Investment, Duration, and Exit Strategies for Corporate and Independent Venture Capital Backed Startups∗ Bing Guo†, Yun Lou‡, and David P´erez-Castrillo§ September 30, 2011

Abstract We propose a model of investment, duration, and exit strategies for startups that takes into account first, the high level of uncertainty regarding returns from the investment in the startup, second, the more accurate information in the hands of insiders, and finally, the discount rate of the partners in the startups. According to our theoretical analysis, CVC backed startups stay longer in the market before exit and they invest more than those financed by IVCs. While longer duration leads to a higher likelihood of an exit through acquisition, a larger investment increases the probability of an IPO exit. These predictions find strong empirical support, using venture capital data from U.S.

JEL Classification: G32, G24. Keywords: Startups, Corporate Venture Capital, Independent Venture Capital, Investment Amount, Duration, Exit Strategy, IPO, Acquisition. ∗

We thank Albert Banal-Esta˜ nol, Gary Dushnisky, Mar´ıa Guti´errez, In´es MachoStadler, Philipp Meyer, Pau Olivella-Cunill, Pedro Rey-Biel, Jo Seldeslachts and participants at the MOVE workshop on venture capital, the 2011 LSE Alternative Investments Research Conference and the Symposium of Industrial Organization and Management Strategy in Chengdu 2011 for their helpful suggestions. We are grateful to AGAUR, research projects ECO2009-07616, 2009SGR-169, and INFOINNOVA (03513A), Barcelona GSE, and ICREA Academia for the financial support. P´erez-Castrillo is MOVE fellow. † Universidad Carlos III de Madrid, Business and Administration Department, [email protected]. ‡ London Business School, Accounting Department, [email protected]. § Universitat Aut` onoma de Barcelona, Economics Department, [email protected].

1

1

Introduction

Entrepreneurs and venture capitalists make investment decisions and choose the length of their involvement in a startup to maximize the chances of success and the value of their venture. They also look ahead and plan for a strategy of cashing in on their company allowing, in particular the venture capitalists, to liquidate their shares. Planning an exit strategy is as important as figuring out how to start the enterprise. There are two main exit routes for a successful startup: the company can go to an Initial Public Offering (IPO) or it can be sold to an existing firm (Acquisition).1 Under an IPO, the venture achieves a stock market listing so that it can receive additional financing for its projects and the insiders can eventually sell their shares to the public. If the startup is acquired, the insiders get immediate cash in return from their shares. The optimal exit route for startups depends on multiple factors, such as expected profitability of the venture; level of uncertainty; asymmetry of information between insiders and outsiders (potential buyers, new investors);2 possible conflicts of interest among insiders;3 venture capital characteristics, etc. Some of these factors are affected by the partners’ investment and duration decisions. Understanding the main trade-offs faced by startups at the exit stage is crucial because it allows to see how venture capitalists and entrepreneurs divest their companies, and also because of the impact of the (anticipated) exit strategy on the decisions taken at the onset of the venture. In this paper, we abstract from possible internal conflicts among insiders and we propose a model of investment, duration, and exit taking into account first, the high level of uncertainty regarding returns from the investment in the startup, second, the more accurate information in the hands of insiders, and finally, the discount rate of the partners in the startups. In our model, the level of investment of a startup influences its expected value. We assume that a higher investment leads to a more favorable distribution of the set of potential values. Furthermore, the decision on the 1

Two other exit routes that are not so commonly used are Management Buy-out and Refinancing (or secondary sale); see, for example Schwienbacher (2009). 2 See Cumming and MacIntosh (2003) for a discussion about the information asymmetries between sellers and potential purchasers of startups. 3 For a recent review of papers that analyze internal conflicts, see for example MachoStadler and P´erez-Castillo (2010)).

2

duration of the startup before exit, that is, the length of the relationship, affects the market information about the successful probability of the venture. Indeed, the level of uncertainty concerning the actual value of a startup is very high. Some of the uncertainty is resolved during the development stage and the market has access to that information. We assume that the level of the potential value of the venture at the time of the exit will be known by every market participants. Nevertheless, the insiders have more precise information about the expected profitability of the startup because they learn the probability of success. Whether the outsiders can be informed of such probability depends on how long the startup stays in the market before exit. We show that independently on the level of information received by the potential acquirers, the ventures whose probability of success is higher are more likely to try an IPO while those with lower probability prefer looking for an acquirer. Moreover, the likelihood of going to IPO increases with the potential value of the startup, if that value is high enough. Startups with low potential value are liquidated. We link the startup exit strategy with the investment decision and with the market level of information. First, a higher investment level brings about both, a higher likelihood of successful exit and a higher rate of IPO exits among the successful ones. Second, the IPO exit rate is lower when the outsiders receive more precise information, that is, when they are informed about the success probability. Finally, we analyze the optimal investment and duration decision of the startup. We show, in particular, that both the level of investment and the duration of the venture decrease with the discount rate of the startup. Our analysis allows to shed light on the discussion concerning the differences in behavior between the startups that receive financing from Corporate Venture Capital funds (CVCs) and those that are financed only by Independent Venture Capital funds (IVCs). Unlike the traditional IVCs, CVCs are private equity funds invested by large corporations. Therefore, while the sole objective of IVCs is financial return on capital, a very important goal of most CVC programs is strategic: the development of new, related business (see for instance Sykes, 1990; Yost and Devlin, 1993; Dushnitsky and Lenox, 2006; Hellmann, Lindsey and Puri, 2008). According to this strategic argument, CVC backed startups are more likely to exit by acquisition because the affiliated company of the CVCs may be particularly interested in the venture. This argument has 3

been confirmed by both theoretical studies (Hellmann, 2002; Riyanto and Schwienbacher, 2006) and empirical studies using European dataset (Cumming, 2008). Based on some questionnaires, Siegel, Siegel and MacMillan (1988) and Sykes (1990) find that the percentage of CVC backed startups that are acquired is higher than that of IVC backed ventures. However, it has been shown that the number of startups acquired by the company behind the CVC funds is small (Maula and Murrey, forthcoming) and our own analysis using the VentureXpert database confirms that the percentage of startups acquired by companies related to CVC investors is around 5%. Moreover, Gompers and Lerner (2000) and Chemmanur and Loutskina (2008) find that startups with CVC investments exit more often through IPO rather than by acquisition. We argue that another important difference between CVC and IVC funds is that IVC funds care more about quick exits than CVCs; that is, IVC backed startups have higher discount rate than those backed by CVCs. Indeed, compared to IVC funds, CVCs have more unused resources such as technology and marketing resources (Sahaym, Steensma and Barden, 2010; Basu, Phelps and Kotha, 2011; Da Gbadji, Gailly and Schwienbacher, 2011). Moreover, IVC managers’ payment is more based on financial returns and their ability to raise additional funds depends on their reputation, which is influenced by their history of successes (Gompers, 1996; Dushnitsky and Shapira, forthcoming). Therefore, they have strong incentives to cash their return from profitable projects early. According to our model, the difference in the discount rate between IVC and CVC backed startups results in different behavior between the two types of venture. First, a lower discount rate for CVC backed startups implies a higher investment level and longer duration. Second, although the identity of the VC fund does not have a direct effect on the choice of exit, it does have two strong indirect effects: on the one hand, the larger investment decided by CVCs leads to more IPO exits; on the other hand, the longer duration induces more exits through acquisition. Therefore, the two forces that we identify go in opposite directions. Finally, a larger investment implies a higher success rate for CVC backed startups. We empirically test our main results using data on 4801 US startups from the period 1969 to 2008. First, we find that startups financed by CVC funds invest around 25% more than those financed by IVC funds. Moreover, the effect is doubled if the syndicate leader of a startup is a CVC fund. Second, CVC backed startups do stay longer before the exit than IVC backed 4

startups. Third, one percent increase in the level of investment significantly increases the probability of IPO exit by 0.065%. Fourth, one percent increase in the duration of the venture significantly decreases the likelihood of IPO exit by 0.017%. Fifth, we show that, after controlling for the duration effect and for the level of investment, there is no significant difference in the rate of IPO exits between IVC and CVC backed startups. In fact, we observe that the presence of CVC investors has a positive, although not significant, effect on the IPO exit rate. All the previous empirical results strongly support the predictions of the theoretical model. Finally, we use an enlarged dataset to test the theoretical results concerning the influence of the level of investment, the duration and the fund’s characteristics on the successful rate. As suggested by our model, duration and fund’s characteristics do not have a significant influence in the rate of successful exists. However, the effect of investment, while positive, is also not significant. The rest of the paper is organized as follows. In Section 2, we introduce the model. In Section 3, we develop the analysis of the optimal exit strategy. In Section 4, we analyze the impact of investment and duration decisions on the likelihood of IPO and on the success rate. In section 5, we study the optimal investment and duration decisions. In Section 6, we derive the empirical hypotheses concerning the differences in behavior between CVC and IVC backed startups suggested by our theoretical results. In Section 7, we describe the dataset, which we use to empirically test the hypotheses in Section 8. Finally, Section 9 concludes. All the proofs are included in the Appendix.

2

The Model

We analyze the optimal investment and duration decisions and exit strategy of startups (S). In our model, startups’ decisions at any stage aim to maximize the expected discounted profits. The main characteristics of the model are the following. The first decisions taken by a startup concern the level of investment I and the duration of the venture d. We consider that these decisions are made at the beginning of the life of the startup and we do not take into account their dynamic aspect, which is not relevant for our purpose. The level of investment has a positive impact on the expected quality of the venture while the duration influences the amount of information that will flow to the market.

5

The optimal investment and duration decisions depend on the discount rate r of the startups. The parameter r will play an important role in our analysis because, in the empirical sections of the paper, we will differentiate between startups receiving CVC funding and those receiving only IVC funding. As we argued in the Introduction, CVC funds tend to be more patient than IVC funds. When the startup makes its first decisions, there is a high degree of uncertainty with respect to both, the potential value of the venture V and the probability p of being able to realize this value. Part of the uncertainty is resolved as the startup develops. All the market participants will be able to observe some of the information, but the insiders will acquire more precise information on the expected quality of the project. In our model, we reflect this asymmetry in the information between insiders and outsiders in a simple way. When it is revealed, everybody can observe the potential value V . Moreover, the insiders always learn the probability p. However, the precision of the information received by the outsiders about p depends on the duration of the venture: the longer d, the more precise the outsiders’ information. The venture requires additional financing C to possibly achieve the value V . Hence, if the potential value V is low, the startup will be liquidated (this is the first exit option). If, on the contrary, continuing the venture is profitable, then the startup will either look for a firm (an acquirer) interested in adding the venture into its business, or it will go to an initial public offering (IPO). In the first case, the acquirer will offer a deal to the startup that will reflect the expected value of the business and the bargaining power of the parties. Then, the acquirer will integrate the venture into its organization and, when it confirms that it is worthwhile doing it, it will make the additional financing to obtain V . In case the startup tries an IPO, then the market investors will go through a thorough analysis concerning all possible aspects of the startup. The market investors will make a careful auditing of the corporate valuation, market prospection and so on. The outcome of the analysis will be a new signal on the profitability of the startup that we model also in a simple way: either the market makers are convinced that the startup will be successful with probability 1 (High signal), or they will still not be able to assess it with certainty (Low signal). All these processes are costly and the startup needs to cover the cost. The market investors will make an offer to the startup owners in case of a High signal.

