Venture Capital and New Business Creation

Journal of Banking & Finance 37 (2013) 46954710
Contents lists available at ScienceDirect
Journal of Banking & Finance

journal homepage: www.elsevier.com/locate/jbf
Venture capital and new business creation q

Alexander Popov a,, Peter Roosenboom b,1
a
b
European Central Bank, Financial Research Division, Germany

Department of Finance, Rotterdam School of Management (RSM), Erasmus University Rotterdam, Netherlands
a r t i c l e
i n f o
Article history:
Received 5 August 2011
Accepted 12 August 2013
Available online 4 September 2013
JEL classication:
G24
L26
M13
a b s t r a c t
Using a comprehensive database of rms from 21 European countries over the period 19982008, we nd
that venture capital investment has a positive effect on the rate of new business creation. This is especially true in countries with higher entry costs, higher protection of intellectual property rights, and
lower taxes on capital gains. Our results suggest that, controlling for country and industry characteristics,
venture capital is benecial to bringing new ideas to the marketplace in the shape of new companies.
2013 Elsevier B.V. All rights reserved.
Keywords:
Venture capital
New business creation
Finance abd Growth
1. Introduction
It is generally accepted that access to nance is an important
determinant of new business creation and growth (Rajan and Zingales, 1998; Aghion et al., 2007). However, banks are often reluctant to nance small new rms because of high uncertainty,
information asymmetry, and agency costs (Beck et al., 2005). In
comparison, venture capitalists are specialized to overcome these
problems through the use of staged nancing, private contracting,
screening, and active monitoring (Hellmann, 1998; Gompers and
Lerner, 1999, 2001a; Kaplan and Stromberg, 2001), and are therefore more likely to nance early stage and technology companies
q
A previous version of the paper was circulated as On the Real Effects of Private
Equity Investment: Evidence from New Business Creation. We thank Marco Da Rin,
Philipp Hartmann, Ulrich Hege, Florian Heider, Josh Lerner, Simone Manganelli,
Marina Martynova, Eric Nowak, Enrico Perotti, Jose Luis Peydro, Richard Rosen, Per
Stromberg, Krishnamurti Subramanian, and Greg Udell, as well as seminar
participants at the European Central Bank, the 2008 EFA meeting, the RICAFE2
third conference, the 11th Symposium on Finance, Banking and Insurance, the 2009
MFA meeting, the Federal Reserve Bank of Chicago, and the Third Searle Symposium
on the Economics and Law of the Entrepreneur. We also thank Lieven Baert, Kim
Bonnema, Edvardas Moseika, Peter Myhailovski, Fiametta Rossetti, and Teng Wang
for outstanding research assistance. The opinions expressed herein are those of the
authors and do not necessarily reect those of the ECB or the Eurosystem.
Corresponding author. Tel.: +49 69 1344 8428.
E-mail addresses: Alexander.Popov@ecb.int (A. Popov), proosenboom@rsm.nl (P.
Roosenboom).
1
Tel.: +31 10 408 1255.
0378-4266/$ - see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jbankn.2013.08.010
than banks (Carpenter and Petersen, 2002; Schwienbacher, 2008;

Cosh et al., 2009). Recent research (Mollica and Zingales, 2007;
Samila and Sorenson, 2011) shows that rm entry increases in
US regions that attract more venture capital (VC). However, there
is no empirical evidence as to the ability of VC to replicate this success in an international context.
In this paper, we ll this gap by providing the rst comprehensive study of the effect of venture capital on new business creation
in 21 European countries. This question is highly relevant to economic policy makers given that they often perceive venture capital
as an important contributor to the rising leadership of US rms in
high technology industries (Gompers and Lerner, 2001b). Hoping
to rival this success, the European Union stimulates venture capital
investment in an attempt to make Europe a hotbed for entrepreneurship (Aernoudt, 1999; Gilson, 2003). We exploit cross-sectional and longitudinal variation in the supply of venture capital
across countries and industries in order to determine whether
the availability of venture capital stimulates new business creation, and which characteristics of the regulatory and business environment strengthen or hinder the effect of venture capital on the
formation of new rms.
Our paper adds to a remarkably limited eld of research on the
effect of venture capital on aggregate economic growth rather than
on rm-level performance. Among the few studies on the subject,
Kortum and Lerner (2000) and Hirukawa and Ueda (2008) show
that venture capital investment in the United States is associated
with more innovation as measured by patent counts and patent
4696
A. Popov, P. Roosenboom / Journal of Banking & Finance 37 (2013) 46954710
citations at the industry level. Tang and Chyi (2008) nd that venture capital investment enhances aggregate productivity growth.
Our paper analyzes another such channel through which venture
capital affects economic growth, namely, the rate of new business
creation.
There are three mechanisms suggested by the literature via
which venture capital should lead to higher rates of new business
incorporation. First, venture capitalists may directly assist the birth
of new rms. Keuschnigg (2004) develops a model in which the
entrepreneurs own wealth constitutes a binding constraint, and
so venture capitalists stimulate new business creation by ensuring
that good ideas receive funding (seed capital) even when conceived
by entrepreneurs without substantial assets. In addition, venture
capitalists may raise the rms early-life survival chances and
growth through value-added services such as mentoring entrepreneurs, hiring executives, formulating strategies, and helping the
companies they nance establish themselves in the marketplace
(Sahlman, 1990; Kaplan and Stromberg, 2001; Hellmann and Puri,
2002; Bottazzi and Da Rin, 2002).2
Second, nascent entrepreneurs may recognize the need for capital in the future and only establish rms when they have reasonably high expectations of obtaining such funding. Sevilir (2010)
develops a model in which the availability of new rm nancing
through venture capitalists makes it more desirable for employees
to exert effort, generate a new business idea, and start their own
rm. This implies that the availability not just of seed and startup capital, but of VC capital at later nancing stages should matter
too for rm entry (Samila and Sorenson, 2011).
Third, rms may be engaged in entrepreneurial spawning or
in spin-offs. Gompers et al. (2005) examine the propensity of publicly traded rms to create new venture backed rms. They nd
that younger public rms located in main hubs of venture capital
activity are the most likely to create new ventures. The employees
of these rms are more likely to start their own business because of
their exposure to the entrepreneurial process and due to working
in a network of entrepreneurs and venture capitalists. While we
investigate the effect of venture capital on rm entry, distinguishing among the various channels through which this effect is realized is beyond the scope of the paper.
The literature has distinguished entry into an industry from
new business creation. The rst accounts for the migration of rms
across industries, while the second emphasizes pure entrepreneurship (de novo rms). We focus on the second approach and dene
entry as the incorporation of a previously nonexistent rm in the
respective industry and country. To that end, we use data from
Amadeus, a comprehensive database of corporations across a number of developed and transition countries in Europe, which allows
us to calculate the share of new rms to total rms in each industry
for the period 19982008. We combine that data with industry-level data on venture capital investment in Europe from VentureXpert. This allows us to study the contribution of venture capital
to new business creation over the longest period for which both
rm entry and venture capital investment can be calculated. In
addition, we capture a full business cycle, encompassing the peak
of the dot-com bubble, the slowdown in VC fund-raising in the
early 2000s, and the resurgence of VC activity in the mid-2000s.
Finally, we address in two different ways the potential endogeneity induced by the fact that the supply of venture capital may in
itself depend on the demand for it by new rms. First, we employ a
panel approach which allows us to eliminate the effect of timeinvariant country and industry level left-out variables. Second,
2
Cumming et al. (2005) show that venture capitalists that provide nancial and
management expertise to entrepreneurial rms raise signicantly more capital from
investors than venture capitalists that only provide marketing and administrative
expertise.
we use the variation across countries and over time in buyout

fund-raising and in pension reforms as instruments to identify
the supply of venture capital. The logic behind this approach is that
the size of buyout funds and pension funds is correlated with risk
capital investment, while at the same time the general demand of
institutional investors for alternative assets should not depend on
entrepreneurship. Our results remain robust to these
specications.
We nd that the rate of new business creation increases in
countries and industries with sizeable venture capital investment.
An increase in venture capital investment by a factor of 7.2 (the difference between an industry at the 25th and an industry at the
75th percentile of VC investment) leads to an increase in the share
of new rms for the medium-entry industry by between 3% and
19%, depending on the estimation approach. This nding is robust
to a variety of data issues, as well as to using venture capital
investment measured over different time periods. Crucially, the
evidence that VC stimulates new business creation does not disappear once we address the endogeneity of the supply of venture
capital, implying that our results are not driven by investment
responding to a higher demand for VC. We nd that the effect of
venture capital is not inuenced by its high correlation with other
types of nance, such as bank credit, or by the sensitivity of venture capital-intensive industries to alternative market and regulatory developments. In general, we nd that venture capital works
better in countries with higher entry costs, higher protection of
intellectual property rights, and lower taxes on capital gains. This
implies that in a cross-country context, VC helps nascent entrepreneurs overcome the monetary cost of establishing new rms, and it
is more effective in countries where the return to investment in
intangible capital is higher. Finally, the effect of VC on new business creation is robust to controlling for other standard determinants of new business creation, notably barriers to entry.
The paper proceeds as follows. In Section 2 we summarize the
data. Section 3 describes the empirical methodology. Section 4 presents the empirical results. Section 5 concludes with the main ndings of the paper.
2. Data
This section describes the two main data sources used in the
empirical analysis. We rst describe our concordance key, necessary to match the data sources, then the data on de novo rm creation, and nally the data on VC investments.
2.1. Concordance
The relevant data are initially available in different industrial
classications. The original venture capital data from Thomson
VentureXpert contain information about deal value as well as each
portfolio companys industry afliation codes using Thomson VentureXperts own VE Primary Industry Sub-Group 3 and SIC codes.
However, for 13.8% of the deals, the SIC industry afliation
information is missing. For these cases we developed a unique concordance key to translate these companies VE Primary Industry
Sub-Group 3 to a SIC code. The concordance key is constructed
based on the most frequently observed SIC code from all deals in
that VE Primary Industry Sub-Group 3 realized in 21 European
countries from 1998 until 2008. By using this key, we are able to
assign all target companies to a SIC code. Aggregate values of venture capital invested in each industry are then calculated for each
year and for each country. This procedure is based on Hirukawa
and Ueda (2008).
The data on the share of new rms come in NACE Rev. 1.1 format, which is the industrial classication used by Amadeus. To
4697

