Contents lists available at ScienceDirect

0
0
18
1 week ago
PDF Preview
Full text
(1)Journal of Business Venturing 25 (2010) 155–172. Contents lists available at ScienceDirect. Journal of Business Venturing. The role of top management team human capital in venture capital markets: Evidence from first-time funds☆ Rebecca Zarutskie ⁎ Duke University, Fuqua School of Business 1, Towerview Drive, Durham, NC 27708, United States. a r t i c l e. i n f o. Article history: Received 1 May 2007 Received in revised form 15 May 2008 Accepted 19 May 2008 Keywords: Venture capital Human capital First-time funds Investment management. a b s t r a c t This paper examines whether the human capital of first-time venture capital fund management teams can predict fund performance and finds that it can. I find that fund management teams with more task-specific human capital, as measured by more managers having past experience as venture capitalists and by more managers having past experience as executives at start-up companies, manage funds with greater fractions of portfolio company exits. I also find that fund management teams with more industry-specific human capital in strategy and management consulting and, to a lesser extent, engineering and non-venture finance manage funds with greater fractions of portfolio company exits. Perhaps counter-intuitively, I find that fund management teams that have more general human capital in business administration, as measured by more managers having MBAs, manage funds with lower fractions of portfolio company exits. Overall, measures of task- and industry-specific human capital are stronger predictors of fund performance than are measures of general human capital. © 2008 Elsevier Inc. All rights reserved.. 1. Executive summary The rapid growth of venture capital as a source of financing for start-up companies and as a fraction of institutional investors’ portfolios has made gaining an understanding of the determinants of venture capital investment performance an important objective. Prior studies have made progress in documenting some general features of venture capital investment performance.1 However, there remain many questions about the underlying mechanisms behind the documented features of venture capital investment performance. In particular, how venture capitalists’ human capital affects the performance of the investments they make and how differences in human capital of venture capital fund management teams may give some funds a performance advantage are two important outstanding questions. In this paper, I collect data on the educational and work histories of venture capitalists who start first-time venture capital funds and use this information to test several hypotheses about the relation between venture capital fund management team human capital and venture capital fund performance. If venture capitalists’ skills in managing a venture capital fund’s investments matter we should see statistically significant correlations between measures of venture capitalist human capital and the performance of their funds. I test several hypotheses about the impact of task-specific, industry-specific and general human capital on venture capital fund performance.. ☆ A previous version of this paper was circulated under the title “Do venture capitalists affect investment performance? Evidence from first-time funds”. I thank seminar participants at Carnegie Mellon, Duke, the first RICAFE2 conference at the London School of Economics, the 2007 American Finance Association Meetings and the 2007 Western Finance Association Meetings for their comments and Sridhar Arcot, Rudiger Fahlenbrach, Simon Gervais, David Hsu, Laura Lindsey, Manju Puri, David Robinson, Catherine Schrand. Dean Shepherd (the associate editor) and two anonymous referees for their suggestions. I thank Joseba Celaya, Adrian Cighi and Ling Luo for research assistance. All errors are mine. ⁎ Tel.: +1 919 660 7981; fax: +1 919 660 8038. E-mail address: rebeccaz@duke.edu. 1 See, for example, Cochrane (2005), Kaplan and Schoar (2005), Phalippou and Gottschalg (2006) and Hwang, Quigley and Woodward (2005). 0883-9026/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.jbusvent.2008.05.008.

(2) 156. R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. I find that the strongest predictors of first-time venture capital fund performance are what I term task-specific and industry-specific measures of human capital rather than general measures of human capital. In particular, first-time funds whose management teams possess more prior venture capital investing experience and more prior experience managing start-up companies exhibit greater fractions of portfolio company exits. I also find that funds whose management teams possess more prior work experience in management and strategy consulting, non-venture finance and professional science and engineering experience greater fractions of portfolio company exits. I find that having more general human capital within a venture capital fund management team, as measured by managers’ education in particular areas such as business, law and science and engineering, does not robustly predict better performance. However, I do find, perhaps counter-intuitively, that management teams with more general human capital in business obtained through MBAs perform on average worse than other fund management teams. A possible explanation for this result is that there is an oversupply of individuals of possessing MBAs relative to those with other educational backgrounds who are typically candidates to enter the venture capital industry. Finally, I show that the human capital measures that predict the performance of venture capital funds also predict whether the management team is able to raise a follow-on fund. The findings support the idea that differences in venture capitalist human capital partially explain the documented heterogeneity and persistence in venture capital fund performance (e.g., Kaplan and Schoar (2005)). The findings also suggest which measures of human capital matter and point us towards future research that can shed further light on how venture capitalists with different types of human capital affect the outcomes of the investments they make. 2. Introduction A central question in the financial economics literature is whether there are differences in the abilities of investment managers exhibit superior investment performance (e.g., Fama, 1970). A central question in the management and strategy literature is how top management teams affect a firm’s decisions and subsequent performance (e.g., Hambrick and Mason, 1984). Fundamentally related to both of these literatures is the labor and organizational economics literature on human capital (e.g., Becker, 1964). While there is a large empirical management literature that investigates the role of human capital on organizational outcomes (e.g., Beckman and O’Reilly, 2007; Bertrand and Schoar, 2003; Gimeno, et al., 1997; Pennings et al., 1998; Wiersema and Bantel, 1992), there is comparatively little research in financial economics on the role that human capital plays in explaining differences in investment performance, despite a fairly large empirical literature which looks for predictability in investment performance more generally (e.g. Brown and Goetzmann, 1995; Elton, et al., 1993; Jensen 1968).2 This is surprising since making and managing investments is a research- and information-based activity which requires a large amount of human effort. A natural place to look for evidence that differences in investor human capital, or skill, predict investment performance is professionally managed investment vehicles, such as mutual funds, hedge funds and venture capital and private equity funds. The researcher can often observe the identities of the managers of these investment vehicles as well as measures of investment performance of these variables. This paper investigates the role the human capital of venture capital fund management teams plays in predicting the performance of venture capital funds. The paper’s contributions are fourfold. First, I contribute to the financial economics literature on investment performance predictability. Additionally, because I focus on fund-level performance I am able to examine whether human capital measures which predict the performance of first-time funds also predict whether a follow-on fund is raised, enabling one test of whether human capital differences can explain persistence in venture capital fund performance documented in previous work. Second, I am able to test several new hypotheses about how the human capital of top management teams affects firm performance. In particular, I test new hypotheses about how two types of specific human capital, which I call task-specific and industry-specific human capital, are related to first-time venture capital fund performance. The hypotheses I test are related to those tested in a recent study by Dimov and Shepherd (2005) which examines the influence of general and specific, broadly defined, human capital on the performance of venture capital firms. Because I collect a different set of manager biographical variables, I test a different, more nuanced, set of hypotheses about the roles of specific and general human capital types, such as task- and industry-specific human capital, in venture capital fund performance. Third, my larger data sample allows me to control for a variety of other factors that may influence venture capital fund performance in order to more accurately assess the influence of management team human capital on fund performance. In so doing, I am able to test a number of previously tested secondary hypotheses about the impact of non-human capital measures on venture capital fund performance in a more recent sample of venture capital funds. Finally, by focusing my analysis on first-time venture capital funds, I generate some novel statistics on the education and employment histories of venture capitalists who form first- time funds and how these venture capitalists join together in teams. Understanding who forms first-time funds and what determines which funds succeed is important not only for investors seeking to invest in such funds but also for understanding which types of venture capitalists may be needed to create a successful venture capital market in regions where such markets are nascent or nonexistent. The rest of the paper is structured as follows. Section 3 discusses the hypotheses to be tested and the theoretical literature motivating them. Section 4 introduces the data and describes the characteristics of venture capitalists raising first-time funds. Section 5 presents the empirical analysis and the main findings. Section 6 discusses the main findings and suggests future directions for research. Section 7 concludes.. 2 Two notable exceptions are empirical studies on the role investor human capital plays in predicting investment performance are Chevalier and Ellison (1999) and Golec (1996) in the setting of the mutual fund industry..

