A Theory of Corporate Venture Investing

First Draft: September 1997 This Draft: July 1998 A Theory of Corporate Venture Investing Thomas Hellmann Graduate School of Business Stanford University Abstract: While start-ups are of strategic importance to them, established corporations seem to play only a modest role in financing these start-ups. I develop a theoretical model that examines the importance of a strategic motivation for making venture capital investments. In a benchmark model with perfect contracting, the entrepreneur always prefers a strategic corporate investor. But if investors can take non-contractible actions the entrepreneur may prefer an independent venture capitalist. Corporate investors may be unable to commit not to shirk with support, not to exercise self-interested control, or not to invest in a rival internal venture. Outcomes depend critically on the extent to which a start-up complements or cannibalizes the profits of the established corporation. For correspondence: Graduate School of Business, Stanford University, Stanford, CA 94305-5015; Email: hellmann@leland.stanford.edu; Tel: (650) 723-6815; Fax: (650) 725-9932 I would like to thank Bharat Anand, Chris Barry, Severin Borenstein, Robert Burgelman, Marco DaRin, Serdar Dinc, Joshua Gans, Robert Gertner, Ronald Gilson, Paul Gompers, Joshua Lerner, Kevin Murdock, John Roberts, Garth Saloner and seminar participants at Berkeley, Columbia, Northwestern, Rochester, Stanford, and UCLA for their helpful comments. All remaining errors are mine. Introduction The last twenty years have witnessed the spectacular rise of the US venture capital industry. While the amount of funds committed to venture capital in 1977 was below $1 billion, it has risen to over 45 billion by 1997. Venture Capital prides itself for the financing of many highly innovative start-up companies, such as Cisco, Genentech, or Netscape, that have changed the structure of their industries. Examining the institutional features of venture capital, observers consistently emphasize the close involvement of the investors with the companies they finance. If investors play an active role in the development of these start-up companies, the identity of the investor becomes an important determinant of the venture process. In particular, venture capitalists could be independent investors, or they could belong to some established corporations.1 A priori, one may think that established corporations are natural candidates to engage in venture capital. Being an active player in a technology and/or product area related to that of a start-up company might be an incentive for them to become investors. Moreover, since many start-up companies innovate in markets related to those of established corporations, the established firm may be particularly keen to obtain a participating stake in those start-ups. Yet a remarkable feature of the US venture capital industry is that it is largely dominated by independent venture capital funds. Even though the available statistical evidence is very imprecise and often hard to interpret – thus I relegate its discussion to the appendix – overall it still suggests that established corporations seem to play only a relatively small role in the financing of entrepreneurial companies. The business press also documents how a number of established corporations tried to become active players in venture capital, and how many of these efforts were discontinued because of poor performance (Gleba, 1994, Yost and Devlin, 1993). In a case study of Apple’s venture capital program, Hellmann, Milius and Risk (1995) document in detail the difficulties encountered by this particular corporate venture capital Another possible affiliation is with exiting financial institutions, in particular banks. I discuss those in Hellmann (1997). 1 1 fund. The overall tone of this literature is that relative to independent venture capitalists, corporations face additional difficulties in practicing venture capital.2 3 These observations bring out a number of interesting and unexplored questions. How does the identity of a venture capitalist matter? How does a strategic motivation affect an investor’s behavior? Under what circumstances will an entrepreneur prefer to accept funding from such a strategic investor? And what can explain some of the apparent difficulties that corporations have in investing in early-stage entrepreneurial companies? In this paper I will develop a simple theoretical framework to address these questions. I consider an entrepreneur who is seeking funding for a new venture and who can decide to accept funding from a corporate investor or an independent venture capitalist. One can obviously obtain any kind of result by exogenously assuming differences in the behavior, ability, information or cost structure of these two types of investors. In this paper, instead, I take the approach of deriving endogenous differences in investor behavior that are derived solely from a more fundamental difference in their identities. In particular, the only primitive difference is the assumption that the corporate investor has a so-called ‘core business’ in a field that is related to the activities of the entrepreneur. From this primitive difference I show how the corporate investor, in addition to financial returns, also cares about the strategic impact that the new venture will have on its core business. This in turn will affect behavior of the investor and influence the entrepreneurs’ choice of investor. Obviously, this may be a premature assessment, given that the venture capital industry is still young and its institutional structure is still in flux. In fact, recently a number of wellknown corporations including Microsoft or Intel have tried themselves at venture capital, though it is still too early to tell how well they fared (Alster 1997, Darlin, 1996). 3 Obviously, corporations spend considerable amounts on R&D. The difference is that R&D constitutes an internal mechanism of creating new businesses (as well as improving old businesses), while venture capital allows a corporation to participate in the innovations that are generated by the market at large. Established corporations also frequently acquire entrepreneurial companies. The difference there is that acquisitions typically occur at a later stage in the development of new companies. Indeed, independent venture capitalists consider the sale of their portfolio companies to established corporations as one of the main paths to liquidity. Thus, the point is not that established corporations play a small role in financing of innovation at large, but more specifically 2 2 Central to the analysis is an understanding of the strategic objectives of the corporate investor.4 I assume that, if successful, the new venture generates an externality on the profits of the established company’s core business. This externality may be thought of as the net effect (or ‘reduced form’) of a potentially complex interaction between the corporation’s core business, the new venture and their environment. This externality may be positive, indicating that the new venture is a complement to the established company’s core business, or it may be negative, indicating that the new venture is a substitute (i.e., it ‘cannibalizes’ profits of the core business). This externality naturally generates a different objective function for the corporation (relative to the independent venture capitalists): in addition to the financial returns, the corporate investor is concerned about the effect of the entrepreneurial venture on the corporation’s core business. How does the existence of a strategic motive change the behavior of the corporation as an investor? For this it is useful to first understand the role of a venture capitalist in the financing of entrepreneurial companies. In addition to providing financing, investors can take a number of actions that have a significant effect on the performance of these companies. Gorman and Sahlman (1989) provide survey evidence showing that in addition to providing finance, venture capitalists can assist with strategic planning, management recruiting, operations planning or introductions to potential customers and suppliers. And the work of Gompers (1995), Hellmann and Puri (1998), Kortum and Lerner (1998) and Lerner (1994, 1995a) provides econometric evidence for the importance of monitoring, support and control activities by venture capitalists.5 that they seem to play a small role in financing external entrepreneurs at an early ‘venture’ stage. 4 The business literature that discusses corporate venture investing puts considerable emphasis on the strategic motives of corporate investors. Yost and Devlin find that 93% of their sample of corporate venture capitalists listed strategic objectives as one of their main objectives. The existence of these strategic objectives is also very apparent in the case studies of Hellmann, Milius and Risk (1995) and Kanter et al. (1990). Obviously, the focus on related business and strategic objectives implies that the analysis does not apply well to ‘unrelated’ venture capital investments for diversification purposes. 5 See also Barry (1994), Bygrave and Timmons (1992), Fenn, Liang and Prowse (1995) and Sahlman (1990). 3 If an entrepreneur chooses a corporate investor she will care about what support she can expect. Kanter et al. (1990) document in some detail how Analog Devices Enterprises, the corporate venture capital arm of Analog Devices Inc., provided support to its portfolio companies. But as noted by Hardymon, DeNino and Salter (1983) the strategic interest of the corporate investor may create different incentives than those of independent venture capitalists. Indeed, the literature on corporate venturing argues that in many cases the existence of strategic motives causes the entrepreneur to distrust the corporate investor. Block and McMillan (1993), for example, note that: “…this distrust has two components: The first is the entrepreneur’s suspicion that the corporation will steal their ideas. The second is their fear that even if the corporations don’t steal ideas, they will control their ventures to satisfy corporate objectives at the expense of the ventures’ well being.” In a survey of corporate venture capitalists, Siegel, Siegel and McMillan (1988) find that the amount of conflict is likely to increase the more the corporation pursues strategic objectives, and that the strategic orientation of the corporate investor seems to have a negative effect on the performance of the new ventures. While the business literature frequently mentions these conflicts of interest, there is also evidence of entrepreneurs clearly benefiting from the support of a corporate investor. Kotkin (1989) describes how an equity investment by Compaq Computer in Conner Peripheral lead to a relationship where Compaq used Conner’s disk drives in their computers. He notes that: “[Conner] received much more than money. It also got a major customer, not to mention marketing knowledge and technical expertise…” Looking at this more anecdotal evidence, it is apparent that investors can take actions that affect the performance of the entrepreneurial venture, and that these actions have an important non-contractible component. As a consequence, when choosing an investor, the entrepreneur needs to be concerned about the incentives of that investor. As a benchmark I show that in a first-best world where all relevant actions of the investor are contractible (and incentives therefore unimportant), the entrepreneur prefers to accept funding from the corporate investor. This is because under perfect contracting the two 4 parties can always implement the efficient contract that internalizes the externality that the new venture imposes on the company’s core business. This is true regardless of whether the new venture is a complement or a substitute. Against this benchmark, I examine how investors behave if their actions are not contractible. The first variant of the model examines the incentives for an investor to add value by supporting the new venture when support is privately costly. I show that if the new venture is a complement, the corporate investor provides more support than an independent venture capitalist, and the entrepreneur indeed prefers the corporate investor. This result echoes a notion frequently encountered in the business literature that corporate investments are particularly effective when the corporation and the entrepreneurs can be put into a “win-win” situation.6 If the new venture is a substitute, however, the corporate investor always provides less support. Although the corporate investor would like to commit to a higher level of support ex-ante, this is not credible ex-post. As a consequence the entrepreneurs chooses the venture capitalist for most parameter values.7 This result has an intriguing ‘Schumpeterian’ interpretation: if the new venture destroys some of the existing value of the corporation, the venture capitalist has a comparative advantage at financing this creative destruction, because he is not linked to the existing rents. The problem for the corporation is that it cannot commit to creatively self-destruct. An example that illustrates the spirit of this result is GO, a start-up venture in pencomputing. The founder Jerry Kaplan (1995) describes his interaction with IBM as a strategic partner. After agreeing to invest, IBM failed to provide any value-added to GO. Indeed, IBM even undermined the strategic investment when it decided to push its own operating system OS/2 into pen-computing. The troubled relationship with IBM stands in marked contrast to interactions of GO with their venture capitalist. John Doerr, a partner Kanter et al. (1990) for example notes that one of the key evaluation criteria for a corporate investor is to look for “good opportunities for complementarity with existing core businesses.” 