- Top of page
- I. Literature and Hypothesis Development
- II. Methodology and Data
- III. Empirical Results
- IV. Conclusion and Discussion
This study empirically evaluates the certification and value-added roles of reputable venture capitalists (VCs). Using a novel sample of entrepreneurial start-ups with multiple financing offers, I analyze financing offers made by competing VCs at the first professional round of start-up funding, holding characteristics of the start-up fixed. Offers made by VCs with a high reputation are three times more likely to be accepted, and high-reputation VCs acquire start-up equity at a 10–14% discount. The evidence suggests that VCs' “extra-financial” value may be more distinctive than their functionally equivalent financial capital. These extra-financial services can have financial consequences.
A central issue for early-stage high-tech entrepreneurs is obtaining external resources when the assets of their start-up are intangible and knowledge-based. Particularly for entrepreneurs without an established reputation, convincing external resource providers such as venture capitalists (VCs) to provide financial capital may be challenging. The literature contains two main lines of research for overcoming this problem. One research stream has concentrated on designing institutional structures to permit financing early-stage ventures. This contractual- and monitoring-based approach is aimed at solving potential agency problems between investors and entrepreneurs (e.g., Admati and Pfleiderer (1994), Lerner (1995), Hellmann (1998), and Kaplan and Strömberg (2001, 2002, 2003)). A second research stream has suggested that when the quality of a start-up cannot be directly observed, external actors rely on the quality of the start-up's affiliates as a signal of the start-up's own quality (e.g., Megginson and Weiss (1991), Biglaiser (1993), and Stuart, Hoang, and Hybels (1999)). This certification-based approach may help legitimate start-ups and entrepreneurs without a prior track record.
While the first research stream emphasizes the VC's problem (designing the appropriate mechanisms), the second highlights the entrepreneur's problem more directly (affiliating with highly reputable partners), and serves as an antecedent to this study. VC certification value, together with their value-added services such as recruiting executive managers (Hellmann and Puri (2002)), have led analysts in the descriptive literature to write: “It is far more important whose money you get [as an entrepreneur] than how much you get or how much you pay for it” (Bygrave and Timmons (1992, p. 208)) and “From whom you raise capital is often more important than the terms” (Sahlman (1997, p. 107)). These views clearly indicate that VCs have different value-added potential and that VC represents more than strict financial capital to entrepreneurs. In contrast, the extant academic literature has not emphasized VC heterogeneity, implicitly treating VCs as one uniform class so that reputation differences among VCs are obscured (see Gompers (1996) and Kaplan and Schoar (2003) for exceptions). As well, whereas much of the previous literature has concentrated on the benefits to certification (e.g., Megginson and Weiss (1991) and Stuart et al. (1999)), the costs of affiliating with prominent actors have not been systematically analyzed empirically. For example, the prescriptive advice to start-up entrepreneurs of affiliating with the highest status partner possible (Stuart et al. (1999)) seems strong given that calculations of returns to such action that do not take into account the costs of affiliation may be overstated. Indeed, demand for affiliation with reputable actors is likely to vary with the cost of such association. More generally, because affiliation with reputable partners confers performance benefits, such association cannot be freely accessed, for otherwise certification agents would not have incentives to invest in acquiring a reputation in the first place (Shapiro (1983)).1
Consequently, this paper explores two interlinked questions: Is there a market for affiliation with reputable partners? If so, what are the prices for such affiliation? Entrepreneurial demand for affiliation with VCs provides an excellent empirical setting to explore these questions for two reasons. First, because VCs can certify and start-ups need to be certified, the exchange nature of the relationship provides a natural marketplace for affiliation. Second, due to the tremendous increase in the supply of venture capital in the second half of the 1990s,2 the situation of “money chasing deals” makes observing a menu of price offers by VCs with varying reputation more likely—a necessary condition for identifying the market for affiliation.
The empirical analysis investigates proxies for VC reputation which explain the variation in offers accepted and valuations offered to start-ups at a point in time, while holding start-up characteristics fixed. To implement this methodology, I developed a novel, hand-collected data set of 148 financing offers (both those accepted and declined) made to a group of 51 early-stage high-tech start-ups. The estimated effects are both statistically and economically significant. A financing offer from a high-reputation VC is approximately three times more likely to be accepted by an entrepreneur. As well, highly reputable VCs acquire start-up equity at a 10–14% discount.