(3) R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. 157. 3. Theory and hypotheses 3.1. Human capital theory, upper echelon theory and venture capital markets Any hypothesis that investment manager human capital should predict investment performance implicitly invokes upper echelon theory (e.g., Hambrick and Mason, 1984; Finkelstein and Hambrick, 1996) and the resource-based view of the firm (Barney, 1991; Peteraf, 1993). A fundamental premise of upper echelon theory is that top management teams matter for firm performance. In the case of a venture capital fund characteristics of the fund management team should be able to predict the performance of the fund. According to the resource-based view of the firm, there must be some limit to the ability of some fund managers to obtain the human capital known to improve investment performance in order for certain fund management teams to perform better than others. Motivated by work in upper echelon theory, an empirical literature has developed to assess the role top management teams play in firm outcomes. This literature is closely related to the labor economics literature on human capital. According to upper echelon theory team processes that drive decision-making matter for firm performance. Since the empiricist typically cannot observe these team-level processes directly, he must rely on observable team characteristics that proxy for them. Many of the management team characteristics studied in the context of upper echelons are measures of educational level, educational specialty and work background, classical measurements of human capital. Human capital theory argues that, to the extent these variables proxy for scarce skill or skills that are costly to acquire, these measures should be correlated with better firm performance. Measures of the collective human capital of a management team are arguably correlated with the kinds of processes and dynamics the team experiences and should also, therefore, be correlated with firm performance. For example, Finkelstein and Hambrick (1990) and Miller (1991) examine the role of management team tenure on firm performance while Bantel and Jackson (1989) also examine management team tenure heterogeneity along with management team age and education level on firm innovation. In addition, management team composition and heterogeneity along demographic variables as well as networking or social capital of management teams have been examined in the context of firm performance and other outcomes (e.g., Wiersma and Bantel, 1992; Boeker, 1997; Pennings et al., 1998). Another line of inquiry related to both the human capital and upper echelon literatures is an empirical literature which explores the relation between measures of entrepreneurs’ backgrounds on the performance of the firms they start. For example, Amason, Shrader and Tompson (2006), Beckman, Burton and O’Reilly (2007) and Chowdhury (2005) each examine the relation between top management team characteristics in new ventures and the subsequent performance of those new ventures. The present study adds to the empirical literatures on top management teams and human capital by testing several hypotheses about the relation between types of specific and general human capital of fund management teams on venture capital fund performance. The distinction between general and specific human capital has been made in several studies, beginning with Becker (1964). While definitions have varied from one setting to another, typically general human capital is defined as skills that can be generally applied across most firms and settings and specific human capital is specific to a particular time or setting, such as human capital specific to a firm, industry or task. In this paper, I make the distinction amongst task-specific, industry- specific and general human capital of venture capital fund managers. While the definitions of these human capital types are unique to the particular empirical setting of this paper, the hypotheses I test are related to existing theories on the role of types of human capital within the firm. For example, Gibbons and Waldman (1999, 2004) put forth a model in which task-specific human capital is developed and determines the wage and promotion structure inside the firm. Lazear (1995) develops a model of job-specific human capital, which can be used also be used to motivate the measures of task-specific human capital I employ. Further, the idea of industryspecific human capital is discussed in Kletzer (1989) and Neal (1995). In the next subsection, I formalize the hypotheses I test and relate them to existing theory. 3.2. Primary hypotheses In this paper I assess the impact of three different types of manager human capital, task-specific, industry-specific and general human capital, on the performance of venture capital funds. In particular, I define task-specific human capital as human capital specific to two primary tasks of participants in the venture capital markets — managing a venture capital fund and running a startup company. Using motivating theory from Gibbons and Waldman (1999, 2004) and Lazear (1995), I argue that venture capital fund manager teams with more experience in these two tasks should be more likely to manage better performing funds. Fund managers with experience managing funds will likely have learned skills necessary for running a fund through trial and error — a learning process less likely to be obtained through other work or educational avenues. Fund managers who have managed funds previously should be more experienced and be better able to run a first-time fund due to a combination of a better understanding of which companies to fund, better access to these good companies as well as knowing how to actively manage those investments. Likewise, since venture capitalists must screen for good managers and start-ups when choosing in which companies to invest and may also assist companies in selecting managers after investing in those companies, having experience in the task of managing a start-up may aid the performance of venture capital fund managers in these tasks. Once again the trial and error learning that a fund manager may achieve by starting and running his or her own firm may allow him or her to acquire skills not easily obtained elsewhere. While fund returns have become recently available to researchers, they are usually done so under the condition of fund anonymity. If a fund’s identity is masked, it is impossible to collect additional information on the human capital of its fund.

(4) 158. R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. managers. I must, therefore, rely on alternative measures of fund performance that serve as proxies for fund returns. The metric of fund performance I use is the fraction of the fund’s portfolio companies that have been exited, either through an IPO or an acquisition. Since venture capital funds can only return cash to their investors when a company is exited, the fraction of companies that exit should be positively correlated with the fund’s return. Previous studies have shown that the correlation coefficient between the fraction of companies exited and the funds liquidation IRR is around 0.6 (e.g., Hochberg, Ljungqvist and Lu, 2007). In the analysis, I find the main empirical results are robust to the use of alternative proxies for fund returns, such as the ability of a fund management team to raise a follow-on fund.3 The first hypothesis that task-specific human capital will be positively related to venture capital fund performance is: Hypothesis 1. Venture capital fund management teams with more task-specific human capital, in the form of a greater fraction of fund managers with past experience (a) being a venture capital fund manager and (b) being a manager in a start-up company will have a greater fraction of portfolio companies that exit. The second type of specific human capital, which I call industry-specific human capital, is also motivated by theories such as Gibbons and Waldman (1999) and Lazear (1995) as well as work by Kletzer (1989) and Neal (1995), which posit that experience in a particular job or industry should enhance a worker’s productivity in that job, regardless of the firm for which he is employed. In the case of industry-specific human capital, the impact on venture capital fund performance will come through experience in tasks and skills learned in the prior industries in which fund managers worked, rather than skills learned directly from the tasks of venture investing and managing start-ups. I choose the three most common industries in which first-time venture capital fund mangers have previously worked – strategy and management consulting, non-venture finance and professional science and engineering – to measure industry-specific human capital. While these three industries are the three most common, it is not clear that each should matter in the same way for fund performance. In particular, experience in strategy and management consulting will likely be correlated with fund managers have more skills in business management and business strategic decisions based on the problems they faced as consultants. However, fund managers with past experience in non-venture finance may be better at helping their portfolio companies obtain alternative sources of financing. Finally, fund managers with prior experience in science and engineering should have an advantage in selecting and advising high-tech companies or companies developing new products. The second primary hypothesis, that industry-specific human capital should be positively related to fund performance, follows: Hypothesis 2. Venture capital fund management teams with a greater fraction of fund managers having worked (a) as strategy and management consultants, (b) in non-venture finance and (c) as industrial engineers or professional scientists will have a greater fraction of portfolio companies that exit. The venture capital markets are particularly well-suited for an investigation of the roles of industry-specific human capital of managers since venture capitalists are active investors who often become involved in the governance and strategic decisions of the companies that their firms finance by sitting on the companies’ boards of directors or in helping their companies identify good managers or advisors. Indeed, past empirical work has documented some of these value-added activities of venture capitalists. (e.g., Bottazzi et al., 2007; Gompers and Lerner, 2001; Gorman et al., 1990; Hellmann and Puri, 2002; Kaplan and Stromberg, 2001, Lerner, 1995; Sapienza, 1992). As argued for hypothesis 2, since most venture capitalists work in non-venture industries prior to becoming venture capitalists, it is likely that experience in these industries is also relevant for venture capital fund management. Experience in solving business and management problems generally is likely a skill learned in strategy and management consulting that may also be valuable as a venture capital fund manager. Likewise, experience in non-venture finance should be particularly important in later stage companies when understanding and enabling a company’s access to alternative sources of non-venture finance become more important.4 As a company grows and develops positive cash flows, it is often useful for it to have alternative sources of financing to venture capital. In addition, as a company gets older, exit via acquisition or IPO becomes more likely. Fund managers with non-venture finance industry experience should be more able to help their portfolio companies find alternative sources of financing as well as line up exit partners, such as investment banks underwriters or potential acquirers. Finally, having experience as a professional scientist or engineer may be particularly valuable for fund managers that focus on hightech investments. Having experience as a professional scientist or engineer should be useful in allowing a fund manager to identify and advise companies that are developing products in areas in which the fund managers have technical expertise. Because experience in non-venture finance and professional science and engineering may be more important for certain kinds of investments (e.g., later stage investments for non-venture finance and high-tech investments for professional scientists and engineers) relative to experience in management and strategy consulting, I posit the third and final hypothesis related to specific human capital in this study: Hypothesis 3. (a) Venture capital fund management teams with a greater fraction of fund managers having worked in nonventure finance will have a greater fraction of portfolio companies that exit when a greater fraction of the companies are laterstage investments. (b) Venture capital fund management teams with a greater fraction of fund managers having worked as professional scientists and engineers will have a greater fraction of portfolio companies that exit when a greater fraction of the companies are high-tech companies.. 3 In unreported results, I also find that my results are robust when I use the fraction of firms that go public as my measure of fund performance. However, the results are stronger when I use the fraction of total exits, both IPOs and acquisitions, consistent with this measure being the more accurate proxy for fund returns. 4 Recent work by Dimov, Shepherd and Sutcliffe (2007) shows that venture capitalists with prior non-venture finance experience, indeed, are less likely to make early-stage investments..

