Web Appendix for: Venture Capital in Europe and the Financing of Innovative Companies

Laura Bottazzi and Marco Da Rin 1 Università Bocconi and IGIER, and Università di Torino and IGIER

Published in:

Economic Policy, 34, Spring 2002

CONTENTS

This Web Appendix contains 7 parts. Part 1 provides additional descriptive statistics for our dataset. Part 2 provides the details on our probit regression that predicts which firms get venture capital backing (these were refereed to in Section 4 of the main text). Part 3 provides additional regressions, which we comment in the main text, where we control for R&D. Part 4 provides the results from two alternate methodologies for gauging the impact of venture capital on the performance of innovative firms (matching and difference in differences). Part 5 is our data appendix. And finally, Part 6 provides a complete list of references.

1 We thank Erik Berglöf, Jean-Bernard Chatelain, Jan van Ours, Henri Pagès, Guido Tabellini, and participants to the Economic Policy 33rd Panel Meeting for valuable comments. Detailed suggestions by Richard Baldwin (the editor) helped us improve the quality of the paper. Veronica Guerrieri, Giuseppe Maraffino, Gaia Narciso, and Battista Severgnini provided excellent research assistance. We also thank all the companies that provided us with data and prospectuses. Financial support from Fondation Banque de France, from the Italian Ministry of University and Research (MIUR), and from Università Bocconi (Ricerca di Base) is gratefully acknowledged. All errors remain our own.

1. ADDITIONAL DESCRIPTIVE STATISTICS

Here we provide three tables additional information from our dataset. The first lists the frequency of investments by venture capital firm, the second shows the sectoral distribution of VC investments and the third shows corporate growth statistics.

Table A-1. Frequency of venture capital investment

Venture capitalist Nationality Number of

investees 3i Group INT(UK) 27 Apax Partner INT(UK) 12 Galileo Partners F 12 Gold-Zack AG D 12 Technologie-Beteiligungsgesellschaft D 12 ABN Amro Ventures INT(NL) 10 BNP Developpement F 10 Financiére Natexis F 10 Atlas Ventures INT(US) 9 Group de Rothschild INT(F) 8 Soffinova Partners F 8 TBG-Technologie-Beteiligungsges. D 8 Banexi Ventures Partners F 7 Dassault Developpement F 7 CDC Innovation F 6 Sofimac Partners F 6 Techno Venture Management Gmbh D 6 Commerz Beteilugungsgesellschaft D 5 Concord Effekten AG D 5 DEWB-Deutsche Effekten und Wech.-B. D 5 IKB Beteiligungsgesellschaft D 5 Knorr Capital Partner AG D 5 TechnoStart GmbH D 5 Vertex Management LTD INT(SIN) 5 AXA F 4 Bank Austria A 4 Banque de Vizille (Group CIC) F 4 HSBC Private Equity INT(UK) 4 IRDI de Midi F 4 Schroeder INT(UK) 4 Thompson Clive & Partners INT(UK) 4 Alpinvest F 3 BUWB Bayerische D 3 CEA Capital Partner GmbH & Co. Beteiligungs KG D 3 Europ@web (Groupe Arnault) F 3 TFG Venture Capital AG D 3 UCA Unternehmer Consult AG D 3

Table A-1 (continued)

Financiere Vecteur F 3 LBB Beteiligungsgesellschaft D 3 Paribas F 3 Partcom SA (group CDC) F 3 Private Equity Partners I 3 Sal. Oppenheim jr & Cie. D 2 S-UBG AG D 2 Advent International INT(US) 2 Alta Berkeley US 2 Apollo Invest F 2 BayBG Bayerische D 2 BB-Kapitalbeteiligungsges. D 2 bmp AG D 2 BW-Venture Capital D 2 DG-Private Equity D 2 Epicea SA F 2 IBB Beteiligungsgesellschaft D 2 Initiative and finance F 2 KB Partners, LLC US 2 Mivtah Shamir IL 2 Pechel Industries F 2 Pino Venture Partners I 2 Saarländische Kapitalbeteiligungsges. D 2 SNBV Partecipations F 2 Sogginove (groupe Societé Generale) F 2 Sopromec F 2 TechnoStart GmbH D 2 Transconnect GmbH D 2 Value Management & Research D 2 Ventech F 2 Venture-Capital Baden-Wuerttenberg D 2 WeHaCo Kapitalbeteiligungs GmbH D 2 West-LB D 2 Other venture capitalists with one investment 122 Notes: INT stands for ’international,’ i.e. a venture capitalist with active offices in more than three countries. In brackets the nationality of the headquarters. Source: Authors’ dataset.

Table A-2. Venture capital, sectoral specialization

Euro.NM VC-backed % % Biomed 8 12 Traditional (manufacturing and services) 4 3 ITSIS (IT Services, Internet, Software) 57 53 Media & Entertainment 10 8 Technology 16 17 Telecommunications 5 7

Source: Authors’ dataset.

