Competition and Specialization in the VC Market: A Non-monotonic Relationship

Staged financing in the VC market—financing to milestones—is a central part of the process. Instead of providing all of the necessary capital upfront, VCs invest in stages so as to keep the project under control. (We offer more institutional details in Sect. 1.1.) For example, Sequoia capital invested in Dropbox in early funding rounds, while it partnered with Zoom in late stages. Sequoia retained the position in Dropbox after early investment and stayed out of later-stage funding rounds. Insight Partners—the investor in Twitter and Tumblr—is known for funding larger-size growth-stage companies.

For an excellent overview and analysis of this industry, we refer the reader to Lerner et al. (2012).

In our data, a VC investor in a given entrepreneurial company participates, on average, in 2.6 rounds and 1.4 stages (we offer more details in Sect. 4). This suggests that while a VC usually participates in multiple rounds of financing (possibly as a signal to new investors that the company is achieving its milestones), it may discontinue the investment in a subsequent stage (the way we have defined the stages in this paper). This makes the notion of stage specialization more meaningful: in the sense that participation in a subsequent stage (for example) does not happen automatically but instead is a decision that needs to be made: taking carefully into account the benefits and costs. But more important, the expertise of the partners in the VC firm—which is an organizational structure decision that is taken prior to the investments—affects the composition of investments across the different stages. For instance, Patzelt et al. (2009) show that VC firms with higher proportion of the top management team members with science/engineering education and entrepreneurial experience are more likely to specialize in early stage investments.

Kanniainen and Keuschnigg (2004) is a notable exception. They focused on the trade off between the size of portfolio and the value of managerial advice.

Sequential moves reflect the fact that stage specialization is a longer-run organizational structure decision, as compared to the equity share (“price”) decision.

Hence, we can choose parameter values—e.g., V high enough—to guarantee that \(\alpha _s^{*}<1\).

There are two complementary sub-markets: the market for early-stage investments and the market for late-stage investments. Some VC firms may be active in both markets (generalists) and some in only one (specialists). The symbol n refers to the total number of VC firms.

Cooperation can be tacit. If it is about the stage participation decision, each VC firm realizes that a deviation from a specialist to a generalist will trigger similar responses by all the other VC firms in the future and (if it can) refrains from deviating that way. Similarly, VC firms realize that by competing to attract ENs the VCs end up offering to accept low equity shares for themselves; and so, if they can, the VCs tacitly agree to mitigate such competition and keep higher equity shares for themselves.

The proof is in Appendix A. We assume that V is high enough so that the market is covered. We also need a high V in order to rank the profits of different deviations in the Appendix. Moreover, and without loss of generality, we set \(t = 1\) and \(r = 0\). Although the cost of funds r affects the equity share offers, it does not affect the profits due to the covered market assumption (or the inelastic aggregate demand, which is inherent in the Salop model): there is a fixed number of ENs that seek funding, and this number does not depend on the equity shares. All that the equity share does is to determine how these ENs are allocated across the VC firms in the market.

See also Fig. 1 where we plot the critical discount factors (we have set \(V=5/2\)).

Note that there exists \(\delta >2/3\) such that VC firms are specialists (but compete in equity shares) even when n is very high. This is because when there are many VC firms, the market share of each one is very small, before and after a deviation, which implies that the incentive to deviate to a generalist is very low. In contrast, when VC firms cooperate in equity shares a deviation to a lower equity share increases the deviator’s market share and profits significantly. The critical \(\delta\) in this case tends to one as \(n\rightarrow \infty\).

Equilibrium VC profit is decreasing in n for any given discount factor \(\delta\).

Our model is symmetric in many aspects. One such aspect is that either all VC firms are generalists or specialists. Symmetry facilitates our goal of allowing for an arbitrary number of firms in order to analyze the impact of competition, since it would be impossible to derive clean theoretical predictions with asymmetric firms. Of course, in reality (and in our data) different VC firms have different ‘degrees’ of specialization in the same market. Therefore, our theoretical model should be viewed as an approximation of reality and judged for the empirical validity of its predictions.