(5) R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. 159. Having defined the two types of specific human capital I measure and described how they are predicted to be related to venture capital fund performance, I turn to predictions about the relation between general human capital and venture capital fund performance. I measure general human capital as education in several fields of study at university – science and engineering, business administration and law.5 These are the three most common educational backgrounds of venture capitalists in my data sets. Having more general knowledge in science and engineering may aid fund managers in selecting and advising their portfolio companies, especially those in high-tech industries, for which having an understanding of the underlying product and technology is important. Having general knowledge in law may aid fund managers in understanding and crafting contractual agreements with their portfolio companies since these venture capitalists will have more knowledge of such contractual arrangements as well as perhaps better legal contacts. Finally, having more general human capital in business may aid fund managers in both selecting and advising portfolio companies. I depart from previous studies which include some fields of study as measures of specific human capital (e.g., Dimov and Shepherd, 2005). While focusing study on an area is specific in one sense compared to studying a broader range of subjects, relative to the industry- and task-specific human capital measures discussed above, having studied an area such as engineering or business or law is more general than having specific work experience in those fields. Thus, I use these educational measures as measures of human capital that is general relative to the measures of work- experience-based specific human capital. I also measure the quality, or reputation, of the general human capital possessed by the venture capital fund managers. Since all venture capitalists in my sample attended university, I measure the reputation of the university they attended. In particular, I measure whether a venture capitalist attended an ivy league university, regarded by many as having a “higher reputation” student body and faculty. 6 Measuring the reputation of educational human capital is another form of testing for the effect of more general human capital on venture capital fund performance. It has been argued (e.g., Chevalier and Ellison, 1999) that having studied at a reputable institution should be correlated with greater general human capital to the extent that better education is received at such institutions.7 I choose to measure reputation of a university as whether or not the university is in the set of “ivy league” universities, traditionally the most competitive to which to gain admission, as a proxy for quality of general human capital quality since my data sample allows me to generate this variable. Thus, the general human capital measures I employ, while measuring particular types of general human capital, are all general relative to the work-related specific human capital measures discussed in Hypotheses 1 to 3. The fourth hypothesis I test is as follows: Hypothesis 4. Venture capital fund management teams with a greater fraction of fund managers having an educational degree (a) in science and engineering, (b) in law (JD), (c) in business (MBA) and (d) from an ivy league university will have a greater fraction of portfolio companies that exit. 3.3. Secondary hypotheses An advantage of my data sample is that I can control for factors other than the human capital of fund managers that may affect a venture capital fund’s performance. In this subsection, I posit five additional hypotheses about the impact other fund-level and market-level factors may have on venture capital fund performance that have been tested in other setting for which I also test in my analysis. First, venture capital funds that are larger and have more resources should be able to invest in more portfolio companies and better enable them to reach exit. In particular, funds with more assets under management will have more money to make followon rounds of financing in order to get their funds to exit in addition to possibly attracting better portfolio companies. The positive impact of fund size, however, may decrease after a certain point since if a fund becomes too large fund managers may begin to make more marginal investments in order to invest the entire fund. Thus, we should expect a positive relationship between the natural logarithm of a fund’s assets and its performance. We should also expect a positive relationship between the number of fund managers and fund performance since more fund managers will provide more human capital in general to the fund and provide more labor to monitor and screen investments. Second, the industrial sectors in which a fund invests may have an impact on the exit rates of its portfolio companies, in particular, if certain industrial sectors experience hot or cold cycles in the public markets over the sample period. To control for these industrial sector effects, I control for the fraction of portfolio companies in a particular industrial sector, with no a priori hypothesis on the directional impact of that sector on the fund’s performance.8 In addition, the degree to which a fund has. 5 All of the venture capitalists in my data sample attended university. I measure general human capital by recording the types of fields studied by venture capitalists. In particular, I measure general human capital in science and engineering by recording whether a venture capitalist has any degree in science and engineering. I measure general human capital in business by recording whether a venture capitalist has an MBA. I measure general human capital in law by recording whether a venture capitalist has a JD. 6 This measure is related to the measure of general human capital employed by Chevalier and Ellison (1999) to examine the impact of university reputation on mutual fund manager performance. In their study, they use the average SAT score of admitted students as a measure of the quality of a university attended by mutual fund mangers. 7 It may also be the case that having attended a high reputation university increases general human capital by providing a fund manager access to a larger network of individuals later on in life rather than by providing more general human capital in the classroom. 8 I use the six major industrial classifications of VentureXpert, one of the primary data sets used in the analysis, to control for the industrial composition of fund’s portfolio companies..

(6) 160. R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. concentrated its investments in a particular industrial sector may also impact the performance of the fund. Both previous theoretical and empirical studies have documented that organizations and managers which specialize in a particular activity exhibit better performance (e.g., Amit, Brander et al., 1998; Dimov and DeClerq, 2006; Gupta and Sapienza, 1992). I, therefore, also control for the degree to which a fund’s portfolio companies are concentrated in industrial sectors by controlling for Herfindahl– Hirschman index of industry shares in the fund. Third, the number of syndicate partners a venture capital fund has may impact the fund’s performance. In particular, if a venture capital fund has a greater number of syndicate partners, the better it may perform, since its larger group of syndicate partners may provide more access to deals that the syndicate partners source and since more syndicate partners can provide second opinions on deal evaluation and monitoring (e.g., Bygrave, 1987; Dimov and DeClerq, 2006). I measure the average number of syndicate partners a venture capital fund has in the rounds in which it first invests in each of its portfolio companies and use this measure as a control variable in the empirical analysis. Fourth, the stage of the companies in which a venture capital fund invests may affect the performance of the fund. In particular, if a fund invests in more companies that are early stage investments, the probability of those early stage companies reaching an exit is lower than if the venture capital fund invested in these companies when they had reached a later stage since new firms are riskier than firms with some track record (e.g., Amit et al., 1998). I control for the fraction of a fund’s portfolio companies are early stage investments in the empirical analysis. Finally, the competitive environment for venture capital deals may affect the performance of venture capital funds. Past studies have shown that the more money that flows into the venture capital industry the higher the price paid for deals and the more marginal are deals that are undertaken (e.g., Gompers and Lerner, 2000). Thus, we would expect an inverse relation between the amount of competition in the venture capital industry as measured by the amount of venture capital raised by all funds and the likelihood that a venture capital fund’s companies are exited. By controlling for these five additional drivers of venture capital fund performance, I am better able to test the four primary hypotheses on human capital in the analysis below. 4. Data I use two data sources in my analysis. First, I use the Thomson Financial/Venture Economics VentureXpert data set to identify venture capital funds, the portfolio companies in which they invest and the outcomes of those investments. The basic unit of observation in VentureXpert is a financing deal, or round. VentureXpert records the identities of the participating venture capital funds in each round as well as the portfolio company receiving the investment. The data set also records the outcomes of the portfolio companies receiving venture capital, including whether they went public, were acquired, were shut down, or are still active investments. Second, I use a hand-collected data set containing the work and educational histories of the individual venture capitalists managing the venture capital funds identified in VentureXpert. I use this database to form the measures of top management team human capital which I use to test the primary hypotheses posited in Section 3. 4.1. Sample selection I restrict my sample of venture capital funds along the following dimensions: First, I only include funds whose managing firms are based in the United States and which are classified as “Private Equity Firms Investing Own Capital”. The impact of venture capitalists’ human capital on the performance of first-time venture capital funds connected to banks, corporations or governments may be different than its impact on funds managed by independent investment firms due to differing incentives and resources of being connected to a larger organization. Second, I restrict the sample to include only funds that were raised between 1980 and 1998. The typical life span of a venture capital investment is around three to five years and the typical life span of a venture capital fund is around ten years. Funds make most of their investments within three years of the fund’s start. Therefore, funds that were started after 1998 may not have had enough time to exit their investments, making comparisons between the performance of these younger funds and older funds difficult. Third, I select only funds that are classified as venture capital funds and exclude those classified as buyout funds. I focus only on venture capital funds, rather then both venture capital and buyout funds, since the types of investments these two types of fund make can be very different and thus skill sets that are likely required for fund success will vary between venture capital and buyout funds. Fourth, I restrict the sample to include only venture capital funds that invested in five or more portfolio companies and which have non-missing size information. Imposing these sample selection criteria leaves a sample of 1184 venture capital funds. Finally, I restrict my sample to include only first-time funds. I define a first-time fund if it is the first fund reported as managed by a venture capital firm and has a vintage year of no more than two years after the founding date of the managing venture capital firm. This final sample selection criterion leaves a sample of 318 first-time venture capital funds. Table 1 Panel A presents a longitudinal view of this sample of 318 venture capital funds. It reports average statistics on fund size, number of portfolio companies per fund, and the fraction of these portfolio companies that exit via IPO or acquisition. The funds contributing to an average in a particular year are the funds that were closed in that year, or have that year as their “vintage year”. The sample average fund size is 52 million year 2000 dollars and the average number of portfolio companies in which a fund invests is 22. The average fraction of funds’ portfolio companies that exit, via IPO or acquisition, is 0.54. The average size of a firsttime fund is smaller than the average venture capital fund size over the sample period, $52 million versus $82 million, though the.

