to Claessens and Laeven (2003), we replace <strong>the</strong> US industry characteristics with industrycharacteristics <strong>from</strong> o<strong>the</strong>r regions in some speci…cations.In Table 3, we compare <strong>the</strong> industry entry rates in <strong>the</strong> US to an aggregated measure<strong>of</strong> European entry <strong>from</strong> <strong>the</strong> Amadeus-EVCA dataset. In more than half <strong>of</strong> <strong>the</strong> industries,entry in <strong>the</strong> US is at least marginally higher than in Europe as a whole.2.4 Sample constructionWe employ <strong>the</strong> same sample selection procedure to <strong>the</strong> Amadeus dataset as Klapper et al.(2006) to see whe<strong>the</strong>r <strong>the</strong> e¤ect <strong>of</strong> <strong>private</strong> <strong>equity</strong> <strong>investment</strong> survives <strong>the</strong> inclusion <strong>of</strong> barriersto entry in a sample where <strong>the</strong>se barriers were already shown to be signi…cant. We initiallyuse <strong>the</strong> 2001 edition <strong>of</strong> Amadeus and limit <strong>the</strong> sample to <strong>the</strong> years 1998 and 1999. The year2000 is not used due to some incompleteness <strong>of</strong> <strong>the</strong> information in each Amadeus editionconcerning <strong>the</strong> previous year. The years prior to 1998 are not used due to <strong>the</strong> well-knownsurvivorship problems <strong>of</strong> <strong>the</strong> Amadeus database: when a …rm ceases existing, Amadeus keepsa record <strong>of</strong> it for 4 years, and <strong>the</strong>n takes it out <strong>of</strong> <strong>the</strong> database. Consequently, while eachyearly addition <strong>of</strong> Amadeus contains data for more than 4 years back, <strong>the</strong> sample <strong>of</strong> …rmsone will …nd reported for <strong>the</strong> time-period 4 years and more prior to <strong>the</strong> year <strong>of</strong> issuing, willnot include many …rms who existed in that year, but exited <strong>the</strong> market after that. Using <strong>the</strong>data indiscriminately will <strong>the</strong>refore induce survivorship bias and misrepresent <strong>the</strong> volume <strong>of</strong>entry, and so we focus our attention to <strong>the</strong> years 2 and 3 prior to <strong>the</strong> year <strong>the</strong> database wasissued. In later robustness tests, in which we use <strong>the</strong> 2008 edition <strong>of</strong> Amadeus, we similarlyfocus our attention on <strong>the</strong> sample <strong>of</strong> …rms in 2006-2007.Firms in <strong>the</strong> …nal dataset we use are also required to have basic accounting informationon <strong>the</strong> variables we are interested in (year <strong>of</strong> incorporation, employment, assets, etc.). Thisapproach excludes phantom …rms created for tax purposes. We drop …rms that report onlyconsolidated statements in order to avoid double-counting …rms and subsidiaries abroad.We exclude industries where <strong>the</strong> activities are country-speci…c, namely agriculture, forestry,…shing, and mining.We also exclude utilities and post and telecommunications, whichECBWorking Paper Series No 1078August 200915
tend to be heavily regulated and/or state-owned, and <strong>the</strong> …nancial services sector because,arguably, …nancial …rms are subject to speci…c regulations which do not apply for o<strong>the</strong>r …rms(for example, initial capital requirements). Finally, we exclude <strong>the</strong> public sector, education,<strong>the</strong> social sector, <strong>private</strong> households, and activities that cannot be classi…ed. We are leftwith 37 NACE industries.At <strong>the</strong> country level, we exclude <strong>the</strong> East European and South-East European countriesdue