7 The exception is a situation where the new venture has low financial and private returns to the entrepreneur. In this case the corporate investor may find it worthwhile to outbid the venture capitalist in order to prevent the venture capitalist from inducing excessive cannibalization. 5 6 of the highly reputed Kleiner, Perkins, Caufield and Byers fund, was the main venture capitalist involved with GO, and he strongly supported GO throughout its period of existence.8 Closely related to the investor’s incentive to provide support is the issue of how an investor may want to influence or control strategic decisions in the new venture. Such control is particularly relevant and problematic when there is a trade-off between what is good for the new venture and what is good for the company’s core business. Hellmann, Milius and Risk (1995), for example, document how Apple faced the problem that entrepreneurs were concerned with being taken advantage of in times of conflict. There was a clear temptation for Apple to force its portfolio companies to favor the Apple platform over the IBM-PC platform. While this might help Apple’s core business, it might not be in the best interest of the entrepreneurial company.9 The second variant of the model addresses the issue of control. In this case the private actions of the investor involve a trade-off between the success of the new venture and the extent to which the new venture will complement (or not cannibalize) the corporation’s core business. I show that the corporate investor always chooses a strategy that is ‘too complementary’ while the venture capitalist chooses a strategy that is ‘too cannibalizing.’ In this case it is not so much the level of the externality (i.e., complementarity or substitutability), but the extent to which an investor will attempt to distort the externality that determines the entrepreneur’s choice of investor. An interesting implication of this analysis is that a certain amount of ex-post interference by the corporate investor is actually efficient. If, however, the amount of ex-post interference becomes excessive the corporate investor may again lose his ability to finance the entrepreneurial venture ex-ante. This result thus suggests that some of the observed conflicts of interest between entrepreneurs and corporate investors may well be part of a (second-best) efficient equilibrium. While GO, like many other venture capital investments, eventually failed, this example illustrates the different willingness by the venture capitalist to provide critical inputs. 9 Apple tried to solve this problem in a peculiar way. It created such strong commitment devices not to take advantage of entrepreneurs that in the end the venture capital division was unable to achieve any significant strategic benefits at all. 6 8 Another issue that concerns entrepreneurs when approaching a corporate investor is that even if the corporation is not a current competitor, it may become one in the future. A clear example is DEC’s investment in Mips Computer System (McQuade and GomesCasseres, 1992). Mips was developing a new generation of semiconductor technology based on the so-called RISC architecture. At the time that DEC invested in Mips it said that it would discontinue its research project on RISC. Two years later, however, DEC introduced the so-called Alpha chip that used RISC architecture that was competing with the Mips chip.10 In the final variant of the model I examine a situation where the entrepreneur is concerned that the corporation may invest in a rival internal venture. I show that while the new venture may seem to be a complement to the corporation, it may become an effective substitute if the corporation decides to engage in the rival internal venture. If the entrepreneur anticipates this, she will avoid the corporate investor. Interestingly, the fact that the corporation is a potential competitor may make it worse off. If there are rents to investing in the entrepreneurial company and if the rival internal venture is only marginally profitable, the corporate investor would prefer to commit not to invest internally, in order to maintain his credibility as an external strategic investor. An interesting aspect of this model is that the entrepreneur’s choice of an investor may sometimes deter the corporations from investing internally. I show that this entry deterrence effect may go either way. In some cases, the entrepreneur chooses the corporate investor in order to dissuade it from making a second rival bet on a related opportunity. In other cases, however, the entrepreneur deters entry by accepting funding from an independent venture capitalist. In this case, it is the venture capitalists high level of support that dissuades the corporation from investing in the internal rival venture. 10 There is some ambiguity about whether DEC discontinued and re-initiated or simply never discontinued its internal R&D on the RISC architecture. In other cases, the issue of potential competition is explicitly recognized. Kaufman (1996) quotes a partner at AT&T Ventures, the corporate venture capital arm of AT&T, saying: “It’s certainly possible that some people at AT&T at some point will say ‘We’ve invested in an unusually interesting company, shouldn’t we consider competing against it?’ ” 7 In contrast to the results of the first variant of the model, the second and third variant have the interesting implication that even if the new venture is a complement the corporate investor may not be chosen for lack of credibility. Together, the three variants of the model endogenously explain some of the behavioral differences between a corporate and an independent investor. The model does not attempt to explain all the potential differences between these two types of investors (or indeed, additional differences with these classes of investors). Instead, it hopes to provide a framework that derives a limited number of differences endogenously.11 To develop the model I draw on a number of theoretical and empirical papers. The recent corporate finance literature emphasizes the importance of investor’s incentives (see, for example, Diamond (1991) or Rajan (1993)). The literature on innovation recognizes the importance of the existing rent structure on the incentive to innovate (Gilbert and Newberry, 1982, Reinagum, 1989).12 Bharat and Galetovic (1998) examine the importance of weak intellectual property rights on the ability of a corporation to invest in new ventures (see also Anton and Yao (1994, 1995)). Huang and Xu (1998) examine the importance of soft-budget constraints in a related context. Gans and Stern (1998) examine the incentives of an established company to engage in a patent race with a startup, when the established company can potentially acquire the start-up at a later stage. This paper also relates to the work of Teece (1992), that argues that collaborations between innovators and established companies are more viable in the presence of complementarities.13 The remainder of the paper is organized as follows. Section 1 develops the basic model. Section 2 examines the case where investors can provide non-contractible support to the new venture. Section 3 examines the implications of investor control. Section 4 considers the case where the corporation may also engage in a rival internal venture. The 11 12 In the conclusion I return to some of the aspects omitted by this approach. This literature, however, largely ignores the moral hazard issues central to this paper. See Holmström (1989) for an interesting exception in a somewhat different context. 13 Apart from the references already mentioned, useful studies of the process of corporate venture investing include Bleicher and Paul (1987), Burgelman (1984), Burgelman and Sayles (1986), Fast (1981), Lerner (1995b), Rind (1981), Sykes (1986, 1990), Sykes and Block (1989) and Zahra (1996). 8 conclusion summarizes the testable predictions of the model and discusses future research avenues. Section 1: The basic model Consider a risk-neutral world with no discounting. An entrepreneur (EN) wants to start a new venture and has secure property rights or unique skills to manage it. The new venture succeeds with probability q, generating a verifiable return π, or it fails, generating no returns.14 Apart from any monetary gains, the EN derives a private benefit from success, the monetary equivalent of which is β.15 The EN has no wealth and requires an investment of L. She has access to a perfectly competitive pool of venture capitalists (VCs). Alternatively she may accept financing from a unique corporate investor (CI). I denote the investor share by α.16 I am interested in a scenario where it is impossible for an investor to buy the new venture up-front, or more generally to make any transfer payment to the EN. For this I assume that there is an adverse selection problem ex-ante that prevents any investor to pay the EN up-front for her idea.17 Apart from this adverse selection problem, all parties have symmetric information. 14 q may represent a wide variety of risks, such as technology risk, development risk, the risk of market acceptance or the probability of winning a patent race. 15 This can be motivated in at least three ways. First, consistent with much of the finance literature, the EN may have the ability to enjoy some (additional) ‘on-the-job consumption’ in the event that the company is successful. Second, the success of the new venture is likely to increase the market’s perception of the EN’s human capital and β may represent the increased rents from that. Finally, it may simply be that the EN derives utility directly from the success of the new venture, be it personal satisfaction, or the enjoyment of creating value to other parties such as consumers. Brockhaus (1982) summarizes some of the evidence on the motivation of entrepreneurs. 16 Since π is a constant the capital structure of the new venture is irrelevant. W.l.o.g. I assume that α represents an equity stake. Admati and Pfleiderer (1994) and Hellmann (1994) discuss issues of capital structure in venture capital contracts. 