(7) R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. 161. Table 1 Venture capital fund summary statistics. 1980– 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1998 Panel A — First-time U.S. venture capital funds raised between 1980 and Fund size 51.9 1.8 38.4 39.3 32.5 40.0 35.6 (millions 2000 $) Companies 22.0 44.2 34.6 31.3 26.5 25.3 24.6 per fund 0.54 0.65 0.60 0.55 0.62 0.58 0.58 Fraction of companies exited per fund Number of funds 318 12 25 23 24 31 17 Panel B- First-time Fund size (millions 2000 $) Companies per fund Fraction of companies exited per fund Number of funds. 1998 28.2. 32.2. 41.5. 17.5. 32.6. 5.9. 46.7. 42.1. 59.1. 66.3. 65.2. 85.7. 101.8. 17.9. 13.9. 18.1. 14.7. 10.5. 5.0. 12.3. 11.2. 19.3. 21.9. 18.3. 16.6. 12.9. 0.62. 0.55. 0.49. 0.60. 0.57. 1.00. 0.55. 0.54. 0.46. 0.51. 0.46. 0.48. 0.42. 20. 19. 8. 10. 4. 1. 3. 6. 7. 20. 26. 35. 27. 43.2. 59.1. 69.6. 62.6. 92.9. 104.5. U.S. venture capital funds raised between 1980 and 1998 with collected venture capitalist histories 61.9 83.9 66.9 54.6 40.6 36.5 55.3 28.7 40.4 49.5 17.4 35.3 5.9 61.3 22.9. 78.3. 43.7. 35.5. 32.8. 30.2. 26.7. 20.9. 13.9. 20.0. 15.7. 11.3. 5.0. 15.0. 11.8. 19.3. 22.7. 19.3. 16.6. 12.9. 0.53. 0.68. 0.65. 0.58. 0.62. 0.55. 0.64. 0.60. 0.54. 0.60. 0.58. 0.55. 1.00. 0.68. 0.52. 0.46. 0.52. 0.47. 0.48. 0.40. 222. 3. 8. 11. 14. 19. 7. 16. 13. 6. 7. 3. 1. 2. 5. 7. 19. 25. 32. 24. The sample includes venture capital funds based in the United States and raised between 1980 and 1998 with at least five portfolio companies and non-missing fund size managed by independent venture firms as recorded in VentureXpert. Average fund size, portfolio companies per fund, and portfolio company exits per fund are reported for the full sample period and by fund vintage year.. average number of companies in which first-time funds invest and the fraction of those companies that are exited is closer to the full sample averages based on the 1184 venture capital funds, 22 versus 23 and 0.54 versus 0.56. 4.2. Venture capitalist biographical information I use the names of the executives recorded in VentureXpert as working for venture capital firms as the starting point for identifying the venture capitalists who raise and manage the first- time venture capital funds in my sample. I am interested in identifying the individuals responsible for making decisions about in which portfolio companies the venture capital fund invests, with whom to syndicate investments, and how much money to invest in each portfolio company. There are two challenges to identifying fund managers from the set of individuals listed in VentureXpert as working for each venture capital firm. The first challenge is distinguishing fund managers, i.e. individuals with the decision-making ability in the fund, from individuals who are primarily engaged in support activities. The second challenge is identifying the individuals who were fund managers during the period over which a first-time fund was managed. I take a two-step approach identifying which individuals are first-time venture capital fund managers. First, I check if any of the individuals identified in VentureXpert served as board members for any of the fund’s portfolio companies. If they have, I classify these individuals as fund managers, since serving as a board member and monitoring and advising portfolio companies is the role of a fund manager; the venture capitalist or fund manager who is the “lead,” or responsible decision maker, for a deal often takes a board seat on the portfolio company. Second, I classify individuals with the same job title as the individuals holding board seats as fund managers as well. It is important to note here that I exclude peripheral individuals, such as “entrepreneurs-in-residence” and “venture partners” who are people who are connected to the fund but do not act as fund managers. Such people may be called upon to serve as CEO of a portfolio company or, provide occasional advice to the fund managers or add advertising value to the fund but who do not engage in active management of the fund. This screening process identifies a set of individuals who at some point may have been fund managers for one of the funds managed by a venture capital firm. I further identify the individuals who were fund managers of the first fund raised and managed by the venture capital firm by collecting information on the date an individual joined the venture capital firm. At this point, I also supplement the names included in VentureXpert with additional names that I find in my data collection effort. In some cases, founding partners of a venture capital firm are actually listed as “Founding Partners” or “Founders”. For each of the venture capitalists I identify as first-time fund managers, I further hand- collect information on the universities they attended, the degrees they attained, including major field of study, and the firms for which they worked and the positions held with these firms. I also collect information on the dates during which the individual held these positions. To collect this data, I first visit the websites of the managing firms of the first-time venture capital funds in my sample, if they are still in existence. For individuals who are still working for the venture capital firms, I collect information from the biographies listed on these websites. I also use a biographical search engine called ZoomInfo to collect additional information on my set of first- time fund managers. ZoomInfo collects information on individuals working for companies in the U.S. and Canada by crawling the websites of these companies and caching appearances in old WebPages over the past several years of individuals included in its database of professionals. I search both on the names of any individuals listed in VentureXpert and on the name of the venture capital fund and managing firm. ZoomInfo records biographical information such as schooling and work history..

