Corporate Venturing, Allocation of Talent, and Competition for Star Managers

We provide new rationales for corporate venturing (CV), based on competition for talented managers. As returns to venturing increase, (cid:133)rms engage in CV for reasons other than capturing these returns. First, higher venturing returns increase managerial compensation, to which (cid:133)rms respond by increasing the power of incentives. Managers increase e⁄ort, prompting (cid:133)rms to reallocate them to new ventures, where the marginal product of e⁄ort is highest. Second, as returns to venturing become large, CV emerges as a way to recruit/retain managers who would otherwise choose alternative employment. We derive several testable empirical predictions about the determinants and structure of CV. ( JEL Codes : G24, G32, M13, M12)


Introduction
At its peak in 2000, corporate venturing (CV) represented $16.2 billion of investments in the U.S. 1 At the time, 16 cents out of every dollar invested in new ventures came from corporate venturing …rms, who participated in about 25% of all deals. Corporate venturing investments had risen dramatically with entrepreneurial activity during the internet boom, up from $1 billion in 1997, and evaporated just as quickly afterwards, down to $1.2 billion in the …rst 9 months of 2002 (Taylor, 2003). Despite their importance, the nature of corporate venturing and its pro-cyclical ‡uctuations with entrepreneurial activity are still largely unexplained. 2 Is corporate venturing simply a way for …rms to capture some of the high returns that they observe in entrepreneurial ventures -a "gravy train"rationale -or is it part of a broader corporate strategy? In this paper we propose two novel and complementary explanations for corporate venturing that are based on competition for talent. We present a model where a …rm and a venture capitalist (VC) compete for the recruitment of a star manager who has an idea and a unique skill to run a new venture. The star may choose the VC's o¤er, in which case the VC …nances her project, and she manages the new venture. 3 Alternatively, she may choose the …rm's o¤er, in which case her task depends on the …rm's organizational structure: if organized as a corporation, the …rm will assign the star the task of managing its main line of business; while if organized for corporate venturing it can …nance the star's project and let her manage the new venture. The main di¤erence between the new venture and the …rm's main business is that the star's marginal product of e¤ort is higher in the new project than in the more routine main business.
When returns to venturing are low, …nancing the star's project is not lucrative for the …rm, and competition from the VC is low for the same reason. The …rm prefers to organize as a corporation, recruiting the star to manage its main business. As returns to venturing increase, the relative attrac- 1 We de…ne corporate venturing as the …nancing and development of new business ventures by large established companies, either inside (intrapreneurship) or outside (corporate venture capital) the corporate structure. In the United States, over 200 corporations were listed in the 2002 Directory of Corporate Venturing as investing as active corporate venture capitalists. Corporations also invest in venture capital through specialized institutions such as venture capital funds. At the end of 2001, corporations were the second largest source of capital to venture capital funds, after endowments and foundations, with total commitments of about $35 billion (Goldman and Russell, 2002). Although some of these investments, often organized through partnerships, are sometimes included in the de…nition of corporate venturing, they are not part of our de…nition here. 2 As is well known, entrepreneurial activity went through a similar boom-and-bust cycles over the same time period: venture capital investment went from $14.6 billion in 1997 to $105 billion in 2000, down to $17.3 billion in the …rst 9 months of 2002 (Brander and Bettignies, 2006). This pro-cyclicality is not new. In the previous two venture capital "waves," in the late 1960s and early 1970s, and then again in the late 1970s and early 1980s, success in venture capital spurred corporate venturing investments which quickly shrank at the end of the booms, in 1973 and 1987 respectively (Gompers andLerner, 1998, Gompers, 2002). tiveness of corporate venturing increases as well for two reasons. First, investing in a new venture enables the …rm to capture some of the higher returns associated with venturing. This is the "gravy train" e¤ect mentioned above. Second, the VC's willingness to bid for the star also increases with returns to venturing, and the star's equilibrium compensation goes up. The …rm responds to higher compensation costs by o¤ering higher-powered incentives, which in turn elicit higher managerial e¤ort.
Foresighted …rms may thus choose corporate venturing as a way of allocating stars where the marginal product of their e¤ort is highest, i.e. in new ventures. We call this the "incentives" rationale for corporate venturing.
