly focused on the secondarily educated onconstruction.3.4 The Determinants of Out-Commuting FlowsThese results suggest that cross-border commuting flows are small in the EU, but thatthere is also some variance among regions. Furthermore results on the structure ofcommuting flows suggest that cross-border commuters - in contrast to internal29 Unfortunately due to the low number of cross-border commuters to the candidate countries from the EU.This group cannot be separately analysed in this case.30 As shown by a number of studies (e.g. Huber 2008) the employment structure of the NMS 12 is still muchmore strongly focused on manufacturing and medium skill level workers than that of the EU 1531 This choice was primarily motivated by our aim to obtain magnitudes of flows that can still be considered tobe representative.WIFO 44

commuters - in the EU 27 are in general not better qualified than non-commuters and aredrawn more than proportionally from manufacturing workers, males and the age group ofthe 20 to 29 year olds. These characteristics apply even more strongly to the crossbordercommuters from and to the NMS 12 than to commuters from and to the EU 15.While these results are largely consistent with the findings from earlier case studies inthe literature, they also suggest that cross border commuters – in contrast to migrants –are not as strongly positively selected on education but stem primarily from theintermediate qualification level of the educational spectrum.In particular, the heterogeneity of cross-border commuting flows found across differentsending regions, raises the question of the causes for these differences. This section thusaddresses this question by means of a regression analysis. 32 We focus on explaining theextent of outgoing cross-border commuting on a place to place basis. Furthermore thisanalysis will be used to discuss to what degree cross-border commuting flows from theNMS 12 to the EU 15 (as well as in the opposite direction) are lower than cross-bordercommuting flows within the EU 15 on account of existing institutional barriers to mobility.The dependent variable in this regression is the share of commuters (in total employmentat the place of residence) moving from one of the 218 NUTS 2-regions of the EU 27countries 33 considered in this paper to one of 31 European receiving countries in 2006 34 .This variable is regressed on a number of characteristics of the receiving and sendingregion. In particular, we use (the logged) differences in the unemployment rate and percapita GDP at exchange rates 35 between the sending region and the receiving countrysince we expect that commuters will predominantly move from low income, highunemployment regions to high income and low unemployment regions. Furthermore wealso use the distance between the sending regions’ and the receiving country’s capital aswell as dummies for commuting flows that occur across neighbouring countries andbetween countries that share the same language as proxies for commuting costs acrossregions. Given our descriptive results we expect that distance should have a negativeimpact on commuting flows, while neighbouring countries and regions sharing the samelanguage should experience higher place to place moves. Furthermore, since ourdescriptive evidence also suggests that moves from the Czech Republic to Slovakia are32 An econometric problem that arises in the estimation of the regression model is that we do not observe anycommuting for 87% of all possible sending region – receiving country observations in our data. Thus ourdata has a very large number of zero flows, which leads to biased results if standard least squarestechniques are used. One of the possible alternatives suggested by Nowotny (2007) is to estimate themodel by means of a censored regression model, such as the tobit model. This model endogenouslyhandles the fact that observations in the data cannot become smaller than 0 and thus provides a consistentestimate of the sample parameters. We apply this method here.33 In principle we would prefer to estimate the model on the basis of region to region flows rather than onregion to country flows. This is, however, precluded on account of data quality issues discussed in the firstsection of this chapter.34 Apart from the EU27 countries we also consider Switzerland, Norway, Turkey and Croatia as potentialreceiving countries, on account of the relatively high share of commuters going to these regions and theirvicinity to the EU27.35 We prefer using the GDP at exchange rates rather than the GDP at purchasing power parity which is oftenused in the migration literature, because in contrast to migrants, commuters tend to consume at the placeof residence and can thus change the income generated at the place of work at official exchange rates.WIFO 45

exceptionally high for historic reasons, we also include a dummy for commuters betweenthe Czech and the Slovak Republics, which we expect to have a positive impact oncommuting flows.In addition we include a number of sending region characteristics. These aside from afamily of sending country dummies (which are intended to capture potential differencesin institutions across sending countries) include the (log of the) share of manufacturing,service and construction employment in the sending region and the share of internalcommuters commuting from the region. These are included in the regression because thedescriptive analysis suggests that cross-border commuters are strongly concentrated inconstruction and manufacturing and because the presence of a strong centre of attractionfor commuters within a country in the vicinity of a region (as would be indicated by highinternal commuting shares) may reduce the share of cross-border commuters. We thusexpect that the coefficient of the internal commuting variable should be negative.Finally, we also include a series of dummy variables for whether the cross-bordermigration flow is from an EU 15 to a NMS 12 country, from the NMS 12 to the EU 15,from EU 15 to non-EU countries and NMS 12 to non EU countries. These variables are thevariables of interest when asking whether current NMS 12 to EU 15 cross-bordercommuter flows are lower than flows within the EU 15 on account of institutionalregulations. Since for these variables flows from EU 15 countries to other EU 15 countriesare the base category, a statistically negative coefficient on the NMS 12 to EU 15 dummyvariable implies that these migration flows are significantly lower than flows acrossborder, while an insignificant coefficient would imply that the hypothesis that these twoflows have already assumed similar magnitudes can be rejected.The results of these estimates (in Table 3.6) suggest that cross-border commuting isindeed significantly associated with differences in GDP per capita between the sendingand the receiving region. The coefficient implies that for a region where some commutingis observed a 1% increase in the difference in the GDP per capita between the sendingand receiving country increases bilateral cross border commuting flows by 0.25percentage points. By contrast the differences in the unemployment rates between thesending and receiving has an insignificant impact on bilateral commuting flows, whiledistance – as expected has a significant negative impact.Aside from this we also find that - as expected - two regions located in neighbouringcountries (given a positive commuting flow) may expect an by 0.9 percentage pointhigher commuting share than regions located in non-neighbouring countries and thatcross-border commuting flows between two regions in different countries that share thesame language or are located in the Czech and Slovak Republic are all else equal by 0,5percentage point and 1.5 points, respectively, higher than cross-border commuting flowsbetween other regions, which reconfirms much of our descriptive analysis. Similarly ahigher share in both manufacturing and service sector employment leads to higher crossborderout commuting shares.WIFO 46

Table 3.6: Regression results for place to place out migration sharesEU 27 EU 15 EU 12Coefficient Std. Error Coefficient Std. Error Coefficient Std. ErrorLn gdp difference -0.00252 ** 0.00123 0.00254 * 0.00149 0.00119 *** 0.00037Ln unemployment rate difference 0.00056 0.00063 0.00075 0.00101 0.00010 0.00036Ln distance -0.00085 *** 0.00036 -0.00180 *** 0.00067 -0.00035 **** 0.00010Share of internal commuters -0.00672 0.00541 -0.00633 0.00648 -0.01071 0.00737Manufacturing employment share 0.00021 *** 0.00007 0.00014 0.00032 0.00006 ** 0.00003Construction employment share 0.00033 0.00024 0.00003 0.00042 0.00057 ** 0.00023Service employment share 0.00020 ** 0.00008 0.00017 0.00026 0.00005 ** 0.00002Neighboring region 0.00913 *** 0.00310 0.01166 *** 0.00427 0.00398 *** 0.00079Same language 0.00531 *** 0.00130 0.00665 *** 0.00175Czech/Slovak Republic 0.01464 *** 0.00564 0.01143 ** 0.00543EU 15 to EU 12 -0.00468 *** 0.00158 -0.00682 0.00236EU 12 to EU 15 -0.00205 0.00264 0.00303 *** 0.00076EU 15 to non-EU -0.00176 0.00115 -0.00250 0.00149EU 12 to EU 12 -0.00932 ** 0.00377EU 12 to non EU -0.00614 ** 0.00300 0.00114 * 0.00064Log Likelyhood 1,690.20 744.04 1,184.66Observations 6,600 5,010 1,590Notes: Dependent Variable: Share of Region to Country flows in employed at region of residence. Table reports coefficients of a tobitregression analysis of bilateral commuting flows, *** (**) (*) signifies significance at the 1% (5%( (10%) level respectively. Coefficient=coefficient estimates S.E: = standard ErrorSource: LAMO household surveys 2004/05 and 2006/07, WIFO-calculationsFinally, and probably most interestingly in terms of this study, we find that while – aftercontrolling for other factors influencing cross-border commuting - flows from both EU 15and NMS 12 countries to non EU countries are significantly lower than cross-bordercommuting flows within the EU 15, there is no significant difference between EU 15 toEU 15 and NMS 12 to EU 15 flows any more. This implies that our regression resultscannot reject the hypothesis that the current number of cross-border commuters fromthe NMS 12 to the EU 15 (after controlling for other factors impacting on cross-bordercommuting) are of the same magnitude as could be expected in an unregulated regimeas applies to the EU 15 today. This thus questions forecasts that argue strongly for adramatic increase in cross border commuting after the end of derogation periods on thefreedom of movement of labour. Our results, however, also suggest that even aftercontrolling for GDP per capita differences and other factors important for cross bordercommuting the flows from the EU 15 to the NMS 12 are substantially lower than whatmay be expected from within EU 15 cross-border commuting flows. Here the estimatedcoefficient suggests that cross-border commuting flows in this direction are by about 0.5percentage points too low relative to within EU 15 levels.While thus relative to the benchmark of within EU 15 flows the NMS 12 to EU 15 crossbordercommuting flows do not seem to be significantly lower, Columns 2 and 3 of table3.6 take this analysis of the differences in the determinants of cross-border commuting alittle bit further, by estimating the regression separately for EU 15 and NMS 12commuting flows. Here we find that these differences are relatively mild. They apply onlyWIFO 47

to the role of income differentials (which are more strongly associated with cross bordercommuting flows but with a smaller marginal effect in the NMS 12) and to the role ofsectoral specialisation. Higher shares of manufacturing, construction and services in thesending region increase cross-border commuting only in the NMS 12.3.5 ConclusionsThis chapter describes extent and structure of cross border commuting in the EU 27 inorder to address a number of questions:First of all, we were interested in the extent of cross-border commuting in the EU 27. Wefind that this in general is limited to individual border regions and has a relatively lowmagnitude when considering the overall European labour market. In the two yearsobserved cross-border commuters accounted for only 0.5% of total employment in theEU. In particular cross-border commuting is of relevance in a small number of borderregions, located at the external border of the EU, the German-French and French Belgianborders, on the Austro-German border, at the Czech-Slovak border, in the Balticcountries and in Western Hungary as well as the German Polish border and potentiallysouthern Sweden, which are mostly characterised by strong linguistic, historic orinstitutional ties, only. In these regions usually more than 1% of the employed commuteacross borders and may surpass the 5% mark in exceptional cases. For most other borderregions outside these "hot spots” out-commuting is below 0.5% of the employed. In sumthus results on the extent of commuting suggest that cross-border commuting flows aresmall in the EU, but that there is also some variance among regions.There are also some differences in the importance of cross-border commuting betweenthe EU 15 and NMS 12. In particular, when considering inbound cross-border commutingwe find that, as a percentage of the employed in the country of work, NMS 12 countriesreceive much fewer cross-border commuters than EU 15 countries. In addition whenconsidering outbound cross-border commuting we find that cross-border commuting fromthe NMS 12 is strongly oriented towards the EU 15 countries rather than non-EUcountries. This can be explained by the fact that most non-EU countries that are closeenough to the NMS 12 to be destinations for cross-border commuting have substantiallylower income levels than the NMS 12. By contrast, outbound cross-border commuting inthe EU 15 is more strongly oriented to non-EU countries rather than to the NMS 12.Again, this can be explained by the differences in income levels to nearby non-EUcountries.Second of all we were interested in the structure of commuting flows in the EU withrespect to gender, age, occupations and education of cross-border commuters. Ourresults suggest that cross-border commuters - in contrast to internal commuters in theEU 27 - are not in general better qualified than non-commuter and are drawn more thanproportionately from manufacturing workers, males and the age group of the 20 to 29year olds. Furthermore, these characteristics apply even more strongly to cross-bordercommuters from the NMS 12 than to commuters from the EU 15. While these results areWIFO 48

largely consistent with the findings of earlier case studies in the literature, they alsosuggest that cross border commuters – in contrast to migrants – are not as stronglypositively selected on educational criteria, but stem primarily from the intermediatequalification level.Finally, we were interested to what degree cross border commuting flows from the NMS12 to the EU 15 may be considered as too low relative to cross-border commuting amongEU 15 countries. While our results in this respect are subject to a rather unsatisfactorydata situation, our finding suggest that - after controlling for other influences on crossbordercommuting - flows from the NMS 12 to the EU 15 are not significantly smallerthan those among the EU 15 countries, while flows from the EU 15 to the NMS 12 aresignificantly lower than those among the EU 15. The primary difference in the factorsdetermining cross-border migration in the NMS 12 and the EU 15 seems to be a closerassociation of cross-border commuting with the industrial specialisation in the NMS 12than the EU 15.WIFO 49

4 The CENTROPE Region: Economic BackgroundThe findings of this study so far thus indicate that the regional impact of cross-borderlabour mobility may be particularly strongly felt in urban agglomerations (due to theconcentration of migrants in these regions) and border regions (on account of crossborder commuting). For the remainder of the study we thus focus on the Austrian-Hungarian-Czech-Slovak border region, which has been considered a primary example ofa border regions that may be strongly affected by cross-border labour mobility afterenlargement.In terms of NUTS 3 regions this so called "CENTROPE" region covers the Austrianprovinces of Burgenland, Lower Austria, and Vienna, the Czech region of South Moravia,Trnava and Bratislava in Slovakia as well as the Hungarian counties of Gyõr-Moson-Sopron and Vas. Its territory measures over 44,000 square kilometres and it has apopulation of around 6.5 million inhabitants. In some cases, however, EUROSTAT sourcesdo not provide NUTS3 level data. Thus in this chapter we sometimes also will use NUTS 2level data. Here, CENTROPE covers the Austrian provinces of Burgenland, Lower Austriaand Vienna; the Czech South East planning region; Bratislava and Western Slovakia inSlovakia as well as Western Transdanubia in Hungary (see Figure 4.1). When operatingon this NUTS 2 level data CENTROPE thus covers a territory of 66,000 square kilometresand has 8.5 million inhabitants.Figure 4.1:The CENTROPE RegionSouth EastCZ062Lower AustriaWestern SlovakiaBratislavaViennaBurgenlandWest TransdanubiaAT125AT124AT126 SK010 SK021AT130AT123AT127AT121AT112AT122HU221AT111AT113HU222Source: Regiograph, WIFONUTS 2 level definitionNuts 3 level definitionIrrespective of the data used, this border region is considered one of the most importanttransnational economic areas at the former Eastern borders of the European Union.Located at the intersection of four countries, comprising two capital cities (Vienna andBratislava) as well as several further major cities (Brno and Gyır) and covering some ofthe most dynamic regions in the Central and East European countries as well as some ofthe most prosperous regions within the EU (Vienna), CENTROPE is considered a region ofconsiderable economic potential and a region that could potentially experienceWIFO 50

substantial cross-border labour mobility after accession, on account of its vicinity to theborder and high urbanisation.Figure 4.2: Market Potential in the European NUTS 2 RegionsMarket Potential 2002

urbanisation. While CENTROPE is in many respects closely linked to the economies of the"twin – capitals" of Vienna and Bratislava, it is not a typical central region in theEuropean context. Its settlement structure on average is not governed by large cities.Much rather – as in much of Central Europe - medium sized towns dominate.Figure 4.3: Growth of Market Potential in the European NUTS 2 RegionsGrowth of Market Potential1995-2002

located further to the East. These regions have also grown more rapidly than the EUaverage in the last decade. Thus CENTROPE is also located very close to those regions inthe EU that have experienced the fastest growth of market potential in the last decades(see Figure 4.3).The best characterisation of CENTROPE is thus that of a region comprised by strongcentres located at the intersection and border of two economically very differentterritories of the EU. It is a "transitory” region, in which good accessibility from theeconomic centres of Western Europe and from the rapidly growing Eastern Europeancountries shape comparative advantages. These – as is documented by a number ofrecent spectacular foreign direct investments – in general lie in a strong industrial base(in particular in ancillary industries such as automotive components), a strong orientationon medium skill and niche products and rapid technological catching up and low wagecosts (in particular in the Eastern part of CENTROPE).Figure 4.4: GDP per capita 2005 at PPS45,00040,00039,77435,00033,12430,00025,00023,080- - Centrope20,00015,00019,87715,25215,98613,614EU 25 average14,60410,0005,0000BurgenlandLower AustriaViennaSouth MoraviaGyor-Moson-SopronVasBratislavaTrnavaSource: Eurostat.4.1 Economic development of CENTROPE4.1.1 GDP and GDP per capitaThis implies that the region is characterised by sharp internal disparities. Due to thelegacies of the communist regimes the main dividing line within the region was - and stillis - the division between the new member states and Austria. While in the Austrian partsper capita GDP approaches or exceeds the EU average, all of the CENTROPE regions inWIFO 53

the new member states - except for Bratislava - currently qualify for Objective 1 status;their GDP per capita is much below the EU 25 average. In the richest region of CENTROPE(Vienna) GDP per capita was at 163% of the average (in 2005), in the poorest region(Vas) it was at 56% of the average (see Figure 4.4).However, not all differences in CENTROPE follow purely national lines. For instance thecapital city of Bratislava can claim a per-capita-GDP that is comparable to the Austrianregions and is above the EU average; Burgenland, on the other hand, has been anObjective 1 region until recently; its GDP per capita is not only below the EU-average butalso below the CENTROPE average.Thus, while there are important national differences within CENTROPE, there is a secondimportant division line between the large urban centres and more rural regions.CENTROPE’s favourable economic position, with a GDP per capita slightly above the EUaverage,mainly goes back to the "twin cities” of Vienna and Bratislava, while some morerural regions in both the Eastern as well as the Western parts of CENTROPE are clearlylagging behind in this respect.Figure 4.5: GDP growth 1995/2004Average annual change in %1210.511.0109.888.48.36EU25 average43.52.72.8- - Centrope20BurgenlandLower AustriaViennaSouth MoraviaGyor-Moson-SopronVasBratislavaTrnavaSource: Eurostat, WIFO-calculationsWhile the new member states regions may be considered poorer than the Austrianregions, they are more dynamic. GDP growth rates in the Czech, Hungarian and Slovakregions of CENTROPE ranged between 8.3% and almost 11.0% and clearly outperformedthe Austrian regions (with growth rates between 2.7% and 3.5%). The rapid catching-upprocess of the Central and Eastern European countries thus makes the eastern part ofWIFO 54

CENTROPE more dynamic than the European average. Most recent data from Eurostat(see Figure 4.5) sources suggests that the regions which have shown above-EU averageGDP growth within CENTROPE in the last decade are all located outside Austria. Thebelow-EU average growth performance of the CENTROPE region is thus due primarily tothe below average performance (and high weight) of the Austrian CENTROPE regions,while the Eastern part of CENTROPE has been characterised by an extremely rapidcatching up process, with most of the CENTROPE regions also growing substantially fasterthan their respective national average.4.1.2 Specialisation and sectoral structureThese dividing lines within CENTROPE illustrated above are also reflected by theeconomic structure of the region. Focusing on the sectoral employment and gross valueadded (GVA) shares in agriculture, industry and services in the NUTS 3 regions ofCENTROPE (see Table 4.1) indicates that in total the structure of CENTROPE does notdiffer significantly from the EU average sectoral structure. The share of agriculture andindustry in GVA are both by 0.1 percentage points lower in CENTROPE than in the EUaverage and the share of services is by 0.2 percentage points higher. These smalldifferences, however, mask the substantial structural heterogeneity within CENTROPE,which once again reflects the dividing lines between the new member states and Austriaon the one hand, and the urban regions and other regions on the other hand.Table 4.1: Sectoral Structure of GVA and Employment in CENTROPE *(NUTS 3 level 2004)Share of Agriculture Share of ManufacturingShare of ServicesStructural Change 1(2000-2004)GVA Empl GVA Empl GVA Empl GVA EmplPercentEU 27 2.2 - 26.2 - 71.6 - 2.0 -South Moravia 4.0 4.7 35.6 37.5 60.3 57.8 2.1 1.5Czech CENTROPE 4.0 4.7 35.6 37.5 60.3 57.8 2.1 1.5Gyır-Moson-Sopron 4.4 5.5 45.5 40.3 50.1 54.2 7.5 1.9Vas 4.8 5.2 44.1 45.1 51.2 49.7 6.8 1.7Hungarian CENTROPE 4.5 5.4 45.0 42.2 50.5 52.5 7.2 1.3Burgenland 5.8 6.8 30.4 35.0 63.8 58.3 0.3 1.0Lower Austria 3.8 4.3 35.7 36.7 60.6 59.1 1.1 1.7Vienna 0.2 0.6 16.3 14.6 83.5 84.7 2.2 2.6Austrian CENTROPE 1.7 2.6 23.7 25.3 74.6 72.1 1.8 0.8Bratislava 0.9 1.6 23.5 22.7 75.6 75.7 3.5 2.7Trnava 5.7 6.1 49.0 39.5 45.3 54.4 0.4 2.6Slovak CENTROPE 2.3 3.2 30.9 28.7 66.7 68.1 2.7 1.3CENTROPE 2.1 3.3 26.1 29.6 71.8 67.1 1.6 1.1Notes: *excluding extra-territorial organizations and bodies, 1 measured by the index of structural change, which isdefined as half the sum of the changes in sectoral shares in the time period 2000 to 2004 with 0 implying no structuralchange and the maximum value being 100.Source: EUROSTAT, WIFO-calculationsIn general, with the exception of Bratislava, the share of manufacturing in GVA andemployment is higher in the CENTROPE regions of the new member states than in theAustrian part of CENTROPE. Only in Lower Austria, which is considered an industrialregion in the Austrian context, the share of industry in employment and unemploymentWIFO 55

attains a level comparable to that of the less heavily industrialised among the newmember state regions of CENTROPE (such as Southern Moravia). In addition, in most ofthe more heavily industrialised regions within CENTROPE (such as Trnava and theHungarian regions) the share of industry in GVA exceeds 40%. The exception to this ruleis Bratislava, which (as its "twin city” Vienna) has a high share of services in both GVAand employment (and a low share in both agriculture and industry). Still, tertiarisation ismuch less advanced in Bratislava compared to Vienna, with a difference in the share ofservice employment of more than 10%.In addition, most of the new member states regions of CENTROPE (in particular Trnavaand - to a lesser extent - the Hungarian CENTROPE regions) have a slightly higher shareof agriculture in GVA and employment (which ranges at over 5% for employment andover 4% for GVA shares) than the Austrian regions. However, rural Austrian CENTROPEregions such as Burgenland approach (or even exceed) these shares – thus manifestingthe second line of division between the urban and more rural regions of CENTROPE.In summary, CENTROPE is not only characterised by significant disparities in terms ofeconomic development, but also in terms of sectoral specialisation. The eastern part ofCENTROPE is characterised by a substantially higher share of manufacturing in bothemployment and GVA, while service sectors tend to be underrepresented. This isreconfirmed when moving to NUTS 2 level data (Table 4.2). That the lower service sectorshare in GVA in the new member state CENTROPE regions applies to almost all servicesectors, but is most pronounced in real estate and business services, thus pointing toparticular structural deficits in these activities the new member states’ regions ofCENTROPE.Table 4.2: Sectoral Structure of GVA in CENTROPE (2004-NUTS II Level)Czech SouthEastWestTransdanubiaBurgenland Lower Austria Vienna Bratislava WesternSlovakiaPercentAgriculture 5.6 4.8 5.8 3.8 0.2 0.9 6.0Industry 31.8 37.9 20.5 26.8 11.2 19.2 39.8Construction 7.2 5.0 9.9 8.9 5.1 4.3 6.2Trade 11.5 7.7 9.7 12.9 16.6 15.8 12.4Hotels and Restaurants 1.8 2.3 4.3 3.0 2.8 1.3 1.3Transport 9.6 6.9 5.6 7.4 9.0 12.4 6.4Financial Services 1.5 2.0 5.2 3.8 8.0 11.9 2.0Real Estate 13.4 13.5 13.9 13.3 23.6 17.3 12.7Public Administration 5.3 7.7 8.4 6.1 6.5 8.1 4.5Education 5.1 5.0 6.5 5.4 4.7 2.8 3.5Health Services 4.3 4.5 6.7 5.7 6.0 2.1 3.4Other Public and Private Services 2.9 2.8 3.2 2.8 6.1 3.9 1.7Private Households 0.0 0.0 0.3 0.3 0.2 0.0 0.0Notes: *(excluding extra-territorial organizations and bodies)Source: EUROSTAT, WIFO-calculationsWIFO 56