(8) 162. R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. I am able to collect biographical information for the founding management teams for 222 of my sample of 318 first-time venture capital funds. Table 1 Panel B presents longitudinal information for my sub-sample of first-time funds for which I have venture capitalist biographical information. The averages in Panel B are similar to those in Panel A, although there are a larger a number of first-time funds in the earlier part of the sample period with missing biographical information than in the later part of the sample period. This is due to the fact that more venture capital managing firms that started funds in the 1980s are no longer in existence and it is therefore more difficult to collect information on the individuals who managed these funds. The first-time funds with collected biographical information are slightly larger than the average fund. To ensure that the main results in the paper are not driven by survivorship bias, I repeat the regression analysis detailed below on the sub-sample of firsttime funds raised in the 1990s, when there are fewer missing first-time funds and find that the main findings are robust to this analysis. 4.3. Summary statistics Table 2 summarizes the human capital measures based on the hand-collected biographical data for the 482 individual venture capitalists managing the final sample of 222 first-time venture funds. The first column of Table 2 reports average statistics across all 482 venture capitalists. Each of the variables in the first column of Table 2 are dummy variables which equal one if an individual venture capitalist possesses a particular characteristic. The second column of Table 2 reports average statistics by first-time venture capital fund. Each of the variables beginning with “Frac” measures the fraction of venture capital fund managers in a given fund who possess a particular human capital characteristic. Appendices A, B and C contain definitions of all of the variables summarized in Table 2. The variables summarized in the second column of Table 2 and defined in Appendices B and C are the variables that will be used in the regression analysis below. The variables defined in Appendix B correspond to the primary hypotheses (Hypotheses 1 to 4) described in Section 3.1. The variables defined in Appendix C correspond to the secondary hypotheses described in Section 3.2. Focusing on the first column of Table 2, we see most of the general capital of individual venture capitalists comes in the form having studied science and engineering, business administration and from having attended a high reputation university. 58% of individual venture capitalists have MBAs. 33% studied engineering or science in college and 37% attended an ivy league university. A large fraction of those who attended an ivy league university also got their MBAs from an ivy league university. 19% of venture capitalists attended Harvard and almost all of them also got an MBA there, or 16%. 14% of venture capitalists attended Stanford, but only a little over half, 9%, also got their MBAs there. Only a small percentage have PhDs in science or law degrees, 7 and 8% respectively. Turning to the individual venture capitalist work histories in the first column of Table 2, we see that the largest fraction of firsttime fund venture capitalists have task-specific human capital from having prior venture capital investing experience, about 44%. About 15% have task-specific human capital from having previously founded and managed start-up companies. The next largest fraction of venture capitalists have industry-specific human capital from having previously worked in non-venture finance, 29%. 16% of venture capitalists have industry-specific human capital from having worked as management or strategy consultants and 9% from having worked as professional engineers. When we focus on the fund-level averages across first-time fund management teams in the second column of Table 2, we see that the fractions of fund management teams that have a particular characteristic roughly line up with the fractions of individuals venture capitalists with a particular characteristic. This implies that venture capitalists with a certain characteristic are not disproportionately teaming up together in smaller or larger funds. The average top management team size is 2.17, indicating that most first-time venture capital funds are lean organizations.9 The second column of Table 2 also summarizes the control variables mentioned in the discussion of secondary hypotheses in Section 3.2. The average number of syndicate partners in the first round a fund invests is 3.46, and the average HHI concentration of portfolio companies across the six VentureXpert industries is 0.41. Table 3 presents correlation matrices for the individual venture capitalist human capital variables in Panel A and for the variables measuring the fraction of venture capital fund managers with a particular human capital characteristic in Panel B. The correlation matrix in Panel A tells us which venture capitalists tend to possess different types of human capital simultaneously. The correlation matrix in Panel B tells us which types of venture capitalists tend to team up when starting first-time funds. There are several interesting facts that emerge from Table 3. First, focusing on Panel A, there appear to be “types” of individual venture capitalists in terms of their educational and work histories. In particular, venture capitalists who have MBAs also tend to also have attended ivy league universities, in particular Harvard University, and tend to have work experience in both venture capital and non-venture finance. The second type of venture capitalist studies science and engineering and is more likely to have been a manager at a start-up and to not have attended an ivy league university. Next, when we examine Panel B of Table 3, we see that venture capitalists tend to form fund management teams with other venture capitalists who possess similar characteristics. Venture capitalists of the first “type”, i.e. those with an MBA, ivy league education, and venture capital and finance experience, tend to form teams with other venture capitalists with these. 9 The average number of partners is smaller than in several previous studies because I focus on first-time funds, which typically have between 1 and 4 founding partners. Approximately 20% of the first-time fund management teams consist of only one person. All of the results in the paper are robust to eliminating this sub-sample of first-time funds with only one fund manager from the analysis..

(9) R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. 163. Table 2 Venture capitalist human capital measures — summary statistics. Individual venture capitalist variables Individual educational history variables SciEngDegree SciEngPhD MBA JD IvyDegree HarvardDegree StanfordDegree IvyMBA HarvardMBA StanfordMBA Individual work history variables PastVC PastStartupExec PastConsultant PastFinance PastEngineer Number of venture capitalists Fund-level variables Fund management team educational history variables FracSciEngDegree FracSciEngPhD FracMBA FracJD FracIvyDegree FracHarvardDegree FracStanfordDegree FracIvyMBA FracHarvardMBA FracStanfordMBA Fund management team work history variables FracPastVC FracPastStartupExec FracPastConsultant FracPastFinance FracPastEngineer Fund-level and market-level control variables Other fund-level variables Log (Fund size) Number of founders AvgFRSyndicate IndustryHHI FracSeedStage FracBiotech FracComm FracComp FracMed FracElec FracNontech Market-level variables Log(Lagged fund inflows) Fund year 1980–1984 Fund year 1985–1989 Fund year 1990–1994 Fund year 1995–1998 Number of venture capital funds. 0.39 0.07 0.58 0.08 0.37 0.19 0.14 0.24 0.16 0.09 0.44 0.15 0.16 0.29 0.09 482. 0.39 0.06 0.59 0.08 0.37 0.20 0.12 0.25 0.17 0.08 0.42 0.15 0.17 0.31 0.09. 3.56 2.17 3.46 0.42 0.51 0.07 0.17 0.39 0.15 0.07 0.14 8.44 0.25 0.22 0.08 0.45 222. The sample includes first-time U.S venture capital funds raised between 1980 and 1998 and managed by independent venture firms identified in VentureXpert and with collected venture capitalist histories. Variables in the first column are dummy variables equal to one if a venture capitalist possesses a particular characteristic. Variables beginning with qFracq in the second column measure the fraction of venture capitalists managing a fund who possess a particular characteristic. Please see Appendix A, B and C for more detailed variable definitions. Means are reported.. characteristics. Venture capitalists of the second “type”, i.e. those with a science and engineering background, non-ivy league education and start-up experience, tend to form teams with other similar venture capitalists. Thus, while there is heterogeneity in the types of venture capitalists who start first-time funds, it appears that venture capitalists with a particular set of human capital attributes seek out other venture capitalist with the same attributes. One might interpret this fact as evidence that.

(10) 164. R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. Table 3 Correlation matrices. 1. 2. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 1.00 0.63 0.02 0.75 0.57 0.01 0.16 - 0.12 0.07 0.12 0.01. 1.00 - 0.10 0.68 0.90 - 0.11 0.14 - 0.07 0.06 0.07 - 0.10. 1.00 - 0.14 - 0.12 0.78 0.17 0.05 0.03 0.00 0.19. 1.00 0.75 - 0.18 0.09 - 0.08 0.08 0.09 - 0.01. 1.00 - 0.13 0.08 - 0.04 0.03 0.05 - 0.08. 1.00 0.17 - 0.01 0.09 0.08 0.13. 1.00 - 0.09 - 0.06 - 0.01 0.05. 1.00 - 0.06 - 0.17 0.06. 1.00 0.00 0.04. 1.00 - 0.03. 1.00. Panel B — Correlations between top management team human capital variables 1. FracSciEngDegree 1.00 2. FracSciEngPhD 0.36 1.00 3. FracMBA -0.02 - 0.21 1.00 4. FracJD -0.18 - 0.12 - 0.17 1.00 5. FracIvyDegree 0.04 0.03 0.28 - 0.01 1.00 6. FracHarvardDegree -0.06 - 0.10 0.28 0.10 0.68 7. FracStanfordDegree 0.14 0.03 0.17 - 0.06 0.05 8. FracIvyMBA -0.01 - 0.12 0.47 - 0.06 0.79 9. FracHarvardMBA -0.01 - 0.09 0.38 - 0.01 0.63 10. FracStanfordMBA 0.09 - 0.01 0.26 - 0.05 - 0.02 11. FracPastVC -0.03 - 0.01 0.15 - 0.06 0.18 12. FracPastStartupExec 0.09 - 0.09 - 0.24 0.16 - 0.11 13. FracPastConsultant 0.03 - 0.07 0.05 0.12 0.16 14. FracPastFinance -0.08 - 0.06 0.13 0.01 0.12 15. FracPastEngineer 0.40 0.22 - 0.10 - 0.08 - 0.02 16. Fund portfolio company exit fraction 0.15 0.03 - 0.09 - 0.05 0.21. 1.00 - 0.05 0.70 0.90 - 0.13 0.19 - 0.13 0.15 0.15 - 0.10 0.12. 1.00 - 0.08 - 0.03 0.81 0.16 0.11 0.03 - 0.01 0.23 0.04. 1.00 0.79 - 0.15 0.13 - 0.11 0.12 0.08 - 0.08 0.13. 1.00 - 0.12 0.12 - 0.10 0.10 0.12 - 0.11 0.14. 1.00 0.16 0.07 0.12 0.02 0.16 0.04. 1.00 - 0.16 - 0.09 - 0.11 0.06 0.14. 1.00 0.01 - 0.22 0.05 0.09. 1.00 0.04 0.03 0.12. 1.00 - 0.01 0.06. 1.00 0.03. Panel A — Correlations between individual venture capitalist 1. SciEngDegree 1.00 2. SciEngPhD 0.33 1.00 3. MBA 0.03 - 0.26 4. JD -0.17 - 0.08 5. IvyDegree 0.07 - 0.01 6. HarvardDegree -0.03 - 0.11 7. StanfordDegree 0.19 0.02 8. IvyMBA 0.00 - 0.13 9. HarvardMBA 0.00 - 0.09 10. StanfordMBA 0.13 - 0.08 11. PastVC 0.03 0.00 12. PastStartupExec 0.13 - 0.07 13. PastConsultant 0.00 - 0.02 14. PastFinance -0.12 - 0.10 15. PastEngineer 0.35 0.15. 3. 4. 5. human capital variables. 1.00 - 0.16 0.34 0.31 0.16 0.49 0.37 0.26 0.18 - 0.16 0.07 0.15 - 0.01. 1.00 0.01 0.08 - 0.05 - 0.07 - 0.02 - 0.03 - 0.02 0.05 0.02 - 0.05 - 0.09. Panel A reports the correlation matrix for the individual venture capitalist human capital variables. Please see Appendix A for definitions of these variables. Panel B reports the correlation matrix for the fund-leveltop management team human capital variables. Please see Appendix B for definitions of these variables.. specialization in a particular type of human capital in a fund management team is deemed more valuable than diversification across different types of human capital. The final row of Table 3 Panel B presents the correlations between the independent variable of interest, the fraction of a first-time fund’s portfolio companies that exit, and the human capital variables, foreshadowing the main findings below. 5. Empirical analysis and results I now turn to the empirical tests of the hypotheses posited in Section 3. I regress the fraction of portfolio companies in which a fund invests that exit, my proxy for fund returns, on fund-level top management team human capital measures detailed in Section 3.2 and other fund-level and market-level controls detailed in Section 3.3 as in Eq. (1). n X. FracExiti = b0 +. m X. bj FraVCCharj;i + j=1. p X. bk Xk;i + k=1. bh Zh;t + ei :. ð1Þ. h=1. The subscript i indexes each fund in the sample; the subscript t indexes each fund-year in the sample. The main variables of interest in testing Hypotheses 1 to 4 are the FracVCCharj,i variables that are summarized in the second column of Table 2 and defined in Appendix B. The variables Xk,i are fund-level controls – the number of fund managers, the natural logarithm of the size of the fund (in constant year 2000 millions of dollars), the Herfindahl–Hirschman index of a fund’s portfolio companies by industry, the fraction of a fund’s portfolio companies that are early stage investments, and the fraction of a fund’s portfolio companies that are in each of the six VentureXpert industries. The variables Zh,t are time-varying marketlevel controls — the lagged natural logarithm of venture capital fund inflows per year (in constant year 2000 millions of dollars) and dummy variables for whether a fund was raised between 1985 and 1989, between 1990 and 1994 or between 1995 and 1998..