17 Specifically, assume that there is a large pool of ‘impostors’ that can pretend to have a venture similar to that of the EN but with a zero probability of success. These impostors have a reservation utility zero, so that they will only enter if offered a positive payment. No investor will thus ever make any up front payments. Below I also show that one can use moral hazard, rather than an adverse selection, to justify why the investor doesn’t make any transfer payments ex ante. 9 The timing of the game is as follows. At date 0 the VCs and the CI make offers αv and αc to the EN. She chooses her preferred bid. Between date 0 and 1 the selected investor may take some action ‘a’ that has some associated costs c(a). At date 1 the project is either successful or not and returns are paid out. To model the strategic objectives of the CI, I assume that the CI’s core business’ profits are affected by the success of the EN. Specifically θ denotes the change in the CI’s core business’ profits that is caused if the EN succeeds. Put differently, θ is the net difference in the utility of the CI in case that the NV succeeds, not including any possible gains from an equity position in the NV. If θ > 0 the new venture is a complement to the CI’s core business, and if θ < 0 the new venture is a substitute. The parameter θ allows us to give a concrete meaning to the notion of a strategic investment motive. Apart from considering any effects on the value of its investment in the NV, the CI will also consider any ‘strategic’ effects on its core business, namely the change in profits, as measured by θ. The specification is meant to be very general, so as to capture a wide variety of situations. The new venture may have an impact on the demand for the CI’s core products, by either cannibalizing it or complementing it.18 It may have an impact on the cost structure of the CI.19 The model is also sufficiently general to accommodate an interpretation where managers don’t always act in the best interest of the shareholders. Most price and quantity competition models suggest that θ is negative, as the industry changes from a (possibly differentiated) oligopoly of n firms to one with n+1 firms. Positive demand externalities are typically obtained from models where households consume a bundle of products or services and where consumption of one good enhances the value of another. Software and hardware, for example, may have positive demand externalities (Katz and Shapiro, 1994). The specification also captures Tushman and Anderson’s (1986) notion of competence-enhancing and competence-destroying innovations. 19 If the new venture reduces cost of the CI, such as by providing a cheaper input, the CI considers the new venture a complement. If it raises the cost, such as by raising the price of a scarce common input, the CI considers the new venture a substitute. There may also be indirect effects: if the new venture lowers the cost of a competitor to the CI, the CI will view the new venture as a substitute. 18 10 For example θ may represent the change in the human capital value of the incumbent management team.20 For simplicity I will treat θ as an exogenous variable. But the model is also fully consistent with a broader interpretation where θ is the outcome of a possible complex endogenous interaction between the CI’s core business and the NV. All we need to assume for this broader interpretation is that these actions happen at (or after) date 1, and that VC and the CI can bargain efficiently at date 1.21 While θ may thus be affected by the actions of all the parties at date 1, at date 0 these parties know that they can negotiate the most efficient arrangement at date 1. For example, if a monopoly is more efficient than a duopoly, then the parties can get together at time 1 and create a monopoly by rearranging ownership accordingly. With Coasian bargaining, all that matters at date 0 is thus the distribution of rents in case that the NV succeeds or fails. This distribution is fully characterized by the parameters α, β, θ and π.22 23 Throughout the paper I assume that there are no intrinsic differences in the information, costs or capabilities of the CI or the VC. The CI and the VC may, however, Another related interpretation is that θ represents the change in the organizational capabilities of the corporation. Henderson and Clark (1990) argue that architectural changes that destroy existing capabilities of the organization are most likely to be resisted by incumbent firms. 21 In section 3 I relax this further and examine a situation where θ can also be influenced by actions that happen prior to date 1. 22 If we have Coasian bargaining at date 1, then the two parties always allocate the asset to the party that implements the more efficient outcome. Subsequent to the allocation of ownership rights, there may be any vector of endogenous actions x involving the CI, the VC, the EN, and possibly even other parties. I denote the equilibrium of this game by x*. If, for example, joint profits are maximized when the CI controls the NV, then the CI’s (ex-post) utility is given by uC = uB(x*) + uN(x*) – π in case that the NV succeeds, where uB and uN are the profits of the core business and the NV respectively, and π is the price of the acquisition. In this case θ depends on the endogenous action x* of all the players, since θ = uB(x*) + uN(x*) – π - uB(no NV). Alternatively, if the CI was the original investor then its utility at date 1 is given by uC = uB(x*) + uN(x*) – (1-α)π so that the difference in utility between success and failure of the NV is given by uB(x*) + uN(x*) – (1-α)π - uB(no NV) = θ + απ. In both cases θ, π and α provide a sufficient characterization of the effect of success on the CI’s utility. 20 11 choose to take different actions, denoted by a ∈ A. These actions will affect variables such as q(a) or θ(a), but these variables have the same functional forms, irrespective of whether the CI or the VC finances the EN. This assumption may seem ‘unrealistic’ at first because there are many potential differences between a corporate and independent venture investor. The point of this paper, however, is not to assume these differences, but rather to derive them from more primitive assumptions about the identity of the investor. The most natural way of deriving differences is thus to start with a ‘level playing field,’ where investors are a priori the same. Put differently, the model focuses on deriving behavioral differences between the VC and the CI, as opposed to assuming exogenous differences in their ability or information. It is also straightforward to extend the analysis of the paper by also introducing some exogenous differences. Define the utilities of the CI and a representative VC as UC(αc) = q (αcπ + θ)- L c and UV(αc) = 0 if the CI finances the EN, and UC(αv) = qθ and UV(αv) = q (αvπ)- L - c if the VC finances the EN. The EN’s utilities are given by UE(αc) = q ((1-αc)π + β) and UE(αv) = q ((1-αv)π + β). Throughout the paper I assume that L is sufficiently low so that the VC is willing to finance the EN.24 The bidding game between the CI and the VCs can be solved as follows. The EN always accepts the bid that provides her with the highest utility. A VC faces perfect competition and makes a bid that gives him either zero expected profits or the minimal amount of efficiency rents.25 In order to get his bid accepted the CI has to make an offer that matches the EN’s utility from accepting a competitive bid from a representative VC. For simplicity I assume that if the EN is indifferent between an offer from the CI and the For simplicity I assume that β, π and θ are all fixed parameters, but nothing changes in the model if these are random variables that are resolved at date 1. 24 Moreover I assume that the participation constraint of the EN is always satisfied. 25 We will see that because the amount of equity held by the investor affects support, the EN may find it in her interest to give the investor some minimal equity in order to increase support. The implied rents cannot be competed away because of the inability of investors to make ex-ante payments to the EN. 23 12 VC, the EN always chooses the CI. The CI makes such a matching offer whenever his utility of financing the EN is higher than his utility if the VC were to finance the EN. The following result establishes a benchmark for the main analysis of the paper. Result 1: If all actions are verifiable then the CI finances the EN for all θ. In the appendix I formally state and prove all the results of the model. If all the actions of the investor are verifiable, then the EN prefers to obtain financing from the CI. The intuition is that if the CI finances the EN the contract will always specify the action that internalizes any externality on the CI’s core business. But if the VC finances the deal, the contract will in general specify a different action. Since the total surplus generated by CI financing is higher than under VC financing (and strictly higher whenever the VC’s optimal action generates a strictly lower sum of utilities) the CI is willing to bid more for financing the EN. The important point to note is that this result holds for all values of θ, both positive and negative: the nature of the externality does not matter in a first-best world. The critical assumption of result 1 is that the action ‘a’ can be contractually specified. In the following sections we ask whether or when the CI’s ability to internalize the externality on his core business provides him with a competitive advantage when certain actions are not contractible. The benchmark case is obviously not meant to represent all the differences between a corporate and independent investor. It is the theoretical outcome when there is a level playing field between these two types of investors and there is perfect contracting. As such, it will be useful to evaluate the importance of incentive and commitment problems that the different types of investors are likely to face. Section 2: Investor support In this section I examine which investor is more likely to finance the EN in the presence of a non-contractible action. In particular I look at the investor’s incentives to ‘support’ the entrepreneurial company. By ‘support’ I try to capture a broad class of 13 activities that investors can take to increase the value of the new venture.26 These activities are by and large subtle so that it would be difficult to specify them in a contract. If support is not verifiable the entrepreneur relies on the goodwill - i.e., the incentives - of the investor. I model support in the simplest possible way by assuming that the probability of success q depends on the private effort of the investor. In particular q = q0 + q1 where q0 is a fixed component and q1 the variable component that is determined by the investor’s support. The investor has a private cost of providing support. For simplicity I use a quadratic specification where the private costs are given by c(a) = c(q1) = (q1)2/(2K).27 The larger K, the lower the costs of providing support, so that K measures the ease of providing support. If q1 were verifiable then the CI and the EN would agree to set q1 = q1FB = K(π + θ + β) (where FB stands for first-best), and result 1 would apply. If q1 is not verifiable, the optimal (second-best) contract can only specify the allocation of equity αc or αv. Given αc and αv, the CI and the VC choose q1 according to: q1(c) = K (αc π + θ) (CI) (VC) q1(v) = K αv π αc and αv are endogenously determined in the bidding game. For any given αv, the CI needs to offer some αc such that UE(αc) ≥ UE(αv). If θ > 0 then the CI can afford to take more equity than the VC and still get the deal, i.e., αc > αv. This is because if he were to simply match the VC’s equity offer he would promise the EN a higher probability of 26 Investors may provide useful guidance and mentoring to inexperienced entrepreneurs. They may improve the likelihood of success by assisting in the recruiting of key managers, they may use their business contacts to broker relationships with potential suppliers, customers or other related parties. They may also be useful in monitoring performance, certifying the company with other investors and possibly exercising valuable corporate governance. 27 Hölmstrom and Milgrom (1987) show how the linear quadratic model can be derived from a dynamic model where the agent - the investor in our case - controls a Brownian motion. 