On the Efficiency of Bank-Affiliated Venture Capital

Transcription

1 On the Efficiency of Bank-Affiliated Venture Capital Douglas J. Cumming Schulich School of Business York University 4700 Keele Street, Toronto (Canada) Phone: Samuele Murtinu Institute of Economic Policy Catholic University of Milan Largo A. Gemelli 1, 20123, Milan (Italy) Phone: This draft: July 13, 2016

2 On the Efficiency of Bank-Affiliated Venture Capital Abstract This paper examines the effect of bank-affiliated venture capital (BVC) on portfolio companies across seven Western European countries from 1991 to Regardless of controls for selection and matching between BVCs and portfolio firms, the data indicate that BVC deals are more likely in larger VC syndicates, and in countries with lower earnings aggressiveness, higher market capitalization, and for local deals and international deals among countries with similar legal systems. The data indicate that BVCs have a positive impact on portfolio firm sales and efficiency (sales/total assets), but do not have an effect on debt to total assets, return on assets, and have a negative effect on net cash flows to total assets. Further, the data indicate BVCs have a positive impact on the probability of portfolio firms achieving an IPO and acquisition. However, BVCs also strongly increase the liquidation risk of their portfolio companies. JEL codes: G21, G24, G32, G33, G34, L26 Keywords: bank, venture capital, syndication, Europe, matching, performance 1

3 1. Introduction In the last fifteen years the most important commercial banks diversified their core business and extended its scope toward the venture capital (VC) world (Andrieu 2013; Andrieu and Groh 2013; Fang, Ivashina, and Lerner 2013; Tykvová 2016). Nowadays one of the most important sources of equity capital for high-tech entrepreneurial ventures especially in Europe is in fact represented by banks. 1 Despite the relevance of this phenomenon, very little is known about the financial and strategic objectives of bank-affiliated VC funds (BVCs), their syndication strategies and patterns, and the impact of BVCs on the performance of high-tech entrepreneurial ventures. This paper aims at filling important gaps in the extant finance literature by investigating the role of banks in VC syndicates, and how their impact creates value for portfolio firms using a brand-new unique European Commission-sponsored longitudinal dataset: the VICO dataset. BVCs always syndicate with limited partnership independent venture capitalists (IVCs) in early stages, and as such, we compare the effect of BVCs in a syndicate relative to non-vc backed firms. We use time-varying cross-country exogenous shocks related to financial reforms and bank crises to test the reliability of our estimates to omitted variables. The VICO dataset includes data on young independent medium- and high-tech VC-backed firms and comparable non-vc-backed potential investees that are headquartered in seven European countries (Belgium, Finland, France, Germany, Italy, Spain and the UK). Our sample of VC-backed and matched non-vc-backed firms fairly represents the population of entrepreneurial ventures in the abovementioned seven Western countries, and is more comprehensive than other VC datasets such as Thompson, as affirmed by a report of the European Parliament (2012). We answer several key questions related to BVC. First, which are the main determinants at country-, industry-, portfolio firm-, and syndicate-level explaining the choice of BVCs to join syndicates? In particular, do banks enter the VC industry in countries with better accounting disclosure of private firms, debt enforcement, creditor rights, discharge from pre-bankruptcy indebtedness, and where financial markets are more developed? Do BVCs target larger, older and more innovative firms? Are syndicates involving BVCs larger? Do the geographical and institutional distances within the syndicate and between the syndicate members and portfolio firms play a role? Do the experience of syndicate members and their governance type matter? Second, which is the impact of syndicates involving BVCs on portfolio firm performance? In particular, we analyze on 1 In 2013 banks account for the 18.9% of the total fundraising in Southern Europe, while such percentage ranges between 3.6% and 5.3% in other EU countries (with the exception of the UK). The contribution of banks to the total VC fundraising in Europe was about 14% in 2007, while it decreased below 10% since 2010 (source: the European Venture Capital Association - EVCA). 2

4 which performance variables the potential impact of syndicates involving BVCs takes place: sales value, debt on total assets ratio (DTA), net cash flow on total assets ratio (NCFTA), return on total assets (ROA), and asset utilization ratio. These variables capture the overall potential effect of BVC: output effects (sales value), effects on financial structure, investments/liquidity, operating performance and efficiency. Further, we test whether such potential impact comes from a better screening of portfolio companies 2 and/or a monitoring effect by means of value addition in the postinvestment period. To this latter, does the monitoring effect occur in the early stages after VC funding or in the late stages? Third, we investigate whether the presence of BVCs influences the investment outcome. Do BVCs impact the likelihood that portfolio companies exit through IPOs, acquisitions or liquidations? The results of our empirical investigation may be summarized as follows. First, BVCs usually prefer larger syndicates led by an IVC fund that target portfolio companies in countries with relatively higher accounting disclosure of private firms, and with relatively more developed financial markets. We also find a preference for local deals or at least for countries with similar legal systems. These results may be driven by a non-random double unobserved selection: i) portfolio companies that expect high horizontal agency costs with traditional VCs are more likely to match with VC syndicates involving a passive investor such as a BVC fund that dilutes the stake of the leading active investor; and ii) VC syndicates whose portfolio is too risky are more likely to match with passive investors such as BVCs who allow syndicates to share the portfolio risk and at the same time not to bear agency costs with the leading investors. We used three different methodologies. First, following Sørensen (2007) we use a probit model on both realized and potential dyads between VCs and portfolio companies, and estimate the likelihood that a realized dyad includes a BVC. Second, we follow Bottazzi, Da Rin, and Hellmann (2008) and estimate two-step and maximum likelihood Heckman procedure on all potential dyads by using country pairs VCs country and portfolio companies country as exogenous identification restriction because BVCs are more likely to invest locally than traditional VCs, so exploiting their ability to collect soft information (Coval and Moskowitz 2001; Mayer, Schoors, and Yafeh 2005). Third, we resorted to the methodology proposed by Ackerberg and Botticini (2002) and used in the VC context by Bottazzi, Da Rin, and Hellmann (2008). The basic idea is that the matching between the company i and the investor j depends on both the local availability of VCs and the geographic distribution of portfolio companies being this latter exogenous and thus each geographic market has a different matching process. Following Bottazzi, Da Rin, and Hellmann (2008), we build 147 interaction terms (7 countries*7 industries*3 investment 2 "Screening" is slightly different from "sorting" (Sørensen 2007). Screening refers to the investors ability to select better companies when there is information asymmetry between investors and potentially investable companies. Sorting refers to a full information market where better-quality investors are matched with better-quality companies. 3

5 stages), and use such interaction terms as instruments for all VC variables included in our linear probability model. Results of these three methodologies confirm the above results. Second, syndicates involving BVCs have a large positive impact on portfolio companies' sales value: the estimated average magnitude in the post-investment period is +47%. When investigating the dynamics of such sales performance improvements, we find a short-term impact that ranges from +34% to +43% depending on the selected short term window and a long-term impact ranging between +56% and +63%. In terms of DTA, the impact is not statistically significant. The effect on firm liquidity as proxied by NCFTA is negative and with a greater magnitude in the short-term (between -21% and -23%) than in the long-term. Interestingly, even though the magnitude of the coefficient is very high (ranging from +95% to +109%) there is no statistically significant impact on firm operating performance, neither in the short nor in the long run. In terms of firm efficiency as measured by means of the asset utilization ratio the impact of BVC is positive and statistically significant, and it ranges from +6% in the short-term to +9% in the long-term. Overall, the above results indicate that VC syndicates involving BVCs improve firm efficiency through an improvement of product market performance (sales value). This latter may be attributed to enhanced investments, because of lower cash constraints (i.e. negative effect on NCFTA). We contribute to the VC literature by showing a selection effect of BVCs in terms of portfolio companies liquidity. More specifically, the liquidity of portfolio companies (as measured by NCFTA) prior to receiving VC funding is 25% lower than that of non-vc-backed firms. This finding may be attributed to the strategic objectives of BVCs (Hellmann, Lindsey, and Puri 2008), i.e. enhance demand of debt capital from portfolio companies that need liquidity to invest and grow at a stable pace and that would borrow additional debt in the future from the parent bank of the BVC fund. Even though we follow Chemmanur, Krishnan, and Nandy (2011) in estimating fixed effects regressions with leads to control for selection bias, we further test the reliability of our results by means of a procedure aimed at controlling for potential time-varying unknown selection bias. The procedure suggested by Chetty, Friedman, and Rockoff (2014) tests the correlation between VC variables and variables omitted from our models, which are potentially correlated with our dependent variables. As to the omitted variables, we extract the reform measures implemented in the seven countries included in our sample in the areas Start-up financing, Seed financing, Access to finance in general and Access to finance for start-ups in general from the MICREF database, developed by the Directorate General for Economic and Financial Affairs (DG ECFIN) of the European Commission. Further, we use the data on systemic banking crises provided by Laeven and Valencia (2013). These two variables financial reforms and bank crises should be strong predictors of our dependent variables (Bekaert, Harvey, and Lundblad 2005; Chava and Purnanandam 4

6 2011; Dell Ariccia, Detragiache, and Rajan 2008; Hackbarth, Haselmann, and Schoenherr 2015; Kerr and Nanda 2009; Rice and Strahan 2010). Results of this alternative procedure are fully in line with our baseline results. We also design an identification strategy to estimate the net effect that BVCs add to syndicates led by IVCs. When compared with a matched sample of IVC backed firms, our findings show that syndicates involving BVCs have a large positive impact on portfolio companies' sales value: the estimated average magnitude in the post-investment period is +38%, and the estimated BVCs net effect is +6%. In terms of DTA, firm liquidity, ROA and firm efficiency there is no statistically significant impact. Overall, the above results indicate that VC syndicates involving BVCs improve firm efficiency through an improvement of product market performance (sales value). However, the real value addition of BVCs is only related to sales value. Our results do not fully confirm the passive nature of BVCs: in fact we find that BVCs shape syndication strategies, and add value to portfolio companies commercialization performance. Third, as to the investment outcome ideally one would like to have data on fund returns. In the absence of publicly available data on fund returns, we follow the extant VC literature (e.g. Gompers, Kovner, Lerner, and Scharfstein 2008; Hochberg, Ljungqvist, and Lu 2007; Sørensen 2007) and use positive (IPOs, acquisitions) and negative exits (liquidations) as proxies of fund performance. As highlighted by Gompers and Lerner (2001) exit is the most important performance metric for VCs, and VCs key decisions are based on expectations about timely exits and associated capital gains (Black and Gilson 1998; Cumming 2008; Giot and Schwienbacher 2007). We use multinomial logit regressions and find that syndicates involving BVCs have a large positive impact on the likelihood of positive exits BVC-backed companies have a higher likelihood to go public (+132%) and to be acquired (+138%) than non-vc-backed companies respectively while there is no statistically significant impact on firm liquidation. These results hold even after controlling for other factors that influence exit. We still use the above identification strategy to estimate the net effect that BVCs add to syndicates led by IVCs. When compared with a matched sample of IVC backed firms, we find that syndicates involving BVCs have a large positive impact on the likelihood of positive exits (+124%), and the estimated net effect of BVCs is non-negligible (+32%). As to firm liquidation, our estimates show that syndicates involving BVCs strongly increase the liquidation risk of portfolio companies: the overall effect is +271%, whose +168% is the estimated net effect of BVCs. The rest of this article is organized as follows. Section 2 reviews the literature. Section 3 describes the data. Section 4 provides an analysis of the determinants of BVCs in VC syndicates. Sections 5 and 6 present analyses of the impact of BVCs on portfolio firm performance and exit, respectively. Section 7 shows results of BVC backing after pre-estimation matched sample with 5

7 IVC-backed companies. Section 8 provides robustness checks. Concluding remarks follow in Section Literature review VC has been traditionally considered as the most suitable and effective financial mode to bridge the funding gap suffered by high-tech entrepreneurial ventures (Sahlman 1990; Gompers and Lerner 2001). Different from traditional financial investors (e.g. commercial banks), VCs not only provide money but also value-adding services, such as strategic planning, marketing, finance and budgeting, and human resource management, in which high-tech entrepreneurial ventures typically lack internal capabilities (Casamatta 2003; Schmidt 2003; Ueda 2004). The recent VC literature highlighted that there is considerable heterogeneity among VCs (Chemmanur and Chen 2014) in terms of experience, industry, stage and geographic specialization, and type of governance. The extant literature has mainly focused on "Silicon Valley-style" IVCs (Puri and Zarutskie 2012), VCs owned and managed by big corporations (Colombo and Murtinu 2016; Gompers and Lerner 2000), and VCs sponsored, owned and/or managed by government bodies (Cumming, Grilli, and Murtinu 2014). However, very little is known about BVCs. Some papers (Ueda 2004; Winton and Yerramilli 2008) investigated the choice of venture managers between debt capital provided by banks and VC financing. Ueda (2004) showed that the lower is the level of collateral, the higher is the growth potential (and so the venture's business risk), and the tighter is the protection of intellectual property rights, and the higher is the likelihood that the focal entrepreneur will choose VC financing. Winton and Yerramilli (2008) model the entrepreneurial choice between bank capital and VC as influenced by the risk and return of portfolio companies' cash flows. To the best of our knowledge, Andrieu and Groh (2012) is the unique study that focused on the entrepreneurial strategy to choose between IVCs and BVCs. According to their theoretical model, BVCs are less (more) likely to divest early (provide more rounds of equity capital). Other studies (e.g. Bottazzi, Da Rin, and Hellmann 2008) show that BVCs are less capable in monitoring activities, diversify more their investments (i.e. lower volatility), and are more likely to hold larger funds than IVCs do Banks in VC syndicates The very few studies that deal with BVC do not take into account the characteristics of VC syndicates involving BVCs. Under the assumption that the strategic objective of BVCs is to enhance demand of debt capital from their portfolio companies (Hellmann, Lindsey, and Puri 2008), BVCs need that their portfolio companies show a low default risk and a stable operating performance in 6

8 order to borrow debt from the parent bank of the fund. Even though some target companies show a too high risk for BVC investments, such companies would show a stable operating performance in the future and thus a high likelihood to demand debt if they receive a proper "treatment" today, i.e. monitoring and value-adding services. Given the passive nature of BVCs (Bottazzi, Da Rin, and Hellmann 2008), BVCs need an agent who is willing to actively monitor and coach the portfolio company and thus improves its operating performance. Alternatively, this agent must show better screening abilities than the bank (Lerner 1994a; Sorenson and Stuart 2001) in targeting entrepreneurial ventures with the most suitable characteristics as to ensure a stable operating performance in the future. A potential candidate is a specialized IVC investor. As highlighted by Chemmanur, Krishnan, and Nandy (2011), IVC-backed companies show a higher operating performance than non-vc-backed companies because of both an effective screening ability of IVCs (Casamatta and Haritchabalet 2007; Gompers and Lerner 2004) and value-adding services provided by IVCs. 3 Hence, through the syndication with IVCs, BVCs can increase the demand of bank debt by portfolio companies, and share the overall risk of their investment portfolio with IVCs. 4 The dilution of equity investment is not a problem for BVCs because there is not a competition between IVCs and BVCs in serving portfolio companies as future lenders. 5 Thus, the typical trade-off decision to syndicate based on benefits and costs to relinquish part of the firm value to other investors (Das, Jo, and Kim 2011; Espenlaub, Khurshed, and Mohamed 2015) is not the main issue for BVCs. However, when syndicating with traditional VCs, banks may face transaction costs and principalprincipal conflicts (Chahine, Arthurs, Filatotchev, and Hoskisson 2012). Such conflicts may increase the risk of BVC-IVC syndicated investments, and BVCs could not bear such risk. Thus, the structure and the composition of the syndicate is a key strategic decision for BVCs. We expect that the number of VCs included in the syndicate, their experience, and their specialization may increase the incentives of a BVC to join the syndicate. Gompers, Mukharlyamov, and Xuan (2016) find that similar VCs in terms of ethnic, educational, or experience background are more likely to syndicate together, but this syndicate homophily decreases the likelihood of investment success especially in early stage deals. Chemmanur, Hull, and Krishnan (2011) and Cumming, Knill and Syvrud (2015) show that syndicates composed of local and international VCs are more successful than pure international and pure local ones. In fact, international VCs show disadvantages when investing 3 It is worth noting that evidence about the screening ability of IVCs is not confirmed in Europe. Colombo and Murtinu (2016) and Croce, Martí, and Murtinu (2013) show that the operating performance of VC-backed companies is not statistically different from that of non-vc-backed companies before VC financing. Bottazzi, Da Rin, and Hellmann (2008) do not find selection effects driven by investor activism. Cumming, Grilli, and Murtinu (2014) show that their results on exit performance are not influenced by VCs screening activity. 4 Theoretical VC literature has investigated the relationship between the optimal size of VCs portfolio and the value added to portfolio companies (e.g. Fulghieri and Sevilir 2009). 5 For a theoretical model see Hellmann (2002). 7