Table 1
Summary statistics: New business creation, by country.
Country
Total rms
% New rms (1 and 2 years old)
Amadeus coverage
Austria
Belgium
Czech Republic
Denmark
Finland
France
Germany
Greece
Hungary
Iceland
Ireland
Italy
Netherlands
Norway
Poland
Portugal
Slovakia
Spain
Sweden
Switzerland
UK
117,142
286,657
115,833
101,958
115,587
709,453
712,738
21,046
180,030
19,276
98,099
638,080
407,940
132,919
96,289
251,412
36,450
604,610
201,875
285,318
1,057,165
9.23
6.25
6.71
12.23
7.78
7.34
7.41
9.87
7.68
14.77
6.49
5.20
4.80
11.85
6.34
6.58
8.90
7.01
8.73
1.59
11.98
61.8
96.2
43.4
96.7
81.9
66.1
42.0
89.3
55.8
29.1
64.5
57.9
48.2
100.0
29.3
55.0
18.4
49.3
55.5
36.8
95.7
Total
6,189,877
7.89
65.1
Note: This table summarizes the data on total rms and on new business creation averaged over the period 19982008, for all countries in the sample. Data come from
Amadeus. The fraction of new rms is calculated as the number of rms 2 or less years old over all rms in a particular industry, and then aggregated over industries, where
each industry is weighted by its relative share in a countrys overall employment. The last column compares the number of rms in Amadeus to the number of rms in
Eurostat.
achieve concordance, all SIC codes are matched to NACE Rev. 1.1
codes through the NAICS industrial classication. See Appendix A
for how we construct the concordance key, and Appendix B for
the concordance key itself.
2.2. Amadeus database
The rm-level data come from the Amadeus database. Amadeus
is a commercial pan-European database provided by Bureau van
Dijk, containing nancial information on over 10 million public
and private companies in 38 European countries. The database
contains detailed rm-level accounting data for a number of nancial ratios, activities, and ownership. Initially received from over 50
different vendors across Europe, the data are then transformed into
a single format enabling comparison across countries. The focus of
the Amadeus database is on nancial information. In addition to
that, Amadeus provides rm-level information on year of incorporation and employment. We use the year of incorporation to calculate the age of the rm, hence the share of new rms in each
industrycountry-year. The variable we create is referred to as
Entryijt, and it denotes the share of rms less than 2-years old in
country i in industry j in year t over the longest period for which
data is available, 19982008. We only count the rms that are at
least 1 full year of age to reduce measurement error. Finally, Amadeus uses the 3-digit NACE industry classication standard, which
we aggregate at the 2-digit level in order to have a sufcient number of rms in each industry for each country. Appendix A describes the sample selection process we use.
Table 1 summarizes the Amadeus data by country. We record
sizeable variation across countries. In particular, high-entry
countries like Denmark, Iceland, and Norway, have two to three
times the average new business creation rates of low-entry
countries like Italy, the Netherlands, and Belgium. Switzerland is
an outlier, with average new business creation at 1.6%. The reason
is that small rms are not required to le, for which reason we later drop all observations from Switzerland in robustness tests. The
third column of Table 1 also summarizes the Amadeus coverage,
namely, the ratio of the number of rms in each country present in

Amadeus to the number of rms in each country according to
Eurostat. The coverage ranges from a low of 18.4% in Slovakia to
a high of 100% in Norway.
In Table 2, we report aggregated entry rates per SIC-NACE
industry for our European sample over the 19982008 period.
Clearly, the variation in industry entry rates is substantial, with
rms younger than 2 years constituting on average only 5.66% of
total rms in industry 3. Textile and Apparel, but 11.69% in
industry 20. Real estate; renting of machinery and equipment;
computer and related services; research and development; other
business activities; other services activities.3
2.3. VentureXpert
The venture capital investment data come from Thomson VentureXpert. The Thomson VentureXpert database contains information for all venture capital deals realized in 21 European countries
(Austria, Belgium, Czech Republic, Denmark, Finland, France,
Germany, Greece, Hungary, Iceland, Ireland, Italy, Netherlands,
Norway, Poland, Portugal, Slovakia, Spain, Switzerland, Sweden,
and UK) over a 11 year period from January 1, 1998 to December
31, 2008. Venture capital investment in the database includes various denitions of early stage nancing and excludes all forms of
later stage transactions by private equity rms, such as buyouts,
mezzanine, turnaround, or distressed debt investments. These
investments can be made by venture capitalists from the same
country in which the portfolio rm is located but also by foreign
venture capitalists. We downloaded all venture capital deals from
Thomson VentureXpert for each country for the period 19982008,
in current euros, and re-calculated them in constant euros using
3
Appendix C reports nancial information on new rms incorporated in 2006/
2007. Appendix D reports ownership information for these rms. These ndings
conrm that the rms in our sample are small, young companies with few employees
that are typically owned by private individuals (founders and their family). Appendix
E shows the trends in start-up activity aggregated across countries and industries.
4698
Table 2
Summary statistics: New business creation, by industry.
2-Digit NACE rev 1.1 code and industry name
Manufacturing
1. Food products and beverages
2. Tobacco products
3. Textiles and apparel
4. Leather products
5. Wood products
6. Paper and paper products
7. Publishing and printing
8. Coke and petroleum products
9. Chemicals and chemical products
10. Rubber and plastic products
11. Other non-metallic mineral products
12. Primary metals and fabricated metal products
13. Machinery and equipment n.e.c.; ofce machinery and
computers; electrical machinery n.e.c.; radio, television,
and communication equipment; medical, precision, and
optical instruments
14. Motor vehicles and other transportation equipment
15. Furniture and manufacturing n.e.c.
Construction
16. Construction
Trade
17. Sale, maintenance and repair of motor vehicles and
motorcycles; wholesale trade, except for motor vehicles;
retail trade, except of motor vehicles and motorcycles
Hotels and restaurants
18. Hotels and restaurants
Transportation
19. Land transport; water transport; air transport; supporting
and auxiliary transport services and travel agencies
Services
20. Real estate; renting of machinery and equipment;
computer and related services; research and development;
other business activities; other services activities
Table 3
Summary statistics: Venture capital data, by country.
% New rms
Country
Austria
Belgium
Czech Republic
Denmark
Finland
France
Germany
Greece
Hungary
Iceland
Ireland
Italy
Netherlands
Norway
Poland
Portugal
Slovakia
Spain
Sweden
Switzerland
UK
6.47
10.86
5.66
6.11
7.34
6.59
7.14
11.66
6.76
6.39
6.28
7.30
6.81
8.05
6.26
9.79
Total
8.70
9.94
9.18
11.69
Note: This table summarizes the data on new business creation by industry, averaged over 19982008, for all countries in the sample. The fraction of new rms is
calculated as the number of rms 2 or less years old over all rms in a particular
industry. Data come from Amadeus.
2006 as baseline year for the purpose of the empirical analysis.

While VentureXpert contains data on VC deals back to 1990,
19982008 is the longest period for which the database on rm entry we use - Amadeus offers extensive coverage of rms.4 Aggregate values of venture capital invested in each industry are then
calculated for each countryindustry-year using the SIC-NACE classication developed (see Appendix B).
Table 3 summarizes the venture capital data by country, in million of current euros. It contains information on the number of
deals per country over the period 19982008, as well as on the
average annual volume of total venture capital investment and
on the ratio of that investment to GDP. Clearly, venture capital
investment varies markedly across countries. For example, Hungary, Iceland, Ireland, the Netherlands, Sweden, and the UK attracted over the period in question an average of more than 0.1%
of venture capital investment to GDP, approaching US levels in that
respect. At the same time, countries like Italy, Poland, and Portugal
each had an average ratio of venture capital investment to GDP of
less than a sixth of the UK one. In terms of total investment, venture capital disbursements are clustered too: France, Germany,
and the UK accounted for around 2/3 of all investment over the
period. Overall, around 96.9 billion of current euros were invested
in venture capital nance over the period, in a total of 21,413 deals.
The average European venture capital deal over the period thus involved 4.53 million euro worth of investment. For comparison,
4
Nevertheless, we use VC data as far back as 1995 to construct lagged VC variables
in some exercises.
VC deals
VC deals volume
VC/GDP in %
339
622
101
790
990
4,016
2,593
43
194
26
607
568
1115
480
311
286
29
886
1358
545
5514
1215.1
2648.8
369.4
2217.4
1113.5
19728.4
10856.5
578.4
875.3
120.7
2755.4
3836.3
6699.1
1678.3
514.9
294.3
62.2
5073.2
4275.6
2240.4
29776.2
0.0459
0.0803
0.0633
0.0865
0.0525
0.0827
0.0337
0.0739
0.1609
0.1935
0.1787
0.0226
0.1059
0.0825
0.0284
0.0183
0.0413
0.0494
0.1115
0.0568
0.1274
21,413
96929.5
0.0786
Note: This table summarizes the data on venture capital investment, aggregated
over the period 19982008, as well as over industries. In the rst column, the total
number of deals involving venture capital investment is reported. In the second
column, the total volume of venture capital nance involved over the period is
reported, in millions of current euros. Data come from VentureXpert. In the third
column, the average annual ratio of venture capital investment to GDP in the
country is reported. Data come from EVCA.
over the same period Thomson VentureXpert covers 56,403 venture capital deals in the US involving a total of 544.4 billion USD.
Table 4 provides a time-dimension look at the venture capital
data. The most venture capital deals recorded were at the peak
of the dot-com bubble, in 2000 and 2001, whereas by 20032004
venture capital investment had fallen by more than two thirds
from its peak, only to recover again in subsequent years. In relation
to GDP, venture capital in Europe in 2000 approached average US
levels, and its decline following the burst of the dot-com bubble
was equally impressive. It is worth noting that part of the recovery
in VC investment in the mid-2000s was due to a sequence of regulatory changes favorable to venture capital, mainly reforms of
pension funds laws, allowing autonomous pension funds to invest
in risk capital. In particular, in 2003 EU-wide Directive 2003/41/EC
liberalized the pension fund regime for all EU member countries,
preventing governments from imposing any but modest quantitative restrictions to the share of assets a pension fund could devote
to risk capital.5 Overall, this development is similar to the one in the
US, where in 1978, the Department of Labour claried that investments in venture capital funds by pension funds do not violate the
prudent man rule in the Employee Retirement Income Security Act
(ERISA), leading to a large increase in VC disbursement in subsequent years.6
See Appendix F for all denitions and sources of all variables
used in the paper.
3. Empirical methodology
We use a panel regression where the main unit of observation is
the industrycountry-year:
Entryijt b0 b1 Venture capitalijt b2 Shareijt b3 Di

b4 Dj b5 Dt eijt
5
6
See EVCA yearbook 2004 for details.