4.1.3 Education, R&D and high technology resourcesSimilarly, the structure of the labour force and infrastructure endowments differsignificantly across CENTROPE regions. Aside from national differences in educationsystems these differences are also closely associated with urbanisation: In general,CENTROPE is characterised by a highly qualified workforce that has its strongholds in thesecondary and upper secondary education levels. In particular in the regions of the CzechRepublic and Slovakia over 90% of the workforce have a completed secondary education.The share of population with a tertiary education is, however, below the Europeanaverage in all regions but the capital cities of Vienna and Bratislava, where around aquarter of the workforce has completed tertiary education. High shares of the workforcewith only a completed primary education can only be found in some of the Austrianprovinces. Infrastructure endowments, accessibility and innovation indicators tend tofollow these patterns. In particular, indicators of R&D activity (such as R&D expenditures,patents per 1000 inhabitants) and infrastructure quality are clearly above the EUaverages for the large agglomerations (in particular Vienna and Bratislava), but not forthe more peripheral regions. 37Figure 4.6: Structure of the Workforce in CENTROPE 2006100%90%80%15.6 15.724.715.3 14.628.513.117.724.570%60%50%40%65.3 67.156.978.971.064.980.1 70.449.330%20%10%0%19.1 17.2 18.45.814.36.6 6.812.026.2BurgenlandLower AustriaViennaSouth MoraviaWestTransdanubiaBratislavaWestern SlovakiaCentropeEU 27low skill medium skill high skillNote: High skill –ISCED groups 0-2, Medium Skill – ISCED Groups 3-4, High Skill – ISCED Groups – 5 or moreSouce: Eurostat37 Among the CENTROPE regions, however, both the capital city of Bratislava and the other CENTROPE regionsrank below Vienna in terms of R&D expenditure. This suggests that cross border co-operation in R&D couldpotentially create additional value added to the region.WIFO 57

4.2 Cross border flows4.2.1 Cross–border enterprise co-operationOne area of substantial progress in recent years was economic integration: Tradebetween the CENTROPE countries has grown well above the EU average since the 1990’s,and by now for each of the CENTROPE countries the other countries in the region belongto the list of the most important trade partners. In addition, Austria and particularly thecity of Vienna have profited substantially from Austrian foreign direct investments goingto the new member states of the European Union (e.g. in the banking sector). Theseinvestments have changed the long term capital account in Austria over the last 20years. While at the end of the 1980’s Austria was a net importer of capital, since 2004Austria has a capital account surplus, with Austrian firms investing more abroad thanforeign firms invest in Austria.Table 4.3: Cross-border Enterprise Cooperation in the CENTROPE regionForm of Enterprise RelationshipsBought/Founded enterprisesBoughtEnterpriseFoundedEnterprisePart ofEnterpriseDelivery NetworksBuy Products Buy ServicesOthersOtherCooperationAt least oneOf any kindNIn percentVienna 6.7 3.0 4.2 16.6 10.4 16.6 25.0 404Lower Austria 3.4 1.0 3.8 14.3 7.8 9.2 18.4 293Burgenland 5.1 0.0 2.5 12.7 10.1 17.7 29.1 79Czech CENTROPE 2.0 0.3 0.6 10.6 4.2 9.2 21.6 357Slovakian CENTROPE 6.3 1.2 3.9 15.6 14.8 25.4 35.9 256Hungarian CENTROPE 0.6 0.6 1.2 20.7 12.4 20.1 30.8 169Total CENTROPE 4.2 1.3 2.8 14.9 9.4 15.4 25.6 1,558Notes: N= Sample sizeAbsoluteSource: LAMO, Huber et al, 2007This increased cross-border co-operation in the enterprise sphere is also documented inrecent questionnaire based evidence on cross border enterprise co-operation inCENTROPE (see: Huber et al, 2007, and Table 4.3). According to this evidence aroundone quarter of all enterprises in CENTROPE have at least one cross border relationshipwith another enterprise in the form of (partial) ownership, delivery or other forms of cooperation.Furthermore, a more detailed analysis of the cross border enterprise networkssuggests that:1. Cross-border delivery networks and other forms of co-operation are wellestablished in CENTROPE by now. Around 15% of the interviewed enterprisesstated that they have bought products from suppliers from across the border and9% have bought services from such suppliers. This form of co-operation isparticularly common in the Hungarian CENTROPE. Around 15% of the enterpriseshold other forms of co-operation, which may range from loose forms of cooperationto R&D networks as well as contractual forms of co-operation likefranchising.WIFO 58

2. Relationships based on ownership seem to be somewhat less frequent by contrast.Less than 5% of all interviewed enterprises stated that they (partially) owned orfounded an enterprise in another country of CENTROPE. Furthermore, theserelationships are more concentrated in the capital city regions. Since mostheadquarters are located in capital city regions, this form of co-operation isparticularly akin to these regions. Thus the highest share of such relationships isfound in Vienna and in the Slovak part of CENTROPE.4.2.2 Cross–border labour market mobilityWhile integration in the enterprise sphere is progressing rapidly, the CENTROPE region isstill less integrated compared to regions within one country. In particular, cross-borderexchange in the labour market (migration and commuting) still remains limited due toexisting institutional impediments and bottlenecks in infrastructure. This can beexemplified using migration data from Austria (see Table 4.4). In total, less than 11% ofthe foreign workers (and less around 1.5% of employees) in Austria are from theCENTROPE countries. The only region where workers from the CENTROPE countriesrepresent a sizeable group of the labour market is Burgenland. This is also due to aspecial institutional arrangements (the so called "Grenzgängerabkommen”) betweenAustria and Hungary, which allows Hungarians to commute to Austrian border regionsaccording to a quota.Table 4.4: Employees from CENTROPE-countries employed in Austria in 2006Vienna Lower Austria Burgenland Total AustrianCENTROPEAbsoluteHungary 2,582 2,817 5,543 10,942Czech Republic and Slovakia 4,212 5,177 561 9,950Total Foreign employees 123,950 54,312 10,705 188,967Total Employees 763,871 541,863 86,248 1,391,982PercentTotal CENTROPE in % of foreigners 5.5 14.7 57.0 11.1Total CENTROPE in % of Employees 0.9 1.5 7.1 1.5Source: Austrian Social Security, WIFO-calculations4.3 Labour Market Development of CENTROPE4.3.1 The structure of employment and unemployment rates in the NUTS 2regions of CENTROPEConsidering the labour market in a European context, CENTROPE can be considered aregion with relatively low unemployment rates and intermediate or slightly higher labourWIFO 59

market participation. Vienna, Vas, Trnava and South Moravia had unemployment ratesbetween 7% and 9%, with Vienna and Trnava showing an unemployment rate above theEU25 average in 2006 and South Moravia and Vas having an unemployment rate whichwas around 1 percentage point below the EU-average. All other CENTROPE regions hadunemployment rates substantially below the EU average (of 8.3%), ranging between 4%and 5% (Figure 4.7).Thus, relative unemployment rates follow the standard lines along which regionaldisparities develop in CENTROPE to a much lesser degree than indicators of regionaldevelopment. In particular there is no clear indication that the CENTROPE regions of thenew member states of the European Union have unambiguously higher or lowerunemployment rates than the Austrian CENTROPE regions. Both the region with thelowest unemployment rate (Gyır-Moson-Sopron, which, together with Lower Austria, hadan unemployment rate of 4.3% in 2006) and the region with the highest unemploymentrate in CENTROPE (Trnava, 8.8%) are located in the new member states of the EU.Figure 4.7: Unemployment rate 2006In %1098.88.8EU 2587- - Centrope7.17.46545.04.04.34.63210Burgenland Lower Austria Vienna South Moravia Gyor-Moson-SopronVas Bratislava TrnavaSource: EurostatIn addition, there is no clear indication of a urban – rural unemployment rate differentialin CENTROPE. Vienna is one of the regions with one of the highest unemployment rates inCENTROPE, while Bratislava is one of the regions with relatively low unemployment andthe regional unemployment rate is largely independent of sectoral specialisation. Thecoefficient of correlation of the regional unemployment rate with the share of agriculture,industry and services in total employment in the region is very low (with -0.03, -0.08 andWIFO 60

0.08, respectively), which suggests that there is no close (linear) relationship betweenregional unemployment rates and sectoral specialisation.Analysing labour market developments in more detail suggests that regional labourmarket disparities in CENTROPE are closely related to a number of more latent nationaland institutional differences between the countries. For example, regional employmentrates suggest that in a number of regions of CENTROPE low unemployment rates areaccompanied by low employment rates. This implies that low unemployment is due to lowlabour market participation. Especially in the Hungarian regions, despite below averageunemployment rates, employment rates are below the EU 25 average (of 64.6%). InWestern Slovakia employment rates are the lowest among the CENTROPE regions,despite high unemployment rates (see Figures 4.8 and 4.9).In general, however, the average employment rate in CENTROPE is 65.9% and thusexceeds the European average of 64.6% in all regions but Western Transdanubia(62.1%) and Western Slovakia (62.3%;). The highest shares were to be observed inLower Austria (71.5%), Bratislava (69.8%) and Burgenland (69.5%).Another difference in the regional labour markets of CENTROPE, which may distortregional unemployment rates, relates to the share of part time employment. This mayhave an impact on unemployment rate statistics, since a larger share of part timeemployed -ceteris paribus- implies lower average working hours per employed. Thus for agiven volume of working hours more people will be employed (and fewer unemployed) asthe share of part time employment increases.Table 4.5: Share of Part Time Employment in Total employment in the CENTROPE-regions(2006)Total Male FemalePercentEU 27 17.9 7.6 30.8Burgenland 19.8 4.3 39.4Lower Austria 21.0 5.6 39.8Vienna 21.5 10.6 33.8Czech South-East 5.5 2.6 9.4West Transdanubia 2.7 1.6 4.2Bratislava 3.6 1.7 5.7Western Slovakia 2.5 1.3 4.1Source: EUROSTAT, European Labour Force SurveyWIFO 61

Figure 4.8: Employment Rate 2005In %747271.57069.569.86866.5Centrope66646264.362.862.3EU 25605856Burgenland Lower Austria Vienna South Moravia WestTransdanubiaBratislavaWestern SlovakiaSource: Eurostat.In this respect the labour markets of the new member states of the EU are characterisedby a substantially lower share of part time employment than the EU 15. This differencealso applies to the CENTROPE regions of these countries (see Table 4.5). While in theAustrian CENTROPE the share of part time employed in total employment ranges between21.5% (Vienna) and 19.8% (Burgenland) and is thus above the EU average (thisdifference arises primarily due to the higher share of females in part time employment),the regional labour markets of the new member states have part time employmentshares which are substantially lower than the EU average.Aside from this, the structure of unemployment and employment rates in CENTROPEvaries substantially across regions. Given the low overall unemployment rates, the shareof long term unemployed is relatively high in most of the new member states' regions ofCENTROPE and low in the Austrian CENTROPE. In the year 2006, despite a favourablemacro-economic development of the regions in question, the share of long termunemployed increased in all of the new member states regions of CENTROPE except forBratislava (where a spectacular seven percentage point reduction was achieved), whilethe share of long term unemployed in total employment reduced in all of the AustrianCENTROPE regions. Furthermore, none of the new member state regions had a share oflong term unemployed in total unemployment which was below the EU27 average. TheAustrian CENTROPE regions, by contrast, have a share of long term unemployed in totalemployment that is 10 percentage points below the EU-average. For the new memberstates, this indicates a severe mismatch problem of the qualifications of the unemployedWIFO 62

with the requirements of prospective employers, as would be expected in economies withthe speed of restructuring of the new member state regions of CENTROPE.Table 4.6: Share of Long Term Unemployment in Total Unemployment in theCENTROPE-regions2002 2003 2004 2005 2006PercentEU 27 45.3 46.1 45.3 46.1 45.8Burgenland 19.2 29.0 34.3 29.0 27.6Lower Austria 26.1 27.2 30.4 27.6 27.0Vienna 36.2 39.3 38.0 29.7 34.0Czech South-East 47.1 45.1 48.4 50.3 52.0West Transdanubia 39.0 34.4 38.3 40.1 47.1Bratislava 53.3 46.9 46.7 39.1 55.1Western Slovakia 69.8 66.3 68.5 69.6 72.8Source: EUROSTATFigure 4.9: Employment Share of the Elder 2006, Age 55 to 64, in %605049.9EU 2542.940Centrope33.5 33.636.338.032.83020100Burgenland Lower Austria Vienna South East WestTransdanubiaBratislavaWestern SlovakiaSource: EurostatWIFO 63

Figure 4.10: Unemployment rate of the Younger 2006Age 15 to 24, in %2521.920EU 2517.718.81513.7Centrope108.07.850Lower Austria Vienna South East WesternTransdanubiaBratislavaWestern SlovakiaNote: Due to the small sample size for the Burgenland the Youth unemployment rate for this region isnot reported in Labour Force Survey dataSource: EurostatAdditional labour market problems specific to CENTROPE may be summarised as follows:- Due to a history of early retirements and the downsizing of the labour forceassociated with industrial restructuring, employment rates of the elder (i.e. thoseaged 55 and older) are low relative to the EU level in four of the seven regions(see Figure 4.9). In Bratislava (49.9%) the rate is above the European average of42.5%, in the Czech South East (42.9%) it does so only slightly. In all Austrianregions, where early retirement was particularly popular until recent changes inthe pension system, employment rates of the elderly are around 35%; they areeven lower in West Slovakia with 32.8%.- Aside from low employment rates of the elderly, youth unemployment rates areabove the EU-average in Vienna, the Czech South East and West Slovakia, butbelow this average for the CENTROPE as a whole. In Bratislava and WestTransdanubia regional youth unemployment rates are substantially below the EUaverage and the total youth unemployment rate in CENTROPE is below the EUaverage- Finally, in a number of the new member states’ regions of CENTROPE traditionallylow gender gaps in employment rates have rapidly increased in the last years.However, Trnava is still the only CENTROPE-region where the unemployment rateof females (as well as that of males) is above the European average; in Vienna therate of male unemployed is higher than the European average; in Trnava the sameWIFO 64

applies to the female unemployment rate. In all other CENTROPE regions, bothmale and female unemployment rates are below the average of the EU 25.Figure 4.11: Unemployment Rate by Gender 2005In %1412Men Women Men EU 27Men Centrope Women EU 27 Women Centrope11.5109.58.08.88.6866.15.85.76.46.644.23.54.53.34.74.420Burgenland Lower Austria Vienna South Moravia Gyor-Moson-SopronVas Bratislava TrnavaSource: Eurostat4.3.2. Development of unemployment and employment ratesAside from being relatively favourable, the labour markets of the CENTROPE region havealso been very dynamic in recent years. These dynamics have to a large degree beencharacterised by country specific developments.• Among the CENTROPE regions in particular the Slovakian regions experienced asubstantial decline in regional unemployment rates in 2005: In Trnava theunemployment rate decreased by 2.7 percentage points and in Bratislava thereduction was – 0.7 percentage points. Especially in Trnava declining unemploymentrates also seem to be of a long run nature. Since reaching an all time high of 18.1%in 2001 unemployment rates have continually declined by more than one percentagepoint each year. In the Slovak regions all other labour market indicators alsoimproved most noticeably among the CENTROPE regions: youth unemploymentdecreased in both NUTS 2 regions of the Slovak CENTROPE and employment ratesincreased both for the aggregate as well as for males, females and the elder(although these developments were stringer in Western Slovakia than in Bratislava)WIFO 65

Figure 4.12: Unemployment rates and their development in CENTROPEIn %TotalYounger204018161412353025102086415102502000 2001 2002 2003 2004 2005 200602000 2001 2002 2003 2004 2005 2006Burgenland Lower Austria ViennaSouth East Bratislava Western SlovakiaWest TransdanubiaMenBurgenland Lower Austria ViennaSouth East Bratislava Western SlovakiaWest TransdanubiaWomen2020181816161414121210108866442202000 2001 2002 2003 2004 2005 200602000 2001 2002 2003 2004 2005 2006Burgenland Lower Austria ViennaSouth East Bratislava Western SlovakiaWest TransdanubiaBurgenland Lower Austria ViennaSouth East Bratislava Western SlovakiaWest TransdanubiaSource: EurostatFigure 4.13: Employment rates and their development in CENTROPEIn %TotalElder63556150594557405535533051254920472000 2001 2002 2003 2004 2005 2006152000 2001 2002 2003 2004 2005 2006Burgenland Lower Austria ViennaSouth East Bratislava Western SlovakiaWest TransdanubiaBurgenland Lower Austria ViennaSouth East Bratislava Western SlovakiaWest TransdanubiaMenWomen707065656060555550504545402000 2001 2002 2003 2004 2005 2006Burgenland Lower Austria ViennaSouth East Bratislava Western SlovakiaWest Transdanubia402000 2001 2002 2003 2004 2005 2006Burgenland Lower Austria ViennaSouth East Bratislava Western SlovakiaWest TransdanubiaSource: EurostatWIFO 66

• A moderate decline in unemployment rates was also registered in the Czech SouthEast, where in 2006 reductions in unemployment – after being solely due to thedecline in male unemployment in 2006 – were also achieved with respect tofemale unemployment.• The Hungarian CENTROPE, by contrast, was characterised by a more modestdecrease in unemployment rates in 2006 (by 0.2 percentage points) after havingexperienced a 1.3 percentage point increase in 2005.• Finally, in Austria, too, regional unemployment rates decreased in all of theAustrian provinces, with the largest decrease in Burgenland (-1.0 percentagepoints) and a more moderate decline of (-0.3 percentage points each) in Viennaand Lower Austria.WIFO 67

5 Cross-Border Migration and Commuting Potentials in the CENTROPERegionThe CENTROPE region on account of its vicinity to the border, high regional disparitiesand high urbanisation may thus be considered as a primary example of a border regionthat could be strongly affected by cross border labour mobility after the end oftransitional periods. This as a first question raises the issue of , how many people in theregion may be willing to migrate or commute across border in the regions. The individuallevel data we use to address this question were collected within the scope of the Austrian"Labour Market Monitoring” (LAMO) project (see Hudler-Seitzberger Bittner, 2005, Huber– Mayerhofer – Nowotny – Palme, 2007). The aim of this project was to gain informationon the willingness to commute and migrate in the Central European "CENTROPE” region,which encompasses the eastern provinces of Austria (Vienna, Lower Austria andBurgenland) as well as the southern parts of the Czech Republic (South Moravia, andVysočina) and the western Slovakian (Bratislava and Trnava) and Hungarian regions(Gyır-Moson-Sopron, Vas and Zala) bordering on Austria. The data were collected in twowaves (with the first one taking place between November 2004 and February 2005, thesecond between November 2006 and February 2007) using personal face-to-faceinterviews in the Hungarian, Slovak and Czech regions of "CENTROPE” and (only in thefirst wave) in the Austrian provinces of Vienna, Burgenland and Lower Austria. In bothwaves, 15,791 individuals were interviewed, 11,693 of them living in the "CENTROPE”regions of the new member states (see table 4.1). According to the sampling plan,random quota sampling was applied to the working-age population of age 15 and older.Quotas were set by municipalities following a spatial analysis of the region. Municipalitieswere chosen based on characteristics such as municipality size, population growth andstructure, employment growth and unemployment rates as well as accessibility. Withinthe municipalities, random sampling was applied. 38These data are especially suitable for our analysis for two reasons: first, they consist notonly of information on the willingness to migrate and commute in view of the end of thetransitional period which currently restricts the free movement of workers from theCentral and Eastern European EU member states, but also include a large set of personalvariables which allows us to model mobility decisions based on individual characteristics.In addition to socio-economic characteristics respondents were also asked questionsconcerning their previous migration and commuting experiences, their plans for futurecross-border mobility, their expectations concerning a workplace abroad and their motivefor staying at home or being mobile. This allows us to differentiate between migrationand commuter potentials, but also to analyse the difference in structure between thesetwo groups. Secondly, in the literature the "CENTROPE” region has been repeatedlymentioned as the region at the former external border of the EU that will be most38 The underlying sampling plan was designed on the basis of an in-depth background analysis of the regionalstructure (Krajasits et al., 2005). The survey is representative of the CENTROPE population over 15 yearsof age.WIFO 68

strongly affected by commuting. It can thus act as a model region for analyzing plannedcross-border labour mobility after enlargement.In the interviews, respondents were asked a number of questions concerning their futurecross-border mobility plans. Interviewees were asked "Would it be conceivable for you towork abroad?” to which respondents could answer "yes” or "no”. Furthermore, they wereasked whether they would prefer (1) "daily commuting”, (2) "weekly commuting”, (3)"monthly commuting” or (4) "living and working abroad”. In subsequent questions,respondents were also asked which country they would prefer to work in and if they hadalready taken concrete steps towards working abroad. 39Table 5.1: Sample size of the LAMO household survey by waves and subregionsYear of observation2004-2005 2006-2007 TotalAbsoluteAustria 3,992 3,992Vienna 1,955 1,955Lower Austria 1,675 1,675Burgenland 362 362New EU member states 5,991 5,641 11,632Czech Republic 2,996 2,901 5,897Slovakia 1,550 1,484 3,034Hungary 1,445 1,256 2,701Total 9,983 5,641 15,624Source: LAMO household surveys 2004-2005 and 2006-2007Based on these questions and following Fassmann Hintermann (1997) as well as theliterature on questionnaire-based mobility surveys, various concepts of migration andcommuting potentials were defined and progressively narrowed: Aside from "general"and "expected" migration and commuting potentials, "real" migration and commutingpotentials are also defined.1. Migration potentials:39 During the interview, respondents were asked: "Have you already taken concrete steps to realize your goalof working abroad? ” If they reported to have collected information on their preferred receiving country,taken a training course, sold their belongings, learned the receiving country’s language, applied for a job, awork or residence permit abroad, found an accommodation or do already have a prospective job therespondents were registered as having taken concrete steps towards working abroad.WIFO 69