6

More precisely, the model is the following: At t = 1, the startup decides the level of investment I and the duration d. • The level of I determines the distribution of the potential value of the startup V : the value V follows a distribution function Γ(V ; I), with density function γ(V ; I). For simplicity, � we assume that � ex ante V is uniformly distributed over the interval f (I), V + f (I) , where f (I) is an increasing function of I. The potential value V can only be cashed if at later stages a fixed new funding C is made. After the investment and before all the other decisions are taken, the value of V is realized and it is observable by everybody. • The level of d determines the information learnt by the potential acquirers about a signal p on the likelihood of success, i.e., the probability of realizing V . Also for simplicity, we assume that ex ante p is uniformly distributed over the interval [0, 1]. The startup always learns p. The potential acquirers will learn p with probability h(d), increasing in the duration d. At t = 2, the startup takes the exit decision. It has three possibilities: liquidation, looking for an acquirer, or going to an IPO. • The liquidation value of the startup is always 0. • In case it decides to look for an acquirer, then a deal price is negotiated, depending on the bargaining power of the two parties and on the acquirer’s information. The bargaining power for the startup is denoted by m. In case of acquisition, the acquirer will invest C to realize V if it confirms that the project is successful. • Going to an IPO is the most complex and costly exit route for the startup. We denote by F all the fixed costs due to the IPO process. It leads to one public signal β� on the profitability of the venture, β� ∈ {H, L}. We assume that β is the probability for the market to be able to verify a successful project after receiving the public signal. Therefore, the probability of observing β� = H is equal to βp. In case the signal is H, the competitive market will set a price Z for IPO. If the startup accepts the price, then a successful IPO is carried out. In order to realize V , in addition to Z, the market needs to raise C to cover the 7

Figure 1: The Time Line

remaining investment needs. In case the signal is L, then no offer is issued.4 The time line is captured by Figure 1. We make assumptions on the functions f (I) and h(d) that make sure F that the three exit routes are possible. We assume that Vˆ + f (0) > C + β−m and that f (I) is concave enough, in particular limI−→∞ f (I) ≤ C. Also, the screening is informative enough: β > m. Finally, the venture makes interior choices of I and d if limI−→0 f � (0) = limd−→0 h� (0) = ∞ and h(d) is concave enough.5 4

We assume that a startup that receives a low signal does not get any offer and quits the market. We make this assumption for simplicity. First, for those startups that receive a signal L, the situation is often similar to the lemon’s market in Akerlof (1970)’s model: there is no price under which market profits are non negative (taking into account the startups that accept that offer). Therefore, the assumption that IPOs do not make offers to startups that receive low signals can be sustained as a result of a more general model. Second, the startups may go to the acquisition market (at t = 3) once they fail at IPO, where the acquirers will take into account the new information produced at IPO. This adds some (small) additional profits to those ventures that choose the IPO exit. However, the qualitative results of our analysis do not change if we add this possibility. (For an analysis of the previous extensions, see Guo, 2010). F 5 If β < m, then the IPO is never chosen. If Vˆ + f (0) < C + β−m , then some exit routes are never taken for low investment levels while liquidation never happens for high investment levels if limI−→∞ f (I) > C. However, our qualitative results would remain with changes in the hypotheses. Similarly, if limI−→0 f � (0) = limd−→0 h� (0) < ∞, then

8

We solve the model by backward induction, taking into account that there may exist asymmetric information among the participants. Therefore, we use sequential equilibrium as the solution concept, since it combines subgame perfection ideas with Bayesian updating.

3

The Analysis of the Optimal Exit Strategy

In this section, we start at t = 2, where the duration has already been decided and the investment made at t = 1 is sunk. The potential value for the venture V is realized and observed by all the participants. Moreover, the startup has received the private signal concerning the probability of success p. The potential acquirer may also know p (this happens with probability h(d)) or not. We study the optimal exit strategy in both situations.

3.1

Optimal Exit Strategy with Informed Outsiders

As mentioned in the previous section, the value of the startup in case of liquidation is 0. Also, the deal price in an acquisition corresponds to a share m of the expected value of the venture. Taking into account that the acquirer knows p and that he needs to invest C to realize V when the venture is believed to be successful, the expected value of the venture is p [V − C]. Therefore, if the startup goes to the acquisition market at t = 2, the deal price is mp [V − C], whenever V − C > 0. Consider now a startup characterized by (V, p) that goes through an IPO, with V − C > 0 (otherwise, profits are always non-positive).6 After the � If the realization is β� = L, startup pays F , the market receives the signal β. which happens with probability (1 − βp), it will not receive any offer. If the realization is β� = H, then the competitive market of investors will offer Z = V − C, which the startup will accept. The startup obtains higher expected profits going to an IPO than looking for an acquirer if and only if βp [V − C] − F ≥ mp [V − C] .

(1)

the optimal decision on I and/or d may be the corner solution I = 0 and/or d = 0, which would complicate the analysis without adding new insights. 6 We take the convention that a startup indifferent between being liquidated and not chooses liquidation. Similarly, a startup indifferent between going to an IPO and looking for an aquirer at t = 2 goes to an IPO.

9

Proposition 1, whose proof follows the previous discussion, describes the optimal exit strategy for a startup characterized by (V, p) when outsiders are informed about p, where we denote � � 1 F po (V ) ≡ min ,1 . (2) [β − m] [V − C] Proposition 1. Consider a startup characterized by (V, p) in a situation where potential acquirers have learned p. The startup’s optimal exit strategy is as follows: 1. If V − C � 0, the startup is liquidated and gets the payoff Uo (V, p) = 0. 2. If V − C > 0 and p < po (V ), the startup goes to the acquisition market and gets a deal value Uo (V, p) = mp [V − C]. 3. If V − C > 0 and p � po (V ), the startup invests F and goes to the IPO market. Moreover, (a) if it gets the public signal H, then it receives an offer Z = V − C from the IPO market and it accepts it; (b) if it gets the public signal L, then it does not receive any offer from the IPO market. Therefore, in this case, the startup’s expected payoff is Uo (V, p) = βp [V − C] − F . The basic trade-off between IPO and acquisition is that while the IPO process is very costly, it also allows the startup owners to get a larger share of the value of profitable ventures. Startups with high enough probability of success are ready to pay the cost of the process. To analyze the effect of the different parameters on this trade off, we conclude the analysis of the optimal exit strategy by doing the comparative statics of po (V ) with respect to all 1 F the parameters, for the interior case where [β−m] < 1. This analysis [V −C] highlights the characteristics of the startups and the market that make it more likely to observe exits through IPO or through acquisition. Indeed, a higher po (V ) implies a lower likelihood of exit through IPO. Corollary 1. Consider situations where potential acquirers learn p. Then, the likelihood of IPO increases with V and β and it decreases with F , C, and m. 10

According to Corollary 1, the higher the potential value V (similarly, the lower the additional funding C), the more willing is the startup to go to the IPO market. Given the costly IPO process, only those startups that really benefit from the more competitive IPO market are willing to follow this path. As expected, a higher bargaining power m in the acquisition market leads to less IPO exits. Finally, an efficient IPO process, reflected by a low cost F and powerful screening capability β, makes IPO an appealing exit.

3.2

Optimal Exit Strategy with Uninformed Outsiders

The analysis of the optimal strategy of a startup that looks for an exit when the potential acquirers do not know the value of p has some similarities with the one developed previously. First, if the startup’s potential value V is lower than C it is liquidated. Second, the IPO offer is Z = V − C if it receives a signal β� = H which, in particular, implies that the startup will accept the offer. Finally, potential profits from IPO versus Acquisition increase with the value of p; therefore, there will be a cut-off equilibrium value poo (V ) (that can possibly be equal to 0 or 1) above which the startup goes to IPO. The main new aspect when the value of p is unknown by the potential acquirers is that the price that they may offer does not depend on the real value of p but on its expected value from the point of view of the acquirer, which is a function of the startup equilibrium behavior. The deal that a potential acquirer will make to a startup that approaches it at t = 2 will not be based on the true probability but on the expected (equilibrium) value of p, which is poo2(V ) . Therefore, the deal price at t = 2 will be m poo2(V ) [V − C]. Similar to the informed outsiders’ case, a startup whose probability of success is equal to poo (V ) must be indifferent (if poo (V ) ∈ (0, 1)) between going to IPO and looking for an acquirer. Therefore, an interior poo (V ) is characterized by βpoo (V ) [V − C] − F = m

poo (V ) [V − C] . 2

Equation (3) implies that the cut-off value is � � 1 F � poo (V ) ≡ min � ,1 . β − m2 [V − C]

(3)

(4)

For completeness, we state the equilibrium behavior of startups when outsiders are uninformed as to the value of p in Proposition 2. 11

Proposition 2. Consider a startup characterized by (V, p) in a situation where potential acquirers have not learned p. The startup’ equilibrium exit strategy is as follows: 1. If V −C � 0, the startup is liquidated and gets the payoff Uoo (V, p) = 0. 2. If V − C > 0 and p < poo (V ), the startup goes to the acquiring market and gets a deal value Uoo (V, p) = m poo2(V ) [V − C]. 3. If V − C > 0 and p � poo (V ), the startup invests F and goes to the IPO market. Moreover, (a) if it gets public signal H, then it receives an offer Z = V − C from the IPO market and it accepts it; (b) if it gets public signal L, then it does not receive any offer from the IPO market. Therefore, in this case, Uoo (V, p) = βp [V − C] − F . Moreover, at equilibrium, the likelihood of IPO increases with V and β and it decreases with F , C, and m. The intuitions behind Proposition 2 are the same as those explained after Proposition 1 and Corollary 1.

4

The Impact of Investment and Duration on the Likelihood of IPO and on the Rate of Success

We now analyze how the optimal exit strategy of a startup is influenced by its investment and duration strategies.

4.1

The Investment Effect

The level of investment affects the likelihood of an IPO exit. A higher investment level implies a shift in the distribution of V towards higher values. As shown in propositions 1 and 2 (see also Figure 2), the higher the value V , the more likely it is that the exit happens through an IPO rather than through acquisition, independently on whether potential acquirers have learned p. Therefore, a higher investment level should imply a higher IPO rate among 12

successful stories. Proposition 3 states this result. Proposition 3. The likelihood of IPO among the successful exits increases with I both, when the potential acquirers have learned p and when they have not learned p. Moreover, the shift in the distribution of V towards higher values implied by a larger investment should also lead to higher likelihood of successful exit. Proposition 4 states this result, whose proof is immediate. Proposition 4. The rate of successful exits increases with I.

4.2

The Duration Effect

Longer duration means that the market has more precise information which, in our model, is reflected by a higher probability for the public to know the successful probability p. The next proposition investigates whether an IPO exit is more likely with informed or with uninformed outsiders. The proof of the proposition is straightforward from equations (2) and (4). Proposition 5. The probability of going to IPO is higher when the potential acquirers have not learned p than when they have learned p. More precisely, poo (V ) < po (V ) whenever poo (V ) ∈ (0, 1), and po (V ) = 1 whenever poo (V ) = 1. Figure 2 helps to explain Proposition 5. It depicts the optimal exit strategy highlighted in propositions 1 and 2. For high values of p and V , IPO is the optimal exit route independently on the potential acquirers’ information. Similarly, going to the acquisition market is always the optimal startups’ strategy for low values of p and V (provided V > C). In the intermediate (shadow) region of Figure 2, startups go to IPO when the outsiders have not learned p, while they prefer going to the acquisition market if outsiders have observed p. The intuition for the existence of the intermediate region in Figure 2 is the following. When the outside acquirers observe the true value of p, they offer a deal according to p. However, when they do not observe the true successful probability, they can only offer a deal according to the expected probability, which is independent of the true value p. Consider a startup whose realized probability is po (V ), that is, it is indifferent between IPO and acquisition if information about p is public. The deal it will obtain in the 13

Figure 2: Optimal Exit Strategy

acquisition market if information is not public is lower, as it is based on the expected probability. Therefore, it would rather go to IPO than look for an acquirer. As a consequence, uninformed markets are more likely to lead to IPO exits. Longer duration leads to more information about p for outsiders. According to Proposition 5, this implies, ceteras paribus, a reduction in the likelihood of IPO exit. We state this result in the following corollary. Corollary 2. The likelihood of IPO among the successful exits decreases with d. We note that, according to our model, the duration has no effect on the level of V . Therefore, the successful rate is not influenced by the duration d.