14 success.28 But if θ < 0 then the CI can take less equity than the VC and still get the deal, i.e., αc < αv. But the lower ownership stake further dilutes the CI’s incentives to support the new venture. This makes the CI a less efficient investor. Indeed, the following result establishes under what circumstances the CI is willing to outbid the VC to finance the EN. Result 2: Suppose investors provide non-contractible support. If the new venture is a complement (θ > 0) then the CI always finances the EN. If the new venture is a substitute (θ<0) then the VC finances the EN, unless the new venture has sufficiently low financial and private returns (i.e., L is sufficiently large and β is sufficiently small). The key insight from result 2 is that with non-contractible support the CI may no longer get to finance the EN. If the new venture is a substitute (θ < 0) it may be that the ability of the CI to internalize the externality on his core business becomes a hindrance. In particular, it prevents the CI from providing an adequate level of support to the EN. Exante the CI would like to commit to a high level of support, but ex-post he cannot promise to deliver it. The CI can try to make up for his deficiency by offering the EN a larger ownership stake, but this only works for a limited range of parameters. If the EN strongly cares about the success of the new venture (high β) she will be unwilling to accept the lower probability of success that comes with CI financing. Put differently, the ownership stake that the CI needs to give to the EN to induce her to accept CI financing is so large that the CI does not find it worthwhile anymore to finance the EN. With θ<0, this result holds for most parameter values. To understand the exceptions consider Figure 1, which shows q1 as a function of L. The two horizontal lines indicate the first best levels, q1FB, when β is large and small. The two upward sloping lines indicate the actual levels of q1(c) and q1(v) as a function of L, where q1(c) is always below q1(v). If β is large then both the VC and the CI provide too little support relative to Formally, if αc = αv = α, then q1(c) = K(απ+θ) > Kαπ = q1(v), and thus UE(CI,α) >UE(VC,α). 28 15 the first best. But since the VC always provides more support he is the more efficient investor. If β is small then, for sufficiently large values of L, the VC provides too much support, relative to the first best. Support is excessive as it does not take into account the cannibalization of the CI’s core business. In this case the corporate investor is willing to outbid the VC in order to prevent excessive cannibalization. 29 If the new venture is a complement (θ>0) the above reasoning is essentially reversed. In this case the CI can always commit to a higher level of support. This even allows him to take a larger ownership stake, further increasing his incentives to provide support. The VC cannot promise to provide an equivalent amount of support, and the CI always finances the EN.30 The model also makes some prediction about the valuation of the equity investment. We saw that if θ > 0 the CI offers a lower valuation (lower α) than the VC: the EN accepts the lower valuation because she can expect a higher level of support from the CI. If θ < 0 then the CI must offer to pay a higher valuation (higher α) than the venture capitalists in order to compensate for the lack of support. He will only be successful in wooing the EN if β is sufficiently small and L sufficiently large. The model so far emphasizes the moral hazard problem for the investor, i.e., the VC and the CI have different incentives to provide support to the EN, and this difference can explain when the VC is more efficient than the CI. While the investor’s support has an important impact on the success of the new venture, it may be argued that the EN’s effort should also have an impact on the likelihood of success, and that this may sometimes be a more important determinant of success. It seems therefore worthwhile to 29 Note also that this result does not depend on the existence of private benefits. Even with β = 0 we can find a critical value of L, such that above (below) it the CI (VC) finances the NV. 30 This result clarifies some confusion in the business literature that tries to resolve whether a high degree of ‘relatedness’ between the entrepreneurial and established company benefits or harm the process of corporate venture investing (Sorrentino and Williams, 1995). Relatedness should be viewed as the absolute value of θ. Result 2 shows that a high degree of relatedness can indeed be helpful or harmful for corporate investing. What matters is not just relatedness, but the ‘direction’ of relatedness, namely whether the companies are complements or substitutes. 16 consider an extension of the model where there is not only the ‘support’ of the investor, but also effort by the EN that affects the performance of the new venture.31 In the appendix I extend the present model to allow for a two-side moral hazard problem where the probability of success depends on the private efforts of both the investor and the entrepreneur. I show that, irrespective of the relative importance of the investor’s support versus the EN’s effort, result 2 continues to hold. The intuition is as follows. If both the investor and the EN need incentives then there is a ‘team problem’ where (for most parameter values) both the EN and the VC are providing too little effort. If the equity share α were the same for the CI and the VC, then the CI has stronger (weaker) incentives if θ > 0 (θ < 0). In this case the CI is a more (less) useful ‘team partner’ because he has a stronger (weaker) motivation. In the appendix I show that this intuition carries forward even after αc and αv are determined endogenously (and also for those parameter values where it isn’t true that both the EN and the VC provide too little effort). Another attractive feature of the model with two-sided moral hazard is that it provides an alternative explanation for why investors cannot buy the new venture at date 0. Even without the adverse selection problem the investor would not want to buy the new venture, since this would undermine the EN’s incentives to add value to the new venture. In the appendix I show that, unless L is small, investors give the EN only equity, even if ex-ante transfer payments do not lead to adverse selection. Another way of interpreting this result is that both parties prefer that the entrepreneur be compensated with equity rather than wages, because equity provides better incentives to the entrepreneur. Section 3: Investor control The model of the previous section tries to capture in a simple way the notion that the strategic concern induces the CI to behave differently than a VC. For this I examined the incentive of the investor to add value to the new venture. One assumption of the 31 See also Hellmann (1994, 1998) for a more extensive discussion of this problem in the specific context of venture capital contracts. 17 model was that the externality of the new venture on the CI’s core business, as represented by θ, is exogenous. As I discussed in the introduction, however, some of the problems with corporate venture investing concern the corporation’s desire to alter the strategic impact of the new venture on its core business. In terms of the model, this means that θ might be affected by private actions of the investor, i.e., that θ is endogenous. Whenever the demand or cost of the two firms’ products are related, there will be strategic decisions for the new venture that have an impact on the CI’s core business. These types of decisions impact not only the new venture itself, but also the CI’s core business. The new venture may have to decide on how to differentiate its product from that of the CI.32 It may have to make some technological choices that affect the compatibility of its product with the product of the CI. Or, if the new venture is a supplier to the CI, it may have to decide on the extent of asset specific investments. When decisions impact both the new venture and the CI’s core business, the CI will have a strategic interest in controlling the decisions of the new venture. Again, we ask the question how this strategic motive affects the CI’s likelihood of financing the EN. Our model allows us to address this question in a very simple, ‘reduced-form’ manner. Suppose that there are actions that affect both the expected returns of the new venture and its externality on the corporation’s core business, i.e., there are actions a ∈ A that affect both θ and q. For any subset of actions that give the same value of θ, the investor will always choose the action a* that gives the highest possible q. 33 For every θ there is thus an associated q(a*(θ)), so that we can trace q as a function of θ. This means that we can think of the choice of the action ‘a’ as the choice of θ. For simplicity I assume that q is a smooth concave function of θ with dq/dθ = 0 for some θ0. Above θ0 there is a binding trade-off between the expected returns of the new venture and its externality on In a Hotelling model, for example, a location closer to the CI’s location implies a lower level of differentiation and thus a higher degree of substitutability. 33 This assumes that αcπ + θ ≥ 0. If this condition is not satisfied, the CI is unlikely to finance the EN because he may be choosing actions that minimize, rather than maximize q. 18 32 the CI’s core business. Note that this specification does not depend on whether θ is in the positive or negative range. The problem for the EN is that the choice of an investor implies being exposed to a certain type of control. In this paper I will not attempt to explain the mechanisms through which the investor exercises such control.34 Instead I assume that the investor has some control, and I examine the implications of it. In terms of the model this simply means that the investor gets to choose the actions a (and therefore θ). This choice is assumed to be non-contractible. This seems reasonable, given that the actions essentially represent future business decisions that are difficult to anticipate ex-ante, or, even if anticipated, cannot be objectively verified by a third party. Abstracting for a while from the issues of support modeled in section 2, I state the main result of this section. Result 3: If the degree of complementarity or substitutability (θ) can be controlled by the investor, the VC will always choose too little complementarity (or too much substitutability), i.e., θv < θFB, and the CI will always choose too much complementarity (or too little substitutability), i.e., θc > θFB. The larger the financial returns (small L) and/or the larger the private returns (large β) of the new venture, the more likely the EN will accept the VC rather than the CI as an investor. The intuition for result 3 can be seen from Figure 2, which depicts q(θ), q(θ)(π+θ) and q(θ)(π+θ+β) as a function of θ. The VC maximizes q(θ)αvπ, while the CI maximizes q(θ)(αcπ+θ). The VC always chooses θv = θ0, i.e., the peak of q(θ), irrespective of αv and thus L. The CI’s choice of θ is influenced by L through its effect on αc. The larger the CI’s share αc, the more sensitive it is to the impact on the new venture’s probability of success, and thus the lower θc. Since θc ≥ θFB, it follows that the distortion (θc-θFB) is decreasing in L, suggesting that a higher L makes the CI relatively more efficient. Figure 19 2 also shows θFB for β=0 and β>0: the larger is β the higher the loss in the sum of utilities of the EN and CI of increasing θ beyond θ0, and thus the lower θFB. This shows why for larger values of β the EN prefers the VC. The main message from result 3 is that if the degree of complementarity or substitutability is affected by the investor, the VC will not make an efficient decision, as he fails to internalize the impact on the core business of the CI. The VC and the EN, however, agree on the strategic objectives of the new venture, namely to maximize its expected returns. If the CI finances the EN, a conflict of interest arises. A small amount of interference by the CI is beneficial since the CI internalizes the externality on his core business. But because the success of the new venture has to be shared with the EN, the CI excessively favors his core business when making strategic decisions. The more profitable the new venture (lower L), the lower ownership stake by the CI (lower αc), the bigger the conflict of interest and thus the less likely that the EN will choose the CI as the investor. Moreover, the higher β, the higher the costs of the conflict of interest and again the less likely that the EN will choose the CI as the investor. An important aspect of result 3 is that it holds for any range of θ, positive or negative. This contrasts with result 2, where the NV always is financed by the CI for θ positive. The difference for result 3 is that it is not concerned with the level of θ per se, but rather about the changes – or inefficient distortions - of θ.35 An interesting extension is to consider an alternative timing of the control decision. So far I considered control in the early stage of the venture, where decisions affect the likelihood of success q. One can also think of strategic decisions that are only made once the new venture has proven to be viable. For example, there may be a trade-off 34 See Berglöf (1994) and Hellmann (1998) for models that derive investor control from first principles. The notion of control used in this paper, however, is relatively mild. We can think of it more in terms of influence, rather than stark control. 