The empirical results suggest that entrepreneurs are willing to forego offers with higher valuations in order to affiliate with more reputable VCs. These results are consistent with the idea that VCs act as more than strict financial intermediaries, placing funds from investors to capital-constrained start-ups. If this were not the case, we might expect entry by suppliers of entrepreneurial finance to equilibrate prices for start-up equity across offers to a given firm. However, if VCs differed in the bundle of services and certification they provide to their portfolio companies, which might be thought of as “extra-financial” VC functions, then prices for affiliation might differ. This implies that the VC information network and its certification value may be more distinctive than their financial capital, and so these extra-financial VC functions can have financial consequences, namely, the price at which VCs are able to acquire equity in a given start-up. Indeed, this view is consistent with the stylized fact that VCs experience substantial interindustry variation in financial performance (Kaplan and Schoar (2003)). Consequently, future research exploring variation within the VC industry, especially as it relates to organizational performance, would be interesting.
The remainder of this paper is organized as follows: Section I discusses the relevant prior literature and derives hypotheses about the entrepreneurial market for VC affiliation. Section II describes the methodology and data used to test these hypotheses. Empirical results are discussed in Section III, while a final section concludes with a discussion of the implications and limitations of the study.
III. Empirical Results
- Top of page
- I. Literature and Hypothesis Development
- II. Methodology and Data
- III. Empirical Results
- IV. Conclusion and Discussion
The empirical assignment is straightforward—to test the hypotheses that (1) financing offers from more reputable VCs are more likely to be accepted, and (2) more reputable VCs acquire start-up equity at a discount. This section is therefore organized around empirical tables that demonstrate these relationships in both univariate and multivariate settings.
Table IV shows simple univariate comparisons of conditional means without controlling for fixed firm effects. Panel A describes difference in means tests for accepted versus declined financing offers. While the average pre-money value of accepted offers is $17.7 million, the declined offers averaged $22.1 million (the difference is not statistically significant, however). Accepted offers had higher values of VC reputation relative to nonaccepted offers, as measured four ways. The measures normalized industry deal experience (industry deal experience per year(s) of operation), normalized funds raised (number of funds raised per year(s) of operation), industry reputation rank, and high network resources rating all have higher values for accepted offers relative to declined ones. The differences in means for the latter three variables are statistically significant. Panel B describes the conditional means of relative valuation offered for the upper and lower halves (divided at the median) of normalized industry deal experience, normalized funds raised, industry reputation rank, and high network resources rating. Examining the conditional means of relative valuation offered rather than pre-money valuation in this context is preferred because the former measure incorporates some information about the comparative nature of the offers. The latter measure does not group offers by start-up firms in any way. While the differences in conditional means for the four reputation measures are not statistically significant, each of the relative means is consistent with the argument that more reputable VCs offer a discount to Series A valuation. Specifically, higher measures of VC reputation are associated with lower valuation offers. These univariate tests, while suggestive, do not control for qualities of the start-up, and so the remaining tables present a more systematic, multivariate analysis.