See Gompers and Lerner (1999) for details.
4699
4.1. Venture capital and new business creation: main results
Table 4
Summary statistics: Venture capital data, by year.
Year
VC deals
VC deals volume
VC/GDP in %
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
Total
629
1221
3019
2507
1481
2782
2782
2290
1850
1469
1383
21,413
2311.7
8113.8
19302.9
12535.1
9847.5
6613.2
6678.3
8233.4
9930.0
7774.2
5589.4
96929.5
0.0485
0.1202
0.1991
0.1439
0.1015
0.0908
0.0665
0.1054
0.1214
0.0924
0.0729
0.0786
Note: This table summarizes the data on venture capital investment. The data are
aggregated over countries, as well as over industries, per year over the 19982008
period. In the rst column, the total number of deals involving VC is reported. In the
second column, the total volume of venture capital nance involved over the period
is reported, in millions of current euros. Data come from VentureXpert. In the third
column, the average annual ratio of venture capital investment to GDP is reported.
Data come from EVCA.
As in Klapper et al. (2006), Entryijt denotes the share of rms less

than 2-years old in country i in industry j in year t. Venture capitalijt
denotes venture capital investment in country i in industry j in period t. Because we expect that historical investment will be responsible for contemporaneous rm creation to a larger degree than
contemporaneous VC investment, in the main specication we
use a 3-year VC investment average at the industrycountry level.
We test for sensitivity to this time period by also considering contemporaneous investment, as well as up to 5-year averages. We expect that venture capital is conducive to new business creation,
and hence a positive sign for the main regression coefcient of
interest, b1.
Studies of cross-sector industrial growth consistently predict
that sectors which have already grown fast in the past grow less
in the future (see, for example, Rajan and Zingales (1998)) and
have larger average rm size (see, for example, Cetorelli and Strahan (2006)). By extension, larger sectors should have lower entry
rates. In addition, theories of the industrys life cycle predict that
sectors which are already relatively large should have lower rates
of new business incorporation (see, for example, Klepper (1996)).
Hence, the variable Shareijt, which denotes industry js share of total employment in country i during year t, is included in the regression, and it should capture the different propensity to entry and
growth due to life-cycle specic reasons.
Next, Di is a matrix of country indicator variables controlling for
any market-specic, time-invariant effects on de novo business creation. Dj is a matrix of industry indicator variables controlling for
any industry-specic, time invariant effects on de novo business
creation. Dt is a matrix of year dummies controlling for the effect
of the business cycle on de novo entry which is common to all
countries and industries. In an alternative specication, we replace
Di, Dj, and Dt with Dit and Djt. Dit is a matrix of country-year indicator variables controlling for any market-specic, time-varying effects on de novo business creation. Analogically, Djt is a matrix of
industry-year indicator variables controlling for any industry-specic, time varying effects on de novo business creation. These xed
effects account for convergence phenomena as in Barro and Sala-iMartin (1992). Finally, eijt is the idiosyncratic error.
4. Results
This section reports the results from the main empirical exercise, from the tests which correct for the endogeneity of the VC series, and from the various robustness checks.
In Table 5, column (i) we present the estimates from the basic

OLS regression of entry rates on venture capital investment. The
main explanatory variable is the average VC investment over the
past 3 years in a particular industrycountry pair in each year.
The estimate of b1 is signicantly positive, implying that de novo
business creation is higher in industries with higher venture capital investment. The regression coefcient is 0.031, and its numerical interpretation is the following. Take the median industry in
terms of entry (Publishing and printing). On average, new rms
(1- or 2-years old) constitute 7.1%. Take the difference between
an industry at the 25th and an industry at the 75th percentile of
VC investment. VC investment is on average 7.2 times higher in
the latter. Increasing VC investment in the medium-entry industry
by a factor of 7.2 would increase the share of new rms by 0.223, or
by 3.1%. This simple rst empirical test conrms the hypothesis
that venture capital investment has a real contribution to rm
creation.7
Regarding the main control variable, industry share is negatively, but not signicantly, correlated to rm entry.8
In this basic case, the main explanatory variable has a value of
zero when there has been no VC investment in the past 3 years
in that industrycountry, and the log of actual investment if there
has been any VC investment in an industry over that period. In column (ii), we use an alternative VC measure, namely an indicator
variable equal to 1 if the industrycountry pair has received any
venture capital investment in the past 3 years. In this case, we simply compare entry in industries that receive VC funds to entry in
industries that do not. The estimate of b1 is again positive and signicant, and its magnitude implies that de novo rm creation rates
are on average higher in industries that receive VC funds relative to
industries that do not.
It is a logical question whether the effects we have observed so
far are caused by the high end of VC investment, or average VC
investment accounts for the bulk of the result. In column (iii), we
report the estimates from a model where the variable for
VC investment has been replaced with dummies for whether VC
investment in that particular industrycountry pair falls in the
1st, 2nd, or 3rd tertile of the VC investment distribution.9 We record that the higher-end VC investment (2nd tertile, but especially
3rd tertile) is associated with higher rm entry relative to low-VC
industries. This suggests that most of the positive effect of VC on rm
entry comes from higher rates of new business creation in industries
to which the big money ows.
In the rest of the columns of Table 5, we test the basic result
using different specications of our empirical model. In columns
(iv)(vi), we account for left and right censoring by replacing the
OLS specication with a Tobit one. The rationale behind this is that
entry rates are left-truncated at 0 and right-truncated at 1. The
coefcients on the main variables of interest do not change, and
they are signicant at least at the 10% level.
In our basic specication so far, we do not control for countryspecic developments that vary over time. One potentially
important co-variate is GDP growth which may very well explain
the number of start ups over time and across countries. For example, high growth rates may signal good growth opportunities,
pushing the marginal agent to switch from a salaried job to entrepreneurship. In column (vii), we formally test this hypothesis. We
nd that as expected, GDP growth is positively and signicantly
7
However, before addressing endogeneity concerns, we will be using the word
effect with caution.
8
Using a similar research design as ours, Klapper et al. (2006) also nd that
industry share is negatively but insignicantly related to entry rates.
9
A similar estimation strategy is pursued in Bernstein et al. (2010).
4700
Table 5
Venture capital and new business creation.
(i)
(ii)
(iii)
OLS
(iv)
(v)
(vi)
(vii)
(viii)
Tobit
OLS
0.029
(0.011)
0.025
(0.012)
0.044
(0.013)
0.103
(0.047)
Fraction of new rms

3-Year VC
0.031
(0.012)
0.355
(0.167)
3-Year VC industry
0.320
(0.163)
0.279
(0.165)
0.544
(0.238)
3-Year VC, 2nd tertile

3-Year VC, 3rd tertile
0.218
(0.166)
0.575
(0.225)
0.017
(0.057)
0.003
(0.056)
0.001
(0.056)
0.004
(0.056)
0.040
(0.014)
0.010
(0.063)
4214
0.38
4214
0.07
4214
0.07
4214
0.07
4214
0.35
GDP growth
Share
0.018
(0.057)
Fixed effects
Country industry year
Observations
R2
4214
0.35
0.015
(0.057)
4214
0.35
Country Year Industry Year

4214
0.57
Note: Table 5 reports estimates from OLS regressions (columns (i)(iii) and (vii)(viii)) and from Tobit regressions (columns (iv)(vi)). The dependent variable is the ratio of all
new rms (1 or 2 years old) to total rms in each industrycountry-year, at the SIC-NACE industry level. The variable is calculated using data from Amadeus. 3-year VC
denotes the natural logarithm of average venture capital investment for the respective industry in the respective country over the past 3 years. 3-year VC industry is a
dummy equal to 1 if the respective industry in the respective country had any VC investment in the past 3 years. 3-year VC, 2nd tertile is a dummy equal to 1 if the industry
belongs in the second tertile of VC-receiving industries in terms of total volume, and 3-year VC, 3rd tertile is a dummy equal to 1 if the industry belongs in the third tertile of
VC-receiving industries in terms of total volume. GDP growth is the annual growth of per capita GDP. The omitted category in columns (iii) and (vi) is industries in the 1st
tertile of VC investment. All VC variables are calculated using data from VentureXpert. Share equals the total employment in a given industrycountry-year divided by total
employment in the corresponding country-year, and is calculated using data from Amadeus. All regressions include a constant, industry dummies, country dummies, and year
dummies, not reported. In column (viii) country-year and industry-year dummy interactions are used. Heteroskedasticity adjusted standard errors are reported in
parentheses.
Report signicance at the 10% level.
correlated with the rates of new business creation. Nevertheless,

VC survives as a both economically and statistically signicant
determinant of the ratio of new rms to the total population of
rms.
Finally, we run an even stronger robustness check by replacing
the matrix of country, industry, and year dummies with countryyear and industry-year xed effect interactions, to account for
any unobservable trends at the market and sector level. Eventhough this is a more stringent test than the one in column (i),
the magnitude of the coefcient on the main variable of interest
actually increases, and the effect is statistically signicant at the
1% level (column (viii)).
4.2. Addressing various data issues

Next, in Table 6 we address various potential questions about
the validity of the VC measures we use and the quality of the
underlying Amadeus data. First, the decision to average VC investment over a 3-year period is certainly arbitrary. It might make
more sense to allow for longer gestation periods. Alternatively,
the effect of VC could be immediate, given that a non-negligible
portion of VC investment is seed and start-up capital. Columns
(i) and (ii) address these questions. We nd that it makes sense
to look at longer rather than shorter periods, and that averaging
over 3-years does not lead to much different estimates relative to
using longer time periods, making the main empirical strategy
we have chosen a reasonable one.
We also need to account for initial rm business creation. Venture capital activity only picked up in Europe in the late 1990s, and
so we may be measuring a simple correlation due to the fact that
VC investment has been highest in industries which already had
high entry rates in the beginning of the sample period. In column
(iii), we estimate a version of our main model which includes a
variable equal to the share of new rms to all rms for each indus-
trycountry pair in 1998, and excludes all observations for 1998

from the regression. Variations in entry rates in 1998 do not explain a signicant portion of the variations in entry rates over
the rest of the sample period. While our estimates decrease slightly
in magnitude, they remain statistically signicant.
Next, we address the issue of the varying representativeness of
Amadeus across countries. As demonstrated in Table 1, in some
countries Amadeus covers too few rms relative to the total rms
in the country according to Eurostat. We start by excluding Switzerland, which has comprehensive coverage on venture capital
investment, but its Amadeus coverage is compromised by the fact
that small rms are not required to le. This leads to a very low observed entry rate (1.6% relative to the sample average of 7.9%). We
then exclude the Czech Republic, Germany, Iceland, Netherlands,
Poland, Slovakia, Spain, and Switzerland, for which the ratio of
rms in Amadeus relative to rms in Eurostat is less than 0.50.10
The sample thus reached should represent the best match of Amadeus and VentureXpert data that is possible to construct while
avoiding limited coverage, insufcient observations, and legal
requirements problems. However, our estimates of the effect of venture capital on rm entry (column (iv)) do not suffer much from this
sample selection procedure.11
Next, in column (v) of Table 6, we exclude the sub-sample of 4
transition economies (Czech Republic, Hungary, Poland, and Slovakia). The 1990s were a vibrant period in the economic history of
Central Europe in terms of privatization, and we would like to
eliminate the possibility that a large number of new private rms
are actually old state-owned rms which have been counted only
after they became private.12 Our results are robust to the exclusion
of this set of countries.
10
A similar sample selection is applied to Amadeus by Klapper et al. (2006).

