Venture capital (VC) firms are one class of financial intermediaries that invest their managed funds in venture business companies. VC firms attempt to find promising investment targets and support them to achieve venture business companies’ business goals by providing not only the financial resources but also various kinds of expert advice, and eventually obtain a return from their investments through initial public offerings (IPOs), trade sales, or management buyouts. Invested companies can obtain access to those resources through VC firms, and such series of activities could contribute to their growth.
Given that such investment activities are highly knowledge-intensive, VC firms use a variety of accumulated resources (e.g. screening, monitoring, and advising abilities). While such resources accumulated within the VC firms themselves are necessary for successful investments, better access to resources accumulated ‘outside’ could also be important for their investments. These resources that are accumulated externally include, for example, specific industry knowledge and a pool of potential deals provided by co-investors. It has been reported that VC firms tend to specialise in specific industries so that they can efficiently accumulate industry-specific knowledge (e.g. Hochberg et al. 2007). The fact that they frequently co-invest implies that such resources accumulated outside of their own firms are indispensable for successful investments.
How do VC firms with good track records go on to co-invest?
These illustrations suggest that the pattern of co-investment network formation reflects how the resources held by VC firms are employed, which is of great interest to both practitioners and academic researchers. Do VC firms rely solely on their own resources, or do they employ other firms’ resources to complement their own? If so, what kind of resources are exchanged among them? Does past collaboration matter? Under what economic environment are such mechanisms more viable? In a recent paper, we tackle these questions by empirically documenting the pattern of co-investments among VC firms and discuss its economic implications (Koujaku and Miyakawa 2017). More specifically, we are interested in how firms with better track records in terms of their investment return and/or invested companies’ growth co-invest in future. Following the literature on network science, we call the patterns of co-investment network formations associated with VC characteristics (e.g. better track records of co-investor VC firms) network ‘configurations’, and study the emergence probability of a network with specific configurations given the VC characteristics.
Regarding such configurations, we should note that there are many possible patterns associated with network formation among VC firms. As one plausible case, those with better resources might co-invest only with well-performing firms. It could be the case, for example, that those ‘good’ VC firms can exchange their internal resources efficiently with each other so that they can obtain larger joint surplus. As a simple illustration, one firm holding information about a promising investment opportunity but lacking industry expertise might want to be matched up with other firms with such resources. If this is the case, we would observe positive assortative matching among them in terms of performance. As another case, consider the situation where well-performing firms hold small financing capacity. Such firms do not necessarily need to exchange, for example, deal flows or other expertise, but only need to find someone to satisfy financing needs. If this is the case, we might not observe any assortative matching among the firms in terms of their performance. Thus, it is an empirical question whether there is any mapping pattern from their characteristics to the emergence probability of networks with specific configurations.
The challenges of empirical examination
While the above-mentioned research theme is straightforward, empirical examination encounters a number of challenges. First of all, it is not necessarily obvious how to measure VC characteristics that are meaningful for this co-investment pattern. Suppose a VC firm exhibits good performance in terms of the return obtained from the past investments. This could lead to a reasoning that it owns valuable resources attractive to other firms. Thus, we might expect that this firm would be involved in a large number of future co-investments. However, we should note that this performance measure is constructed only from its perspective and not from other companion VC firms or the target VB company. Even if one specific firm enjoys a high return from its investment, it could still be the case that other companion firms did not do well because, for example, they needed to pay a relatively high price for their investments. In a similar sense, the invested venture business company might face difficulty after accomplishing its IPO. As many studies have reported, VC firms might induce venture business companies to go public even if it would not contribute to the latter’s long-term growth (Hamao et al. 2000, Hellmann et al. 2008). Thus, from the viewpoint of empirical analysis, it is not entirely satisfactory to focus only on the return obtained by one VC firm as the characteristics meaningful for the co-investment pattern. Rather, we would need to employ multi-dimensional VC characteristics, taking into account other co-investor VC firms and invested venture business companies.
Second, we also have to take into account the investment history associated with each VC firm. Having better characteristics would lead to a higher likelihood of joint investments if such characteristics are well recognised by other firms. This could be the case when, for example, the firm and its companion firm had experienced joint investment in the past. This implies that we need to take into account not only each firm’s characteristics, but also its past investment history.
Third, we also need to take into account the economic environment surrounding VC firms and venture business companies. To illustrate, under difficult financing conditions for VC firms, each might be less picky about the quality of its investment partners as far as such collaboration could contribute to securing sufficient funds. Thus, it could be the case that the positive assortativity among VC firms becomes weaker under such worse market condition. This discussion requires us to allow the time-variant feature of the mapping pattern from VC characteristics to the emergence probability of graphs with specific graph configurations.