oooThe "general” migration potential includes individuals who do not currentlywork abroad, but consider seeking a job there (or would consider doing so ifthere were no transitional periods), and would also move their residenceabroad, returning home more often than once a month.The "expected” migration potential consists of those in the general migrationpotential who have either already collected information about their respectivetarget country, have taken training courses, learned the language, applied fora residence or work permit or for a job or who have a confirmed job offer or aplace to live.The "real” migration potential comprises only those in the expected potentialwho have already applied for a residence or work permit or a job or even havea confirmed job offer or a place to live abroad 40 .2. Commuting potentialsoooThe "general commuting potential" includes all persons who currently do notwork abroad, but consider seeking a job there (or would consider doing it ifthere were no transitional restrictions), but who intend to commute from theircurrent residence to their workplace abroad on a daily or weekly basis.Those in the general commuting potential who have either collectedinformation about their respective target country, taken training or languagecourses, applied for a residence or work permit or for a job or who alreadyhave a confirmed job offer belong to the are counted among the "expected”commuting potential.The "real" commuting potential has an even narrower definition. It refers onlyto those from the expected commuting potential who have either applied for ajob or a work permit or already have a confirmed job offer.3. Mobility potentialsFinally, there is also a "general mobility potential", which includes both thegeneral migration and the general commuting potentials. Similar definitions of the"expected" and "real mobility potential" can be derived from the expected and realmigration and commuting potentials 41 .5.1 Migration and commuting potentialsWhen applying these concepts to the LAMO data (table 4.2), the general mobilitypotential for 2006-2007 amounted to approximately 16.6%. As expected, the share ofpersons in the expected mobility potential is much smaller and amounts only to 5.2%,the real mobility potential to 2.1%. Between 2004-2005 (22.3%) and 2006-2007, the40 The real migration potential in this sense has a slightly broader scope than its counterpart in the study byFassmann - Hintermann (1997). The same applies to the real commuting potential.41 The concepts of "mobility potentials" are thus analogous to the "migration potential" concept of the study ofFassmann - Hintermann (1997). However, the "migration potential" in this study includes only those willingto migrate, not commuters.WIFO 70

general mobility potential decreased significantly 42 by more than 5 percentage points.The expected mobility potential also declined significantly by approximately 3 percentagepoints. The real mobility potential, by contrast, remained unchanged at 2.1%.Analysing the migration and commuting potentials separately reveals that the generalmigration potential comprised around 10.9% of the population in the NMS-regions ofCENTROPE age 15 or older in 2006-2007, while 5.6% generally considered commutingacross the border. Here, too, the expected and real potentials are substantially lower:The expected migration potential amounted to 3.8%, the real potential only to 1.3%. Theexpected commuting potential represents 1.4% of the population, the real commutingpotential only 0.8%.The decrease in the general mobility potential registered between 2004-2005 and 2006-2007 is thus mainly attributable to a more than 4 percentage point decline in the generalcommuting potential. The general migration potential also decreased over time, but byonly 1.5 percentage points. The expected migration and commuting potentials were alsosignificantly lower in 2006-2007 than they were in 2004-2005. The minor changes in thereal migration and commuting potentials, however, are not statistically significant.Table 5.2: Migration, commuter and mobility potentials in selected CENTROPE-regionsTotal NMS-regionsCzech Republic Slovakia Hungary2004-2005 2006-2007 2004-2005 2006-2007 2004-2005 2006-2007 2004-2005 2006-2007As a percentage of respondsMigration potentialGeneral 12.4 10.9 10.7 9.9 20.0 12.0 7.5 12.0Expected 4.9 3.8 3.6 3.3 9.4 5.5 2.8 2.7Real 1.4 1.3 1.5 1.0 1.9 2.1 0.8 1.3Commuting potentialGeneral 9.9 5.6 5.1 3.8 17.4 2.7 12.0 13.3Expected 3.0 1.4 1.1 1.2 6.0 0.7 3.9 2.7Real 0.7 0.8 0.3 0.7 1.7 0.3 0.6 1.4Mobility potentialGeneral 22.3 16.6 15.9 13.8 37.4 14.7 19.5 25.3Expected 8.0 5.2 4.7 4.5 15.4 6.2 6.8 5.4Real 2.1 2.1 1.8 1.7 3.6 2.4 1.3 2.6AbsoluteNo. of observations 5,991 5,641 2,996 2,901 1,550 1,484 1,445 1,256Source: LAMO household surveys 2004-2005 and 2006-2007, WIFO-calcuations42 Unless stated otherwise, the following discussion assumes a significance level (probability of error) of5 percent. This means that a statistical test (such as a test for differences in proportions) will detectsignificance of a chance relationship in no more than 5 percent of all cases.WIFO 71

Table 5.3: Migration, commuting and mobility potentials in the Austrian parts ofCENTROPEAustria Vienna Lower Austria BurgenlandAs a percentage of respondents, 2004-2005Migration potentialGeneral 16.4 25.3 7.6 9.1Expected 2.4 3.8 0.9 1.4Real 0.9 1.4 0.3 0.3Commuting potentialGeneral 2.2 1.9 2.3 2.8Expected 0.3 0.3 0.2 0.0Real 0.1 0.1 0.1 0.0Mobility potentialGeneral 18.6 27.2 10.0 11.9Expected 2.6 4.1 1.1 1.4Real 0.9 1.5 0.4 0.3AbsoluteNo. of observations 3,992 1,955 1,675 362Source: LAMO household survey 2004-2005, WIFO-calculationsDespite this decrease, the LAMO household survey reveals a considerable generalpotential for mobility in the NMS. A comparison with the Austrian subregions ofCENTROPE (table 3.3) - for which data was collected in the 2004 – 2005 wave only - putsthese figures into perspective: In 2004-2005, the general mobility potential in Austriawas higher than in the Czech Republic and only insignificantly smaller than in Hungary.The general migration potential was higher than the average of the NMS-regions(22.3%). This shows that the general mobility, migration and commuting potentials arevery broad concepts which express vague wishes rather than real intentions andtherefore must not be equated with actual or future migration: Only a small proportion ofthose who generally consider working abroad will actually do it.Expected and real mobility potentials in Austria, however, are lower than in the new EUmember states: A significantly lower number of persons has taken concrete steps tocommute or migrate abroad. One striking feature is the comparably low internationalcommuting potential in the Austrian CENTROPE subregions, which is mainly attributableto the fact that the neighbouring NMS labour markets are not attractive for mostAustrians (above all due to the lower wage levels) and that for most of them othercountries (like Germany) are beyond acceptable commuting distances. Thus, Austrianswould rather migrate than commute abroad.WIFO 72

The general decline in migration and commuting potentials in the NMS-regions ofCENTROPE was associated with relatively dissimilar developments in the individualcountries:- In 2004-2005 the general mobility potential in Slovakia was nearly twice as highas in Hungary and the Czech Republic. The same applied to the expected and realmobility potentials. However, Slovak data show large disparities between the twowaves of the survey: The general mobility potential in Slovakia shrunk from 37.4to 14.7%. The decrease was particularly pronounced in the general commutingpotential (from 17.4% to 2.7%). Compared to 2004-2005, the general migrationpotential also decreased substantially by 8 percentage points to 12.0% in 2006-2007. Similarly, the expected mobility potential in Slovakia was less than half ofits 2004-2005 value (15.4%) in the 2006-2007 survey. The real mobility potentialdecreased by approximately a third (first wave: 3.6%, second wave: 2.4%).- In Hungary, the general mobility potential showed an opposite development: Duemainly to a higher general migration potential (2004-2005: 7.5%, 2006-2007:12.0%), the general mobility potential increased significantly, from 19.5% to25.3%. The expected mobility potential declined also in Hungary, but notsignificantly (6.8 to 5.4%). A significant rise was observed in the real mobilitypotential, which doubled between 2004-2005 and 2006-2007, from 1.3% in thefirst wave to 2.6% in the second. Thus, Hungary was the only country with morenationals having undertaken concrete steps to work abroad in 2006-2007 than twoyears earlier.- The lowest general mobility potential can be found in the Czech regions. In thesecond wave it declined further, from 15.9 to 13.8%. The changes in the expectedand real mobility potentials ( 0.2 percentage points to 4.5% and 0.1 percentagepoints to 1.7%, respectively) were, however, not statistically significant. 43In general, a (moderate) decline in mobility propensity over time as shown by the datafor the Czech Republic and Slovakia was to be expected. With growing convergence ofincomes and economic standards, coupled with the positive impact of the economic boomon the labour market, the expected benefit from mobility is lower. This contributes toreducing the willingness for cross-border mobility.The importance of economic conditions is reflected by the fact that the development ofmobility in the NMS corresponds to the dynamics of unemployment: As shown in chapter2 between 2004 and 2006, the Slovak regions where interviews took place registered a43 The lower mobility among Czech workers is consistent with findings from studies on internal mobility byFidrmuc − Huber (2007), who have found a low propensity also for internal migration among the Czechpopulation. In the cross-border context, regional differences could also contribute to the low migrationpotential in the Czech Republic: Western Hungary and Western Slovakia are advantaged regions comparedwith their respective national average, so migration within the country would hardly make sense. The Czechregions at the Austrian border, by contrast, are partly disadvantaged in relation to the national average. Asa consequence, migration within a country might be considered an alternative to cross-border migrationWIFO 73

sharp decline in unemployment, which was paralleled by a decrease in the mobilitypotential. The moderate growth dynamics in the Czech Republic were also coupled with afall in unemployment, which partly explains the slight decrease in the Czech mobilitypotential. In the Hungarian regions, where the general mobility potential increasedbetween the two survey waves, unemployment went up between 2004 and 2006.The development of mobility also corresponds to real GDP growth dynamics in the NMS:while in the Slovak subregions the sharp drop in the willingness to move wasaccompanied by strong growth dynamics, the mobility potential in the Czech Republicdeclined only marginally against the background of a more moderate real GDP growth. InHungary, on the other hand, a lower GDP growth in 2007 compared to 2004 wasaccompanied by an increased general mobility potential. Part of the development ofcross-border mobility preferences can thus be explained by economic growth dynamics.Even though the economic development in the NMS contributes a lot to explaining thedynamics of the willingness to be mobile, the extent of the decrease shown in the datafor Slovakia is nevertheless striking. Additional factors could help explain these relativelypronounced changes:- Data always reflect the regime in place at the time of a survey. Against thebackground of Slovakia’s recent EU accession 6 months before the first survey, thewillingness to work in another (EU) country might have exceeded the long-termaverage for a short period before decreasing to its normal level by the time of thesecond survey. However, this effect would apply to all NMS, not only Slovakia. Thetemporary rise may be due to a positive general mood about EU accession inSlovakia 44 , which might have also shown itself as a more positive attitude towardscross-border mobility.- Considering the Slovaks' high willingness to move during the first survey, parts ofthe population who were willing to move may already have done so. As aconsequence, the remaining potential was correspondingly smaller by the time ofthe second survey. This is supported by the fact that in 2006-2007, 6.2% of theSlovaks stated they were already working abroad a significant 5 percentage pointsrise compared to 2004-2005. The corresponding shares in the Czech and theHungarian samples also increased significantly, but less sharply (from 0.7 to 2.0%and from 0.8 to 2.1%, respectively).- As the general potentials basically represent vague intentions, they might havebeen stated more generously in the first survey–when free movement of labourwas not possible due to the transitional periods–than in the second survey, withthe end of the transitional period drawing close. Therefore, the willingness to workin another country may be higher if it is not immediately possible. This44 Approval of the EU accession was particularly high in Slovakia (92.5% in the referendum held in May 2003.Hungary: 83.8%, Czech Republic: 77.3%. See European Commission, 2003).WIFO 74

psychological effect (people are likely to agree to actions in the future than toimmediately possible actions) could have contributed to the lower mobilitypropensity in the second survey (2006-2007) compared with 2004-2005. Thiswould, however, again apply to all countries, not only Slovakia.- Another reason for the decline in mobility propensity might be that, by the time ofthe second survey, more individuals had already made concrete experiences withthe transitional arrangements, which had a negative impact on their futurewillingness to move even after the end of the transitional period. Bureaucraticbarriers or other negative experiences made when searching a job abroad couldchange the general attitude towards cross-border mobility, even if at a later datethese barriers will not exist anymore. This effect, too, should theoretically occur inall countries, not only in Slovakia.- Finally, it cannot be excluded that disparities in data collection methodscontributed to these differences: Although the questionnaires were the same inboth waves, the second survey was performed by a different institute. Differencesin the sample structures might be another reason for the disparities between thetwo waves. However, eliminating differences in sample structure by weighting 45changes mobility potentials only marginally (±1 to 2% for the general mobilitypotential). Sample structure therefore plays only a minor role in explaining thediscrepancy in the mobility potentials between the two waves.5.2 Mobility, migration and commuting potentials towards AustriaFocusing on the mobility potentials directed to Austria as the preferred destination, it canbe seen that Austria is by far the most popular target country: About 40% of the generalmobility potential from the NMS-regions of CENTROPE is directed to Austria. There was,however, a slight decrease in Austria's share from 42.6% (1st wave) to 39.7% (2ndwave). Following Austria, Great Britain and Germany are other popular destinations 46 .Due to its geographical proximity, the majority (first wave: 71.2%, second wave: 64.7%)would rather commute than migrate to Austria. The opposite is true when looking atmobility to other countries: Here, the majority (first wave: 75.1%, second wave: 86.2%)would rather migrate. Austria is therefore above all interesting for potential commuters.The preference for Austria, measured as a share of the general mobility potential, ishighest in Hungary: Around two thirds (first wave: 66.3%, second wave: 62.6%) of thegeneral mobility potential in Hungary is directed to Austria. This is nearly twice as muchas in the Czech Republic (first wave: 36.8%, second wave: 26.3%) and in Slovakia (firstwave: 35.8%, second wave: 30.7%). The reason for this might be found in existinglabour market arrangements: An agreement on frontier workers which facilitates the45 The data of the second wave were weighted by age structure, gender and educational level and therebyadapted to the structure of the first wave.46 For a detailed overview of country preferences, see section 5.2.WIFO 75

labour market access in the border regions of Burgenland for Hungarian workers hasexisted between Austria and Hungary since 1998 47 .Table 5.4: Migration, commuting and mobility potentials from the NMS-regions ofCENTROPE to AustriaNMS Czech Republic Slovakia Hungary2004-2005 2006-2007 2004-2005 2006-2007 2004-2005 2006-2007 2004-2005 2006-2007As a percentage of respondentsMigration potentialGeneral 2.7 2.3 2.6 1.5 2.8 2.4 2.9 4.1Expected 0.9 0.8 0.5 0.4 1.3 1.1 1.4 1.0Real 0.2 0.2 0.1 0.1 0.2 0.3 0.5 0.3Commuting potentialGeneral 6.8 4.3 3.2 2.1 10.6 2.2 10.0 11.7Expected 2.2 1.0 0.8 0.7 3.6 0.6 3.5 2.4Real 0.5 0.6 0.1 0.5 1.2 0.3 0.4 1.1Mobility potentialGeneral 9.5 6.6 5.8 3.6 13.4 4.5 12.9 15.8Expected 3.1 1.8 1.3 1.1 4.9 1.8 4.8 3.4Real 0.7 0.8 0.2 0.7 1.4 0.6 0.9 1.4AbsoluteNo. of observations 5,991 5,641 2,996 2,901 1,550 1,484 1,445 1,256Notes: NMS: new EU member states.Source: LAMO household surveys 2004-2005 and 2006-2007, WIFO-calculations.Although the general preference for Austria as a receiving country remained largelyunchanged between 2004-2005 and 2006-2007, the general mobility potential directedto Austria declined significantly over time (table 3.4): While in 2004-2005, 9.5% of thepopulation in the NMS-regions of CENTROPE aged 15 or older were generally willing towork in Austria, the share was only 6.6% in 2006-2007. The main reason for this declineis that Slovakia’s commuting potential decreased while the general migration potentialremained practically unchanged 48 . The expected (along with the general) mobilitypotential to Austria also decreased significantly between the waves, from 3.1% to 1.8%,while the real mobility potential remained unchanged (2004-2005: 0.7%, 2006-2007:0.8%).47 The agreement regarding frontier workers gives workers in defined border regions labour market access aswell as a residence permit for 6 months, which can subsequently be renewed for another 6 months.48 The decrease of 0.4 percentage points between 2004-2005 and 2006-2007 is not statistically significant.WIFO 76

6 Determinants and Structure of Potential Migration and Commutingin the CENTROPE region6.1 Theoretical aspectsApart from the magnitude of the migration potential within the CENTROPE region, itsstructure is also of relevance. The question of whether and how the population willing tomigrate or commute differs from those unwilling to do so with respect to personalcharacteristics, and whether potential commuters and migrants have divergentdemographic features will be the focus of this section. In this respect economic theoryoffers a variety of approaches on modelling cross-border mobility of labour. As a startingpoint the neoclassical model explains mobility between two regions by real wage gapscaused by differences in factor endowment or technological levels. It predicts thatworkers migrate to regions where they can earn more. As a result of this, real wagesconverge across regions, since they decrease in the receiving and increase in the sendingregion. This is due to the assumption of a decreasing marginal product which in thismodel corresponds to the real wage (additional workers are employed as long as theirmarginal product surpasses their wages). Once wage differentials have been eliminatedin equilibrium, the incentive for mobility no longer exists.Contrary to the simplifying assumptions of the neoclassical model, however, mobility alsogenerates costs. Aside from monetary costs (e.g., those involved in moving residenceand, for commuters, travel costs, costs of job and housing search) or investment inhuman capital (such as learning a foreign language or acquiring additional qualifications),there are also non-pecuniary elements to these costs. These include the loss of personalcontacts, a greater distance to one's family and the loss of location-specific insideradvantages 49 (Fischer et al., 2000, Straubhaar, 2000), which are not transferable toother places of work and residence. Furthermore, bonds with the welfare state, such asnational insurance and transfer systems, can raise the attractiveness of immobility if theyincrease the opportunity cost of mobility 50 .In addition, various types of labour mobility differ both in terms of cost and benefits.Given that migration is the relocation of one's place of work and residence to a localityout of one's original area of residence (and work) either permanently or for a certainperiod of time and commuting is the relocation of solely the workplace to a locality out of49 Location-specific insider advantages can be production or consumption oriented. Production-oriented insideradvantages are, e.g., the knowledge of local standards, values and social manners, knowledge of localproduction technologies, ability to deal with local authorities and interest groups or company-specificknowledge that is only useful at the current workplace. Consumption-oriented insider advantages are, e.g.,the knowledge of prices and the quality of the local opportunities of consumption, amenities or publicservices (such as the educational or health system), see Straubhaar (2000).50 Straubhaar (2000) illustrates this by the following example: Welfare state transfers (such as unemploymentbenefit, social welfare allowance) subsidize immobility, since they may prevent recipients of unemploymentbenefit or social welfare allowance from improving their situation e.g. by migrating to a region with higherlabour demand.WIFO 77

one's area of residence 51 , costs and returns of migration differ from those of commuting.The decision to commute saves search costs for housing and reduces certain nonpecuniarycosts, like the loss of friends and consumption-oriented insider advantages orthe greater distance to family members. Furthermore, mobility costs differ betweencommuters and migrants: Commuting causes higher transportation costs, as the distancebetween a person's residence and their workplace has to be travelled daily (or weekly)while migrants pay a one time travel cost only, but also incur non-pecuniary costs.In addition, for risk averse individuals, commuting offers the opportunity to reduce therisk relating to the expected income in the receiving country without having to bear anymigration costs and, the separation of workplace and place of residence enablescommuters to enjoy the benefits of "two worlds", namely the higher wage levels and/orbetter job opportunities of the place where they work and the higher quality of living,healthier environment and lower real estate and housing prices of their place ofresidence.These differences between migration and commuting can be analysed in terms of arelatively simple model of migration and commuting originally due to Renkow and Hoover(2000). To highlight the interdependence of migration and commuting this modelconsiders two countries where the home country is composed of J regions and the foreigncountry f of only one. Since we focus on the willingness to migrate and commute acrossborders we consider an environment where labour is initially mobile only within but (dueto institutional barriers) immobile across countries, but where cross-border mobility willbecome an option in the future. Furthermore, we assume that the location of workplacesis given exogenously and that individuals faced with an offer for a workplace actmyopically and thus consider wages and land prices in all regions as given. 52Consider an individual k living in region i and working in region j of the home country(i,j∈J) which is asked whether it would be willing to commute or to migrate. We assumethat this individual receives utility from (expected lifetime) income in the region of work(Y j ) and (expected lifetime) amenities in the region of residence (A i ) which also includethe (expected lifetime) disutility arising from the rental price of housing. Furthermore, ifthe place of work and the place of residence of the individual do not coincide (i≠j), theindividual incurs (pecuniary and non-pecuniary lifetime) commuting costs of d ij . Theutility U k S of the individual living in region i and working in region j of the home countrycan then be written as:Uk,S=Y j−d ij+A i+εk,S (1)51 Migration and commuting flows of course also exist within a region, but are not considered in this study.52 While this would be a strong assumption in a general equilibrium analysis, it accords well with the empiricalpart of this study, since the data used are based on a questionnaire on prospective cross-border mobility.Issues of endegoneity are therefore of lesser importance in empirical applications based on actual mobilityoutcomes.WIFO 78

with d ij =0 if i=j and ε k s a random utility component for each individual associated withworking and residing in the home country.When considering moving across borders the individual working in region j and residingin region i expects a job offer associated with a lifetime income Y f from an employerlocated in the foreign country f. In this case accepting this job offer and remainingresident in i (i.e. commuting) would result in a utility of:U k,C =Y f−d if+A i+ε k,C (2)(where d if are the pecuniary and non-pecuniary costs of cross-border commuting and ε k Cis a random utility component associated with commuting from i to f), while acceptingthe offer and migrating to the new workplace abroad would give an expected lifetimeutility of:U k,M =Y f−M if+A f+ε k,M (3)where M if are the (pecuniary and non-pecuniary) costs of migration from i to f, A f is theexpected lifetime utility value of amenities (net of the rental price of housing) when livingabroad, and ε k M is a random utility component associated with migration. For the momentwe impose no restrictions on the random utility components ε k C and ε k M, which can beeither thought of as capturing random heterogeneity in tastes (as in Wall 2001), asreflecting uncertainty concerning living and working conditions in f (see, e. g., Burda1995) or as random draws from a distribution of mobility costs (as in Burda – Funke1993).Equations (1), (2) and (3) can be used to compute the differentials for individual kbetween the utility of staying in the home country (U k S) and the utility of commuting(U k C) or migrating (U k M) to country f:U k,C −U k,S =(Y f−Y j)−(d if−d ij)+(ε k,C −ε k,S ) (4)U k,M −U k,S =(Y f−Y j)−(M if−d ij)+(A f−A i)+(ε k,M −ε k,S ) (5)Equations (4) and (5) show that a higher income differential between the home andforeign countries (Y f - Y j > 0) increases the utility gain from working abroad. Higher crossborderthan within-country costs of commuting (d if - d ij > 0) decrease the utilitydifferential between commuting across the border and staying in the home country. Thesame holds true for the difference (M if - d ij ) in the case of migration: if the costs ofmoving abroad are higher than the lifetime commuting costs at home, the utility gainfrom relocating residence to f is diminished. If the utility of amenities in f is higher thanthe utility arising from amenities in the home country (or the rental price of housingabroad lower than in i), the term A f -A i is positive, which increases the utility gain frommigrating to f.WIFO 79