9 abroad and/or in institutionally distant countries due to the lack of proximity, and local VCs as the bank-affiliated ones help lower information costs and foster the risk allocation strategy (Gompers and Lerner 2004) and the access to local knowledge, networks and business opportunities (Cumming, Knill and Syvrud, 2015). When joining syndicates with international experienced VCs, BVCs may overcome their liability of newness and smallness in the VC arena, and thus maximize capital gains (Cumming, Knill and Syvrud 2015). A syndication activity with prominent VCs may allow BVCs to invest in geographically distant companies (Sorenson and Stuart 2001), and access resources and competencies over and beyond those possessed by syndicate partners by leveraging the global network of their business links. Butler and Goktan (2013) show that also experienced VCs may have an advantage in syndicating with less experienced VCs as BVCs are. These latter are in fact specialized in collecting, producing and disclosing soft information about portfolio firms, which are usually opaque. Such soft information is much more important when portfolio companies do not show revenues, as in the case of smaller and younger companies. In so doing, geographic and institutional proximity with the portfolio company plays a role (Chen, Gompers, Kovner, and Lerner 2010). In fact, geographical and institutional distance between the VC fund and the venture leads to a more difficult and costly monitoring Impact of BVC on portfolio firm performance As highlighted by Bottazzi, Da Rin, and Hellmann (2008), the organizational structure of BVCs is very different from that of traditional VCs, and such structure strongly influences fund objectives and investment behavior, and thus the value provided to portfolio companies (Gompers, Kovner, and Lerner 2009). The main backlash is that IVCs are significantly more involved with their portfolio companies than BVCs. Bottazzi, Da Rin, and Hellmann (2008) show that IVCs help in recruiting outside CEOs, managers and directors, in developing stock option plans, and in the fundraising activity. Moreover, BVCs are believed to face a less high-powered incentive structure than that of IVCs. These latter usually have a fixed management fee (2/2.5% of the committed capital by limited partners) and an over-performance fee (around 20%) in the event of good fund performance, and hurdle rates and/or clawbacks in the event of poor fund performance (Gompers and Lerner 2004). Conversely, BVCs are reputed to have an incentive structure that is invariant across managers and funds, also due to equity concerns inside the bank. Further (and related to the latter point), BVCs might face problems related to manager retention: more skilled managers prefer to work in IVC funds. Finally, while IVCs are usually structured as limited partnerships (Chemmanur, Loutskina, and Tian 2014; Cumming, Grilli, and Murtinu 2014), BVCs are influenced by the pressure of the parent bank in their day-to-day operations. Thus, when syndicating with IVCs, BVCs benefit from the advantages of IVC limited partnerships, especially when the IVC investor is the syndicate 8

10 leader. In addition, BVC-IVC syndicates might leverage the different sources of networks and contacts provided by IVCs and banks, and exploit complementary resources, skills, and expertise (Andrieu and Groh 2012). As suggested by Du (2011), VCs usually prefer syndicate partners that are different from them. To this extent, the monitoring role of banks can serve to certify the quality of the investment target (Barry and Mihov 2015). However, it is worth noting that heterogeneous syndicates may underperform when compared to homogeneous syndicates (Cumming, Grilli, and Murtinu 2014). 3. Data In this work, we use a new European Commission-sponsored longitudinal dataset: the VICO dataset. Even though the dataset is quite new, it has been already used in the finance literature (e.g. Cumming, Grilli, and Murtinu 2014). The VICO dataset includes data on medium- and high-tech VCbacked and comparable non-vc-backed firms 6 that are headquartered in seven European countries (Belgium, Finland, France, Germany, Italy, Spain and the UK). Such firms are less than 20 years old, and they are not controlled by other firms at foundation. As to the industries included in the VICO dataset, they are reported in Table 1. The VICO dataset includes surviving and non-surviving ventures, i.e. liquidated and acquired firms, whether they are VC-backed firms or not. Even though survivorship bias does not exist in the VICO dataset, in unreported regressions we test whether our main results are driven by the fact that high-performing VC-backed firms are more likely to be acquired. 7 As to the VC-backed firms, information were collected through a random extraction from several country-specific proprietary and commercial databases, such as the yearbooks and reports provided by the national VC and private equity associations, websites and financial reports of the VCs, the ZEW Foundation Panel database, Private Equity Monitor, Venture Source, press releases and press clippings, IPO prospectuses, Thomson One, VCPro-Database, and Zephyr. As to the non- VC-backed firms, information were collected through a random extraction from Orbis and countryspecific proprietary databases. As to Orbis, all of the available vintage years were used to include the population of non-surviving firms. 6 Non-VC-backed firms are potential investees, because they were included in the dataset following the same criteria (country, age, independence, industry) used for the inclusion of the VC-backed firms. For more details, see Bertoni and Martí (2011: p. 5). 7 We applied the survivorship bias test proposed by Semykina and Wooldridge (2010), and we estimated our main models with the inclusion of a time-varying inverse Mills ratio (IMR)-type term. The IMR-type term is computed from a series of annual probit models on the likelihood that a firm exits the dataset because of acquisition or liquidation. The coefficient of the IMR-type term is not statistically significant in every regression, and thus there is no survivorship bias is our sample. 9

11 As common in the VC literature (e.g. Cumming, Grilli, and Murtinu 2014) the VICO dataset includes seed, early-stage, late-stage and expansion VC investments, while it does not include LBOs, real estate, distressed debt funds and other private equity investments. Moreover, VC-backed firms included in the dataset were less than 10 years old at the time of first VC investment. For a full description of the variables included in the VICO dataset, and a precise documentation of the data collection procedures and sources at portfolio firm-, investment-, and investor-level. In this work, the observed time period is The VICO dataset is representative of the European population of VC-backed firms (for more details see European Parliament 2012; Cumming, Grilli, and Murtinu 2014), and provides a better coverage of VC investments in Europe than the one provided by available commercial databases. The VICO dataset shows some advantages when compared to the most popular databases used by VC scholars (e.g. Thomson One). These latter usually under-represent VC investments made by nonindependent funds (Ivanov and Xie 2010: p. 135), especially in Europe (Cumming, Grilli, and Murtinu 2014). The VICO dataset includes firms that received their first VC round in the time period Even though this is a limitation of the dataset, it allows us to have a sufficient number of post-investment observations to assess the impact of VC on portfolio firms performance, consistent with Gompers and Lerner (2000), Hochberg, Ljungqvist, and Lu (2007), Nahata (2008), and Nahata, Hazarika, and Tandon (2014). As common in the VC literature (e.g. Ivanov and Xie 2010), we classify portfolio firms as BVC-backed if they are backed in the year of the first VC investment by a bank-affiliated VC fund (potentially syndicated by other VCs), independently from the type of VCs (independent, corporate, bank-affiliated or government-sponsored) that invested in subsequent VC rounds. Portfolio firms that received a first investment by VC investors with missing name, address and/or contact information are removed from the dataset. For the same criterion on the VICO dataset, see Colombo and Murtinu (2016) and Cumming, Grilli and Murtinu (2014). The VICO database includes relevant information on the syndicate size, the lead investor on the basis of the distribution of the amount invested and the equity stakes possessed by syndicated partners, geographic location in which VCs and portfolio firms are headquartered and distance using the average latitude and longitude (Coval and Moskowitz 1999), and VC experience. As to the BVCs, we also extract information about the ownership (governmentowned or private) of the parent banks (source: Bankscope). As to the portfolio companies, the VICO dataset provides information on company name, VAT code, address, foundation year, NACE industry classification, listed status (including year of IPO), status of the company (active, acquired, bankrupt), accounting variables, and patent data (source: PATSTAT). 10

12 As to the exit measures, we adopt the coding procedure of Cumming, Grilli and Murtinu (2014), Gompers, Kovner, and Lerner (2009), Gompers, Kovner, Lerner, and Scharfstein (2008), Hochberg, Ljungqvist, and Lu (2007), Nahata (2008) and Nahata, Hazarika, and Tandon (2014), and consider IPOs and acquisitions as successfully exited VC investments. 8 Similarly to Nahata, Hazarika, and Tandon (2014), we do not have data about deal values. Our sample of acquisition exits includes only trade-sales, and not buyouts of VC shares by other shareholders of the portfolio firm (e.g. founders), and/or by other VCs. These two latter exit measures (buybacks and sales to other VCs) account for about 2% of all VC exits and less than 0.6% of all VC investments in the VICO dataset. In line with Nahata, Hazarika, and Tandon (2014), our results hold upon reclassification of buybacks and sales to other VCs as successful VC exits. We control for macro factors at country-level that facilitate VC success (Nahata, Hazarika, and Tandon 2014). First, we extract the legal distances among countries from Berkowitz, Pistor, and Richard (2003), the country-level index of accounting conservatism provided by Bhattacharya, Daouk, and Welker (2003), the country-level efficiency index of debt enforcement provided by Djankov, Hart, McLiesh, and Shleifer (2008), the country-level index aggregating creditor rights provided by Djankov, McLiesh, and Shleifer (2007), and the two indexes of discharge from prebankruptcy indebtedness provided by Armour and Cumming (2008). Second, we extract data on the country-year market capitalization/gdp ratio from the World Bank in the portfolio company s country. The breakdown by country and industry of bank-affiliated VC (BVC)-backed firms is in Table 1. The most number of BVC investments in our sample is France, Germany, and the U.K., respectively, comprising 64% of the sample. The most common industries in which BVCs are involved include software, biotech/pharmaceuticals, and Internet, comprising 78% of the sample. Table 2 provides definitions for all of the variables in the sample, and descriptive statistics are provided in Table 3 for industry, year of investment, country market and legal conditions, and venture information. Table 3 Panel D shows that sales are on average larger for companies backed by syndicates involving BVCs relative to other types of syndicates, and larger relative to non-vc-backed companies. Debt to total assets is on average smaller among BVC syndicates relative to other types of syndicates, and relative to non-vc backed firms. Net cash flow to total assets is on average not statistically different between BVC syndicates and non-bvc syndicates, and smaller relative to non- VC backed firms. Return on assets is on average not statistically different among BVCs and non- BVC syndicates, but smaller relative to non-vc backed firms. Efficiency sales to total assets is on average lower among BVCs and non-vc syndicates relative to non-vc backed firms. IPOs are 8 Hochberg, Ljungqvist, and Lu (2007) show that such successful exit measures are strongly correlated with VC returns. 11

13 more likely among BVC syndicates than non-bvc syndicates, and relative to non-vc backed IPOs. Acquisitions are equally likely for BVC and non-bvc syndicates, but more likely among BVC syndicates than non-vc backed firms. Liquidations are equally likely among BVC syndicates, non- BVC syndicates, and non-vc backed firms. In the next sections, we examine these relationships in multivariate tests and with an assessment of causality. [Tables 1-3 about here] 4. Determinants of bank involvement in VC syndicates 4.1. Determinants of BVC involvement at the time of first investment First of all it is important to understand the determinants behind the willingness of a BVC fund to join a syndicate. Suppose that there are only two potential syndicates on the VC market (A and B) and one BVC fund. This latter can choose only one of the two syndicates according to the return to join each syndicate. Suppose that the return Y (i.e. exit multiple) of joining the syndicate A is ya and the return associated with the syndicate B is yb. Assume that ya<yb, and that the BVC fund bears a selection effort x (0,1) in the choice of the syndicate. The greater is the selection effort, the higher is the likelihood to get a higher return. Assume that the distribution of the overall risk of the portfolio of target companies of each syndicate is the same. This assumption is crucial to avoid that the difference in the expected value of the BVC fund is a pure function of the risk profile of the two syndicates. This assumption is similar to that used in the model of Das, Jo, and Kim (2011), i.e. any value arising from syndication is due to selection and/or monitoring activities. Thus, the expected return of the BVC fund is Y=(1-x)*yA+x*yB. If the total effort in selection and monitoring activities of the BVC fund is normalized to unity, the monitoring effort provided by the BVC fund to the syndicate is 1-x. If we assume that the probability of a positive exit depends on the monitoring effort of the BVC fund, the expected return is E[Y]=(1-x)[(1-x)*yA+x*yB] and we arrive at the same result of Das, Jo, and Kim (2011). If we assume that the BVC fund is a passive investor in the syndicate (Bottazzi, Da Rin, and Hellmann 2008), the probability of a positive exit does not depend on the monitoring effort of the BVC fund, and the expected return is E[Y]=p[(1-x)*yA+x*yB] where p (0,1) is the monitoring effort (e.g. proxied by the experience) of the syndicate leader and of the other active syndicate members. If we assume that p is a linear function of the selection effort of the BVC fund to find an experienced syndicate (given its passive nature in monitoring), we have that E[Y]=x[(1-x)*yA+x*yB]. Thus, taking the first derivative of the expected return with respect to the selection effort, we have: de[y]/dx=ya-2ya*x+2yb*x. Thus the optimal selection effort of the BVC 12

14 fund to maximize its return is x*=1/2(1-ππ) where ΠΠ=yB/yA. Thus, the higher is ΠΠ, the higher is the importance of the selection effort of the BVC fund. We can model the likelihood that the BVC fund joins the syndicate i to finance the portfolio company j at time t as prob[sijt] = f (Xit; Yjt; Zt), where Sijt is a dummy variable that equals 1 if the BVC fund joins the syndicate i to finance the portfolio company j at time t; Xit is a vector of syndicatespecific variables at time t (i.e. size of the syndicate, total amount provided, number of rounds, experience, type of leadership, staging frequency, geographic and legal distance between the BVC fund and the syndicate members); Yjt is a vector of portfolio company-level variables at time t (i.e. age, size, patent stock); Zt is a vector of controls at time t (industry, and macro factors of the country where the portfolio company operates). We assume that the experience of an investor represents a quality signal for the target company, and the size of the syndicate is a proxy of its commitment/value added to the target company. Having explained the importance to study the determinants of the likelihood that a BVC fund joins a VC syndicate at the time of first investment, we use the following probit model specification: Pr(BVC) = f(syndicate, Portfolio Firm, Controls). (1) Pr(BVC) is a dummy variable that equals one if a BVC fund is part of the VC syndicate, and zero otherwise. VCs tend to syndicate their investments rather than going alone (Gompers and Lerner 2004). As claimed by Lerner (1994a), syndication could be beneficial for several reasons, such as the advantage of a second opinion when screening companies (Casamatta and Haritchabalet 2007; Gompers and Lerner 2004), and the access to complementary resources, skills, networks and expertise of syndicate members (Andrieu and Groh 2012). The vector Syndicate includes the following variables. First, nvc is the size of the syndicate and represents the number of VCs backing the portfolio firm in the first investment year. Nahata, Hazarika, and Tandon (2014) find that larger syndicates lead to a higher likelihood of VC success. Second, VC Diversity controls for VC fund type of governance (i.e. independent, corporate, bankaffiliated, university-sponsored, governmental) and counts the number of sub-groupings of each governance type backing the portfolio firm in the first investment year. Diversity facilitates risk sharing, improved due diligence and value-added advice, mitigates hold-up problems in staged financing, and improves feasible outcomes (Gompers and Lerner, 1999). Third, Rounds is the total number of rounds received by the portfolio firm in the first investment year. The extant VC literature (Cornelli and Yosha 2003; Giot and Schwienbacher 2007; Gompers 1995) has shown that VCs use stage financing to have exit options at every financing round. When the milestones are reached by 13

15 the portfolio company, the uncertainty about the entrepreneurial project becomes lower, and thus the likelihood of a positive or negative exit is more likely (Giot and Schwienbacher 2007). Chemmanur and Tian (2011) find that VC syndicates that use more financing rounds are more likely to reach a positive exit. Fourth, VC Amount is the total VC investment given by all VC syndicate members to the portfolio firm in the first investment year (Cumming, Fleming, and Schwienbacher 2006; Giot and Schwienbacher 2007). As suggested by Barry and Mihov (2015), the incentive of monitoring is an increasing function of the VC amount. Fifth, VC-target Geo Distance is the average geographic distance between the portfolio firm and syndicate members (Giot and Schwienbacher 2007). The location of the entrepreneurial venture matters for the investment outcome. On the one hand, portfolio firms located in more developed countries in terms of business facilities, infrastructures, and start-up finance should perform better than other firms located in less developed countries. On the other hand, Lerner (1995) shows that VCs are more likely to serve on the boards of close portfolio companies. As to the geographical distance between the VC fund and the venture, the less is such distance and the easier is the VC monitoring activity (Giot and Schwienbacher 2007). Further, the less is the distance, the faster should be the VC exit (Cumming and Johan 2006), whether positive (next round) or negative (investment stopped). Wang and Wang (2012) find that the distance between VCs and portfolio companies increases monitoring and transaction costs and hampers the ability to collect information and manage the portfolio company. VC-target Geo Distance2 is the squared term of the average geographic distance between the portfolio firm and syndicate members. Sixth, VC-target Legal Distance is the average legal distance between the portfolio firm and syndicate members (for more details on the metric, see Berkowitz, Pistor, and Richard 2003). Espenalub, Khurshed, and Mohamed (2015, p. 218) claim that the quality and effectiveness of legal systems are crucial to promoting the development of capital markets including IPO and VC markets. Cumming, Fleming, and Schwienbacher (2006) show that efficient legal systems make IPOs more likely, mitigate agency problems between entrepreneurs and outside investors, allow VCs to extract more value at exit, and reduce the time to exit. Gompers and Lerner (1998) and Jeng and Wells (2000) show the importance of the quality of institutions in the fundraising activity. Laws impact not only exit but also the decision to invest. VC-target Legal Distance2 is the squared term of the average legal distance between the portfolio firm and syndicate members. Seventh, we control for the distances within the syndicate. As shown by Schertler and Tykvova (2011), one third of worldwide VC investments involve at least one syndicate member that is headquartered outside the country of the portfolio company. Syndicate Geo Distance is the average geographic distance among the syndicate members. Syndicate Geo Distance2 is the squared term of the average geographic distance among the syndicate members. Eighth, Syndicate Legal Distance is the average legal distance among the syndicate members. Syndicate 14