Table V examines start-up fixed-effects logits of VC offer accepted using Chamberlain's (1980) conditional likelihood method. Specification (5-1) shows that in the bivariate case, high industry deal experience is positively associated with VC offer accepted, at a statistically significant level (5%) and implies a 2.94-fold change in the odds of offer acceptance for a discrete change in this measure of VC reputation. While a more systematic exploration of the robustness of the VC reputation result is found in Table VII, a similar result holds in the bivariate relationship between VC offer accepted and high-normalized funds raised. The reputation result is strengthened when a measure of valuation, relative valuation offered, is included in specification (5-2). Notice the relative importance of the reputation effect over the valuation effect on the likelihood that an offer is accepted. Specification (5-3) includes an additional measure of VC reputation, high network resources rating, and controls for a variety of VC- and terms-of-financing-effects: angel investor, corporate venture capital, financing offered, and equity taken threshold. The high network resources rating measure is meant to capture VC value-added effects through contacts and/or resources that could make an offer more attractive (and can contribute to VC reputation). The estimated coefficient on this variable is positive and statistically significant at the 1% level. The measure angel investor is meant to capture the fact that a knowledgeable angel investor could be a substitute for a reputable VC in providing certification and business development resources, while the corporate VC method of organizing entrepreneurial finance may have implications for the value they can add to portfolio firms (Gompers and Lerner (1999)). Higher levels of financing offered may be a VC offer feature that may make it more attractive, since entrepreneurs may not have to return as many times or as soon for further financing rounds (fund-raising is an activity that may be quite time-consuming for start-up executives). Finally, the reputation result is not sensitive to the choice of a wide range of equity taken threshold levels between 20 and 50% of equity taken in the financing round (unreported regressions).
Notice that start-up characteristics are not included in these specifications. Since start-up characteristics (such as industry representation) are invariant across offers for a given start-up, including these qualities in the regressions does not affect the results. In addition, because financing offers for a given start-up did not span a large time window, variables on financing timing were not included in the regressions. In the pre-test of the survey, I asked respondents about the time window issue. It was my sense based on these interviews that the time window was not open for a long duration, given the start-up financing conditions of the late 1990s. Unfortunately, in the survey, I only noted the date of the realized Series A funding round, so I am unable to empirically document the time window length. Notwithstanding this shortcoming, the main result from Table V is that start-ups in this sample may not be selecting investors primarily on the basis of price and valuation; instead, VC reputation and affiliation effects may indeed be more important.
Table VI presents relative valuation offered start-up fixed-effects OLS regressions. The reported standard errors are robust—having been adjusted for clustering by start-up firm. The pairwise specification with high industry deal experience in (6-1) shows a negative relationship that is statistically significant at the 5% level. As well, the estimated coefficient implies a substantial discount, 14%, on relative valuation offered for a discrete change in the measure of VC reputation.
In (6-2), together with the measure of VC reputation, a dummy variable for VC offer accepted is included as a regressor. Notice that this parameter estimate, while positive (in both (6-2) and (6-3)), does not achieve statistical significance and is small in magnitude. The reputation effect persists and is of a slightly larger estimated magnitude relative to the previous specification. In model (6-3), several additional variables (parallel to those used in the prior table) are introduced. While the economic significance of the reputation result is slightly diminished in this specification, the parameter is estimated more precisely, achieving statistical significance at the 1% level. While the high network resources rating estimate is not statistically significant, it is estimated with a negative coefficient, which is consistent with the main hypothesis tested. The estimated coefficient on equity taken threshold is negative and significant at the 1% level, suggesting that larger equity stakes are associated with price discounts, though as previously mentioned, endogeneity concerns moderate the interpretation of this control variable. As well, the logarithm of financing offered is estimated with a positive, significant coefficient, indicating that the magnitude of funding, including potential liquidity effects, is associated with higher valuation. While robustness checks of the valuation regressions are presented in Table VIII, the results presented in Table VI are consistent with the idea that start-up entrepreneurs pay a premium to accept financing from more reputable VCs.
Because the above-reported results may be an artifact of either the particular measures used or due to selection biases arising from examining the multiple-offers data set, Tables VII and VIII present robustness checks of the reputation results for the offer acceptance and valuation regressions, respectively. The first three columns of Table VII successively employ alternate measures of VC reputation in fixed-effects logits to study the robustness of the positive correlation between VC offer accepted and reputation in similar specifications to (5-3). Specification (7-1) substitutes high-normalized funds raised for high industry deal experience as one of the measures of reputation. While the statistical significance falls to the 10% level, a discrete change in the funds raised measure corresponds to a doubling of the odds that an offer is accepted. Relative to specification (5-3), the estimated coefficient of high network resource rating is very similar in (7-1), both in magnitude and in statistical significance. In (7-2), industry reputation rank substitutes for high industry deal experience as an alternative measure of VC reputation. In this specification, both industry reputation rank and high network resources rating are positive and significant at the 1% level, though the estimated coefficient on relative valuation offered is much larger in comparative magnitude than the reputation measures. In (7-3), boards per general partner is used as an alternative measure of VC resources and is meant to capture the available time that partners in VC firms might have available in mentoring, developing, and connecting start-ups. While that variable is estimated with a nearly zero effect, the other reputation measure used in this specification, high industry deal experience, is estimated with quantitatively similar results (statistically and economically) to those found in Table V. Varying the measure of valuation as a control variable from relative valuation offered to the log of pre-money valuation causes high industry deal experience to fall to the 6% level, but does not alter the economic significance of the estimate (unreported specification).