Table A-3. Corporate growth: descriptive statistics

Pre-IPO Mean Median Min Max S.D. Obs. Assets 113 7 0 44,823 2,082 463 Debt 28 4 0 7,967 370 462 Equity 72 1.3 0 28,213 1,383 416 EBITDA 5.8 0.5 -18.1 1,679 79 456 Leverage 0.7 0.8 0 1 0.2 414 ROA 0.1 0.1 -6.1 5.1 0.7 455 Sales 52 7.5 0 12,344 576 463 Employees 141 68 0 1,538 216 413 Capex 1.8 0.6 -2.8 37 4 342 Foreign sales (%) 0.2 0 0 1 0.3 282 Intangible assets 2.6 0.2 0 30.8 8.6 372 R&D 2 0.8 0 17.4 3 156 R&D intensity 0.8 0.1 0 20 1.7 158 Post-IPO Assets 195 53.1 0.2 13,187 1,013 355 Debt 64.2 14.9 0.2 5,194 302 354 Equity 123 33.5 0 7,959 616 350 EBITDA -17.1 1.8 -7,035 338.9 379 352 Leverage 0.4 0.3 0 1 0.2 349 ROA 5.1 0.04 -140 1,853 99.5 349 Sales 121 30 0 18,287 967 360 Employees 399 194 10 3,210 553 332 Capex 37.7 6.3 -96 3,860 235 293 Foreign sales (%) 0.2 0.1 0 1 10.3 147 Intangible assets 39.5 6 0 2,249 162 299 R&D 13.8 3.3 0 731 65.6 152 R&D intensity 0.4 0.1 0 32 2.6 152

Notes: For ease of comparability, we drop 34 companies with negative equity values in the pre-IPO period, and three companies with negative values in the post-IPO period. Source: Authors’ calculations.

2. WHICH COMPANIES ARE VENTURE-BACKED?

This section investigates the question of which characteristics of a firm are associated with receiving venture capital financing. Theory predicts venture capital to be associated with young, innovative companies that, being at an early stage of development, are characterised by low profitability and a small amount of sales.2 In the main text, we showed that European venture capitalists have been increasing their early stages investment. Therefore we expect our findings to conform to the predictions of the theory.

2 We are not aware of any statistical study of the determinants of venture financing for the US, except for Hellmann and Puri (2000), who look at a sample of venture-backed start-ups and find that those that pursue more radical innovations are more likely to attract venture capital.

We estimate a probit regression in which the dependent variable is a dummy variable that takes value one if a company has obtained venture capital financing. The independent variables are measured before the arrival of the first venture capitalist (’preVC’).3

Unfortunately we cannot use in this analysis all the companies in the data set, since there are some missing observations, and since 30 companies, i.e. 14% of the venture-backed companies in our sample, have received venture capital funding before they started reporting accounting information. Still, this leaves us with 359 companies.

Table A-4 reports the probit regression results. We find that sales negatively affect the probability of obtaining venture capital financing, while leverage has a positive effect, although it is not statistically significant. We control for sectors of activity, but these are found to have no effect. Our findings are consistent with a view of venture capital getting involved with firms which are still at a very initial stage of development and are therefore not yet able to sell. A marginal increase in sales decreases the likelihood of receiving venture financing by 0.3%, an economically small but statistically highly significant result. The positive effect of leverage is consistent with a view of venture capital as an important source of financing. In other words, the ’hard’ side of venture capital goes well along its ’soft’ side. Table A-4. Probit regression-dependent-variable venture capital

thfimfihist

3 Fsuwe4 Ttheob

Independent Marginal increase Coefficient z-statistic Variables in probability Sales(preVC) -0.003 -0.01 *** -2.651 Leverage(preVC) 0.117 0.34 1.482 Constant -0.45 -1.131 Number of obs. 349 Log likelihood -184.82

2 (7)Waldχ 19.74

P-value 0.006 Notes: ‘preVC’ denotes variables measured before the arrival of a venture capitalist. Significance levels are indicated by * (10%), ** (5%), and *** (1%). Huber-White corrected standard errors are used to obtain robust estimates.

Alternative (unreported) specifications have considered the level of debt, its maturity, e amount of assets and a national market effect as possible determinants of venture nancing.4 We have done so to check if companies whose debt is mostly short-term ight be more credit constrained and might therefore look more aggressively for venture nancing. A dummy that takes care of the national effect was also used to capture the gher proportion of French venture-backed companies. All these variables turn out to be atistically insignificant, and in all specifications the quality of the fit worsens.

or non venture-backed firms, we use the average of the pre-IPO values. We also experimented with alternative measures,

ch as measuring variables in the years before the average date of entry of venture capital in venture-backed companies, but found no substantial difference in the results. Hence we stick to the simpler pre-IPO measure. he national effect is measured through dummies which take value one for companies listed in France, Belgium, Italy, and Netherlands. As in all other regressions, the latter three never turn out to be significant, also due the low number of servations.

We also expect venture capital to be involved with more highly innovative companies. To check for this prediction we control for the natural proxy for innovativeness, R&D intensity, measured as the ratio of R&D to assets.5

This reduces the number of observations to 101, since few companies disclose R&D expenditure. The reported R&D figures are however reliable, since they are voluntarily disclosed in the IPO prospectuses as a signal of the company’s quality and are not a legally required item of the accounts. We find that a higher R&D intensity makes a company less likely to receive venture capital financing (see below), but the result is not statistically significant. The amount of sales remains the driving force behind obtaining venture financing, and industry controls are all significant.6

3. STATISTICAL ANALYSIS, CONTROLLING FOR R&D

In the tables below, preVC indicates variables that are measured before the arrival of the venture capitalist, while ’at IPO’ indicates variables that are measured at the time of the IPO. Significance levels are indicated by * (10%), ** (5%), and *** (1%). Huber-White corrected standard errors are used to obtain robust estimates.