16 Legal Distance2 is the squared term of the average legal distance among the syndicate members. Ninth, we control for VC experience (Nahata 2008). Espenlaub, Khurshed, and Mohamed (2015) suggest that VCs differ in terms of abilities, expertise and specialization, and all of these three characteristics may be proxied by VC experience (Krishnan, Ivanov, Masulis, and Singh 2011). For instance, younger VCs usually grandstand (Gompers 1996). Conversely, more experienced VCs have wider networks (Hochberg, Ljungqvist, and Lu 2007) and better access to deal flow (Kaplan and Schoar 2005), also because of a less costly and larger fundraising (Nahata 2008). As claimed by Giot and Schwienbacher (2007) and Sørensen (2007), more experienced VCs are able to provide more value to their portfolio companies. In our model specification, we control for several types of experience. First, Local VC is a dummy variable that equals one if at least one of the syndicate members is headquartered in the same country where the portfolio firm operates, and zero otherwise. Nahata, Hazarika, and Tandon (2014) find that the presence of a local VC partner mitigates the liability of foreignness of foreign VCs and thus enhances venture success. In particular, local VCs help to screen and monitor portfolio companies. Second, US-based VC is a dummy variable that equals one if at least one of the syndicate members is headquartered in the US, and zero otherwise. Nahata, Hazarika, and Tandon (2014) claim that US-headquartered VCs are dominant players in the VC industry, and US-style VC contracts are more efficient (Kaplan, Martel, and Strömberg 2007). Nahata, Hazarika, and Tandon (2014) also find that the joint presence of local VCs and USheadquartered VCs in a syndicate lead to higher likelihood of investment success. While the former lowers the liability of foreignness, the latter bring experience. Third, VC Country Experience is a dummy variable that equals one if at least one of the syndicate members has already invested in the same country where the portfolio firm operates, and zero otherwise. VC Industry Experience is a dummy variable that equals one if at least one of the syndicate members has already invested in the same industry of the portfolio firm, and zero otherwise. VCs that have more international and crossindustry experience are able to develop more connections (Cumming, Knill, and Syvrud 2015). Bottazzi, Da Rin, and Hellmann (2008) and Du (2011) claim that industry experience allows VCs to be more actively involved in the management of their portfolio companies. Finally, we control for the type of leadership. GVC Leader is a dummy variable that equals one if the lead investor of the syndicate is a government-managed VC investor, and zero otherwise. CVC Leader is a dummy variable that equals one if the lead investor of the syndicate is a corporate VC investor, and zero otherwise. The vector Portfolio Firm includes the following variables. First, Total Assets is the logarithm of total assets of the portfolio firm in the first investment year. As shown by Puri and Zarutskie (2012), VC-backed firms are larger than non-vc-backed ones. Second, Age and Age2 are the logarithm of 15

17 firm age (measured by the years since firm foundation) and its squared term in the first investment year. As suggested by Bottazzi, Da Rin, and Hellmann (2008), younger companies need more support and advice from VCs. Third, Patent Stock is the logarithm of the patent stock in the first investment year (with yearly depreciation equal to 0.15). As highlighted by Giot and Schwienbacher (2007), VC targets are always characterized by high information asymmetries, and thus patents can act as a quality signal. The vector Controls includes a set of 2-digit SIC industry dummies, and year dummies. Moreover, we control for macro factors at country-level that facilitate VC success (Nahata, Hazarika, and Tandon 2014). Earnings Aggressiveness is a country-level index of accounting conservatism provided by Bhattacharya, Daouk, and Welker (2003, Table 2); Debt Enforcement is a country-level efficiency index of debt enforcement provided by Djankov, Hart, McLiesh, and Shleifer (2008; Table 2); Creditor Rights is a country-level index aggregating creditor rights provided by Djankov, McLiesh, and Shleifer (2007; Table 1); Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies (Armour and Cumming, 2008); Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies (Armour and Cumming, 2008); Market Capitalization is the country-level ratio between the stock market capitalization and the GDP (source: World Bank). As shown by Levine and Zervos (1998), Rajan and Zingales (2003), Beck, Levine, and Loayza (2000), and Wurgler (2000), stock market conditions positively impact on the value VCs can extract from portfolio companies, and finally on their exit dynamics (Cumming, Fleming, and Schwienbacher 2006; Espenlaub, Khurshed, and Mohamed 2015; Gompers and Lerner 1999; Jeng and Wells 2000; Lerner 1994b; Rhodes-Kropf, Robinson, and Viswanathan 2005; Shleifer and Vishny 2003). Black and Gilson (1998) claim that favorable IPO market conditions foster VC exits. Giot and Schwienbacher (2007) and Michelacci and Suarez (2004) find that more liquid markets impact all VC exit types. In fact, as suggested by Nahata, Hazarika, and Tandon (2014) and Espenlaub, Khurshed, and Mohamed (2015), stock market liquidity may also favor trade-sales through enhanced availability of equity capital. [Table 4 about here] Probit results are shown in Table 4. In column (1) we only insert the variables included in the vector Controls. In column (2) we add the variables included in the vector Portfolio Firm, and in column (3) we also add the variables included in the vector Syndicate. By focusing on the third 16

18 column, we find that BVCs usually prefer syndicates that invest in countries with relatively higher accounting disclosure of private firms, and with relatively more developed financial markets. The coefficient and the associated marginal effect of Earnings Aggressiveness is negative and statistically significant at 1%, while the coefficient of Market Capitalization is positive and statistically significant at 5%. Further, BVCs are attracted by larger syndicates: the coefficient of nvc is positive and statistically significant at 1%. The magnitude of the associated marginal effect at mean is +12% (significant at 1%). In terms of monitoring, we find an U-shaped relationship in the average legal distance between the portfolio firm and syndicate members, and the statistical significance of the coefficients of VC-target Legal Distance and VC-target Legal Distance2 is at 5% and 10% respectively. More interestingly, we find an inverted U-shaped relationship in the average legal distance within the syndicate: the statistical significance of the coefficients of Syndicate Legal Distance and Syndicate Legal Distance2 is at 5%. BVCs prefer syndicates led by IVCs, in fact the coefficients of the variables GVC Leader and CVC Leader are negative and statistically significant at 1%. In terms of marginal effects, GVC (CVC) leadership decreases the likelihood that BVCs join the syndicate by 16% (15%) both statistically significant at 1%. Finally, syndicates involving a USheadquartered VC fund seem to avoid the presence of BVCs (the marginal effect is -17%, significant at 1% confidence level) Selection bias Probit results in Table 4 (first three columns) may be driven by unobserved selection. In our context, such selection might be twofold. First, portfolio companies that fear horizontal agency costs with traditional VCs are more likely to match with VC syndicates involving a passive investor like a BVC fund. The presence of a BVC fund dilutes the stake and thus the activism of the leading investor. Second, VC syndicates whose portfolio companies are too risky are more likely to match with passive investors like BVCs. These latter allow to share the risk of the portfolio; moreover, BVCs are passive investors and thus the likelihood that agency costs between BVCs and leading investors bear is low. If this double selection is non-random, there may be a correlation between the error term in eq. (1) and the variables included in the vectors Syndicate and Portfolio Firm, leading to biased estimates. To check the reliability of our results, we used three different methodologies. First, we perform the methodology proposed by Sørensen (2007). 9 His strategy to identify selection effects in the matching process between VCs and portfolio companies is to consider both realized and potential dyads. As explained in the Introduction (see footnote 2) Sørensen (2007) identifies a 9 The classical instrumental variables (IV) approach is not suitable in our context: all variables included in the vector Syndicate are potentially endogenous. As to applications in VC literature, see e.g. Bottazzi, Da Rin, and Hellmann (2008), Hellmann and Puri (2002) and Hellmann, Lindsey, and Puri (2008). 17

19 screening effect through a full information market, and thus needs to exploit the characteristics of all economic agents and interactions among their choices. Specifically, we built all the potential dyads between VCs and portfolio companies and identified the realized ones. Then, we estimate the likelihood that a realized dyad includes a BVC fund. In line with Bottazzi, Da Rin, and Hellmann (2008) we avoid a Bayesian estimating approach. We drop all dyads with missing information, and thus we have 152,800 potential dyads. Probit results are reported in Table 4, column (4), and confirm that BVCs target portfolio companies in countries with relatively higher accounting disclosure of private firms, and with relatively more developed financial markets. It seems that BVCs are attracted by larger syndicates: the coefficient of nvc is positive but statistically significant at only 10%. We find a negative relationship between the average legal distance between the portfolio firm and syndicate members, but the statistical significance of the coefficient of VC-target Legal Distance is only at 10%. The coefficients of the variables GVC Leader and CVC Leader are negative and statistically significant at 5%. Second, we follow Bottazzi, Da Rin, and Hellmann (2008) and estimate two-step (Table 4, column (5)) and maximum likelihood Heckman procedure (Table 4, column (6)) on all potential dyads. For the selection equation we use country pairs VCs country and portfolio companies country fixed effects as exogenous identification restrictions because BVCs are more likely to invest locally than traditional VCs, so exploiting their ability to collect soft information (Coval and Moskowitz 2001; Mayer, Schoors, and Yafeh 2005). Results are fully in line with those shown in the first four columns of Table 4. Third, we used the methodology proposed by Ackerberg and Botticini (2002). Such methodology has already been used in the VC context by Bottazzi, Da Rin, and Hellmann (2008). The basic idea is that the matching between VCs and portfolio companies depends on all investorlevel and company-level characteristics that may be unobserved (or observed with error measurements). More specifically, Bottazzi, Da Rin, and Hellmann (2008) argue that the matching between the company i and the investor j depends on both the local availability of VCs and the geographic distribution of portfolio companies, being this latter exogenous. Under this assumption, each geographic market has a specific matching process. Following Bottazzi, Da Rin, and Hellmann (2008), we build 147 interaction terms given by the products of 7 countries (i.e. domestic markets), 7 industries (see Table 1) and 3 investment stages. We use such interaction terms as instruments for all the variables included in the vector Syndicate. Similarly to Bottazzi, Da Rin, and Hellmann (2008), we firstly estimate an IV probit model but such model does not achieve numerical convergence (as in the work of Bottazzi, Da Rin, and Hellmann 2008). Thus, we estimate a linear probability model. 18

20 Results are shown in Table 4 (column (7)) and are fully in line with the results shown in the first six columns of Table Impact of BVC on portfolio firm performance 5.1. Accounting for selection bias The estimation of BVC backing on firm performance must take into account of potential reverse causality problems. As suggested by Gompers and Lerner (2001), Sørensen (2007) and Nahata (2008) among others, the causality between VC and portfolio firm performance could be bidirectional, i.e. VC-backed firms may have an higher performance than non-vc-backed ones before the first VC investment because of a selection bias. Following Chemmanur, Krishnan, and Nandy (2011), the impact of BVC on portfolio firm performance is estimated by specifying the following panel data model: PPPPPPPPPPPPPPPPPPPPPP ii,tt = SSSSSSSSSSSSSSSSSS ii,tt + 2 PPPPPPPPPPPPPPPPPP FFFFFFFF ii,tt + αα 3 CCCCCCCCCCCCCCCC ii,tt + ββ 1 BBBBBB PPPPPPPP ii,tt + ββ 2 BBBBBB PPPPPP ii,tt + ηη ii + εε ii,tt. (2) PERFORMANCEi,t is the performance of venture i at time t. As performance measures, we alternatively use the following variables: logarithm of sales value (Paglia and Harjoto 2014; Puri and Zarutskie 2012), debt on total assets (Schmidt 2003), net cash flow on total assets (Gompers 1995), return on assets (Khrishnan, Ivanov, Masulis, and Singh 2011), and firm efficiency being this latter measured as the annual sales value divided by total assets (Ang, Cole, and Lin 2000). All accounting variables are winsorized at the 1 st and 99 th percentiles. The vector Syndicate includes the VC size (nvc) and represents the yearly number of VCs backing the portfolio firm i at time t, the number of sub-groupings of each VC governance type backing the portfolio firm (VC Diversity), and the total VC investment given by all VC syndicate members to the portfolio firm i at time t (VC Amount). The vector Portfolio Firm includes the logarithm of total assets of the portfolio firm i at time t (Total Assets), the logarithm of firm age (measured by the years since firm foundation) and its squared term (Age and Age2), and the logarithm of the patent stock at time t (Patent Stock). The variables included in the vector Controls are a set of 2-digit SIC industry dummies and year dummies, Earnings Aggressiveness, Debt Enforcement, Creditor Rights, Discharge, Discharge Years, and Market Capitalization. Finally, ηi is a firm-specific unobserved effect, and εi,t is an i.i.d. error term. The arguments of the logarithmic functions are augmented by one. BVC PRE i,t is a dummy variable that equals one for BVC-backed firms in years t-1 and t, with t representing the year in which the portfolio firm received its first VC investment, and 0 otherwise. BVC POST i,t is a dummy variable that equals one 19

21 for BVC-backed firms from the year after the first VC investment onwards, and 0 otherwise. As is common in the VC literature (Chemmanur, Krishnan, and Nandy 2011; Croce, Martí, and Murtinu 2013; Cumming, Grilli, and Murtinu 2014), BVC POST i,t does not switch back to 0 when the VCs exit the portfolio firm's equity. We estimate equation (2) by means of fixed effects (FE) regressions with standard errors clustered at venture level. As explained by Chemmanur, Krishnan, and Nandy (2011), FE estimation removes the time-invariant unobserved heterogeneity; moreover, the term BVC PRE i,t controls for selection effects. To study the dynamics of the impact of BVC investments on firm performance, we substitute in the Equation (2) the term BVC POST i,t with two dummy variables that distinguish the first two years after the first VC investment (BVC POST_SHORT i,t) from the time period starting from the third year after the first VC investment (BVC POST_LONG i,t). The marginal value added provided by VCs should be in fact higher in the early stages after VC backing and declining in the subsequent years (Black and Gilson, 1998). Equation (3) follows: PPPPPPPPPPPPPPPPPPPPPP ii,tt = SSSSSSSSSSSSSSSSSS ii,tt + 2 PPPPPPPPPPPPPPPPPP FFFFFFFF ii,tt + αα 3 CCCCCCCCCCCCCCCC ii,tt + ββ 1 BBBBBB PPPPPPPP_SSSSSSSSSS ii,tt + ββ 2 BBBBBB PPPPPPPP_LLLLLLLL ii,tt + ββ 3 BBBBBB PPPPPP ii,tt + ηη ii + εε ii,tt. (3) In Tables 5-9 we set other thresholds when distinguishing the short-term and the long-term impact of BVC. More specifically, BVC POST_SHORT i,t equals 1 from t+1 to t+3 (column (3)), t+4 (column (4)), t+5 (column (5)), with t representing the year of first VC investment, and 0 otherwise; while BVC POST_LONG i,t equals 1 from the fourth year (column (3)), the fifth year (column (4)) or the sixth year after the year of first VC investment (column (5)), and equal 0 otherwise (for the same approach on the same dataset, see Colombo and Murtinu 2016). Results of equations (2) and (3) are reported in Tables 5 (sales value), 6 (debt on total assets), 7 (net cash flow on total assets), 8 (ROA), and 9 (firm efficiency). As shown in Table 5, syndicates involving BVCs have a large positive impact on portfolio companies' sales value: after considering the selection effect, the estimates reported in column (1) indicate a positive impact (statistically significant at 5%). The estimated average magnitude in the post-investment period is +47%. 10 When investigating the dynamics of such sales performance improvements, we find a short-term impact that ranges from +34% to +43% depending on the selected short term window and a long-term impact ranging between +56% and +63%. As to the control variables, we find a positive contribution of the 10 The null hypothesis that syndicates involving BVCs have no impact on portfolio firm performance has been tested by means of the following Wald test: i) BVC POST i,t - BVC PRE i,t = 0. 20

22 country stock market capitalization/gdp ratio (significant at 1%). As to the variables included in the vector Portfolio Firm we find a positive impact of firm size on sales performance (significant at 1%) and an inverted U-shaped relationship between firm age and sales value. [Table 5 about here] In terms of DTA (Table 6), the impact is not statistically significant. This result is in line with the findings of Megginson, Meles, Sampagnaro, and Verdoliva (2016). They use a sample of 1,593 US firms that went public in the period, and find that VC-backed companies show less financial risk of distress after IPO than matched non-vc-backed companies. The authors explain this evidence with the screening activity pursued by VCs, who select companies with a lower financial distress risk. However, companies backed by more reputable VCs show a higher risk of financial distress because of more leverage. As to the effect of syndicates involving BVCs on firm liquidity (Table 7) as proxied by NCFTA this is negative and statistically significant at 1% (column (1)). The average estimate magnitude in the post-investment period is 21%. When investigating the dynamics of liquidity reductions, we find that such negative impact of BVCs has a greater magnitude in the short-term (between -21% and -23%) than in the long-term. As to the screening effect, we find that the liquidity of portfolio companies prior to receiving VC funding is 25% lower than that of non- VC-backed firms. As to the variables included in the vector Portfolio Firm we find a negative impact of patent stock (significant at 5%), i.e. entrepreneurial firms that invest in R&D and patent their technology are more likely to use internal finance (for more details see Carpenter and Petersen 2002). As to the variables included in the vector Syndicate we find a positive (negative) and statistically significant impact of syndicate size (VC amount), i.e. entrepreneurial firms backed by larger syndicates have less internal finance constraints but entrepreneurs are obligated to invest VC money. [Tables 6 and 7 about here] As to ROA (Table 8), even though the magnitude of the coefficients is very high (ranging from +95% to +109%) there is no statistically significant impact of syndicates involving BVCs on firm operating performance, neither in the short nor in the long run. In terms of firm efficiency (Table 9) as measured by means of the asset utilization ratio after considering the selection effect, the average impact of BVC in the post-investment period is positive and statistically significant at 1% and its magnitude is +7%. When investigating the dynamics of such efficiency improvements, we find an impact that ranges from +6% in the short-term to +9% in the long-term. As to the control 21