Thus far, the analysis has not taken into account the possibility of a selection bias as a result of only using the multiple-offers data, though descriptive data from Table I suggest no statistical differences in the key observable start-up characteristics in the subsamples of the data corresponding to single versus multiple offers. Had we observed the alternate option for entrepreneurs that factually received single offers, would the results persist? Because establishing that counterfactual is difficult, two-stage Heckman (1979) regressions are presented where in the first stage, a probit of the likelihood of multiple offers is estimated using qualities of the start-up. These estimates are used in a second-stage fixed-effects regression of VC offer accepted (in Table VII) and relative valuation offered (in Table VIII) as an adjustment for possible selection effects.
The first-stage regression in Tables VII and VIII includes start-up covariates of the likelihood of receiving multiple offers: the natural logarithm of the pre-Series A number of start-up employees, L initial employees; a dummy equal to 1 if the start-up does not have assigned patents, zero patents; dummy variables for the following industry sectors: Internet industry (including infrastructure, services, and retail subsegments); health science industry (biotechnology and medical devices); computer industry (software and hardware); and year of Series A financing dummies for Year 1998, Year 1999, and Year 2000.6 Second-stage fixed-effects Heckman linear probability estimates of VC offer accepted are reported in the final two columns of Table VII.7 While the bivariate specification including high industry deal experience (7-4) is estimated more precisely (significant at the 1% level) relative to its counterpart in (5-1), the economic magnitude of the estimate is diminished. Meanwhile, the fully specified model (7-5) yields estimates of similar statistical significance to its counterpart in (5-3), though again with diminished economic significance levels.
Table VIII explores the robustness of the valuation results. A parallel specification structure to that used in the previous robustness table is employed. Specification (8-1) substitutes high-normalized funds raised as one of the measures of VC reputation. While the high-normalized VC funds raised variable is estimated with a positive (though insignificant) coefficient, recall that univariate comparisons in Table IV indicate that high-normalized VC funds raised was negatively correlated with relative valuation offered (although the difference was not statistically significant). In the multivariate regression, the prior funds variable may be picking up some countervailing effects, such that VCs with more prior funds raised are able to raise subsequent funds of larger sizes.8 The resulting relaxation in VC liquidity may have a confounding effect on this proxy for VC reputation.
Specification (8-2) utilizes an alternate measure of VC reputation, industry reputation rank. The variable estimate is negative and statistically significant at the 5% level. However, an objection to using this measure of reputation is that it is subject to entrepreneurial recall bias and/or ex post rationalization by the survey respondent. As an imperfect control for these potential effects, a dummy variable for VC offer accepted is included in the specification because the accepted offer is likely to be the chief candidate for recall and retrospection biases. Although the reputation measure is statistically significant at conventional levels, we should interpret the result cautiously because of the limitations of this measure.
In both specifications (8-1) and (8-2), high network resources rating, a measure of VC services to and resources for the start-up (and an important contributor to VC reputation), is estimated with a negative (though insignificant) coefficient, a finding consistent with the results from Table VI.9 Specification (8-3) varies this measure of VC resources to boards per general partner, and while the measure reaches statistical significance at the 10% level, the economic effect is insignificant. Importantly, note that the high industry deal experience proxy for VC reputation is robust (though reduced in statistical significance due to some degree of collinearity with boards per general partner). In an unreported regression, the log of pre-money valuation was used as an alternate measure of valuation. The estimates of high industry deal experience were robust to this variation. As well, introducing specifications with dummy variables for the most frequently appearing VC firms in the sample did not alter the main results (unreported regressions). These indicator variables may be appropriate if we believe that the pricing behavior of a handful of VCs is driving the results (over 100 distinct VC firms are represented in the sample, however).