The same is true when we only estimate our model using data on countries with
Amadeus coverage higher than two thirds.
12
This cannot be inferred from Amadeus.
11
4701

Table 6
Venture capital and new business creation: Data issues.
(i)
(ii)
(iii)
(iv)
(v)
(vi)
Fraction of new rms

5-Year VC
(vii)
(viii)
Entrepreneurial activity
0.040
(0.013)
0.031
(0.011)
1-Year VC
0.025
(0.012)
3-Year VC
0.034
(0.016)
0.030
(0.013)
0.029
(0.012)
0.385
(0.219)
1-Year VC, EVCA

Initial entry
Share
0.019
(0.057)
Fixed effects
Observations
R2
4214
0.38
0.021
(0.057)
4214
0.38
0.009
(0.031)
0.014
(0.063)
0.049
(0.078)
0.082
(0.064)
0.459
(0.226)
0.027
(0.057)
Country Year
4214
0.39
2624
0.34
3434
0.38
3805
0.34
219
0.62
125
0.62
Note: This table reports estimates from OLS regressions. The dependent variable is the ratio of all new rms (1 or 2 years old) to total rms in each industrycountry-year, at
the SIC-NACE industry level (columns (i)(vi)), and in each country-year (column (vii)); data from Amadeus are used. In column (viii), the dependent variable is dened as the
percentage of 1864 population who are either a nascent entrepreneur or ownermanager of a new business; data from the Global Entrepreneurship Monitor are used. 5year VC denotes the natural logarithm of average annual venture capital investment for the respective industry in the respective country over the past 5 years. 3-year VC
denotes the natural logarithm of average venture capital investment for the respective industry in the respective country over the past 3 years. 1-year VC measures total
venture capital investment for the respective industry in the respective country over the current year. All VC variables are calculated using data from VentureXpert. 1-year
VC, EVCA measures total venture capital investment in the respective country over the current year, using data from EVCA. Initial entry denotes the fraction of 1 and 2 years
old rms in the respective SIC-NACE industry in 1998; observations in 1998 are excluded. In column (iv), the countries for which the share of rms in Amadeus relative to
Eurostat is less than 0.50 (Czech Republic, Germany, Iceland, Netherlands, Poland, Slovakia, Spain, and Switzerland) are excluded. In column (v), all transitional economies
(Czech Republic, Hungary, Poland, and Slovakia) are excluded. In column (vi), the UK and Greece are excluded. In column (vii), VC data at the country level from EVCA are
used. Share equals the total employment in a given industrycountry-year divided by total employment in the corresponding country-year, and is calculated using data from
Amadeus. All regressions include a constant and various combinations of industry dummies, country dummies, and year dummies, not reported. Heteroskedasticity adjusted
standard errors are reported in parentheses.
In column (vi) we exclude the country with the most VC investment (the UK) and the country with the least VC investment
(Greece). The UK market accounts, on average, for a quarter of all
VC deals in the sample, and for one third of total VC disbursements.
In Greece, on the other hands, there are only 6 non-zero industryyear observations over the sample period. We want to make sure
that our results are not driven by the industrial composition in
one particular economy. We also want to eliminate a country with
too little venture capital that may be biasing the results if this
country happens to have low entry. Once the UK and Greece are
dropped from the sample, the estimate of the effect of VC on
new business creation continues to be signicant.
Next, we address data issues related to the quality of the VC
data we use. While we have chosen to use data on VC disbursements from Thomson VentureXpert (TVE), the European Venture
Capital Association (EVCA) is another potential source of such data.
While EVCAs data is supposedly more accurate, especially in the
earlier years of our sample period, EVCA only started reporting
VC investment data at the sectoral level by country of destination
in 2007. Before 2007, it only reported VC investment data at the
sectoral level by country of management. Nevertheless, to mitigate
concerns about the potential inferior accuracy of the TVE data, in
column (vii) we replace our preferred countryindustry-year VC
investment data from TVE with country-year VC investment data
from EVCA. Our estimates imply that regardless of the VC data
source, venture capital does have a signicant effect on the rate
of new business creation.
Finally, we address data issues related to the new business creation data we use in the paper. Our original variable for new business entry is constructed by taking all rms in Amadeus younger
than two years and dividing by the number of total rms, for each
countryindustry-year. There are potential problems related to
this approach that we already referred to; for instance, Amadeuss
coverage of rms varies by country, and our sample selection procedure may discard a number of true start ups. To account for that
fact, we now employ data on entrepreneurship by another data
source, the Global Entrepreneurship Monitor (GEM), a survey of
the degree of entrepreneurial activity for a large number of countries. Among others, GEM reports an indicator called Total Early
Stage Entrepreneurial Activity dened as the percentage of
1864 population who are either a nascent entrepreneur or ownermanager of a new business. The two sub-components of this
indicator are dened as follows. A nascent entrepreneur is dened as actively involved in setting up a business they will own
or co-own; this business has not paid salaries, wages, or any other
payments to the owners for more than three months. An owner
manager of a new business is dened as owning and managing a
running business that has paid salaries, wages, or any other payments to the owners for more than three months, but not more
than 42 months. Because GEM does not differentiate by industry,
we run our main specication at the country level. Column (viii)
suggests that our main result is not affected by the denition of
new business creation.
4.3. Endogeneity and selection
The empirical methodology chosen is traditionally prone to
endogeneity problems. A measured positive regression coefcient
on the variable of interest does not automatically imply causality;
it could be that venture capital investment is driven by industries
with high entry rates, or it could be that a set of omitted variables
is jointly driving both the propensity to enter and the propensity to
invest in start-up companies. One simple way to account for the
endogeneity of current VC investment is to use lagged values of
VC investment. Such a denition of VC investment should be less
correlated with the current industry structure, and hence should
4702
partially address the concern that the effects we measure are due
to VC investors reacting to current opportunities (see Bernstein
et al., 2010).
However, lagged VC variables are not a perfect solution to the
endogeneity problem because entrepreneurial opportunities and
industry dynamics are likely to be correlated along longer periods.
Therefore, it is preferable to nd an instrumental variable which is
correlated with VC investment but not with entrepreneurial opportunities, and use it to extract the endogenous element of VC investment. While it has been generally agreed that the countrys legal
origin is a strong predictor of the degree of legal regulation and
by extension the quality of the nancial system nowadays (La Porta et al., 1998), in the case of venture capital investment, there are
two problems which reduce legal origin to a suboptimal instrument. First, the exogenous component of the legal system is likely
more strongly correlated with traditional types of nance than
with venture capital investment, which violates the relevance
condition. Second, legal systems most probably affect entry via
channels other than venture capital investment, like barriers to entry, which violates the exclusion restriction.
One alternative approach suggested by the VC literature (Kortum and Lerner, 2000; Hirukawa and Ueda, 2008; Samila et al.,
2008), is to use the size of pension funds and buyout fund-raising
as instruments for VC investment. The rationale for the latter is
that the magnitude of domestic buyout fund-raising is correlated
with VC fund-raising and with subsequent VC investment because
more interest among institutional investors to invest in private
equity leads to more interest in VC as well. The relevance condition
for this instrument is satised by the fact that when institutional
investors commit to private equity as an asset class, this affects
simultaneously fundraising for buyouts and for VC. At the same
time, being targeted at later-stage investment, it should be orthogonal to the rates of new business creation, satisfying the exclusion
restriction. Regarding the former, as pointed out in Gompers and
Lerner (1999), the 1979 clarication of the ERISA by the US Department of Labour led to a vefold increase in VC investment in the
next two decades. European law also offers variation over time
and across countries in that respect: while the UK has been allowing pension funds to invest in risk capital since pre-1991, for many
European countries restrictions were only lifted by the EU-wide
Directive 2003/41/EC in 2003, which eliminated restrictions on
the investment behavior of pension funds, only allowing national
governments some discretion on, for example, maximum amounts
pension funds are allowed to invest in risk capital markets.13
(However, in one case Sweden the Directive was followed by a
national legislative ban on investment in risk capital by pension
funds, deeming such investments too risky). Finally, how much pension funds would invest in reality is, of course, a function of their asset size.
Therefore, in an attempt to address the endogeneity problem,
we create two instruments. First, we use data from Thomson VentureXpert on domestic fund-raising by buyout funds, and create a
variable 3-year lagged buyout funds equal to the 3-period lagged
amount of domestic buyout funds raised. This variable should constitute a VC supply-shifter, for the reasons described in the previous paragraph.14 We use domestic fund-raising by buyout funds to
better reect the degree to which domestic investors are interested
in investing in the private equity asset class. In case domestic
investors are interested in investing in buyout funds we argue that
they also interested in investing in venture capital funds as part of
the private equity asset class. This increases the supply of VC funds.
We lag this variable because money that is raised in the current year
takes a few years to get invested in companies.
The variable we have generated is a country-level variable common to all industries in a particular country-year. To account for
the fact that our unit of observation is the countryindustry-year,
we create a variable 3-year lagged VC share which denotes the
3-period lagged VC investment in a particular countryindustryyear as a share of total VC investment in a country-year. Then,
we interact it with 3-year lagged buyout funds. The resulting
interaction variable varies at the countryindustry-year level,
relating the amount of buyout funds raised to the VC-intensity of
a particular industry.
Second, we follow Kortum and Lerner (2000) in using pension
fund asset size and liberalization events as an instrument. We
rst create a variable Pension reform, which is an indicator
variable equal to 1 if the respective country has already liberalized risk capital investment rules for pension funds in that particular year. We also acquire information on the assets of
domestic pensions funds from Eurostat, and calculate the 3-year
lagged value of the ratio of each countrys total pension fund assets to GDP in order to create a variable 3-year lagged pension
funds. Finally, we interact these two variables with a variable
pre-reform VC which is equal to the average VC investment
in a particular countryindustry prior to the pension reform in
that country. Similar to the approach in Kortum and Lerner
(2000), this resulting instrumental variable should constitute a
VC supply-shifter: larger pension funds (in ratio to GDP) should
result in higher VC investment in the future, but only if pension
funds are allowed to invest in risk capital, and only in industries
that were of interest to venture capitalists in the past.
The latter instrument is subject to two caveat. For one, it is possible that the laws regulating risk-capital investment by pension
funds were enacted following pressure from a growing VC industry, which would make them endogenous to the size of venture
capital investment and hence violate the exclusion restriction.
However, our investigation into the genesis of these laws conrms
that the formal motive expressed during the legislative process
was universally diversication of risk. For two, a pension reform
need not result immediately in an inow of funds from pension
funds into private equity funds, especially if the enactment of a
pension directive (like the EU-wide Directive 2003/41/EC) has been
anticipated. However, the data implies a structural break around
2003, the year in which 10 of the 21 countries in the sample liberalized their pension funds investment regime. For example, while
an average of 188.3 billion euro annually were raised from pension
funds during the 19982003 period, the same number stood at
290.6 billion euro annually during the 20042008 period.15 In
addition, over the same period pension funds become the foremost
institutional investor with respect to private equity fund raising, at
23% of total funds raised, surpassing banks who had been the largest
institutional investor before the 2003 EU-wide Directive.
In Table 7, we report the estimates from the 2SLS procedure.
Columns (i)(iii) report the estimates from the rst stage where
we have regressed the main VC variable used in the paper on the
buyout funds-based instrument (column (i)), on the pension reform-based instrument (column (ii)), as well as on both instruments (column (iii)). In the case of both instruments, the sign is
positive and the effect is signicant at the 1% statistical level,
implying that the relevance condition is satised. The F -statistics
is higher than the critical value required for the IV estimates to
have no more than 10% of the bias of the OLS estimates (see Stock
and Yogo (2005), for details).
13
See Appendix G for details.