23 variables, we find a positive contribution of the country stock market capitalization/gdp ratio (significant at 5%). As to the variables included in the vector Portfolio Firm we show an inverted U- shaped relationship between firm age and efficiency. [Tables 8 and 9 about here] Overall, the above results indicate that VC syndicates involving BVCs improve firm efficiency through an improvement of product market performance (sales value). This latter may be attributed to enhanced investments, because of lower internal and external cash constraints (i.e. negative effect on NCFTA). These results are fully in line with the finance literature highlighting a strong impact of VC on product commercialization (Hellmann and Puri 2000): VCs in fact do not provide full funding upfront but instead portfolio companies must achieve a milestone (i.e. usually a certain amount of sales value) to get the next stage of money (Tian 2011). Hence, portfolio companies need to invest in better-quality workforce (Chemmanur, Krishnan, and Nandy 2011; Colombo and Murtinu 2016), and this is doable due to lower financial constraints (Bottazzi 2009). Syndicates involving BVCs behave as much as traditional VCs, confirming what claimed by Bottazzi (2009): bank VC firms [ ] are less likely to monitor their firms frequently or to sit on the board of directors (p. 44) and interact less frequently with the firms they invest in (p. 48). As to the screening of portfolio companies we confirm the lack of screening effects in Europe in terms of sales value, DTA, operating performance and firm efficiency (Colombo and Murtinu 2016; Croce, Martí, and Murtinu 2013). However, we contribute to the VC literature by showing a selection effect of BVCs in terms of portfolio companies liquidity. More specifically, the liquidity of portfolio companies (as measured by NCFTA) prior to receiving VC funding is 25% lower than that of non-vc-backed firms. This finding may be attributed to the above-explained strategic objectives of BVCs (Hellmann, Lindsey, and Puri 2008), i.e. enhance demand of debt capital from portfolio companies that need liquidity to invest and grow at a stable pace and that would borrow additional debt in the future from the parent bank of the BVC fund. Even though we follow Chemmanur, Krishnan, and Nandy (2011) in estimating fixed effects regressions with leads to control for selection bias, we further test the reliability of our results by means of an alternative procedure (explained in Sections 5.2) Selection on unobservables To further control for potential time-varying unknown selection bias, we follow Chetty, Friedman, and Rockoff (2014) and test the correlation between the VC variables in eq. (2) and 22

24 variables omitted from eq. (2), being these latter potentially correlated with our performance variables. The estimates shown in Tables 5-9 reliably estimate the impact of syndicates involving BVCs on firm performance if and only if a key assumption holds: VC variables in our models are not correlated with unobserved variables that influence firm performance. As to the unobserved variables, firstly we extract the reform measures implemented in the seven countries included in our sample in the areas Start-up financing, Seed financing, Access to finance in general and Access to finance for start-ups in general from the MICREF database, developed by the Directorate General for Economic and Financial Affairs (DG ECFIN) of the European Commission, in collaboration with the Joint Research Centre in Ispra and the Directorate General for Enterprise and Industry. Such reforms are shown in Table 10. Secondly, we use data on systemic banking crises provided by Laeven and Valencia (2013). While country dummies and year dummies in our models allow to control for time shocks in a specific country, a decline in credit stock and deposits may happen with a delay because loans are rolled over and held at book values in banks balance sheets for several years before they are written off (Laeven and Valencia 2013: p. 227). [Table 10 about here] These two variables financial reforms and bank crises should be strong predictors of firm performance. Hackbarth, Haselmann, and Schoenherr (2015) study the effect of the 1978 Bankruptcy Reform Act that shifted the bargaining power from debtholders to shareholders in case of corporate distress. The authors find that such reform affects the returns of distressed companies. Kerr and Nanda (2009) find a positive impact of US banking deregulations on entrepreneurship and churning among new entrants. Bekaert, Harvey, and Lundblad (2005) show that equity market liberalizations lead to an average 1% increase in the annual real economic growth, and this effect is greater in countries with an higher quality of institutions. Zarutskie (2006) studies the impact of the 1994 Riegle-Neal Interstate Banking and Branching Efficiency Act, which led to an increased competition in the US banking markets. The effect was an increase in firms financial constraints, as testified by a lower use of debt, a smaller corporate size, and an higher ROA. This effect is negatively moderated by firm age. Bertrand, Schoar, and Thesmar (2007) study the impact of the banking deregulation in France in the 1980s. They find that firms in industries more dependent on external finance are more likely to pursue restructuring activities. The aggregate result is a better allocative efficiency across firms and an increase in competition. Rice and Strahan (2010) study the effect of banking competition on small firm finance. They find that differences in US states' branching restrictions affect credit supply, and 23

25 in US states more open small firms are more likely to borrow from banks and could borrow money at lower interest rates (from 80 to 100 basis points) than those available to firms in less open states. Chava and Purnanandam (2011) find that negative shocks to the banking industry affect the performance of their borrowers. Further, this corporate effect is stronger in countries where publicdebt markets are less developed. The aggregate effect is a reduction in the credit supply and higher loan interest rates. Dell Ariccia, Detragiache, and Rajan (2008) confirm such aggregate effect and find that banking crises hit industries more dependent on external finance such as VC and countries with less financial development and less access to foreign finance. As to the procedure, firstly we run OLS regressions for every year included in our sample where the dependent variable is one of the dependent variables in eq. (2) alternatively. The independent variables are the financial reforms and the bank crises above explained. It is worth noting that there is no reverse causality between firm performance and these two independent variables (see more explanations about this approach in the VC literature in Colombo and Murtinu 2016). For each dependent variable for every year, we get the predicted performance of firm i. In Table 11, we reestimate our models in Tables 5-9 with the inclusion of the predicted performances. After controlling for the predicted portfolio firm performance, the impact of syndicates involving BVCs is +47% (significant at 5%) on sales value, not significant on DTA, -20% (significant at 1%) on NCFTA, not significant on ROA, and +7% (significant at 5%) on firm efficiency. Given that the variables about financial reforms and bank crises have strong predictive power for portfolio firm performance (with the exception of DTA) and they are not correlated with the BVC variables, we conclude that the degree of bias in our main models in Tables 5-9 is small. When comparing the effects of syndicates involving BVCs in Tables 5-9 (column (1)) and Table 11 it is clear that the relationship between BVC backing and firm performance does not seem to be driven by sorting on omitted variables about financial reforms and bank crises. [Table 11 about here] 6. Impact of BVC on investment outcome As to the investment outcome ideally one would like to have data on fund returns. In the absence of publicly available data on fund returns, we follow the extant VC literature (e.g. Gompers, Kovner, Lerner, and Scharfstein 2008; Hochberg, Ljungqvist, and Lu 2007; Sørensen 2007) and use positive (IPOs, acquisitions) and negative exits (liquidations) as proxies of fund performance. Gompers and Lerner (2001) point out that exit is the most important performance metric for VCs. Every VC decision is based on the likelihood of a timely exit because of the possibility to realize a 24

26 capital gain (Barry, Muscarella, Peavy, and Vetsuypens 1990; Brav and Gompers 1997; Giot and Schwienbacher 2007). As suggested by Espenlaub, Khurshed, and Mohamed (2015), among the alternative exit types, IPO is usually considered the most successful and the more profitable one, followed by trade-sale. In this study we consider these two positive exit types and also liquidations. In Table 12, we estimate the impact of BVC backing on the likelihood of an IPO, an acquisition or a liquidation by means of a multinomial logit. We estimate the following model: Exit Type = f(syndicate, Portfolio Firm, Controls). (4) Exit Type is a categorical variable and assumes four values: 1 for IPOs, 2 for acquisitions, 3 for liquidations, and 0 for the baseline category (companies that did not go public, were not acquired and are still in operation). It is worth noting that Exit type is always equal to zero for the baseline category of firms. For all the other companies, Exit Type is zero in all years prior to the first exit event (IPO, trade-sale or liquidation) and it assumes a value of one, two or three in the year of the first exit event, depending on the exit type. Exit Type is set to missing in the years after the first exit event. The vectors Syndicate, Portfolio Firm and Controls are the same as those described in Section 4.1. We find that syndicates involving BVCs have a large positive impact on the likelihood of positive exits BVC-backed companies have a higher likelihood to go public (+132%, significant at 1%) and to be acquired (+138%, significant at 1%) than non-vc-backed companies, respectively while there is no impact on firm liquidation. The economic significance of these effects are also shown graphically in Figures 1-3. These results hold even after controlling for other factors that influence exit. [Table 12 and Figures 1-3 about here] 7. Pre-estimation matched sample with IVC-backed companies Given that BVCs always syndicate with limited partnership IVCs in early stages, we classify IVC- and BVC-backed companies according to the type of governance of the VC fund that provided the first VC investment. Following the VC literature (e.g., Ivanov and Xie 2010; Colombo and Murtinu 2016), we define BVC-backed companies as companies that were initially backed by a BVC fund, potentially syndicated with one or more IVCs. IVC-backed were instead firstly backed by one or more IVCs. Separately for IVC-backed and BVC-backed companies, we built a matched sample of non-vc-backed companies by means of propensity score matching methodology in the year of the 25

27 first VC investment. 11 Propensity scores are obtained by means of probit models, where the independent variables are firm age, total assets, country and industry dummies. This set of regressors is that used by Puri and Zarutskie (2012), with the only difference being that as geographic controls we use European countries. PPPPPPPPPPPPPPPPPPPPPP ii,tt = SSSSSSSSSSSSSSSSSS ii,tt + 2 PPPPPPPPPPPPPPPPPP FFFFFFFF ii,tt + αα 3 CCCCCCCCCCCCCCCC ii,tt + ββ 1 IIIIII PPPPPPPP ii,tt + ββ 2 IIIIII PPPPPP ii,tt + ββ 3 BBBBBB PPPPPPPP ii,tt + ββ 4 BBBBBB PPPPPP ii,tt + ηη ii + εε ii,tt. (2) IVC PRE i,t is a dummy variable that equals one for IVC-backed firms in years t-1 and t, with t representing the year in which the portfolio firm received its first VC investment, and 0 otherwise. IVC POST i,t is a dummy variable that equals one for IVC-backed firms from the year after the first VC investment onwards, and 0 otherwise. BVC PRE i,t is a dummy variable that equals one for BVC-backed firms in years t-1 and t, with t representing the year in which the portfolio firm received its first VC investment, and 0 otherwise. BVC POST i,t is a dummy variable that equals one for BVC-backed firms from the year after the first VC investment onwards, and 0 otherwise. As is common in the VC literature (Chemmanur, Krishnan, and Nandy 2011; Croce, Martí, and Murtinu 2013), BVC POST i,t does not switch back to 0 when the VCs exit the portfolio firm's equity. We estimate equation (2) by means of fixed effects (FE) regressions with standard errors clustered at the target firm level. As explained by Chemmanur, Krishnan, and Nandy (2011), FE estimation removes the time-invariant unobserved heterogeneity; moreover, the terms IVC PRE i,t and BVC PRE i,t control for selection effects. In Table 13 we show results on sales value (column (1)), debt on total assets (column (2)), net cash flow on total assets (column (3)), ROA (column (4)), and firm efficiency (column (5)). Our findings show that syndicates involving BVCs have a large positive impact on portfolio companies' sales value: the estimated average magnitude in the post-investment period is +38%, and the estimated BVCs net effect is +6%. In terms of DTA, firm liquidity, ROA and firm efficiency there is no statistically significant impact. Overall, the above results indicate that VC syndicates involving BVCs improve firm efficiency through an improvement of product market performance (sales value). However, the real value addition of BVCs is only related to sales value. Our results do not fully confirm the passive nature of BVCs: in fact we find that BVCs shape syndication strategies, and add value to portfolio companies commercialization performance. 11 Matching methods have been used in many VC works (Chemmanur, Khrishnan, and Nandy 2011; Megginson and Weiss, 1991; Puri and Zarutskie, 2012). We match without replacement because of the ten-times larger sample of non- VC-backed companies in our dataset. 26

28 [Table 13 about here] In Table 14, we still use the above identification strategy to estimate the net effect that BVCs add to syndicates led by IVCs. When compared with a matched sample of IVC backed firms, we find that syndicates involving BVCs have a large positive impact on the likelihood of positive exits (+124%), and the estimated net effect of BVCs is non-negligible (+32%). As to firm liquidation, our estimates show that syndicates involving BVCs strongly increase the liquidation risk of portfolio companies: the overall effect is +271%, whose +168% is the estimated net effect of BVCs. [Table 14 about here] 8. Additional Robustness Checks 8.1. IV approach We re-estimate equation (2) without leads (BVC PRE i,t) by means of an instrumental variables (IV) two-stage least squares methodology to test the robustness of our main results (for the econometric explanation about the exclusion of BVC PRE i,t see Colombo and Murtinu 2016). The choice of the instrument is driven by the economic theory. In particular, we use the country s government composition. As shown by Serdar Dinc and Erel (2013), right-wing governments prefer that local firms that are targets of M&A processes remain owned by domestic investors that the foreign ones. However, right-wing governments usually favor pro-market policies, so increasing the development of local financial markets (Bortolotti, Fantini, and Siniscalco 2003). We use the countrylevel government composition given by the ratio between cabinet posts of right-wing parties and total cabinet posts (source: Armingeon, Knöpfel, Weisstanner, and Engler 2014). It is worth noting that we use the percentage of cabinet posts of right-wing political parties and not only a dummy left/right. In fact, as shown by Julio and Yook (2012), country-level political uncertainty leads firms to reduce investment expenditures. Relating the abovementioned arguments to our setting, foreign (local) VCs should be less (more) likely to invest in countries with strong right-wing majorities. At the same time, the higher is the political uncertainty, the lower should be the incentive of firms to look for VC to finance their investments. Results are available upon request from the authors. A second alternative instrument is driven by the extant theoretical and empirical literature about finance and institutions (Beck, Demirgüç-Kunt, and Levine 2003; La Porta et al., 1998, 1999). As claimed by Guiso, Sapienza, and Zingales (2003, p. 226): There is hardly an aspect of a society s life that is not affected by religion. Hilary and Hiu (2009) show that religion is linked to risk aversion, and this relationship also impacts organizational behavior. In particular, the authors find that an higher 27

29 level of religiosity positively impacts risk aversion (see also Kumar, Page, and Spalt 2011). Stulz and Williamson (2003) show that differences in a country s religion explain cross-sectional variations in creditor rights and investor protection. We may think that entrepreneurs in religious countries have lower rational expectations on their future investments and growth patterns, and thus today such entrepreneurs are less likely to look for VC. Further, banks may be less likely to enter VC syndicates and prefer to pursue their core business. At the same time, there is no reverse causality between the likelihood that the firm i receives BVC today and the religion of its country (Acemoglu, Johnson, and Robinson 2005). As to the metric of our instrument, we refer to La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1999, Appendix B). 12 Whichever the instrument used, the estimation is composed of two steps. In the first step, we estimate the probability of BVC backing. We do estimate this first step equation by means of a linear model to avoid the "forbidden regression" bias (Angrist and Pischke 2008, p. 190). In the second step, we estimate the equation (2) without leads using the predicted value at the first step. The values of the Kleibergen and Paap (2006) statistic always reject the null hypothesis of weak identification in both the LM and F versions, and IV results are fully in line with those in Tables Bayesian Model Averaging Evidence Through the Bayesian model averaging procedure developed by Leamer (1978) we deal with the typical uncertainty related to empirical model specifications. More specifically the covariates included in our models are based on a priori assumptions. Following Danilov and Magnus (2004), we test for the presence of auxiliary unnecessary regressors in our main model specifications. In particular we focus on the statistically significant results in Table 5 (column (1)), Table 7 (column (1)), Table 9 (column (1)), and Table 13 (columns (1) and (5)). To note that we focus on linear model specifications only, because of the properties of the above procedure (for more details see Magnus, Powell, and Prüfer 2010). A potential problem in our main models could be the presence of unobserved non-pairwise correlations among regressors that inflate our findings. To this extent, the identification of unnecessary auxiliary regressors allows us to improve the statistical properties of estimated parameters and thus get more reliable inference. More in detail, we identified the explanatory variables we want to keep in our main model specifications, i.e. IVC Pre, IVC Post, BVC Pre, BVC Post, year and industry dummies, and the constant. All other variables are assumed as potentially unnecessary. Then, we compute the weighted 12 We are conscious that this variable has some weaknesses, such as its country-level time-unvarying nature. However, as highlighted by the extant literature (e.g., Beck, Demirgüç-Kunt, and Levine 2003; La Porta et al., 1998, 1999), such variable does not change much from year to year. 28