14

5

Analysis of the Optimal Investment and Duration Decisions

We address now the optimal initial decisions by the startup at t = 1. The analysis of Section 3 allows computing the expected income Uo (V, p) or Uoo (V, p) of a startup whose potential, publicly known value is V and whose probability of success is p, depending on the level of information by the outsiders. We now calculate the expected profits for a given duration d and a given investment I, denoted as U (d, I), which requires taking the expectation of the expected income over the possible values of V (whose distribution function Γ(V ; I) depends on I) and p: � � −rd U (d, I) = e [h(d)Uo (V, p) + [1 − h(d)] Uoo (V, p)] dpdΓ(V ; I) − I (5) V p −rd = e [h(d)EUo (I) + (1 − h(d))EUoo (I)] − I where we denote EUo (I) =

� � V

EUoo (I) =

� � V

Uo (V, p)dpdΓ(V ; I),

(6)

Uoo (V, p)dpdΓ(V ; I).

(7)

p

p

We can interpret EUo (I) as the expected profits at the time of exit of a startup that has invested I at t = 0 and whose realized probability p is known by potential acquirers. Similarly, EUoo (I) is the startup’s expected profits if the realization of p is unknown by outsiders. The probability that the information is known by the potential acquirers, h(d) increases with the duration. Moreover, the longer d, the lower the expected profits at t = 0 due to the discounting. Lemma 1 shows that EUo (I) is always higher than EUoo (I), that is, the startup’s expected profits are higher when the potential acquirers learn p than when they do not. This result, interesting by itself, will be useful in the analysis of the optimal duration and investment decisions. Lemma 1. EUo (I) ≥ EUoo (I) for every I > 0, and the inequality is strict whenever poo (V + f (I)) < 1. The intuition behind Lemma 1 is the following. The asymmetry of information between startups and potential acquirers makes it more profitable 15

for some ventures (those whose successful probability lies in the interval (poo (V ), po (V ))) to go to the IPO market although they would obtain a better deal in the acquisition market if the information were symmetric. Therefore, in expectation, startup’s profits are higher when information about p reaches the potential acquirers, that is, EUo (I) ≥ EUoo (I). The optimal duration and investment decisions (d∗ , I ∗ ) are interior. Therefore, they satisfy the first-order conditions ∂U (d∗ , I ∗ ) = ∂U (d∗ , I ∗ ) = 0. In ∂d ∂I Proposition 6, we study the effect of the discount rate r on (d∗ , I ∗ ). Proposition 6. The optimal duration d∗ and investment I ∗ decrease with the discount rate r. The result in Proposition 6 is quite intuitive. If the participants in the startups care less about the future, they invest less and stay a shorter period of time in the ventures. Proposition 6 allows us to discuss the effect of r on the likelihood of IPO exit. The discount rate r indirectly influences the likelihood of IPO exit through the investment effect and the duration effect. A lower discount rate r means both higher investment level and longer duration which, according to Proposition 3 and Corollary 2, have opposite impacts on the likelihood of IPO exit. On the one hand, a larger investment leads to more exits through IPO. On the other hand, a longer duration implies a better informed potential acquirer, which leads to more acquisitions.

6

CVC vs IVC Backed Startups: Empirical Implications from the Theoretical Model

The analysis of the previous sections allows us to contribute to the discussion of the differences between startups that receive funds from CVC and those that only receive IVC funding. CVCs are typically less compelled to recover the investment earlier. We associate this difference with a lower discount rate for startups that receive CVC funding. According to our analysis, the difference in the discount rate between CVC and IVC backed startups has testable implications on their strategies. We use our theoretical results to propose empirical hypotheses concerning the differences in investment amount, duration before exit, exit strategy, and successful rate between CVC and IVC backed 16

startups. First, one implication of our model (see Proposition 6) is that startups with lower discount rate choose higher investment levels. Therefore, we should observe higher investments in CVC backed startups than in IVC backed startups. With more unused resources and having strategic aim, CVC funds invest more in the startup projects. We state this empirical implication in the following hypothesis. Hypothesis 1: CVC backed startups receive a higher investment amount than IVC backed startups. Second, Proposition 6 also implies that CVC backed startups (having lower discount rate) choose to stay longer before exit. Since CVC funds are more patient than IVC funds, we expect that CVC invested startups have a longer duration before exit than those backed by IVC funds. This is reflected in Hypothesis 2. Hypothesis 2: CVC backed startups have a longer duration than IVC backed startups. Third, our theoretical model indicates an indirect impact of VC funds’ characteristics on startups’ exit strategy. As we mentioned before, startups with CVC backing invest more in their projects than those with IVCs. The theoretical model predicts that a higher investment level increases the probability of an IPO exit (see Proposition 3). Furthermore, a startup with CVC backing also stays longer and, after Corollary 2, longer duration reduces the probability of IPO exits. According to our model, once we take into account the effect of investment and duration, the characteristics of the VC funds do not play a role in the choice of the exit strategy. We will test the theoretical results through Hypotheses 3, 4 and 5. Hypothesis 3: Investment amount has a positive effect on the probability of IPO exit. Hypothesis 4: Duration of a startup before exit has a negative effect on the probability of IPO exit. Hypothesis 5: CVC backed startups have the same probability of IPO exit as IVC backed startups.

17

Fourth, higher investments imply higher successful exit rates (see Proposition 4) while the duration and the characteristics of the fund do not have any direct effect on the successful rate. Hypotheses 6, 7 and 8 state these implications from our theoretical model. Hypothesis 6: Investment amount has a positive effect on the probability of successful exit. Hypothesis 7: Duration of a startup before exit has no effect on the probability of successful exit. Hypothesis 8: CVC backed startups have the same probability of successful exit as IVC backed startups.

7

Data Description

We obtain the relevant data, i.e., investment amount, IPO date, acquisition date, investment rounds, number of investors, investors’ type, IPO price, Acquisition deal value, VC fund size, etc, from the VentureXpert database. To test the first five hypotheses, we use a dataset that contains 4801 successful startups in the US market from 1969 to 2008. We describe this dataset in detail. In Subsection 7.5, we will comment on the enlarged dataset that includes unsuccessful startups and successful startups with other types of exit besides IPO and Acquisition, which we use to test the last three hypotheses. Our sample of successful startups covers 63 industries in the US. Table 1 provides an overview of the industry composition of the sample (by twodigit SIC code). In the table, we just include industries with more than 20 observations. We observe concentration of industries with SIC code 28, 35, 36, 38 and 73. These codes correspond to Chemical, Electronic and Business Service related industries, where venture capital investments are more common.

7.1

Dependent Variables

We use the dataset with successful startups to test the effect of VC funds’ characteristics (CVC or IVC) on investment amount, startups’ duration and 18

exit strategies. Therefore, we need three dependent variables: the total investment amount, startups’ duration before exit, and the exit rule (IPO or Acquisition). To measure the investment amount of a startup, we sum up the round level investments to get the total investment amount. The duration before exit is measured by the number of days. It is calculated as the difference between the exit date (IPO date or Acquisition date) and the date at which a startup receives the first investment from venture capital firms. Finally, the exit strategy of a startup is indicated by a dummy variable. It is equal to 1 if a startup exits through an IPO and 0 if it exits through an Acquisition.

7.2

Independent Variables

The theoretical model predicts the influence of different VC funds’ characteristics on startups’ decisions and exit behavior. The most important independent variable is whether a startup is financed with CVC funds or IVC funds. There are two definitions of CVC backed startups in the literature (Toldra, 2010). Under the first definition, a startup is defined as CVC financed if all the investments are from CVC funds. The second defines a startup as CVC backed if at least one of the investors is a corporate venture fund. We use two variables to measure the VC fund characteristics. Firstly, we use the second definition in Toldra (2010) and create a dummy variable that is equal to 1 if a startup receives at least one investment from CVCs. If all the investments are from IVC funds, the startup is IVC financed and the dummy variable is equal to 0.7 Secondly, we use a continuous variable to measure the funds’ characteristics: we calculate the percentage of investment made by CVC funds out of total investment in each startup.

7.3

Controlling Variables

Based on the previous literature, we use a set of controlling variables to estimate the effect of VC funds characteristics on the investment amount of startup projects, their duration before exit, and exit strategies. It includes the average VC fund size, average VC fund age, total number of investment rounds, VC syndicate size, syndicate leader characteristics (i.e., whether the leader is a CVC fund or an IVC fund), industry market-to-book value, the 7

We can not use the first definition of CVC in Toldra (2010) because the number of startups that are fully funded by CVCs is very small.

19

relationship between CVC funds and their invested startups, 3 months and 6 months MSCI return8 before the exit date, the industry fixed effect, and the year fixed effect. We also take into account the relative controlling power between the entrepreneur and the VC funds through the Later Stage Dummy variable that takes value of 1 if the startup is at expansion or later stages at the exit. As stated by Smith (2005) and Schwienbacher (2009), VC funds guarantee themselves more controlling rights at the beginning of their involvement and for seed financing than at late-stage financing. Hence, exiting at expansion or later stages proxies for higher entrepreneurs’ controlling power.9 Table 2 provides the definitions of all the variables and Table 3 summarizes the basic statistics of these variables.

7.4

Correlations across Variables

Table 4 reports the correlation matrix. It represents the correlations across the main variables. The variables in our estimation are not highly correlated, except the 3 months and 6 months MSCI returns. In particular, the correlation between Duration and Investment Amount is 0.07. This low correlation suggests low multicollinearity of the independent variables.

7.5

Enlarged Dataset with Unsuccessful Startups

The last three hypotheses estimate the influence of the level of investment, the duration and the fund’s characteristics on the successful rate. We test these hypotheses with an enlarged dataset which includes both successful and unsuccessful startups in the U.S. before the year 2009. We define a startup as a “Failure” if the company status is “Defunction” in the database of VentureXpert. On the contrary, a startup is a “Success” if the status is “Acquisition”, “Merger”, “Went Public”, “LBO”, etc. The enlarged dataset has a similar industry distribution to the one which only contains startups successfully exited through IPO or Acquisition. Around 65% of the startups have a successful exit and 35% are defined as “Defunction” in the dataset.10 8

The Morgan Stanley Capital International (MSCI) constructs a free float-adjusted market capitalization weighted index that measures the equity market performance of developed and emerging markets. We use the MSCI ACWI (All Country World Index) Index of the United States in the paper. 9 We use this dummy variable because we can not measure directly the relative controlling power between the entrepreneurs and VC funds due to limited data. 10 The successful rate of the startups in our dataset is very high. This might due to the restrictions of the database. VentureXpert database does not have the information of all the failure cases in the U.S. market.

20

We create a new dependent variable F ailurei for the empirical estimation. It is a dummy variable equal to 1 if the startup is a “Failure” as was defined before and 0 if it is a “Success”. The independent variables are the same as those described in Section 7.2. We exclude some controlling variables, such as the industry market-to-book value, fund age, due to data insufficiency. From VentureXpert, we can not observe the date that startups are considered as “Defunction”. Hence, all the controlling variables that need the exact date are excluded in the regressions of the last three hypotheses. The duration of the startups in this case is defined as the number of days between the first and the last investment dates.