(11) R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. 165. Table 4 First-time fund performance regressions. Predicted sign Fund-level educational history variables FracSciEngDegree + FracMBA. +. FracJD. +. FracIvyDegree. +. Fund-level work history variables FracPastVC. +. FracPastStartupExec. +. FracPastConsultant. +. FracPastFinance. +. FracPastEngineer. +. Other fund-level variables Log(Fund size). +. Number of managers. +/-. AvgFRSyndicate. +/-. IndustryHHI. +. FracSeedStage. -. FracBiotech. +/-. FracComm. +/-. FracComp. +/-. FracMed. +/-. FracElec. +/-. Market-level variables Log(Lagged fund inflows). -. Fund year 1985-1989. +/-. Fund year 1990-1994. +/-. Fund year 1995-1998. +/-. Constant N Adjusted R2. (1). (2). 0.06 (1.90) - 0.05 (- 1.92) - 0.05 (- 1.03) 0.04 (0.86). ⁎⁎. 0.07 (2.58) 0.12 (3.19) 0.10 (2.48) 0.00 (0.09) - 0.06 (- 2.15). ⁎⁎⁎. 0.02 (2.29) 0.00 (- 0.42) 0.03 (5.04) - 0.06 (- 1.25) - 0.13 (- 2.17). ⁎⁎. - 0.07 (- 7.11). 1.01 (10.07) 222 0.32. ⁎⁎. ⁎⁎⁎ ⁎⁎⁎. ⁎⁎. ⁎⁎⁎. ⁎⁎. ⁎⁎⁎. ⁎⁎⁎. (3). 0.04 (1.63) - 0.04 (-1.75) - 0.04 (-0.87) 0.03 (0.71). ⁎. 0.05 (1.95) 0.11 (3.36) 0.09 (2.21) 0.00 (-0.08) - 0.05 (-1.79). ⁎⁎. 0.02 (2.44) 0.00 (-0.49) 0.03 (3.55) - 0.06 (-1.29) - 0.16 (-2.67) 0.23 (2.09) 0.10 (1.91) 0.10 (1.76) 0.02 (0.20) 0.04 (0.25). ⁎⁎⁎. - 0.08 (-7.46). ⁎⁎⁎. 1.06 (10.29) 222 0.32. ⁎⁎. ⁎⁎⁎ ⁎⁎. ⁎. ⁎⁎⁎. ⁎⁎⁎ ⁎⁎ ⁎ ⁎. ⁎⁎⁎. 0.03 (0.94) - 0.04 (- 1.70) - 0.05 (- 0.93) 0.03 (0.71). ⁎. 0.05 (1.87) 0.11 (3.57) 0.09 (2.42) - 0.01 (- 0.26) - 0.04 (- 1.10). ⁎. 0.03 (2.85) 0.00 (- 0.35) 0.03 (3.91) - 0.05 (- 1.04) - 0.17 (- 2.79) 0.26 (2.55) 0.15 (2.80) 0.16 (2.42) 0.06 (0.47) 0.07 (0.53). ⁎⁎⁎. - 0.06 (- 4.65) 0.06 (2.46) - 0.01 (- 0.20) - 0.05 (- 1.37) 0.80 (5.97) 222 0.33. ⁎⁎⁎. ⁎⁎⁎ ⁎⁎. ⁎⁎⁎. ⁎⁎⁎ ⁎⁎ ⁎⁎ ⁎⁎. ⁎⁎. ⁎⁎⁎. The sample includes first-time U.S. venture capital funds raised between 1980 and 1998 and managed by independent venture firms identified in VentureXpert and with collected venture capitalist biographical histories. The dependent variable is the fraction of a fund's portfolio companies that exit, either via an IPO or an acquisition. Reported regression coefficients are estimated using OLS . T-statistics adjusted for clustering by fund-year are reported in parentheses. ⁎⁎⁎, ⁎⁎, and ⁎ indicate one-tailed significance for coefficients with unambiguously predicted signs (i.e., + or -) and two-tailed significance for coefficients with ambiguously predicted sign (i.e., +/-) at the 1%, 5%, and 10% levels, respectively. Please see Appendices B and C for definitions of the variables reported in this table.. I also estimate regressions with the same independent variables as in Eq. (1) but with the log odds ratio, LN (PercentExiti /(1 — PercentExiti)) , as the dependent variable. Doing so constrains the predicted values to range between zero and one. The results are very similar to the results from estimating Eq. (1). Since the vast majority of values PercentExiti takes are within the zero to one interval, rather than being equal to zero or one, the predicted values that are generated from estimating Eq. (1) are all between zero.