In addition, our tests also imply that this variable is only weakly correlated with
the rate of new business creation (q = 0.09).
14
15
We note, however, that these numbers refer to funds raised for total private
equity investment, not just for VC investment.
4703

Table 7
Venture capital and new business creation: Endogeneity and selection issues.
(i)
(ii)
(iii)
(iv)
(v)
(vi)
(vii)
(viii)
Fraction of new rms

2SLS
OLS
First stage
Second stage
0.195
(0.055)
3-Year VC
Log 3-year lagged buyout funds 3-year lagged VC share
0.919
(0.068)
0.217
(0.040)
Log 3-year lagged pension funds Pension reform Pre-reform VC
0.671
(0.216)
0.185
(0.052)
0.033
(0.013)
1.128
(0.089)
0.202
(0.039)
0.226
(0.093)
0.111
(0.072)
0.115
(0.099)
0.063
(0.014)
0.013
(0.014)
0.021
(0.057)
2975
0.59
9.17
0.001
2750
0.71
3755
0.32
4214
0.36
3-Year VC High GDP per capita

3-Year VC Low GDP per capita
Share
0.028
(0.067)
0.077
(0.078)
Fixed effects
F-statistics
F-statistics (p-value)
Hansen J
Hansen J (p-value)
Observations
R2
184.76
0.001
3883
0.78
29.32
0.001
2975
0.52
0.006
(0.085)
0.038
(0.064)
98.40
0.001
2750
0.65
3883
0.71
Note: This table reports estimates from 2SLS regressions (columns (i)(vi)) and from OLS regressions (columns (vii)(viii)). Columns (i)(iii) report the estimates from a rststage regression where the dependent variable is 3-year VC, which denotes the natural logarithm of average venture capital investment for the respective industry in the
respective country over the past 3 years. Data come from VentureXpert. The dependent variable in columns (iv)(viii) is the ratio of all new rms (1 or 2 years old) to total
rms for each countryindustry-year at the SIC-NACE industry level. The variable is calculated using data from Amadeus. Log 3-year lagged buyout funds is the 3-period
lagged average volume of private equity funds raised in the respective country-year, expected to be allocated to buyouts. Data come from VentureXpert. Log 3-year lagged
pension funds is the 3-period lagged asset size of autonomous pension funds in this particular country-year. Data on pension fund size come from Eurostat. 3-year lagged VC
share is the 3-period lagged VC allocation in a particular industrycountry-year as a share of all VC allocated in that country-year. Pension reform is a dummy equal to 1 if
domestic pension funds are allowed to invest in risk capital in this particular country-year. See Appendix G for pension fund-related events. Pre-reform VC is the average VC
investment in a countryindustry before the pension reform in the country. In column (iv), VC is instrumented using Log 3-year lagged buyout funds 3-year lagged VC. In
column (v), VC is instrumented using Log 3-year lagged pension funds Pension reform Pre-reform VC. In column (vi), VC is instrumented using both instruments. In
column (vii), the top 10% industries in terms of relative size are excluded. In column (viii) High GDP per capita is a dummy equal to 1 if GDP per capita is in the top half of its
distribution. Low GDP per capita is a dummy equal to 1 if GDP per capita is in the bottom half of its distribution. Data come from the Penn Tables. Share equals the total
employment in a given industrycountry-year divided by total employment in the corresponding country-year, and is calculated using data from Amadeus. All regressions
include a constant, industry dummies, country dummies, and year dummies, not reported. Heteroskedasticity adjusted standard errors are reported in parentheses.
In columns (iv)(vi), our main measure of VC is instrumented

jointly using the buyout-based instrument (column (iv)), the pension funds-based instrument (column (v)), as well as on both
instruments (column (vi)). We nd that the estimate of b1 is still
positive and in all cases signicant at the 1% level, and the economic magnitude of the effect is almost triple the OLS effect reported in Table 5, column (i), when both instruments are used
(column (vi)).16 Numerically, the IV estimate (0.185) implies that
increasing VC investment in the medium-entry industry by a factor
of 7.2 would increase the share of new rms by 1.332, or by 18.8%.
While conceptually it is clear that the general demand of institutional investors for alternative assets, including venture capital,
should not depend on entrepreneurship, we also test formally the
correlation between the instruments and the second stage error
term, that is, the probability that they violate the validity of the
exclusion restriction (see Hansen (1982), for details). In column
(vi), the Hansen J-statistic does not exceed the critical value required
to reject the null hypothesis that both instruments are valid.
While the results from these tests are satisfactory, certain endogeneity concerns can still be raised. For example, it could be that
countries with large industries with high natural entry may have
16
However, the sample is now smaller than in Table 5 due to the exclusion of
Finland, Italy, Ireland, and the UK, where the pension reform took place before 1995
and so no comprehensive data on pre-reform VC investment exist.
both higher entrepreneurial culture (resulting in more entry

regardless of venture capital investment) and have higher levels
of venture capital investment (due to higher demand for all types
of nance). One method to control for that possibility is to exclude
the industries that are in the right tail of the industry size distribution. We restrict our sample to the industries that are in the bottom
tertile, bottom two tertiles, or outside the top 10% of their countrys industries in terms of size. When we do that, in all cases we
get estimates which are still positive and signicant, similar to
magnitude to the ones estimated previously, and they do not vary
much regardless of the denition of small industries that we use.
The coefcient after the exclusion of the top 10% is reported in column (vii). The effect of VC on rm entry still stands.
Another consideration is that economic development is correlated with VC investment, and so we need to control for GDP per
capita. It could be the case that poor countries are naturally more
adverse to entrepreneurial activity (and hence, there is naturally
less entry), and so there could also be stricter rules guiding private
investment in risk capital, as the local legislatures would be adverse to business activity in general. The above argument would
imply that if it is a selection problem, venture capital investment
could have a larger effect in poorer countries where VC investment
is more limited. This is not the case; in fact as indicated in column
(viii), venture capital is relatively more conducive to business creation in richer countries, implying no omitted variable bias.
4704
Table 8
Venture capital and new business creation: Alternative proxies for nance.
(i)
(ii)
(iii)
(iv)
0.031
(0.012)
0.031
(0.012)
0.030
(0.012)
0.115
(0.396)
0.017
(0.020)
0.105
(0.279)
0.017
(0.057)
4141
0.35
Fraction of new rms

0.030
(0.012)
0.274
(0.335)
3-Year VC
3-year VC industry Private credit/GDP
3-year VC industry Access to nance
0.020
(0.016)
3-year VC industry Investors protection

Share
0.018
(0.057)
0.016
(0.057)
0.145
(0.272)
0.021
(0.057)
Fixed effects
Observations
R2

4141
0.35
4214
0.35
4214
0.35
Note: This table reports estimates from OLS regressions. The dependent variable is the ratio of all new rms (1 or 2 years old) to total rms for each countryindustry-year, at
the SIC-NACE industry level. The variable is calculated using data from Amadeus. 3-year VC denotes the natural logarithm of average venture capital investment for the
respective industry in the respective country over the past 3 years. 3-year VC industry is a dummy equal to 1 if the respective industry in the respective country had any VC
investment in the past 3 years. All VC variables are calculated using data from VentureXpert. Private credit/GDP measures the ratio of private credit allocated by commercial
banks to country total GDP. Data come from the IMF-IFS. Access to nance measures general access to external nance by households and businesses. Data come from
Finance for All? The World Bank, 2007. Investors protection measures the degree of legal protection of private investment in the country. Data come from the Doing Business
Database. Share equals the total employment in a given industrycountry-year divided by total employment in the corresponding country-year, and is calculated using data
from Amadeus. All regressions include a constant, industry dummies, country dummies, and year dummies, not reported. Heteroskedasticity adjusted standard errors are
reported in parentheses.
Table 9
Venture capital and new business creation: Sensitivity to industry and country characteristics.
(i)
(ii)
(iii)
(iv)
(v)
(vi)
0.025
(0.012)
0.031
(0.012)
0.030
(0.012)
0.024
(0.012)
0.051
(0.015)
1.912
(0.415)
0.606
(0.469)
0.024
(0.057)
4141
0.35
Fraction of new rms

3-Year VC
R&D intensityUS 3-year VC
Capital intensity
US
0.018
(0.013)
0.564
(0.299)
0.087
(0.053)
0.411
(0.189)
3-year VC
0.051
(0.015)
Financial dependenceUS Private credit/GDP
1.827
(0.408)
Labor intensityUS Labor regulations
0.025
(0.057)
0.018
(0.057)
0.791
(0.464)
0.017
(0.057)
4141
0.35
4214
0.35
4214
0.35
EntryUS Entry cost

Share
0.022
(0.057)
Fixed effects
Observations
R2
4214
0.35
0.026
(0.057)
4214
0.35
respective industry in the respective country over the past three years. R&D intensityUS measures the per-industry median R&D usage per worker. Capital intensityUS
measures the per-industry median capital investment per worker. Financial dependenceUS measures the per-industry median share of operating expenses nanced
externally. Labor intensityUS measures per-industry median number of workers divided by total assets. The variables are calculated using data on mature Compustat
companies for the 19901999 period. All industry benchmarks are calculated at the SIC-NACE industry level. EntryUS is the ratio of new rms to total rms in the US. Data
come from Dun and Bradstreet. Private credit/GDP measures the ratio of private credit allocated by commercial banks to country total GDP. Data come from the IMF-IFS.
Labor regulations measures the degree of protection of workers. Data come from Botero et al. (2004). Entry cost measures the monetary cost of establishing a limited
liability company normalized by per-capita GDP, Data come from the Doing Business Database. Share equals the total employment in a given industrycountry-year divided
by total employment in the corresponding country-year, and is calculated using data from Amadeus. All regressions include a constant, industry dummies, country dummies,
and year dummies, not reported. Heteroskedasticity adjusted standard errors are reported in parentheses.
4.4. Alternative proxies for nance

We next account for the possibility that venture capital investment is a mere proxy for other types of nance or for return to
investment. For instance, countries with higher venture capital

investment will tend to have better developed banking sectors,
so our measure of venture capital investment may partially be capturing the effects of credit markets on entry via business loans.
4705

Table 10
Venture capital and new business creation: What makes VC work?
(i)
(ii)
(iii)
(iv)
(v)
(vi)
0.017
(0.005)
0.001
(0.001)
0.047
(0.014)
0.336
(0.189)
0.018
(0.011)
0.016
(0.058)
4214
0.36
Fraction of new rms

3-Year VC Entry cost
0.008
(0.004)
3-Year VC Labor regulations
0.001
(0.001)
0.048
(0.012)
3-Year VC Int. property protection
0.323
(0.161)
3-Year VC Capital gains tax
0.037
(0.057)
0.013
(0.057)
0.026
(0.011)
0.017
(0.057)
4214
0.35
4214
0.35
4214
0.35
3-Year VC Human capital