Beyond a certain threshold level of returns to venturing, recruiting the star would become too expensive for the …rm, if it were organized as a corporation. Faced with constant bene…ts but rising compensation costs, it could no longer match o¤ers made by the VC, who would "steal" away the star manager. We show that when returns to venturing are high enough for this problem to arise, corporate venturing emerges as a solution, enabling the …rm to hire/retain talented managers who would otherwise take another job. We call this the "recruitment/retention" rationale for corporate venturing.
The key di¤erence between the gravy train and incentives rationales on the one hand, and the recruitment explanation on the other, is that they take place at di¤erent levels of returns to venturing. Indeed, at low and moderate levels, the …rm anticipates that it can successfully recruit the star regardless of its organizational form, and therefore its organizational choice is more about how to best allocate the star across its activities. When the gravy train and incentives e¤ects become su¢ ciently large, the optimal allocation of talent switches from the main business (corporation) to a new venture (corporate venturing).
In contrast, at high levels of returns to venturing, the …rm anticipates that if it organizes as a corporation, it will not recruit the star. This adds both a cost and a bene…t to the relative attractiveness of corporate venturing. On the one hand, by organizing for corporate venturing, the …rm has to pay a large compensation cost in order to recruit the star, a cost which would be avoided if it organized as a corporation and did not recruit. In our stylized model this cost exactly o¤sets the gravy train and incentives e¤ects mentioned above, as competition for talent ensures that the star extracts all expected rents from the new venture. On the other hand, organizing for corporate venturing and allowing the star to develop her idea inside the …rm may be a way to prevent her from developing it somewhere else, e.g. with a VC.
Suppose that the new venture is correlated to the …rm's main business, e.g. through spillovers, which could be positive or negative. These spillovers will likely bene…t the main line of business less, or hurt it more, if the venture is developed by the VC than if it is developed under the umbrella of corporate venturing, because unlike the …rm, the VC will not internalize the impact of the new venture on the …rm's main line of business. Hence, at high levels of returns to venturing, although the gravy train and incentives e¤ects disappear, a bene…t from recruitment and retention of star employees emerges as the …rm attempts to internalize new venture spillovers on its main business.
Despite growing interest, only a few papers have explicitly addressed the question of corporate venturing theoretically. 4 Anton andYao (1994, 1995), as well as Anand, Galetovic and Stein (2004), examine this theme in the context of weak property rights. Hellmann (2002) argues that an entrepreneur should seek corporate venturing funding when the venture project is complementary to the parent corporation's main line of business. Amador and Landier (2003) investigate whether innovation is to be implemented inside incumbent …rms or outside by venture capitalists, in a model of entrepreneurial optimism. Gromb and Scharfstein (2003) analyze how the tradeo¤ between information and incentives a¤ects …rms' choice between intrapreneurship and direct venture capital …nancing. 5 Hellmann (2006) explores a …rm's decision to encourage or discourage exploration of new ideas by employees, and shows that it depends chie ‡y on the allocation of intellectual property rights and the relative value of innovation inside and outside the …rm.
The thrust of our contribution to this literature is to place competition for talent at the forefront of the analysis. 6 This enables us to highlight several new insights and empirical implications. First, we provide the two new explanations for corporate venturing and its pro-cyclical ‡uctuations that we mentioned: one based on managerial incentives and the optimal allocation of talent; the other based 4 For an excellent review of this literature, and of employee innovation in general, see Hellmann (2006). 5 Gromb and Scharfstein (2003) consider the "safety net" that intrapreneurship provides to entrepreneurs, who can get their corporate job back in case of failure. Intrapreneurship provides lower incentives than VC …nancing because of the safety net, but it comes with an informational advantage about the quality of the failed intrapreneur. In their paper, as well as in Landier's (2006), the market perception of the reasons for failure plays a central role in reaching the equilibrium level of entrepreneurship. Dix and Gandelman (2003) also look at a similar choice between intrapreneurship and corporate venture capital, but the tradeo¤ is based on somewhat di¤erent informational asymmetries. 