(12) 166. R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. and one. I choose to report estimates of Eq. (1), rather than estimates using the log odds ratio, since interpretation of the coefficients as marginal effects is more straightforward. 5.1. Fund performance regressions Table 4 reports regression results for Eq. (1) estimated using the sample of 222 first-time VC funds. T-statistics adjusted for clustering at the fund-year level are reported in parentheses. The specification in column 1 of Table 4 regresses the fraction of a fund’s portfolio companies that are exited on the control variables that are directly related to Hypotheses 1, 2 and 4. The last two specifications in columns 2 and 3 or Table 4 add additional fund-level and market-level control variables that do not directly relate to Hypotheses 1, 2 and 4 and to the secondary hypotheses in Section 3.3 but which control for possible industry-wide and time variation in the performance of venture capital funds. In particular, the second specification adds control variables which measure the fraction of a fund’s portfolio companies that are in each of the six VentureXpert industry categories. The last specification adds dummy variables for the time periods in which the first-time funds were raised. Examining the estimated coefficients in Table 4, we see that the coefficients on the fund management team human capital variables are similar across all three specifications. Controlling for additional industrial sector and time variation in the likelihood that a portfolio company will go public or be acquired in the second and third specifications does not affect the estimated coefficients on the top management team human capital measures. However, because the number of control variables in the last two specifications gets large, at 20 and 23 covariates respectively, relative to the sample size of 222 funds, the statistical significance of several of the variables is reduced. The coefficients in Table 4 strongly support Hypothesis 1, that task-specific human capital gained from previously having been both a venture investor and from previously managing a start-up should be positively related to a venture capital fund’s portfolio company exit fraction. We see that the coefficients on the variables measuring the fraction of managers with past venture investing experience and the fraction of managers with past experience as an executive at start-up are positive and statistically significant. If a top management team has members who all have past venture investing experience the fund’s predicted exit fraction increases by around 0.06 relative to a fund whose management team has no members with past venture investing experience. If the fraction of a fund’s management team with experience as executives in a start-up increases from zero to one, the fund’s exit fraction increases by 0.12. Hypothesis 2 is partially supported in Table 4 by the coefficient on the fraction of managers with management consulting experience. An increase in the fraction of fund managers with past experience in consulting from zero to one, increases the fund’s exit fraction increases by around 0.10, consistent with Hypothesis 2(a). However, industry-specific expertise gained by working in non-venture finance has no impact on fund performance, inconsistent with Hypothesis 2(b), and industry-specific expertise gained by working as an industrial engineer or professional scientist has a negative impact on fund performance, though the coefficient on this variable becomes insignificant when we control for time variation in fund performance in Column 3, inconsistent with Hypothesis 2(c). Recall that Hypothesis 3 in Section 3.2 posits that industry-specific human capital in nonventure finance and professional science and engineering should matter more for funds which specialize more in later stage and high-tech investments respectively. In the next subsection, I explore whether this is the case in the data. Table 4 also presents mixed evidence in support of Hypothesis 4, that more general human capital positively predicts fund performance. We see that the coefficients on the educational history variables are fairly weak statistically relative to the work history variables. The amount of general human capital as measured by having a degree form an ivy league university does not significantly predict the fraction of portfolio company exits. Having more general human capital in science and engineering, as measured by having a degree in the area, positively predicts the fraction of portfolio company exits in the first specification; an increase in the fraction of managers with a science or engineering degree from zero to one raises the exit fraction by 0.06.10 However, the statistical significance of the coefficient on this measure of general human capital is eliminated when I control for industrial sector and time variation in fund performance. The strongest and most robust coefficient on the educational history variables is the coefficient on the fraction of fund managers with an MBA. However, contrary to Hypothesis 4 which posits that more general human capital should be positively correlated with the fraction of portfolio company exits, the coefficient on the fraction of managers with an MBA is negative.11 An increase in the fraction of fund managers with an MBA from zero to one decreases the fraction of portfolio companies that exit by around 0.05. Thus, the evidence on the impact of more general human capital on venture capital fund performance is mixed, with only support for the hypothesis that venture capital fund management teams with more general human capital in the form of science and engineering education manage better performing funds (i.e., Hypothesis 4(a)). Overall, the analysis in Table 4 suggests that measures of task- and industry- specific human capital are stronger predictors of first-time venture capital fund performance than are measures of general human capital. In the following subsection, I further explore Hypothesis 3, that measures of industry-specific human capital may matter more for certain types of venture capital fund investments by interacting the fraction of fund managers with industry-specific human capital in non-venture finance and science and engineering with the fraction of a fund’s portfolio companies that are later stage and high tech, respectively. But before doing so, I briefly summarize the coefficients of relevance to the secondary hypotheses 10 Controlling for whether fund managers have an advanced degree (e.g., PhD) in science and engineering does not add any explanatory power above the base effect of having any university degree in science and engineering. 11 When I control for the impact of on fund performance when fund managers have MBAs from “high reputation” institutions, i.e., ivy league universities, Harvard and Stanford, I find that there is no impact of having an MBA on fund performance when the MBA is from a “high reputation” institution..

(13) R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. 167. Table 5 First-time fund performance regressions — interacting science and finance experience with fund investments type. Predicted sign Fund-level educational history variables FracSciEngDegree. +. FracSciEngDegree⁎FracHightech. +. FracMBA. +. FracJD. +. FracIvy. +. Fund-level work history variables FracPastVC. +. FracPastStartupExec. +. FracPastConsultant. +. FracPastFinance. +. FracPastFinance⁎FracLaterStage. -. FracPastEngineer. +. FracPastEngineer⁎FracHightech. +. Other fund-level variables FracLaterStage. +. FracHightech. +/-. N Adjusted R2. 222 0.34. - 0.06 (-1.09) 0.23 (2.20) - 0.04 (-1.90) - 0.04 (-0.81) 0.03 (0.71). ⁎⁎ ⁎⁎. 0.05 (1.87) 0.12 (4.05) 0.11 (2.55) - 0.08 (-1.89) 0.14 (1.92) - 0.04 (-2.36) 0.09 (2.59). ⁎⁎. 0.55 (2.01) 0.13 (1.54). ⁎⁎. ⁎⁎⁎ ⁎⁎⁎ ⁎⁎ ⁎⁎ ⁎⁎ ⁎⁎⁎. The sample includes first-time U.S. venture capital funds raised between 1980 and 1998 and managed by independent venture firms identified in VentureXpert and with collected venture capitalist biographical histories. The dependent variable is the fraction of a fund's portfolio companies that exit, either via an IPO or an acquisition. Reported regression coefficients are estimated using OLS. T-statistics adjusted for clustering by fund-year are reported in parentheses. ⁎⁎⁎, ⁎⁎, and ⁎ indicate one-tailed significance for coefficients with unambiguously predicted sign (i.e., + or -) and two-tailed significance for coefficients with ambiguously predicted sign (i.e., +/-) at the 1%, 5%, and 10% levels, respectively. Other variables included in the regression but not reported are log(fund size), number of managers, average first round syndicate size, fund industry HHI, fraction of portfolio companies that are seed stage, log(lagged fund inflows), and a constant. Please see Appendices B and C for definitions of the variables reported in this table.. posited in Section 3.3. As posited and consistent with prior studies, I find that larger funds have a greater fraction of portfolio company exits. The coefficient on Log(Fund size) is positive and statistically significant. Having more fund managers does not positively, or negatively, predict fund performance. The hypothesis that funds that specialize more in certain industrial sectors should have more exits, is not supported in this data sample, unlike in previous studies; the coefficient on IndustryHHI is insignificant and negative. I also do not find evidence that funds with more syndicate partners in the rounds in which they first invest in companies, contrary to prior empirical work. The coefficient on AvgFRSyndicate is positive and significant; an increase in the number of syndicate partners of one increases the fraction of portfolio companies that exit by 0.03. Finally, as posited I find that funds with a larger fraction of early stage investments have lower exit fractions and funds started in times when there has been a lot of venture capital raised have lower exit fractions, consistent with prior work. 5.2. Interacting industry-specific human capital with fund portfolio company composition We have seen that measures of task-specific human capital, both prior venture capital investing experience and prior experience managing a start-up, are strongly and positively related to more first-time fund portfolio company exits. However, support of Hypothesis 2 that industry-specific human capital should be positively correlated with more portfolio company exits was mixed, with support for only Hypothesis 2(a). Only funds with more managers with past management and strategy consulting have significantly greater fractions of portfolio company exits. As posited in Hypothesis 3, it is likely that industry-specific human capital in science and engineering and in non-venture finance will be more important for certain kinds of portfolio companies. In particular, science and engineering work experience will likely matter more when more a fund’s investments are high-tech investments. Likewise, non-venture finance experience will likely matter more when more of a fund’s investments are later stage investments when contacts with alternative sources of finance become more critical for portfolio company success..