When deciding on the willingness to work abroad, the household also considers the utilitydifference between migration and commuting, which is given by:U k,M −U k,C =(A −A )−(M −d )+(ε k,M −ε k,C )f i if if(6)From equation (6) it follows that there is a positive utility differential between migrationand commuting (U k M - U k C > 0) if the foreign country offers more amenities (or a lowerrental price of housing), such that (A f - A j > 0) and if migration is associated with lower(pecuniary and non-pecuniary) costs (M if - d if > 0).Apart from the direct utility gains arising from the income, mobility cost or amenitydifferentials discussed above, the differences in the random utility components inequations (4), (5) and (6) also determine the choice between willingness to stay, migrateor commute. DefiningΩ =(Y −Y )−(d −d )CS f j if ijΩ MS=(Y f−Y j)−(M if−d ij)+(A f−A i)Ω MC=(A f−A i)−(M if−d if)as the "direct utility gains” from commuting vs. staying (Ω CS ), migration vs. staying (Ω MS )and migration vs. commuting (Ω MC ) 53 , and the "random utility gains” between commutingand staying, and between migration and staying asξ k,C =ε k,C −ε k,Sξ k,M =ε k,M −ε k,Sequations (4) to (6) can be rewritten as:U k,C −U k,S =Ω CS+ξ k C (7)U k,M −U k,S =Ω MS+ξ k M (8)U k,M −U k,C =Ω MC+(ξ k M-ξ k C) (9)According to equations (7) to (9), an individual’s willingness to commute or migrateacross borders depends on the realizations of the random utility variables ξ k M and ξ k C.This is also shown in Figure 6.1 where the optimum choices between the willingness tocommute, migrate and stay are depicted for given "direct” utility gains Ω CS and Ω MS withξ k M drawn on the horizontal axis and ξ k C on the vertical axis. An individual will not be53 Note that ΩMC can be computed as the difference between Ω MS and Ω CSWIFO 80

willing to work abroad if Ω CS -(ξ k M+ξ k C)Figure 6.1: Optimal Choices between Commuting, Migration and Staying in dependence ofξ k,C and ξ k,M for a given realization of Ω CSand Ω MSThe figure also illustrates that – as compared to a model in which only cross bordermigration is possible – the availability of commuting as a travel mode increases thelikelihood that a given person will be willing to become mobile, while at the same time itreduces the likelihood of a person being willing to migrate. To see this, notice that in theabsence of the possibility to commute all persons with Ω MS >-ξ k M will be willing to migrate,while all others will remain immobile. Thus the possibility of commuting will allow allpersons with Ω MS -ξ k C to become mobile (and to commute while all personsfor whom Ω MS >-ξ k M and Ω MC >-(ξ k M+ξ k C) will be willing to migrate rather than to commuteacross borders. Thus, for a relevant range of parameters individuals that would beunwilling to migrate may become mobile if the additional possibility of commuting isavailable, while – because for a further set of parameters cross-border commuting will bethe superior mode of labour to mobility – some individuals will now commute even if theywould migrate rather than stay if commuting were not available (see also Renkow –Hoover 2000)WIFO 81

Under the assumption that ξ k S, ξ k M and ξ k C follow a trivariate normal distribution with anarbitrary variance-covariance matrix Σ, equations (7) to (9) define a standardmultinomial probit model (see Maddala 1983). 54 This implies that the probability of theindividual being willing to commute from i to f (P C ), the probability of her being willing tomigrate to f (P M ) and the probability of her being immobile and staying in i (P S ) are givenby:P Fehler! (10)CP MFehler! (11)P SFehler! (12)Table 6.1: Direction of partial derivatives of the probabilities to commute, migrate andstay with respect to model variablesY f−Y jA f−A id ijd ifM ifP C + – + – +P M + + + + –P S – – – + +Taking the derivatives of these equations we find that the comparative statics of ourmodel follow those generally found by more complicated search theoretic models, whichconsider the possibility of commuting (e. g. Rouwendal, 1998, van Ommeren – Rietveld –Nijkamp 2000). As in these models higher wage differentials between receiving and thesending region will increase both the probability of being willing to commute and tomigrate, while reducing the probability of being immobile. By contrast, higher differencesin the benefits of residing in the sending and receiving region decrease the probabilitiesto be willing to commute and to remain immobile but increase the probability of beingwilling to migrate. Finally, individuals currently commuting in their home country (i. e.individuals with d ij >0) will be more likely to consider working abroad (because theopportunity cost of staying in the home country are higher) while the impact of increasedcommuting and migration costs has negative effects on the likelihood to be willing tocommute or migrate, respectively, and a positive impact on the choice to stay. Table 6.1summarizes the signs of the partial derivatives of the probabilities (10)–(12) with respectto model variables.54 In contrast to the multinomial logit model, where the assumption on the distribution of the error terms forcesthe covariance between choices to be zero (the so called Independence of Irrelevant Alternativeshypothesis, IIA), this formulation allows an arbitrary covariance structure across choices. As in themultinomial logit formulation, however, the parameters can only be identified relative to a base category,furthermore since utility levels are not identified the variance of one of the error terms must be set equal tounity.WIFO 82

6.2 Explanatory VariablesWe use three sets of explanatory variables to empirically model the willingness tocommute and migrate. First, we use a set of individual level variables which are intendedto capture differences in income opportunities. Among these are the age of theinterviewee, which can be expected to have a negative effect on both the willingness tomigrate and to commute (as the potential gain in lifetime income is higher for youngerindividuals), gender as well as highest completed education. Furthermore, we include adummy variable for individuals who have not yet finished their education ("student”). Tocontrol for language skills we also include dummy variables for the knowledge of German,English, another foreign language or no foreign language at all.To control for potential effects of social deprivation (see, e. g., Stark – Taylor 1991) onthe willingness to migrate and commute, we also include a variable measuring the(subjective) social status of the respondent relative to his/her reference group. Weconstruct this measure by taking the difference between the individual’s evaluation of herpersonal overall living conditions on an 11 point scale (with 1 representing the bestconceivable social status and 11 the worst conceivable social status) and those of theirfriends and relatives. The measure is thus negative if the individual assesses her ownsocial status as being higher than that of her peers, and positive if the individual feelsdeprived of her peers’ social status (see Stark – Taylor 1991) .Our second set of variables captures individual differences in the costs to commute or tomigrate. We include a dummy if the respondent is currently commuting to the currentplace of residence ("commuter”) as a proxy for the costs of commuting, since our modelsuggests that respondents who are currently commuting should be more willing tomigrate and commute, and dummy variables for marital status ("single”) and thepresence of children in the household ("kids”), since a number of contributions suggestthat persons living in larger households will face higher costs of migration (such as jobsearch costs or schooling for other household members) than single households (seeMincer 1978). Furthermore, we include variables which measure whether the respondentshave family members or friends residing abroad as proxies for potential network effects(see Straubhaar 2000), which can help to reduce mobility (as well as job search) costssignificantly. We also control whether the individual herself has already worked abroad byincluding a dummy variable that takes on the value 1 if previous mobility exists.As a measure of the direct costs associated with commuting or migrating abroad, weemploy the road distance between the interviewees municipality of residence 55 to thenearest EU15 border crossing in kilometres 56 as a proxy for distance to the nearestpotential workplace in the EU 15. Distance has proven to be uniformly the most importantfactor in explaining both migration as well as commuting patterns in many countries55 Note that the distance variable is still identified in this specification despite the inclusion of regional dummiessince it is defined on a regionally more disaggregated level.56 The distance was obtained using the route planner of the Austrian Motorists Association (ÖAMTC).WIFO 83

(see Fields 1979), and a negative effect on the willingness to migrate can therefore beexpected.Third, we include a family of regional dummy variables at the NUTS 3 level of disaggregationto control for characteristics (such as amenities) of the region of residence as wellas a dummy for interviews conducted during the second wave in 2006/07. Interactionterms of these region and wave dummies are included to account for regional differencesin changes in the macro-economic environment.Table 6.2: Mobility potentials by ageGeneralExpectedRealObservations2004/05 2006/07 2004/05 2006/07 2004/05 2006/07 2004/05 2006/07In percent of persons interviewedUp to 25 years 1 54.5 41.7 22.1 12.2 4.1 3.0 1,052 1,0002 42.9 44.6 48.5 41.9 33.6 25.226 – 35 1 30.2 21.4 10.9 7.2 3.9 3.6 1,186 1,2252 26.8 28.0 27.0 30.2 35.9 37.036 – 45 1 19.4 12.6 4.9 4.4 1.8 2.3 1,122 1,0392 16.3 14.0 11.5 15.8 15.6 20.246 – 55 1 12.3 7.6 3.9 1.9 1.1 1.2 1,141 1,0612 10.5 8.7 9.2 6.9 9.4 10.956 – 65 1 4.5 4.3 1.8 1.3 0.7 0.7 965 8432 3.2 3.9 3.6 3.8 5.5 5.0Over 66 years 1 0.8 1.7 0.2 0.8 0.0 0.4 525 4732 0.3 0.9 0.2 1.4 0.0 1.7AbsoluteObservations 1,336 935 478 291 128 119 5,991 5,641PercentUp to 25 years 31.4 20.1 27.6 13.9 23.3 10.026 – 35 47.2 35.9 41.1 27.3 28.3 36.436 – 45 60.6 57.3 60.0 50.0 55.0 58.346 – 55 65.7 60.5 61.4 50.0 41.7 53.856 – 65 51.2 36.1 29.4 36.4 42.9 50.0Over 66 years 25.0 37.5 0.0 0.0 0.0 0.0AbsoluteTimes mentioned 596 318 182 78 42 43Notes: 1 = relative row frequency, row sum = 100; 2= relative column frequency, column sum = 100.Absolute6.2.1 Individual level variables to capture income differentialsThe upper part of Table 6.2 shows mobility potentials of the NMS-regions of CENTROPEfor both waves by age group. Each category is described of the different mobilitypotential concepts (general, expected and real mobility potential) by two figures: Thefirst figure indicates the share of persons willing to be mobile (either through commutingor migration), the second figure beneath it, gives the share of persons in the respectiveWIFO 84

mobility potential. 57 Furthermore, also the number of observations for each of the wavesis given. This table confirms the higher willingness to be mobile of younger workers. Thegeneral willingness to work abroad tends to be higher among young people than amongolder people: Less than 5% of the over-55-year-olds can imagine working abroad. Bycontrast, the percentage among persons up to 25 years is 54.5% (1st wave) and 41.7%(2nd wave). A similar picture also emerges in the expected and real mobility potentials.Even though the propensity to be mobile decreased in nearly all age cohorts between2004/05 and 2006/07, around two-thirds of mobile persons in both waves are 35 yearsold or younger.The lower part of Table 6.2 shows that the willingness to commute generally increaseswith age. Very young persons would rather be willing to migrate in both waves, while themajority of persons over 35 would rather commute across borders. The highest share ofpotential commuters can be found in the age group of 46 to 55-year-olds with 65.7%(1st wave) and 60.5% (2nd wave).Table 6.3:Mobility potentials by genderGeneralExpectedRealObservations2004/05 2006/07 2004/05 2006/07 2004/05 2006/07 2004/05 2006/07In percent of persons interviewedMen 1 27.5 18.7 10.0 6.2 2.7 2.6 2,950 2,8072 60.8 56.1 61.7 59.5 62.5 62.2Women 1 17.2 14.5 6.0 4.2 1.6 1.6 3,041 2,8342 39.2 43.9 38.3 40.5 37.5 37.8AbsoluteObservations 1,336 935 478 291 128 119 5,991 5,641Share of commuters in mobility potential in percentMen 45.0 35.8 40.0 29.5 32.5 39.2Women 44.1 31.7 35.0 22.9 33.3 31.1AbsoluteObservations 596 318 182 78 42 43Notes: 1 = relative row frequency, row sum = 100; 2= relative column frequency, column sum = 100.AbsoluteSource: LAMO household surveys 2004/05 and 2006/07, WIFO-calculations.The survey results thus suggest that the willingness to migrate or commute correlatesnegatively with age, and older workers if mobile would rather commute than migrate.The only deviation is the slightly higher migration propensity of workers over 55 years of57 Thus, the first line of the table indicates that 54.5% of the 1,052 persons aged up to 25 years interviewed inthe first wave would consider working abroad. The second line indicates that 42.9% of the 1,336 persons inthe category of general mobility potential of the first wave are 25 years old or younger. The sum of thesecond line is 100% in every column.WIFO 85

age. The willingness to commute within this group differs only in the general mobilitypotential of the second wave and in the expected mobility potential of the first wave. 58Looking at the mobility potentials by gender (see Table 6.3) shows that the generalmobility propensity of women is significantly lower at both points in time than that ofmen. This difference is statistically significant also in the expected and real mobilitypotentials, and is consistent with the findings of earlier studies (Eliasson 2003). Bycontrast, there is no significant difference between men and women as regards thedecision to commute or to migrate. The share of commuters in mobility potentialdecreased among both women and men between 2004/05 and 2006/07.Table 6.4:Mobility potentials by foreign language knowledgeGeneral Expected Real Observations2004/05 2006/07 2004/05 2006/07 2004/05 2006/07 2004/05 2006/07In percent of persons interviewedNo foreign language knowledge 1 7.3 6.4 2.2 0.7 0.3 0.6 1,197 9672 6.5 6.6 5.4 2.4 3.1 5.0Foreign language knowledge 1 26.1 18.7 9.4 6.1 2.6 2.4 4,794 4,6742 93.5 93.4 94.6 97.6 96.9 95.0AbsoluteAbsoluteObservations 1,336 935 478 291 128 119 5,991 5,641Share of commuters in mobility potential in percentNo foreign language knowledge 64.4 67.7 65.4 100.0 25.0 100.0Foreign language knowledge 43.2 31.6 36.5 25.0 33.1 32.7AbsoluteObservations 596 318 182 78 42 43Notes: 1 = relative row frequency, row sum = 100; 2= relative column frequency, column sum = 100Source: LAMO household surveys 2004/05 and 2006/07, WIFO-calculationsPersons without foreign language skills have also a much lower willingness to migrate orcommute than persons with foreign language skills. In 2004/05 only 7.3% of all personswithout foreign language skills were generally willing to work abroad, while the shareamong persons with foreign language skills was 26.1%. As regards the data of the 2ndwave and the expected and real mobility potentials, this difference is also statistically58 Younger persons willing to migrate or commute show a lower degree of preference for Austria as adestination country: The share of persons in the general mobility potential that would prefer to work inAustria is more than 50% among those over 35 years, but only 31% among the group of the under-26-year-olds at the time of the first wave of interviews. This share decreased further to 25.9% by 2006/07.However, this does not mean that it is mostly older workers that would like to commute or migrate toAustria: More than half of the mobility potential directed towards Austria is 35 years or younger. In theremaining general mobility potential (directed at other countries), the share of below-35-year-olds ishowever around 80% (1 st wave: 77.7%, 2 nd wave: 83.2%).WIFO 86

significant. Only around 5% of persons willing to migrate or to commute do not speakany foreign languages.The share of commuters among those in the general mobility potential without foreignlanguage skills is significantly higher than among those with foreign language skills: Asexpected, persons that do not speak any foreign language would rather commute thanmigrate, because cross-border commuting requires lower foreign language skills thanmigration. This is also the case for the expected mobility potential in 2004/05, while inthe remaining potentials, the share of commuters in the group of persons without foreignlanguage skills must be interpreted cautiously, due to the low number of observations.Finally, highest completed education is used as a measure of qualification (Table 6.5). Itcan be seen that the general mobility potential of NMS-region consists mainly of personswith secondary and tertiary education. However, there are differences in the compositionof the mobility potential between 2004/05 and 2006/07: The share of persons withhigher education in the general mobility potential decreased from 42.4% to 34.9%.Nonetheless, the mobility potentials consist up to two-thirds of persons with higherqualifications.Table 6.5:Mobility potentials by highest completed level of educationGeneral Expected RealObservations2004/05 2006/07 2004/05 2006/07 2004/05 2006/07 2004/05 2006/07In percent of persons interviewedElementary school 1 22.3 29.5 7.8 5.8 1.0 1.3 792 5392 13.2 17.0 13.0 10.7 6.3 5.9Apprentice/vocational school 1 15.8 10.9 4.5 3.1 1.3 1.6 1,983 2,2262 23.4 25.9 18.6 24.1 19.5 29.4Upper secondary school 1 25.9 17.3 9.8 6.0 2.7 2.1 2,187 1,8802 42.4 34.9 44.8 38.5 46.1 33.6College/university 1 27.1 20.9 11.0 7.8 3.5 3.7 1,029 9962 20.9 22.2 23.6 26.8 28.1 31.1AbsoluteObservations 1 1,336 935 478 291 128 119 5,991 5,641Share of commuters in mobility potentialElementary school 30.5 34.6 25.8 38.7 37.5 71.4Apprentice/vocational school 53.0 47.5 52.8 42.9 36.0 57.1Upper secondary school 45.1 28.2 35.0 19.6 32.2 30.0College/university 43.0 26.9 38.9 17.9 30.6 16.2AbsoluteObservations 596 318 182 78 42 43Notes: 1 = relative row frequency, row frequency = 100; 2= relative column frequency, column sum = 100AbsoluteSource: LAMO household surveys 2004/05 and 2006/07, WIFO-calculationsThis positive qualification structure reflects the distribution of education levels in thesending countries. Comparing the general willingness to migrate or commute across theeducational levels shows that on the one hand, the willingness to migrate or commute ishighest among those persons that completed only elementary school education (1stwave: 22.3%, 2nd wave: 29.5%) and lowest among persons who have completed anapprenticeship or a vocational school (1st wave: 15.8%, 2nd wave: 10.9%). On the otherhand, the general willingness to work abroad increased slightly among those with higherWIFO 87

education levels. There are thus signs for both a negative as well as a positive selectionin general mobility potential. The mobility potential is therefore not unambiguouslypositively or negatively selected, but rather shows signs of a bipolar selection.There are, however, relatively large differences by country: The general willingness tomigrate or commute is significantly higher among Czech citizens with lower educationallevels (elementary schooling) and higher qualification (college, university) in both wavesthan among persons with completed apprenticeships, vocational schools or uppersecondary education. By contrast, a generally higher willingness to migrate or commutecan be found among Slovak citizens with a low level of qualification and a rather lowdegree of willingness to migrate or commute among persons with vocational training(apprenticeship/vocational school). Generally, one may speak of a negative selection inthe Slovak group in both waves. No concrete result can be derived for Hungary: While inthe first wave, the general willingness to migrate or commute among persons withelementary school education was significantly lower than average, it was significantlyhigher in 2006/07. The Willingness to migrate or commute within the remainingeducational levels hardly changed between the two time periods. 59The willingness to commute is highest among those in the general mobility potential whohave completed an apprenticeship or vocational training. The highest degree ofwillingness to migrate in the general mobility potential of the 1st wave is found amongpersons with elementary school only. By contrast, persons willing to be mobile withhigher levels of education would rather migrate than commute.6.2.2 Individual level variables affecting migrations costsFurthermore, the distribution of mobility potentials by family status (Table 6.6) showsthat single persons are more mobile: 35.4% could imagine working abroad in 2006/07(vs. 13.2% of married persons or individuals living in a partnership), and singlesaccounted for 65.0% of the general mobility potential. The expected and real mobilitypotentials are also higher among single persons than among persons living in59 Differentiating by preferred target countries there are considerable differences in the distribution of theeducational levels between those who would prefer to work in Austria and those who do not state Austria astheir first country of preference. The data of the first wave show that the share of persons who havecompleted apprenticeships or vocational schools in the mobility potential directed towards Austria issignificantly higher (29.7%) than in the remaining mobility potential (18.8%), while the share of those withtertiary education was significantly lower (18.1% vs. 23.0%) than in the mobility potential not directedtowards Austria. Data from the second wave also indicate a significantly higher share of persons withcompleted apprenticeships or vocational secondary schools in the mobility potential directed towardsAustria (37.7% vs. 18.1%). This contrasts with a significantly lower share of persons with secondary(29.4% general mobility potential directed towards Austria vs. 38.5% mobility potential not directedtowards Austria) and tertiary education (15.9% vs. 26.4%). What is remarkable is the significant decline inthe number of persons with secondary education over time: While the share of persons with secondaryschool education in the mobility potential directed at Austria in 2004/05 was still 40.1%, it dropped to only29.4% in 2006/07. The share of persons with tertiary education was also lower in 2006/07 than two yearsbefore. Austria thus seems to be losing appeal for workers with higher qualifications. This group rathershows an increasing preference for other countries. Thus, the relatively restrictive transition arrangementsin Austria and other countries might have caused a shift in country preferences among higher qualifiedmigrants and commuters towards countries that did not make use of the transitional arrangements, likeGreat Britain or Ireland.WIFO 88

partnerships. This supports the hypothesis that the reduced flexibility of multipersonhouseholds influences mobility decisions. Looking at the mobility potential by familystatus and type of mobility shows that – as could be expected – there is a higherwillingness to migrate among single persons. Those who are married or living in apartnership would rather commute abroad.Table 6.6:Mobility potentials by family status and childrenGeneral Expected Real Observations2004/05 2006/07 2004/05 2006/07 2004/05 2006/07 2004/05 2006/07In percent of persons interviewedNot single 1 13.2 9.7 3.6 2.8 1.3 1.5 3,539 3,4162 35.0 35.3 27.0 32.3 35.9 42.0Single 1 35.4 27.2 14.2 8.9 3.3 3.1 2,452 2,2252 65.0 64.7 73.0 67.7 64.1 58.0AbsoluteObservations 1,336 935 478 291 128 119 5,991 5,641In percent of persons interviewedNo children 1 22.7 18.4 8.3 6.0 2.3 2.2 3,243 3,2222 60.3 63.4 62.2 66.0 62.2 60.5Children 1 23.2 14.1 7.9 4.1 2.2 1.9 2,086 2,4172 39.7 36.6 37.8 34.0 37.8 39.5AbsoluteObservations 1,220 935 434 291 119 119 5,329 5,639Share of commuters in mobility potential in PercentNot single 63.7 50.3 58.9 41.5 43.5 58.0Single 34.3 25.1 30.4 19.8 26.8 20.3AbsoluteObservations 596 318 182 78 42 43Share of commuters in mobility potential in PercentNo children 36.1 29.7 29.6 22.4 24.3 22.2Children 55.4 41.5 50.0 35.4 48.9 57.4AbsoluteObservations 534 318 162 78 40 43Notes: 1 = relative row frequency, row frequency = 100; 2= relative column frequency, column sum = 100AbsoluteSource: LAMO household surveys 2004/05 and 2006/07, WIFO-calculationsThe hypothesis that children are a barrier to mobility was not supported by the data ofthe first wave (Table 6.6): there is no statistically significant difference in the willingnessto migrate and commute between persons without and those with children younger than15 years of age living in the same household 60 . However, the data of the 2nd wave showsthat childless persons have a significantly higher willingness to work abroad.60 As the question of children in the household was not answered by all persons interviewed, only 5,329 (1stwave) and 5,639 observations (2nd wave) are available for evaluation. The question was: "How manychildren up to 15 years of age live in your household?" Therefore, the children are not necessarily theinterviewed persons’ own children.WIFO 89

