Proceedings in pdf format. - Sociotechnical Systems Engineering ...

departures for 1995 are taken from the World ResourcesDatabases (WRI, 2000). There are two major problemswith this dataset. Firstly, for some countries, the reporteddata are arrivals and departures for tourism only. Forother countries, the data are arrivals and departures for allpurposes. Unfortunately, it is impossible to correct forthis. Secondly, there are missing observations,particularly with regard to departures.For arrivals, 181 countries have data but 26 do not.We filled the missing observations with a statisticalmodel, viz.,(Equation 1)−7 −3 2 −5ln A = 5.97+ 2.05⋅ 10 G + 0.22T − 7.91⋅ 10 T + 7.15⋅ 10 C + 0.80ln Yi i i i i i0.97 0.96 0.07 2.21 3.03 0.09N = =2139; Radj0.54where A denotes total arrivals, G is land area (in squarekilometres); T is annual average temperature for 1961-1990 (in degrees Celsius) averaged over the country, Ctis length of coastline (in kilometres), and Y is per capitaincome. i indexes the country of destination. This modelis the best fit to the observations for the countries forwhich we do have data. The total number of touristsincreases from 55.2 million (observed) to 56.5 million(observed + modelled). The 26 missing observationsconstitute only 2% of the international tourism market.For departures, the data problem is more serious: 107countries report but 99 do not; 46.5 million departuresare reported, against 56.5 million arrivals, so that 18% ofall international tourists have an unknown origin. Wefilled the missing observations with a statistical model,viz.,(Equation 2)Di−3 2 −2ln = 1.51− 0.18Ti + 4.83⋅10 Ti − 5.56⋅ 10 Bi + 0.86lnYi − 0.23ln GiP 17.05 0.17 16.82 4.22 0.09 0.13iN = =299; Radj0.66where D denotes departures (in number), P denotespopulation (in thousands) and B is the number ofcountries with shared land borders. i indexes the countryof origin. This model is the best fit to the observationsfor the countries for which we do have data. This leadsto a total number of departures of 48.2 million, so wescaled up all departures by 17% so that the total numberof observed and modelled departures equals the totalnumber of observed and modelled arrivals.For most countries, the volume of domestic touristflows is derived using 1997 data contained in theEuromonitor (2002) database. For some other countries,we rely upon alternative sources, such as nationalstatistical offices, other governmental institutions or tradeassociations. Data are mostly in the form of number oftrips to destinations beyond a non-negligible distancefrom the place of residence, and involve at least oneovernight stay. For some countries, data in this formatwere not available, and we resorted to either the numberof registered guests in hotels, campsites, hostels etc., orthe ratio between the number of overnight stays and theaverage length of stay. The latter formats underestimatedomestic tourism by excluding trips to friends andrelatives; nevertheless, we included such data forcompleteness, relying on the fact that dropping them didnot lead to any dramatic change.In general, the number of domestic tourists is less thanthe regional population. ´However in 22 countries,residents were domestic tourists more than once per year.An examination of the characteristics of such countriesshows that these are in general rich countries, endowedwith plenty of opportunities for domestic tourism andlarge (or at least medium-sized). This definition fits inparticular Scandinavian countries (e.g., 4.8 domestictourists per resident in Sweden) but also Canada,Australia, and the USA. In the USA, the combination ofa large national area, a large number of tourist sites, highincome per capita contribute to explain why, on average,an average American took a domestic holiday 3.7 timesin 1997. Distance from the rest of the world is alsoimportant, and this is most probably the explanation forthe many domestic holidays in Australia and NewZealand.We filled the missing observations using tworegressions. We interpolated total tourist numbers, D+H,where H is the number of domestic tourists, using(Equation 3)Di+ Hiln = − 1.67+0.93lnYiP0.83 0.10iN = R =263;adj0.60Note that (3) is not limited from above. The number oftourists may exceed the number of people, which impliesthat people take a holiday more than once a year. Notethat we measure population numbers in thousands. Theparameters imply that people with an income of $10,000per person per year take one holiday per year.The ratio of domestic to total holidays wasinterpolated using(Equation 4)Hi− − − −ln = − 3.75+ 0.83⋅ 10 ln Gi + 0.93⋅ 10 ln Ci + 0.16⋅10 Ti − 0.29⋅10TiD + H 1.19 0.42 0.30 0.32 1.11i−7( Yi)+ 0.16− 4.43⋅10 lnY0.12 1.24N = R =i263;adj0.36i1 1 1 3 2The individual temperature parameters are notstatistically significant from zero at the 5% level, butthey are jointly significant. “Observations” for 1995 werederived from 1997 observations by dividing the latter bythe population and per capita income growth between1995 and 1997, correcting the latter for the incomeelasticity of (3) and (4). The income elasticity ofdomestic holidays is positive for countries with lowincomes but falls as income grows and eventually goesnegative. Qualitatively, this pattern is not surprising. Invery poor countries, only the upper income class haveholidays and they prefer to travel abroad, also becausedomestic holidays may be expensive too (cf. Equation 6).As a country gets richer, the middle-income class haveholidays too, and they first prefer cheap, domesticAnnual Proceedings of Vidzeme University College “ICTE in Regional Development”, 2006105