(14) 168. R. Zarutskie / Journal of Business Venturing 25 (2010) 155–172. Table 6 Probability of raising a follow-on fund. Dependent variable. Predicted sign. Dummy = 1 if fund raises a follow-on fund (1). Fund-level educational history variables FracSciEngDegree. +. FracMBA. +. FracJD. +. FracIvyDegree. +. 0.06 (0.17) - 0.09 (-0.28) - 0.55 (-1.07) 0.46 (1.36). Fund-level work history variables FracPastVC. +. FracPastStartupExec. +. FracPastConsultant. +. FracPastFinance. +. FracPastEngineer. +. Other fund-level variables Log(Fund Size). +. Number of founders. +/-. AvgFRSyndicate. +/-. IndustryHHI. +. FracSeedStage. -. Market-level variables Log(Lagged fund inflows). -. Constant N Pseudo R2. 0.43 (1.10) 1.13 (1.72) 0.94 (2.30) 0.11 (0.28) - 0.53 (2.02). 0.03 (0.17) 0.38 (2.52) 0.04 (0.59) 0.26 (0.31) - 0.36 (-0.54). - 0.42 (-1.87) 3.82 (2.00) 222 0.19. ⁎ ⁎⁎. ⁎. ⁎⁎. ⁎ ⁎⁎. The sample includes first-time U.S venture capital funds raised between 1980 and 1998 and managed by independent venture firms identified in VentureXpert and with collected venture capitalist biographical histories. The dependent variable is a dummy equal to one if a first-time fund raises a follow-on fund. The specification is a probit models estimated using maximum likelihood. Coefficients and z-stats adjusted for clustering by fund-year are reported. ⁎⁎⁎, ⁎⁎, and ⁎ indicate one-tailed significance for coefficients with unambiguously predicted sign (i.e., + or -) and two-tailed significance for coefficients with ambiguously predicted sign (i.e., +/-) at the 1%, 5%, and 10% levels, respectively. Please see Appendices B and C for definitions of the variables reported in this table.. I interact the FracPastEngineer variable with a variable called FracHighTech, which measures the fraction of the first-time fund’s portfolio companies that are in the computer, biotech, medical, communications or electronics industries. According to Hypothesis 3, we should expect the coefficient on this interaction to be positive and statistically significant. Likewise, I interact the FracPastFinance variable with a variable called FracLaterStage, which measures the fraction of the first-time fund’s portfolio that are later-stage investments. According to Hypothesis 3, we should also expect the coefficient on this interaction to be positive and statistically significant. I also interact the variable SciEngDegree with the variable FracHighTech to test whether having a degree in science and engineering has a stronger positive effect on the fraction of portfolio company exits when more of the portfolio companies are in high-tech industries. I re-estimate Eq. (1) including these three new interactions terms. The estimated coefficients and t-statistics, adjusted for clustering by fund-year, are reported in Table 5. The coefficients on both the FracSciEngDegree⁎FracHighTech and FracPastEngineer⁎FracHighTech are positive and statistically significant. This indicates that when fund managers have both more general and industry-specific human capital in science and engineering, their positive influence on fund company exits is more important when more of those investment are in high-tech fields in which such expertise is likely more valuable. The coefficient on FracPastEngineer is still negative and statistically significant, reflecting the negative coefficient we also observed for this variable in Table 4. The raw effect on a fund’s exit fraction of having an increase in industry-specific human capital in science and engineering is -0.04. However, if all of the fund’s portfolio companies are high-tech companies, this impact on fund exit fraction becomes positive at 0.05 (i.e., -0.04 plus the coefficient on the interaction FracPastEngineer⁎FracHighTech of 0.09). The significant.

Mostrar más

Nuevo documento

Ejercicios propuestos I
son adem´as tangentes a los lados de un ´angulo agudo por la parte interior del mismo, por lo que sus centros estar´an situados sobre la bisectriz de dicho ´angulo.. Halla el valor del
Números racionales e irracionales
Æ Números IRRACIONALES: Los números irracionales, al contrario que los racionales, no pueden expresarse como cociente de números enteros, luego no pueden ser ni decimales exactos ni
NÚMEROS REALES: Unidad Didáctica
  Los  números  racionales  también  contienen  a  los  números  que  tienen  expresión  decimal  exacta  0’12345  y  a  los  que  tienen  expresión  decimal 
Apuntes de Trigonometría III
Veamos que dado un ángulo cualquiera comprendido entre 90º y 360º, existe otro ángulo en el primer cuadrante con razones trigonométricas iguales, en valor absoluto, a las del dado...
Apuntes DE TRIGONOMETRÍA
Teorema del seno y del coseno Estos teoremas se utilizan para resolver triángulos no rectángulos, en los que no podemos aplicar ni el teorema de Pitágoras, ni las razones
Las espirales en el arte
La influencia de saber griego en el cual todo había de poder representarse con regla y compás en Durero marcaron la forma que obtuvo el método para resolver la espiral áurea, las cuales
Dossier cuarto op b
− Adquirir métodos y herramientas para resolver problemas del cálculo de probabilidades − Estudiar los diferentes casos que se pueden presentar a la hora de contar el número de
Cálculo Algebraico
Las operaciones de suma y de multiplicaci´on se extienden a este nuevo conjunto, y la resta queda bien definida entre cualquier par de n´umeros enteros.. Si bien la resta es una
RESUMEN de inecuaciones VA
Desigualdades no lineales Ejemplo Resolver la desigualdad x 2 > 2x Soluci´on: Cuarto paso: para la u ´ltima fila, multiplicamos todos los signos de cada columna:... Valor Absoluto
Inecuacionesysistemas resueltas
Estudiamos el signo de la función en cada uno de estos intervalos ¿Verifica la inecuación?... SISTEMAS DE

Etiquetas

Documento similar

Fully dynamic and memory-adaptative spatial approximation trees
These data structures are hybrid schemes, with the full features of dynamic spatial approximation trees and able of using the available memory to improve the query time.. It has been
0
0
12
Curcumin inhibits proliferation and metastasis of human hypopharyngeal carcinoma cells
In conclusion, these data suggested that curcumin can effectively suppress proliferation and inhibit the invasion and metastasis of FaDu cells via the induction of apoptosis, the
0
0
1
Netconf Element Management System . Design and Implementation
Conclusion This work was based on the design and implementation of Network Element Management System NEMS multi-user server to manipulate the configuration on those devices with Netconf
0
0
4
Solving a big-data problem with GPU: the network traffic analysis
In this paper we present the causes of increasing data volumes circulating on the network, data analysis and monitoring current techniques, and the feasibility of combining data mining
0
0
10
Big data analytics in intensive care units: challenges and applicability in an Argentinian hospital
Most of the CDSS systems work with non continuous data see Section Medical data and equipment, i.e., the data normally used by them are saved on clinical stories as: clinical data,
0
0
7
CBS: constraint-based approach for scheduling in bluetooth networks
Due to dynamic topology changes, different bandwidth available and unpredictable interference of media in Bluetooth networks, the congestion of data flow will inevitably emerges on the
0
0
7
Pre-school contributions to future achievements
Data Colombia has a mandatory educational system based on …ve years of primary and six years of secondary three of basic and three of middle education.. During these eleven years,
0
0
20
Compositional and structural characterisation of Ni-phyllosilicates in hydrous silicate type Ni-laterite deposits
The results of this multiscale study provide new data on the mode of occurrence, mineralogy and mineral chemistry at the micrometre and nanometre scale, the local environment of Ni,
0
0
171
Ensemble Case Based Learning for Multi-Agent Systems
With the purpose of studying the effect of distributed data and distributed decision making in learning problems, in this work we have focused on MultiAgent Case Based Reasoning Systems
0
0
299
Essays in the economics of health and wellbeing
The large sample sizes of these data sets allows me to follow cohorts - defined by state and year of birth - over time and analyze how the local unemployment rate around the time of
0
0
204
Genetic diversity of " brain genes" across worldwide
This new approach together with the flood of new polymorphism data in the form of SNPs available for genotyping, new data on the Linkage Disequilibium in the human genome in different
0
0
155
Lessons from Facebook: The Effect of Social Network Sites on College Students’ Social Capital
Hypotheses and Research Questions With the existing literature on social network sites and social capital in mind, this study uses original survey data to test the effects of using
0
0
39
DELE ENROLMENT APPLICATION FORM 2019 Enrolment conditions 2019
ADDITIONAL INFORMATION ON THE PROTECTION OF PERSONAL DATA In accordance with General Regulation UE 679/2016 on Data Protection and current legislation, we inform you that the personal
0
0
5
Food and Chemical Toxicology 50 (2012) 4221–4231 Contents lists available at SciVerse ScienceDirect
The lifespan of the control group of animals corresponded to the mean rat lifespan, but as is frequently the case with most mammals including humans WHO, 2012, males on average died
0
0
11
Contents lists available at ScienceDirect
Type 2/3, little GlcCer accumulation, no Gaucher cells, show systemic inflammation, anemia and leukopenia, and skin abnormalities Type 1 GD, normal life span, massive infiltration of
0
0
11
Show more