Schwienbacher (2013) takes the VC stage specialization decisions as given and develops a theoretical model that seeks to understand the generalist versus specialist tradeoff that an early-stage entrepreneur who looks for VC funding will face. VCs that specialize in early stage add more value, but are more reluctant to finance later rounds; the opposite is true for generalists. Kang et al. (2011) develop a decision model for VC portfolio choice in which a VC chooses between fund specialization and diversification in the presence of potential knowledge spillovers among projects. None of these models, however, examine the effect of competition on these decisions.

See Lerner (1995) and Gompers and Lerner (1999) for discussions about the VentureXpert database and data coverage.

Stage refers to the development status of an entrepreneurial company (e.g., seed, early, late etc.). VCs usually transfer capital to entrepreneurs through multiple funding rounds. A particular stage can entail multiple investment rounds. Entrepreneurs may or may not advance to the next development stage to receive a new round of investment.

We exclude investment rounds that involve individuals, undisclosed VC firms, angel investors, and buyout funds. We focus on entrepreneurial companies in the contiguous U.S. states and exclude those in Hawaii, Alaska, or Puerto Rico. A VC firm could manage multiple funds. Like Gompers et al. (2009), we focus on VC specialization at the VC firm level. A VC firm could manage multiple funds and each fund is usually designated for a specific type of investment. It is easy for a VC firm to change the scope of investments by raising new funds.

Hochberg et al. (2010) also use the interaction of industry and Metropolitan Statistical Area (MSA). Results that are based on the industry-MSA market are very similar to the state market and are available upon request.

We focus on specialization at the VC firm level instead of at the fund level since a VC fund is usually dedicated to invest in companies that are in particular stages of development and industries. Diversification at the VC firm level can be achieved by directing resources to different funds under its management. In addition, VC competition primarily takes place at the VC firm level.

The growth stage includes both the expansion and later stages, as defined by the VentureXpert database. Since there is no clear distinction between expansion and later stages, we combine them into a single category. About 53% of other-stage VC rounds are bridge loans and 22% are open market purchases.

Alternatively, we also consider different stage classifications that include: (1) using the five stages (seed/startup, early, expansion, later, and other) as defined by the VentureXpert; and (2) excluding deals in “other” stage. They do not change our empirical results.

The dollar value that each VC firm has invested in a given round is not readily available. Instead, VentureXpert reports the total round amount (made by all investors that participated in that round) and the total amount that a VC firm has invested in a given entrepreneurial company. We impute a VC firm’s round investment value by distributing the total investment in a portfolio company into each round in which the VC firm has participated according to round amounts. For example, suppose an entrepreneur received $1 million in total from \(VC_{i}\) that participated in rounds 1 and 4. If the total amounts raised in rounds 1 and 4 are $2.5 million and $7.5 million dollars, respectively, then the imputed investment for \(VC_{i}\) in round 1 is \(1*2.5/(2.5+7.5)=0.25\) million and in round 4 it is \(1*7.5/(2.5+7.5)=0.75\) million.

We decided to include low-volume VC firms in our later analysis for two reasons: first, there are many of them: about 70% of observations. Second, the return to specialization may be especially important for low- volume VC firms, which are usually more resource-constrained than are high-volume VC firms. As we discuss shortly, the non-monotonic relationship between specialization and competition is robust to the exclusion of low-volume VC firms.

\({\textit{hhi}}\) is highly correlated with \({\textit{dhhi}}\), with a correlation coefficient of 0.93.

\({\textit{exp}}\) and \({\textit{cumexp}}\) enter our specification in logarithms. We add 1 to all logged variables, so as to avoid a problem with 0 values.