Share
0.011
(0.057)
Fixed effects
Observations
R2
4214
0.35
0.017
(0.057)
4214
0.35
respective industry in the respective country over the past 3 years. The variable is calculated using data from VentureXpert. Entry cost measures the monetary cost of
establishing a limited liability company normalized by per-capita GDP in the respective country. Data come from the Doing Business Database. Labor regulations measures
the degree of protection of workers. Data come from Botero et al. (2004). Int. property protection measures the degree of protection of intellectual property in the respective
country. Data come from the Heritage Foundation. Capital gains tax measure the average tax on capital gains in the respective country. Data come from the PriceWaterhouseCoopers Worldwide Taxes. Human capital measures the average years of schooling in the respective country. Data come from the Barro and Lee (2000). Share equals
the total employment in a given industrycountry-year divided by total employment in the corresponding country-year, and is calculated using data from Amadeus. All
regressions include a constant, industry dummies, country dummies, and year dummies, not reported. Heteroskedasticity adjusted standard errors are reported in
parentheses.
Also, countries with dynamic VC industries will tend to have higher

investor protection, and so again our estimates may be contaminated by the effect of the expected return on investment on entry.
Most importantly, the volume of nance may matter less than
access to nance per se. Hence, our VC variables might be picking
up the effect on new business entry by innovative companies of
easier access to all kinds of nance, including consumer loans
and mortgages. In all of those cases, our estimates would be biased.
Therefore, in Table 8 we proceed to measure the effect of venture capital investment on entry alongside the effect of nance in
general. In column (i), we show the estimates of a regression which
includes a measure of private credit by commercial banks,
normalized by GDP, and interacted with an indicator equal to 1 if
the industry has received any VC investment in the past 3 years.
This measure is widely accepted as a good proxy for a range of
nancial issues, like access to business loans and depth of the
Table A2
Industry conversion key: SIC and NACE.
SIC-NACE industry
SIC 1987 code
NACE Rev 1.1

code
1. Food products and beverages

2. Tobacco products
3. Textiles and apparel
4. Leather products
5. Wood products
6. Paper and paper products
7. Publishing and printing
8. Coke and petroleum products
9. Chemicals and chemical products
10. Rubber and plastic products
11. Other non-metallic mineral products
12. Primary metals and fabricated metal products
13. Machinery and equipment n.e.c.; ofce machinery and computers; electrical machinery n.e.c.; radio, television, and
communication equipment; medical, precision, and optical instruments
14. Motor vehicles and other transportation equipment
15. Furniture and manufacturing n.e.c.
16. Construction
17. Sale, maintenance and repair of motor vehicles and motorcycles; wholesale trade, except for motor vehicles; retail
trade, except of motor vehicles and motorcycles
18. Hotels and restaurants
19. Land transport; water transport; air transport; supporting and auxiliary transport services and travel agencies
20. Real estate; renting of machinery and equipment; computer and related services; research and development; other
business activities; other services activities
20
21
22, 23
31
24
26
27
29
28
30
32
33, 34
35, 36, 38
15
16
17, 18
19
20
21
22
23
24
25
26
27, 28
2933
37
25, 39
15, 16, 17
5057, 59
34, 35
36
45
50, 51, 52
58, 70
4145, 47, 48
65, 67, 72, 73, 75, 76, 78, 79,
81, 87, 89
55
6064
7074, 92,
93
Note: This table reports the conversion key used in the paper to match SIC and NACE industrial specications. Matching is done through 6-digit NAICS industrial classes. See
US Census Bureau for original classications.
4706
Table A3
Financial information on new rms incorporated in 2006/2007.
Operating revenue in 2006

Operating revenue in 2007
Net income in 2006
Net income in 2007
Employees in 2006
Employees in 2007
Table A5
Trends in start-up activity. Source: Amadeus.
Mean
Median
Number of observations
Year
Fraction of new rms
281
368
10
12
3.39
3.95
100
132
0
1
2
2
127,597
289,374
163,979
377,017
82,745
188,949
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
5.68
4.41
7.25
8.35
10.05
8.96
10.05
9.20
5.94
7.12
8.63
Note: This table shows nancial information on all new rms incorporated in the
years 2006 and 2007. Only rms that have provided records either on operating
revenue, net income, or employees are taken into account (therefore number of
observations vary). Financial data on operating revenue and net income are shown
in thousands of euros. Additionally, nancial data variables are winsorized at the
5th and 95th percentiles. Data is retrieved from Amadeus DVD database (updated
on December, 2008).
Table A4
Ownership structure of new rms incorporated in 2006/2007.
Shareholder type
Number of
observations
Percentage
of total
Individual(s) or Family(ies)
Industrial Company(ies)
Financial Company(ies), Mutual & Pension
Fund/Trusts, Private Equity rm(s)
Unnamed shareholders
Foundations and research institutes
Bank(s) and Insurance Company(ies)
State, Public Authority
411,823
30,813
9183
90.53
6.77
2.02
891
902
830
462
0.2
0.2
0.18
0.10
Note: This table shows the ownership structure by shareholder type for new rms
incorporated in the years 2006 and 2007. New rms are selected only if they have
least one recorded shareholders. Data is retrieved from Amadeus DVD database
(updated on December, 2008).
nancial sector affecting the ability of nancial players to gain access to investment opportunities (Rajan and Zingales, 1998; Beck
et al., 2008). The effect of private credit on new business entry is
not statistically signicant, implying that debt nance is not so
important in VC-intensive industries. The effect of VC remains positive and signicant, and so it is not through VC-intensive industries that private credit affects new business creation.
Next, we account for the fact that private credit is also a volume
measure and thus an imperfect proxy for access to nance. Therefore, we employ a formal proxy for access to nancial services (column (ii)) taken from the World Banks Finance for All? Policies
and Pitfalls in Expanding Access which is a composite indicator
measuring the percentage of the adult population with access to
an account with a nancial intermediary. While this index captures
mere access to business loans, it is a better measure than the volume of private credit of how easy it is to access nancial services in
general. The correlation between the two measures is 0.71, implying that they are highly but not perfectly correlated, and so the formal index could indeed be capturing more access issues than
private credit. While it is positively (but insignicantly) correlated
with new business incorporation, including it in the regression
does not eliminate the effect of venture capital on rm entry.
We also look at investors protection. Rajan and Zingales (2003)
argue that the absence of regulation protecting investors could act
as a barrier to new rm creation. The right measure of nancial
development, the argument goes, would capture not only the ease
with which any entrepreneur or company with a sound project can
obtain nance, but also the condence with which investors anticipate an adequate return. The previous two measures we used
would then be a poor proxy for this investor condence, and we
next proceed to incorporate in our regression a direct measure of
the degree to which individual investments are protected by the legal system in the country. The indicator we employ is a composite
Note: This table shows the trends in start-up activity

aggregated across countries and industries. Fraction of
new rms is dened as the number of rms 2 years or
younger as a fraction of the total rms, at the 2-digit
NACE rev. 1.1 level.
of the quality of three indices: transparency of transactions, liability for self-dealing, and shareholders ability to sue ofcers and
directors for misconduct. This index has a positive (but insignicant) effect on entry of rms in VC-intensive industries, pointing
to the fact that venture capitalists do take into account how the degree of legal protection impacts the expected return to individual
investments in start-up companies. Tellingly, the effect of venture
capital investment on business entry survives this extension of the
basic model (column (iii)).
Finally, we perform a horse race in which we include all country-level measures used in Table 8 so far (column (iv)). We nd
that all other proxies for access to nance have an insignicant effect on de novo rm creation in VC-intensive industries. Importantly, venture capital nance continues to affect positively rm
entry, with a similar order of magnitude.
4.5. Sensitivity to industry and country characteristics
Another concern we need to address is that sensitivity to venture capital investment may be correlated with standard industry
determinants of entry that have been suggested by the literature,
and venture capital investment itself may be picking up the effect
of standard country-level determinants of entry. In particular, it
has been pointed out that entry rates are affected negatively by
nancial dependence, and positively by technological opportunities, among else.17 We want to therefore make sure that the effects
we are measuring are not driven by other industry characteristics
correlated with what makes an industry attractive to venture capitalists. For example, the positive effect of venture capital investment
on entry might be partially driven by the fact that the VC-intensive
industries may also be more dependant on external nance. Hence,
higher entry in those could be caused by the availability of credit
in general, which may be highly correlated with venture capital
investment. We also want to check whether venture capital investment affects entry for industries that are likely to be affected by targeted investment, like R&D intensive industries.18 We do not look at
industry characteristics which have been shown to have an ambiguous or slow effect on rm entry, like protability.
At the country level, Rajan and Zingales (1998) show that nancial development exerts a disproportionately higher effect on new
business creation in industries dependent on external nance.
Klapper et al. (2006) show that entry barriers are associated with
relatively lower entry in industries that have higher natural entry
(and so the relaxation of barriers to entry is relatively more bene17
See Geroski (1995) for a summary of the empirical evidence on that.

Such rms typically have the fastest IPO exits too (see Giot and Schwienbacher,
2007).
18
4707
Table A6
Variables: Denitions and sources.
Variable
Venture capital data
Venture capital
Venture capital
deals
1-year VC
3-year VC
5-year VC
3-year VC industry
3-year VC, 1st
tertile
3-year VC, 2nd
tertile
3-year VC, 3rd
tertile
Industry-level data
Fraction of new
rms
Initial entry
Share
EntryUS
Capital intensityUS
R& D intensityUS
Financial
dependenceUS
Labor intensityUS
Denition and source

Venture capital investment, in 2006 euros, allocated to seed, start-up, development, early, balanced, expansion and later stage investments; by
country. Source: VentureXpert
Number of venture capital investment deals, allocated to seed, start-up, development, early, balanced, expansion and later stage investments; by
country. Source: VentureXpert
Venture capital investment in a countryindustry over the past year. Source: VentureXpert
Average volume of venture capital investment in a countryindustry over the past 3 years. Source: VentureXpert
Average volume of venture capital investment in a countryindustry over the past 5 years. Source: VentureXpert
A dummy variable equal to 1 if the respective countryindustry had any VC investment in the past 3 years. Source: VentureXpert
A dummy variable equal to 1 if the industrycountry belongs in the rst tertile of VC-receiving industries in terms of total volume in the past
3 years. Source: VentureXpert
A dummy variable equal to 1 if the industrycountry belongs in the second tertile of VC-receiving industries in terms of total volume in the past
A dummy variable equal to 1 if the industrycountry belongs in the third tertile of VC-receiving industries in terms of total volume in the past
Number of rms 2 years or younger as a fraction of the total rms, at the 2-digit NACE rev. 1.1 level. Source: Amadeus
Number of rms 2 years or younger as a fraction of the total rms, at the 2-digit NACE rev. 1.1 level, in 1998. Source: Amadeus
Fraction of employment in each countryindustry-year over total employment in that country-year, at the 2-digit NACE rev. 1.1 level. Source:
Amadeus
Entry rates for US corporations. Calculated for 2-digit NACE industries (original data on a 4-digit SIC level). Average for the years 19981999.
Source: Dun & Bradstreet
Measure of physical capital usage, equal to the industry-level median of the ratio of physical capital used to sales. The numerator and
denominator are summed over all years for each rm before dividing. Computed for all US rms for the period 19901999. Calculated for 2-digit
NACE industries (original data on a 4-digit SIC level). Source: Compustat
Measure of dependence on research and development, equal to the industry-level median of the ratio of R& D expenses to sales. The numerator
and denominator are summed over all years for each rm before dividing. Computed for all US rms for the period 19901999. Calculated for 2digit NACE industries (original data on a 4-digit SIC level). Source: Compustat
Industry-level median of the ratio of capital expenditures minus cash ow over capital expenditures. The numerator and denominator are
summed over all years for each rm before dividing. Cash ow is dened as the sum of funds from operations, decreases in inventories, decreases
in receivables, and increases in payables. Capital expenditures include net acquisitions of xed assets. This denition follows Rajan and Zingales
(1998). We compute this measure for all US rms for the period 19901999. Calculated for 2-digit NACE industries (original data on a 4-digit SIC
level). Source: Compustat
Measure of labor usage, equal to the industry-level median of the ratio of number of employees to total assets. The numerator and denominator
are summed over all years for each rm before dividing. Computed for all US rms for the period 19901999. Calculated for 2-digit NACE
industries (original data on a 4-digit SIC level). Source: Compustat
Country-level variables
1-year VC, EVCA
Venture capital investment in a country over the past year. Source: EVCA.
Entrepreneurial
Percentage of 1864 population who are either a nascent entrepreneur or ownermanager of a new business. Source: Global Entrepreneurship
activity
Monitor
GDP per capita
GDP divided by the population in each country-year. Source: Penn Tables
GDP growth
Annual growth of per-capita GDP. Source: Penn Tables
Private credit over
Ratio of domestic credit to the private sector scaled by GDP, in each country-year. Source: International Monetary Funds International Financial
GDP
Statistics (IMF-IFS)
Access to nance
General index of access to external nance by households and businesses, averaged over 19952005. Source: Finance for All? The World Bank,
2007
Investors
Index of the degree of protection of investors, calculated as an average of three indices: transparency of transactions, liability for self-dealing, and
protection
shareholders ability to sue ofcers and directors for misconduct, averaged for 19982008. Source: Doing Business Database (WB)
Int. property
Index of degree of protection of intellectual property rights, averaged for 19982008. Source: Heritage Foundation
protection
Entry cost
Number of procedures to register a business, averaged for 19982008. Source: Doing Business Database (WB)
Labor regulations
Index of protection of workers. Source: Botero et al. (2004)
Capital gains tax
Measure of the marginal tax on capital gains, averaged for 19982008. Source: PriceWaterHouseCoopers Worldwide Taxes (19992008)
Human capital
Average years of schooling for an individual in the respective country, averaged for 19982008. Source: Barro and Lee (2000)
Buyout funds
Private equity funds raised, whose expected allocation is a buyout investment, for each country-year. Source: Thomson VentureXpert
Pension funds
Assets held by all pension funds in the respective country normalized by GDP, for each country-year. Source: Eurostat
cial in those). Finally, the propensity to enter can also be affected