Finally, in (8-4) and (8-5), fixed-effects Heckman regressions are reported using the entire sample of single and multiple offers in an effort to address potential selection issues. In both the bivariate and the fully specified equations, the results are very similar to those reported in Table VI—while the economic significance of the results is unchanged, the precision of the estimates is slightly enhanced.10
A final robustness check suggested that the hypothesized affiliation effects could be found using within-industry variation, though these results are not formally reported because they are merely suggestive. With the caveat that the categories of “Internet” (that includes Internet infrastructure, Internet services, and Internet retailing) and “non-Internet” (that includes biotechnology, medical devices, communications, and computer software and hardware) are very coarse groupings, the measures of VC reputation (high industry deal experience and high network resources rating) are positively associated with VC offer acceptance and negatively associated with relative valuation offered, although these relationships tend to hold more strongly for the non-Internet subsample and less so for the Internet subsample.11 These results are based on parsimonious specifications (keeping the limited sample size issue in mind); however, due to the nature of the data set, no conclusions about whether these results are due to time period effects can be made.
To conclude the empirical analysis, it is interesting to compare these results to a simple cross-sectional OLS analysis of the natural log of pre-money valuation on all accepted offers, done as if information on the bundle of declined offers were not available. The results, presented in Table IX, are striking.
In (9-1), a bivariate regression, high industry deal experience, is estimated with a positive coefficient, which is significant at the 1% level. When several start-up qualities are included in specification (9-2), the VC reputation result persists, disappearing in statistical significance only with the inclusion of VC characteristics (9-3), though the reputation measure is still estimated with a positive coefficient in that specification. As previously mentioned, problems of unobserved heterogeneity likely bias these estimates.
IV. Conclusion and Discussion
- Top of page
- I. Literature and Hypothesis Development
- II. Methodology and Data
- III. Empirical Results
- IV. Conclusion and Discussion
I have tested and confirmed the proposition that entrepreneurs are willing to accept a discount on the valuation of their start-up in order to access the capital of VCs with better reputations. These results help deepen our understanding of the market for affiliation by presenting empirical evidence that affiliation is an ordinary economic good for which actors seeking association will face a price-reputation trade-off. This finding is consistent with the view that VCs' reputation (which in turn depends on their experience, information network, and direct assistance to the portfolio firms) may be more distinctive than their functionally equivalent financial capital. These conclusions are drawn from an analysis of multiple offers to a set of start-ups, which allows a high degree of statistical control. Because the characteristics of the start-up can be held constant, only differences in VC reputation across financing offers explain inter-offer variation in offer acceptance and price for start-up equity.
One may wonder why prices charged by competing VCs to acquire the equity in a given start-up can be differentiated in equilibrium, given free entry. VCs with higher reputations may be able to sustain their higher prices (rather than having competition equilibrate prices) as a result of investments in reputation being costly (Shapiro (1983) and Megginson and Weiss (1991)). Consequently, while financial capital per se is not a differentiated good, the reputation of VCs providing the financial capital can be a source of differentiation among VC organizations. The findings in this article are consistent with Kaplan and Schoar's (2003) recent evidence of substantial cross-sectional variation and persistence in VC fund performance.
Several alternate explanations to the empirical pattern have been considered throughout the empirical analysis. First, more savvy VC firms might have foreseen the coming public market downturn and offered lower prices as a result. However, entrepreneurs receiving multiple offers would not necessarily have to accept offers from such VC firms, and so this explanation does not seem consistent with the observed empirical pattern. A second alternative hypothesis is that the term sheet covenants across offers for a given start-up may have differed. Indeed, the price VCs offer for equity may not be the only factor that matters when entrepreneurs select a VC firm, and other dimensions of the term sheet may not be “priced in” to the offered valuations. While surveyed entrepreneurs were asked for a copy of their term sheet offers, very few complied with this request. However, Suchman (1995) provides some evidence of convergence in VC financing agreements over time as a result of using the same law firms. As well, while Kaplan and Strömberg (2003) find that covenants in VC contracts differ by stage of start-up development, the offered terms of financing for a given start-up across VC firms may not be as variable. Nevertheless, different VCs probably prefer different terms, and offered terms are likely to vary even for a given start-up in a particular time period. As a result, comparing offers primarily on the basis of price is an inherent limitation of the present study.