However, children seem to affect the choice of mobility mode significantly. As expected,the general migration propensity among those without children is significantly higherthan among persons with children: The latter would rather commute than migrate. Thisdifference also applies to other mobility concepts. Thus, one can state that there aresignificant differences between childless persons and persons with children with regard tomigration potentials: The former express a higher willingness to migrate than the latter.Table 6.7:Mobility potentials by networksGeneral Expected Real Observations2004/05 2006/07 2004/05 2006/07 2004/05 2006/07 2004/05 2006/07In percent of persons interviewedNo networks 1 15.0 10.2 3.9 1.9 0.9 0.5 3,682 2,8622 41.9 31.5 30.0 18.5 26.6 12.8Networks 1 35.6 24.6 15.4 9.1 4.4 4.0 2,157 2,5722 58.1 68.5 70.0 81.5 73.4 87.2AbsoluteObservations 1,320 923 474 287 128 117 5,839 5,434Share of commuters in mobility potential in percentNo networks 46.3 34.7 40.1 24.5 26.5 26.7Networks 43.8 33.5 37.7 27.4 35.1 37.3AbsoluteObservations 592 313 182 77 42 42Notes: 1 = relative row frequency, row frequency = 100; 2= relative column frequency, column sum = 100AbsoluteSource: LAMO household surveys 2004/05 and 2006/07, WIFO calculationsRelatives and friends abroad can contribute to lowering the costs of mobility: Networkscan, for example, help with housing, work or bureaucratic hurdles, etc. Therefore,persons with friends or relatives already working abroad should be more willing tomigrate or commute due to these network effects than persons without networks. This isalso confirmed in our data (Table 6.8). Among those with networks 61 the general mobilitypropensity is twice as high as among those that do not have network contacts. Thisdifference becomes even clearer when looking at the expected and real mobilitypotentials, with the mobility propensity of persons with networks being four to eighttimes higher.Networks are not only important for migrants, but also for commuters: In both wavesthere is a significant difference in the willingness to work abroad in the general, expectedand real migration and commuting potentials between those with and without networks.However, the existence of networks does not seem to have any significant influence onthe choice of mobility mode: The share of potential commuters among those without61 Unfortunately, there is no information on whether these persons live in the preferred/intended destinationcountry or in another country.WIFO 90

networks does not differ significantly from that of individuals with network contacts.Networks are therefore equally important for both commuters as well as migrants.Table 6.8:Mobility potentials by previous cross-border mobilityGeneral Expected Real Observations2004/05 2006/07 2004/05 2006/07 2004/05 2006/07 2004/05 2006/07In percent of personsNo previous mobility 1 20.1 15.0 6.5 3.7 1.2 1.1 5,439 4,9632 82.1 79.5 74.4 63.2 51.6 45.4Previous Mobility 1 48.8 38.2 25.0 21.3 12.7 12.9 488 5022 17.9 20.5 25.6 36.8 48.4 54.6AbsoluteObservations 1,333 935 476 291 128 119 5,927 5,465Share of commuters in mobility potential in percentNo previous mobility 46.4 34.6 41.0 24.5 33.3 40.7Previous Mobility 36.1 31.8 29.5 30.8 32.3 32.3AbsoluteObservations 594 318 181 78 42 43Notes: 1 = relative row frequency, row frequency = 100; 2= relative column frequency, column sum = 100.AbsoluteSource: LAMO household surveys 2004/05 and 2006/07, WIFO calculationsPersons who have worked abroad before are also more willing to migrate than those whonever worked outside of their home country. This is particularly obvious when looking atthe real mobility potential: Around 13% (1st wave: 12.7%, 2nd wave: 12.9%) of allpersons who have worked abroad at some point have already taken preparatory steps todo so again in the future. Around half (1st wave: 48.4%, 2nd wave: 54.6%) of the realmobility potential consists of persons who have already worked abroad even though thisgroup accounts for less than one tenth (1st wave: 8.2%, 2nd wave: 9.2%) of the sample.Investigating the decision to commute or to migrate by previous mobility, there is asignificant difference in the general and expected mobility potential of the 1st wave:Individuals who never worked abroad are more willing to commute than persons withprevious mobility experience. This indicates that commuting is a method to reduceuncertainty regarding living and income conditions abroad. This uncertainty can beexpected to be lower among persons who had already worked abroad before, becausethey have already reduced their uncertainty in a previous mobility step. Therefore, theynow show a higher willingness to migrate. By contrast, those who never worked abroadare rather willing to commute in order to reduce this uncertainty.6.3 Estimation resultsDescriptive analysis thus suggests that in particular, the presence of kids or a spouse inthe household is a more serious impediment for the willingness to migrate than for thewillingness to commute. This suggests that kids and a partner in the household increasecross-border migration costs more strongly than cross-border commuting costs.WIFO 91

Furthermore, gender differences in the willingness to commute are larger than for thewillingness to migrate (although women are both significantly less willing to commuteand to migrate), and the willingness to migrate reduces much more strongly with agethan does the willingness to commute. While migrants may thus be considered to becomposed mainly of young singles, commuters are more likely to be older and have afamily. Furthermore, commuters may be expected to be even more strongly selected bygender. Furthermore descriptive analysis also suggests that both those willing tocommute as well as those willing to migrate are disproportionately often drawn from thetwo extremes of the educational distribution, and are thus often either highly or lesseducated.Table 6.9 presents the regression results of a multinomial probit estimation of thedeterminants of the willingness to migrate and commute, in order to analyse to whatdegree these results are influenced by potential colinearities between different variables.These results suggest a number of differences between potential cross-border migrantsand commuters. In particular, age has a stronger negative effect on the probability ofbeing willing to migrate (the "propensity to migrate”, measured relative to the probabilityof being unwilling to migrate or to commute, i. e. the probability to stay) than on theprobability of being willing to commute (the "propensity to commute”). However, thedifference in the coefficients is only significant at the 10% level. The presence of kids inthe household is insignificant for potential commuters, but significantly negative forpotential migrants and the dummy variable for single households is significantly higherfor potential cross border migrants than for potential cross-border commuters. This thussuggests that the presence of children and a spouse in the household increases crossbordermigration costs more strongly than commuting costs.The results, however, also suggest that women are generally less mobile than men.Especially their probability of being willing to commute (relative to staying) isconsiderably lower. This thus lends support to the argument (see Madden 1981, White1986, Clark, Huang and Withers 2003) that women are less likely to commute or, whencommuting, travel shorter distances due to higher opportunity costs of time spentcommuting. Furthermore, as expected, distance is a stronger deterrent to cross-bordercommuting than to cross-border migration, while English language knowledge increasesthe probability of being willing to migrate and at the same time reduces the probability ofbeing willing to commute. Knowledge of other languages than German and Englishincreases only the probability of being willing to migrate and has no significant influenceon the propensity to commute across the border. Knowledge of the German language bycontrast increases the probability of being willing to commute significantly more than thewillingness to migrate. These stylized facts can be explained by the fact that Germanspeaking countries are the only ones that can be reached by daily commuters from theregion surveyed in our sample.WIFO 92

Table 6.9: Multinomial probit regression of willingness to be mobileVariable Migrant Commuter Mean (Std. dev.)Age -0.033*** -0.022*** 39.615(0.004) (0.005) (13.090)Age 0.043* 0.117*** 0.370(0.023) (0.027) (1.557)Distance -0.000 -0.007*** 50.834(0.002) (0.002) (28.797)Student 0.004 0.031 0.103(0.112) (0.157) (0.304)Single 0.585*** 0.300*** 0.383(0.099) (0.115) (0.486)Female -0.173*** -0.390*** 0.511(0.075) (0.092) (0.500)Kids -0.241*** 0.100 0.425(0.085) (0.100) (0.494)Vocational educ. 0.012 0.050 0.366(0.160) (0.184) (0.482)Secondary educ. 0.089 -0.004 0.373(0.143) (0.175) (0.484)Tertiary educ. 0.273* 0.164 0.180(0.164) (0.199) (0.384)English 0.461*** -0.237*** 0.357(0.097) (0.118) (0.479)German 0.177*** 0.459*** 0.452(0.080) (0.101) (0.498)Other foreign lang. 0.206*** -0.119 0.646(0.095) (0.112) (0.478)Network 0.739*** 0.712*** 0.435(0.082) (0.099) (0.496)Previous mobility 0.988*** 0.772*** 0.097(0.091) (0.111) (0.296)Commuter -0.191*** 0.126 0.243(0.095) (0.102) (0.429)Second wave -0.4 -0.762*** 0.511(0.248) (0.386) (0.500)Constant -2.521*** -2.384*** —(0.306) (0.361) —Log-likelihood -2,146.622Observations 9,063 9,063Notes: ***significant at 1%, **significant at 5%, *significant at 10% level. Standard errorsin parentheses. Region dummies and region–wave interactions not reportedSource: LAMO household surveys 2004/05 and 2006/07, WIFO calculationsFurthermore, the presence of networks and previous mobility experience increase thewillingness to commute less strongly than the willingness to migrate, while persons whofeel deprived relative to their reference group (of friends and relatives) are more likely toWIFO 93

e willing to commute rather than being willing to migrate. This can be explained by thefact that people may choose to work abroad and then use the higher income to increasetheir social status relative to their reference group at home, rather than also moving theirresidence abroad, which may entail changing their reference group with a prioriambiguous effects on social status.Table 6.10: Marginal effects and discrete change in probabilities of commuting, migrationand staying by independent variables.Willingness to Migrate Willingness to Commute No Willingness to be mobileContinuous variablesMarginaleffectStandardErrorMarginaleffectStandardErrorMarginaleffectStandardErrorAge -0.001*** (0.000) -0.001*** (0.000) 0.001*** (0.000)Deprivation 0.001 (0.001) 0.002*** (0.001) -0.003*** (0.001)Distance 0.000 (0.000) -0.001*** (0.000) 0.001** (0.000)Dummy variablesStudent 0.000 (0.004) 0.001 (0.003) -0.001 (0.005)Single 0.021*** (0.004) 0.005** (0.003) -0.026*** (0.005)Female -0.005** (0.002) -0.007*** (0.002) 0.012*** (0.003)Kids -0.008*** (0.003) 0.002 (0.002) 0.005 (0.003)Vocational educ. 0.000 (0.005) 0.001 (0.004) -0.001 (0.006)Secondary educ. 0.003 (0.005) 0.000 (0.003) -0.003 (0.006)Tertiary educ. 0.010 (0.007) 0.003 (0.004) -0.013 (0.008)English 0.017*** (0.004) -0.005** (0.002) -0.012*** (0.005)German 0.005* (0.003) 0.009*** (0.002) -0.014*** (0.003)Other foreign lang. 0.006** (0.003) -0.003 (0.002) -0.004 (0.004)Network 0.025*** (0.003) 0.014*** (0.002) -0.039*** (0.004)Previous mobility 0.054*** (0.008) 0.020*** (0.005) -0.074*** (0.010)Commuter -0.006** (0.003) 0.003 (0.002) 0.003 (0.003)Second wave -0.012 (0.008) -0.015** (0.009) 0.027** (0.012)Notes: Marginal effects and discrete probability changes computed at mean of independent variables based onmultinomial probit regression (see table 4.).*** significant at 1%, **significant at 5%, *significant at 10% level. Regiondummies and region–wave interactions not reportedSource: LAMO household surveys 2004/05 and 2006/07, WIFO-calculationsIn addition, the results point to a number of further interesting facts. First, the coefficientsof the education dummy variables are insignificant throughout, which suggests thatboth potential commuters and potential migrants are neither positively nor negativelyselected. Second, the dummy variable for interviews conducted in the second wave issignificantly negative only for the willingness to commute. This suggests that in the timeperiod from 2004/2005 to 2006/2007, only the willingness to commute across bordershas fallen in the regions under investigation. Third, contrary to our theoretical expectations,we find that commuting in the home country reduces (rather than increases) thewillingness to migrate. Furthermore, while it has the expected positive impact on thewillingness to commute, it is insignificant in the commuting equation.WIFO 94

While these results point to strong differences between potential commuters, migrantsand stayers which are rooted primarily in differences in the costs of commuting andmigration, the coefficients reported in table 6.9 have the interpretation of increases inrelative probabilities. These do not necessarily lend themselves to assessing thequantitative impact of the variables. Therefore we also computed (for continuousvariables) marginal effects on as well as (for dummy variables) discrete changes in theprobabilities of being willing to commute, willing to migrate, or stay which are reported intable 6.10.The results in table 6.10 suggest that the willingness to commute increases with previousmobility experience and the presence of friends or family abroad a well as Germanlanguage knowledge. According to the results an otherwise average person with previousexperience of mobility has a 2 percentage point higher probability of being willing tocommute than an average person without such an experience. Similarly, the presence ofnetworks abroad increases the probability of being willing to commute abroad by 1.4percentage points, while knowledge of the German language increase this probability by0.9 percentage points.Figure 6.2: Marginal effects of age on the probabilities of migration and commuting vs.stayingNote: Marginal effects calculation based on multinomial probit regression (see table 4.8). Shadedareas represent 95% confidence interval of marginal effectFor the willingness to migrate abroad by contrast, aside from previous experience withworking abroad and the presence of network effects – which have substantially higherWIFO 95

marginal effects (of 5.4 and 2.5 percentage points) than for the willingness to commute –being single and knowledge of English also have an important impact. Otherwise averagesingles are by 2.1 percentage points more willing to migrate than individuals living in apartnership and knowledge of the English language increases the probability to commuteby 1.7 percentage points.Finally, for the probability of being unwilling to migrate or commute, the discrete changesin probability are highest for individuals with previous experience of working abroad(–7.4 percentage points), networks (–3.9 percentage points), singles (–2.6 percentagepoints) and for persons who know English (–1.2 percentage points) or German (–1.4percentage points).Figure 6.3: Marginal effects of deprivation on the probabilities of migration andcommuting vs. stayingNote: Marginal effects calculation based on multinomial probit regression (see table 4.8). Shadedareas represent 95% confidence interval of marginal effectTable 6.10 also reports the marginal effects of the continuous variables age, deprivationand distance on the probability of being willing to migrate, commute and stay. Thesemeasure the percentage point increase in the probability of being willing to commute,willing to migrate or to stay arising from an incremental change for an otherwise averageindividual. Because of the nonlinear fashion of the estimator, marginal effects are notconstant for all values of the continuous variables. Therefore, the marginal effects of age,deprivation and distance on the probabilities of being willing to migrate and being willingto commute were evaluated for all observed values of the continuous variables. TheWIFO 96

esults are depicted in figures 6.2 to 6.4. In addition, 95% confidence intervals were alsoincluded for the marginal effects as shaded areas, with darker sectors markingoverlapping confidence intervals of the marginal effects on Pr(I=M) and Pr(I=C).Figure 6.2 shows that the marginal effects of age on the probabilities of being willing tomigrate and commute are significantly negative. The marginal effect is larger for thepropensity to migrate than for the propensity to commute: the probability of being willingto migrate is highest for young individuals (about 0.055 for an average person age 18)but decreases sharply with age. The negative impact of an incremental year of age on thewillingness to migrate is highest for young individuals, and declines with increasing age.The marginal effect on the willingness to commute on the other hand is also highest foryoung individuals, but varies much less with age.Figure 6.4: Marginal effects of distance on the probabilities of migration and commutingvs. stayingNote: Marginal effects calculation based on multinomial probit regression (see table 4.8). Shadedareas represent 95% confidence interval of marginal effectThe marginal effect of the relative deprivation variable on the probability of being willingto commute (see figure 6.3) increases steeply with subjective deprivation: the effect islargest for highly deprived individuals. On the other hand, the marginal effect on thepropensity to migrate is only significant at low deprivation values, and becomesinsignificant for deprived individuals. Finally, as can be seen from figure 4.4, the marginaleffect of distance on the propensity to commute is highest for individuals living close tothe EU 15, but the impact of an incremental kilometre of distance to the EU 15 declinesWIFO 97

the further away the individual lives from the border. The marginal effect of distance onthe willingness to migrate on the other hand is insignificant and thus statistically notdistinguishable from zero.WIFO 98

7 Motives, expectations and preferences of potential migrants andcommuters in the CENTROPE RegionThe LAMO data also offer insights into the motives, preferences and expectations ofpotential migrants and commuters. This is of interest because especially before easternenlargement it was often argued that it is not only economic motives which drive crossbordermobility but also non-economic motives such as opportunities for education ortraining or networks abroad. The literature on migration typically distinguishes betweenpull factors, i.e., features of the recipient country (such as high wages or betterconditions of living), and push factors, i.e., characteristics of the sending country (suchas the political or economic situation), with the relative importance of these factors formigration flows being under dispute. Data such as those obtained by the LAMO projectenable identification of the relative importance of these factors.In addition, econometric studies, which attempted to quantify the migration flows afterEU enlargement based on the historical distribution of migrants and commuters, alwayscame to the conclusion that Austria and Germany would receive the largest share ofmigrants and that most of them would settle in the urban regions of the recipientcountries. Instead, the imposed transitional periods have caused a diversion of migrationflows from the new EU member states to the EU 15: in the first two years afterenlargement, substantially more people moved to Ireland and the United Kingdom (dueto these countries granting free access to their labour markets) than had been expectedex ante. In contrast, the number of people moving to Austria and Germany wassubstantially lower due to the transitional periods. The data used in this deliverable cangive an idea on these shifts in mobility preferences.Finally, another argument in the enlargement discussion was that the expectations ofpotential migrants regarding income and working conditions in the recipient countrieswere, at least in part, unrealistic and could give rise to "irrational" migration. On thisissue the survey data can again provide an assessment of actual expectations bypotential migrants.7.1 Mobility motivesLooking at the motives given by those considering cross-border migration or commuting(Figure 7.1), economic pull factors such as better earnings, a higher standard of living orbetter working conditions abroad rank highest in both waves. Good employmentprospects in the recipient region were also among the top five reasons stated. Of theclassical push factors, only the lack of "improvement of the economic situation in thehome country" was found among the five main reasons given by individuals consideringmigration in the first wave. This motive however lost significance in the second wave inrelative and in absolute terms. This could be attributed to improving labour market andeconomic conditions in the home countries of the prospective migrants and commuters.Many interviewees also seek new experiences abroad. The relative and absoluteimportance of this factor even increases over time: in 2006-2007 it was among the fiveWIFO 99

top-ranking motives. In contrast, "classical” motives like family reunion rank at the lowerend of the scale in both waves.Figure 7.1: Motives for moving abroad2004-2005Better earningsHigher living standardBetter working conditionsGood employment prospectsNo economic improvement at homeGet new experience and meet new peoplePositive experience by othersBetter career prospectsBad political and economic situation at homeBetter further education/training opportunities abroadGreater personal and political freedomFriends or acquaintances are already abroadBad environmental conditions at homeLost jobTotalSlov akiaHungaryCzech RepublicFamily or relatives are already abroadFeel discriminated against1.0 1.5 2.0 2.5 3.0 3.5 4.02006-2007Better earningsHigher living standardBetter working conditionsGood employment prospectsGet new experience and meet new peoplePositive experience by othersBetter career prospectsNo economic improvement at homeBad political and economic situation at homeBetter further education/training opportunities abroadGreater personal and political freedomFriends or acquaintances are already abroadLost jobFamily or relatives are already abroadBad environmental conditions at homeFeel discriminated againstTotalSlov akiaHungaryCzech Republic1.0 1.5 2.0 2.5 3.0 3.5 4.0Source: LAMO household surveys in 2004-2005 and 2006-2007, WIFO calculations. - Base: general mobilitypotential. Categories: 1 "does not matter ", 2 "less important ", 3 "important", 4 "very important ".Generally thus, traditional economic pull factors provide the main motives for mobility.This is further confirmed by the fact that some of the traditional push factors, such as jobloss, discrimination or education/training, rank at the lower end of the scale. Adeteriorating environmental situation or weak political and economic conditions in thehome country are of average importance only.WIFO 100

Figure 7.2: Motives for staying in the home country2004 -2005Have family, friends and acquaintances at homeThis is where I am at home and where I know my way aboutOwn a house, garden or other real propertyWorking abroad has no attractions for meKnow no foreign languageAge, health or other personal reasonsHave no contacts abroadHave a good job at homeCosts are too high abroadToo much troubleRed tape too complicatedDifficult to get a work permitTotalSlov akiaHungaryCzech RepublicFear of a strange surroundingBad experience of others abroadXenophobia1.0 1.5 2.0 2.5 3.0 3.5 4.02006-2007Have family, friends and acquaintances at homeThis is where I am at ho me and where I know my way aboutOwn a house, garden or other real propertyWorking abro ad has no attractions for meKnow no foreign languageAge, health or o ther perso nal reaso nsHave no contacts abro adHave a go od jo b at homeCosts are too high abro adTo o much troubleRed tape too co mplicatedDifficult to get a wo rk permitTotalSlov akiaHungaryCzech RepublicFear of a strange surro undingB ad experience of o thers abro adXeno phobia1.0 1.5 2.0 2.5 3.0 3.5 4.0Source: LAMO household surveys in 2004-2005 and 2006-2007, WIFO calculations. - Base: persons outside thegeneral mobility potential. Categories: 1 "does not matter ", 2 "less important ", 3 "important", 4 "very important".Motives for cross-border mobility are ranked similarly by the potentially mobile in allcountries of the CENTROPE region as well as by commuters and migrants 62 . Nevertheless,Hungarian respondents tended to accord push factors (such as a bad political and62 In this, the potentially mobile from the new EU member states are distinct from those in developingcountries, where similar surveys have found an overwhelming prevalence of push factors.WIFO 101

economic situation or the lack of economic improvements) more importance in 2006-2007 than in 2004-2005. In Slovakia on the other hand, pull factors (such as bettereducation/training opportunities abroad, more personal and economic freedom, ornetworks) were of greater importance than in the average of the CENTROPE regions.In sum, those willing to migrate or commute in the CENTROPE regions of the new EUmember states constitute a group that is strongly drawn by the better economicconditions in the recipient region, while the political and economic situation back homeand personal reasons (except for family reasons) parameters of considerable importancein other surveys appear to exert less of an impact on the decision to become mobile.When considering those unwilling to move the motivational situation is entirely different.In both surveys (see Figure 7.2), key motives for non-mobility are personal factors andnon-monetary costs, such as the fear of losing family and personal networks, the feelingof affinity to one's home country and knowledge of relevant local factors. This highlightsthe importance of location-specific insider advantages as an explanatory factor for nonmobility,as well as the relevance of uncertainty as a major barrier to mobility. Amongthe monetary factors identified were real estate assets (ownership of a house, home orgarden, etc.) or the lack of investments in human capital, like foreign language skills.Personal factors are thus the greatest barrier to mobility in the CENTROPE regions of theNMS, while less importance is accorded to institutional barriers, such as the difficulty ofgetting a work permit.7.2 Choice of country and region of work7.2.1 Country preferencesWhile earlier econometric studies (e.g. Boeri Brücker, 2001) concluded that about twothirds of the NMS’ migration potential plan to migrate to Germany or Austria, this is nolonger the case according to the results of the LAMO survey. According to this data theproportion of those willing to migrate to Germany and Austria is about 40% (first wave:40.7%; second wave: 39.5%). On the other hand, the share of potential migrantspreferring the United Kingdom is substantially higher (by about a quarter) than in earliersurveys (first wave: 20.9%; second wave: 24.3%). This shift is less obvious whenlooking at the general mobility potential since this includes also a large share of potentialcommuters, where the share of those who prefer Austria as their preferred target countryis about 65% in both waves and thus strikingly high 63 .These differences can be attributed to a number of factors:- First, contrary to Austria and Germany, the UK did not make use of transitionperiods on the labour market.WIFO 102