Table A-5. Probit regression-dependent variable venture capital

Independent Marginal

increase Coefficient t-statistic

Variables in probability R&D intensity (preVC) -0.05 -0.15 -0.763 Sales (preVC) -0.006 -0.02 * -1.816 Leverage (preVC) 0.221 0.75 1.522 Constant -3.04 *** -4.688 Number of obs. 101 Log likelihood -54.02

2 (7)Waldχ 8.7

P-value 0.001

5 Our results do not vary if we control for R&D expenditure. 6 The finding that R&D has a negative (albeit statistically weak) effect is only apparently at odds with the theory, however, since we are measuring variables before the arrival of the venture capitalist. The result is in fact consistent with venture capital selecting companies at an early stage of development, when R&D intensity is still low. However, one could conceivably think of alternative interpretations, for example that companies whose R&D intensity is higher are less credit constrained and therefore less likely to look for venture financing.

Table A-6. Cox regression-dependent variable time-to-listing

Independent variables Hazard ratio t-statistic R&D (at IPO) 1.00 *** 7.01 Venture capital 0.83 -1.22 Leverage (at IPO) 1.07 0.17 ROA (at IPO) 1.00 -1.30 France 0.34 *** -5.56 Age*97 1.01 *** 7.29 Age*98 1.01 *** 6.31 Age*99 1.00 * 1.70 Age*100 1.00 *** -3.13 Number of obs. 186 Log likelihood -752.98

2 (14)Waldχ 194

P-value 0.00

Table A-7. Robust regression-dependent variable funds raised over assets

Independent variables Coefficient t-statistic R&D intensity (at IPO) -0.005 -0.424 Sales/Assets 0.065 1.574 Venture Capital 0.49 1.241 Leverage (at IPO) 2.717 ** 2.327 ROA (at IPO) -1.184 *** -3.416 Age -0.005 *** -2.510 France -1.883 *** -4.165 Constant 3.642 ** 2.010 Number of obs. 180

(12,167)F 4.07 P-value 0.000

Table A-8. Robust regression-dependent variable employment growth

Independent variable Coefficient t-statistic R&D (at IPO) 11.59 *** 4.037 Venture capital -27.94 -1.134 ROA (at IPO) -7.82 -0.256 Leverage (at IPO) 185.51 *** 2.948 Foreign sales -46.11 * -1.829 Age 0.03 0.811 France -69.68 *** -2.548 Constant 77.12 1.204 Number of obs. 117

(11,105)F -4.43 P-value 0.000

Table A-9. Robust regression-dependent variable sales growth

Independent variables Coefficient t-statistic R&D (at IPO) 2.04 *** 5.041 Venture capital -7.90 *** -2.435 ROA (at IPO) -3.90 * -1.366 Leverage (at IPO) 18.10 ** 2.015 Foreign sales -1.22 -0.343 Age 0.03 ** 2.351 France -1.10 -0.301 Constant 9.54 1.263 Number of obs. 143

(11,131)F 5.31 P-value 0.000

4. ROBUSTNESS CHECKS FOR OUR REGRESSION ANALYSIS

Section 5 of the main text tried to evaluate the impact of venture capital financing on the companies listed on Euro.nm. The ’evaluation problem’, as it is known in the econometric literature, is the problem of correctly measuring the effect of a ’cure’ – such as a policy reform or a training program – on some variables (see Blundell and Costas Dias (2000)). The problem in evaluating a cure is that both observable and unobservable variables may be present, which might bias the estimates if not properly accounted for.

In our case the correct approach to assess the effect of venture capital (the ’cure’) should look at certain companies and compare their reaction when they do and when they do not receive venture financing. Unfortunately this is not possible as our companies are either venture-backed or non venture-backed, and receiving venture capital is not a random event. The issue is then how to construct the right counterfactual.

In the impossibility of obtaining experimental data, researchers have adopted different methods of evaluation. We consider two different methodologies. One approach is known as the matching method, and mainly addresses the issue of bias due to incorrect control for observable variables. The second approach is known as the difference in differences method, and it is particularly useful in removing unobservable individual effects and common macro effects. We thus re-evaluate the effect of venture capital assessing whether our previous estimates are subject to these biases.

4.1. The ’matching’ method

What we have tried to measure in our analysis is the effect of being venture-backed on sales and employment growth, on the amount raised at IPO, and on the time-to-listing. Denote any of these variables with Y. The analysis then entails measuring the average effect (τ) of venture capital on venture-backed companies:

( / ) ( /i ivc vcE Y i VC E Y i NVCτ = ∈ − ∈ )

)

(1)

where the first element on the right hand side measures the expected value of venture capital on variable Y conditional on company i being venture-backed, while the second term measures the expected value of the same variable, were company i without venture capital. The problem is that this last term is not observable. It is impossible to see the characteristics of venture-backed companies in the absence of venture capital. In other words we lack a proper control sample. Being backed by venture capital is in fact not random, as we have observed when we have estimated the probability of receiving venture financing. As a consequence, the assignment process to venture capital might be determined by (observable) variables (Xi) that potentially affect the outcome Y as well. If that effect turned out to be important, our previous estimates might be biased and our conclusions flawed.