30 average of conditional pooled OLS estimates of all potential specifications we can model by means of exclusion of the potentially unnecessary regressors. The exclusion of unnecessary regressors increases the precision of our estimates, at the expense of a higher likelihood of omitted variable bias. The criterion to still include a variable in our model specification is to look at the estimated a posteriori probability: a variable is reliably correlated with the dependent variable if the absolute value of the estimated t-ratio is greater than one. In Table 15 we apply the Bayesian model averaging procedure to our main model specifications. As to the estimates in column (1), after the estimation of 1024 potential model specifications, two characteristics of VC syndicates seem auxiliary: nvc and VC Amount show t- ratios (a posteriori probabilities) of and 0.13 (0.11 and 0.02) respectively. Hence, we reestimate our model in Table 5 (column (1)) by means of a FE regression with standard errors clustered at venture level, and with the exclusion of the two auxiliary regressors. After considering the selection effect, we find that syndicates involving BVCs have an average positive impact on portfolio companies' sales value, whose estimated magnitude is still +47% (statistically significant at 5%). In the same way, we find that syndicates involving BVCs negatively impact firm liquidity (as in Table 7, column (1)), and the average estimated magnitude is 22% (significant at 1%), while the average impact of BVC on firm efficiency is still positive (almost +8%) and statistically significant at 1% (as in Table 9, column (1)). Also the results on portfolio companies sales value and efficiency with preestimation matched sample with IVC-backed companies (Table 15, columns (4) and (5)) are fully in line with those shown in Table 13 (columns (1) and (5)). [Table 15 about here] 8.3. Bank ownership Univariate t-tests in Table 16 comparing BVCs affiliated to government-owned banks (GOVBVCs) with those affiliated to private ones (PRIVBVCs) show that PRIVBVCs prefer syndicates that invest in countries with relatively lower accounting conservatism of private firms and debt enforcement, and in larger portfolio companies through larger amounts of money than syndicates involving GOVBVCs. These latter also show a lower legal distance within the syndicate and towards portfolio companies, and always (never) include a local (corporate) investor. Nonparametric tests show strong differences in the distributions of portfolio companies country and syndicates amount of money between PRIVBVCs and GOVBVCs. However, the multivariate tests show very similar effects on portfolio firm performance. As such, we do not explicitly present these differences, but they are available on request. 29

31 We carried out additional tests pertaining to small versus large portfolio firms, and old and new firms by age, and segregated the samples. Likewise, we have carried out matching analyses of the samples by portfolio firm size and age. These results are likewise analogous to those reported above and available on request. [Table 16 about here] 8.4. Limitations, Extensions, and Future Research Despite the new large international sample that enables an assessment of the efficiency of BVCs, our data have a variety of limitations. For example, we do not have information on security design, contractual veto and control rights, board seats, and ownership percentages, and theory is consistent with the view that these terms could affect portfolio firm efficiency (Kirilenko, 2001). Likewise, we do not have evidence on the extent of due diligence and the specific nature of valueadded activities taken by BVCs and other investors to enhance the value of their investee firms. Some prior work on topic involving much smaller samples is suggestive that these factors matter for investee outcomes in different contexts. Further work on BVCs versus other types of VCs, and studies on the efficiency of VC more generally for all types of VCs, could make use of such information as new data become available. Further, our evidence is based on data from Western Europe, and further work could examine other countries to assess whether or not institutional differences in other settings give rise to material differences in the impact of BVCs on entrepreneurial firms. 9. Conclusions This paper examined the effect of bank-affiliated venture capital (BVC) on portfolio companies across seven Western European countries from 1991 to BVCs syndicate with limited partnership independent venture capitalists. We compare the effect of BVCs in a syndicate on several measures of portfolio company performance relative to non-vc backed firms, and we control for the potential presence of corporate, government, and university VCs within the syndicate and over time. We control for selection and matching between BVCs and portfolio firms, and carry out a number of robustness checks in ascertaining the impact of bank VC on portfolio firm efficiency. The data examined indicate that that BVC syndicates are larger, more likely in countries with lower earnings aggressiveness, higher market capitalization, and for local deals and international 30

32 deals among countries with similar legal systems. Accounting for such selection and matching, the data indicate that BVCs have a positive impact on portfolio firm sales and efficiency, but do not have an effect on debt to total assets, return on assets, and has a negative effect on net cash flows to total assets. Further, the data indicate BVCs have a positive impact on the probability of portfolio firms achieving an IPO or an acquisition. This evidence highlights an overall positive impact of BVCs in syndicates based on large sample international evidence from Western Europe. We have highlighted some limitations about our dataset, including unobservables in respect of the detailed cash flow and control rights in venture capital contracts, and specific actions taken by venture capitalists in improving portfolio firm performance. Future research could explore these dimensions of contract formation and actions of investors, and whether they are materially different for BVCs versus limited partnership VCs. Further, evidence in other regions around the world in developed and developing countries could provide insight into the generalizability of the results in this paper to other institutional contexts. Within Europe, it is worth exploring in future work the impact of the Brexit decision of June 2016 on the role of BVCs in different European countries, and the changing landscape of sources of capital around Europe for startups. This type of evidence will enable a better understanding about capital allocation and efficiency in start-up finance to better guide future academic work, as well as offer insights into best practice for practitioners and policymakers. 31

33 References Acemoglu, D., Johnson, S., & Robinson, J. A. (2005). Institutions as a fundamental cause of longrun growth. Handbook of economic growth, 1, Ackerberg, D. A., & Botticini, M. (2002). Endogenous matching and the empirical determinants of contract form. Journal of Political Economy, 110(3), Andrieu, G. (2013). The impact of the affiliation of venture capital firms: A survey. Journal of Economic Surveys, 27(2), Andrieu, G., & Groh, A. P. (2012). Entrepreneurs' financing choice between independent and bankaffiliated venture capital firms. Journal of Corporate Finance, 18(5), Ang, J. S., Cole, R. A., & Lin, J. W. (2000). Agency costs and ownership structure. Journal of Finance, 55(1), Angrist, J. D., & Pischke, J. S. (2008). Mostly harmless econometrics: An empiricist's companion. Princeton university press. Armingeon, K., Knöpfel, L., Weisstanner, D. & Engler, S. (2014). Comparative Political Data Set I Bern: Institute of Political Science, University of Berne. Armour, J., & Cumming, D.J. (2008). Bankruptcy Law and Entrepreneurship, American Law and Economics Review 10, Barry, C. B., & Mihov, V. T. (2015). Debt Financing, Venture Capital, and the Performance of Initial Public Offerings. Journal of Banking & Finance. Barry, C. B., Muscarella, C. J., Peavy, J. W., & Vetsuypens, M. R. (1990). The role of venture capital in the creation of public companies: Evidence from the going-public process. Journal of Financial economics, 27(2), Beck, T., Demirgüç-Kunt, A., & Levine, R. (2003). Law, endowments, and finance. Journal of Financial Economics, 70(2), Beck, T., Levine, R., & Loayza, N. (2000). Finance and the Sources of Growth. Journal of Financial Economics, 58(1), Bekaert, G., Harvey, C. R., & Lundblad, C. (2005). Does financial liberalization spur growth?. Journal of financial Economics, 77(1), Berkowitz, D., Pistor, K., & Richard, J. F. (2003). Economic development, legality, and the transplant effect. European Economic Review, 47(1), Bertoni, F., & Martí, J. (2011). Financing entrepreneurial ventures in Europe: the VICO dataset. Available at SSRN Bertrand, M., Schoar, A., & Thesmar, D. (2007). Banking deregulation and industry structure: Evidence from the French banking reforms of The Journal of Finance, 62(2), Bhattacharya, U., Daouk H., & Welker, M The world price of earnings opacity. The Accounting Review, 78: Black, B. S., & Gilson, R. J. (1998). Venture capital and the structure of capital markets: banks versus stock markets. Journal of financial economics, 47(3), Bortolotti, B., Fantini, M., & Siniscalco, D. (2004). Privatisation around the world: evidence from panel data. Journal of Public Economics, 88(1), Bottazzi, L. (2009). The role of venture capital in alleviating financial constraints of innovative firms. EIB Papers, 14(2), Bottazzi, L., Da Rin, M., & Hellmann, T. (2008). Who are the active investors?: Evidence from venture capital. Journal of Financial Economics, 89(3), Brav, A., & Gompers, P. A. (1997). Myth or reality? The long run underperformance of initial public offerings: Evidence from venture and nonventure capital backed companies. The Journal of Finance, 52(5), Butler, A. W., & Goktan, M. S. (2013). On the role of inexperienced venture capitalists in taking companies public. Journal of Corporate Finance, 22,

34 Carpenter, R. E., & Petersen, B. C. (2002). Is the growth of small firms constrained by internal finance?. Review of Economics and statistics, 84(2), Casamatta, C. (2003). Financing and advising: optimal financial contracts with venture capitalists. The Journal of Finance, 58(5), Casamatta, C., & Haritchabalet, C. (2007). Experience, screening and syndication in venture capital investments. Journal of Financial Intermediation, 16(3), Chahine, S., Arthurs, J. D., Filatotchev, I., & Hoskisson, R. E. (2012). The effects of venture capital syndicate diversity on earnings management and performance of IPOs in the US and UK: An institutional perspective. Journal of Corporate Finance, 18(1), Chava, S., & Purnanandam, A. (2011). The effect of banking crisis on bank-dependent borrowers. Journal of Financial Economics, 99(1), Chemmanur, T. J., & Chen, Z. (2014). Venture capitalists versus angels: The dynamics of private firm financing contracts. Review of Corporate Finance Studies, 3(1-2), Chemmanur, T. J., Hull, T., & Krishnan, K. (2011). Do Local and International Venture Capitalists Play Well Together? The Geography of International Venture Capital Investments. Working paper, Boston College. Chemmanur, T. J., Krishnan, K., & Nandy, D. K. (2011). How Does Venture Capital Financing Improve Efficiency in Private Firms? A Look Beneath the Surface. Review of Financial Studies, 24(12), Chemmanur, T. J., Loutskina, E., & Tian, X. (2014). Corporate venture capital, value creation, and innovation. Review of Financial Studies, 27(8), Chemmanur, T. J., & Tian, X. (2011). Peer Monitoring, Syndication, and the Dynamics of Venture Capitalist Interactions. SSRN Working Paper Series. Chen, H., Gompers, P., Kovner, A., & Lerner, J. (2010). Buy local? The geography of venture capital. Journal of Urban Economics, 67(1), Chetty, R., Friedman, J. N., & Rockoff, J. E. (2014). Measuring the Impacts of Teachers I: Evaluating Bias in Teacher Value-Added Estimates. The American Economic Review, 104(9), Colombo, M. G., & Murtinu, S. (2016). Venture Capital Investments in Europe and Portfolio Firms' Economic Performance: Independent versus Corporate Investors. Journal of Economics & Management Strategy, forthcoming. Cornelli, F., & Yosha, O. (2003). Stage financing and the role of convertible securities. The Review of Economic Studies, 70(1), Coval, J. D., & Moskowitz, T. J. (1999). Home bias at home: Local equity preference in domestic portfolios. Journal of finance, Coval, J. D., & Moskowitz, T. J. (2001). The Geography of Investment: Informed Trading and Asset Prices. The Journal of Political Economy, 109(4), Croce, A., Martí, J., & Murtinu, S. (2013). The impact of venture capital on the productivity growth of European entrepreneurial firms: Screening or value added effect?. Journal of Business Venturing, 28(4), Cumming, D. (2008). Contracts and exits in venture capital finance. Review of Financial Studies, 21(5), Cumming, D., Fleming, G., & Schwienbacher, A. (2006). Legality and venture capital exits. Journal of Corporate Finance, 12(2), Cumming, D. J., Grilli, L., & Murtinu, S. (2014). Governmental and independent venture capital investments in Europe: A firm-level performance analysis. Journal of Corporate Finance, in press, doi: /j.jcorpfin Cumming, D., Knill, A., Syvrud, K. (2015). Do international investors enhance private firm value? Evidence from venture capital. Journal of International Business Studies, forthcoming. Das, S. R., Jo, H., & Kim, Y. (2011). Polishing diamonds in the rough: The sources of syndicated venture performance. Journal of Financial Intermediation, 20(2),

35 Dell'Ariccia, G., Detragiache, E., & Rajan, R. (2008). The real effect of banking crises. Journal of Financial Intermediation, 17(1), Danilov, D., & Magnus, J. R. (2004). On the harm that ignoring pretesting can cause. Journal of Econometrics, 122(1), Djankov, S., Hart, O., McLiesh, C., & Shleifer, A. (2008). Debt Enforcement around the World. Journal of Political Economy, 116(6), Djankov, S., McLiesh, C., & Shleifer, A. (2007). Private credit in 129 countries. Journal of financial Economics, 84(2), Espenlaub, S., Khurshed, A., & Mohamed, A. (2015). Venture capital exits in domestic and crossborder investments. Journal of Banking & Finance, 53, European Parliament. (2012). Potential of Venture Capital in the European Union. Directorate General for Internal Policies. Policy Department A: Economic and Scientific Policy Industry, Research and Energy. Fang, L., Ivashina, V., & Lerner, J. (2013). Combining banking with private equity investing. Review of Financial Studies, 26(9), Fulghieri, P., & Sevilir, M. (2009). Size and focus of a venture capitalist's portfolio. Review of Financial Studies, 22(11), Giot, P., & Schwienbacher, A. (2007). IPOs, trade sales and liquidations: Modelling venture capital exits using survival analysis. Journal of Banking & Finance, 31(3), Gompers, P. A. (1995). Optimal investment, monitoring, and the staging of venture capital. Journal of finance, Gompers, P. A. (1996). Grandstanding in the venture capital industry. Journal of Financial economics, 42(1), Gompers, P., Kovner, A., & Lerner, J. (2009). Specialization and success: Evidence from venture capital. Journal of Economics & Management Strategy, 18(3), Gompers, P., Kovner, A., Lerner, J., & Scharfstein, D. (2008). Venture capital investment cycles: The impact of public markets. Journal of Financial Economics, 87(1), Gompers, P., & Lerner, J. (1998). Venture Capital Distributions: Short Run and Long Run Reactions. The Journal of Finance, 53(6), Gompers, P., & Lerner, J. (1999). An analysis of compensation in the US venture capital partnership. Journal of Financial Economics, 51(1), Gompers, P., & Lerner, J. (2000). The determinants of corporate venture capital success: Organizational structure, incentives, and complementarities. In Concentrated corporate ownership (pp ). University of Chicago Press. Gompers, P., & Lerner, J. (2001). The venture capital revolution. Journal of economic perspectives, Gompers, P. A., & Lerner, J. (2004). The venture capital cycle. MIT press. Gompers, P. A., Mukharlyamov, V., & Xuan, Y. (2016). The cost of friendship. Journal of Financial Economics, 119(3), Guiso, L., Sapienza, P., & Zingales, L. (2003). People's opium? Religion and economic attitudes. Journal of monetary economics, 50(1), Hackbarth, D., Haselmann, R., & Schoenherr, D. (2015). Financial distress, stock returns, and the 1978 Bankruptcy Reform Act. Review of Financial Studies, hhv009. Hellmann, T., Lindsey, L., & Puri, M. (2008). Building relationships early: Banks in venture capital. Review of Financial Studies, 21(2), Hellman, T., & Puri, M. (2000). The interaction between product market and financing strategy: The role of venture capital. Review of Financial studies, 13(4), Hilary, G., & Hui, K. W. (2009). Does religion matter in corporate decision making in America?. Journal of Financial Economics, 93(3), Hochberg, Y. V., Ljungqvist, A., & Lu, Y. (2007). Whom you know matters: Venture capital networks and investment performance. The Journal of Finance, 62(1),