35 If we add the problem of support from section 2, we can see how the two actions interact. The tendency of the CI to make decisions that increase θ now has an additional effect that is actually positive, namely that a higher θ increases the CI’s incentives to support the new venture. In this case both the efficient level of θ and the incentive of the investors to distort θ affect the EN’s choice of an investor. 20 between θ and π. In this case the CI can always wait for the resolution of the uncertainty about success or failure. If the new venture is successful the CI may acquire it at this later stage and then implement the efficient choice of θ.36 It is straightforward to show that as long as there are no private actions in the early stage it doesn’t matter whether the VC or the CI makes the date 0 investment. But if we reintroduce the support problem of section 2, an outcome similar to result 2 obtains. Indeed, it is possible to show that the CI makes an early stage investment (at date 0) only if the new venture is a complement, but prefers to wait for a later-stage acquisition (at date 1) if the new venture is a substitute.37 This result may explain why corporations play a more significant role in acquiring later-stage venture capital backed companies than they play in early-stage venture investing. For example, in discussing why the large pharmaceutical companies don’t undertake their own biotechnology research, The Economist (1995) notes: “R&D is in some ways the perfect thing to buy in from small outside suppliers. Far from benefiting from economies of scale, laboratories tend to suffer from internal conflict as they become bigger.” Section 4: Internal versus external venturing So far I have analyzed the viability of corporate venture investments by examining a model where the CI is identical to the VC in his ability to invest in the new venture, but where the strategic motive of the CI alters his behavior as an investor. In this section I extend the model to introduce one fundamental difference between the CI and the VC. By virtue of operating in a related line of business, the CI may be in a position to invest in an internal venture that is a rival to the EN’s new venture. If the CI can invest in a rival internal venture, some interesting interactions between the EN’s and the CI’s decisions may arise. The CI’s incentive to add value to the EN’s new venture depends on whether or not CI invests simultaneously in a rival internal 36 Because uncertainty about the viability of the project is resolved, the adverse selection problem from date 0, that prevented early-stage acquisitions, should also be resolved at this later stage. 21 venture. Moreover, the CI’s incentive to actually invest in the rival internal venture might depend on whether or not the CI gets to finance the EN. This introduces a new dimension to the EN’s choice between a VC and the CI. In addition to the usual concern about support, the EN is now also considering how her choice between the VC and the CI might deter the CI from engaging in the rival internal venture. For the model I now assume that CI has the ability to engage in a rival internal venture that is an imperfect substitute to the new venture.38 I denote the rival internal venture by IV, its probability of success by qI, its profits in case of success by πI, and its costs by LI.39 NV denotes the EN’s new venture. I model the NV and the IV as imperfect substitutes, where the probability of success of one is negatively affected by the probability of success of the other. Formally, let qN = pN - cN qI and qI = pI - cI qN, where 0 0 and Θ < 0. The outcome of the game depends on whether the CI can commit himself to an investment decision. In game-theoretic terms, commitment is represented by a game where the CI chooses about the IV before the EN chooses the investor.41 Figures 3a and 3b depict the game tree (at date 0) for the games with and without commitment. The game with commitment has a simple solution. The EN prefers the CI if he does not invest in the IV, but she prefers the VC if the CI does invest. The CI has a choice of either investing internally and forgoing the benefit of becoming a strategic investor externally, or he commits not to invest internally and partners with the EN. This trade-off is affected by the profitability of the IV, which can be parameterized by its cost LI. In the ˆ appendix I show that there exists a critical value LCO (where CO stands for commitment) I below which the CI decides to invest in the IV, and above which he prefers not to invest in it. 40 This assumption considerably simplifies the analysis, since the EN’s preferences are solely determined by the amount of support provided by the investor. The insights of the model would also carry over to a more general specification. 23 Consider next the game without commitment. Once the EN has chosen the CI as an investor, the CI may still want to invest in the IV. Again, there exists a critical value ˆ LCI below which the CI will invest in the IV. Similarly, once the EN has chosen the VC I ˆ as an investor there exists a critical value LVC below which the CI will invest in the IV. I Given these choices, the EN has to decide whether or not to accept the CI or the VC as her financier. In the appendix I solve this game. Figures 4a and 4b describe the set of all possible outcomes. They show the outcome of the game as a function of the cost of the IV (LI). The outcomes of the game without commitment are shown above the line, and the outcomes of the game with commitment are shown below the line. There are four cost ranges that I call low, intermediate low, intermediate high and high. If the cost of the IV is low the CI always wants to invest, and if the cost of the IV is high the CI never wants to invest. In both these cases commitment is not an issue. But for the two intermediate cost ranges there may be differences between the games with or without commitment. Consider the case where the cost of the IV is ‘intermediate low.’ In this case the CI cannot commit not to invest in the IV. Anticipating this the EN chooses the VC as a financier. But if the cost of the IV is only ‘intermediate low’ as opposed to ‘low’ the CI would like to commit ex-ante not to invest in the IV, in order to have the opportunity to finance the EN. The return on the internal investment is actually not worth the loss of credibility as an external investor. The inability to commit not to invest internally hurts the CI in this case.42 Consider next the case where the cost of the IV is ‘intermediate high.’ In this case the EN’s strategy of choosing an investor includes a calculation that the appropriate choice of investor might prevent the CI from investing in the IV. 41 An equivalent distinction is that in the commitment case the IV is verifiable and can therefore be contracted upon: in the no commitment case, however, IV is not verifiable. 42 This problem is closely related to the problem of a ‘narrow business strategy’ (Rotemberg and Saloner, 1994). The ability of a firm to engage in a broad set of activities prevents it from being sufficiently committed to any one of them. In this model there is a trade-off between internal and external venturing. In order to be credible as an ‘external’ strategic investor the CI would have to commit to a ‘narrow strategy’ that prevents him from making rival internal investments. 24 In the first scenario (Figure 4a) the EN accepts financing from the CI. Once the CI owns a stake in the NV he no longer finds it worthwhile to invest in the IV. Without the IV he also delivers a high level of support. In this scenario the EN deters entry by ‘coopting’ the CI, i.e., by accepting him as a strategic investor. In the second scenario (Figure 4b) the EN cannot induce the CI to drop the IV by making him a strategic investor. However, by aligning with a VC the EN can expect a high level of support. She becomes a sufficiently strong competitor that the CI no longer finds it worthwhile to pursue the IV. In this case the EN deters entry by ‘confronting’ the CI, i.e., by accepting the VC as an investor. For a small range of parameters the commitment problem for ‘intermediate low’ costs also applies to ‘intermediate high’ costs. In this case the EN accepts the VC, but the CI still invests in the IV. This happens in Figure 4a when the CI invests in the IV and the ˆ ˆ VC in the EN in the range [ LCI , LVC ]. In the appendix I show that there exists θ* > 0 such I I that θ ≥ θ* is a sufficient condition for this case not to occur. I will use this condition for the remainder. An interesting question is under what circumstances the EN is more likely to coopt or to compete with the CI. In the appendix I show that the greater the ease of providing support (high K), the more likely the EN will co-opt the CI. The intuition is that if the cost of providing support is low, the CI can create significant value by providing support to the NV. This makes the opportunity cost of investing in the IV large. Under these circumstances the EN has ample opportunities to influence the CI’s investment decision by co-opting him through a strategic investment. I also show that the higher θ, the more likely the EN will co-opt the CI. The intuition is that the CI can offer a higher level of support. Result 4 summarizes the main findings of this section. Result 4: Suppose that the NV is a complement (θ>0) and that if the CI invests in the IV it becomes a net substitute (Θ<0). Then, for sufficiently large β, a.) there exists a range where the cost of the IV is ‘intermediate low’ so that the CI would like to commit ex-ante not to invest in it, but always wants to invest in it ex-post. The 25 CI loses his credibility as a strategic investor and the cost of losing credibility exceeds the expected benefit from investing in the IV. b.) there exists a range where the cost of the IV is ‘intermediate high’ so that i.) either the EN chooses to ‘co-opt’ the CI by making him a strategic investor; once a strategic investor, the CI no longer wants to invest in the IV. ii.) or the EN chooses to ‘confront’ the CI by accepting the VC as a financier. Once the VC backs the NV, the CI no longer finds it worthwhile to invest in the IV. iii.) or, if θ is sufficiently small, it may be that the EN chooses to ‘confront’ the CI by accepting the VC as a financier, but the CI still finds it worthwhile ex-post to invest in the IV. Ex-ante, however, the CI would like to commit not to invest in the IV. c.) The more elastic the support (high K), and the smaller the strategic externality (low θ), the more likely the EN will confront rather than co-opt the CI. Conclusion This paper examines a model where a corporation considers making a strategic investment in an entrepreneurial company. It has to compete with independent venture capitalists that do not take into account the externality that the entrepreneurial company imposes on the established corporation. With perfect contracting the entrepreneur always prefers the established corporation as an investor, but if investors can take noncontractible actions the entrepreneur may prefer an independent venture capitalist. The model generates a number of testable predictions. An entrepreneurial company is more likely to be financed by a corporate investor if it is a complement rather than a substitute to the corporation’s core business. If it is a substitute, however, the corporate investor is more likely to finance a deal if the deal is less promising, requiring the investor to take a larger portion of the equity. If the entrepreneurial firm is a complement (substitute) the corporate investors offers a lower (higher) valuation than an independent venture capitalist. The larger the opportunity for the corporate investor to 26 control the strategy of the entrepreneurial venture, the more likely the entrepreneur will chose an independent venture capitalist. Finally, the corporate investor is more likely to make a strategic investment in a particular area when it