To exploit the information we get about Y from non venture-backed companies, which act as our control sample, we assume that, conditional on Xi, the value of Y (time to listing (TTL), amount raised at IPO, sales or employment growth) and the fact of being venture-backed are independent. Under this assumption equation (1) can be re-written as:7

(2) ( / , ) ( / ,i i

vc i vc iE Y i VC X E Y i NVC Xτ = ∈ − ∈ We can now estimate equation (2) non-parametrically. In order to correctly measure

the effect of venture capital we need to estimate the two terms on the right hand side of equation (2), matching each venture-backed company with a non venture-backed company with the same characteristics Xi. In other words, we need to find a way to compare observations with similar Xi. Only in this case the different behaviour of the two companies can be correctly attributed solely to the presence of venture capital.

When Xi is high dimensional the estimation strategy may become unfeasible. Still, one can resort to matching companies not on the values of Xi but on a function of Xi. We do this through the ’propensity score’ method, which we illustrate in detail in the Web Appendix. We then proceed to estimate the average effect of being venture-backed, taking into account that by using the matching method we reduce the bias due to specification error but possibly at the cost of losing efficiency. In other words we could obtain estimates with a lower statistical significance.

Table A-10 reports the estimates obtained with this method for our four variables. The estimated values of τ provide measures of the effect of venture capital different from those we obtained in the previous sections. Venture capital is now found to slightly (but statistically significantly) increase TTL. As regards the average effect of venture capital on the amount raised at IPO the matching method suggests a negative effect, casting some doubts on the positive effect of venture capital which we found in the main text.

7 This assumption is known in the literature as ’conditional independence.’

However, this method confirms our finding of a negative effect of venture financing on the post-IPO growth of sales and employment. The standard errors of amount raised, employment and sales growth (but not that of TTL) in fact become considerably larger, but, as we observed, this is not surprising for this type of estimates.

Overall, we conclude that the inference from our regression analysis may not be robust to possible specification biases with respect to observable variables in the case of TTL and amount raised. In the case of sales and employment growth, instead, we find reasons to remain confident on the robustness of our findings. Table A-10. Nonparametric stratification estimates: Average venture capital effect

4.1

τ t-ratio Time-to-listing 0.42 *** 5.276 Amount raised (over assets) -10.19 0.639 Employment growth -21.478 0.434 Sales growth -36.678 0.513 Notes: Significance levels are indicated by * (10%), ** (5%), and *** (1%).

.1. The stratification estimator Here we discuss in more detail the implementation of the stratified estimation method

above. When the dimensionality of the vector Xi becomes unwieldy, one possibility is to make use of the ’propensity score’ method, i.e. to estimate the probability (p(Xi)) of obtaining venture capital conditional on observables Xi. Observations with the same propensity score will then have the same distribution of the full vector of observables Xi This methodology entails a two-step strategy. We first estimate the propensity score, and subsequently the conditional average of the variable venture capital.

In order to estimate the average effect of venture capital we need first to estimate the propensity score for each observation of our sample. We use a probability model where we introduce the following covariates (Xi), which we measure at IPO: return on assets (ROA), sales, the dummy for foreign sales,, and controls for country and sector of activity.

Then, following Dehejia and Wahba (1998) we stratify the estimated propensity scores for venture-backed and non venture-backed firms into five blocks (’bins’) of equal score range (0-0.2, … 0.4-0.6). Ideally we would like to have the same frequency of venture-backed and non venture-backed firms in each bin. We check whether we succeed in balancing the covariates within each stratum by testing for the equality of the first and second moments of covariates within each stratum. If there are no differences we accept that specification. Otherwise we split the block and test again. Six observations for venture-backed companies are discarded since their propensity scores were lower than the minimum value of the propensity scores associated to non venture-backed companies.

The stratification estimator is then nothing more than a weighted average of the difference in means for the variable Y, across the discrete bins produced by the propensity score estimation:

1 1 1( )b b

VC VC NVCb i iVC NVCb

i VC i NVCb b

N Y YN N N

τ∈ ∈

= −

∑ ∑ ∑

where VCb and NVCb are the sets of VC backed and non VC backed observations in each bin and , the corresponding number of observations. The relative variance can also be computed as:

VCbN NVC

bN

Venture capital is considered to be the most appropriate form of financing for innovative firms in high-tech sectors. Venture capital has greatly developed over the last three decades in the United States, but much less so in Europe, where policy makers are striving to help channel more funds into this form of financial intermediation. We provide the first assessment of venture capital in Europe. We document its development in the 1990s, also providing a conceptual framework for this analysis. Comparing the evolution and structure of European and American venture capital, we find the wedge between these two industries to be large and growing. We then look at the involvement of venture capital with some of Europe’s most innovative and successful companies: Those listed on Europe’s ’new’ stock markets. Venture capital is effective in helping these firms overcome credit constraints, and thus to be born in the first place. Using a unique, hand collected data set from the listing prospectuses and annual reports of these companies, we then find European venture capital to have a limited effect on their ability to raise funds, grow, and create jobs. We conclude that public support of the European venture capital industry should look at both its growth and at its maturation.