Table 7.2: Targets of commuters and migrants, by regions and education levelsElementaryschoolApprenticeship/technical collageSecondary school University Totallevel) 1(university entrance2004-2005Capital 25.2 13.6 12.9 10.7 14.6Other urban area 9.8 7.5 6.4 6.9 7.3Near-border rural area 14.6 23.1 9.3 4.4 11.9Don't know/don't care 50.4 55.8 71.4 78.0 66.2Commuting potential in percentCapital 16.7 7.2 14.5 15.8 12.9Other urban area 5.6 8.4 3.5 10.8 6.5Near-border rural area 31.5 50.0 34.4 25.0 36.6Don't know/don't care 46.3 34.3 47.7 48.3 44.0Mobility potential in percentCapital 22.6 10.2 13.6 12.9 13.8Other urban area 8.5 8.0 5.1 8.6 7.0Near-border rural area 19.8 37.4 20.6 13.3 22.9Don't know/don't care 49.2 44.4 60.7 65.2 56.32006-2007Migration potential in percentMigration potential in percentCapital 30.8 18.9 34.6 44.7 33.2Other urban area 7.7 7.1 12.4 10.5 10.0Near-border rural area 11.5 25.2 14.1 11.8 15.4Don't know/don't care 50.0 48.8 38.9 32.9 41.3Commuting potential in percentCapital 9.1 5.2 18.5 23.2 12.9Other urban area 5.5 4.3 3.3 1.8 3.8Near-border rural area 56.4 53.0 50.0 35.7 49.7Don't know/don't care 29.1 37.4 28.3 39.3 33.6Mobility potential in percentCapital 23.3 12.4 30.1 38.9 26.3Other urban area 6.9 5.8 9.8 8.2 7.9Near-border rural area 27.0 38.4 24.2 18.3 27.1Don't know/don't care 42.8 43.4 35.9 34.6 38.7Notes:Base: general potentials.Source: LAMO household surveys in 2004-2005 and 2006-2007, WIFO-calculations.- Second, foreign language education in the NMS changed substantially over thepast years, with English taking over German as the primary foreign languagetaught in school. As a result, younger migrants (which make up a substantial partof the migration potential) are more fluent in English than in German.- Finally, public debate in Germany and Austria before EU accession may havecaused some potential migrants to feel unwelcome in these countries, which againcould have contributed to the change in country preferences.63 Nevertheless this is not really surprising considering that these surveys were conducted in the borderregions of CENTROPE.WIFO 103

Figure 7.3: Motives for country preference by recipient country: Austria, United Kingdomand overall2004- 2005Knowledge of the languageGood payGeographical proximityFamily members, relatives, friendsor acquaintances already live thereQuiet, secure and politically stablecountryOpportunities for furthereducation/trainingGreat BritainAustriaTotalOther reasonsGood labour marketSimilar culture/mentalityEasy to get residence/work permit0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.02006-2007Good payGeographical proximityKnowledge of the languageFamily members, relatives, friendsor acquaintances already live thereOpportunities for furthereducation/trainingOther reasonsGreat BritainAustriaTotalGood labour marketQuiet, secure and politically stablecountrySimilar culture/mentalityEasy to get residence/work permit0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0Source: LAMO household surveys in 2004-2005 and 2006-2007, WIFO calculations. − Base:general mobility potential. Multiple choices allowedThe motives for the choice of target countries confirm some of these hypotheses. Thus, acomparison of motives for choosing the United Kingdom and Austria respectively (seeWIFO 104

Figure 7.3) shows that those who prefer Austria do so mainly because of its geographicalproximity (which obviously more important for commuters than for potential migrants seeFigure 7.4) and its high wage level. All other motives, such as language skills, residentfamily members, relatives or friends, education or training opportunities as well as therelative easiness of obtaining a residence or work permit seem to speak for the UnitedKingdom.The survey thus points at a shift in country preferences, especially among potentialmigrants, which can be attributed to the transition periods, but also to other factors. Thistrend has grown stronger over time, as evidenced by the recent growth in migration flowsfrom the NMS to the UK.7.2.2 Regional preferencesApart from country preferences, the LAMO survey also provides information regardingregional preferences within the respective target countries (see Tables 7.2 and 7.3).Potential migrants tend to prefer urban regions (capital or other city/town), whilepotential commuters are more likely to also consider rural regions near the border. This isobvious from the data for 2006-2007, where the percentage of potential migrants whowant to move into urban regions has more than doubled since 2004-2005, while thefigure has remained approximately constant among potential commuters. However, thisresult is affected by differences in the proportion of respondents who did not specifyconcrete target regions between the waves. Especially Slovak respondents did not alwaysspecify regional preferences in the first wave (75.5%). Similarly, more than half of thosesurveyed in Hungary in 2004-2005 gave no information on their target region preference,compared to about one in three respondents in the Czech Republic. Differences betweenwaves, as shown in Tables 7.2 and 7.3, must therefore be interpreted with caution.Studying the target preferences by educational level one can conclude that individualswith a vocational training (apprenticeship) show a significant preference for rural areas,while higher-skilled workers (secondary and tertiary education) clearly prefer urbanregions as a target for cross-border mobility, especially in 2006-2007. 64Significant differences in regional preferences can also be found between countries (Table7.3): a considerable part of the Hungarian population (potential commuters and potentialmigrants of both waves) would consider working in rural areas near the border while thegeneral migration and mobility potential from the Czech Republic clearly prefers urbanregions. These results appear to be influenced also by the existence of an Austrian-Hungarian agreement on cross-border commuting, as a result of which a quota ofHungarian workers is allowed to work in the Austrian province of Burgenland.64 However, the change that occurred in regional preferences between the two waves as shown in Table 7.2needs to be interpreted with due caution (see above).WIFO 105

Table 7.2: Targets of commuters and migrants, by regions and education levelsElementaryschoolApprenticeship/technical collageSecondary school University Totallevel) 1(university entrance2004-2005Capital 25.2 13.6 12.9 10.7 14.6Other urban area 9.8 7.5 6.4 6.9 7.3Near-border rural area 14.6 23.1 9.3 4.4 11.9Don't know/don't care 50.4 55.8 71.4 78.0 66.2Commuting potential in percentCapital 16.7 7.2 14.5 15.8 12.9Other urban area 5.6 8.4 3.5 10.8 6.5Near-border rural area 31.5 50.0 34.4 25.0 36.6Don't know/don't care 46.3 34.3 47.7 48.3 44.0Mobility potential in percentCapital 22.6 10.2 13.6 12.9 13.8Other urban area 8.5 8.0 5.1 8.6 7.0Near-border rural area 19.8 37.4 20.6 13.3 22.9Don't know/don't care 49.2 44.4 60.7 65.2 56.32006-2007Migration potential in percentMigration potential in percentCapital 30.8 18.9 34.6 44.7 33.2Other urban area 7.7 7.1 12.4 10.5 10.0Near-border rural area 11.5 25.2 14.1 11.8 15.4Don't know/don't care 50.0 48.8 38.9 32.9 41.3Commuting potential in percentCapital 9.1 5.2 18.5 23.2 12.9Other urban area 5.5 4.3 3.3 1.8 3.8Near-border rural area 56.4 53.0 50.0 35.7 49.7Don't know/don't care 29.1 37.4 28.3 39.3 33.6Mobility potential in percentCapital 23.3 12.4 30.1 38.9 26.3Other urban area 6.9 5.8 9.8 8.2 7.9Near-border rural area 27.0 38.4 24.2 18.3 27.1Don't know/don't care 42.8 43.4 35.9 34.6 38.7Notes:Base: general potentials.Source: LAMO household surveys in 2004-2005 and 2006-2007, WIFO-calculations.In addition, the LAMO questionnaire offers an opportunity to assess the preferences ofpotential commuters regarding maximum commuting times. This is especially interestingwhen considering that previous estimates of the commuting potential (see, e.g., Huber,2001, Birner and Huber, 1999, Alecke and Untiedt, 2001) assumed that most of thepotential commuters would commute a maximum of 180 to 240 minutes per day. Thedata used here indicate that (see table 7.3) some 85% of potential daily commuters inthe first wave would accept commutes of up to 150 minutes while only 15.7% wouldundertake a daily commute of more than 2.5 hours. At the time of the second survey in2006-2007, this share was down to 9.0%, however, this decline is not statisticallyWIFO 106

significant). Among potential weekly commuters, about 80% would accept a maximumweekly commute of up to 6 hours in 2004-2005, while the proportion of those willing toaccept commutes of more than 6 hours significantly declined from 20.2% in 2004-2005to 12.7% in 2006-2007.Table 7.3: Maximum commuting times, daily and weekly commutersDaily commutersWeekly commuters2004-2005 2006-2007 2004-2005 2006-2007PercentUp to 1 hr 27.5 44.1 26.6 29.51 to 2.5 hrs 56.8 46.9 53.2 57.8More than 2.5 hrs 15.7 9.0 20.2 12.7Notes:Base: general potentials.Source: LAMO household surveys in 2004-2005 and 2006-2007, WIFO-calculations.7.3 Length and timing of migrationIn previous studies it was often not possible to assess the period within which themobility potential might be realised. Table 7.4 depicts the periods within which potentialmigrants want to take up work abroad. Since it is reasonable to assume that individualswith more concrete mobility intentions also have more concrete ideas with regard totiming, we distinguish by mobility potentials. This allows more differentiated observations,but also leads to very low numbers of observations from the real mobilitypotential 65 , so that any further subdivision (e.g. by sending or receiving countries) wouldnot be useful.The findings are in line with a priori expectations. Among those in the general mobilitypotential, mobility intentions are rather vague: 44.8% of respondents in 2004-2005, whowere in the general mobility potential but had not yet taken any preparatory stepstowards taking up work abroad (i.e. those in the general migration potential withoutthose in the expected mobility potential), were unable to state when they intended tomove abroad. Nevertheless, the data also show that the mobility intentions took moreconcrete forms between 2004-2005 and 2006-2007: the share of undecided in thegeneral mobility potential fell significantly to 29.0%.65 Altogether the real migration potential consists of just 128 observations in the first wave, and only 119 inthe second wave.WIFO 107

Table 7.4: Preferences on when to start working abroad, by mobility potentials2004-2005General 1 Expected 2 Real General (total)In the next six months 4.7 16.7 46.5 13.0In 6 to 12 months 12.2 13.8 19.8 13.5In 1 to 2 years 15.8 17.6 11.6 15.8In 3 to 5 years 15.8 19.5 4.7 15.5After 5 years 12.2 8.1 2.3 9.9Don't know yet 39.4 24.3 15.1 32.3Commuting potential in percentIn the next six months 7.2 20.0 40.5 12.6In 6 to 12 months 8.2 11.4 16.7 9.6In 1 to 2 years 14.0 20.7 9.5 15.3In 3 to 5 years 13.8 13.6 9.5 13.4After 5 years 6.3 6.4 4.8 6.2Don't know yet 50.5 27.9 19.0 43.0Mobility potential in percentIn the next six months 5.9 18.0 44.5 12.8In 6 to 12 months 10.3 12.9 18.8 11.8In 1 to 2 years 14.9 18.9 10.9 15.6In 3 to 5 years 14.8 17.1 6.3 14.6After 5 years 9.3 7.4 3.1 8.2Don't know yet 44.8 25.7 16.4 37.12006-2007Migration potential in percentMigration potential in percentIn the next six months 3.5 8.8 36.8 8.8In 6 to 12 months 7.9 27.7 31.6 15.2In 1 to 2 years 23.8 25.5 18.4 23.5In 3 to 5 years 25.0 16.1 7.9 20.9After 5 years 14.6 3.6 0.0 10.4Don't know yet 25.2 18.2 5.3 21.2Commuting potential in percentIn the next six months 5.8 22.9 53.5 14.2In 6 to 12 months 12.9 2.9 20.9 12.9In 1 to 2 years 18.8 28.6 7.0 18.2In 3 to 5 years 16.3 8.6 4.7 13.8After 5 years 10.8 8.6 0.0 9.1Don't know yet 35.4 28.6 14.0 31.8Mobility potential in percentIn the next six months 4.3 11.6 42.9 10.6In 6 to 12 months 9.8 22.7 27.7 14.4In 1 to 2 years 21.9 26.2 14.3 21.7In 3 to 5 years 21.7 14.5 6.7 18.5After 5 years 13.2 4.7 0.0 9.9Don't know yet 29.0 20.3 8.4 24.8Notes:1 Excluding expected potential. - 2 Excluding real potentialSource: LAMO household surveys in 2004-2005 and 2006-2007, WIFO-calculations.Base: general mobility potential.WIFO 108

Table 7.5: Preferences regarding period of stay after taking up work abroad, by mobilitypotentials2004-2005General 1 Expected 2 Real General (total)In the next six months 4.7 16.7 46.5 13.0In 6 to 12 months 12.2 13.8 19.8 13.5In 1 to 2 years 15.8 17.6 11.6 15.8In 3 to 5 years 15.8 19.5 4.7 15.5After 5 years 12.2 8.1 2.3 9.9Don't know yet 39.4 24.3 15.1 32.3Commuting potential in percentIn the next six months 7.2 20.0 40.5 12.6In 6 to 12 months 8.2 11.4 16.7 9.6In 1 to 2 years 14.0 20.7 9.5 15.3In 3 to 5 years 13.8 13.6 9.5 13.4After 5 years 6.3 6.4 4.8 6.2Don't know yet 50.5 27.9 19.0 43.0Mobility potential in percentIn the next six months 5.9 18.0 44.5 12.8In 6 to 12 months 10.3 12.9 18.8 11.8In 1 to 2 years 14.9 18.9 10.9 15.6In 3 to 5 years 14.8 17.1 6.3 14.6After 5 years 9.3 7.4 3.1 8.2Don't know yet 44.8 25.7 16.4 37.12006-2007Migration potential in percentMigration potential in percentIn the next six months 3.5 8.8 36.8 8.8In 6 to 12 months 7.9 27.7 31.6 15.2In 1 to 2 years 23.8 25.5 18.4 23.5In 3 to 5 years 25.0 16.1 7.9 20.9After 5 years 14.6 3.6 0.0 10.4Don't know yet 25.2 18.2 5.3 21.2Commuting potential in percentIn the next six months 5.8 22.9 53.5 14.2In 6 to 12 months 12.9 2.9 20.9 12.9In 1 to 2 years 18.8 28.6 7.0 18.2In 3 to 5 years 16.3 8.6 4.7 13.8After 5 years 10.8 8.6 0.0 9.1Don't know yet 35.4 28.6 14.0 31.8Mobility potential in percentIn the next six months 4.3 11.6 42.9 10.6In 6 to 12 months 9.8 22.7 27.7 14.4In 1 to 2 years 21.9 26.2 14.3 21.7In 3 to 5 years 21.7 14.5 6.7 18.5After 5 years 13.2 4.7 0.0 9.9Don't know yet 29.0 20.3 8.4 24.8Notes:1 Excluding expected potential. - 2 Excluding real potentialSource: LAMO household surveys in 2004-2005 and 2006-2007, WIFO-calculations.Base: general mobility potential.WIFO 109

In spite of becoming more concrete, individuals in the general mobility potential typicallyshowed very vague ideas as to the time of migration. It can be expected that the most ofthem would migrate in the short term only if and when they receive an attractive offerfrom a prospective employer abroad.The situation is different for the potentially mobile who have already taken preparatorysteps towards working abroad and who are thus in the expected mobility potential.Although even in this group one out of four (first wave) and one out of five (secondwave) respondents had no concrete plans regarding the preferred timing of migration,over 30% stated that they plan to start working abroad within the next year. Accordinglythere is a substantial pool of individuals in the NMS, who may be mobilised at shortnotice, but generally speaking, a larger part has mobility preferences which are ratherfocused on the medium to long term.Those who have already taken more concrete preparatory steps towards working abroad(those in the real migration potential), are characterised by highly concrete, short-termmobility plans: 44.5% (first wave) and 42.9% (second wave) indicated that theyintended to migrate within the next six months; another 20 to 30% planned to leavewithin one year. A majority of this group may thus be included in the group of those thatcan be mobilised at short notice.Concerning the length of the mobility period (Table 7.5), some 20 to 25% of respondents(regardless of the mobility concept) stated that they intend to stay up to two years, sothat about one out of four or five potential migrants plans only a short-term stay. Yetabout 30% of those in the general migration potential and about 35% of those assignedto the expected or real migration potential intend to stay abroad for as long as possible.Adding the 5 to 10% who intend to stay for up to ten years or until retirement there is aquite substantial long-term mobility potential in contrast to the findings of many previoussurveys in the mid-1990s of 40 to 45% of the general mobility potential. The distributionof preferences hardly changed between the two waves.Interestingly, preferences for short-term mobility are slightly less distinct amongpotential commuters than among potential migrants. To compensate, preferences for along-term move are slightly higher among potential commuters. Even though thisobservation is based on a small sample, it is still surprising considering that migration(and subsequent remigration back home) comes at a higher cost while commuters canswitch between jobs at home and abroad at relatively low cost. On the other hand,migrants may prefer short-term mobility in order to evade non-monetary costs (such asthe loss of friends or location-specific insider advantages at home).The claim that migrants from the NMS intent on working abroad solely to become eligiblefor social insurance benefits, which was frequently asserted in the public debate beforeEU enlargement, seems to be of rather minor importance here: Only 1 to 2% want tostay only until they become eligible for social welfare.WIFO 110

5.4 Expectations concerning type of workThe LAMO project also surveyed expectations regarding potential workplaces abroad.Some 30 to 50% of the potentially mobile expect to get a "better" job abroad than athome, and another 20 to 30% hope for a job of the "same quality" than the one theyheld at home. Furthermore, about 40% of the interviewees expect to be employedaccording to their skill level. A comparison between the two waves shows (table 7.6) thatin 2006-2007 both the share of those in the general mobility potential who expect abetter job abroad and of those who expect their job to be worse has significantly risenover 2004-2005.Table 7.6: Expectations regarding the "quality" of the job abroadMigrantsCommutersTotal2004-2005 2006-2007 2004-2005 2006-2007 2004-2005 2006-2007Work qualityPercentSame as at home 19.7 22.0 20.8 28.0 20.2 24.1Irrelevant 13.8 19.3 15.3 20.8 14.4 19.8Better 36.8 48.1 33.1 50.3 35.1 48.9Worse 5.4 8.3 4.9 5.3 5.2 7.3Accords with the skill level 43.4 37.1 41.9 40.6 42.7 38.3Does not accord with the skill level 10.4 13.8 13.1 12.3 11.6 13.3Notes: Base: general mobility potential. Multiple choices permitted.Source: LAMO household surveys in 2004-2005 and 2006-2007, WIFO-calculationsTable 7.7: Preferences regarding sector of employment abroad, by mode of mobilityMigrantsCommutersTotal2004-2005 2006-2007 2004-2005 2006-2007 2004-2005 2006-2007PercentAgriculture and farming 9.2 13.9 15.4 20.4 12.0 16.1Fishing 1.2 1.3 2.7 1.9 1.9 1.5Mining 0.4 0.3 0.8 0.9 0.6 0.5Manufacturing 11.8 14.1 20.5 21.1 15.6 16.5Electricity, gas and water supply 3.1 0.5 3.0 1.3 3.1 0.7Construction 13.0 11.2 14.6 23.0 13.7 15.2Trade 10.7 13.0 13.6 12.9 12.0 12.9Hotels and restaurants 24.6 27.2 21.1 23.0 23.1 25.8Transport and communications 8.0 8.3 9.7 5.0 8.8 7.2Financial intermediation 13.2 3.4 6.7 0.9 10.3 2.6Real estate and business services 9.2 6.2 5.9 4.1 7.7 5.5Public administration 4.7 3.2 2.9 0.9 3.9 2.5Education 6.2 8.9 5.9 3.5 6.1 7.1Health, veterinary and social services 10.3 10.7 11.6 11.6 10.9 11.0Other services 19.2 14.3 16.3 9.7 17.9 12.7Private households 10.8 11.7 14.4 14.5 12.4 12.6Extraterritorial organisations 10.5 5.7 7.0 1.6 9.0 4.3Others/don’t know/no reply 25.1 12.5 19.6 8.2 22.7 11.0Notes: Base: general mobility potential. Multiple choices permitted.Source: LAMO household surveys in 2004-2005 and 2006-2007, WIFO-calculationsWIFO 111

Table 7.8: Wage expectations for work abroadElementaryschoolApprenticeship/technical collageSecondary school(university entrancelevel)UniversityTotal2004-2005Same 0.8 2.0 0.0 0.6 0.7Higher by 50% 4.9 5.4 4.5 3.8 4.6About double 22.0 31.3 18.6 18.9 21.82 to 3 times as high 15.4 22.4 23.5 36.5 24.74 to 5 times as high 8.1 20.4 16.7 18.9 16.5More than 5 times as high 2.4 4.8 7.1 6.3 5.7Don't know/no reply 46.3 13.6 29.6 15.1 26.1Commuting potential in percentSame 0.0 0.0 1.2 1.7 0.8Higher by 50% 5.6 3.0 3.1 3.3 3.4About double 20.4 25.9 21.1 30.8 24.32 to 3 times as high 31.5 47.0 35.9 40.8 39.64 to 5 times as high 13.0 12.7 20.7 11.7 15.9More than 5 times as high 3.7 6.0 3.5 4.2 4.4Don't know/no reply 25.9 5.4 14.5 7.5 11.6Mobility potential in percentSame 0.6 1.0 0.5 1.1 0.7Higher by 50% 5.1 4.2 3.9 3.6 4.0About double 21.5 28.4 19.8 24.0 22.92 to 3 times as high 20.3 35.5 29.1 38.4 31.44 to 5 times as high 9.6 16.3 18.5 15.8 16.2More than 5 times as high 2.8 5.4 5.5 5.4 5.1Don't know/no reply 40.1 9.3 22.8 11.8 19.62006-2007Migration potential in percentMigration potential in percentSame 0.0 1.6 0.9 5.9 2.1Higher by 50% 8.7 2.4 6.4 7.9 6.3About double 16.3 16.5 24.8 25.7 21.92 to 3 times as high 36.5 45.7 31.2 28.9 34.54 to 5 times as high 5.8 19.7 17.5 13.8 15.1More than 5 times as high 1.0 1.6 5.1 5.9 3.9Don't know/no reply 31.7 12.6 14.1 11.8 16.2Commuting potential in percentSame 1.8 0.0 0.0 1.8 0.6Higher by 50% 3.6 4.3 6.5 7.1 5.3About double 21.8 28.7 28.3 39.3 29.22 to 3 times as high 47.3 33.9 40.2 30.4 37.44 to 5 times as high 18.2 22.6 17.4 17.9 19.5More than 5 times as high 1.8 4.3 1.1 0.0 2.2Don't know/no reply 5.5 6.1 6.5 3.6 5.7Mobility potential in percentSame 0.6 0.8 0.6 4.8 1.6Higher by 50% 6.9 3.3 6.4 7.7 6.0About double 18.2 22.3 25.8 29.3 24.42 to 3 times as high 40.3 40.1 33.7 29.3 35.54 to 5 times as high 10.1 21.1 17.5 14.9 16.6More than 5 times as high 1.3 2.9 4.0 4.3 3.3Don't know/no reply 22.6 9.5 12.0 9.6 12.6Notes: Notes: base is the general mobilitySource: LAMO household surveys in 2004-2005 and 2006-2007, WIFO-calculationsWIFO 112