\({\textit{size}}\) enters our specification in logs. Besides the variables reported in Table 1, we have experimented with a large set of controls (e.g., fraction of syndicated deals, fund growth rate etc.). However, these variables explain little variation in stage specialization and are not included in our specifications.

We also experimented with variables such as state population and GDP. Unfortunately, they perform poorly in predicting market structure, once the cross-sectional variations are limited by fixed effects.

The results are similar if we use un-lagged demand variables.

Following Gompers and Lerner (2000), we used the public-market pricing multiples to proxy for private market investment climates. Specifically, we calculate the average book-to-market ratio with the use of CompStat data and map them to the six VC industries. Hochberg et al. (2010) suggest that incumbent could easily reduce entry by refusing to syndicate with entrants. The fraction of syndicated deals could proxy for (the inverse of) entry barriers in a market.

Hochberg and Westerfield (2016) examine industry and geographical specialization.

There are many market structures that could generate \(\mathrm{HHI}=0.41\). For example, three firms with shares being 0.2, 0.26 and 0.54 could generate \(\mathrm{HHI}=0.41\). As can be seen in Table 1, the mean HHI is 0.33, while the mean number of VC firms is 15.5. Therefore, we have to interpret the HHI number with caution. The message here is that the effect of competition on specialization in concentrated markets is different from that in competitive markets.

Our results based on the number of VC firms in a market are also robust to the exclusion of inactive VC firms. As defined by Hochberg et al. (2010), a VC firm is considered actively investing in the target market if it funded at least five deals in the past five years.

We have also performed the analysis by (1) focusing on the first investment in a given entrepreneur and excluding all follow-on investments, and (2) by excluding all deals in “other” stage as defined by the VentureXpert database. The results remain qualitatively the same. We thank an anonymous referee for suggesting these robustness checks.

Our results also remain the same if we focus on VCs with at least 20 deals across all markets in a given year.

Additional investigations also suggest that our results are not driven by a particular state, industry or time period. For example, we find very similar results if we limit our time window to account for structural breaks in a VC firm’s focus.

In CA, NY, and MA, the quadratic term is sometimes imprecisely estimated, due to lack of variation in market structure. Markets are highly competitive in these states, and most of these markets belong to the upward-sloping region of the inverted-U curve.

To economize on space, we do not report these results in this paper but they are available upon request.

There is evidence in the VC literature that specialization produces ‘better’ outcomes. For instance, Gompers et al. (2009) empirically document a higher probability of an IPO among more industry-specialized VC firms. What remains to be seen is whether stage specialization has a similar effect.

Several papers in the VC literature have examined the effectiveness of public initiatives that aim to boost entrepreneurial activity. These initiatives are likely to have an impact on the structure of VC markets. Lerner (2002) discusses the efficiency of ‘public venture capital’ programs that make equity or equity-like investments in young firms, or encourage other intermediaries to make such investments. An important question is how public authorities should structure these programs so as to ensure their greatest effectiveness. The European Union—in an effort to boost the European VC market and increase the VC supply—has enacted policy initiatives, such as the 1998 Risk Capital Action Plan, that includes measures that are aimed at increasing stock market openness and/or labor market flexibility, tax incentives and so on (see, e.g., Cumming et al., 2017; Grilli & Murtinu, 2014). Da Rin et al. (2006), focusing on 14 European countries between 1988 and 2001, find no evidence that public policies that increase the supply of VC funds, have any effect on high-tech and early stage investments. What matters more is the opening of ‘new’ stock markets for entrepreneurial companies, taxation (see also Keuschnigg and Nielsen 2002) and labor market policies. Samila and Sorenson (2011) found that the supply of venture capital positively affects firm starts, employment, and aggregate income in local markets.

\(\pi ^d\) is a multiple of 4 because the deviating firm captures ENs on both sides of 0 and across both stages. Also, given the linear transportation cost, the ENs that are located between k/n and \(((k+1)/n)/2\) are also indifferent between the two VC firms. We assume that they patronize the VC that is located at k/n.