4.2. The ’difference in differences’ method

Another popular method of evaluation is called ’difference in differences’ (DID), and is helpful in addressing possible evaluation biases due to the effects of unobservable variables that could be driving the difference in behavior of the two groups we are trying to compare, venture-backed and non venture-backed companies. The name of the DID estimator comes indeed from the fact that it compares the difference in the average behaviour before and after the IPO for the eligible group (venture-backed companies) with the behaviour before and after the IPO of the control group (non venture-backed companies). Notice that we compare behaviour around the IPO since we need to pin down the effect of venture capital (the ’cure’) on how treated and untreated companies react to a common external shock (the IPO).

We can apply the DID estimator only in the case of employment and sales growth, since time-to-listing and amount raised only occur at IPO. We indicate the estimator with θDID, which measures the growth of venture-backed in excess to that of non venture-backed companies. Abstracting from any regressor besides venture capital indicator, we write:

where YVC and YNVC are the mean values of the variables for the venture-backed and the non venture-backed companies, respectively, and t0 and t1 represent the pre- and post-IPO periods.

Under the assumptions of common time effects across groups and of no composition changes within each group, θDID measures the average effect of the IPO by removing the unobservable individual effects and common macro effects. In fact, by differentiating the mean value of Y inside the brackets we eliminate the common individual effect without affecting the common time effect, which can be eliminated by differentiating the two brackets.

Table A-11 shows the results of the DID estimator for the post-IPO growth in employment and sales for the two groups of companies. Our estimated effects confirm the sign of the coefficient of our previous estimates, as well as that of the stratification estimator, although both their value and their significance is now higher. There are two possible weaknesses of the DID estimator. One is due to the lack of control for unobservable (temporary) individual specific components that might influence the behaviour of the two groups. The DID estimator might then over-estimate the effect of the cure. This is a possible an explanation of the higher values in A-11 than in Table A-10. A second weakness of the DID estimator is that the assumption of common macro effect across companies. If the two groups have some characteristics that distinguish them and make them react differently to the common shock, we may get inconsistent estimates. Table A-11. Difference in differences estimates: Average effect of venture capital

1 0 1 0( ) (VC VC NVC NVC

DID t t t tY Y Y Yθ = − − − )

θ t-statistic Employment growth -106.38 * -1.84 Sales growth -127.18 -1.08 Notes: Significance levels are indicated by * (10%), ** (5%), and *** (1%).

5. DATA APPENDIX

We collect our information from all available issuing prospectuses and annual Reports or Euro.nm listed companies. Data are codified for all available years prior to the IPO (from prospectus information, which typically contains information for the three years before listing), as well as for all available years after the IPO (from subsequent annual reports).

Some companies choose to end their fiscal years in a month different from December. In these cases if the company’s fiscal year ends between January and June we consider it as ending the previous December, otherwise we consider it ending the following December. For the year 2000 in the few cases where the annual report was not yet issued at the time of writing, we rely on end-year official announcements of results.

Since 1999 most information is reported in euros. For earlier years we use monthly averages of the exchange rate between the ECU and national currencies to obtain an euro-equivalent. For subsequent years we use euro conversion rates.

5.1. Financial and business variables

We codify several financial variables: assets, debt, equity, and EBITDA (earnings before interest, taxes, depreciation and amortization), from which we compute leverage (debt divided by debt plus equity) and ROA (return over assets, computed as EBITDA over assets). EBITDA is a common measure of a firm’s profitability that does not depend on its financial policy and tax regime.

We computed the amount of capital raised at IPO as the issue price times the number of shares sold at IPO, except those sold by existing shareholders, but including the amount of greenshoe actually used. The amount of shares issued under the greenshoe over-allotment option is taken from the web site of the stock exchange for the German and Italian markets, and for the French market until the end of 1999. For French IPOs in 2000 we contacted directly issuing companies.

We take from the prospectus and annual reports information about a company’s business and strategy. We codify the following variables: Sales (total revenue from sales of goods and services), Employees (at year end), Capex (capital expenditure), Foreign sales share (share of foreign sales over total sales), Intangible assets (the stock of goodwill, patents, software and advertising), and R&D (current expenditure in research and development. From these data we can compute R&D intensity (R&D over Sales). Patents would be a natural measure of a firm’s innovation strategy. Unfortunately, the nature of the data makes it difficult to use them sensibly since patent applications take about 18 months to be released by the European Patent Office, so we are only starting to get patent data for the post-IPO period.

5.2. Sectoral attributions

Each company in the data set is assigned to a sector. The procedure we use is based on the sectoral attribution of Datastream, which are derived from the classification of the Financial Times. We use the following seven sectors: BIOMED, FINSER, ITSIS (comprising Internet, IT services, and software), MEDIA \& ENTERNAINEMENT, TECHNOLOGY, TELECOMMUNICATIONS, and TRADITIONAL (products and services).

