1.1 – Corporate private equity

8Corporate venturing can be defined as an inter-firm collaboration between a large established company and an innovative start-up, where the former takes an equity stake in the latter. The collaboration provides the start-up access to the capital, experience, and other resources of the large firm, while helping the large firm to overcome its non- entrepreneurial culture, characteristic to big corporations, hindering innovations (Siegel, Siegel and MacMillan, 1988) hence revitalize and improve its strategic and financial performance (Chesborough 2002; Dushnitsky and Lenox 2005, 2006).

9The objectives of corporations participating in corporate venturing are twofold. On the one hand, a CPE unit of a corporation is expected to generate financial returns on their investments, as other PE firms do. This implies screening the market for promising target companies, helping them to solve financial and organizational problems by giving them access to corporation’s capital, using their own network to matching them with suppliers and customers, providing strategic advice, closely monitoring management activities. In return they expect that the investment will pay off in due time, because the growth in value of the portfolio firm will allow the corporation successfully exit from the portfolio firm through IPO, M&A or LBO. As Röhm (2018) noticed “[f]rom a financial point of view, the short-term inefficiencies and costs of CVC initiatives can be compensated for, if corporations understand the use of CVC as a long-term instrument” Dushnitsky and Lenox (2006) found that CVC increase the corporation’s growth opportunities (measured by Tobin’s Q). Zahra and Hayton (2008) investigated the relationship between a firm’s external venturing activities and its financial performances and found that investments made through CVC funds are positively associated with a corporation’s ROE and revenue growth.

10On the other hand, unlike independent PE, corporate private equity generally have another set of objectives of a strategic nature. To keep its position in its market an established company must stay innovative by constantly exploring new emerging opportunities and developing new capabilities. Through external corporate venturing the company learns about new technologies developed by their portfolio firms, so that it can adopt appropriate technology for building up a new business related to its core activities (Dushnitsky and Lenox, 2006; Hellmann, Lindsey and Puri, 2008). Winters and Murfin (1988) mentioned several examples where the detailed knowledge of venture company activities obtained by involvement in venture capital had influenced the strategic planning of major corporations. In addition, companies actively engaged in CVC have to constantly screen promising technologies, which help them staying up to date (Chesbrough and Tucci, 2004). Furthermore, CVC investments are often syndicated, therefore the strategic benefits of CVC may as well include opportunities to observe and learn about co-investors’ operational processes, their know-how and their capabilities (Heimeriks and Duysters, 2007). More specifically, it helps the company learn how their partners conduct their corporate venture investments and how they internalize innovative ideas championed by syndicate partners and their investees (Gulati, 1999).

11This second agenda of CPE investment may influence CPE behaviour and therefore affect the probability of successful exit from their portfolio firms. In presence of a trade-off between financial and strategic objectives, a CPE firm may act differently than an independent PE firm does. This, for instance, may happen at the pre-investment stage – the CPE may go for riskier or less rewarding investment if the target firm has a technology that may have a strategic value to the CPE’s parent company business. Also at the post-investment stage, a CPE provider might try to impose the portfolio firm a strategy that is congruent to its parent company even if it is sub-optimal from the portfolio firm’s view. Zahra (1996) studies the differences in technological strategies between CVC-backed and independent start-ups his finding suggests that although corporate-sponsored ventures spent more heavily on R&D, stressed basic R&D and patent more, they pursue less ambitious and less innovative technological strategies than do independent ventures.

12On the other hand, when the objectives of portfolio firm coincide with those of the CPE’s parent company, having such an investor may give the portfolio firm access to the resources that it may not otherwise have. Park and Steensma (2011) have shown that a CVC is beneficial for start-ups when they use parent company product development, manufacturing, legal, sales, distribution and customer service, to commercialize new products. New ventures that require high development costs benefit the most from such a help. Chemmanur, Loutskina and Tian (2013) provided statistical evidence that firms financed by CVC in the US are more innovative, considering the number of patents granted, and the number of citation of these patents.

13The strategic dimension of corporate venturing and its consequences for the behaviour of CVC can be seen through the lens of resource-dependence theory (Pfeffer and Salancik, 1978). Both sides, the start-ups and the investing corporations, seek resources needed for their survival in a competitive market. For the start-ups these resources may include capital, management skills and experience, knowledge of the up- and down- stream markets, access to suppliers and customers. For the corporate investors, besides the financial motive, a CPE provides an opportunity to learn about new technologies and new capabilities, which may potentially happen to be future game changers.