8

Empirical Analysis and Results

We first look at whether startups backed by CVC funds and those financed by IVC funds have different behavior. We provide some basic statistic differences between CVC and IVC financing in Panel A of Table 5. We observe significant differences between startups with CVC funds and with IVC funds in most variables. CVC backing implies a significantly higher investment than IVC backing. The average investment per venture for both exit strategies is around 50 million USD for CVC backed startups while it is only around 21 million USD for those only backed by IVCs. This fact has already been highlighted by previous literature (Gompers and Lerner, 2000). Moreover, there is a large gap between the mean duration of the two types of venture. The mean duration for CVC backed startups is 1929 days, compared to 1649 days for IVC backed startups. We also find that CVC financing leads to more investment rounds. CVC backed startups exit at later investment stages (i.e. more exits at expansion or later stages than exits at seeds or early stages). Compared to IVC funds, CVC funds are older. However, we have not found any significant difference in IPO exit rate and VC fund size between the two types of VC funds.

8.1

Fund’s Characteristics and Investment Strategy

The difference in the investment amount might come either from differences in the type of projects in which the funds invest (selection bias), or from the intrinsic characteristics of the type of fund, such as the discount rate. We use the dataset to confirm that the selection bias does not seem important: it is only when CVCs enter the startups that there is a change in the investment 21

amount. This analysis is included in Panel B and Panel C of Table 5. Table 5 (Panel B) depicts the number of startups that receive funds from the two types of VCs. The columns stand for different investment rounds and the lines are groups of startups, differentiated by the round in which CVC investors enter. We provide results until group 8, in which CVCs enter the project at the eighth investment round.11 The highlighted numbers represent the number of survival startups when CVC investors join in the venture. For example, the first line (Gr.0) describes the group of startups that only receive IVC funds. There are 2778 of them, out of which 2117 also receive second round financing, 1492 receive third round financing, and so on. The third line (Gr.2) includes the group of startups that start receiving CVC funds at the second investment round. There are 415 of them, out of which 291 also receive third round financing, and so on.12 It is worth highlighting that most CVC backed startups start receiving CVC financing at very early stages. One third of them (548 out of 1792) receive CVC funds at the first round, and almost 55% of them get CVC financing at the first two rounds. This is somehow at odds with previous findings suggesting that CVCs often enter at later investment rounds (Hellmann, Lindsey and Puri, 2008; Dushnitsky and Shapira, forthcoming; Masulis and Nahata, 2009). More interestingly, Table 5 (Panel C) shows the average investment per round and per group. Before CVCs enter the ventures, the investment amount is similar for all groups. For example, startups in Group 0 (that never receive CVC funds) invest almost 5.4 million USD in round 1, comparable with the 5.47 million USD of those that will receive CVC backing in round 2. However, these numbers are quite lower than 8.90 million USD received by startups backed by CVCs at round 1. A similar effect appears for all the rounds. Hence, before CVC investors join in the ventures, IVCs invest in similar projects, suggesting no obvious selection bias among the projects. The investment levels are significantly increased when CVCs enter into the startups. To see whether the intrinsic characteristics of the type of fund has an effect on the investment decision, we test Hypothesis 1 using the following 11

There are startups where CVCs only enter after the eighth round. Since the number of these ventures is small, we don’t show the details of those cases. 12 The number just before 415 should have been 415. However, it is only 401 due to missing data. A similar problem appears in other lines.

22

model: ln Investi = α0I + α1I F und� sCharacteristics +

7 �

αkI Zk + �i

(8)

k=1

In equation (8), Investi is the total investment amount at the startup level. F und� sCharacteristics measures whether the startup is financed by CVC funds or IVC funds. Zk is a set of controlling variables, including number of investment rounds, VC fund size, VC syndicate size, syndicate leader characteristics, industry market-to-book value, industry and year fixed effect. To obtain a robust estimation of how venture capital funds influence startups’ investment amount, we have estimated four models. Model 1 We use the dummy variable CV C for the funds’ characteristics. Model 2 It still uses the dummy variable CV C and it also includes the controlling variable of CVC strategic relationship. The variable measures whether the corporation behind CVC funds is a potential competitor to the invested startups or not. It is included according to Masulis and Nahata (2009). They indicate that because of the strategic aim of CVC funds, they can be competitors to the startups in the future. Therefore, startups ask for a higher investment from CVC funds than from IVC funds in order to be compensated for potential market competition. Model 3 It includes an additional controlling variable: the average fund age across all the investing funds in a startup. More mature VC funds have more experience, better reputation and richer resources. Therefore, they can invest more into startups. Model 4 It uses the percentage of investment made by CVCs (CV C per) as an indicator of CVC backed startups. A higher value means that the startup is more CVC oriented. The results of an OLS regression on the models are reported in Table 6. The four models provide similar estimated results and they give strong support to Hypothesis 1. CVC funds have a significantly positive impact on the total investment amount of startups. Startups financed by CVC funds invest 25% more than those financed by IVC funds if VC fund’s characteristic is attributed by the dummy variable. Moreover, if the syndicate leader of a startup is a CVC fund, the startup receives an additional 25% increase in its level of investment. Similarly, if the CVC investment amount as a percentage of the total investment for a startup increases by 1%, the startup 23

receives 0.28% additional investment. Moreover, more investment rounds, larger fund size, more mature VC funds, larger syndicate size and higher industrial market-to-book value lead to more investment in startups. We do not find any significant effect of the corporate venture capital funds’ relationship with the startups (competitive or not) on startups’ investment amount.

8.2

Funds’ Characteristics and Duration

Our theoretical model predicts that startups financed by CVC funds stay longer in the market before exit than those financed by IVC funds. The following survival model describes the estimation for Hypothesis 2. Duration = α0D + α1D F und� sCharacteristics +

6 �

α k Zk + � i

(9)

k=1

The dependent variable in Equation (9) is the duration of the startups before exit, which is measured by the number of days between the exit date and the first investment date. The independent variable indicates whether the startup is invested by only IVC funds or by CVC funds. To estimate the duration model, we have the same controlling variables as in the Equation (8) and the industry fixed effect. As before, we have estimated four regression models. We use the dummy variable CVC to indicate that a startup is financed by CVC funds for Models 1 to 3. In Models 2 and 3, we add the controlling variables that measure the competition relation between the startups and the parent company of CVC funds and the VC fund age respectively in the regression. We then use the CVC percentage investment as the dependent variable in Model 4. We estimate the survival model in a parametric way and assuming a Weibull distribution of the residual values.13 The hazard rates of the regression are shown in Table 7. As is suggested by Hypothesis 2, CVC backed startups do stay longer before the exit than IVC backed startups. Moreover, startups receiving finance from larger and more mature VC funds have a longer duration. Interestingly, more investment rounds reduce the duration for startups. 13

The assumption of a Weibull distribution is based on the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC). We also tried the semi-parametric estimation (COX estimation) of the survival model. However, the post estimation PH test rejected the proportional hazards assumption of the COX estimation.

24

Finally, the impact of the relationship between CVCs and startups on duration is not significant. This, together with the similar finding concerning the level of investment, suggests that the behavior of CVC funds in the startups in the same industry as the parent corporation is the same as their behavior in those startups in other industries.

8.3

Funds’ Characteristics and Exit Strategy

The indirect impact of the characteristics of venture capital funds (via investment and via duration before exit) on the exit strategy for startups is conveyed by Hypotheses 3 to 5. The following simple regression summarizes the test of the hypotheses: Exiti = α0 + α1Ex F und� sCharacteristics + α2Ex ln Duration(Days)i +

α3Ex

ln Investi +

9 �

αkEx Zk + �i

(10)

k=1

The dependent variable Exiti is a dummy variable, with value of 1 if it is IPO exit and 0 for an Acquisition exit. Invest and Duration(Days) have the same definition as in Equations (8) and (9). The set of controlling variables is similar to the previous regression, except that we include three additional controlling variables: 3-month and 6-month MSCI index, and later stage dummy variable. The first two variables are used to control the stock market conditions before the exit date: a strong stock market before the exit date may increase the probability of an IPO exit. The later stage dummy variable is a proxy for the relative control between entrepreneurs and VC funds.14 If startups exit at some later stages (i.e. the expansion or the later stage), entrepreneurs have more controlling power. Hence, it is a dummy variable equal to 1 for expansion or later stage, and 0 for seed or early stage. To check the robustness of our results, we estimate six models based on Equation (10). Model 1 The first estimated model is an OLS regression that uses the dummy variable CV C to measure the funds’ characteristics. Model 2 It includes the average fund age as controlling variable. Model 3 It considers the percentage of CVC investment out of the total investment amount as the measure of the funds’ characteristics. 14

Aghion and Bolton (1992) and Cumming (2008) point out that the relative controlling power between the entrepreneur of a startup and VC funds influences the exit strategy. Entrepreneurs prefer IPO exit while VC funds could vote for Acquisition exit.

25

Model 4 Since the dependent variable in the estimation model is a dummy variable, we use a Logistic estimation in Model 4 to check the robustness of our OLS estimation. Model 5 We test the possibility of a quadratic effect of the duration on the IPO likelihood for startups in model 5.15 Model 6 It estimates the effect of funds’ characteristics on startups’ exit strategy, without the control of duration effect or investment effect. The results are described in Table 8. The effect of CVC fund’s characteristics on the IPO exit is not statistically significant, using either the CVC dummy variable or the percentage investment of CVC as an explanatory variable. This conclusion provides support for Hypothesis 5. However, CVC funds have an influence on the exit strategy through the startups’ investment amount and the duration. Investment amount has a significantly positive effect on the probability of IPO. We observe that one percent increase in the investment amount increases the probability of IPO exit by 0.068%. Moreover, longer duration means a significantly lower probability of IPO. One percent increase in the duration will decrease the probability of IPO by 0.019%.16 These results are robust for all the first four models, with both OLS estimation and Logistic estimation. They provide strong support to Hypotheses 3 and 4. The results of Model 6 confirm that the duration effect and the investment effect explain the impact of VC funds’ characteristics on the exit strategy. Specifically, the investment effect dominates the duration effect, because CVC funds increase the likelihood of IPO exits when neither duration nor investment effect is controlled. Table 8 also provides indirect support to the idea that the discount rate has a strong influence on the startups’ decisions. We note that the effect of VC fund size on the probability of IPO is not statistically significant when both the duration and investment effects are controlled. However, if the two effects are not controlled, we observe that larger VC funds significantly increase the probability of IPO exits. The effect of VC fund size is particularly interesting because it seems reasonable that the discount rate of a fund decreases with its size. Therefore, in the framework of our theoretical model, the impact of an increase in the size of the fund should be similar to the impact of receiving financing from CVC instead of IVC funds. This claim is 15

To capture the quadratic effect of the duration on the IPO likelihood for start-ups, we use the square of duration at year level. 16 Cumming and MacIntosh (2003) also find a negative (although not significant) impact of duration on the likelihood of IPO exit versus acquisition exits.

26

also empirically validated. Fund size has a positive effect on both investment and duration decisions (see Tables 6 and 7). Moreover, similar to the CVC dummy, fund size does not have any direct effect in the probability of IPO exits; its effects are indirect through the investment and duration decisions. These empirical findings provide further support to our theoretical model. We now comment on the other effects. First, if the syndicate is led by CVC funds, then IPO exit is more likely. Second, a stronger stock market for 3 months before the exit date also leads to more IPO exits. Third, the likelihood of IPO exist decreases with the strength of the stock market for 6 months before the exit date. Fourth, the Later Stage Dummy has a significantly positive effect, that is, as the entrepreneurs have more controlling power over the startups than VC managers, more IPO exits are observed. This result is consistent with the findings of Cumming (2008). Finally, the startups exit more often through IPO if they are in the same industry, defined by 4-digit SIC code, as the parent company behind CVC funds. This suggests that CVCs may push not to sell the startups to potential competitors of the parent company. Based on Model 5, we also find that the duration has a quadratic effect on the probability of IPO exit. According to this result, the negative effect of duration on the probability of IPO exit is stronger when the duration is short than when it is long. To better understand the quadratic effect, we run similar regressions for subsamples of our dataset. They suggest that the effect of duration on IPO exit has two regions: it is linear (and significant) for startups whose duration is shorter than 7 years while it is not significant (neither linear nor quadratic) when duration is longer than 7 years. In our dataset, the average duration for startups is around 4 − 5 years; more than 80% of the startups in our sample stay less than 8 years.