We also employ an alternative sectors classification, following the sectors attributions introduced in May 2000 by the Neuer Markt, and we obtain very similar results as the one we report in the main text. The Neuer Markt assigns each company to the sector that generates the largest share of its earnings. For companies listed in markets other than the Neuer Markt we attribute sectors by looking at the business description contained in the issuing prospectus, and we augment the Neuer Markt classification with two further

sectors: Manufacturing (other than high-tech products) and (traditional) Services. The nine resulting sectors are: Biomed (which includes biotechnology—pharmaceutical products and services based on applications of genetics—and medtech-medical goods and health care services), Financial services (banking, insurance and brokerage, usually provided through the internet), Industrial Services (innovative services for industrial firms), ITSIS (infrastructure for the internet, internet services, IT services, internet, software), Manufacturing (goods other than high-tech), Media and entertainment, Services (marketing services, wholesale and retail distribution, business consulting, logistic services), Technology (high-tech products and services), and Telecom.

5.3. Venture capital data

In order to identify venture capital financing we could not rely on a simple procedure. A venture capital firm is not a bank, which can be readily identified as such. Since no license or professional registry exists, one must resort to a number of ’identifying conditions.’ For each company we proceeded to identify from the listing prospectus the venture capitalists which provided financing, starting from a pool of more than 300 ’candidates,’ i.e. financiers which were not founders, nor individuals, friends and families, or strategic partners (i.e. other companies). For each company we used the information about its ownership structure at the time of the IPO, which lists all its shareholders and details their holdings before and after the IPO. This helped us identify which venture capitalists have been involved with the company, and the extent of their shareholdings. A potential limitation of relying on data at IPO is that a venture capital might have already exited the company. This turns out to be a rare event, since the venture capital can profit much more by remaining until IPO. Moreover, the US experience shows that when a venture capital sells before IPO, it usually sells to another venture capitalist, in which case the presence of a venture capitalist at IPO would still show up. Direct inspection of the history section of all prospectuses revealed that a venture capitalist exited before IPO only in a couple of cases.

The process of identification of venture capitalist consisted of several steps. First, We identified venture capitalists by using the directories of the European Venture Capital Association and the directories of the British, Belgian, Dutch, French, German, Italian, Israeli, and US national venture capital associations. The venture capitalists members of one of these associations form our ‘tier 1,’ or core, group of venture capitalists, and they total 123. We then proceed to using other sources in order to identify venture capitalists which are not members of an association. Using online directories of venture capitalist (mainly at regional level), web sites, IPO prospectuses, and press sources we check if each ’candidate’ is defined as an ’actual’ venture capitalist. These ’self declared’ venture capitalists form our ’tier 2’ group, which comprises 75 more venture capitalists. Overall, our sample reveals that 198 venture capitalists were involved with companies listed on Euro.nm. Ten of these were national branches of international venture capital groups, like 3i group, the British private equity and venture capital firm. We report the results we

obtain using the broader definition of venture capitalist (i.e. those falling in either tiers), as our results do not change if we restrict the definition to the core.

For each venture capitalist we measure from the prospectus the share of equity held at IPO and the amount that is sold at that time. Finally, we then looked at information about the date of entry of venture capitalists, i.e. the date at which they first contributed funds to the company. If more venture capitalist are involved with a company, we use as date of entry that of the first of them which got involved with the company. The entry date could often be identified from the company’s own history description. In all the other cases we directly contacted the company. Overall, we could assign the entry date in 90% of the cases.

5.4. Price data

We take our price data from Datastream. We collect price data for all the companies in our data set, excluding the sixteen ones in the financial services sector. We use daily closing prices, which are corrected for stock splits and changes in nominal value of the underlying stocks.

5.5. Definitions of the variables

The following is a list of definitions for all the variables we use. AGEIPO is the age of a company at the date of its IPO. To determine a company’s

date of birth we employ the earliest evidence of business activity in the listing prospectus, which need not coincide with the date of incorporation. In fact, several companies were born as partnerships or limited companies before incorporating.

TIME-TO-LISTING (TTL) is the time elapsed from a company’s foundation to its IPO. For companies born before the opening of the Euro.nm market in which they list, TTL is the time elapsed from the opening of the Euro.nm market and their IPO. Two companies which went public in the same month but which are listed in two different markets may then have two different TTL, since different markets opened at different times.

VC is a dummy variable that takes the value 1 if a company has received venture capital financing, and 0 otherwise.

AMOUNT measures the capital raised at IPO. It equals the issue price times the number of shares sold at IPO (except those sold by existing shareholders, but including the greenshoe).

BIOMED, FINSER, ITSIS (comprising Internet, IT services, and software), MEDIA & ENTERNAINEMENT, TECHNOLOGY, TELECOMMUNICATIONS, and TRADITIONAL (products and services), are dummy variables which take the value 1 if the company operates in that industry and 0 otherwise.

We use the following financial variables: ASSETS is current total asset.

DEBT is the sum of commercial and financial debt. EQUITY is total shareholders’ equity. EBITDA is earning before interest, taxes, depreciation and amortization. DEBT-TO-EQUITY is DEBT divided by EQUITY. LEVERAGE is DEBT divided by DEBT plus EQUITY. ROA is EBITDA over ASSETS.