by labor regulations: if the industries that have naturally higher
entry also are the most labor-intensive ones, then lower entry rates
may be the result of stricter labor regulations rather than the
unavailability of start-up nancing. We need to make sure that
such characteristics of the business environment do not contaminate the effect of VC, which would be the case if VC investment
is strongly correlated with those, and if the industries that are of
interest to venture capitalists are also naturally more sensitive to
such alternative developments.
In columns (i) and (ii) of Table 9, we examine the effect of venture capital on industries with higher technological opportunities,
proxied by R& D intensity and capital intensity.19 In terms of the

empirical specication, we simply add to the measure of 3-year
average VC investment at the countryindustry-year level its interaction with industry measures of median capital investment and
R&D investment per worker, respectively. We nd that the effect
of venture capital investment is statistically higher in both capital
intensive and in R&D intensive industries, suggesting that in a Euro-
19
All industry natural intensities are calculated using US data on mature
Compustat rms averaged over the period 19902000. See Appendix F for detail.
4708
Table A7
Changes in prudential rules concerning risk capital investment by pension funds.
Country
Year enacted
Type of change
Austria
Belgium
Czech Republic
Denmark
2003
2003
2003
2003
Finland
France
Germany
Greece
Hungary
1995
2003
2003
2003
1998
Iceland
Ireland
1993
Italy
1993
Netherlands
1993
Norway
Poland
2001
Portugal
Slovakia
2003
2002
Spain
2003
Sweden
2003, 2005
Switzerland
2000
UK
Pre-1991
EU-wide Directive 2003/41/EC

An action plan for risk capital by the Danish government enables owners of private pension funds to spend a share of their funds
on direct investment in non-listed companies
New legislation facilitates the channeling of domestic savings into private equity funds and the new national pension scheme
now involves a privately managed element
None
The government requests the pension fund industry to increase its investment in Irish unquoted equities as a means of
developing small and medium sized businesses. The pension fund industry and government jointly undertake a study of the
equity requirements of SMEs and also the performance of unquoted equities over the past 20 years. The study concludes that the
provision of funds to established VCs with good track records is an activity which the members of the pension fund industry
should participate in
On 8 April 1993, as a part of the Pension Reform, a specic law on private pension plans is approved. One of its requirements is
that closed-end funds must invest between 40% and 80% of the total value of the fund in unquoted companies
A new fund is initiated by the Dutch government for larger investments in larger industrial rms. The fund, partly nanced by
banks, insurance companies, and pension funds, amounts to some D 1 billion
None
A law comes into place providing Polish investment institutions with new types and structures of investment funds, such as
securitisation funds. The law also permits the soliciting of funds from the Polish nancial institutions, such as pension funds and
insurance companies, for investment into private equity
The government initiates pension and tax reforms, in part intended to produce measures to allow pension funds to invest in risk
capital
Spanish pension funds are allowed to invest up to 30% of their assets in free transferable securities issued by audited private
equity entities, provided that the investment does not imply a direct or indirect control over the investor entity and there are no
economic links with the investor entitys shareholders or directors
EU-wide Directive 2003/41/EC. A new legislation enters into force stating that pension funds and employment saving trusts may
not invest in partnerships or act as a general partner in a limited liability partnership. The reason behind is that such investments
are deemed too risky for the beneciaries of the trusts
Swiss pension fund regulations are revised as of April 1, 2000 in order to permit Swiss pension funds to invest in both Swiss and
foreign VC funds
Pension funds are allowed to invest in risk capital
pean context VC is complementary to both R&D and to capital

investment in stimulating rm entry.
In columns (iii)(v), we run our main regression, accounting for
three other industry characteristics dependence on external nance, labor intensity, and propensity to entry in combination
with nancial development, labor market rigidities, and entry barriers, respectively. We nd that all of these characteristics of the
business environment matter for entry exactly as predicted. For
example, higher private credit is associated with higher rm entry
in nancially dependent industries (column (iii)). In addition to
that, higher labor market rigidities (column (iv)) and higher barriers to entry (column (v)) are associated with lower rates of new
business creation in industries that are more labor-intensive (proxied by the workers-to-assets ratio for Compustat rms over the
1990s) and that have higher rates of new business creation (proxied by the average share of new entry in the respective US industries over the 1990s), respectively. Importantly, in all cases
higher venture capital investment keeps its independent effect
on entry.
Finally, the effect of venture capital on entry survives another
horse race: when all interactions from columns (iii)(v) are included in the regression, the effect of venture capital still stands.
4.6. Other characteristics of the business environment
In Table 10, we turn to the characteristics of the business environment that can in theory augment the effect of VC investment on
entry. The large cross-country dataset we work with is ideally
suited to answer this question. We interact venture capital investment with the rest of the relevant country-level variables we have
so far been working with - namely, barriers to entry (proxied by

the number of procedures necessary to start a business), labor regulations (proxied by how difcult it is to hire and re workers),
intellectual property rights protection, tax on capital gains, and human capital (proxied by average years of schooling).
We nd that the effect of venture capital on new business creation depends on entry costs. While one would expect that the
ability to raise external funds will enable entrepreneurs to start
new business more easily in countries with lower barriers to entry,
institutional substitutabilities could cancel that effect if access to
nance tends to decrease the marginal cost of meeting certain regulations (Ahlin and Pang, 2008). We nd that the second effect
dominates in the sense that VC helps nascent entrepreneurs overcome the monetary costs of setting new companies (columns (i)
and (vi)). Venture capital investment has a stronger effect on de
novo rm creation in countries with less strict labor regulations.
In this case, VC is complementary to labor market exibility,
implying that the benets of VC on new business creation can only
be properly realized if hiring workers is not prohibitively expensive or difcult. However, the effect is not signicant. We also nd
that the effect of VC on rm entry is stronger in countries with
higher intellectual property rights protection (columns (iii) and
(vi)). One explanation for this effect is that stronger protection of
intellectual property rights ensures a higher return to investment
in intangible capital. In countries with lower levels of protection
venture capitalists and nascent entrepreneurs may be deterred
by the fact that the system will not protect innovative products
adequately. This result broadly conrms a strand of research which
has documented the value of VC-university hubs in fostering entrepreneurship (see Gompers and Lerner, 1999). We also nd that
venture capital investment has a stronger effect on new business

creation in countries with lower taxes on capital gains (columns
(iv) and (vi)). This regularity relates to previous research which
has found that a larger portion of venture funds raised is allocated
to early-stage nance in countries where the returns to entrepreneurship are higher (Da Rin et al., 2006). Finally, variations in
schooling across countries do not explain a signicant portion of
the variation in the sensitivity of de novo rm entry to VC investment in the horse race (column (vi)).
5. Conclusion
We use a large panel of 21 countries and 20 industries over the
19982008 period to identify the impact of venture capital investment on entrepreneurship. We use the Amadeus database, which
includes data on about three million private rms all across Europe
over the 19982008 period, and match that data to data from VentureXpert on the size and number of venture capital deals. We nd
that venture capital investment has a sizeable effect on new
business creation: numerically, an increase in venture capital
investment by a factor of 7.2 (the difference between an industry
at the 25th and an industry in the 75th percentile of VC investment) leads to an increase in the share of new rms for the medium-entry industry by between 3% and 19%, depending on the
estimation approach. This effect is relatively higher in countries
with higher entry costs, higher protection of intellectual property
rights, and lower taxes on capital gains. This effect holds for different proxies for VC investment and for various partitioning of the
sample. Most importantly, the results stand when we correct for
the endogeneity of the venture capital investment series by using
an IV procedure in which the variation in buyout fund-raising
and in prudential regulation guiding the investment behavior of
domestic pension funds is used as an instrument to identify the
supply of VC funds. The effect of venture capital investment on entry is generally robust to accounting for other industry characteristics, as well as for other characteristics of the business
environment that have been suggested by the literature as important determinants of new business creation, most notably barriers
to entry.
This paper is, to our knowledge, the rst attempt to empirically
link venture capital investment to industry entry in a large crosscountry setting. Our results strongly suggest that venture capital
investment is conducive to bringing new ideas to the marketplace
in the shape of young companies. In this paper we have nothing to
say about the private return to this process, but as far as aggregate
economic effects are concerned, VC investment does seem to generate value through fostering entrepreneurial activity in the economy. A number of important questions remain unanswered due to
the nature of our data. For example, what is the relative importance of the different channels via which venture capital affect entry? Is it more anticipatory considerations where nascent
entrepreneurs are aware of future nancing needs and are more
likely to decide to start their own business at a time when more
venture capital is available? Or is it entrepreneurial spawning
where venture-backed rms are more likely to create other new
ventures? Future research can greatly contribute by addressing
those questions.
Appendix A. Industry concordance key and Amadeus data
process
The U.S. Census Bureau provides a concordance key between SIC
and NAICS and between NACE Rev. 1.1 and NAICS. To each NAICS 6digit class, we match the corresponding SIC and NACE Rev. 1.1 classes. This results in a non-unique matching between 2-digit SIC and
4709
2-digit NACE Rev. 1.1 classes. Therefore, we collapse 52 SIC codes