Several additional issues associated with the data used for this study result in interpretational concerns. First, are the findings simply an artifact of the sample used? Although this group of companies might be of higher quality relative to average start-ups (assuming that the decision to be involved with MIT is a signal of quality), the sample, while modest at 148 offers, may represent a conservative test of the hypothesized effect. High quality entrepreneurs have their own reputations and established networks, which would tend to obviate the need to pay a premium to access capital from more experienced funding sources. In any case, employing start-up fixed effects makes this quality issue less important for the purposes of the empirical analysis. Indeed, the unique timing of the study in an environment in which many VCs were “chasing after deals” allowed identification of the market for affiliation—though it does not necessarily address the applicability of these results to other time periods or other relationships. It is difficult, unfortunately, to speculate on that answer.
A second issue is whether the results are produced from the competitive effect associated with studying a sample of multiple offers. This proposition is also hard to evaluate, however, due to the difficulty of establishing a counterfactual to single-offer situations. On a related note, a deeper understanding of the process leading to multiple offers would be desirable. The manifold processes generating offers (some of which are unobserved, such as entrepreneurial charisma), as well as the disparate bargaining processes leading to offers, makes the fixed-effects methodology attractive in controlling for unobserved heterogeneity. Nevertheless, had I collected information about the sequencing of offers, I might have been able to gain some empirical insight into the process generating multiple offers. Sequencing data may have also helped in beginning to empirically disentangle the pure affiliation effect from the VC value-added effect.
Two issues related to offers and how they might affect the interpretation of the results are also worth discussing. First, what if entrepreneurs “shopped” their deal to other VCs without receiving a formal term sheet? Unfortunately, I do not have the history of how many pitches entrepreneurs made to different VC groups without receiving formal offers. The results may be biased if (1) informal offers were leveraged to negotiate more favorable formal offers, and (2) VCs with a less established reputation were more willing to revise their valuations upward relative to VCs with more established reputations. Unfortunately, data constraints prohibit this analysis; however, survey respondents were instructed to provide information on final formal term sheets only (including informal offers would have subjected the sample to entrepreneurial interpretation of what constituted an informal offer). A second issue is whether some of the high valuation offers were withdrawn by VCs as the negotiations became more serious. Again, systematic data are unfortunately unavailable to address this question. If withdrawals came from across the full distribution of VCs, this would not bias the results. If, on the other hand, withdrawals were systematically from less-reputable VCs, the study would be biased toward finding the results (and the opposite would be true if reputable VCs tended to withdraw offers). Given the market environment (“money chasing deals”) in which these data were collected in the second half of 2000 and my conversations with survey respondents, however, I believe that offer withdrawal was not a pervasive phenomenon in this sample. However, because I cannot rule out these two issues, they represent caveats to the study.
A final issue is interpreting what start-ups are buying. For example, industry deal experience in the start-up's sector can proxy for both the scope of the VC's information network as well as the VC's ability to evaluate deals. Unfortunately, the data in this study do not provide a clean way of disentangling these effects. As well, higher ability entrepreneurs may be taking a lower offer in an effort to signal quality (Spence (1974)). While the empirical setting and the documented empirical patterns make this explanation unlikely, the proposition cannot be ruled out.
Looking to the future, while this study does not test the ex post performance implications of selecting a particular VC, it would be interesting to do so. For example, did start-ups accepting funding from more reputable VCs receive higher step-ups in valuation in subsequent rounds? Did they achieve an IPO faster or deliver products to the market more quickly?12 Nevertheless, the findings in this study are consistent with the theory that entrepreneurs who are tied into more connected networks at reputable VC firms expect to come across more opportunities for start-up growth, but must pay a premium for such access.