We use the following variables which reflect a company’s business situation and strategic choices: SALES is total revenue from sales of goods and services. EMPLOYEES is the total number of employees at year end. CAPEX is capital expenditure, i.e. investment in tangible and intangible fixed assets. FOREIGN SALES SHARE is the share of foreign sales over total sales. INTANGIBLE ASSETS equals the capitalized amount of goodwill, patents, software and advertising. R&D is current expenditure in research and development. R&D INTENSITY is R&D over ASSETS.

6. COMPLETE REFERENCES

To facilitate further research by specialist readers, we here provide the full list of references from the main text and the Web Appendix.

Audretsch, David, and Roy Thurik (eds.)[1999], Innovation, industry evolution, and employment,

Cambridge, Cambridge University Press. Admati, Anan, and Paul Pfleiderer [1994], ’Robust Financial Contracting and the Role of Venture

Capitalists,’ Journal of Finance, 49 (2) 371-402. Bank of England [2000], ’Finance for Small Firms: A Seventh Report,’ Bank of England (http:\\

www.bankofengland.co.uk) Barry, Christopher, Chris Muscarella, John Peavy, and Michael Vetsuypens [1990], ’The Role of

Venture Capitalists in the Creation of Public Companies,’ Journal of Financial Economics, 27 (2) 447-71.

Becker, Rolf, and Thomas Hellmann [2000], ’The Genesis of Venture Capital: Lessons from the German Experience,’ mimeo, (http:\\faculty-gsb.stanford.edu/hellmann).

Benjamin, George, and Ellen Sandles [1998], ’Angel Investors: Culling the Waters for Private Equity,’ Journal of Private Equity, 1 (3), 41-59.

Bergemann, Dirk, and Hulrich Hege [1998], ’Venture Capital Financing, Moral Hazard and Learning,’ Journal of Banking and Finance, 22 (2), 703-735.

Berglöf, Erik [1994], ’A Control Theory of Venture Capital,’ Journal ofLaw, Economics and Organizations, 10 (1) 247-67.

Black, Bernard, and Ronald Gilson [1998], ’Venture Capital and the Structure of Financial Markets: Banks versus Stock Markets,’ Journal of Financial Economics, 47 (1) 243-77.

Blass, Asher, and Yishay Yafeh [2000], ’Vagabond Shoes Longing to Stray: Why Foreign Companies List in the United States,’ Journal of Banking and Finance, 25 (3) 555-72.

Blundell, Steve, and Monica Costas Dias [2000] ’Evaluation Methods for Non-experimental Data,’ mimeo, University College London, (http:\\www.ucl.ac.uk).

Bottazzi, Laura, and Marco Da Rin [2001], ’Europe’s ’New’ (Stock) Markets: Euro.nm, 1997-2000,’ IGIER Working Paper n.171 (http:\\www.igier.uni-bocconi.it).

Brav, Alon, and Paul Gompers [1997], ’Myth or Reality? The Long-run Underperformance of Initial Public Offering: Evidence from Venture and Nonventure-backed Companies,’ Journal of Finance, 52 (4) 1791-1820.

BVCA - British Venture Capital Association [1999], ‘The Economic Impact of Venture Capital in the UK,’ (http:\\www.bvca.co.uk).

Casamatta, Catherine [2000], ’Financing and Advising: Optimal Financial Contracts with Venture Capitalists,’ mimeo, Universitè de Toulouse.

Committee of Wise Men [2001], ’Final Report of the Committee of Wise Men on the Regulation of European Securities Markets,’ Brussels. (http:\\europa.eu.int/comm/internal market/en).

Cornelli, Francesca, and Oved Yosha [1998], ’Stage Financing and the Role of Convertible Debt,’ mimeo, London Business School (http:\\www.lbs.ac.uk}.

Davis, Steven, John C. Haltiwanger, and Scott Schuh [1996], Job Creation and Destruction, Cambridge, MIT Press.

Dehejia, Robert, and Steven Wahba [1999], ’Causal Effects in Non-Experimental Studies: Re-evaluating the Evaluation of Training Programs,’ Journal of the American Statistical Association, 94 (3) 1053-62.

Euro.nm [1999], ’The Successful Alliance of Growth Markets in Europe.’ European Commission [1994], Green Paper on Innovation, Brussels.

(http:\\europa.eu.int/en/record/green/gp9512). European Commission [1998], Risk Capital: A Key to Job Creation in the European Union,

Brussels. (http:\\europa.eu.int/comm/off/com/riskkal). ECVA - European Venture Capital Association [1996], ’The Economic Impact of Venture Capital

in Europe,’ a study conducted by Coopers & Lybrand for EVCA. (http:\\www.bvca.org). EVCA - European Venture Capital Association [1998], ’Priorities for Private Equity: Realising

Europe’s Entrepreneurial Potential,’ (http:\\www.bvca.org). EVCA - European Venture Capital Association [2001], ’2000 Investment Benchmarks Report,’

(http:\\www.bvca.org). Freeman, John [1998] ’Venture Capital and Growth Business in Germany,’ mimeo, University of

California at Berkeley. Gale, Douglas, and Martin Hellwig [1985], ‘Incentive-Compatible Debt Contracts: The One-