14Dependence of a start-up from the resources of the corporation allows the latter extend its power over the former and limit its choice of actions. However, such dependence is potentially mutual because the start-up controls corporation’s access to new technology. This reverse direction of power relationship is absent in relationships between independent PE firms and their portfolio companies. It is, however, difficult to say a priori whether the corporate PE gets higher or lower degree of control over the new venture than an independent PE investor would have in the same situation, because on the one hand the corporation controls more resources which the venture needs, on the other hand private investor does not depend on the access to the venture’s technology.

15There is another, related, aspect in which CPE firms’ behaviour may differ from the behaviour of an independent PE. Independent PE firms typically invest money from institutional investors and risk their reputation among such investors, while CPE firms invest money of their parents companies, thus do not have the same incentives as their independent counterparts. As a result, we may expect the independent PE firms be more meticulous in the selection of their investment targets and have stricter control rules for their portfolio firms. Chemmanur, Loutskina and Tian (2013) suggests that CVC investors have a greater tolerance for failure, measured by the amount of time that VC allow start-ups to have before stopping their investment, when their portfolio company is in difficulty. In similar vein, Guo, Lou and Pérez-Castrillo (2015) conclude that CVC investors are more patient than independent VC investors with respect to timing of financial returns: on average the CVC-backed start-ups remain longer in portfolio of their investors, and CVCs keep investing in distressed companies longer than do independent VCs.

16To sum up the discussion on the difference between motives and behaviour of corporate and independent private equity firm, we formulate the following hypothesis.

• H₁: The probability of success of a portfolio company differ between corporate and independent private equity firms.

1.2 – The role of networks

17The resource-based view of the firm considers a firm as a bundle of productive resources and its performance depends on the attributes of the resources it can procure (Penrose, 1959). However, even the largest firms have limited resources within their organizational boundaries, and therefore firm’s links to other organizations play vital role providing access to the resources of their partners, and through them to the resources of the partners of the partners and so on. Thus, networks are important and “corporate social capital” defined as “processes of forming and mobilizing social actors’ network connections within and between organizations to gain access to other actors’ resources” (Knoke, 1999). In the context of PE firms, such resources would include capital, managerial skills and experience but, foremost, information.

18The role of networks of personal connections and/or inter-organizational links in producing valuable information and thereby reducing informational frictions has also been acknowledged in the financial literature (Allen and Babus, 2010). For example, mutual fund portfolio managers often place larger bets on firms with which they are connected through their personal networks, and on such holdings, they perform consistently better relative to their non-connected holdings (Cohen, Frazzini and Christopher, 2008). In M&As, connections between acquirer and target firms provide the acquirer with an advantage, based on better information on true value of the target, over the non-connected and therefore less-informed bidders (Cai and Sevilir, 2012). In syndicated loan markets, borrowers lending from syndicates composed from lenders occupying central network positions pay lower costs (Godlewski, Sanditov and Burger-Helmchen, 2012). In IPO markets, book managers’ position in the network of underwriters has significant effects on IPO pricing and placement (Chuluun, 2015).

19Networks play an important role in entrepreneurial finance. Investors in new ventures have to operate in the environment with high risks and information asymmetries. The latter arises both before investment, in the context of initial screening when CPE firms select target portfolio firms, as well as at post-investment stage, in the context of monitoring the start-ups in which the CPE firms have invested. Start-up’s innovativeness exacerbates the uncertainty and asymmetry of information between the start-up and the investors. Private equity firms often invests collectively together with other private equity firms (corporate or independent), forming PE syndicates (Awounou and Dubocage, 2016). By redistributing the risk of start-up failure among the partners, syndication reduces exposure to the risk for an individual firm (Gompers, 1996; Brander, Amit and Antweiller, 2002).

20However, syndication also serves another, more subtle, function. Observing each partners’ willingness to invest in potentially promising deals, PE firms can pool correlated signals on the quality of those deals and thereby select better investments in situations extreme uncertainty about the viability and return potential of investment proposals (Wilson, 1968; Sah and Stiglitz, 1986). Thus, in a syndicate partners PE firms do not only share the risks, but also knowledge, experience and sensitive informal information concerning portfolio companies (Bygraves, 1987; Lerner, 1994, Manigart, Lockett, Meuleman, Wright, Landström, Bruining, Desbrières and Hommel, 2006), and provide the latter with the “non-financial value-added resources” from the PE firms’ network (Dubocage and Rédis, 2016). Furthermore, syndication appears to be a relevant mechanism to mitigate agency problems at the exit time as it improves IPO performance (Awounou and Dubocage, 2016). Since many PE firms participate in more than one syndicates, the emerging network of PE syndication serves a function of an information network that allows investors get timely access to important information about start-ups they are funding.