can commit not to make a rival internal investment in this area. While these predictions are testable in principle, an empirical examination would have to face a number of challenges. In particular, it may be difficult to measure the extent of complementarity or cannibalization. Conceptually there are two kinds of measures. First, one can try to estimate ex-post whether the successful entry of an entrepreneurial company has a positive or negative effect on its actual and potential corporate investors. Second, one could try to extract from interviews and source documents whether market participants consider the entrepreneurial firm to be a complement or substitute. The first approach may be more objective, although the second may be closer to what the model predicts to be the actual driver of behavior.43 The model makes some simplifying assumptions. In particular, the entrepreneurial company is assumed to have a single investor, and there is only one corporation that is a potential strategic investor. The model also excludes issues of asymmetries of information and the staging of finance (see Sahlman, 1990). More generally, the model does not attempt to explain all the potential differences between corporate and individual investors. Instead, it hopes to provide a methodology to analytically derive a limited number of endogenous differences. This way of thinking may then hopefully provide new ways of thinking about further differences. For example, it is sometimes argued that not having any talented venture capitalists is one of the problems hampering corporations. Rather than accepting this statement at face value (i.e., taking it as an exogenous difference) we may ask why corporations are unable to develop, attract or retain the most talented venture capitalists? While there may be a variety of reasons, the arguments developed in this paper may provide a first insight. If corporations pursue strategic objectives that may interfere with the maximization of returns for the entrepreneurial companies, then it may become harder to provide incentives (both at the 43 See also Athey and Stern (1996) for the issue of measuring complementarities. 27 level of the individual deal and the overall career path) to their venture capitalists. Future research could then examine this notion more carefully. This last example also points toward further intriguing avenues for future research. Another said reason why corporations seem to have difficulties to retain the most talented venture capitalists is the abundance of ‘internal politics.’ This paper only focuses on the external process between the corporation and the entrepreneur. And while it provides some clues about the likely nature of internal conflicts - between protagonists of the new venture versus the core business - it would be interesting to formally model the internal processes. One may also think of this as a problem of organizational design. Indeed, anecdotal evidence suggests that corporations differ in terms of the leeway they give to their managers in charge of corporate venturing. Some corporations have a separate venture capital subsidiary that has considerable autonomy while in other corporations top management exerts close control over the corporate venturing activities. Recently, some corporations have even taken the extreme route of ‘outsourcing’ all of their venture capital investments to an independent venture capital fund.44 Understanding the organizational differences would be a fascinating line of research. 44 Examples include the Adobe venture capital fund, managed by Hambrecht and Quist, and the Java fund, managed by Kleiner Perkins, Caufield and Byers, that has a set of Java-friendly investors lead by Sun Microsystems. 28 References Admati, A., and P. Pfleiderer, 1994, “Robust Financial Contracting and the Role of Venture Capitalists” Journal of Finance, 49:2, June, 371-402. Alster, N., 1997, “Venture Financing: tech firms try their hands” Investor’s Business Daily, October 22nd, p. A10 Anton, J., and D. Yao, 1994, “Expropriation and Inventions: Appropriable Rents in the Absence of Property Rights” American Economic Review, March, 84:1, 190-209. Anton, J., and D. Yao, 1995, “Start-ups, Spin-offs, and Internal Projects” Journal of Law, Economics and Organization, 11:2, 362-378. Athey, S. and S. Stern, 1996, “An Empirical Framework for Testing Theories About Complementarities” mimeo, MIT. Barry, C., 1994, “New Directions in Research on Venture Capital Finance” Financial Management, 23:3, 3-15. Berglöf, E., 1994, “A Control Theory of Venture Capital Finance” Journal of Law, Economics and Organization, 10:2, 247-267. Bharat, A. and A. Galetovic, 1998, “Weak Property Rights and Hold-up in R&D” unpublished working paper, Harvard Business School Bleicher, K. and H. Paul, 1987, “The External Corporate Venture Capital Fund -- A Valuable Vehicle for Growth” Long Range Planning, 20:6, 64-70. Block, Z., and I. MacMillan, 1993, Corporate Venturing, Harvard Business School Press, Boston, Massachusetts. Brockhaus, 1982, “The psychology of the entrepreneur” chapter 3 in Encyclopedia of Entrepreneurship, eds. C. Kent, D. Sexton and K. Vesper, Prentice Hall, Englewood, New Jersey. Burgelman, R., 1984, “Designs for Corporate Entrepreneurship In Established Firms” California Management Review, 26:3, 154-166. Burgelman, R., and L. Sayles, 1986, Inside Corporate Innovation: Strategy, Structure and Managerial Skills The Free Press, New York. Bygrave W., and J. Timmons, 1992, “Venture Capital at the Crossroads” Harvard Business School Press, Boston, Massachusetts. 29 Darlin, D., 1996, “Intel’s palace” Forbes, September, 42-43 Deger, R., 1996, “Valley of the Angels” Venture Capital Journal, March, 40-41. Diamond, D., 1991, “Monitoring and Reputation: The Choice between Bank Loans and Directly Placed Debt” Journal of Political Economy, 99:4, 689-721. Fast, N., 1981, “Pitfalls of Corporate Venturing” Research Management Magazine, 2124. Fenn, G., N. Liang and S. Prowse, 1995, “The Economics of Private Equity Markets” Staff Study 168, Board of Governors of the Federal Reserve System. Gans, J. and S. Stern, 1998, “Incumbency and R&D Incentives: Licensing in the Gale of Creative Destruction” unpublished working paper, MIT Sloan School Gilbert, R., and D. Newberry, 1982, “Preemptive patenting and the persistence of monopoly” American Economic Review, 72, 514-526. Gleba, D., 1994, “Corporate Venturing” Upside, September Issue, p.65 Gompers, P., 1995, “Optimal Investment, Monitoring, and the Staging of Venture Capital” Journal of Finance, 50:5, 1461-1489. Gorman, M., and W. Sahlman, 1986, “What do venture capitalists do?” Journal of Business Venturing, 4, 231-248. Hardymon G., M. DeNino and M. Salter, 1983, “When corporate venture capital doesn’t work” Harvard Business Review, May-June, 114-122. Hellmann, T., 1994, “Financial Structure and Control in Venture Capital” chapter 2, Ph.D. Dissertation, Stanford University. Hellmann, T., 1997, “Venture Capital: A Challenge for Commercial Banks” Journal of Private Equity, Fall, 49-55 Hellmann, T., 1998, “The Allocation of Control Rights in Venture Capital Contracts” Rand Journal of Economics, 29:1, Spring, 57-76 Hellmann, T., S. Milius and G. Risk, 1995, “Apple Computer - Strategic Investment Group” Stanford Graduate School of Business Case study, S-SM-21. Hellmann, T., and M. Puri, 1998, “The Interaction Between Product Market and Financing Strategy The Role of Venture Capital” unpublished working paper, Stanford University 30 Henderson, R. and K. Clark, 1990, “Architectural Innovation: The Reconfiguration of Existing Product Technologies and the Failure of Established Firms” Administrative Science Quarterly, 35, 9-30. Holmström, B., 1989, “Agency Costs and Innovation” Journal of Economic Behavior and Organization, 12, 305-327. Holmström, B. and P Milgrom, 1987, “Aggregation and Linearity in the Provision of Intertemporal Incentives” Econometrica, 55:2, March, 303-328 . Huang, H., and C. Xu, 1997, “Financing Mechanism and R&D Investment” unpublished working paper, London School of Economics Kanter, R., J. North, A. Bernstein and A. Williamson, 1990, “Engines of Progress: Designing and Running Entrepreneurial Vehicles in Established Companies” Journal of Business Venturing, 5, 415-430. Kaplan, J., 1995, Start-Up: A Silicon Valley Adventure, Houghton Mifflin Company, Boston. Katz, M. and C. Shapiro, 1994, “Systems Competition and Network Effects” Journal of Economic Perspectives, 8:2, Spring, 93-115. Kaufman, S., 1996, “Silicon Valley’s hot product: capital” San Jose Mercury News, Monday September 30th, Section E. Kortum, S. and J. Lerner, 1998, “Does Venture Capital Spur Innovation” unpublished working paper, Harvard Business School Kotkin, J., 1989, “Natural Partner: A new source of start-up financing” in Inc. Magazine, June, 67-80. Lerner, J., 1994, “Venture Capitalists and the decision to go public” Journal of Financial Economics, 35, 293-316. Lerner, J., 1995a, “Venture Capitalists and the Oversight of Private Firms,” Journal of Finance, 50:1, March, 301-318. Lerner, J., 1995b, “Xerox Technology Ventures: March 1995” Harvard Business School Case Study 9-295-127. Liebeskind, J., 1997, “Ownership, Governance and Incentives in Human-Capital Intensive Firms: Evidence from the biotechnology Industry” unpublished working paper, University of Southern California 31 McQuade, K., and B. Gomes-Casseres, “Mips Computer Systems” Harvard Business School, Case Study 9-792-055. Rajan, R., 1992, “Insiders and Outsiders: The Choice between Informed and Arm’sLength Debt” Journal of Finance, 47:4, September, 1367-1400. Reinganum, J., 1989, “The Timing of Innovation: Research, Development, and Diffusion” chapter 14 in Handbook of Industrial Organization, vol. 1, Schmalensee, R., and R. Willig, Elsevier Publishers. Rind, K., 1981, "The Role of Venture Capital in Corporate Development" Strategic Management Journal, 2, 169-180. Rotemberg, J., and G. Saloner, 1994, “Benefits of Narrow Business Strategies” American Economic Review, 84:5, December, 1330 - 1349. Sahlman, W., 1990, “The structure and governance of venture capital organizations” Journal of Financial Economics, 27, 473-521. Siegel, R, E. Siegel and I. MacMillan, 1988, “Corporate Venture Capitalists: Autonomy, Obstacles, and Performance” Journal of Business Venturing 3, 233-247. Sorrentino, M. and M. Williams, 1995, “Relatedness and Corporate Venturing: Does It Really Matter?” Journal of Business Venturing 10, 59-73. Sykes, H., 1986, “The anatomy of a corporate venturing program: factors influencing success” Journal of Business Venturing, 1, 275-293. Sykes, H., 1990, “Corporate Venture Capital: Strategies for Success” Journal of Business Venturing 5, 37-47. Sykes, H., and Z. Block, 1989, “Corporate Venturing Obstacles: Sources and Solutions” Journal of Business Venturing 4, 159-167, Elsevier, New York. Teece, D., 1992, “Strategies for capturing the financial benefits from technological innovation” chapter 7 in Technology and the Wealth of Nations Rosenberg, N., R. Landau, and D. Mowery, Stanford University Press, 1992. The Economist, “Biotechnology mergers: Unseemly coupling” May 13th 1995 issue, 6669. Tushman, M. and P. Anderson, 1986, “Technological Discontinuities and Organizational Environments” Administrative Science Quarterly, 31, 439-465. 32 Wetzel, W., 1987, “The Informal Venture Capital Market: Aspects of Scale and Market Efficiency” Journal of Business Venturing, 2, 299-313. Winters, T., and D. Murfin, 1988, “Venture capital investing for corporate development objectives” Journal of Business Venturing, 3, 207-222. Yost, M. and K. Devlin, 1993, “The State of Corporate Venturing” Venture Capital Journal, June, 37-40. Zahra, S., 1996, “Technology Strategy and New Venture Performance: A Study of Corporate-Sponsored and Independent Biotechnology Ventures” Journal of Business Venturing 11, 289-321. 