8.4

Funds’ Characteristics and Successful Exit Rate

Hypotheses 6 to 8 summarize the indirect effect of funds’ characteristics (via the investment effect) on the successful exit rate of startups. We test them through the following equation: F ailurei = α0 + α1S F und� sCharacteristics + α2S ln Duration(Days)i +

α3S

ln Investi +

5 �

αkS Zk + �i

k=1

27

(11)

According to the hypotheses, α1S and α2S are expected to be insignificant. However, we expect a negative but significant coefficient α3S . The estimation results are shown in Table 9. Model 1 includes the results when we use CVC dummy variable as a measure of the Funds’ Characteristics. The results with the independent variable of CVC percentage investment are in Model 2. We find that funds’ characteristics, duration, investment amount and whether CVC is the syndicate leader, are negatively but not significantly correlated with the liquidation probability. These results support Hypotheses 7 and 8. However, we do not find support for Hypothesis 6. The only significant impact is from the company strategic relationship on the failure rate. If the startup and the parent company of CVC funds are potential competitors, the probability for the startups to be liquidated is lower. Therefore, CVC backed startups are at least as successful as those backed by IVCs. Those CVC-backed startups having no potential competition from the parent companies perform significantly better than those only receiving investment from IVCs. Venture capital fund characteristics indirectly influences startups’ successful exit rate through their strategic relationship. These results are consistent with the findings of Gompers and Lerner (2000). Our theoretical predictions are closer to the findings of Chemmanur and Loutskina (2008). Using a different subsample of the same database, Chemmanur and Loutskina have found that CVC investments in the startups lead to a higher but not significant successful exit rates. It is a higher investment level that increases the successful rate, as is predicted by our theoretical model. Other empirical studies also confirm that CVC backed startups perform better than IVC backed startups. For example, Dushnitsky and Shapira (forthcoming), show that CVC backed startups exhibit significantly better performance as measured by the rate of successful portfolio exits. The increase in the successful exit rate ranges from 9.7% to 20% depending on CVC managers’ incentives.

8.5

Sensitivity Test

To check the sensitivity of our estimated results, we estimate the previous regressions by removing the startups in the Business Service industry (SIC2 = 73), which contains almost 50% startups in our sample. Table 10 provides the estimated results for the subsample. Our results are qualitively robust, although the effect of CVC funds on the duration and the duration effect on the exit strategy are not statistically significant.

28

8.6

Robustness Test for Possible Endogeneity

One concern regarding our empirical analysis is that the investment amount, duration period before exit, the exit strategy and the VC fund characteristics may all depend on the potential value of the startups (parameter V in the theoretical model).17 This would mean that the choice of investment strategy, duration strategy and the exit strategy of startups can also be triggered by the quality of the startups. Since we don’t have direct measure of the potential value V of a startup, we use two methods to keep the effect of V constant in our estimation. First, we split our dataset into two subsamples: startups that finally exited through the IPO market and those that chose to go to the acquisition market. According to our theoretical model, the startups with IPO exits have higher value than those with Acquisition exits. Within each subsample, the startups have similar value. We estimate again the investment strategy and the duration before exit (see Equations (8) and (9)) for the set of startups that chose IPO-exit and for the set that went to Acquisition-exit separately. The estimated results are shown in Tables 11 and 12. For the investment strategy, the results indicate that the previous main conclusions still hold for both IPO-exit startups and Acquisition-exit startups. In the two subsamples, we find that CVC backed startups receive a significantly higher investment amount. One interesting difference is that, for Acquisition-exit startups, the strategic relationship between the startups and CVC funds does play a role. This suggests that, if the startups and the parent companies of the CVC funds are in the same industry, i.e. they are potential competitors, CVCs invest more in the ventures that end up being acquired. For the duration strategy, CVC investment significantly increases the duration for startups in the acquisition market. Interestingly, we have found that the VC fund characteristics is not important in deciding the duration before exit for startups in the IPO market. This result matches with our theoretical predictions: according to our model, the decision on the duration before exit (that is, the level of information) only matters if the startups go to the acquisition market. Second, we use a less stringent definition for the strategic relationship 17

We recall that our analysis in Section 8.1 already suggested that no obvious selection bias between CVC and IVC projects exit.

29

between startups and VC funds (potential competitors or not). If the startup and the parent company of the CVC fund are in the same industry, the latter is likely to have more information about the potential value of the venture than if the partners operate in different industries. We have controlled this effect using a 4-digit SIC code (SIC4) in Sections 8.1 to 8.4. In Tables 6 to 9 we have shown that, after controlling for the information effect, we still find that CVC significantly influences investment and duration, while it does not influence the exit strategy. To make sure that these results are not due to the fact that the measure we have used is too stringent, we now redefine the variable of CVC Strategic Relation using the 2-digit SIC code. The dummy variable is equal to 1 if the startup and the parent company of CVC fund have the same 2-digit SIC code, and 0 otherwise. The regression results are in Tables 13 to 15. There is no significant difference between the results of using SIC4 code and those of using SIC2 code. This means that our results are robust for either definition of the strategic relationship between CVCs and ventures.

9

Conclusion

In this paper, we have studied the optimal initial and exit decisions by startups. In particular, we have focused on the difference in behavior between CVC backed startups and IVC backed startups. In our theoretical model, the difference between CVC and IVC financing is attributed to different discount rates. We have assumed that (for example because of strategic objectives) CVC funds are less hurried to exit than IVC funds. Therefore, startups backed by CVC funds have a lower discount rate than those backed by IVCs. We have found that CVC backed startups have longer duration before exit and larger investment level than those financed by IVCs. These properties, in turn, lead to higher successful exit rates and to two opposite impacts on the likelihood of an IPO exit. Longer duration, implying more information in the acquisition market, increases the probability that the startup looks for an acquirer. On the contrary, higher investment level, increasing the value of the startups, encourages more IPO exits. The theoretical results have been then empirically tested with data from VentureXpert database. Our empirical study indicates that CVC financing do imply longer duration and larger investment level than IVC funding. Moreover, the effect of venture capital funds’ characteristics on startups’ exit strategy can be explained through the investment and duration deci30

sions. Shorter duration as well as larger investment level significantly lead to a higher likelihood of IPO exit. Once these two effects are taken into account, whether the venture capital fund is corporate or independent does not have a significant influence on the startup exit decision.

31

Appendix Proof of Proposition 3 Proof. We notice that the rate of IPO exits over the total successful exits is either � 1 [1 − po (V )] γ(V ; I)dV (12) F 1 − Γ(C; I) V ≥C+ β−m or

1 1 − Γ(C; I)

�

V ≥C+ β−Fm

[1 − poo (V )] γ(V ; I)dV.

2

(13)

� ˆ Given that V is uniformly distributed over the interval f (I), V + f (I) , γ(V ; I) 1 = . 1 − Γ(C; I) Vˆ + f (I) − C

�

Therefore, � � 1 [1 − po (V )] γ(V ; I)dV = F 1 − Γ(C; I) V ≥C+ β−m   � Vˆ +f (I) 1 ∂  � � [1 − po (V )] dV  = F ∂I Vˆ + f (I) − C C+ β−m � � � Vˆ +f (I) � � � � � f (I) [1 − po (V )] dV > 0, � �2 1 − po (Vˆ + f (I)) Vˆ + f (I) − C − F C+ β−m ˆ V + f (I) − C ∂ ∂I

�

where the inequality holds because f � (I) > 0 and [1 − po (V )] is an increasing function of V . Therefore, the expression (12) is increasing in I. A similar argument allow to prove that (13) is also increasing in I. Proof of Lemma 1. Proof. Propositions 1 and 2 imply that �� � po (V )

EUo (I) =

=

�

V ≥C

V ≥C

�

0

mp(V − C)dp +

�

1 po (V )

�

[βp(V − C) − F ] dp dΓ(V ; I)

� �� 1 1 2 β(V − C) − (β − m) po (V ) (V − C) + F (1 − po (V )) dΓ(V ; I) 2 2 32

and

� � 1 poo (V ) EUoo (I) = m (V − C)dp + [βp(V − C) − F ] dp dΓ(V ; I) 2 V ≥C 0 poo (V ) � � �� � 1 1 2 = β(V − C) − (β − m) poo (V ) (V − C) + F (1 − poo (V )) dΓ(V ; I). 2 V ≥C 2 �

��

poo (V )

Therefore, EUo (I) ≥ EUoo (I) if 1 1 (β − m) po (V )2 (V −C)+F (1 − po (V )) ≤ (β − m) poo (V )2 (V −C)+F (1 − poo (V )) . 2 2 (14) Equation (14) holds equality if poo (V ) = 1 (and then, po (V ) = 1 as well). Otherwise, denote j(p) ≡ 12 (β − m) p2 (V − C) + F (1 − p). Then, j � (p) = (β − m) p(V − C) − F < 0 for all p < min {po , 1}, given the definition of po . Therefore, j(poo (V )) > j(po (V )) whenever poo (V ) < 1, that is, (14) holds with strict inequality when poo (V ) < 1 for some V in the support of the distribution Γ(d; I), that is, when poo (Vˆ + f (I)) < 1. Proof of Proposition 6 Proof. Maximizing Equation (5), d∗ and I ∗ are characterized by ∂U ∗ ∗ (d , I ) = 0 ∂d

(15)

∂U ∗ ∗ (d , I ) = 0. (16) ∂I We differentiate Equations (15) and (16) and solve the system to obtain that, at (d∗ , I ∗ ), ∂I ∗ ΛI =− (17) ∂r ∆ ∂d∗ Λd =− , (18) ∂r ∆ where ∂ 2U ∂ 2U ∂ 2U ∂ 2U ΛI = − ∂d2 ∂I∂r ∂I∂d ∂d∂r ∂ 2U ∂ 2U ∂ 2U ∂ 2U Λd = − ∂I 2 ∂d∂r ∂I∂d ∂I∂r and � 2 �2 ∂ 2U ∂ 2U ∂ U ∆= − . 2 2 ∂I ∂d ∂I∂d 33

Before the analysis of the second derivatives of the function U (d, I) = e−rd [h(d)EUo (I) + (1 − h(d))EUoo (I)] − I, we analyze the functions EUo (I) and EUoo (I). � 1 2 EUo (I) = (β − m)po (V ) (V − C) + F (1 − po (V )) dΓ(V ; I). V ≥C V ≥C 2 (19) 1 Taking into account that dΓ(V ; I) = Vˆ , the first term of the right-hand side of (19) is � Vˆ +f (I) �2 1 β �ˆ β(V − C)dΓ(V ; I) = V + f (I) − C 2 4Vˆ C � � 1 F and, given that po (V ) = min (β−m) (V −C) , 1 , we split the right-hand side of (19) in two parts: � C+ F β−m 1 1 F2 (β − m)(V − C) dV = 2 Vˆ 4(β − m)Vˆ C �

and �

1 β(V −C)dΓ(V ; I)− 2

�

�

� �� F2 F 1 +F 1− dV = F 2(β − m)(V − C) (β − m)(V − C) Vˆ C+ β−m � � � �� � � � F2 F F F − log Vˆ + f (I) − C − log + Vˆ + f (I) − C − . β−m β−m 2(β − m)Vˆ Vˆ Vˆ +f (I)

�

Therefore,

� �2 F � β �ˆ 3F ˆ EUo (I) = V + f (I) − C − V + f (I) − C − + 4 (β − m) 4Vˆ Vˆ � � � �� � F2 F log Vˆ + f (I) − C − log . (20) β−m 2(β − m)Vˆ � � Similarly, taking into account that poo (V ) = min β−1 m (V F−C) , 1 , we ( 2) obtain: � �2 β �ˆ (β − m)F 2 F ˆ F EUoo (I) = V + f (I) − C − � − V + f (I) − C − � � 2 m β− 4Vˆ Vˆ 4 β − 2 Vˆ � � �� � � (β − m)F 2 F ˆ + f (I) − C − log log V + � � m 2 ˆ β − m2 2 β− 2 V � � � �� � F2 F ˆ log V + f (I) − C − log . � � β − m2 β − m2 Vˆ 34

m 2

�

� − (21)

From (20) and (21),   � � � 2 f (I)  β ˆ F � � EUo� (I) = V + f (I) − C − F + 2 ˆ Vˆ 2(β − m) V + f (I) − C (22) � � f (I) β ˆ � EUoo (I) = [V + f (I) − C] − F + 2 Vˆ 2(β −

� βF 2 . m 2 ˆ ) (V + f (I) − C) 2

(23)

� As it is intuitive and easy to check, EUo� (I) > 0 and EUoo (I) > 0 always. We now analyze the sign of the second derivatives of the function U (., .).