21Bygraves (1988) studying co-investment patterns among venture capital firms in the US noticed that the VC firms financing high innovative technology companies form a dense tightly coupled network by contrast with the VC firms financing low innovative technology companies that form a loosely connected network. He argued that the VC firms investing in innovative start-ups usually face higher uncertainty and therefore had to rely more on external information provided by the VC network, than do their counterparts who invest in less innovative ventures. Sorenson and Stuart (2001) studied the geographical distribution of venture capital in the US and found that even though VC investments tend to be localized, the VC syndication network can mitigate the negative effect of distance. By channelling information about the start-ups the network extends geographical reach of the VC firms which are well positioned in this network.

22Hochberg, Ljungqvist and Lu (2007) used the data on 3’469 US venture capital funds to study the effect of VC firm position in the network of VC syndication on the success of its portfolio firms. They found that VC funds whose parent VC firms occupy central positions in the network have higher fraction of a fund’s of portfolio firms that have been successfully exited via IPO or M&A, than do VC funds with less central parents.

23With rare exceptions the abovementioned earlier research has focused on the US with its mature well-developed venture capital market, and even though we expect to find similar patterns in French private equity industry, we must test whether such intuition is supported by the data and, therefore, our second hypothesis is as follows.

• H₂: The probability of success of a portfolio company increases with the centrality of the PE firm in the PE syndication network.

24While previous research cited above concludes that the network position of the VC, and more generally the network position of the PE firm, may have an effect on its performance, the literature is largely silent on whether such an effect differs between the types of PE firms.

25Given our interest in CPE, we may formulate several reasons why the effect of the PE network on the success of the CPE firm’s investments may differ from its effect on the investments of independent PE firms.

26First, while for independent PE firms investing in start-ups is their raison d’être, from the point of view of the parent corporations start-up financing is only one innovation strategy among other strategies, and often of a secondary importance. In addition, as mentioned above, the reputation concerns related to companies’ failures for independent PE firms are likely to be higher than for CPE firms. As a result, investment in portfolio firms does not receive the same scrutiny in a CPE firm, as it does in an independent PE firm, because the corporation either does not have or does not want to devote sufficient human resources for such task than an independent PE firm would do. [4] Under such conditions, access to external information, expertise and managerial skills of the other firms is more important for CPE firms than for independent PE firms.

27On the other hand, CPE firms, being units of their parent corporations, have direct access to many other resources (capital, technology, relationships with buyers and suppliers, etc.) and therefore do not depend on the PE network to the extent as the independent PE firms.

28Therefore we may expect to see that the magnitude of the effect of PE network differ between corporate and independent PE firms.

• H₃: The effect of the centrality on the probability of success of a portfolio company differ between corporate and independent PE firms.

2.1 – Network and centrality measures

30In order to construct networks of PE firms we make an assumption common in the empirical research on financial networks discussed in the previous section: we assume that two investors are connected, if they have participated in the same PE syndicate. Since most PE firms participate in many investments, PE firms and their connections constitute a network, and such a network can be studied using methods of social network analysis. We also assume that relationships between investors are symmetric and have the same strength, thus the structure of our network can be adequately described using an unweighted undirected graph with vertices being investors and edges being co-investment connections among them.

31We employed three popular measures of network centrality, originally introduced by Freeman (1979), which have been widely used in earlier studies on financial networks: degree, closeness and betweenness centrality measures.

2.2 – Sample

37Our sample came from Thomson One Banker database and covers private equity (PE) investments in French companies from 1989 to 2016. The initial sample that we use to construct the network of PE syndication is based on the information on investments of 800 PE investors. [5] We then proceed cleaning the data by disambiguating names of the PE firms and eliminating the duplicates. The final sample includes 478 PE investors into 426 portfolio companies. The breakdown of the investors by their type is shown in Table 1.

38Although we limit our attention to performance of French independent and corporate PE firms, our assumption is that the information (as well as other relevant resources) flowing through co-investment network can pass equally well any type of investor (not only private and corporate PE firms). Therefore, the syndication network has been constructed including all types of investors (independent, corporate, financial, and public) covered by Thomson One Banker that have invested in French companies whether they are located in France or abroad. The network centrality measures defined above were calculated for such extended network.

39The constructed syndication network consisting of 478 vertices (investors) and 3’163 edges (co-investments) is well connected and contains the largest connected component (also called “giant component”) of 454 vertices (95% of all sample), 5 components of size 3 and 2 and 10 isolate vertices.

40Since our dependent variable measures the number of successful exits from portfolio companies, our data is prone to right-censoring, because PE firms still have not exited from some recent investments by the time we have extracted the data from Thomson One Banker. To limit possible bias we limit our analysis to those investments from which PE firm have exited. As a result, the 88 PE investors who had only active investments (i.e. no past investment from which they have exited) by 2016 had to be excluded from our sample.