and the 39 NACE codes into 20 industry classes such that no information is lost. This results in several aggregated industrial classes;
for example, Industry 13 (Machinery and equipment n.e.c.; ofce
machinery and computers; electrical machinery n.e.c.; radio, television, and communication equipment; medical, precision, and
optical instruments) now includes 3 SIC and 5 NACE Rev. 1.1 classes. See Appendix B for the resulting concordance key.
We employ a similar sample selection procedure to the Amadeus dataset as Klapper et al. (2006) and Da Rin et al. (2011). From
each Amadeus DVD, the last year is not used due to lags in data collection and hence intrinsic incompleteness of the information. The
years prior to three years before the current year are not used due
to the survivorship problems of the Amadeus database: when a
rm ceases existing, Amadeus keeps a record of it for 4 years,
and then takes it out of the database. Consequently, while each
yearly addition of Amadeus contains data for more than 4 years
back, the sample of rms one will nd reported for the time-period
4 years and more prior to the year of issuing, will not include many
rms which existed in that year, but exited the market after that.
For example, the data for rms in 2000 as reported in the 2004
database will not include rms that exited in 2000 or before. Using
the data indiscriminately will therefore induce survivorship bias
and misrepresent the volume of entry, and so we should focus
our attention to the years 23 prior to the year the database was
issued. In practice, for a set of 11 databases over 19982008, we focus in each one on the year two years before the year of issuance
(that is, we use the data on 2002 from the 2004 database, and so
on).
We drop rms that report only unconsolidated statements in
order to avoid double-counting rms and subsidiaries abroad.
We exclude subsidiaries. We exclude industries where the activities are country-specic, namely agriculture, forestry, shing and
mining. We exclude Luxembourg due to insufcient availability
of data on venture capital investment, although this country is
covered by Amadeus. We also exclude utilities and post and telecommunications, which tend to be heavily regulated and/or
state-owned, and the nancial services sector because nancial
rms are subject to specic regulations which do not apply for
other rms (for example, initial capital requirements). We do not
exclude proprietorships and partnerships. Finally, we exclude the
public sector, education, the social sector, private households,
and activities that cannot be classied. We are left with 37 NACE
industries. The nal sample consists of an annual average of
6,189,877 rms in 21 countries: Austria, Belgium, Czech Republic,
Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Netherlands, Norway, Poland, Portugal, Slovakia, Spain,
Sweden, Switzerland, and UK.
Appendix B
Table A2.
Appendix C
Table A3.
Appendix D
Table A4.
Appendix E
Table A5.
4710
Appendix F
Table A6.
Appendix G
Table A7.
References
Aernoudt, R., 1999. European policy towards venture capital: myth or reality?
Venture Capital 1, 4758.
Aghion, P., Fally, T., Scarpetta, S., 2007. Credit constraints as a barrier to the entry
and post-entry growth of rms. Economic Policy 22, 731779.
Ahlin, C., Pang, J., 2008. Are nancial development and corruption control
substitutes in promoting growth? Journal of Development Economics 86,
414433.
Barro, R., Lee, J.-W., 2000. International data on educational attainment: updates
and implications. Oxford Economic Papers 53, 541563.
Barro, R., Sala-i-Martin, X., 1992. Convergence. Journal of Political Economy 100,
223251.
Beck, T., Demirg-Kunt, A., Maksimovic, V., 2005. Financial and legal constraints to
growth: does rm size matter? Journal of Finance 60, 137177.
Beck, T., Demirguc-Kunt, A., Laeven, L., Levine, R., 2008. Finance, rm size, and
growth. Journal of Money, Credit, and Banking 40, 13791405.
Bernstein, S., Lerner, J., Sorensen, M., Stromberg, P., 2010. Private Equity and
Industry Performance. NBER Working paper 15632.
Botero, J., Djankov, S., La Porta, R., Lopez-de-Silanes, F., Shleifer, A., 2004. The
regulation of labour. Quarterly Journal of Economics 119, 13391382.
Bottazzi, L., Da Rin, M., 2002. Venture capital in Europe and the nancing of
innovative companies. Economic Policy 17, 229270.
Carpenter, R., Petersen, B., 2002. Capital market imperfections, high-tech
investment, and new equity nancing. Economic Journal 112, F54F72.
Cetorelli, N., Strahan, P., 2006. Finance as a barrier to entry: bank competition and
industry structure in local US markets. Journal of Finance 61, 437461.
Cosh, A., Cumming, D., Hughes, A., 2009. Outside entrepreneurial capital. Economic
Journal 119, 14941533.
Cumming, D., Fleming, G., Suchard, J.-A., 2005. Venture capitalist value-added
activities, fundraising and drawdowns. Journal of Banking & Finance 29, 295
331.
Da Rin, M., Nicodano, G., Sembenelli, A., 2006. Public policy and the creation of
active venture capital markets. Journal of Public Economics 90, 16991723.
DaRin, M., DiGiacomo, M., Sembenelli, A., 2011. Entrepreneurship, rm entry, and
the taxation of corporate income: evidence from Europe. Journal of Public
Economics 95, 10481066..
Geroski, P., 1995. What do we know about entry? International Journal of Industrial
Organization 13, 421440.
Gilson, R., 2003. Engineering a venture capital market: lessons from the American
experience. Stanford Law Review 55, 10671103.
Giot, P., Schwienbacher, A., 2007. IPOs, trade sales and liquidations: modelling
venture capital exits using survival analysis. Journal of Banking & Finance 31,
679702.
Gompers, P., Lerner, J., 1999. The Venture Capital Cycle. Cambridge University Press,
Cambridge, MA.
Gompers, P., Lerner, J., 2001a. The Money of Invention. Harvard Business School
Press, Boston, MA.
Gompers, P., Lerner, J., 2001b. The venture capital revolution. Journal of Economic
Perspectives 15, 145168.
Gompers, P., Lerner, J., Scharfstein, D., 2005. Entrepreneurial Spawning: public
corporations and the formation of new ventures, 19861999. Journal of Finance
60, 577614.
Hansen, L.P., 1982. Large sample properties of generalized method of moments
estimators. Econometrica 50, 10291054.
Hellmann, T., 1998. The allocation of control rights in venture capital contracts.
RAND Journal of Economics 29, 5776.
Hellmann, T., Puri, M., 2002. Venture capital and the professionalization of start-up
rms: empirical evidence. Journal of Finance 57, 169197.
Hirukawa, M., Ueda, M., 2008. Venture Capital and Industrial Innovation. CEPR
Discussion paper 7089.
Kaplan, S., Stromberg, P., 2001. Venture capitalists as principals: contracting,
screening, and monitoring. American Economic Review 91, 426430.
Keuschnigg, C., 2004. Venture capital backed growth. Journal of Economic Growth 9,
239261.
Klapper, L., Laeven, L., Rajan, R., 2006. Entry regulation as a barrier to
entrepreneurship. Journal of Financial Economics 82, 591629.
Klepper, S., 1996. Entry, exit, growth and innovation over the product life cycle.
American Economic Review 86, 562583.
Kortum, S., Lerner, J., 2000. Assessing the contribution of venture capital to
innovation. RAND Journal of Economics 31, 674692.
La Porta, R., Lopez de Silanes, F., Shleifer, A., Vishny, R., 1998. Finance and law.
Journal of Political Economy 106, 11131155.
Mollica, M., Zingales, L., 2007. The Impact of Venture Capital on Innovation and the
Creation of New Businesses. University of Chicago Working paper.
Rajan, R., Zingales, L., 1998. Financial dependence and growth. American Economic
Review 88, 559586.
Rajan, R., Zingales, L., 2003. The great reversal: the politics of nancial development
in the 20th century. Journal of Financial Economics 69, 550.
Sahlman, W., 1990. The structure and governance of venture capital organizations.
Journal of Financial Economics 27, 473521.
Samila, S., Sorenson, O., 2011. Venture capital, entrepreneurship, and regional
economic growth. Review of Economics and Statistics 93, 338349.
Samila, S., Sorenson, O., 2008. Venture Capital, Entrepreneurship, and Regional
Economic Growth. Yale University working paper.
Schwienbacher, A., 2008. Innovation and venture capital exits. Economic Journal
118, 18881916.
Sevilir, M., 2010. Human capital investment, new rm creation, and venture capital.
Journal of Financial Intermediation 19, 483508.
Stock, J., Yogo, M., 2005. Testing for weak instruments in linear IV regression. In:
Andrews,. W.K., Stock, J.H. (Eds.), Identication and Inference for Econometric
Models: Essays in Honor of Thomas Rothenberg. Cambridge University Press,
Cambridge, MA.
Tang, M.-C., Chyi, Y.-L., 2008. Legal environments, venture capital, and total factor
productivity. Contemporary Economic Policy 26, 468481.

Venture Capital and New Business Creation

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Venture Capital and New Business Creation

Uploaded by

Copyright:

Available Formats

Journal of Banking & Finance 37 (2013) 46954710

Contents lists available at ScienceDirect

Journal of Banking & Finance

Venture capital and new business creation q

European Central Bank, Financial Research Division, Germany

than banks (Carpenter and Petersen, 2002; Schwienbacher, 2008;

A. Popov, P. Roosenboom / Journal of Banking & Finance 37 (2013) 46954710

we use the variation across countries and over time in buyout

A. Popov, P. Roosenboom / Journal of Banking & Finance 37 (2013) 46954710

% New rms (1 and 2 years old)

namely, the ratio of the number of rms in each country present in

A. Popov, P. Roosenboom / Journal of Banking & Finance 37 (2013) 46954710

2006 as baseline year for the purpose of the empirical analysis.

Entryijt b0 b1 Venture capitalijt b2 Shareijt b3 Di

See EVCA yearbook 2004 for details.

A. Popov, P. Roosenboom / Journal of Banking & Finance 37 (2013) 46954710

4.1. Venture capital and new business creation: main results

As in Klapper et al. (2006), Entryijt denotes the share of rms less

In Table 5, column (i) we present the estimates from the basic

A. Popov, P. Roosenboom / Journal of Banking & Finance 37 (2013) 46954710

Fraction of new rms

3-Year VC, 2nd tertile

Country industry year

Country Year Industry Year

Report signicance at the 10% level.

Report signicance at the 5% level.

Report signicance at the 1% level.

correlated with the rates of new business creation. Nevertheless,

4.2. Addressing various data issues

trycountry pair in 1998, and excludes all observations for 1998

A similar sample selection is applied to Amadeus by Klapper et al. (2006).

A. Popov, P. Roosenboom / Journal of Banking & Finance 37 (2013) 46954710

Fraction of new rms

1-Year VC, EVCA

Country industry year

Report signicance at the 10% level.

Report signicance at the 5% level.

Report signicance at the 1% level.

A. Popov, P. Roosenboom / Journal of Banking & Finance 37 (2013) 46954710

See Appendix G for details.

A. Popov, P. Roosenboom / Journal of Banking & Finance 37 (2013) 46954710

Fraction of new rms

Log 3-year lagged pension funds Pension reform Pre-reform VC

3-Year VC High GDP per capita

Country industry year

Report signicance at the 10% level.

Report signicance at the 5% level.

Report signicance at the 1% level.

In columns (iv)(vi), our main measure of VC is instrumented

both higher entrepreneurial culture (resulting in more entry

A. Popov, P. Roosenboom / Journal of Banking & Finance 37 (2013) 46954710

Fraction of new rms

3-year VC industry Investors protection

Country industry year

Report signicance at the 1% level.

Report signicance at the 10% level.

Report signicance at the 5% level.

Fraction of new rms

Financial dependenceUS Private credit/GDP

Labor intensityUS Labor regulations

EntryUS Entry cost

Country industry year

Report signicance at the 10% level.

Report signicance at the 5% level.