Given these three concerns, we employ the exponential random graph model (ERGM) developed in network science literature (e.g. Snijders 2002, Robins et al. 2007), which can simultaneously account for the multi-dimensional performance measures, various matching configurations, and time-variant feature of the network formation.
A limited number of studies have examined the economic implication of co-investment and the dynamics of VC networks. As a prominent study in this field, Hotchberg et al. (2010) found that VC firms formulate networks as a barrier to entry. Also, Hotchberg et al. (2007) found that those with higher centrality in the network perform well. These studies succeed in illustrating the roles of VC co-investment networks, but it has not been clear how such networks are formulated. In this context, the most related study to ours is Hotchberg et al. (2015), which examined the determinants of network formation and confirmed that the resource exchange motive is more important than the homophily motive. Given these backgrounds, applying the exponential random graph model to unique investment data of Japanese VC firms over the last two decades, we empirically examine the relationship between VC firm performance and the dynamics of their co-investment networks.
Our findings are summarised as follows.
- First, from the estimation results of the ERGM, we find that VC firms’ co-investment network formation is not independent from VC characteristics.
- Second, more specifically, their past experiences of co-investments contribute to higher likelihood of future co-investments among them not only when they have gained higher return from their past co-investments, but also when the jointly invested venture business companies experience higher growth after IPOs. These results are stably observed over the data periods.
- Third, such positive assortativity in terms of the returns obtained from their co-investment became significantly weaker after the great financial crisis in 2007-2009 (see Figure 1), which is consistent with another result that VC firms’ co-investment network formation became less dependent on VC characteristics over the same periods.
These results jointly suggest that the worse financial market condition made the network structure less stiff. This could be partly motivated by VC firms’ need for financing.
- Fourth, and somewhat puzzling, the positive assortativity in terms of jointly invested venture buiness companies’ growth became stronger after the great financial crisis. It could be the case, for example, that there are elite networks of VC firms holding resources contributing to venture buiness companies’ long-term growth, and it became more difficult for the former without resources to join such elite networks after the Global Crisis.
Figure 1 Impact of past co-investment return on future co-investments
Note: Return is measured over [t-6, t-3], network is measured over [t-3, t]
Directions for future research
Given that the current study is still in its developing stage, we are planning to expand this research to a number of directions. First, we need to enrich the list of configuration so as to identify the central driver(s) of VC network formation. Among the potential set of configurations, it would be important to consider the homophily and heterogeneity in terms of VC firms’ characteristics. Second, we need to examine if the observed positive assortativity among these firms in terms of their performance actually leads to better performance in the future. If this is the case, we can confirm the ‘rich gets richer’ pattern in the VC industry. Third, we can use the exponential random graph model to study more complicated network configuration corresponding to, for example, more than two firms. Given the importance of accommodating more and more startup companies, which presumably contribute to vital economic conditions, it is important to examine these research questions by using the analytical framework that we employ.
Editors’ note: The main research on which this column is based first appeared as a Discussion Paper of the Research Institute of Economy, Trade and Industry (RIETI) of Japan.
Hamao, Y, F Packer and J R Ritter (2000), “Institutional Affiliation and the Role of Venture Capital: Evidence from Initial Public Offerings in Japan”, Pacific-Basing Finance Journal 8: 529–558.
Hellmann, T, L Lindsey and M Puri (2008), “Building Relationships Early: Banks in Venture Capital”, Review of Financial Studies 21: 513–541.
Hochberg, Y, L Lindsey and M Westerfield (2015), “Resource Accumulation through Economic Ties: Evidence from Venture Capital”, Journal of Financial Economics 118: 245-267.
Hochberg, Y V, A Ljungqvist and Y Lu (2007), “Whom You Know Matters: Venture Capital Networks and Investment Performance”, Journal of Finance 62: 251-301.
Hochberg, Y V, A Ljungqvist and Y Lu (2010), “Networks as a Barrier to Entry and the Competitive Supply of Venture Capital”, Journal of Finance 65: 829-859.
Koujaku, S and D Miyakawa (2017), “Venture Capital Networks: An analysis using the exponential random graph model“, RIETI Discussion Paper Series 17-E-084.
Robins, G, P Pattison, Y Kalish and D Lusher (2007), “An introduction to exponential random graph (p*) models for social networks”, Social Networks 29(2): 173-191.
Snijders, T A B (2002), “Markov Chain Monte Carlo Estimation of Exponential Random Graph Models”, Journal of Social Structure 3.