33 Appendix Discussion of the statistical evidence on corporate venture capital There are very few statistics about the venture capital activities of corporations, and the existing statistics have to be interpreted with extreme caution. The National Venture Capital Association (NVCA) suggest that ‘corporate industrial’ venture capitalist accounted for $0.5 billion dollars out of $45.6 billion in committed capital in 1997.45 This number considerably underestimates the amount of capital made available by corporations in a number of ways. First, the methodology used by the NVCA excludes so-called ‘evergreen funds, i.e., funds with an unlimited life time horizon. An older methodology included the evergreen funds (as well as some other minor changes). For corporate venture capital the average amount of funds reported for the years 1984-1993 is on average 81% lower with the new methodology, relative to the old methodology. The equivalent number for all other venture capital funds is 22%. The change from the old to the new methodology thus disproportionately affected the measurement of corporate industrial venture capital funds. Applying these ratios to the 1997 data suggests that corporate venture capital would constitute approximately $2.6 billion out of a total of $62.1 billion of committed capital (4.25%) under the old methodology, instead of the $0.5 billion out of a total of $45.6 billion (1.1%) actually reported under the new methodology. This is obviously an ad-hoc calculation, since it assumes that the ratio continues to apply, i.e., that evergreen funds grew at the same rate as the overall venture capital market. Second, it only measures the amount of capital made available through a formally organized venture capital fund. Apart from the organized segment, there also exists a large amount of ad hoc venture investing both by corporations and by non-corporate investors, and no aggregate statistics exist for this informal sector. Using sample evidence, a study by Winters and Murfin (1988) estimated that, relative to the number of corporations with an explicit venture capital fund, there were two to three times as many corporations involved in some sort of ad hoc venture capital investing. Obviously, one would also like to compare this to the non-corporate informal segment. Estimates of the amount of funding provided by independent private individuals, who are sometimes called angels, range from twice (Wetzel, 1987) to seven times (Deger, 1996) the amount of funding provided by the organized venture capital market. Again, it is also 45 The NVCA breaks out commitments by four categories: independent, investment banks, corporate financial and corporate industrial. Corporate financial includes mainly commercial banks and insurance companies, which I consider different from corporate (industrial) venture capital. Note also that these figures does not include the money that corporate pension funds invest in venture capital. Pension funds, however, should be thought of as purely financial investors, given that ERISA laws prevent them from making investments for any strategic purpose. Finally, note that the amount of capital committed is a stock, not a flow measure. In any year a venture capital funds is unlikely to invest all of its committed capital. NVCA statistics indicate that the overall ratio of disbursement to capital committed is around 10%. 34 questionable that these ratios continue to apply as the formal venture capital market continues to grow. One of the sectors that is commonly thought to attract a particularly large amount of corporate venture capital is biotechnology. But even there the role of corporations seems to be limited. Liebeskind (1997) examines the IPO filings of some 81 California biotechnology companies over the period 1974-1996. She finds that on average 10.82% of the pre-IPO equity is held by corporations, compared to 33.72% held by independent venture capital firms. The statistics of the European Venture Capital Association do not break down capital commitments by the type of venture capital firm. They do, however, provide a breakdown of the sources of venture capital by type of investor. In 1996 corporate investors invested 325 million ECU (≈$413 million) in venture capital out of a total of 7969 million ECU (≈$10.1 billion), thus constituting about 4.1% of the total funding of European venture capital. Since corporate venture capital funds typically do not take in money from other sources, but some of the money from corporations may be given to independent venture capital firm, the amount of corporate venture capital is likely to be less. It should also be noted that the European definition of venture capital is broader than that of the US. Preparation for result 1: For this appendix it is useful to define ‘joint utilities’ as the sum of the two contracting parties’ utilities and ‘total utilities’ as the sum of all three parties. Consider any action ‘a’ and any equity split α then JUC(a) = UE(a,αc) + UC(a,αc) = q(a) (π(a) + θ(a) + β(a))- L - c(a) JUV(a) = UE(a,αv) + UV(a,αv) = q(a) (π(a) + β(a))- L - c(a) TUC(a) = UE(a,αc) + UC(a,αc) + UV(a,αc) = q(a) (π(a) + θ(a) + β(a)) - L - c(a) TUV(a) = UE(a,αv) + UC(a,αv) + UV(a,αv) = q(a) (π(a) + θ(a) + β(a))- L - c(a) Result 1: If ‘a’ is contractible the CI always finances the EN. Proof of result 1: With competition the VC offers a contract such that av maximizes JUV and αv (which is transferable utility when ‘a’ is verifiable) yields UV(av,αv) = 0. In case of indifference I assume that the EN chooses the CI. Suppose the CI offers a contract where ac maximizes JUC and αc yields UE(ac,αc) = UE(av,αv), then the CI will get to finance the EN. To see that the CI would want to make this offer, note that JUC(ac) ≥ JUC(av) Û q(ac) (π(ac) + θ(ac) + β(ac))- L(ac) - c(ac) ≥ q(av) (π(av) + θ(av) + β(av))- L(av) - c(av). Using UE(ac,αc) = UE(av,αv) Û q(ac) ((1-αc)π(ac) + β(ac)) = q(av) ((1-αv)π(av) + β(av)) and UV(av,αv) = 0 Û 35 q(av) (αvπ(av) + θ(av))- L(av) - c(av) = 0 in this equation yields q(ac) (αcπ(ac) + θ(ac))L(ac) - c(ac) ≥ q(av) θ(av) Û UC(a,αc) ≥ UC(a,αv), which shows that the CI would want to make the above offer. Preparation for result 2: In this model the EN would not necessarily want to take too much of the equity in order to preserve the investor’s incentives to support the new venture. The linear quadratic specification ensures that UE is concave in αv and αc. Differentiating UE(αc) = (q0 + K(αcπ+θ))((1-αc)π+β) w.r.t. αc and setting it equal to zero yields q0 1 αc = ( π + β − θ − ) . Similarly αv is given by the same equation, but with θ = 0. 2π K The amount of equity retained by the VC is given by UV(αv) = 0 as long as αv ≥ αv. From UV(αv,L) = 0, αv is a function of L. There exists Lv, so that αv ≥ αv Û L ≥ Lv. The amount of equity retained by the CI is given by αc satisfying UE(αc) = UE(αv), as long as αc ≥ αc. There exists Lc, so that αc ≥ αc Û L ≥ Lc. Note that αc is decreasing in θ, so that αc < αv if and only if θ > 0. It follows that Lc < Lv if and only if θ > 0. The reverse is true for θ < 0. For simplicity I prefer to avoid corner solutions, so I impose some mild conditions on 1 − q0 K.46 The largest possible value of q is q = qFB, so that qFB ≤ 1 Û K ≤ . For θ > π + β +θ 0 all q’s are always positive. For θ < 0 (but π+θ+β ≥ 0) the smallest possible value of q is given by q0+K(αcπ+θ) = q0+K(π+θ+β) which is positive. 47 It is also more elegant to think of q1 as positive, in which case we may want to impose K(αcπ+θ) = K(π+θ+β) ≥ 0 q0 Û K≥ . π + β +θ Throughout the paper I assume that the new venture is sufficiently profitable to be financed by the VC. In this part of the model this means L ≤ L v ≡ (q0+q1(v))π q1(v)2/(2K) where q1(v) evaluated at αV = 1. If L sufficiently large αc may also reach the constraint αc = 1. If this occurs we define L c so that we have UE(αc=1) = UE(αv) for L = L c and some αv ≤ 1. Define ∆ ≡ TUC - TUV. ∆ measures the efficiency of the CI relative to the VC. The analysis can, however, be extended straightforwardly to allow for corner solutions. The case π+θ+β < 0 can be safely ignored since in this case the parties would have an incentive to renegotiate and not operate the new venture. 47 46 36 Lemma 1: Suppose Lc ≤ L ≤ L c and Lv ≤ L ≤ L v , then (i) qc and qv are increasing in L (ii) qc is increasing in θ (iii) If θ > (<) 0 then qc > (<) qv (iv) If θ > (<) 0 then αc > (<) αv (v) the EN accepts the CI (VC) as the investor if and only if ∆ ≥ 0 (<0). Proof of Lemma 1: (i) αv is determined byU V (α v ) = q 0α v π + 0.5K (α v π ) 2 − L = 0 , so that αv is increasing in L. αc is given from UE(αc) = UE(αv) and since UE(αc) is decreasing in αc and UE(αv) is decreasing in αv, αc is increasing in L. From the first order equations we immediately see that qc and qv are increasing in αc and αv respectively and thus increasing in L. (ii) This is immediate from the first order condition. (iii) and (iv) Consider θ > 0 and consider any αc and αv satisfying UE(αc) = UE(αv): if αc = αv were true then UE(αc) > UE(αv) since the CI provides more effort, but since UE(αc) is decreasing in αc it follows that αc > αv. Applying this condition and θ > 0 in the firstorder conditions implies qc > qv. The reverse is true for θ < 0. (v) If Lv ≤ L ≤ L v the VC makes zero profits. αv needs to satisfy UV = qv (αvπ)- Lv = 0. In order to have his bid accepted, the CI only needs to match the utility that the EN received from a VC. If Lc ≤ L ≤ L c , αc needs to satisfy UE(αc) = qc ((1-αc)π + β) = qv ((1-αv)π + β) = UE(αv). The CI finds it in his interest to make such an offer whenever48 UC(αc) = qc (αcπ + θc)- Lc ≥ qv θv. Using the first two equalities in the third inequality yields qc (αcπ + θc)- Lc + qc ((1-αc)π + β) ≥ qv θv + qv (αvπ)- Lv + qv ((1-αv)π + β) Û TUC(αc) ≥ TUC(αv) Û ∆ ≥ 0. Result 2: (i) If θ ≥ 0 then the CI finances the EN (ii) If θ < 0 and β+θ > 0 then the VC finances the EN $ (iii) If θ < 0 and β+θ ≤ 0 then ∃ L ≥ Lc so that CI (VC) finances the EN whenever L > $ (<) L . We distinguish between θc and θv in this proof to show that this proof is also valid for result 3. 48 37 Proof of result 2: Suppose first that Lc ≤ L ≤ L c and Lv ≤ L ≤ L v , then (i) We have TUC = (q0+q1(c))(π+θ+β)-L- (q1(c))2/2K and TUV = (q0+q1(v))(π+θ+β)-L(q1(v))2/2K. If θ > 0 then for any L we have q1(v) < q1(c) ≤ qFB. Since TUC and TUV are both concave in q1 it follows that TUC > TUV, i.e. ∆ > 0. (ii) If θ < 0 and β + θ > 0 then for any L we have q1(c) < q1(v) < qFB. Since TUC and TUV are both concave in q1 it follows that TUC < TUV, i.e. ∆ < 0. (iii) If θ < 0 and β + θ < 0 then for any L we have q1(c) < qFB but q1(v) >, = or < qFB. We ∂α v ∂∆ ∂∆ ∂α c ∂α v ∂∆ ∂α v take the derivative of ∆ w.r.t. L: = + . Since > 0 we can ∂L ∂α c ∂α v ∂L ∂α v ∂L ∂L focus on the three other terms. Define Xv = (q0/K) + αvπ + θ, Xc = (q0/K) + αcπ + θ, Yv = θ + β + (1-αv)π, Yv = β + (1-αc)π then ∂∆/∂αc = KπXc, ∂∆/∂αv = - KπXv and ∂α c ( X v − Yv ) ∂∆ ∂α v ( Yc X v − Yv X c ) = . It follows that = > 0 since for θ < 0, Yc > ∂L ∂L X c − Yc ∂α v X c − Yc $ $ Yv, Xc < Xv and Xc - Yc > 0. The critical value L is simply found from ∆( L ) = 0 or from the appropriate corner solution. Consider next the case of L > L c . If θ > 0 and L rises above L c then UE(αc=1) > UE(αv) since αv increases while αc remains at 1. The CI continues to finance the EN. If θ < 0 then at L = L v we have αc < αv =1 so that αc never hits the constraint αc = 1 in the relevant range of L. If θ > 0 and L < Lv. Below Lv the VC cannot drop αv, making it even more unattractive to the EN. The CI thus finances the EN. Finally, if θ < 0 and L < Lc, then below Lc the CI cannot lower αc but the VC continues to lower αv. The CI cannot match the VC, i.e., UE(αc) < UE(αv). The VC thus finances the EN. Preparation for extension of result 2: Suppose now that q = q0 + q1 + q2 where q0 is a fixed component, q1 is the variable component that can be affected by the investor and q2 the component that can be affected by the EN. I use a quadratic cost of effort q12/(2K1) for the investor and q22/(2K2) for the entrepreneur, where K1 and K2 reflect the respective ease of providing effort. For a given αc and αv, the levels of effort provided by the EN, VC and CI are given by q1c = K1(αcπ+θ), q1v = K1αvπ, q2c = K2((1-αc)π+β) and q2v = K2((1-αv)π+β). The optimal choices of αc and αv are limited by the fact that the EN may voluntarily relinquish shares to the investor. In particular we have ∂UE(αc)/∂αc ≤ 0 and ∂UE(αv)/∂αv ≤ 0. The minimal amount of equity, denoted by αc, is thus given by αc ε [0,1] satisfying ∂UE(αc)/∂αc = 0 Û ( K − K 2 )( π + β) − q 0 − K1θ (or αc = 1 if the right-hand side term exceeds 1). αc = 1 (2 K1 − K 2 ) π 38 Similarly αv, where we set θ = 0 to obtain the formula for αv. With this we can define Lc and Lv as before and we can see that Lc< (>) Lv whenever θ > (<) 0. In this model the investor may also be concerned about the incentives of the EN. There may exist a maximal level of equity above