Table 7.9: Expectations with regard to type of work abroadElementaryschoolApprenticeship/technical collageSecondary school(university entrancelevel)UniversityTotal2004-2005Part-time 21.1 10.2 18.6 13.2 16.2Full-time 61.8 76.9 78.5 81.1 75.9Quasi-freelance/contract for works and services 12.2 17.0 17.0 15.7 15.9Self-employed/entrepreneur 20.3 5.4 15.8 14.5 14.2Commuting potential in percentPart-time 18.5 10.8 21.5 18.3 17.6Full-time 70.4 80.7 71.9 73.3 74.5Quasi-freelance/contract for works and services 14.8 10.2 22.7 24.2 18.8Self-employed/entrepreneur 9.3 7.8 7.0 12.5 8.6Mobility potential in percentPart-time 20.3 10.5 19.9 15.4 16.8Full-time 64.4 78.9 75.5 77.8 75.3Quasi-freelance/contract for works and services 13.0 13.4 19.6 19.4 17.2Self-employed/entrepreneur 16.9 6.7 11.8 13.6 11.72006-2007Migration potential in percentMigration potential in percentPart-time 15.4 11.0 16.2 10.5 13.6Full-time 73.1 74.0 74.4 71.1 73.3Quasi-freelance/contract for works and services 26.0 25.2 27.4 31.6 27.7Self-employed/entrepreneur 12.5 3.1 7.3 13.2 8.8Commuting potential in percentPart-time 14.5 14.8 15.2 16.1 15.1Full-time 69.1 77.4 69.6 76.8 73.6Quasi-freelance/contract for works and services 27.3 21.7 18.5 16.1 20.8Self-employed/entrepreneur 1.8 6.1 6.5 8.9 6.0Mobility potential in percentPart-time 15.1 12.8 16.0 12.0 14.1Full-time 71.7 75.6 73.0 72.6 73.4Quasi-freelance/contract for works and services 26.4 23.6 24.8 27.4 25.3Self-employed/entrepreneur 8.8 4.5 7.1 12.0 7.8Notes: Base: general mobility potential. Multiple choices permittedSource: LAMO household surveys in 2004-2005 and 2006-2007, WIFO-calculationsA majority of the potentially mobile in both waves (24.6%/27.2% of potential migrantsand 21.1%/23.0% of potential commuters) wants to work in the hotel and restaurantbusiness. Another 13.7% (first wave) and 15.2% (second wave) would seek employmentin the construction industry. These are usually the most typical "guest worker" sectors,also in Austria. Still, there is also a large number of individuals who want to work in themanufacturing sector (15.6% in the first wave and 16.5% in the second wave) or inagriculture and forestry (12.0% and 16.1% respectively). Only a few can imagineworking in the public sector (public administration and schools). The only non-marketservices sector to have some attraction is the health sector. The preferences are thusdominated by the "typical guest worker sectors", hotels and restaurants, construction,manufacturing and farming and forestry.The LAMO survey also contains information on reservation wages required to makepotential commuters and migrants take up a job abroad (Table 7.8) and preferences fordifferent types of work (Table 7.9). A majority of the potentially mobile would workabroad only if paid a substantially higher wage than at home: two out of three expecttheir wage abroad to be multiple times higher than what they earn now, with a relativemajority (31.4% in the first wave and 35.5% in the second wave) expecting wagesabroad to be two to three times their current wage. There is hardly any differenceWIFO 113

etween the waves of observations and between commuters and migrants. Similarly,differences between educational levels are small, especially since a large share ofrespondents in the general mobility potential could not or would not respond to thisquestion, which makes it difficult to draw a general conclusion.Thus, although the preference for migration is coupled with the expectation ofsubstantially higher wages, these expectations appear to be quite realistic given thewage differentials between the NMS and the EU15. Only about 3 to 5% of the potentialcommuters and migrants interviewed (5.1% in the first wave, 3.3% in the second wave)expect wages to increase more than fivefold, an expectation which should be difficult tomeet.Moreover, three out of four potential migrants and commuters would prefer a full-timejob (see Table 7.9), while only 15% would also work part-time. The proportion of thepotentially mobile who would accept only a full-time job is greatest in Hungary. Therewas a significant rise in the share of those who would work abroad on the basis of aquasi-freelance employment contract from 2004-2005 to 2006-2007: The proportionincreased to more than a quarter of all in the general mobility potential. The willingnessto work as a freelancer increased particularly in the Czech Republic and Hungary, while itdeclined in Slovakia.WIFO 114

8 ConclusionsThis deliverable analysed the regional distribution of migration and cross-bordercommuting in the EU27 using European Labour Force data. Furthermore a case study ofmigration and commuting potentials in one of the border regions, which can be deemedto be most affected from these flows (the border region of the new member states toAustria) was conducted by using the first two waves of the LAMO household surveyconducted in the CENTROPE region in 2004-2005 and 2006-2007.With respect to the regional structure of migration in the EU we find the largest localclusters of migrants in the EU 15 in the Île de France as well as Inner and Outer Londonand a markedly different settlement structure of migrants relative to natives: 23.9% ofall migrants would have to change their region of residence in order to achieve a uniformdistribution of migrants across EU-15 countries. Migrants from the NMS-8 show a lowerdegree of concentration than those from Bulgaria and Romania or the candidatecountries, while they are more regionally concentrated than migrants from othercountries. The biggest local clusters of NMS migrants can be observed in the Londonareas and Vienna. Looking at individual sending countries, Polish migrants show thelowest tendency to cluster regionally among migrants from the NMS. Furthermore, wefind that low skilled migrants with primary education are much more concentrated thanmigrants with secondary or tertiary education, which confirms earlier findings.The concentration of migrants did not differ substantially between migration cohorts:those who moved during the last 10 years are about as concentrated as those whomigrated earlier. However, the target regions of more recent migration waves areconsiderably different from those of earlier cohorts. This applies in particular to migrantsfrom the NMS-8, where the different institutional regimes since accession have shiftedthe target country structure of migration, which also affects the regional patterns ofmigration. Although the geographical concentration increased for more recent cohorts ofmigrants from the NMS-8, the correlation of local concentrations across time is rather lowand even insignificant for some CEE countries. However, a regression analysis showsthat—even after controlling for geographic and economic characteristics of the regions—ethnic networks do play a significant role in explaining the locational choice of migrants.With respect to the extent of cross-border commuting in the EU 27 we find that this ingeneral is limited to individual border regions and has a relatively low magnitude whenconsidering the overall European labour market. In the two years observed cross-bordercommuters accounted for only 0.5% of total employment in the EU. In particular crossbordercommuting is of relevance in a small number of border regions, located at theexternal border of the EU, the German-French and French Belgian borders, on the Austro-German border, at the Czech-Slovak border, in the Baltic countries and in WesternHungary as well as the German-Polish border and potentially southern Sweden, which aremostly characterised by strong linguistic, historic or institutional ties, only. In theseregions usually more than 1% of the employed commute across borders and in individualcases cross-border commuting may surpass the 5% mark. For most other border regionsWIFO 115

outside these "hot spots” out-commuting is below 0.5% of the employed. In sum theextent of commuting is small in the EU, but there is some variance among regions.There are also some differences in the importance of cross-border commuting betweenthe EU 15 and NMS 12. In particular, inbound cross-border commuting as a percentage ofthe employed in the country of work, is substantially lower in the NMS 12 than the EU 15countries. In addition outbound cross-border commuting from the NMS 12 is stronglyoriented towards the EU 15 countries rather than non-EU countries. This can be explainedby the fact that most non-EU countries that are close enough to the NMS 12 to bedestinations for cross-border commuting have substantially lower income levels than theNMS 12. By contrast, outbound cross-border commuting in the EU 15 is more stronglyoriented to non-EU countries rather than to the NMS 12. Again, this can be explained bythe differences in income levels.Our results also indicate that cross-border commuters - in contrast to internal commutersin the EU 27 - are not in general better qualified than non-commuter and are drawn morethan proportionately from manufacturing workers, males and the age group of the 20 to29 year olds. Furthermore, these characteristics apply even more strongly to cross-bordercommuters from the NMS 12 than to commuters from the EU 15. While these results arelargely consistent with the findings of earlier case studies in the literature, they alsosuggest that cross border commuters – in contrast to migrants – are not as stronglypositively selected on educational criteria, but stem primarily from the intermediatequalification level.Finally, - while our results in this respect are subject to a rather unsatisfactory datasituation, our findings also imply that after controlling for other influences on crossbordercommuting - flows from the NMS 12 to the EU 15 are not significantly smallerthan those among the EU 15 countries, while flows from the EU 15 to the NMS 12 aresignificantly lower than those among the EU 15. The primary difference in the factorsdetermining cross-border migration in the NMS 12 and the EU 15 seems to be a closerassociation of cross-border commuting with the industrial specialisation in the NMS 12than the EU 15.The case study of the CENTROPE region indicates that 10.9% of the interviewed in theCENTROPE regions of the Czech Republic, Hungary or Slovakia expressed the wish tomigrate to one of the EU 15 countries in the future (and thus belonged to the generalmigration potential). Furthermore, 3.8% of the population in the region were willing tomigrate and had either already collected information about their respective targetcountry, taken training courses, learned the language, applied for a residence or workpermit or for a job, or already had a confirmed job offer or a place to live and thereforebelonged to the expected migration potential. 1.3% of the population had applied for awork permit and or already had a job offer abroad (real migration potential) in 2006-2007.An additional 5.6% of the population in the region under consideration expressed thewish to commute to the EU 15 in the future (and thus belonged to the generalWIFO 116

commuting potential). 1.4% of the population in the region were willing to commute andhad either already collected information about their respective target country, takentraining courses, learned the language, applied for a residence or work permit or for a jobor already had a confirmed job offer (i.e. belonged to the expected commuting potential).Finally, 0.8% of the population had applied for a work permit and or already had a jobabroad (real commuting potential) in 2006-2007.Relative to the first wave of interviews in 2004-2006 this represents a decrease in themigration potential of between 1.5 percentage points (for the general migrationpotentials) and 0.1 percentage points (for the real migration potential). Furthermorecommuting potentials declined more strongly for the general and expected commutingpotentials, while the real commuting potential increased slightly.A comparison with the Austrian subregions of this border region - for which data wascollected in the 2004-2005 wave only – suggests that the general migration potential inAustria is as high as in the average of the NMS-regions. We interpret this as indicationthat the general migration and commuting potentials are very broad concepts whichexpress vague wishes rather than real intentions and therefore must not be equated withactual or future migration: Only a small proportion of those who generally considerworking abroad will actually do it.The expected and real mobility potentials (migration potential plus commuting potential)in the Austrian part of this border region are thus lower than in the new EU memberstates (by 2.6 percentage points for the expected and by 1,2 percentage points for thereal mobility potentials), primarily on account of the fact that commuting from Austria tothe new member states is less attractive than commuting from the new member states toAustria because of wage differences.In addition, the general decline in migration and commuting potentials in the NMSregionswas associated with relatively dissimilar developments in the individual countries:- The general mobility potential in Slovakia decreased from 37.4 to 14.7%. Thisdecrease was particularly pronounced in the general commuting potential (from17.4% to 2.7%). Compared to 2004-2005, the general migration potential alsodecreased substantially by 8 percentage points to 12.0% in 2006-2007. Similarly,the expected mobility potential in Slovakia was less than half of its 2004-2005value (15.4%) in the 2006-2007 survey. The real mobility potential decreased byapproximately a third (first wave: 3.6%, second wave: 2.4%).- In Hungary, the general mobility potential showed an opposite development: Duemainly to a higher general migration potential (2004-2005: 7.5%, 2006-2007:12.0%), the general mobility potential increased significantly, from 19.5% to25.3%. The expected mobility potential declined also in Hungary (6.8 to 5.4%). Asignificant rise was observed in the real mobility potential, which doubled between2004-2005 and 2006-2007, from 1.3% in the first wave to 2.6% in the second.WIFO 117

Thus, Hungary was the only country with more nationals having undertakenconcrete steps to work abroad in 2006-2007 than two years earlier.- The lowest general mobility potential can be found in the Czech regions. In thesecond wave it declined further, from 15.9 to 13.8%. The changes in the expectedand real mobility potentials ( 0.2 percentage points to 4.5% and 0.1 percentagepoints to 1.7%, respectively) were however not statistically significant.We also use these data to analyse the determinants and structure of potential commutersand migrants. Our descriptive as well as econometric evidence suggests that inparticular, the presence of kids or a spouse in the household is a more seriousimpediment for the willingness to migrate than for the willingness to commute. Thissuggests that kids and a partner in the household increase cross-border migration costsmore strongly than cross-border commuting costs. Furthermore, gender differences in thewillingness to commute are larger than for the willingness to migrate (although womenare both significantly less willing to commute and to migrate), and the willingness tomigrate reduces much more strongly with age than does the willingness to commute.While migrants may thus be considered to be composed mainly of young singles,commuters are more likely to be older and have a family. Furthermore, commuters maybe expected to be more strongly selected by gender.We also find that both those willing to commute as well as those willing to migrate aredisproportionately often drawn from the two extremes of the educational distribution, andare thus often either highly or less educated. When, however, including education in amultivariate regression analysis we find that education has no significant effect on boththe willingness to migrate and to commute, which we take to imply that - at least in theregion analysed - potential migrants as well as potential commuters are neither positivelynor negatively selected.The willingness to migrate decreases much more rapidly with distance to the nearestpotential workplace abroad than the willingness to migrate while the latter is positivelyinfluenced by English and other foreign language knowledge. The willingness to commuteis, however, more strongly associated with German language knowledge. In addition, thewillingness to migrate is also more strongly influenced by the presence of networks andprevious experience of working abroad than the willingness to commute.Finally, when analysing the changes in the preferences associated with the willingness tomigrate and commute, we find that in contrast to similar research conducted beforeaccession there is a striking difference with respect to the choice of country of work ofpotential migrants. In our data the proportion of those willing to migrate to Germany andAustria is about 40% (first wave: 40.7%; second wave: 39.5%) and thus substantiallylower than in previous studies. On the other hand, the share of potential migrantspreferring the United Kingdom is substantially higher than in earlier surveys (first wave:20.9%; second wave: 24.3%). This shift is most obvious when looking at the migrationpotential but less so when looking at potential commuters, where the share of those whoWIFO 118

prefer Austria as their preferred target country is about 65% in both waves of thequestionnaire.Comparing the motives for choosing the United Kingdom and Austria, respectively, showsthat those who prefer Austria do so mainly because of its geographical proximity (whichis obviously more important for commuters than for potential migrants) and its highwage level. All other motives, such as language skills, resident family members, relativesor friends, education or training opportunities as well as the relative easiness of obtaininga residence or work permit seem to speak for the United Kingdom.Furthermore we also find that:- Those willing to migrate or commute in the regions of the new EU member statesanalysed constitute a group that is strongly drawn by the better economicconditions in the recipient region, while the political and economic situation backhome and personal reasons (except for family reasons) appear to exert less of animpact on the decision to become mobile- The key motives for non-mobility, by contrast, are primarily personal factors andnon-monetary costs, such as the fear of losing family and personal networks, thefeeling of affinity to one's home country and knowledge of relevant local factors.Among the monetary factors identified real estate assets (ownership of a house,home or garden, etc.) or the lack of investments in human capital, like foreignlanguage skills, belong to the most important deterrents for mobility. Lessimportance is accorded to institutional barriers, such as the difficulty of getting awork permit.- Potential migrants tend to prefer urban regions (capital or other city/town) as aregion of work, while potential commuters are more likely to also consider ruralregions near the border. This is due to the fact that 91% of the potential dailycommuters are not prepared to accept commuting times in excess of 2.5 hoursdaily.- The relative majority of potential commuters expects substantial wage increases(of about double the amount earned currently) from mobility, would like to stay inthe receiving country for as long as possible and would like to work in regular fulltime jobs.WIFO 119

9 ReferencesAiginger, Karl. 2000. "Europe's Position in Quality Competition." Working Paper andBackground Report for "The European Competitiveness Report 2000", DG Enterprise,Brussels.Altzinger, W. 2006. "On the Earnings of Austrian Affiliates in the New EU MemberCountries." University of Economics and BA, Vienna; Department of Economics, presentedat the NOEG Annual Meeting May 5–6.Altzinger, W. 2004. "Grenzüberschreitende Direktinvestitionen" in BMWA, ÖsterreichsAußenwirtschaft, Jahrbuch 2003/2004, BMWA.Amuendo-Dorantes, C. and S. de la Rica, Immigrant’s Responsiveness to LaborMarket Conditions and Its Implications on Regional Disparities: Evidence from Spain, IZADiscussion Paper No. 1557, Institute for the Study of Labor, Bonn, 2005.Andersson, P. and E. Wadensjö, The Employees of Native and Immigrant Self-Employed, IZA Discussion Paper No. 3147, Institute for the Study of Labor, Bonn, 2007.Arbeiterkammer Österreich. 2005. "Ein Jahr EU-Erweiterung: Trends und Fakten." –Eine umfassende Analyse aus Sicht der Bundesarbeiterkammer, AK, Vienna.Astrov, Vasily. 2004. "Wirtschaftsentwicklung in den MOEL durch deutliche Konvergenzgeprägt." WIFO-Monatsberichte: Jg. 77, Heft 5.Astrov, Vasily. 2005. "Unterschiedliche Einflüsse beschleunigen Wachstum in denMOEL." WIFO-Monatsberichte: Jg. 78, Heft 5.Astrov, Vasily. 2006. "Wirtschaftswachstum hält in den MOEL an." WIFO-Monatsberichte: Jg. 79, Heft 5.Bandyopadhyay, S., Coughlin, C.C. and H.J. Wall, Review of International Economics16(1), 2008, pp. 199-213.Bartel, A.P., Where do the new U.S. immigrants live?, Journal of Labor Economics 7(4),1989, pp. 371-391.Bauer, T., Epstein, G. and I.N. Gang, What are Migration Networks?, IZA DiscussionPaper No. 200, Institute for the Study of Labor, Bonn, 2000.Bauer, T., Epstein, G. and I.N. Gang, Herd Effects of Migration Networks? The LocationChoice of Mexican Immigrants in the U.S., IZA Discussion Paper No. 551, Institute for theStudy of Labor, Bonn, 2002.Bauer, T., Epstein, G. and I.N. Gang, Enclaves, language and the location choice ofmigrants, Population Economics 18(4), 2005, pp. 649-662.WIFO 120

Beer, E. et al. 2003. "Die Erweiterung der Europäischen Union mit 1. Mai 2004: Waswird sich für uns ändern." Kommentar zu Beitrittsvertrag aus Sicht der österreichischenArbeitnehmerInnen, Arbeiterkammer Österreich, Vienna.Bertrand, M., Luttmer, E.F.P. and S. Mullainathan, Network effects and welfareculture, The Quarterly Journal of Economics 115(3), 2000, pp. 1019-1055.Bernotat K.; Snickars F. 2002 Regional Development expectations in the Öresundregion – travel patterns and cross-border mobility, paper resented at the 42 nd Congressof the European Regional science Association, Dortmund, 2002Biffl, Gudrun. "SOPEMI" Report on Labour Migration, Austria, mehrere Jahrgänge.Biffl, Gudrun et al. 1997. "Ökonomische und strukturelle Aspekte derAusländerbeschäftigung in Österreich." Studie im Auftrag des AMS, WIFO, Vienna.Biffl, Gudrun, and Julia Bock-Schappelwein. "Zur Niederlassung von Ausländerinnenund Ausländern in Österreich." WIFO, various years.Birner; A., Peter Huber, and P. Winkler. 1999. "Schätzung des Potentials anEinpendlern und Arbeitsmigranten aus den MOEL und regionale Arbeitsmarktwirkungen."Biffl, Gudrun, Peter Huber, and Ewald Walterskirchen. 2006. "Auswirkungen derÜbergangsregelungen im Bereich der Freizügigkeit und der Dienstleistungsfreiheit."WIFO, Vienna.Blonigen, B. April 2005. "A Review of the Empirical Literature on FDI Determinants."NBER Working Paper, (11299).Blom, S., Residential concentration among immigrants in Oslo, International Migration37(3), 1999, pp. 617-639.Boeri, T., and H. Brücker. 1999. (Hrsg.) "The Impact of Eastern Enlargement onEmployment und Wages in the EU Member States." Manuskript, European IntegrationConsortium, Deutsches Institut für Wirtschaftsforschung, DIW.Borjas, George J. Ethnicity, neighborhoods and Human Capital Externalities, AmericanEconomic Review, Vol 85, No 3, pp 365 -390, 1995Borjas, G.J., Immigration and Welfare Magnets, NBER Working Paper No. 6813, NationalBureau of Economic Research, Cambridge, 1998.Borjas, G.J., The Economic Analysis of Immigration, in O. Ashenfelter and D. Card (eds.)Handbook of Labor Economics, Volume 3, Elsevier, 1999.Brandel F., H. Hofer, and K. Pichelmann. 1994. "Verdrängungsprozesse amArbeitsmarkt." Institut für Höhere Studien Forschungsbericht No. 345.WIFO 121

Breuss, Fritz., and Fritz Schebeck. 1998. "Kosten und Nutzen der EU-Osterweiterungfür Österreich", WIFO-Monatberichte, 71(11): 741–750.Breuss, Fritz. 2001. "Makroökonomische Auswirkungen der EU-Erweiterung auf alte undneue Mitglieder." WIFO-Monatsberichte 74(11). 655–666.Breuss, Fritz. März 2002a. "Die Kosten der Nicht-Erweiterung der EU." WIFO-Studie,WIFO, Vienna.Breuss, Fritz. 2002b. "Benefits and dangers of EU-Enlargement." Empirica, 29(3), 245–247Breuss, Fritz, and Margit Schratzenstaller. 2004 "Unternehmenssteuerwettbewerbund internationale Direktinvestitionen." WIFO-Monatsberichte, 8/2004: 645–653.Breuss, F., G. Fink, and P. Haiss. 2004. "How Well Prepared Are the New MemberStates for the European Union?" Journal of Policy Modeling (26).Breuss, Fritz, Serguei Kaniovski, and Margit Schratzenstaller.8/2004."Steuerreform 2004/05 – Maßnahmen und makroökonomische Effekte." WIFO-Monatsberichte: 627–643.Breuss, Fritz. 2006. "Ostöffnung, EU-Mitgliedschaft, Euro-Teilnahme und EU-Erweiterung – Wirtschaftliche Auswirkungen auf Österreich." WIFO Working Papers, 270.Brücker, Herbert, and T. Boeri et al. 2000. The Impact of Eastern Enlargement onEmployment and Labour Markets in the EU Member States. Final Report, EuropäischeKommission, Brüssel.Buch T., Niebuhr A., Schmidt T.D. Stuwe, M, Grenzpendeln in der deutsch dänischenGrenzregion, IAB-regional, Nr. 4/2008, 2008Burda, M. 1995. Migration and the option value of waiting, Economic and Social Review27(1): 1–19.Cameron, G., and J. Muellbauer. 1998. "The housing market and regional commutingand migration choices", Scottish Journal of Political Economy, 45(4): 420–446.Card, David and Ethan G. Lewis (2005) The Diffusion of Mexican Immigrants Duringthe 1990s: Explanation and Impacts, NBER Working Paper 11552Cardak, Buly A. and James Ted McDonald (2004) Neighbourhood effects preferenceheterogeneity and immigrants educational attainment, Applied Economics, pp 559-572Carrington, W.J., Detragiache, E. and T. Vishwanath, Migration with endogenousmoving costs, The American Economic Review 86(4), 1996, pp. 909-930.WIFO 122