Period Problem,’ Review of Economic Studies, 52 (3), 647-63. Gompers, Paul [1994], ’The Rise and Fall of Venture Capital,’ Business and Economic History, 23

(2) 1-26. Gompers, Paul [1995], ’Optimal Investment, Monitoring, and the Staging of Venture Capital,’

Journal of Finance, 50 (4) 1461-90. Gompers, Paul, and Josh Lerner [1997] ’Risk and Reward in Private Equity Investments: The

Challenge of Performance Assessment,’ Journal of Private Equity 1 (2), 5-12. Gompers, Paul, and Josh Lerner [1998] ’What Drives Venture Capital Foundraising?,’ Brookings

Papers on Economic Activity Microeconomics, 149-92. Gompers, Paul, and Josh Lerner [2000] ’Can Corporate Venture Capital Succeed?: Organizational

Structure, Complementarities, and Success,’ in Randall Morck (ed.) Concentrated Ownership, Chicago, University of Chicago Press for the National Bureau of Economic Research, 17-50.

Gorman, Michael, and William Sahlman [1989], ’What do Venture Capitalists do?’ Journal of Business Venturing, 4 (1), 231-48.

Hart Oliver, and John Moore [1998], ’Default and Renegotiation: A Dynamic Model of Debt,’ Quarterly Journal of Economics, 113 (1), 1-41.

Hellmann, Thomas [1998], ’Venture Capital and the Allocation of Control Rights,’ Rand Journal, 29 (1) 57-76.

Hellmann, Thomas [2000], ’Venture Capitalists: The Coaches of Silicon Valley,’ in Chong-Moon Lee, William F. Miller, Marguerite Gong Hancock, and Henry S. Rowen (eds.) The Silicon Valley Edge: A Habitat for Innovation and Entrepreneurship, Stanford, Stanford University Press.

Hellmann, Thomas [2002], ’A Theory of Corporate Strategic Investment,’ forthcoming, Journal of Financial Economics.

Hellmann, Thomas, Laura Lindsey, and Manju Puri [1999] ’Banks in the Venture Capital Market: An Empirical Analysis,’ mimeo, Stanford University GSB.

Hellmann, Thomas, and Manju Puri [2000], ’The Interaction between Product Market and Financing Strategy: The Role of Venture Capital,’ Review of Financial Studies, 13 (4) 959-84.

Hellmann, Thomas, and Manju Puri [2002], ’Venture Capital and the Professionalization of Start-ups: Empirical Evidence,’ forthcoming, Journal of Finance.

Jeng, Leslie, and Philippe Wells [2000], ’The Determinants of Venture Capital Funding: Evidence Across Countries,’ Journal of Corporate Finance, 6 (1),241-89.

Kaplan, Steven, and Per Stromberg [2000] ’Financial Contracting Theory Meets the Real World: An Empirical Analysis of Venture Capital Contracts,’ CEPR Discussion Paper 2421. (http:\\www.cepr.org)

Kiefer, Nicholas [1988], ’Economic Duration Data and Hazard Functions,’ Journal of Economic Literature, 26 (2), 646-79.

Kortum, Samuel, and Josh Lerner [2000], ’Assessing the Contribution of Venture Capital to Innovation, Rand Journal, 31 (4), 674-692.

Lerner, Josh [1994a], ’Venture Capitalists and the Decision to go Public,’ Journal of Financial Economics, 35 (1) 293-316.

Lerner, Josh [1994b], ’The Syndication of Venture Capital Investments,’ Financial Management, 23 (3), 16-27.

Lerner, Josh [1995], ’Venture Capitalists and the Oversight of Private Firms,’ Journal of Finance, 50 (1) 301-18.

Meggison, William, and Kathleen Weiss [1991], ’Venture Capital Certification in Initial Public Offerings,’ Journal of Finance, 46 (3) 879-903.

Michelacci, Claudio, and Javier Suarez [2000], ’Business Creation and the Stock Market,’ working paper, CEMFI. (http:\\www.cemfi.es).

NVCA - National Venture Capital Association [1998] ’Eighth Annual Economic Impact of Venture Capital Study,’ a study conducted by Coopers & Lybrand and Venture One, NVCA, Arlington.

OECD [1998], Fostering Entrepreneurship, Paris, OECD. Pagano, Marco [1993], ’The Flotation of Companies in the Stock Market: A Coordination Failure

Model,’ European Economic Review, 37 (3) 1001-1125. Pagano, Marco, Ailsa Roell, and Josef Zechner [1999], ’The Geography of Equity Listing: Why

Do European Companies List Abroad?’ CSEF Working Paper n.28. Prowse, Stephen [1998] ’Angel Investors and the Role of Angel Investments,’ Journal of Banking

and Finance, 22 (6) 785-92. Repullo, Rafael, and Javier Suarez [2000], ’Venture Capital Finance: A Security Design

Approach,’ CEPR Discussion Paper 2097. (http:\\www.cepr.org). Sahlman, William [1990], ’The Structure and Governance of Venture-capital Organizations,’

Journal of Financial Economics, 27 (2) 273-521. Zahra, Shaker [1996], ’Techonology Strategy and New Venture Performance: A Study of

Corporate-Sponsored and Independent Biotechnology Ventures,’ Journal of Business Venturing, 11 (2), 289-321.