to insu¢ cient availability <strong>of</strong> data on <strong>investment</strong> by <strong>private</strong> <strong>equity</strong>, although <strong>the</strong>secountries are covered by Amadeus. The same applies to Luxembourg. Likewise, we excludeSwitzerland, which has comprehensive coverage on <strong>private</strong> <strong>equity</strong> <strong>investment</strong>, but itsAmadeus coverage is compromised by <strong>the</strong> fact that small …rms are not required to …le. Finally,we use Eurostat to con…rm to what extent Amadeus is representative <strong>of</strong> <strong>the</strong> …rm sizedistribution in <strong>the</strong> respective country. We <strong>the</strong>n exclude Iceland, Ireland, and Portugal, forwhich <strong>the</strong> ratio <strong>of</strong> employment in …rms with more than 250 employees in Amadeus to employmentin …rms with more than 250 employees in Eurostat is less than 0.5, and/or forwhich <strong>the</strong> di¤erence between <strong>the</strong> share <strong>of</strong> small …rms (10-50 employees) in Amadeus and inEurostat is more than 0.25. The sample thus reached represents <strong>the</strong> best match <strong>of</strong> Amadeusand EVCA data that is possible to construct while avoiding limited coverage, insu¢ cientobservations and country-speci…c industry scope and legal requirements problems. The …nalsample consists <strong>of</strong> 2,788,680 …rms for 1998-1999 in 16 countries: Austria, Belgium, Czech Republic,Denmark, Finland, France, Germany, Greece, Hungary, Italy, Ne<strong>the</strong>rlands, Norway,Poland, Spain, Sweden, and UK. Firm-level information is <strong>the</strong>n aggregated at <strong>the</strong> industrylevel and matched with <strong>the</strong> EVCA data to create a dataset consisting <strong>of</strong> 571 2-digit NACErev. 1.1 industry-country data points (37 industries in 16 countries, with 21 data pointsmissing). Due to our sample selection method, we do not use 3 <strong>of</strong> <strong>the</strong> 17 EVCA industries(o<strong>the</strong>r electronics, …nancial services, and agriculture).Table 4 gives <strong>the</strong> conversion key <strong>from</strong> <strong>the</strong> 2-digit NACE rev 1.1 industries into <strong>the</strong> EVCAindustries used in robustness regressions. Note that out <strong>of</strong> <strong>the</strong> 37 NACE rev. 1.1 industries,12 fall exclusively under 1 EVCA class, for 10 more than 70% <strong>of</strong> <strong>the</strong> 3-digit subclasses16 ECBWorking Paper Series No 1078August 2009
- Page 1 and 2: Working Paper SeriesNo 1078 / augus
- Page 3 and 4: © European Central Bank, 2009Addre
- Page 5 and 6: AbstractUsing a comprehensive datab
- Page 7 and 8: environment leaves the main results
- Page 9 and 10: development and restructuring on sm
- Page 11 and 12: to industries which are more R&D in
- Page 13 and 14: of destination.We …rst recalculat
- Page 15: 2.3 US industry-level entry dataAs
- Page 19 and 20: speci…c e¤ects. Critically, we a
- Page 21 and 22: …rms only (less than 10 employees
- Page 23 and 24: higher for industries with higher e
- Page 25 and 26: percentage of the adult population
- Page 27 and 28: cost of …nancing to venture capit
- Page 29 and 30: 4.6 RobustnessWe conduct an extensi
- Page 31 and 32: is the same as the gap between priv
- Page 33 and 34: This paper is, to our knowledge, th
- Page 35 and 36: [12] Cetorelli, N., and P.E. Straha
- Page 37 and 38: [35] Petersen, M., and R. Rajan, 19
- Page 39 and 40: Table 2Summary statistics on entry,
- Page 41 and 42: Table 42-digit NACE rev. 1.1 indust
- Page 43 and 44: Table 6Private equity timing and fi
- Page 45 and 46: Table 8Private equity and alternati
- Page 47 and 48: Appendix 1. Variables: definitions
- Page 49 and 50: Share shadow economyShare of the in
- Page 51: 1068 “Asset price misalignments a