Clark, William A V; Withers, Suzanne Davies (1999) Changing Jobs and ChangingHouses: Mobility Outcomes of Employment Transitions, Journal of Regional Science, 39:4,653-73Chiswick, B.R., Lee, Y.L. and P.W. Miller, The Determinants of the GeographicConcentration among Immigrants: Application to Australia, IZA Discussion Paper No.462, Institute for the Study of Labor, Bonn, 2002.Chiswick, B.R. and P.W. Miller, Do enclaves matter in immigrant adjustment?, Citys &Community 4(1), 2005, pp. 5-35.Crozet, Mathieu (2001) Do Migrants Follow Market Potentials, manuscript, University ofParis 1 Pantheon – SorbonneCutler, David M and Edward L. Glaeser, (1997) Are Ghettos Good or Bad? QuarterlyJournal of Economics, vol. 112, no. 3, pp. 827-72DaVanzo, J. Repeat migration in the United States: who moves back and who moveson? Review of Economics and Statistics 65(4): 552–559.Duncan, O.D. and B. Duncan, A methodological analysis of segregation indexes,American Sociological Review 20(2), 1955, pp. 210-217.Edin, P.-A., Fredriksson, P. and O. Åslund, Ethnic Enclaves and the Economic Successof Immigrants—Evidence from a Natural Experiment, CEPR Discussion Paper No. 2729,Centre for Economic Policy Research, London, 2001.Edin, Per-Anders, Peter Fredrikson and Olof Alsund (2001a) Settlement Policies andthe Economic Success of Immigrants, CEPR Discussion Paper No 2370Egger, Peter. "Außenhandelsbeziehungen zwischen Österreich und den 5 MOEL." inÖROK: Regionale Auswirkungen der EU-Integration der Mittel- und OsteuropäischenLänder, ÖROK Schriftenreihe Nr. 146/IS 67–131.Egger, Peter. 2001. "Außenhandel im erweiterten Integrationsraum, PreparityTeilprojekt 4." WIFO.Eliasson, K., U. Lindgren, and O. Westerlund. 2003. "Geographical labour mobility:migration or commuting?" Regional Studies 37(8): 827-837.Epstein, Gil S (2002) Informational Cascades and Decision to Migrate, IZA WorkingPaper 445Epstein, G., Informational Cascades and Decision to Migrate, IZA Discussion Paper 445,Institute for the Study of Labor, Bonn, 2002.Epstein, G. and I.N. Gang, The influence of others on migration plans, Review ofDevelopment Economics 10(4), 2006, pp. 652-665.WIFO 123

European Commission. 2006. Enlargement two years after: An economic Evaluation,European Economy, Occasional Papers, Brüssel.Europäische Kommission. 2006. Bericht über die Anwendung der im Beitrittsvertrag2003 festgelegten Übergangsregelungen, Mitteilung der Kommission an den Rat daseuropäische Parlament, den europäischen Wirtschafts- und Sozialausschuss und denAusschuss der Regionen, Brüssel.Fassmann, H., and C. Hintermann. 1997. Migrationspotential Ostmitteleuropa.Struktur und Motivation potentieller Migranten aus Polen, der Slowakei, Tschechien undUngarn, ISR-Forschungsbericht 15, Institut für Stadt- und Regionalforschung, Vienna.Feldkircher, Martin. Regional Convergence within the EU 25: a Spatial EconometricAnalysis, Workshops: Proceedings of OeNB Workshops, New regional Economics inCentral European Economies: The Future of CENTROPE, 101 - 120.Fidrmuc, J. 2004. Labor mobility during transition: evidence from the Czech Republic,Zeitschrift für Arbeitsmarktforschung, 73(4): 437–450.Fields, Gary S. (1979) Place to Place Migration: Some new Evidence, Review ofEconomics and Statistics, Vol. LXI, pp. 21 – 32.Fischer, P., E. Holm, G. Malmberg, und T. Straubhaar. 2000., "Why do people stay?Insider advantages and immobility." HWWA Discussion Paper 112, Hamburgisches Welt-Wirtschafts-Archiv, Hamburg.Fritz, Oliver et al. 2005. "Case study of West-Transdanubia." WIFO-Projekt, WIFO.Gann, D.M. 1994 "Innovation in the Construction Sector", in Dodgson, M., Rothwell, R.(eds.), The Handbook of Industrial Innovation, Aldershot: 202–212.Grandner, Thomas. 1998. "The Impact of an EU Enlargement on Austria’s LaborMarket." in Karl Pichelmann (Hrsg.): The Economoic Consequences of EasternEnlargement of the European Union – The Austrian View, IHS, März.Greve; B. Rydberg M. 2003 Cross Border Commuting in the EU: Obstacles and Barriers,Country Report: the Sonderylland-Schlesswig Region, University of Roskildens, ResearchPapers 10/03Greve; B. Rydberg M. 2003 Cross Border Commuting in the EU: Obstacles and Barriers,Country Report: the Oresund Region, University of Roskildens, Research Papers 11/03Gottholmseder G., Theurl E. (2007) Determinants of Cross-Border Commuting: DoCross-Border Commuters within the Household Matter?, Journal of Borderland Studies,Vol. 22(2), pp 97-112.WIFO 124

Gottholmseder G., Theurl E. (2006) Nicht-PendlerInnen, BinnenpendlerInnen,GrenzpendlerInnen – Eine sozi-ökonomische Charakterisierung am Beispiel derPendlerregion Bodenseeraum, Wirtschaft und Gesellschaft, 2/2006.Gross, D.M. and N. Schmitt, The role of cultural clustering in attracting newimmigrants, Journal of Regional Science 43(2), 2003, pp. 295-318.Hazans M. (2003) Commuting in the Baltic States: Patterns, Determinants and Gains,Zentrum für Europäische Integrationsforschung, Working Paper B02/2003, BonnHeitmueller, A., Coordination failures in network migration, The Manchester School74(6), 2006, 701-710.Helmstedt, K. 1997. "Ausländische Direktinvestitionen in den Oststaaten."Bundesministerium für Wirtschaftliche Angelegenheiten, 1997.Hofer, H., K. Pichelmann, K., and A.U. Schuh. 1998. "Price and Quantity Adjustmentsin the Austrian Labor Market." Institute for Advanced Studies, Economic Series 57,Vienna.Hofer, H. 1998. "The Impact of Emigration on Host Country’s Wages undUnemployment." in Pichelmann, Karl (Hrsg.): The Economic Consequences of EasternEnlargement of the European Union – The Austrian View, IHS, März.Hofer, H., and Peter Huber. 1999. "Wage and Mobility Effects of Trade on the AustrianLabor Market." Institut für höhere Studien und Österreichisches, Institut fürWirtschaftsforschung, Vienna.Hou, Feng (2005) The Initial Destinations and Redistribution of Canada's MajorImmigrant Groups: Changes over the Past Two Decades, Manuscript, Statistics CanadaHuber, Peter (2002) Migration and Regions – The case of Austria, in Ralph RotteMigration Policy and the Economy: International Experiences, ars et unitas, Neuried,pp157 – 178.Huber, P., Mayerhofer, P., "Structural Change in the CENTROPE Region", in OeNB(ed.), New Regional Economics in Central European Economies: The Future of CENTROPE,Proceedings of OeNB Workshops, No.9, Wien, 2006, 146-179.Huber, Peter., and K. Pichelmann. 1998. "Osterweiterung, struktureller Wandel undArbeitsmärkte", Wirtschaftspolitische Blätter (4):339-349.Huber, Peter, and Gerhard Palme. 2000. "Teilprojekt 3: RegionalwirtschaftlicheEntwicklung in den MOEL, Strukturpolitik und Raumplanung in den Regionen an dermitteleuropäischen EU-Außengrenze zur Vorbereitung auf die EU-Osterweiterung." WIFO,Vienna.WIFO 125

Huber, Peter. 2001. "Auswirkungen der Erweiterung der Europäischen Union aufBeschäftigung und Arbeitsmarkt, Preparity Teilprojekt 9." WIFO, Vienna.Huber, Peter. 2001a. "Auswirkungen der Erweiterung der Europäischen Union auf dasArbeitskräfteangebot, Preparity Teilprojekt 10." WIFO, Vienna.Huber, Peter et al. 2002. "Wiener Wirtschaftsaktivitäten." WIFO, Vienna.Huber, Peter et al. 2003. "Analyse der Niederösterreichischen Wirtschaftsaktivitäten.Teil I: Analyse." WIFO, Vienna.Huber, Peter, and Herbert Brücker. 2003. "Auswirkungen und Ausnutzung vonÜbergangsfristen für die Freizügigkeit der Arbeitskräfte nach der EU-Erweiterung." WIFO,Vienna.Huber, Peter. 2004. "Intra-national Labour Market Adjustment in the CandidateCountries." Journal of Comparative Economics, vol. 32, no. 2: 248-264.Huber, Peter. 2005. "Beschäftigung und Arbeitslosigkeit im Burgenland 1995 bis 2003."WIFO-Studie, Vienna.Huber, Peter, and Peter Mayerhofer. 2006. "Structural Change in the CENTROPERegion", in OeNB (ed.), New Regional Economics in Central European Economies: TheFuture of CENTROPE, Proceedings of OeNB Workshops, No.9, Vienna. 146-179.Hudler-Seitzberger, M., and M. Bittner. 2005., "Labourmarkte Monitoring (LAMO) –Arbeitsmarktmonitoring. Entwicklung, Anwendung und Validierung einesMointoringinstruments zur regelmäßigen Beobachtung der Veränderungen amArbeitsmarkt im Zuge der EU-Erweiterung." Zusammenfassende Projektdarstellung, PaulLazarsfeld Gesellschaft für Sozialforschung, Vienna.Hudler-Seitzberger, M., and M. Bittner. 2005. "Labourmarket Monitoring (LAMO) –Arbeitsmarktmonitoring. Analyse der leitfadenorientierten ExpertInneninterviews zurAbschätzung der Veränderungen am Arbeitsmarkt im Zuge der EU-Erweiterung." PaulLazarsfeld Gesellschaft für Sozialforschung, Vienna.Hudler-Seitzberger, M., and M. Bittner. 2005 "Labourmarket Monitoring (LAMO) –Arbeitsmarktmonitoring. Zusammenfassende Projektdarstellung." Paul LazarsfeldGesellschaft für Sozialforschung, Vienna.Hunya, G., and Jan Stankovsky. July 2004. WIIW-WIFO-Database, "Foreign DirectInvestment in Central, and Eastern Europe with Special Attention to Austrian FDIActivities in this Region", WIFO.Hunya, G. May 2006. wiiw-Database on Foreign Direct Investment in Central, East andSoutheast Europe: Increasing Significance of Repatriated and Reinvested Earnings, wiiw.WIFO 126

Institute for Public Policy Research, EU Enlargement: Bulgaria and Romania – migrationimplications for the UK, IPPR, London, 2006.Jaeger, D.A., Green Cards and the location choices of immigrants in the United States,1971-2000, Research in Labor Economics 27, 2007, 131-183.Keuschnigg, Ch., and W. Kohler. 1998. "Eastern Enlargement of the EU: How Much IsIt Worth For Austria?" CEPR Working Paper, London. (1786).Keuschnigg, Ch., and W. Kohler. 1999. "Eastern Enlargement to the EU: EconomicCosts and Benefits for the EU Present Member States." Final Report, The Case of Austria,Study XIX/B1/9801, Manuscript, Brüssel.Knoll, Norbert. April 2004 "International orientierte Unternehmen in Österreich,Rahmenbedingungen für Steuerungsfunktionen und Forschungskompetenz." WIFO.Kohler, W. 1999. "Trade and wages: What Can Faktor Contents Tell Us?" Working Paper,Universität Linz, Linz.9906.Krajasits, C., G. Neuteufl and R. Steiner. 2003. "Regional Analysis of the CENTROPE-Region." ÖIR, Vienna.Krajasits, Cornelia, Giselheid Neunteufl, and Iris Wach. 2005. "RaumstrukturelleHintergrundanalyse und Auswahlverfahren für Gemeinden, Städte und Unternehemen derLAMO-Region." ÖIR, Vienna.Landesmann. 2001.Lazear, E.P., Culture and language, Journal of Political Economy 107(6), 1999, pp. 95-126.Levine, P.B. and D.J. Zimmermann, An Empirical Analysis of the Welfare MagnetDebate Using the NLSY, NBER Working Paper No. 5264, National Bureau of EconomicResearch, Cambridge, 1995.Lutz, Hedwig, and Ewald Walterskirchen. 2004. "Aktive Beschäftigung in Österreich."WIFO, Vienna.Massy, D., Arango, J., Hugo, G., Kouaouci, A., Pellegriono, A, and J. Taylor,Theories of international migration: a review and appraisal, Population and DevelopmentReview 19(3), 1993, pp. 431-466.Mayerhofer, Peter, and Gerhard Palme. 2001. "Sachgüterproduktion undDienstleistungen: Sektorale Wettbewerbsfähigkeit und regionale Integrationsfolgen."Teilprojekt 6/1 in dies. (Koord.), "PREPARITY – Strukturpolitik und Raumplanung in denRegionen an der mitteleuropäischen EU-Außengrenze zur Vorbereitung auf die EU-Osterweiterung." Forschungsprogramm des WIFO, Vienna.WIFO 127

Mayerhofer, Peter, and Gerhard Palme. (2001a), "Aspekte der regionalenWettbewerbsfähigkeit." Teilprojekt 8 in dies. (Koord.), "PREPARITY – Strukturpolitik undRaumplanung in den Regionen an der mitteleuropäischen EU-Außengrenze zurVorbereitung auf die EU-Osterweiterung", Forschungsprogramm des WIFO, Vienna, 2001.Mayerhofer, Peter, and Gerhard Palme. (2001b), "Wirtschaftspolitische Empfehlungenauf Basis der Analyseergebnisse Teilprojekt 15." in dies. (Koord.), "PREPARITY –Strukturpolitik und Raumplanung in den Regionen an der mitteleuropäischen EU-Außengrenze zur Vorbereitung auf die EU-Osterweiterung", Forschungsprogramm desWIFO, Vienna, 2001.Mayerhofer, Peter, and Gerhard Palme. 2001. "Sachgüterproduktion undDienstleistungen: Sektorale Wettbewerbsfähigkeit und regionale Integrationsfolgen,Preparity Teilprojekt 6/1." WIFO.Mayerhofer, Peter, and Gerhard Palme. 2001a. "Aspekte der Regionalen Wettbewerbsfähigkeit,Preparity Teilprojekt 8." WIFO.Mayerhofer, Peter. 2006. "Wien in einer erweiterten Union. Ökonomische Effekte derOstintegration auf die Wiener Stadtwirtschaft", LIT-Verlag, Vienna.Mincer, J. 1978. "Family migration decisions." Journal of Political Economy 86(5).Miranda, A., Migrant Networks, Migrant Selection, and High School Graduation inMexico, IZA Discussion Paper 3204, Institute for the Study of Labor, Bonn, 2007.OeNB, Pressekonferenz ad Zahlungsbilanz , 2006.http://www.oenb.at/de/stat_melders/presse/Aussenwirtschaft/Zahlungsbilanz/2006/pa_zabil_20060425_oesterreich_etabliert_sich_als_auslandsinvestor.jsp#tcm:14-41064.OeNB, Österreichische Direktinvestitionen im Ausland und ausländischeDirektinvestitionen in Österreich, Stand per Ende 2003", Sonderheft Statistiken, OeNB,2005.Overman, Henry G; and Diego Puga, (2002) Unemployment Clusters across Europe'sRegions and Countries, Economic Policy: A European Forum, no. 34, pp. 115-43Palme, Gerhard et al. 1999. "Regionale Auswirkungen der EU-Integration der MOEL."WIFO, ÖIR, Vienna.Palme, Gerhard, and Martin Feldkircher. 2006. "Economic Challenges in theCENTROPE-Region." Workshops: Proceedings of OeNB Workshops, New regionalEconomics in Central European Economies: The Future of CENTROPE, Vienna, 88-101.Peneder, Michael. 2001. "Entrepreneurial Competition and Industrial Location." EdwardElgar, Cheltenham, UK.WIFO 128

Pfaffermayr, Michael. 1999. "Ownership Advantages, foreign Production andProductivity – Evidence form Austrian Manufacturing." The Review of IndustrialOrganisation 15(4).Pfaffermayr, Michael, and Jan Stankovsky. 1999. "Internationalisierung Österreichsdurch Direktinvestitionen." WIFO.Pichelmann, K. März 1998. "The Economic Consequences of Eastern Enlargement of theEuropean Union." - The Austrian View, IHS.Pollan, Wolfgang. 2000. "Die volkswirtschaftlichen Auswirkungen der Zuwanderung vonArbeitskräften", WIFO-Monatsberichte, 73(2).Portes, Jonathan, and Simon French Simon. 2005. "The Impact of free movement ofworkers from central and eastern Europe on the UK labour market: early evidence."Department of Work and Pensions, Working Paper No. 18, London.Portnov, B.A., The effect of regional inequalities on migration: A comparative analysis ofIsrael and Japan, International Migration 37(3), 1999, pp. 587-615.Punpuing, S. 1993. "Correlates of commuting patterns." A case-study of Bangkok,Thailand, Urban Studies 30(3): 527–546.Renkow, Mitch and Dale Hoover (2000) Commuting, Migration and Rural-UrbanPopulation Dynamics, Journal of Regional Science, 40:2, 261-287Rephann, Terance J and Coomaren P Vencatasawmy, (2000) Determinants of theSpatial Mobility of Immigrants: Evidence from Sweden, Review of Regional Studies, vol.30, no. 2, pp. 189-213Rephann, Terance and Coomaren Vencatsawmy (1998) Determinants of the SpatialMobility of Immigrants in Swedem, Paper presented at the 39 th European Congress of theEuropean regional Science Association, DublinRomaní, J., J. Suriñach, and M. Artís. 2003. "Are commuting and residential mobilitydecisions simultaneous?" the case of Catalonia, Spain, Regional Studies 37(8): 813–826.Rouwendal, Jan (1999) Spatial Job Search and Commuting Distances, Regional Scienceand Urban Economics, 29:4, 491-517Rouwendal, Jan (1998) Search Theory, Spatial Labor Markets, and Commuting, Journalof Urban Economics, 43:1, 1-22Schuh, N. "Ein erster Überblick über die mögliche Bedeutung der Finanzmärkte für dieEntwicklung der CENTROPE-Region", Manuskript, Österreichische Nationalbank.Schündeln, Matthias (2007) Are Immigrants More Mobile Than Natives? Evidence fromGermany, IZA Discussion Paper 3226.WIFO 129

Sieber, Susanne. Aussenhandel 2005, WIFO-Monatsberichte, 5/2006.Sieber, Susanne. 2006. "Direktinvestitionen österreichischer Unternehmen in Ost-Mitteleuropa." WIFO-Monatsbericht 8.Smith, A. 1998. Restructuring the Regional Economy – Industrial and RegionalDevelopment in Slovakia, Studies of Communism in Transition, Edward Elgar,Cheltenham, UK.Stankovsky, Jan, and Yvonne Wolfmayr-Schnitzer. 1996. "Österreich als Standortfür Ostzentralen." WIFO, Vienna.Stankovsky, Jan. 1999. "Österreichs Direktinvestitionen in Osteuropa", WIFO-Monatsberichte, 72(2).Stankovsky, Jan. Februar 2002. "Direktinvestitionen in Osteuropa," in Puwein et al.,Auswirkungen der EU-Erweiterung auf die Wirtschaft Österreichs, WIFO-Studie.Stark, O., Patterns of Labor Migration when Workers Differ in Their Skills, in H. Giersch(ed.) Economic Aspects of International Immigration, Springer Verlag, Berlin, 1994.Straubhaar, T. 2000. "Internationale Migration. Gehen oder Bleiben: Warum gehen sowenige und bleiben die meisten?" HWWA Discussion Paper 111, Hamburgisches Welt-Wirtschafts-Archiv, Hamburg.The Economist Intelligence Unit, "Vienna and Centrope Region: an internationalbusiness perspective, Studie im Austrag der Österreichischen Nationalbank, Vienna,2005.UNCTAD. World Investment Report 2005.Untied, Gerhard et al. 2006. "Auswirkungen der EU-Erweiterung auf Wachstum undBeschäftigung in Deutschland und Ausgewählten EU-Mitgliedstaaten." Manuskript, Gefra,Münster.Van der Velde M., Janssen, M, van Houtum H. (2005) Job Mobility in the Dutch-German Reggional Labour Market: The threshhold of indifference, manuscript,Van Ommeren, J., P. Rietveld, and P. Nijkamp. 1998. "Spatial moving behavior oftwo-earner households." Journal of Regional Science 38(1): 23–41.van Ommeren, Jos; Rietveld, Piet; Nijkamp, Peter (1999) Job Moving, ResidentialMoving, and Commuting: A Search Perspective, Journal of Urban Economics, 46:2, 230-53van Ommeren, Jos; Rietveld, Piet; Nijkamp, Peter (2000) Job Mobility, ResidentialMobility and Commuting: A Theoretical Analysis Using Search Theory, Annals of RegionalScience, 34: 2, 213-32WIFO 130

Walterskirchen, Ewald, and R. Dietz. 1998. "Auswirkungen der EU-Osterweiterung aufden österreichischen Arbeitsmarkt." Vienna, WIFO.White M. 1986 Sex differences in Urban Commuting Patterns, American EconomicReview 76/2, pp. 368 - 372Winters, Paul; de Janvry, Alain; Sadoulet, Elisabeth (2001) Family and CommunityNetworks in Mexico-U.S. Migration Journal of Human Resources, vol. 36, no. 1, , pp. 159-84Wolfmayr-Schnitzer, Yvonne. 1998. (Coordination), "The Competitiveness ofTransition Countries." OECD-Proceedings.Wolfmayr-Schnitzer, Yvonne. 1999. Economic Integration and Specialisation and theLocation of Industries. A Survey of the Theoretical Literature, WIFO.Wolfmayr, Yvonne. 1999. "Economic Integration, Specialisation and the Location ofIndustries. A Survey of the Theoretical Literature." WIF0 Working Papers, 120.Wolfmayr, Yvonne. 2004. "Österreichs Außenhandel mit den EU-Beitrittsländern",WIFO-Monatsberichte, 77(4).Wolfmayr, Yvonne. 2004A. "Außenhandelsstruktur der österreichischen Industrie", inÖsterreichs Außenwirtschaft, Jahrbuch 2003/2004, BMWA.Wolfmayr, Yvonne. 2004B. "Interessante Absatzmärkte und Exportpotentiale für dieösterreichische Industrie", WIFO-Monatsberichte, 77(06): 473-490.Wolfmayr, Yvonne. 2004C. "Österreichs Außenhandel mit den EU Beitrittsländern",WIFO-Monatsbericht, 77(04): 231-249.Wolfmayr, Yvonne et al. März 2006. "Auswirkungen der neuenStrukturfondsverordnungen auf Österreichs Wirtschaftsstandort und Exporte." Vienna,WIFO.Zax, Jeffrey S (1994) When Is a Move a Migration? Regional Science and UrbanEconomics, 24:3, 341-60WIFO 131

Appendix to Chapter 3: List of NUTS 2 codes usedAT12AT13DE11DE30DE71DEA1DEA5ES30ES51ES52ES61FR10FR71FR82ITC4ITD3ITE4LU00NL33SE22UKH1UKI1UKI2Lower AustriaViennaStuttgartBerlinDarmstadtDüsseldorfArnsbergComunidad de MadridCataluñaComunidad ValencianaAndalucíaÎle de FranceRhône-AlpesProvence-Alpes-Côte d'AzurLombardyVenetoLazioLuxembourgSouth HollandSouth SwedenEast AngliaInner LondonOuter LondonWIFO 132