41From the remaining 390 investors we have extracted all French PE firms (independent or corporate). Our study sample has 119 PE firms having legal address in France (out of 159 French PE in the initial sample) of them 21 are corporate PE and 98 are independent PE firms. Further we have excluded two corporate PE firms, which have invested in an unusual number of portfolio firms 63 and 32 (3^rd largest number is 24, while median is 2), to avoid the influence of these two outliers on the results of estimation. Our final sample includes 117 PE firms. Table 2 provides descriptive statistics for the sample.

2.3 – Variables

Control variables:

45The total number of portfolio companies in which PE firm has invested is our offset (or “exposure”) variable i.e. the number of portfolio companies “exposed” to successful exit.

46We measure the experience of a PE firm by its age, number of years from the founding date to 2016. An alternative to that might have been the total number of investments (different targets, different investment rounds) or total funds invested by the PE, however those variables are highly correlated with our offset variable, thus we have decided to include only the firm’s age, keeping in mind that partly the effect of PE’s experience is captured in the offset variable.

47Larger, more active PE firms invest in PE funds more frequently than smaller ones, thus for the size of the PE firm we use the firm’s average number of funds per year of operation, i.e. the total number of funds invested by the PE firm divided by its age.

48Following Hochberg, Ljungqvist and Lu (2007) we also include a variable that characterizes the intensiveness of competition among PE firms by the total inflow of capital to private equity market in the relevant year. We measure the inflow by the total amount of PE investments as the percentage of GDP, and we take the years in which the PE firm invested. Our variable is the average of the PE inflows over the years when the PE made its investments in portfolio companies.

49Finally, we included clustering coefficient, a variable that measures density of connections among the vertices adjacent to the given vertex. The clustering coefficient takes values between 0 and 1, where 0 correspond to a PE firm whose partners do not have ties among them i.e. never invested together, while value 1 correspond to a PE firm whose partners have direct connections i.e. any two partners have at least once invested in the same syndicate.

50Table 3 reports on the correlation between the variables in our sample. As in earlier studies with network centrality measures we observe strong and statistically significant correlation among those variables (Hochberg, Ljungqvist and Lu, 2007; Godlewski, Sanditov and Burger-Helmchen, 2012; Chuluun, 2015). Similarly to those studies, we run separate models for each of the centrality measures.

Table 3

Correlation matrix

(1) (2) (3) (4) (5) (6) (7) PE age (1) PE size (2) -0.202** CPE (3) 0.078 -0.268*** Competition (4) -0.245*** 0.030 -0.001 Clustering (5) -0.059 -0.216** -0.187* -0.020 Degree (6) 0.101 0.228*** 0.124 -0.041 -0.507*** Closeness (7) 0.070 0.205** 0.148 -0.048 -0.381*** 0.715*** Betweenness (8) 0.069 0.258*** 0.145 -0.007 -0.741*** 0.673*** 0.552***

Correlation matrix

^*p<0.1, ^**p<0.5, ^***p<0.01

2.4 – Model

51One approach to study the effect of the network centrality and control variables on success of PE investors is to follow Hochberg, Ljungqvist and Lu (2007) and use an OLS regression with exit rates as the dependent variable in fund-level analysis. Although it works well on the large sample of the US-based VC funds, applying it to our sample would be impractical. In addition to the problem of the modest size of our sample (117 PE firms in our sample vs. 3’469 VC funds in theirs), we must also take into account that most of French PE firms have made only few investment to the date (the median number of portfolio companies per PE in our sample is 2) implying that the exit rates for PE firms in our sample tend to be clustered around its extremes (0 and 1).

52Instead of estimating OLS regression with exit rates, proxies for the probability of successful exit, we decide to estimate that probability directly using a count regression model, the simplest of which is Poisson regression. Such approach is parsimonious and better fit to the nature of the data we analyze.

53In Poisson regression, we model the logarithm of the expected number of successes/failures in a series of experiments as a linear combination of some explanatory variables

54E(Y|x) + Y₀e^θX

55or

56log(E(Y|x)) = log(Y₀) + θX,

57where Y is the dependent variable, the number of events, X is the vector of explanatory variables, θ is the vector of estimated parameters and Y₀ is the offset (or “exposure”) variable which is equal to the number of experiments (exposed units under observation) and coefficient of which is not estimated, but set to be equal to 1.

58All regression reported below were tested using the regression-based test for overdispersion (Cameron and Trivedi, 1990) and none of the estimated models exhibits statistically significant overdispersion.