∂ 2U ∗ ∗ (d , I ) = e−rd [h�� (d) − rh� (d)][EUo (I) − EUoo (I)]. (24) ∂d2 In Equation (24), h� (d) > 0 and h�� (d) < 0. Moreover, Proposition 1 implies 2 that EUo (I) > EUoo (I). Therefore, ∂∂dU2 (d∗ , I ∗ ) < 0. ∂ 2U ∗ ∗ (d , I ) = −d < 0. ∂I∂r

(25)

∂ 2U ∗ ∗ �� (d , I ) = e−rd [h(d)EUo�� (I) + (1 − h(d))EUoo (I)] < 0 ∂I 2

(26)

because EUo�� (I) =

��



2

f (I)  β ˆ F � � − (V + f (I) − C) − F + ˆ 2 ˆ V 2(β − m) V + f (I) − C  

f � (I)2  β − 2Vˆ

�

F2

(β − m) Vˆ + f (I) − C

 �2 

f �� (I) β ˆ [ [V + f (I) − C] − F + Vˆ 2 2(β − f � (I)2 βF 2 [β − . 2Vˆ (β − m )2 (Vˆ + f (I) − C)2

�� EUoo (I) =

βF 2 ]+ m 2 ˆ ) (V + f (I) − C) 2

2

and both



EUo�� (I)

�� < 0 and EUoo (I) < 0 if f (I) is concave enough.

35

∂ 2U ∗ ∗ (d , I ) = −e−rd [h(d)EUo (I) + (1 − h(d))EUoo (I)] < 0. ∂d∂r

We notice that

∂2U ∂I∂d

� EUo� (I) − EUoo (I) =

(27)

� (d∗ , I ∗ ) = −r + e−rd h� (d)[EUo� (I) − EUoo (I)], with m2

f � (I)F 2 4 2 Vˆ (Vˆ +f (I)−C ) 2(β−m)(β− m ) 2

> 0. Therefore, investment

and duration may be complement or substitute, depending on the compar∂2U ison of the two terms. If they are complements, that is, ∂I∂d (d∗ , I ∗ ) ≥ 0, then ΛI > 0 and Λd > 0. If they are substitutes, the same inequalities hold as long as the functions h(d) and f (I) are sufficiently concave, which also implies that ∆ > 0 (it is always positive in any strict maximum). Therefore,

∂I ∗ ∂r

< 0 and

∂d∗ ∂r

< 0, as we wanted to prove.

36

Table 1. Industry Composition of the Sample

Two-Digit SIC Code 13 20 27 28 35 36 38 48 50 51 59 62 63 73 80 87

Industry Name Oil and gas extraction Food and kindred products Printing and publishing Chemicals and allied products Industrial machinery and equipment Electronic and other electronic equipment Instruments and related products Communications Wholesale trade - durable goods Wholesale trade - nondurable goods Miscellaneous retail Security and commodity brokers Insurance carriers Business services Health services Engineering and management services

37

Number of Startups 27 21 24 360 274 510 348 199 57 20 64 20 29 2, 199 127 195

Table 2. Definitions of Variables Variables CVC

Definitions Dummy variable equal to 1 for CVC backed startups and 0 for IVC backed startups

CVC Per

Percentage of investment by CVC in each startup

IPO

Dummy variable equal to 1 for IPO exit and 0 for Acquisition exit

Investment amount

Total investment amount at startup level, measured by disclosed equity amount (USD Million) summed over investment rounds

Duration (Days)

Difference in days between the exit date and the date at which a startup receives the first investment from venture capital firms

Duration (Years)

Duration (Days) divided by 365

Investment rounds

Number of investment rounds for a startup

VC syndicate

Number of venture capital firms that invest in a startup

Syndicate CVC

Dummy variable equal to 1 if the syndicate leader is CVC fund and 0 for IVC fund as the leader

VC fund size

Average size (USD Million) of venture capital funds that finance the startup

CVC strategic relationship

Measure of CVC strategic competitors, dummy variable of 1 if a CVC has the same 4-digit SIC code as its start-up, and 0 otherwise

VC fund age (Years)

Average fund age across all funds invested in a startup

Industry MB

Industry market-to-book value at the year that CVC firm makes the first investment

MSCI 3 mon

MSCI return 0-3 months prior to the exit date

MSCI 6 mon

MSCI return 3-6 months prior to the exit date

Later stage dummy

Dummy variable equal to 1 if a startup is at expansion or later stage at the exit and 0 otherwise

38

Table 3. Summary Statistics

CVC CVC Per IPO Investment amount Duration (Days) Duration (Years) Investment rounds VC syndicate Syndicate CVC VC fund size VC fund age (Years) CVC strategic relationship Industry MB MSCI 3 mon MSCI 6 mon Later stage dummy

No. of Obs. Mean 4801 0.37 4801 0.23 4801 0.35 4801 31.47 4801 1753.26 4801 4.8 4801 4.26 4801 5.81 4801 0.05 4801 211.75 4737 7.98 4801 0.01 4801 13.51 4801 2861.97 4801 2821.1 4801 0.81

39

Std. Dev. 0.48 0.34 0.48 78.04 1197.6 3.28 2.87 4.37 0.22 383.62 5.46 0.12 22.53 1391.96 1395.53 0.39

Min 0 0 0 0.02 10 0.03 1 1 0 0.09 0 0 0.74 113.88 108.83 0

Max 1 1 1 4653.06 12056 33.03 27 35 1 6011.62 45.33 1 432.1 4881.96 4881.96 1

Table 4. Correlation Matrix

VC Fund Age Syndicate CVC CVC Investment Amount Investment Rounds Duration VC Syndicate VC Fund Size CVC Strategic Re. IPO Industry MB CVC per MSCI mon. 3 MSCI mon. 6 Later Stage Dummy

VC Fund Age 1.00 −0.01 0.06∗∗∗ 0.03∗∗ 0.05∗∗∗ −0.02 0.05∗∗∗ 0.03∗∗ 0.02 −0.10∗∗∗ 0.07∗∗∗ 0.04∗∗ 0.21∗∗∗ 0.21∗∗∗ 0.00

CVC Strat. Re.

Synd. CVC

CVC

1.00 0.29∗∗∗ 0.04∗∗∗ −0.04∗∗∗ −0.01 0.01 −0.02 0.13∗∗∗ 0.05∗∗∗ 0.01 0.39∗∗∗ 0.02 0.01 −0.01

1.00 0.18∗∗∗ 0.25∗∗∗ 0.12∗∗∗ 0.46∗∗∗ −0.02 0.15∗∗∗ 0.01 0.06∗∗∗ 0.85∗∗∗ 0.14∗∗∗ 0.14∗∗∗ −0.03∗∗

IPO

VC Fund Age Syndicate CVC CVC Investment Amount Investment Rounds Duration VC Syndicate VC Fund Size CVC Strategic Re. 1.00 IPO 0.03 1.00 Industry MB 0.00 −0.23∗∗∗ CVC per 0.12∗∗∗ −0.01 MSCI mon. 3 0.05∗∗∗ −0.49∗∗∗ MSCI mon. 6 0.05∗∗∗ −0.50∗∗∗ Later Stage Dummy −0.04∗∗∗ −0.01 ∗∗ and ∗∗∗ denote statistical significance at 5% and 1%

40

Invest Amount

1.00 0.19∗∗∗ 0.07∗∗∗ 0.25∗∗∗ 0.14∗∗∗ 0.04∗∗∗ 0.04∗∗∗ 0.03∗∗ 0.16∗∗∗ 0.16∗∗∗ 0.16∗∗∗ 0.01

Industry MB

1.00 0.07∗∗∗ 0.24∗∗∗ 0.26∗∗∗ 0.02 level.

Invest Rounds

1.00 0.46∗∗∗ 0.58∗∗∗ −0.03 0.04∗∗∗ 0.07∗∗∗ −0.02 0.10∗∗∗ 0.01 0.01 −0.02

CVC per

1.00 0.12∗∗∗ 0.12∗∗∗ −0.03∗∗

Duration

1.00 0.29∗∗∗ −0.04∗∗∗ 0.02 −0.05∗∗∗ −0.01 0.04∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.01

VC Synd.

VC Fund Size

1.00 −0.06∗∗∗ 0.06∗∗∗ 0.14∗∗∗ −0.02 0.36∗∗∗ −0.01 −0.01 −0.02

1.00 −0.01 −0.12∗∗∗ 0.07∗∗∗ −0.02 0.22∗∗∗ 0.22∗∗∗ 0.02

MSCI mon. 3

MSCI mon. 6

Later Stage Dummy

1.00 0.99∗∗∗ −0.01

1.00 −0.01

1.00

Table 5. CVCs vs IVCs Panel A. CVC VS. IVC: Summary Statistics IVC CVC Difference t-statistics IPO 0.35 0.36 −0.01 −0.64 Investment amount 20.5 49.88 −29.38 −12.83∗∗∗ Duration 1648.86 1928.57 −279.71 −7.88∗∗∗ Investment rounds 3.71 5.19 −1.48 −17.9∗∗∗ VC syndicate 4.25 8.42 −4.16 −35.93∗∗∗ VC fund size 216.7 203.43 13.27 1.16 VC fund age 7.73 8.39 −0.66 −4.03∗∗∗ Later stage dummy 0.77 0.88 −0.11 −9.46∗∗∗ ∗ ∗∗ , and ∗∗∗ denote statistical significance at 10%, 5% and 1% level.