36 Hollander, S., & Verriest, A. (2016). Bridging the gap: the design of bank loan contracts and distance. Journal of Financial Economics, 119(2), Inderst, R., & Müller, H. M. (2004). The effect of capital market characteristics on the value of startup firms. Journal of Financial Economics, 72(2), Ivanov, V. I., & Xie, F. (2010). Do corporate venture capitalists add value to start up firms? Evidence from IPOs and acquisitions of VC backed companies. Financial Management, 39(1), Jeng, L. A., & Wells, P. C. (2000). The determinants of venture capital funding: evidence across countries. Journal of corporate Finance, 6(3), Julio, B., & Yook, Y. (2012). Political uncertainty and corporate investment cycles. The Journal of Finance, 67(1), Kaplan, S. N., Martel, F., & Strömberg, P. (2007). How do legal differences and experience affect financial contracts?. Journal of financial intermediation, 16(3), Kaplan, S. N., & Schoar, A. (2005). Private equity performance: Returns, persistence, and capital flows. The Journal of Finance, 60(4), Kerr, W. R., & Nanda, R. (2009). Democratizing entry: Banking deregulations, financing constraints, and entrepreneurship. Journal of Financial Economics, 94(1), Kirilenko, A. A. (2001). Valuation and control in venture finance. The Journal of Finance, 56(2), Kleibergen, F., & Paap, R. (2006). Generalized reduced rank tests using the singular value decomposition. Journal of econometrics, 133(1), Krishnan, C. N. V., Ivanov, V. I., Masulis, R. W., & Singh, A. K. (2011). Venture capital reputation, post-ipo performance, and corporate governance. Journal of Financial and Quantitative Analysis, 46(05), Kumar, A., Page, J. K., & Spalt, O. G. (2011). Religious beliefs, gambling attitudes, and financial market outcomes. Journal of Financial Economics, 102(3), La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. (1998). Law and finance. Journal of Political Economy, 106(6), La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. (1999). The quality of government. Journal of Law, Economics, and organization, 15(1), Laeven, L., & Valencia, F. (2013). Systemic banking crises database. IMF Economic Review, 61(2), Leamer, E. E. (1978). Specification searches: Ad hoc inference with nonexperimental data (Vol. 53). John Wiley & Sons Incorporated. Lerner, J. (1994a). The syndication of venture capital investments. Financial management, Lerner, J. (1994b). Venture capitalists and the decision to go public. Journal of Financial economics, 35(3), Lerner, J. (1995). Venture capitalists and the oversight of private firms. The Journal of Finance, 50(1), Levine, R., & Zervos, S. (1998). Stock markets, banks, and economic growth. American Economic Review, Magnus, J. R., Powell, O., & Prüfer, P. (2010). A comparison of two model averaging techniques with an application to growth empirics. Journal of Econometrics, 154(2), Mayer, C., Schoors, K., & Yafeh, Y. (2005). Sources of funds and investment activities of venture capital funds: evidence from Germany, Israel, Japan and the United Kingdom. Journal of Corporate Finance, 11(3), Megginson, W. L., Meles, A., Sampagnaro, G., & Verdoliva, V. (2016). Financial Distress Risk in Initial Public Offerings: How Much Do Venture Capitalists Matter?. Available at SSRN Megginson, W. L., & Weiss, K. A. (1991). Venture capitalist certification in initial public offerings. The Journal of Finance, 46(3), Michelacci, C., & Suarez, J. (2004). Business creation and the stock market. The Review of Economic Studies, 71(2),

37 Nahata, R. (2008). Venture capital reputation and investment performance. Journal of Financial Economics, 90(2), Nahata, R., Hazarika, S., & Tandon, K. (2014). Success in global venture capital investing: do institutional and cultural differences matter?. Journal of Financial and Quantitative Analysis, 49(04), Paglia, J. K., & Harjoto, M. A. (2014). The effects of private equity and venture capital on sales and employment growth in small and medium-sized businesses. Journal of Banking & Finance, 47, Puri, M., & Zarutskie, R. (2012). On the life cycle dynamics of venture capital and non venturecapital financed firms. The Journal of Finance, 67(6), Rajan, R. G., & Zingales, L. (2003). The great reversals: the politics of financial development in the twentieth century. Journal of financial economics, 69(1), Rhodes Kropf, M., Robinson, D. T., & Viswanathan, S. (2005). Valuation waves and merger activity: The empirical evidence. Journal of Financial Economics, 77(3), Rice, T., & Strahan, P. E. (2010). Does Credit Competition Affect Small Firm Finance?. The Journal of Finance, 65(3), Sahlman, W. A. (1990). The structure and governance of venture-capital organizations. Journal of financial economics, 27(2), Schertler, A., & Tykvová, T. (2011). Venture capital and internationalization. International Business Review, 20(4), Schmidt, K. M. (2003). Convertible securities and venture capital finance. The Journal of Finance, 58(3), Semykina, A., & Wooldridge, J. M. (2010). Estimating panel data models in the presence of endogeneity and selection. Journal of Econometrics, 157(2), Serdar Dinc, I., & Erel, I. (2013). Economic nationalism in mergers and acquisitions. The Journal of Finance, 68(6), Shleifer, A., & Vishny, R. W. (2003). Stock market driven acquisitions. Journal of financial Economics, 70(3), Sørensen, M. (2007). How smart is smart money? A two sided matching model of Venture Capital. The Journal of Finance, 62(6), Sorenson, O., & Stuart, T. E. (2001). Syndication networks and the spatial distribution of venture capital investments1. American journal of sociology, 106(6), Stulz, R. M., & Williamson, R. (2003). Culture, openness, and finance. Journal of financial Economics, 70(3), Tykvová, T. (2016). When and Why Do Venture Capital-Backed Companies Obtain Venture Lending? Journal of Financial and Quantitative Analysis, forthcoming Tian, X. (2011). The causes and consequences of venture capital stage financing. Journal of Financial Economics, 101(1), Ueda, M. (2004). Banks versus venture capital: Project evaluation, screening, and expropriation. The Journal of Finance, 59(2), Wang, L., & Wang, S. (2012). Economic freedom and cross-border venture capital performance. Journal of Empirical Finance, 19(1), Winton, A., & Yerramilli, V. (2008). Entrepreneurial finance: Banks versus venture capital. Journal of Financial Economics, 88(1), Wurgler, J. (2000). Financial markets and the allocation of capital. Journal of financial economics, 58(1), Zarutskie, R. (2006). Evidence on the effects of bank competition on firm borrowing and investment. Journal of Financial Economics, 81(3),

38 Tables Table 1. BVC-backed firms across countries and industries BVC-backed firms (1) (2) (3) N. Obs. % Country Belgium Finland France Germany Italy Spain UK Total Industry ICT manufacturing TLC Internet Software Biotechnology & Pharmaceuticals Other Total Legend: the sample includes BVC-backed companies that enter the VICO dataset between 1991 and 2010 that first receive BVC financing in the year they enter the VICO dataset (if such year is not before 1994; see Section 3) or in any subsequent year. In column (1) the number of BVC-backed companies according to country and industry. In column (2) the firm-year observations of BVC-backed companies according to country and industry. In column (3) the percentage of BVC-backed companies according to country and industry. NACE codes of the industries are: ICT manufacturing (30.02, 32, 33); TLC (64.2); Internet (72.60); Software (72.2); Biotechnology & Pharmaceuticals (24.4, 73.1). 37

39 Table 2. Definition of variables Variable name Metric Source Earnings Aggressiveness Debt Enforcement Creditor Rights Discharge Country-level index of accounting conservatism Bhattacharya et al. (2003) Country-level efficiency index of debt enforcement Country-level index aggregating creditor rights Country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; one otherwise. For France it equals 0.5 in the time period Djankov, Hart, McLiesh, and Shleifer (2008) Djankov, McLiesh, and Shleifer (2007) Armour and Cumming (2008) Number of years until discharge from pre-bankruptcy Discharge Years indebtedness if such discharge is available for either sole Armour and Cumming proprietorships or guaranteed debts of closely-held (2008) private companies; if such discharge is unavailable, the variable equals life expectancy minus 40 Market Capitalization Country-level stock market capitalization/gdp ratio World Bank Total Assets Logarithm of total assets of firm i at time t VICO dataset Age Logarithm of age of firm i at time t VICO dataset Age2 Squared term of logarithmic age of firm i at time t VICO dataset Patent Stock Logarithm of the patent stock of firm i at time t (yearly depreciation = 0.15) VICO dataset, PATSTAT nvc Number of VCs backing the firm i at time t VICO dataset VC Diversity Number of sub-groupings of each VC governance type (i.e. independent, corporate, bank-affiliated, universitysponsored, VICO dataset governmental) backing the portfolio firm Rounds Number of rounds received by firm i in the first investment year VICO dataset VC Amount VC funding by all VC syndicate members to firm i at time t VICO dataset VC-target Geo Average geographic distance between firm i and Distance syndicate members in the first investment year VICO dataset VC-target Geo Squared average geographic distance between firm i and Distance2 syndicate members in the first investment year VICO dataset VC-target Legal Average legal distance between firm i and syndicate VICO dataset, Berkowitz, Distance members in the first investment year Pistor, and Richard (2003) VC-target Legal Squared average legal distance between firm i and VICO dataset, Berkowitz, Distance2 syndicate members in the first investment year Pistor, and Richard (2003) Syndicate Geo Average geographic distance among the syndicate Distance members in the first investment year VICO dataset Syndicate Geo Squared average geographic distance among the Distance2 syndicate members in the first investment year VICO dataset Syndicate Legal Average legal distance among the syndicate members in VICO dataset, Berkowitz, Distance the first investment year Pistor, and Richard (2003) Syndicate Legal Squared average legal distance among the syndicate VICO dataset, Berkowitz, Distance2 members in the first investment year Pistor, and Richard (2003) Local VC Dummy that equals one if at least one of the syndicate members in the first investment year is headquartered in VICO dataset the same country of firm i 38

40 US-based VC VC Country Experience VC Industry Experience GVC Leader CVC Leader BVC Pre BVC Post BVC Post Short BVC Post Long Dummy that equals one if at least one of the syndicate members in the first investment year is headquartered in the US Dummy that equals one if at least one of the syndicate members in the first investment year has already invested in the same country of firm i Dummy that equals one if at least one of the syndicate members in the first investment year has already invested in the same industry of firm i Dummy that equals one if the lead investor of the syndicate in the first investment year is a governmental VC investor Dummy that equals one if the lead investor of the syndicate in the first investment year is a corporate VC investor Dummy that equals one for BVC-backed firms in years t-1 and t, with t representing the year in which firm i received its first VC investment Dummy that equals one for BVC-backed firms from the year after the first VC investment onwards Dummy that equals one for BVC-backed firms in the first two, three, four or five years after the first VC investment alternatively Dummy that equals one from the third, fourth, fifth or sixth year after the first VC investment alternatively IVC Pre Dummy that equals one for IVC-backed firms in years t- 1 and t, with t representing the year in which firm i received its first VC investment IVC Post Sales DTA NCFTA ROA Efficiency IPO Acquisition Liquidation Dummy that equals one for IVC-backed firms from the year after the first VC investment onwards Logarithm of revenues from goods/services sold of firm i at time t (k ) Ratio between total financial debt and total assets of firm i at time t. Total financial debt does not include short-term receivables Ratio between net cash flow and total assets of firm i at time t Ratio between after tax net profit and total assets of firm i at time t Ratio between revenues from goods/services sold and total assets of firm i at time t Dummy that equals one if the first exit of firm i in a competing risk scenario is an IPO. Dummy that equals one if the first exit of firm i in a competing risk scenario is an acquisition. Dummy that equals one if the first exit of firm i in a competing risk scenario is a liquidation. VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset VICO dataset Legend: accounting variables are deflated using the consumer price index in 2005 as the reference year (source: Eurostat) and are winsorized at the 1 st and 99 th percentiles. 39

41 Table 3. Descriptive statistics Panel A: BVC-backed Obs. Mean Median Min Max Std. Dev. Earnings Aggressiveness Debt Enforcement Creditor Rights Discharge Discharge Years Market Capitalization Total Assets Age Patent Stock Sales DTA NCFTA ROA Efficiency IPO Acquisition Liquidation

42 Panel B: Other-VCbacked Obs. Mean Median Min Max Std. Dev. Earnings Aggressiveness Debt Enforcement Creditor Rights Discharge Discharge Years Market Capitalization Total Assets Age Patent Stock Sales DTA NCFTA ROA Efficiency IPO Acquisition Liquidation

43 Panel C: Non-VC-backed Obs. Mean Median Min Max Std. Dev. Earnings Aggressiveness Debt Enforcement Creditor Rights Discharge Discharge Years Market Capitalization Total Assets Age Patent Stock Sales DTA NCFTA ROA Efficiency IPO Acquisition Liquidation

44 Panel D: Comparison tests Variable BVC vs. Other VC Mean (Wald test) Other-VC vs. non-vc BVC vs. non-vc BVC vs. Other VC Median Other-VC vs. non-vc BVC vs. non-vc Sales DTA NCFTA ROA Efficiency IPO Acquisition Liquidation *** (0.0903) *** (0.1239) (0.0589) (0.1586) *** (0.0115) ** (0.0031) (0.0028) (0.0037) *** (0.0364) * (0.1316) *** (0.0606) *** (0.0611) *** (0.0046) *** (0.0008) *** (0.0011) (0.0015) *** (0.0836) *** (0.0530) *** (0.0377) *** (0.1478) *** (0.0107) *** (0.0030) ** (0.0026) (0.0034) Legend: Panel A includes VC-backed companies that enter the VICO dataset between 1984 and 2004 that first receive BVC funding in the year they enter the VICO dataset (if such year is not before 1994; see Section 3) or in any subsequent year. Panel B includes VC-backed companies that enter the VICO dataset between 1984 and 2004 that first receive VC funding (other than BVC) in the year they enter the VICO dataset (if such year is not before 1994; see Section 3) or in any subsequent year. Panel C includes non- VC-backed companies. In the first column Earnings Aggressiveness is a country-level index of accounting conservatism; Debt Enforcement is a country-level efficiency index of debt enforcement; Creditor Rights is a country-level index aggregating creditor rights; Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Total Assets is the logarithm of a firm s total assets; Age is the logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock. Sales is the logarithm of revenues from goods/services sold; DTA is the total financial debt on total assets ratio; NCFTA is the net cash flow on total assets ratio; ROA is the after tax net profit on total assets ratio; Efficiency is the sales value on total assets ratio; IPO is a dummy that equals one if the first exit in a competing risk scenario is an IPO; Acquisition is a dummy that equals one if the first exit in a competing risk scenario is an acquisition; Liquidation is a dummy that equals one if the first exit in a competing risk scenario is a liquidation. Panel D reports Wald tests on the difference in mean performance in the first three columns between: i) VC-backed companies that first receive BVC funding and VC-backed companies that first receive VC funding (other than BVC); ii) VC-backed companies that first receive VC funding (other than BVC) and non-vc-backed companies; and iii) VC-backed companies that first receive BVC funding and non- VC-backed companies. In the last three columns Panel D reports the p-value associated to the nonparametric two-sample Kolmogorov-Smirnov tests for equality in the distributions of variables between i) VC-backed companies that first receive BVC funding and VC-backed companies that first receive VC funding (other than BVC); ii) VC-backed companies that first receive VC funding (other than BVC) and non-vc-backed companies; and iii) VC-backed companies that first receive BVC funding and non- VC-backed companies. Standard errors in round brackets. * p <.10; ** p <.05; *** p <

45 Table 4. Determinants of BVC involvement at the time of first investment (1) (2) (3) (4) (5) (6) (7) Probit Probit Probit Probit Heckman (twostep) Heckman (ML) Heckman (twostep) Earnings Aggressiveness *** *** *** *** ( ) ( ) ( ) (8.2503) Debt Enforcement *** *** (0.0054) (0.0100) (0.0128) (0.0054) Creditor Rights (0.0632) (0.1097) (0.1370) (0.0572) Discharge (0.2639) (0.3632) (0.4639) (0.1929) (0.0851) (0.0851) (0.0912) Discharge Years * * * (0.0079) (0.0101) (0.0134) (0.0051) (0.0026) (0.0026) (0.0028) Market Capitalization ** *** ** ** (0.0023) (0.0034) (0.0041) (0.0016) (0.0006) (0.0006) (0.0006) Total Assets *** * * (0.0454) (0.0727) (0.0282) (0.0150) (0.0150) (0.0161) Age (0.3097) (0.3909) (0.1560) (0.0820) (0.0820) (0.0878) Age (0.1377) (0.1753) (0.0690) (0.0357) (0.0357) (0.0381) Patent Stock ** (0.1575) (0.2308) (0.0924) (0.0487) (0.0487) (0.0525) nvc *** * *** *** *** (0.1976) (0.0576) (0.0417) (0.0418) (0.0449) Rounds (0.1192) (0.0308) (0.0242) (0.0242) (0.0259) VC Amount (0.1179) (0.0431) (0.0196) (0.0196) (0.0208) VC-target Geo Distance (0.0005) (0.0002) (0.0001) (0.0001) (0.0001) VC-target Geo Distance VC-target Legal Distance (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) ** * ** ** **

46 VC-target Legal Distance2 (1.2595) (0.4435) (0.2791) (0.2793) (0.2951) * ** ** ** (0.2531) (0.0907) (0.0620) (0.0621) (0.0655) Syndicate Geo Distance (0.0005) (0.0002) (0.0001) (0.0001) (0.0001) Syndicate Geo Distance2 Syndicate Legal Distance Syndicate Legal Distance (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) ** *** *** *** (1.2353) (0.3139) (0.2446) (0.2447) (0.2626) ** *** *** ** (0.3237) (0.0798) (0.0660) (0.0659) (0.0702) Local VC * * * * (1.6780) (0.5771) (0.3159) (0.3152) (0.3337) US-based VC * ** ** ** (0.9064) (0.2808) (0.2050) (0.2050) (0.2204) VC Country Experience (0.3084) (0.1095) (0.0595) (0.0595) (0.0650) VC Industry Experience (0.4643) (0.1704) (0.1015) (0.1015) (0.1082) GVC Leader *** ** *** *** *** (0.3186) (0.1400) (0.0536) (0.0536) (0.0576) CVC Leader *** ** *** *** ** (0.4288) (0.1998) (0.0756) (0.0756) (0.0917) Industry dummies Yes Yes Yes Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Yes Yes Yes Obs Pseudo R Legend: estimates are derived by means of probit regressions in the first investment year (columns (1)-(4)), and two-step (columns (5) and (7)) and maximum likelihood Heckman procedure (column (6)). Earnings Aggressiveness is a country-level index of accounting conservatism; Debt Enforcement is a country-level efficiency index of debt enforcement; Creditor Rights is a country-level index aggregating creditor rights; Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Total Assets is the logarithm of a firm s total assets; Age is the logarithmic firm age; Age2 is the squared logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; Rounds is the number of rounds; VC Amount is the funding provided by all syndicate members; VC-target Geo Distance is the average geographic distance between the portfolio company and syndicate members; VC-target Geo Distance2 is the squared average geographic distance between the portfolio company and syndicate members; VC-target Legal Distance is the average legal distance between the portfolio company and syndicate members; VC-target Legal 45