which the investor does not want to take α, i.e., αc and αv must satisfy ∂UC/∂αc ≥ 0 and ∂UV/∂αv ≥ 0. The maximal value of α, denoted by α c is either given by any αc ε [0,1] satisfying ∂UC/∂αc = 0 Û q 0 + K 2 ( π + β) + ( K 2 − K 1 ) θ αc = , or if that doesn’t exist by α c = 1. Similarly (2 K 2 − K1 ) π for α v ,except that we set θ = 0. Again, we define L c and L v directly from α c and α v . Extension of result 2: (i) (ii) (iii) If θ ≥ 0 then the CI finances the EN $ $ If θ < 0 then ∃ L ≥ Lc so that CI (VC) finances the EN whenever L > (<) L . If θ < 0 and β sufficiently large then the VC always finances the EN. Proof of the extension of result 2: We begin by assuming Lc ≤ L ≤ L c and Lv ≤ L ≤ L v . Suppose that θ > 0 and that the VC offers some αv. If the CI were to offer αc’ = αv then the EN would provide the same effort as with VC financing, and the CI would provide a higher effort than the VC. In this case UE(αc’) > UE(αv) and UC(αc’) > UC(αv) where the second inequality follows from the fact that the CI could choose the same effort level as the VC, but actually chooses an effort level that gives him more utility. The contract αc’ = αv is not the equilibrium, since the CI wants to offer another contract αc < αc’ such that UE(αc) = UE(αv). But if the CI chooses αc over αc’ it is better off than before, i.e., UC(αc’) > UC(αc). It follows that for θ > 0 the CI always finances the EN. Consider now the case where θ < 0. We then have B2 ∂∆ ∂∆ ∂α c ∂α v ∂∆ ∂α v ∂α v = + = (A1 − A 2 ) where ∂L ∂α c ∂α v ∂L ∂α v ∂L ∂L B1 ∂TUC A1 = = ( K1 + K 2 )((1 − α c ) π + β) − K 2 ( π + β + θ) ∂α c ∂TUC A2 = = ( K1 + K 2 )((1 − α v ) π + β) − K 2 ( π + β + θ) + K1θ ∂α v ∂UC B1 = = π( K 1 (α c π + θ) + K 2 ((1 − α c ) π + β) + q 0 − K 2 (α c π + θ)) ∂α c ∂UV B2 = = π ( K 1α v π + K 2 ((1 − α v ) π + β) + q 0 − K 2 α v π) ∂α v 39 From this we get A1B2 - A2B1 = ((K1)2 + (K2)2 - K1K2)(π+β+θ)(αv-αc)π + q0(K1+K2)(αv-αc)π + q0K1(-θ) + K1(-θ)(K1(αcπ+θ)+K2((1-αc) π+β)-K2(αcπ+θ)). The first two terms are positive since αv > αc, which follows from the same reasoning as in Lemma 1. The third term is positive for θ < 0. The fourth term is positive, which follows from the fact that any αc must satisfy ∂UC/∂αc ≥ 0. This establishes that ∆ is $ increasing in L, implying the existence of a critical value L below (above) which the VC 49 (CI) will finance the EN. If L < Lc and or L < Lv then the same reasoning as for the proof of result 2 applies. For L c < L v and θ > 0 then the CI still invests in the EN: even if the CI were to offer αc = αv, which is suboptimal, it is still better off than not financing the EN. Finally, the combination of L c < L v and θ < 0 is impossible.50 This establishes that the CI finances the EN always if θ > 0 and possibly if θ < 0 and L sufficiently large. To see that β must also be sufficiently small simply note that for sufficiently large β, αc = αc = αv = αv = 1: in this case q1c = K1 (π+θ), q1v = K1 π, q2c = K2 β, q2v = K2 β, so that with θ < 0 the VC always finances the EN for all L. In the main text we also noted that with a double moral hazard problem, we do not need to introduce the adverse selection effect to justify that the investor doesn’t simply acquire the EN. If α c < 1 and α v <1 then the CI and the VC never find it in their interest to buy all the equity in the new venture. This implies that we do not need to rely on the adverse selection effect to explain why the investors don’t acquire the new venture at date 0. K 1 ( π + β) − K 2 θ ∂U C ∂U E Moreover, if then the CI would not make + < 0 Û αcπ > K1 + K 2 ∂α c ∂α c It is straightforward to find examples that establish that both ∆ > 0 and ∆ < 0 are both possible. 50 To see this note that for any θ < 0 and any L we have αc(αv) < αv. If α c = 1 then L c < L v is clearly not possible. A necessary condition for α c < 1 is K1 < 2K2. Suppose K1 ≤ K2, then α c ≥ α v so that L c < L v again not possible. If K2 < K1 ≤ 2K2 then we use ( K1 − K 2 ) q0 ∂U E (α c ) ≤ 0 Û αcπ + θ ≥ ((1 − α c ) π + β) − in K1 K1 ∂α c ∂U C = q 0 + K 2 (π + β ) + ( K1 − 2 K 2 )α c π + ( K1 − K 2 )θ to get ∂α c 49 ∂U C K 2 ( K − K 2 ) 2 + K1 K 2 ≥ q0 + 1 ((1 − α c )π + β ) > 0 from which we again conclude ∂α c K1 K1 that L c < L v is impossible. 40 any ex-ante transfer payments to the EN, i.e., the EN would receive only equity. Similarly for the VC where the above condition holds for θ = 0. Preparation for result 3: If the choice of θ is not contractible the VC maximizes q(θ)αvπ so that θv = θ0. The CI, on the other hand, maximizes q(θ)(αcπ+θ) so that the first-order condition for θc is given by q’(θc) (αcπ+θc) + q(θc) = 0. I assume that q(θ) is ‘well-behaved’ so that the secondorder condition is always satisfied, i.e., ∂2UC/∂αc2 = q’’(αcπ+θ)+2q’ < 0. We define Lc as before from αc with ∂UE(αc)/∂αc = 0 or else αc = 0. Again we assume that the EN can always be financed by the VC, i.e., L ≤ L v = q(θ0)π. Result 3: (i) The optimal choices of θ are given by θc > θFB > θv = θ0 $ $ (ii) There exists L ∈ [Lc, L v ] so that the CI (VC) finances the EN if L > (<) L $ $ (iii) For a given L there exists β ≥ 0 so that the CI (VC) finances the EN if β < (>) β . Proof of result 3: (i) This follows immediately from the first order conditions. (ii) Suppose first that L ∈ [Lc, L v ] then since the proof of Lemma 1(v) applies to this model too, we can use ∆ to determine who finances the EN. We have ( q c ’( π + β + θ c ) + q c π ) q c ’ ∂∆ ∂[q c ( π + β + θ c )] ∂θ ∂α c ∂α v = = ( q c ’( π + β + θ c ) + q c π ) = >0 − ( q c ’’(α c π + θ c ) + 2q c ’π) q c ∂L ∂L ∂α c ∂α v ∂L since qc’≡dq(θc)/dθ < 0 and qc’(π+β+θc) + qcπ < qc’(αcπ+θc) +qcπ = 0. The critical value $ $ L is found from ∆( L ) = 0 or one of the two corner solutions. If L < Lc then the CI cannot match the VC’s offers, i.e., UE(αv) > UE(αc). The VC finances the EN in this case, so that $ the lower limit of L is given by Lc (iii) We have (q c ’( π + β + θ c )) q c ’(q c − q v ) ∂∆ ∂θ ∂α c = (q c − q v ) + (q c ’( π + β + θ c )) = (q c − q v ) + <0 − (q c ’’(α c π + θ c ) + 2q c ’π) qc ∂β ∂α c ∂β since qc < qv and qc’ < 0. Result 4: We prove result 4 by solving the game for θ > 0 and Θ < 0. For β large the EN always yields all the equity to the investors, i.e., αc = αv = 1. Applying the usual first-order conditions we have 41 Kπ N , p N ( S 2 ) = Kπ N , 1 − c N cI K (π N + θ − cRπ I ) p N ( S3 ) = , p N ( S 4 ) = K (π N + θ ) . 1 − c N cI From this we get Kπ N c p qN ( S1 ) = − N I , q N ( S 2 ) = Kπ N , 2 (1 − cN cI ) 1 − cN cI K (π N + θ − cIπ I ) c p qN ( S3 ) = − N I , q N ( S 4 ) = K (π N + θ ) , 2 (1 − cN cI ) 1 − c N cI pI c Kπ qI ( S1 ) = − I N 2, qI ( S 2 ) = 0 , 1 − cN cI (1 − cN cI ) pI c K (π N + θ − cIπ I ) qI ( S3 ) = − I , qI ( S 4 ) = 0 . 1 − c N cI (1 − cN cI ) 2 Note that for all S the parameter values also must satisfy 0 ≤ qI(S) ≤ 1 and 0 ≤ qN(S) ≤ 1. The optimized utility of the CI is given by UC(S1) = uc(S1) - LI, UC(S2) = uc(S2), UC(S3) = uc(S3) - LI - LN, UC(S4) = uc(S4) - LN, where Kπ N (θ − cIπ I ) pI (π I − cNθ ) uc ( S1 ) = + , uc( S 2 ) = Kπ N θ , (1 − cN cI )2 1 − c N cI pN ( S1 ) = K (π N + θ − cIπ I ) 2 pI (π I − cNθ − cNπ N ) uc( S3 ) = + , 2(1 − cN cI ) 2 1 − c N cI uc( S 4 ) = K (π N + θ ) 2 2 Consider the game with commitment as in Figure 3a. The EN prefers whatever yields the highest probability of success. If θ> 0 and Θ< 0 we have qN(S1) > qN(S3) and qN(S4) > ˆ qN(S2). The CI prefers S4 over S1 whenever UC(S4) > UC(S1) Û LI < LCO ≡ uc(S1) - uc(S4) I + LN. Figure 4 also shows the outcomes of the commitment game. Consider next the game without commitment as in Figure 3b. If the EN has accepted the ˆ VC, then the CI will invest in IV whenever LI < LVC ≡ uc(S1) - uc(S2). Similarly, if the I ˆ EN has accepted the CI, the CI will still invest in IV if LI < LCI ≡ uc(S3) - uc(S4). I ˆ ˆ ˆ ˆ We have LCO < LVC and LCO < LCI . To see this note that CI always prefers to finance the I I I I EN than have the VC finance it on the same terms (this follows immediately from the zero profit condition of the VC). We have UC(S3) > UC(S1) and UC(S4) > UC(S2). It ˆ ˆ follows that UC(S3) - UC(S1) = uc(S3) - uc(S1) - LN = LCI - LCO > 0 and UC(S4) - UC(S2) = I I ˆVC - LCO > 0. ˆ uc(S4) - uc(S2) - LN = L I I ˆ ˆ ˆ Consider first the case where LVC > LCI as in Figure 4a. If LI < LCI then the EN chooses I I I ˆ between S1 and S3. With Θ < 0 we have qN(S1) > qN(S2). If LI > LVC then the EN chooses I ˆ between S2 and S4. With θ > 0 we have qN(S4) > qN(S3). For the intermediate case LCI < I 42 ˆ LI < LVC the EN is choosing between S2 and S3. I claim that the EN always chooses S2 in I this case. To see this suppose on the contrary that qN(S3) ≥ qN(S2) Û K (π N + θ − cIπ I ) c p c p c c Kπ − Kπ N ≥ N I . Using q2(S1) ≥ 0 Û N I ≥ N I N2 in the 2 (1 − cN cI ) 1 − c N cI 1 − cN cI (1 − cN cI ) previous inequality yields after some transformation θ − cIπ I ≥ −π N cN cI (1 − cN cI ) . Using θ - cI πI < 0 while squaring both sides yields (θ − cIπ I ) 2 ≤ (π N cN cI (1 − cN cI ))2 . ˆ ˆ Straightforward calculations also reveal that LCI < LVC Û I I (θ − cIπ I ) ≥ ((1 − cN cI )θ ) + (π N cN cI ) . Combining these two inequalities yields 2 2 2 (θ − cIπ I ) 2 ≤ (π N cN cI ) 2 (1 − cN cI ) 2 < (π N cN cI )2 + (1 − cN cI ) 2θ 2 ≤ (θ − cIπ I )2 which is a contradiction. It follows that qN(S3) < qN(S2). ˆ ˆ ˆ Consider next the case where LVC < LCI as in Figure 4b. For LI < LVC the EN prefers S1 I I I CI ˆ over S3 as before, and for LI > L then the EN prefers S4 over S2 as before. For the I ˆ ˆ intermediate case LVC < LI < LCI the EN is now choosing between S1 and S4. As shown I I below there exist parameter values such that either will be chosen. We can, however, derive a sufficient condition for S4 to be chosen. Consider Kπ N c p qN ( S 4 ) − qN ( S1 ) = K (π N + θ ) − + N I . Using qI(S1) ≥ 0 I obtain after 2 (1 − cN cI ) 1 − cN cI Kπ N transformation qN ( S 4 ) − qN ( S1 ) ≥ K (π N + θ ) − . A sufficient condition for S4 to (1 − cN cI ) c cπ occur is thus θ ≥ N I N ≡ θ * . (1 − cN cI ) ˆ ˆ ˆ ˆ So far I have shown that if LVC > LI > LCI then S2 results and if LVC < LI < LCI then either I I I I S4 or S1 occurs. To show that all of these cases can actually occur I construct three simple examples. I use the following assumptions: Assumptions Example 1 Example 2 Example 3 K 0.5 0.5 0.5 θ 0.1 1 0.005 πN 1.1 0.9 1 πI 11 11 0.15 cN 0.1 0.1 0.1 cI 0.1 0.1 0.1 pI 0.5 0.5 0.51 LN 0.01 0.01 0.0001 LI 5.036 4.865 0.00007 43 With these assumptions I obtain the following results:51 Results ˆ LCI I ˆVC L I Example 1 5.14 4.93 0.51 0.55 0.0005 0.6 -0.05 0.06 0.45 0.35 Example 2 4.72 5.01 0.41 0.45 0.36 0.95 0.59 0.45 0.75 0.89 Example 3 0.00004 0.00010 0.505 0.500 0.499 0.503 0.00253 0.00250 0.25238 0.25241 qN(S1) qN(S2) qN(S3) qN(S4) UC(S1) UC(S2) UC(S3) UC(S4) ˆ ˆ In all examples LI is in between LCI and LVC . In example 1 the CI prefers S2 over S1 and I I S3 over S4; the EN chooses S2 over S3. In example 2 the CI prefers S1 over S2 and S4 over S3; the EN chooses S4 over S1. In example 3 the CI prefers S1 over S2 and S4 over S3; the EN chooses S1 over S4. This shows that all three cases may actually occur. ˆ ˆ ˆ The S2 equilibrium is associated with LCI > LVC whereas the S4 is associated with LCI < I I I ˆVC . We say that the S4 equilibrium is more likely when LVCI ≡ LVC - LCI is increasing at ˆ ˆ ˆ L I I I I K K (π N + (θ − cIπ I ) ) pI cNπ N 2 ˆ ˆ + LVCI = 0. We have LVCI = (π N + θ 2 ) − , which is I I 2 2(1 − cN cI )2 1 − c N cI ˆ decreasing in K and increasing in θ at LVCI = 0. 2 2 I 51 All calculations were performed using Mathematica and are available from the author upon request. 44 q FIGURE 1 q FB(high β) qv qc q FB(low β) L Lv Lc ^ L - Lc - Lv 45 FIGURE 2 q(θ )(π + θ+ β ) q(θ )( π + θ) q( θ ) θ θv = θ0 θFB β>0 ) ( θFB (β=0 ) θc(high L) θc(low L) 46 FIGURE 3a : WITH COMMITMENT FIGURE 3b : WITHOUT COMMITMENT IV No IV CI decides on internal investment VC CI EN decides on investor VC CI VC CI EN decides on investor IV No IV IV No IV CI decides on internal investment S1 S3 S2 S4 S1 S2 S3 S4 47 Figure 4a: Low elasticity of support Cost of internal venture With Commitment Without Commitment Low IV VC IV VC Intermediate Low No IV CI IV VC Intermediate High No IV CI No IV (IV) CI (VC) High No IV CI No IV CI Figure 4b: High elasticity of support Cost of internal venture With Commitment Without Commitment Low IV VC IV VC Intermediate Low No IV CI IV VC Intermediate High No IV CI No IV VC High No IV CI No IV CI 48