Panel B. CVC VS. IVC: No. of startups per Round No. Gr. 0 Gr. 1 Gr. 2 Gr. 3 Gr. 4 Gr. 5 Gr. 6 Gr. 7 Gr. 8

R1 2778 548 401 344 183 97 50 25 11

R2 2117 422 415 334 178 99 51 23 10

R3 1492 322 291 352 179 96 51 20 9

R4 977 233 188 254 187 95 49 24 9

R5 617 164 114 156 131 100 49 24 9

R6 408 95 68 95 83 54 55 22 10

R7 238 63 39 54 41 30 28 26 11

R8 161 44 24 35 25 19 20 13 11

Panel C. CVC VS. IVC: Investment Amount per startup per Round Invest. Gr. 0 Gr. 1 Gr. 2 Gr. 3 Gr. 4 Gr. 5 Gr. 6 Gr. 7 Gr. 8

R1 5.39 8.90 5.47 3.91 2.97 2.89 2.75 2.45 2.18

R2 6.05 11.34 16.42 9.06 6.28 5.12 5.46 4.41 3.38

R3 7.11 10.91 11.74 18.16 9.53 7.48 5.20 3.60 3.57

R4 6.71 10.95 13.18 14.15 18.26 8.34 3.84 6.51 6.27

41

R5 6.41 9.83 11.37 12.33 14.66 15.54 6.46 5.09 7.26

R6 5.64 6.76 15.23 8.90 11.21 11.41 17.56 10.05 6.94

R7 4.24 4.56 7.13 10.14 14.81 9.19 12.97 14.03 12.31

R8 4.57 4.49 9.37 7.09 8.17 6.89 12.07 17.30 18.43

Table 6: Investment Strategy

Dependent Variable: ln(Investment Amount) Model 1 Model 2 Model 3 Model 4 0.2525∗∗∗ 0.249∗∗∗ 0.2484∗∗∗ CVC (8.08) (7.91) (7.85) 0.2852∗∗∗ CVC per (6.44) 0.1105∗∗∗ 0.1104∗∗∗ 0.1081∗∗∗ 0.1131∗∗∗ Investment Rounds (20.15) (20.13) (19.58) (20.33) ∗∗∗ ∗∗∗ ∗∗∗ 0.1289 0.129 0.1287 0.1311∗∗∗ VC Syndicate (32.81) (32.83) (32.63) (33.3) ∗∗∗ ∗∗∗ ∗∗∗ 0.2545 0.2491 0.2638 0.2458∗∗∗ Syndicate CVC (4.2) (4.09) (4.29) (3.83) 0.35∗∗∗ 0.35∗∗∗ 0.356∗∗∗ 0.356∗∗∗ ln(VC Fund Size) (32.2) (32.21) (32.47) (32.39) 0.0118∗∗∗ 0.012∗∗∗ VC Fund Age (4.96) (5.05) 0.1137 0.1083 0.1504 CVC Strategic Re. (1.05) (0.99) (1.38) ∗∗ ∗∗ ∗∗ 0.0516 0.051 0.0502 0.0497∗∗ ln(Industry MB) (2.55) (2.52) (2.47) (2.44) Year Fixed Effect Yes Yes Yes Yes Industry Fixed Effect Yes Yes Yes Yes The t statistic is in the parentheses. ∗ , ∗∗ and ∗∗∗ denote statistical significance at 10%, 5% and 1% level.

42

Table 7: Duration Strategy

Dependent Variable: Duration (Days) Model 1 Model 2 Model 3 Model 4 1.06∗ 1.06 1.07∗ CVC (1.64) (1.60) (1.78) 1.16∗∗∗ CVC per (2.78) 0.86∗∗∗ 0.86∗∗∗ 0.86∗∗∗ 0.86∗∗∗ Investment Rounds (−21.39) (−21.38) (−21.23) (−20.88) ∗ 0.99 0.99 0.99 0.99∗∗ VC Syndicate (−1.35) (−1.35) (−1.80) (−2.17) 0.899 0.898 0.90 0.86∗∗ Syndicate CVC (−1.48) (−1.5) (−1.43) (−2.02) 1.06∗∗∗ 1.06∗∗∗ 1.06∗∗∗ 1.06∗∗∗ ln(VC Fund Size) (4.77) (4.77) (4.57) (4.54) 1.01∗∗∗ 1.01∗∗∗ VC Fund Age (3.82) (3.89) 1.04 1.01 1.02 CVC Strategic Re. (0.28) (0.1) (0.15) 0.99 0.99 0.98 0.98 ln(Industry MB) (−0.38) (−0.39) (−1.05) (−1.16) Year Fixed Effect No No No No Industry Fixed Effect Yes Yes Yes Yes The Z value is in the parentheses. ∗ , ∗∗ and ∗∗∗ denote statistical significance at 10%, 5% and 1% level.

43

Table 8: Exit Strategy

CVC CVC per ln(Duration)

Dependent Variable: Dummy Variable Model 2 Model 3 Model 4 −0.004 0.034 (−0.33) (0.36) −0.013 (−0.72) −0.018∗∗ −0.019∗∗ −0.019∗∗ −0.405∗∗∗ (−2.39) (−2.46) (−2.49) (−7.44) Model 1 −0.003 (−0.24)

Duration Year (Duration Year)2 0.068∗∗∗ 0.068∗∗∗ 0.602∗∗∗ (12.96) (13.33) (14.45) −0.004 −0.004 −0.163∗∗∗ ln(VC Fund Size) (−0.74) (−0.77) (−4.44) 0.0005 0.0005 −0.002 VC Fund Age (0.48) (0.47) (−0.31) 0.05∗∗ 0.054∗∗ 0.059∗∗ 0.3236∗ Syndicate CVC (1.97) (2.08) (2.19) (1.84) 0.076∗ 0.076∗ 0.076∗ 0.741∗∗∗ CVC Strategic Re. (1.67) (1.64) (1.65) (2.74) 0.44∗∗∗ 0.45∗∗∗ 0.45∗∗∗ 5.53∗∗∗ ln(MSCI mon. 3) (3.56) (3.58) (3.58) (7.29) −0.278∗∗ −0.296∗∗ −0.296∗∗ −7.23∗∗∗ ln(MSCI mon. 6) (−2.33) (−2.44) (−2.44) (−9.55) 0.009 0.01 0.01 −0.394∗∗∗ ln(Industry MB) (0.85) (0.87) (0.88) (−8.26) 0.058∗∗∗ 0.055∗∗∗ 0.055∗∗∗ 0.48∗∗∗ Later stage dummy (4.00) (3.81) (3.79) (4.2) Year Fixed Effect Yes Yes Yes No Industry Fixed Effect Yes Yes Yes No The t statistic is in the parentheses for Model 1, 2, 3, 5 and 6. The Z-value is in the parentheses for Model 4. ∗ ∗∗ , and ∗∗∗ denote statistical significance at 10%, 5% and 1% level. ln(Invest)

0.067∗∗∗ (13.08) −0.005 (−0.9)

44

of IPO Exit Model 5 Model 6 −0.0045 0.052∗∗∗ (−0.35) (4.34)

−0.013∗∗∗ (−3.19) 0.0008∗∗∗ (3.39) 0.069∗∗∗ (13.09) −0.003 (−0.64) 0.0006 (0.63) 0.052∗∗ (2.01) 0.077∗ (1.67) 0.437∗∗∗ (3.48) −0.294∗∗ (−2.43) 0.009 (0.84) 0.053∗∗∗ (3.66) Yes Yes

0.025∗∗∗ (5.32) 0.002∗ (1.68) 0.046∗ (1.75) 0.076 (1.63) 0.429∗∗∗ (3.35) −0.26∗∗ (−2.11) 0.013 (1.11) 0.103∗∗∗ (7.39) Yes Yes

Table 9: Successful Exit Rate Dependent Variable: Failure rate Model 1 Model 2 0.0014 CVC (0.12) 0.0073 CVC per (0.31) −0.005 −0.0049 ln(Duration) (−0.88) (−0.87) −0.0005 −0.0005 ln(Invest) (−0.10) (−0.11) −0.0261 −0.0267 Syndicate CVC (−1) (−1.05) 0.0046 0.0046 ln(VC Fund Size) (1.04) (1.06) −0.3149∗∗∗ −0.3148∗∗∗ CVC Strategic Re. (−5.43) (−5.43) −0.0161 −0.0162 Later stage dummy (−1.19) (−1.19) Year Fixed Effect No No Industry Fixed Effect Yes Yes The t statistic is in the parentheses. ∗ , ∗∗ and ∗∗∗ denote statistical significance at 10%, 5% and 1% level.

45

Table 10: Sensitivity Test Dependent variable CVC

ln(Investment Amount) 0.26∗∗∗ (5.53)

Duration(Days) 1.01 (0.24)

0.107∗∗∗ (14.14) 0.126∗∗∗ (23.81) 0.367∗∗∗ (3.88) 0.39∗∗∗ (24.2) 0.01∗∗∗ (2.73) 0.09 (0.7)

0.87∗∗∗ (−15.04) 1.002 (0.34) 0.913 (−0.85) 1.06∗∗∗ (3.56) 1.01∗∗∗ (3.28) 1.15 (0.91)

0.04 (1.52)

0.95∗ (−1.82)

ln(Duration) ln(Invest) Investment Rounds VC Syndicate Syndicate CVC ln(VC Fund Size) VC Fund Age CVC Strategic Re. ln(MSCI mon. 3) ln(MSCI mon. 6) ln(Industry MB) Later stage dummy Year Fixed Effect Yes No Industry Fixed Effect Yes Yes The t statistic and the Z value are in the parentheses. ∗ ∗∗ , and ∗∗∗ denote statistical significance at 10%, 5% and 1% level.

46

Dummy Variable of IPO Exit −0.0056 (−0.29) −0.008 (−0.75) 0.074∗∗∗ (10.25)

0.076∗ (1.92) −0.01 (−1.25) 0.001 (0.67) 0.08 (1.43) 0.7∗∗∗ (3.47) −0.407∗∗ (−2.07) 0.02 (1.44) 0.05∗∗ (2.18) Yes Yes

Table 11: Investment Strategy for IPO Startups and for Acquisition Startups

Dependent Variable: ln(Investment Amount) IPO Exits Acquisition Exits Model 3 Model 4 Model 3 Model 4 0.2387∗∗∗ 0.2287∗∗∗ CVC (4.17) (4.53) 0.3188∗∗∗ 0.2852∗∗∗ CVC per (3.84) (6.44) 0.1492∗∗∗ 0.1539∗∗∗ 0.0906∗∗∗ 0.0950∗∗∗ Investment Rounds (14.95) (15.38) (14.10) (14.64) ∗∗∗ ∗∗∗ ∗∗∗ 0.1011 0.1020 0.1452 0.1485∗∗∗ VC Syndicate (16.53) (16.72) (27.98) (28.69) ∗∗∗ ∗∗∗ 0.3785 0.3333 0.0768 0.0723 Syndicate CVC (3.74) (3.11) (1.02) (0.93) 0.2970∗∗∗ 0.2970∗∗∗ 1.08∗∗∗ 1.08∗∗∗ ln(VC Fund Size) (14.62) (14.25) (29.10) (29.03) 0.0083 0.0086∗ 0.0130∗∗∗ 0.0130∗∗∗ VC Fund Age (1.62) (1.69) (5.07) (5.09) ∗ −0.1495 −0.1352 0.2318 0.2823∗∗ CVC Strategic Re. (−0.87) (−0.78) (1.70) (2.07) ∗ ∗ ∗∗∗ 0.0826 0.0822 0.08 0.0792∗∗∗ ln(Industry MB) (1.72) (1.70) (3.58) (3.54) Year Fixed Effect Yes Yes Yes Yes Industry Fixed Effect Yes Yes Yes Yes The t statistic is in the parentheses. ∗ , ∗∗ and ∗∗∗ denote statistical significance at 10%, 5% and 1% level.