47 Distance2 is the squared average legal distance between the portfolio company and syndicate members; Syndicate Geo Distance is the average geographic distance among the syndicate members; Syndicate Geo Distance2 is the squared average geographic distance among the syndicate members; Syndicate Legal Distance is the average legal distance among the syndicate members; Syndicate Legal Distance2 is the squared average legal distance among the syndicate members; Local VC is a dummy that equals one if at least one of the syndicate members is headquartered in the same country of the portfolio company; US-based VC is a dummy that equals one if at least one of the syndicate members is headquartered in the US; VC Country Experience is a dummy that equals one if at least one of the syndicate members has already invested in the same country of the portfolio company; VC Industry Experience is a dummy that equals one if at least one of the syndicate members has already invested in the same industry of the portfolio company; GVC Leader is a dummy that equals one if the lead investor of the syndicate is a governmental VC; CVC Leader is a dummy that equals one if the lead investor of the syndicate is a corporate VC. All regressions are estimated with an intercept term. Standard errors in round brackets. * p <.10; ** p <.05; *** p <

48 Table 5. Impact of BVC on portfolio firm performance: sales value (1) (2) (3) (4) (5) FE FE FE FE FE Discharge (0.5254) (0.5249) (0.5249) (0.5251) (0.5252) Discharge Years (0.0142) (0.0141) (0.0141) (0.0141) (0.0141) Market Capitalization *** *** *** *** *** (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) Total Assets *** *** *** *** *** (0.0159) (0.0159) (0.0159) (0.0159) (0.0159) Age *** *** *** *** *** (0.0519) (0.0519) (0.0520) (0.0519) (0.0519) Age *** *** *** *** *** (0.0354) (0.0353) (0.0353) (0.0353) (0.0353) Patent Stock (0.0823) (0.0823) (0.0824) (0.0825) (0.0825) nvc (0.0961) (0.0958) (0.0955) (0.0949) (0.0961) VC Diversity * * * * (0.2329) (0.2399) (0.2417) (0.2351) (0.2347) VC Amount (0.0389) (0.0391) (0.0395) (0.0389) (0.0390) BVC Pre (0.3329) (0.3341) (0.3335) (0.3340) (0.3338) BVC Post (0.3805) BVC Post Short (0.3876) (0.3866) (0.3852) (0.3891) BVC Post Long (0.3907) (0.3914) (0.3866) (0.3764) Year dummies Yes Yes Yes Yes Yes Obs Firms BVC Post BVC Pre BVC Post Short BVC Pre BVC Post Long BVC Pre ** (0.2377) (0.2423) ** (0.2641) (0.2439) ** (0.2632) * (0.2407) ** (0.2569) * (0.2456) ** (0.2447) Legend: the dependent variable is sales value. Estimates are derived by means of FE regressions with firm-clustered robust standard errors. Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Total Assets is the logarithm of a firm s total assets; Age is the logarithmic firm age; Age2 is the squared logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; VC Diversity is the number of sub-groupings of each VC governance type (i.e. independent, corporate, bankaffiliated, university-sponsored, governmental) backing the portfolio firm; VC Amount is the funding provided by all syndicate members; BVC Pre is a dummy that equals one for BVC-backed firms in years t-1 and t, with t representing the year of the first VC investment; BVC Post is a dummy that equals one for BVC-backed firms from the year after the first VC investment onwards; BVC Post Short is a dummy that equals one for BVC-backed firms in i) the first two years after the first VC investment (column (2)), ii) the first three years after the first VC investment (column (3)), iii) the first four years after the first VC investment (column (4)), iv) the first five years after the first VC investment (column (5)); BVC Post Long is a dummy that equals one from: i) the third year after the first VC investment (column (2)), ii) the fourth year after the first VC investment (column (3)), iii) the fifth year after the first VC investment (column (4)); i) the sixth year after the first VC investment (column (5)). Year and industry dummies are included in the estimates (coefficients are omitted in the table). All regressions are estimated with an intercept term. Standard errors in round brackets. * p <.10; ** p <.05; *** p <.01.

49 Table 6. Impact of BVC on portfolio firm performance: DTA (1) (2) (3) (4) (5) FE FE FE FE FE Discharge (7.0153) (7.0164) (7.0167) (7.0160) (7.0158) Discharge Years (0.2233) (0.2233) (0.2233) (0.2233) (0.2232) Market Capitalization (0.0050) (0.0050) (0.0050) (0.0050) (0.0050) Age (0.5627) (0.5638) (0.5642) (0.5646) (0.5644) Age (0.5651) (0.5664) (0.5666) (0.5667) (0.5664) Patent Stock (0.2487) (0.2479) (0.2481) (0.2484) (0.2490) nvc (0.0311) (0.0283) (0.0275) (0.0300) (0.0306) VC Diversity (0.0965) (0.0907) (0.0879) (0.0925) (0.0907) VC Amount (0.0235) (0.0240) (0.0240) (0.0237) (0.0231) BVC Pre (0.5339) (0.5316) (0.5294) (0.5248) (0.5264) BVC Post (0.6154) BVC Post Short (0.5746) (0.5802) (0.5803) (0.5874) BVC Post Long (0.6430) (0.6542) (0.6738) (0.6909) Year dummies Yes Yes Yes Yes Yes Obs Firms BVC Post BVC Pre BVC Post Short BVC Pre BVC Post Long BVC Pre (0.1350) (0.0944) (0.1845) (0.0978) (0.2043) (0.1084) (0.2256) (0.1151) (0.2327) Legend: the dependent variable is DTA. Estimates are derived by means of FE regressions with firm-clustered robust standard errors. Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Age is the logarithmic firm age; Age2 is the squared logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; VC Diversity is the number of sub-groupings of each VC governance type (i.e. independent, corporate, bank-affiliated, university-sponsored, governmental) backing the portfolio firm; VC Amount is the funding provided by all syndicate members; BVC Pre is a dummy that equals one for BVC-backed firms in years t-1 and t, with t representing the year of the first VC investment; BVC Post is a dummy that equals one for BVC-backed firms from the year after the first VC investment onwards; BVC Post Short is a dummy that equals one for BVC-backed firms in i) the first two years after the first VC investment (column (2)), ii) the first three years after the first VC investment (column (3)), iii) the first four years after the first VC investment (column (4)), iv) the first five years after the first VC investment (column (5)); BVC Post Long is a dummy that equals one from: i) the third year after the first VC investment (column (2)), ii) the fourth year after the first VC investment (column (3)), iii) the fifth year after the first VC investment (column (4)); i) the sixth year after the first VC investment (column (5)). Year and industry dummies are included in the estimates (coefficients are omitted in the table). All regressions are estimated with an intercept term. Standard errors in round brackets. * p <.10; ** p <.05; *** p <

50 Table 7. Impact of BVC on portfolio firm performance: NCFTA (1) (2) (3) (4) (5) FE FE FE FE FE Discharge (3.9804) (3.9801) (3.9800) (3.9803) (3.9804) Discharge Years (0.1055) (0.1055) (0.1055) (0.1055) (0.1055) Market Capitalization (0.0005) (0.0005) (0.0005) (0.0005) (0.0005) Age (0.0824) (0.0824) (0.0825) (0.0824) (0.0824) Age (0.0700) (0.0702) (0.0703) (0.0702) (0.0702) Patent Stock ** ** ** ** ** (0.0677) (0.0675) (0.0676) (0.0677) (0.0678) nvc * * * * * (0.0316) (0.0313) (0.0314) (0.0308) (0.0314) VC Diversity (0.0739) (0.0726) (0.0726) (0.0736) (0.0737) VC Amount * * * (0.0175) (0.0175) (0.0176) (0.0176) (0.0175) BVC Pre ** ** ** ** ** (0.1093) (0.1081) (0.1081) (0.1088) (0.1090) BVC Post *** (0.1166) BVC Post Short *** *** *** *** (0.1049) (0.1089) (0.1145) (0.1149) BVC Post Long *** *** *** *** (0.1301) (0.1337) (0.1312) (0.1382) Year dummies Yes Yes Yes Yes Yes Obs Firms BVC Post BVC Pre BVC Post Short BVC Pre BVC Post Long BVC Pre *** (0.0606) *** (0.0602) *** (0.0777) *** (0.0610) ** (0.0825) *** (0.0605) * (0.0866) *** (0.0594) * (0.0960) Legend: the dependent variable is NCFTA. Estimates are derived by means of FE regressions with firm-clustered robust standard errors. Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Age is the logarithmic firm age; Age2 is the squared logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; VC Diversity is the number of sub-groupings of each VC governance type (i.e. independent, corporate, bank-affiliated, university-sponsored, governmental) backing the portfolio firm; VC Amount is the funding provided by all syndicate members; BVC Pre is a dummy that equals one for BVC-backed firms in years t-1 and t, with t representing the year of the first VC investment; BVC Post is a dummy that equals one for BVC-backed firms from the year after the first VC investment onwards; BVC Post Short is a dummy that equals one for BVC-backed firms in i) the first two years after the first VC investment (column (2)), ii) the first three years after the first VC investment (column (3)), iii) the first four years after the first VC investment (column (4)), iv) the first five years after the first VC investment (column (5)); BVC Post Long is a dummy that equals one from: i) the third year after the first VC investment (column (2)), ii) the fourth year after the first VC investment (column (3)), iii) the fifth year after the first VC investment (column (4)); i) the sixth year after the first VC investment (column (5)). Year and industry dummies are included in the estimates (coefficients are omitted in the table). All regressions are estimated with an intercept term. Standard errors in round brackets. * p <.10; ** p <.05; *** p <

51 Table 8. Impact of BVC on portfolio firm performance: ROA (1) (2) (3) (4) (5) FE FE FE FE FE Discharge (1.4482) (1.4485) (1.4482) (1.4476) (1.4460) Discharge Years (0.0394) (0.0394) (0.0394) (0.0394) (0.0394) Market Capitalization (0.0005) (0.0005) (0.0005) (0.0005) (0.0005) Age (0.0569) (0.0569) (0.0570) (0.0570) (0.0571) Age (0.0678) (0.0679) (0.0679) (0.0679) (0.0680) Patent Stock (0.2045) (0.2042) (0.2045) (0.2046) (0.2041) nvc (0.2308) (0.2321) (0.2309) (0.2304) (0.2321) VC Diversity (0.3050) (0.3072) (0.3063) (0.3036) (0.3030) VC Amount (0.1317) (0.1319) (0.1318) (0.1316) (0.1314) BVC Pre (1.2341) (1.2360) (1.2349) (1.2321) (1.2248) BVC Post ** (0.1913) BVC Post Short *** *** ** ** (0.1792) (0.1883) (0.2007) (0.2160) BVC Post Long ** ** ** *** (0.2047) (0.2040) (0.1875) (0.1673) Year dummies Yes Yes Yes Yes Yes Obs Firms BVC Post BVC Pre BVC Post Short BVC Pre BVC Post Long BVC Pre (1.3419) (1.3271) (1.3538) (1.3397) (1.3475) (1.3482) (1.3284) (1.3680) (1.2403) Legend: the dependent variable is ROA. Estimates are derived by means of FE regressions with firm-clustered robust standard errors. Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Age is the logarithmic firm age; Age2 is the squared logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; VC Diversity is the number of sub-groupings of each VC governance type (i.e. independent, corporate, bank-affiliated, university-sponsored, governmental) backing the portfolio firm; VC Amount is the funding provided by all syndicate members; BVC Pre is a dummy that equals one for BVC-backed firms in years t-1 and t, with t representing the year of the first VC investment; BVC Post is a dummy that equals one for BVC-backed firms from the year after the first VC investment onwards; BVC Post Short is a dummy that equals one for BVC-backed firms in i) the first two years after the first VC investment (column (2)), ii) the first three years after the first VC investment (column (3)), iii) the first four years after the first VC investment (column (4)), iv) the first five years after the first VC investment (column (5)); BVC Post Long is a dummy that equals one from: i) the third year after the first VC investment (column (2)), ii) the fourth year after the first VC investment (column (3)), iii) the fifth year after the first VC investment (column (4)); i) the sixth year after the first VC investment (column (5)). Year and industry dummies are included in the estimates (coefficients are omitted in the table). All regressions are estimated with an intercept term. Standard errors in round brackets. * p <.10; ** p <.05; *** p <

52 Table 9. Impact of BVC on portfolio firm performance: efficiency (1) (2) (3) (4) (5) FE FE FE FE FE Discharge ** ** ** ** ** (0.1292) (0.1292) (0.1292) (0.1292) (0.1292) Discharge Years ** ** ** ** ** (0.0035) (0.0035) (0.0035) (0.0035) (0.0035) Market Capitalization ** ** ** ** ** (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) Age *** *** *** *** *** (0.0117) (0.0117) (0.0117) (0.0117) (0.0117) Age *** *** *** *** *** (0.0077) (0.0077) (0.0077) (0.0077) (0.0077) Patent Stock (0.0111) (0.0111) (0.0111) (0.0111) (0.0111) nvc (0.0126) (0.0126) (0.0125) (0.0125) (0.0126) VC Diversity * * * (0.0313) (0.0322) (0.0324) (0.0315) (0.0315) VC Amount (0.0055) (0.0055) (0.0056) (0.0055) (0.0055) BVC Pre (0.0408) (0.0409) (0.0409) (0.0409) (0.0409) BVC Post (0.0454) BVC Post Short (0.0467) (0.0465) (0.0461) (0.0465) BVC Post Long (0.0467) (0.0469) (0.0466) (0.0458) Year dummies Yes Yes Yes Yes Yes Obs Firms BVC Post BVC Pre BVC Post Short BVC Pre BVC Post Long BVC Pre ** (0.0321) (0.0327) ** (0.0353) * (0.0330) ** (0.0353) * (0.0326) *** (0.0346) * (0.0332) *** (0.0336) Legend: the dependent variable is the annual sales value divided by total assets. Estimates are derived by means of FE regressions with firm-clustered robust standard errors. Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Age is the logarithmic firm age; Age2 is the squared logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; VC Diversity is the number of sub-groupings of each VC governance type (i.e. independent, corporate, bank-affiliated, university-sponsored, governmental) backing the portfolio firm; VC Amount is the funding provided by all syndicate members; BVC Pre is a dummy that equals one for BVC-backed firms in years t-1 and t, with t representing the year of the first VC investment; BVC Post is a dummy that equals one for BVC-backed firms from the year after the first VC investment onwards; BVC Post Short is a dummy that equals one for BVC-backed firms in i) the first two years after the first VC investment (column (2)), ii) the first three years after the first VC investment (column (3)), iii) the first four years after the first VC investment (column (4)), iv) the first five years after the first VC investment (column (5)); BVC Post Long is a dummy that equals one from: i) the third year after the first VC investment (column (2)), ii) the fourth year after the first VC investment (column (3)), iii) the fifth year after the first VC investment (column (4)); i) the sixth year after the first VC investment (column (5)). Year and industry dummies are included in the estimates (coefficients are omitted in the table). All regressions are estimated with an intercept term. Standard errors in round brackets. * p <.10; ** p <.05; *** p <

53 Table 10. Reform measures implemented in the seven countries included in our sample Country Year Reform Area Title of reform measure France 2000 Start-up financing Introduction of a Business Start-up Loan (Pret à la Création d'entreprise or PCE) Spain 2000 Start-up financing Provision of early-stage capital to technology-based companies United Kingdom 2001 Start-up financing New business incubation fund United Kingdom 2001 Seed financing New programme of Early Growth Funding United Kingdom 2001 Access to finance in general Regional Venture Capital Fund Belgium 2002 Start-up financing ARKimedes (Flanders) Belgium 2002 Access to finance in general SOWALFIN SME Dome - SME Financial Umbrella Body in Wallonia France 2002 Start-up financing Set up of Co-investment Fund for Start-up Firms (Fonds de Co-investissement pour les Jeunes Entreprise or FCJE) Germany 2002 Start-up financing Introduction of 'micro loans' scheme Spain 2002 Start-up financing Opening of prefential credit lines to risk capital entities United Kingdom 2002 Access to finance in general Extension of Small Firms Loan Guarantee Scheme to previously excluded sectors United Kingdom 2002 Access to finance in general Launch of Bridges Community Development Venture Capital Fund Belgium 2003 Start-up financing Start-Up Fund United Kingdom 2003 Access to finance in general Improving the operation of SME and large company credits Finland 2004 Start-up financing Launch of Start Fund Vera Germany 2004 Seed financing Continuation of EXIST-SEED Germany 2004 Start-up financing ERP-Start-Fund Germany 2004 Start-up financing Act to promote Venture Capital Italy 2004 Access to finance for start-ups in general Loans to financial intermediaries to increase participation on innovative companies United Kingdom 2004 Access to finance for start-ups in general Enhanced tax relief for investments in Venture Capital Trusts Belgium 2005 Access to finance for start-ups in general Belgian Knowledge Centre for SME Financing (BeCeFi) Belgium 2005 Start-up financing Introduction of a micro-credit mechanism (Wallonia) Finland 2005 Start-up financing Revision of eligibility criteria for start-up grants France 2005 Access to finance for start-ups in general Innovation Development Contract and Innovation Guarantee Facility France 2005 Access to finance for start-ups in general Launch of the Alternext market France 2005 Access to finance for start-ups in general Creation of the OSEO Group Germany 2005 Start-up financing The High-Tech Start-Up Fund United Kingdom 2005 Access to finance for start-ups in general Establishment of the Enterprise Capital Funds (ECFs) United Kingdom 2005 Access to finance for start-ups in general Launch of the newly enhanced Small Firms Loan Guarantee (SFLG) scheme Belgium 2006 Access to finance for start-ups in general Creation of B2E Belgium 2006 Start-up financing Creation of the Win win loan (Flanders) 52

54 Belgium 2006 Start-up financing Spin off in Brussels France 2006 Start-up financing Simplification of the Business Start-up Loan Germany 2006 Start-up financing Grants to promote new businesses established by people previously unemployed Germany 2006 Access to finance in general SME innovative Spain 2006 Access to finance for start-ups in general Creation of a new public (ICO) credit line for entrepreneurs Spain 2006 Access to finance for start-ups in general Extension of the micro-credit programme for women United Kingdom 2006 Start-up financing Funding for additional rounds of competition for Enterprise Capital Funds (ECFs) Belgium 2007 Access to finance in general Simplification of R&D support for SMEs Germany 2007 Seed financing EXIST grant for entrepreneurs Italy 2007 Access to finance in general improvement female entrepreneurships Spain 2007 Access to finance for start-ups in general Launch of the CEIPAR programme to support the creation of innovative technology-based firms Belgium 2008 Access to finance in general Credit mediator for SMEs in Wallonia Finland 2008 Access to finance in general Introduction of counter-cyclical loans for SMEs and increase of commitments to Finnish export credit France 2008 Access to finance in general Scheme to support SMEs financing Germany 2008 Access to finance in general Act on Modernisation of the Framework for Private Equity Investors United Kingdom 2008 Access to finance in general Strengthening the Small Firms Loan Guarantee (SFLG) Belgium 2009 Access to finance in general Funding for applied research in SMEs Belgium 2009 Access to finance for start-ups in general Action for starters Finland 2009 Start-up financing Proposal on tax incentive for business angels Finland 2009 Access to finance for start-ups in general Start-up accelerator programme VIGO Italy 2009 Access to finance in general Plafond of 'Cassa Depositi e Prestiti' for SMEs Italy 2009 Access to finance in general Moratorium for SMEs United Kingdom 2009 Access to finance in general Support package to address the cash flow and credit need of SMEs Source: MICREF database 53

55 Table 11. Impact of BVC on portfolio firm performance: selection on unobservables (1) (2) (3) (4) (5) Sales DTA NCFTA ROA Efficiency Discharge ** (0.5263) (6.7987) (3.8587) (1.4521) (0.1295) Discharge Years ** (0.0142) (0.2151) (0.1023) (0.0393) (0.0035) Market Capitalization *** ** (0.0002) (0.0047) (0.0005) (0.0005) (0.0001) Total Assets *** (0.0159) Age *** *** (0.0519) (0.5620) (0.0815) (0.0570) (0.0117) Age *** *** (0.0353) (0.5640) (0.0688) (0.0678) (0.0077) Patent Stock ** (0.0822) (0.2489) (0.0676) (0.2043) (0.0111) nvc * (0.0960) (0.0313) (0.0327) (0.2306) (0.0127) VC Diversity * * (0.2330) (0.0965) (0.0767) (0.3046) (0.0314) VC Amount (0.0389) (0.0233) (0.0176) (0.1315) (0.0055) BVC Pre ** (0.3329) (0.5223) (0.1102) (1.2337) (0.0407) BVC Post *** ** (0.3803) (0.6029) (0.1174) (0.1914) (0.0454) Predicted Y *** * *** *** (0.0196) (0.2152) (0.4650) (0.2533) (0.0333) Year dummies Yes Yes Yes Yes Yes Obs Firms BVC Post BVC Pre ** (0.2377) (0.1336) *** (0.0603) (1.3418) ** (0.0321) Legend: the dependent variables are sales value (column (1)), DTA (column (2)), NCFTA (column (3)), ROA (column (4)) and the annual sales value divided by total assets (column (5)). Estimates are derived by means of FE regressions with firm-clustered robust standard errors. Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closelyheld private companies; Market Capitalization is the stock market capitalization/gdp ratio; Total Assets is the logarithm of a firm s total assets; Age is the logarithmic firm age; Age2 is the squared logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; VC Diversity is the number of sub-groupings of each VC governance type (i.e. independent, corporate, bank-affiliated, university-sponsored, governmental) backing the portfolio firm; VC Amount is the funding provided by all syndicate members; BVC Pre is a dummy that equals one for BVC-backed firms in years t-1 and t, with t representing the year of the first VC investment; BVC Post is a dummy that equals one for BVC-backed firms from the year after the first VC investment onwards; Predicted Y is the predicted value of a series of yearly OLS regressions where the regressors are the reform measures shown in Table 12 (source: MICREF database) and the data on systemic banking crises (source: Laeven and Valencia 2013) and the dependent variable is sales value, DTA, NCFTA, ROA or the annual sales value divided by total assets alternatively. Year and industry dummies are included in the estimates (coefficients are omitted in the table). All regressions are estimated with an intercept term. Standard errors in round brackets. * p <.10; ** p <.05; *** p <

56 Table 12. Impact of BVC on investment outcome: exit performance (1) (2) (3) IPO Acquisition Liquidation Discharge *** *** (0.5894) (0.3091) (0.2830) Discharge Years ** *** *** (0.0186) (0.0099) (0.0118) Market Capitalization * ** (0.0050) (0.0020) (0.0015) Total Assets *** *** *** (0.1127) (0.0420) (0.0489) Age *** (0.8672) (1.1023) (0.7297) Age ** *** (0.2386) (0.2689) (0.1951) Patent Stock *** (0.1276) (0.1370) (0.3903) nvc (0.1775) (0.5419) (0.1291) VC Diversity *** (0.4865) (0.6490) (0.2361) VC Amount *** ** (0.0713) (0.1111) (0.1480) BVC Post *** *** (0.4671) (0.3732) (0.7839) Industry dummies Yes Yes Yes Obs Firms Legend: estimates are derived by means of multinomial logit regressions with firm-clustered robust standard errors. In column (1) the dependent variable is the likelihood of an IPO. In column (2) the dependent variable is the likelihood of an acquisition. In column (3) the dependent variable is the likelihood of a liquidation. Discharge is a country-level dummy that equals zero if discharge from prebankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Total Assets is the logarithm of a firm s total assets; Age is the logarithmic firm age; Age2 is the squared logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; VC Diversity is the number of sub-groupings of each VC governance type (i.e. independent, corporate, bank-affiliated, university-sponsored, governmental) backing the portfolio firm; VC Amount is the funding provided by all syndicate members; BVC Post is a dummy that equals one for BVC-backed firms from the year after the first VC investment onwards. Industry dummies are included in the estimates (coefficients are omitted in the table). All regressions are estimated with an intercept term. Standard errors in round brackets. * p <.10; ** p <.05; *** p <

57 Table 13. Impact of BVC and IVC on portfolio firm performance: pre-estimation matched sample with IVC-backed companies (1) (2) (3) (4) (5) FE FE FE FE FE Discharge *** *** (2.0538) (8.5765) (1.6308) (8.1298) (0.3106) Discharge Years *** *** (0.0554) (0.2332) (0.0447) (0.2202) (0.0084) Market Capitalization ** (0.0008) (0.0055) (0.0006) (0.0007) (0.0001) Total Assets *** (0.0493) Age *** *** (0.1820) (0.3132) (0.1533) (0.1576) (0.0291) Age *** * *** (0.1340) (0.4134) (0.1758) (0.1917) (0.0209) Patent Stock (0.1470) (0.1467) (0.0680) (0.4435) (0.0190) nvc * * (0.0785) (0.1049) (0.0370) (0.0748) (0.0092) VC Diversity (0.1885) (0.0976) (0.0877) (0.1656) (0.0225) VC Amount ** ** (0.0234) (0.0372) (0.0115) (0.0214) (0.0028) IVC Pre (0.1520) (0.1669) (0.1134) (0.1723) (0.0236) IVC Post ** ** ** (0.1650) (0.1844) (0.1412) (0.1782) (0.0259) BVC Pre *** (0.2910) (0.2204) (0.1064) (0.6733) (0.0396) BVC Post *** * (0.3341) (0.1753) (0.0942) (0.1849) (0.0451) Year dummies Yes Yes Yes Yes Yes Obs Firms IVC Post IVC Pre *** (0.1110) BVC Post BVC ** Pre (0.1805) (0.2022) (0.1785) (0.1616) (0.1029) (0.1665) (0.7781) ** (0.0162) (0.0250) Legend: the dependent variable is sales value. Estimates are derived by means of FE regressions with firm-clustered robust standard errors. Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Total Assets is the logarithm of a firm s total assets; Age is the logarithmic firm age; Age2 is the squared logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; VC Diversity is the number of sub-groupings of each VC governance type (i.e. independent, corporate, bankaffiliated, university-sponsored, governmental) backing the portfolio firm; VC Amount is the funding provided by all syndicate members; IVC Pre is a dummy that equals one for IVC-backed firms in years t-1 and t, with t representing the year of the first VC investment; IVC Post is a dummy that equals one for IVC-backed firms from the year after the first VC investment onwards; BVC Pre is a dummy that equals one for BVC-backed firms in years t-1 and t, with t representing the year of the first VC investment; BVC Post is a dummy that equals one for BVC-backed firms from the year after the first VC investment onwards. Year dummies are included in the estimates (coefficients are omitted in the table). All regressions are estimated with an intercept term. Standard errors in round brackets. * p <.10; ** p <.05; *** p <

58 Table 14. Impact of BVC and IVC on exit performance: pre-estimation matched sample with IVCbacked companies (1) (2) IPO/Acquisition Liquidation Discharge *** *** (0.5781) (0.9394) Discharge Years *** (0.0180) (0.0474) Market Capitalization (0.0032) (0.0049) Total Assets *** ** (0.0868) (0.1475) Age (0.9871) (2.1206) Age (0.2669) (0.5659) Patent Stock ** (0.1643) (0.7566) nvc (0.2068) (2.2690) VC Diversity (0.4145) (5.9482) VC Amount * (0.0673) (3.1063) IVC Post *** * (0.3033) (0.5564) BVC Post *** *** (0.3582) (0.8172) Industry dummies Yes Yes Obs Firms Legend: estimates are derived by means of multinomial logit regressions with firm-clustered robust standard errors. In column (1) the dependent variable is the likelihood of an IPO or an acquisition. In column (2) the dependent variable is the likelihood of a liquidation. Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Total Assets is the logarithm of a firm s total assets; Age is the logarithmic firm age; Age2 is the squared logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; VC Diversity is the number of sub-groupings of each VC governance type (i.e. independent, corporate, bankaffiliated, university-sponsored, governmental) backing the portfolio firm; VC Amount is the funding provided by all syndicate members; IVC Post is a dummy that equals one for IVC-backed firms from the year after the first VC investment onwards; BVC Post is a dummy that equals one for BVC-backed firms from the year after the first VC investment onwards. Industry dummies are included in the estimates (coefficients are omitted in the table). All regressions are estimated with an intercept term. Standard errors in round brackets. * p <.10; ** p <.05; *** p <

59 Table 15. Impact of BVC on portfolio firm performance: sales value (1) (2) (3) (4) (5) Sales NCFTA Efficiency Sales Efficiency Discharge [1.00] [0.01] [1.00] [1.00] [1.00] Discharge Years [0.69] [0.00] [1.00] [1.00] [1.00] Market Capitalization [1.00] [0.01] [1.00] [1.00] [1.00] Total Assets [1.00] [1.00] Age [1.00] [0.22] [1.00] [1.00] [1.00] Age [1.00] [0.15] [1.00] [0.04] [0.96] Patent Stock [1.00] [0.01] [1.00] [1.00] [1.00] nvc [0.11] [0.00] [0.10] [1.00] [0.85] VC Diversity [0.90] [0.00] [0.24] [0.01] [0.15] VC Amount [0.02] [0.00] [0.02] [0.01] [0.05] Obs Model space BVC Post BVC Pre ** (0.2269) *** (0.0635) *** (0.0268) ** (0.1735) ** (0.0232) Legend: the dependent variables are sales value (columns (1) and (4)), NCFTA (column (2)), and the annual sales value divided by total assets (columns (3) and (5)). Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closely-held private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Total Assets is the logarithm of a firm s total assets; Age is the logarithmic firm age; Age2 is the squared logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; VC Diversity is the number of sub-groupings of each VC governance type (i.e. independent, corporate, bank-affiliated, university-sponsored, governmental) backing the portfolio firm; VC Amount is the funding provided by all syndicate members. We report the value of the estimated t-ratio, and the estimated a posteriori probabilities (in square brackets). The null hypothesis that syndicates involving BVCs have no impact on portfolio firm performance has been tested by means of the Wald test BVC POST - BVC PRE = 0, where BVC Pre is a dummy that equals one for BVC-backed firms in years t-1 and t, with t representing the year of the first VC investment; and BVC Post is a dummy that equals one for BVC-backed firms from the year after the first VC investment onwards. The above impact is derived by means of FE regressions with an intercept term and firm-clustered robust standard errors. Standard errors in round brackets. * p <.10; ** p <.05; *** p <

60 Table 16. Differences between government-owned and private BVC funds: univariate analysis Variables Earnings Aggressiveness Obs. Private Banks Government-Owned Banks Mean St.Dev. Obs. Mean St.Dev Debt Enforcement Difference Two-sample Kolmogorov -Smirnov test for equality of distribution functions T-test p-value *** ** * Creditor Rights Discharge Discharge Years Market Capitalization Total Assets * Age Patent Stock nvc VC Diversity Rounds VC Amount *** VC-target Geo Distance 129 VC-target Legal Distance *** Syndicate Geo Distance Syndicate Legal *** Distance Local VC ** US-based VC VC Country Experience VC Industry Experience GVC Leader CVC Leader ** Legend: Earnings Aggressiveness is a country-level index of accounting conservatism; Debt Enforcement is a country-level efficiency index of debt enforcement; Creditor Rights is a country-level index aggregating creditor rights; Discharge is a country-level dummy that equals zero if discharge from pre-bankruptcy indebtedness is available for either sole proprietorships or guaranteed debts of closelyheld private companies; Discharge Years is the number of years until discharge from pre-bankruptcy indebtedness if such discharge is available for either sole proprietorships or guaranteed debts of closely-held private companies; Market Capitalization is the stock market capitalization/gdp ratio; Total Assets is the logarithm of a firm s total assets; Age is the logarithmic firm age; Patent Stock is the logarithm of a firm s patent stock; nvc is the size of syndicate; VC Diversity is the number of sub-groupings of each VC governance type (i.e. independent, corporate, bank-affiliated, university-sponsored, governmental) backing the portfolio firm; Rounds is the number of rounds; VC Amount is the funding provided by all syndicate members; VC-target Geo Distance is the average geographic distance between the portfolio company and syndicate members; VC-target Legal Distance is the average legal distance between the portfolio company and syndicate members; Syndicate Geo Distance is the average geographic distance among the syndicate members; Syndicate Legal Distance is the average legal distance among the syndicate members; Local VC is a dummy that equals one if at least one of the syndicate members is headquartered in the same country of the portfolio company; US-based VC is a dummy that equals one if at least one of the syndicate members is headquartered in the US; VC Country Experience is a dummy that equals one if at least one of the syndicate members has already invested in the same country of the portfolio company; VC Industry Experience is a dummy that equals one if at least one of the syndicate members has already invested in the same industry of the portfolio company; GVC Leader is a dummy that equals one if the lead investor of the syndicate is a governmental VC; CVC Leader is a dummy that equals one if the lead investor of the syndicate is a corporate VC. In the last two columns, we report: i) the T-test on the difference between the mean variable of BVCs affiliated to private banks and the mean variable of BVCs affiliated to government-owned banks; and ii) the p-value associated 59

61 to the nonparametric two-sample Kolmogorov-Smirnov tests for equality in the distributions of variables of BVCs affiliated to private banks and of BVCs affiliated to government-owned banks. * p <.10; ** p <.05; *** p <

62 Figures Figure 1: IPO by BVC Legend: firm age and the estimated unconditional hazard of an IPO are on the horizontal and vertical axes respectively. The hazard function is the unconditional instantaneous probability of an IPO, provided that this has not occurred by t. The estimated hazard is calculated as a kernel smooth of the estimated hazard baseline contributions by means of a semiparametric Cox model. 61

63 Figure 2: Acquisition by BVC Legend: firm age and the estimated unconditional hazard of an acquisition are on the horizontal and vertical axes respectively. The hazard function is the unconditional instantaneous probability of an acquisition, provided that this has not occurred by t. The estimated hazard is calculated as a kernel smooth of the estimated hazard baseline contributions by means of a semiparametric Cox model. 62

64 Figure 3: Liquidation by BVC Legend: firm age and the estimated unconditional hazard of a liquidation are on the horizontal and vertical axes respectively. The hazard function is the unconditional instantaneous probability of a liquidation, provided that this has not occurred by t. The estimated hazard is calculated as a kernel smooth of the estimated hazard baseline contributions by means of a semiparametric Cox model. 63