47

Table 12: Duration Strategy for IPO Startups and for Acquisition Startups

Dependent Variable: Duration (Days) IPO Exits Acquisition Exits Model 3 Model 4 Model 3 Model 4 1.01 1.17∗∗∗ CVC (0.14) (3.38) 1.14 1.27∗∗∗ CVC per (1.38) (3.75) 0.89∗∗∗ 0.89∗∗∗ 0.84∗∗∗ 0.84∗∗∗ Investment Rounds (−10.02) (−9.94) (−19.43) (−18.92) ∗∗∗ 1.01 1.00 0.97 0.97∗∗∗ VC Syndicate (0.92) (0.92) (−4.55) (−4.69) 0.99 0.91 0.91 0.87∗∗ Syndicate CVC (−0.10) (−0.76) (−1.02) (−1.47) 1.08∗∗∗ 1.08∗∗∗ 1.06∗∗∗ 1.06∗∗∗ ln(VC Fund Size) (3.41) (3.42) (3.77) (3.82) 1.01 1.01 1.01∗∗∗ 1.02∗∗∗ VC Fund Age (1.45) (1.48) (4.69) (4.78) 0.95 0.94 0.99 1.04 CVC Strategic Re. (−0.24) (−0.32) (−0.04) (0.22) 0.94 0.94 1.04∗ 1.04∗ ln(Industry MB) (−1.51) (−1.63) (1.71) (1.66) Year Fixed Effect No No No No Industry Fixed Effect Yes Yes Yes Yes The Z value is in the parentheses. ∗ , ∗∗ and ∗∗∗ denote statistical significance at 10%, 5% and 1% level.

48

Table 13: Investment and Duration Strategies - SIC2

ln(Investment Amount) Duration (Days) Model 3 Model 4 Model 3 Model 4 0.247∗∗∗ 1.07∗ CVC (7.74) (1.87) 0.2815∗∗∗ 1.16∗∗∗ CVC per (6.32) (2.84) ∗∗∗ ∗∗∗ ∗∗∗ 0.1081 0.1130 0.86 0.86∗∗∗ Investment Rounds (19.56) (20.29) (−21.23) (−20.86) ∗∗∗ ∗∗∗ ∗ 0.1287 0.1311 0.99 0.99∗∗ VC Syndicate (32.63) (33.3) (−1.81) (−2.16) 0.2621∗∗∗ 0.2427∗∗∗ 0.90 0.86∗∗ Syndicate CVC (4.24) (3.77) (−1.37) (−1.96) 0.3561∗∗∗ 0.3561∗∗∗ 1.06∗∗∗ 1.06∗∗∗ ln(VC Fund Size) (32.47) (32.4) (4.57) (4.54) ∗∗∗ ∗∗∗ ∗∗∗ 0.0118 0.012 1.01 1.01∗∗∗ VC Fund Age (4.97) (5.06) (3.83) (3.90) 0.0644 0.1 0.95 0.96 CVC Strategic Re. (0.89) (1.39) (−0.55) (−0.54) 0.0504∗∗ 0.05∗∗ 0.98 0.98 ln(Industry MB) (2.48) (2.46) (−1.04) (−1.14) Year Fixed Effect Yes Yes Yes Yes Industry Fixed Effect Yes Yes Yes Yes The t statistic is in the parentheses. ∗ , ∗∗ and ∗∗∗ denote statistical significance at 10%, 5% and 1% level.

49

Table 14: Exit Strategy - SIC2 Dependent Variable: Dummy Variable of IPO Exit Model 2 Model 3 Model 4 Model 5 −0.0074 −0.0023 −0.0076 CVC (−0.57) (−0.02) (−0.59) −0.0165 CVC per (−0.92) −0.0183∗∗ −0.0191∗∗ −0.0194∗∗ −0.405∗∗∗ ln(Duration) (−2.41) (−2.48) (−2.52) (−7.43) −0.0133∗∗∗ Duration Year (−3.22) 0.0008∗∗∗ 2 (Duration Year) (3.42) 0.0675∗∗∗ 0.0676∗∗∗ 0.0679∗∗∗ 0.6017∗∗∗ 0.0685∗∗∗ ln(Invest) (13.08) (12.96) (13.3) (14.44) (13.1) −0.0045 −0.0038 −0.0039 −0.1642∗∗∗ −0.0033 ln(VC Fund Size) (−0.91) (−0.74) (−0.76) (−4.45) (−0.64) 0.0005 0.0005 −0.0021 0.0007 VC Fund Age (0.51) (0.50) (−0.28) (0.66) 0.0464∗ 0.0493∗ 0.0544∗∗ 0.2843 0.0475∗ Syndicate CVC (1.81) (1.90) (2.02) (1.61) (1.83) 0.069∗∗ 0.0746∗∗ 0.0752∗∗ 0.7086∗∗∗ 0.0759∗∗ CVC Strategic Re. (2.29) (2.44) (2.47) (3.68) (2.48) 0.4374∗∗∗ 0.4451∗∗∗ 0.4452∗∗∗ 5.4843∗∗∗ 0.433∗∗∗ ln(MSCI mon. 3) (3.52) (3.54) (3.55) (7.23) (3.45) −0.2742∗∗ −0.293∗∗ −0.294∗∗ −7.1869∗∗∗ −0.2915∗∗ ln(MSCI mon. 6) (−2.3) (−2.42) (−2.42) (−9.49) (−2.41) 0.0094 0.0097 0.0098 −0.3961∗∗∗ 0.0093 ln(Industry MB) (0.85) (0.86) (0.88) (−8.29) (0.83) 0.0575∗∗∗ 0.055∗∗∗ 0.055∗∗∗ 0.4816∗∗∗ 0.0529∗∗∗ Later stage dummy (3.99) (3.79) (3.76) (4.19) (3.65) Year Fixed Effect Yes Yes Yes No Yes Industry Fixed Effect Yes Yes Yes No Yes The t statistic is in the parentheses for Model 1, 2, 3, 5 and 6. The Z-value is in the parentheses for Model 4. ∗ ∗∗ , and ∗∗∗ denote statistical significance at 10%, 5% and 1% level. Model 1 −0.0058 (−0.45)

50

Model 6 0.0493∗∗∗ (4.04)

0.0248∗∗∗ (5.33) 0.0017∗ (1.72) 0.0414 (1.57) 0.0754∗∗ (2.43) 0.4246∗∗∗ (3.32) −0.2574∗∗ (−2.09) 0.0126 (1.11) 0.1023∗∗∗ (7.37) Yes Yes

Table 15: Successful Exit Rate - SIC2 Dependent Variable: Failure rate Model 1 Model 2 −0.0007 CVC (−0.06) 0.0002 CVC per (0.01) −0.0052 −0.0052 ln(Duration) (−0.92) (−0.92) −0.0004 −0.0005 ln(Invest) (−0.09) (−0.12) −0.03 −0.03 Syndicate CVC (−1.13) (−1.18) 0.0045 0.0045 ln(VC Fund Size) (1.02) (1.03) 0.151∗∗ 0.15∗∗ CVC Strategic Re. (2.09) (2.07) −0.017 −0.017 Later stage dummy (−1.23) (−1.23) Year Fixed Effect No No Industry Fixed Effect Yes Yes The t statistic is in the parentheses. ∗ , ∗∗ and ∗∗∗ denote statistical significance at 10%, 5% and 1% level.

51

References [1] Aghion Ph. and Bolton P. (1992). “An Incomplete Contracts Approach to Financial Contracting”, The Review of Economic Studies, 59, p473494. [2] Akerlof G.A. (1970). “The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism”, The Quarterly Journal of Economics, 84, p488-500. [3] Basu S., Phelps C., and Kotha, S. (2011). “Towards understanding who makes corporate venture capital investments and why”, Journal of Business Venturing, 26, Issue 2, p153-171. [4] Chemmanur T.J. and Loutskina E. (2008). “How do Corporate Venture Capitalists Create Value for Entrepreneurial Firms?” Working paper. [5] Cumming D. and MacIntosh J.G. (2003). “Venture Capital Exits in Canada and the United States”, The University of Torento Law Journal, 53, p101-200. [6] Cumming D. (2008). “Contracts and Exits in Venture Capital Finance”, The Review of Financial Studies, 21, No. 5. [7] Da Gbadji L.A.G., Gailly B., and Schwienbacher A. (2011). “International Analysis of Venture Capital Programs of Large Corporations and Financial Institutions”, Working Paper [8] Dushnitsky G. and Lenox M.J. (2006). “When does Corporate Venture Capital Investment Create Firm Value?”, Journal of Business Venturing, 21, p753-772. [9] Dushnitsky G. and Shapira Z. (forthcoming). “Entrepreneurial Finance Meets Organizational Reality: Comparing Investment Practices and Performance of Corporate and Independent Venture Capitalists”, Strategic Management Journal. [10] Giot P. and Schwienbacher A. (2006). “IPOs, Trade Sales and Liquidations: Modelling Venture Capital Exits Using Survival Analysis”, Working Paper. [11] Gompers P.A. (1996). “Grandstanding in the Venture Capital Industry”, Journal of Financial Economics, 42, p133-156.

52

[12] Gompers P.A. and Lerner J. (2000). “The Determinants of Corporate Venture Capital Success: Organizational Structure, Incentives and Complementarities”, Concentrated Corporate Ownership, (Ed. Morck R.), University of Chicago Press. [13] Guo B. (2010). “Essays on Economics of Organizations”, Chapter 3, PhD Thesis, Universitat Aut`onoma de Barcelona. [14] Hellmann T. (2002). “A Theory of Strategic Venture Investing”, Journal of Financial Economics, 64, p284-314. [15] Hellmann T. (2006). “IPOs, Acquisitions, and the Use of Convertible Securities in Venture Capital”, Journal of Financial Economics, 81, p649679. [16] Hellmann T., Lindsey L., and Puri M. (2008). “Building Relationships Early: Banks in Venture Capital”, The Review of Financial Studies, 21, p513-541. [17] Ivanov V. and Xie F. (forthcoming). “Do Corporate Venture Capitalists Add Value to Start-up Firms? Evidence from IPOs and Acquisitions of VC-backed Companies”, Financial Management. [18] Macho-Stadler I. and P´erez-Castillo D. (2010). “Incentives in University Technology Transfers”, International Journal of Industrial Organization, 28, p362-367. [19] Masulis R.W. and Nahata R. (2009). “Financial Contracting with Strategic Investors: Evidence from Corporate Venture Capital backed IPOs”, Journal of Financial Intermediation, 18, Issue 4 (October), p599631. [20] Maula M. and Murray G. (forthcoming). “Corporate Venture Capital and the Exercise of the Options to Acquire”, R&D Management. [21] Riyanto Y.E. and Schwienbacher A. (2006). “The Strategice Use of Corporate Venture Financing for Securing Demand”, Journal of Banking & Finance, 30, p2809-2833. [22] Sahaym A., Steensma H.K., and Barden J.Q. (2010). “The influence of R&D investment on the use of corporate venture capital: An industrylevel analysis” Journal of Business Venturing, 25, Issue 4, p376-388. [23] Schwienbacher A. (2009). “Venture Capital Exits, in Companion to Venture Capital (Ed. D. Cumming), Wiley/Blackwell. 53

[24] Siegel R., Siegel E., and MacMillan I.C. (1988). “Corporate Venture Capitalists: Autonomy, Obstacles, and Performance”, Journal of Business Venturing, 3, p233-247. [25] Smith D.G. (2005). “The Exit Structure of Venture Capital”, UCLA Law Review, 53, p315-356. [26] Sykes H.B. (1990). “Corporate Venture Capital: Strategies for Success” Journal of Business Venturing, 5, p37-47. [27] Toldra A. (2010). “Venture Capital Syndication and Firm Entry: Theory and Evidence”, Working Paper. [28] Yost M. and Devlin K. (1993). “The State of Corporate Venturing”, Venture Capital Journal, 33, p37-40.

54