6 Amador and Landier (2003) and Anand et al. (2004) also developed models in which a manager/entrepreneur with an innovative idea chooses between contracts o¤ered by an incumbent …rm and a VC. A key feature of Amador and Landier's model is that the entrepreneur is subject to an optimism bias about her idea. The VC's advantage in recruiting her lies in its superior ability to exploit her bias through contracts, while the incumbent …rm has a cost advantage due to potential synergies with existing assets. Interestingly, the authors …nd that the relative recruiting advantage of the VC increases with project value -a variable similar to the "returns to venturing" measure used in our model -and predict that high value projects will be …nanced outside the …rm. In contrast our model suggests that high value projects may still be developed inside the …rm, as it attempts to internalize the spillovers on its main line of business. In Anand et al., the focus is on studying how the strengths of property rights, and the centralization of operations, a¤ect a …rm's ability to recruit a talented manager. They do not analyze the e¤ects of changes in returns to venturing and competition for talent. on recruiting and retaining star managers as a spillover internalization strategy. Second, we highlight the importance of competition for talent as a determinant of …rms'organizational strategy. We show that -by generating the incentives and recruiting e¤ects -competition has an unambiguously positive impact on corporate venturing, inducing …rms to switch to corporate venturing "sooner" as returns to venturing increase. Third, we suggest that spillovers from the new venture to the …rm's main business may have a positive impact on corporate venturing. Interestingly, in contrast with Helmann (2002), corporate venturing may still be optimal even when spillovers are large and negative, as the …rm attempts to preempt the VC's development of the new venture, which would lead to a negative spillover e¤ect of an even greater magnitude. Fourth, we extend the model to adress issues related to "contractual incompleteness."This enables us to examine the impact of intellectual property (IP) rights on the prevalence of corporate venturing, as well as the e¤ects of returns to venturing on the optimal allocation of control rights between the star and the corporate parent. We suggest that weaker IP protection may be favorable to corporate venturing, and that increased venturing returns may induce corporate venturing …rms to allocate more control rights to the star, thereby switching from tighter controlled structures, such as internal ventures, to more autonomous structures, such as corporate venture capital. Finally, we underline the link between competition for talent, economic e¢ ciency, and corporate venturing: we point out that there may be less corporate venturing in equilibrium than is socially desirable, but that competition -by increasing the prevalence of corporate venturing -may improve economic e¢ ciency.
The paper is organized as follows. Section 2 describes the model. Section 3 analyzes competition for star managers when the …rm is organized as a corporation, while section 4 looks at these issues in the context of corporate venturing. In section 5, we characterize the optimal organizational choice and how it is a¤ected by returns to venturing, competition for talent, and spillover e¤ects. Section 6 extends the model to contractual incompleteness and examines the optimal structure of corporate venturing; while section 7 discusses other extensions to the model. Section 8 presents the key empirical implications of the model, and relates them to the empirical literature. Section 9 concludes. The characterization of the main optimization program is in the appendix at the end of the paper; all proofs are in the online supplement.

Basic Setup
A …rm intends to recruit a star manager/entrepreneur, and anticipates competition from other institutions in the "market for talent." 7 This star manager has an idea for a new venture that requires …nancing. The …rm must choose between two organizational structures F 2 fC; CV g, which di¤er primarily in the way they allocate talent.
If organized as a corporation (C), the …rm focuses on its core competency, and devotes all its attention and resources to its existing line of business. If recruited, the star would manage the main business, and her idea for a new venture is not pursued.
If involved in corporate venturing (CV ), the …rm does not focus exclusively on its main business, and is organized so as to be able to pursue new opportunities. This organizational change could take various forms, from setting up a committee in charge of evaluating employees' new venture ideas, to the funding of a full- ‡edged fund devoted to new venture investments, similar to standard venture capital funds. Under this organizational form, conditional on recruiting the manager, the …rm …nances her idea, and the star manages the new venture.
For simplicity, we assume that the main business yields a (dollar) return M with certainty if managed by the star, and M < M , otherwise. The new venture can only be valuable if it is run by the star manager, and it can turn out to be of two types. It may be a "base hit," in which case it yields a base dollar return min (net of initial capital outlay), or a "home run," in which case the return is + , with > 0. The star exerts a non-veri…able e¤ort e which strictly increases the probability of a home run. For simplicity we assume this probability to be e. Hence, the new venture yields an expected return + e (gross of e¤ort cost) when managed by the star, and zero otherwise.
The star is assumed to be risk-neutral and wealth-constrained; and her cost of e¤ort is c(e) = k 2 e 2 . A key di¤erence between the main line of business and the new venture is that the star's marginal product of e¤ort is larger in the latter than in the former. The basic idea is that while new ventures are highly risky and the star can have a large impact on the probability of success, the main line of business may include (relatively) more of a routine job where the star's e¤ort has a lower impact. One could also argue that the management of the new venture likely involves a product that is closer to 7 In this paper, we view managerial talent and entrepreneurial talent as requiring similar skills, which appears to be consistent with the fact that …rms lose a large number of successful managers during times of high entrepreneurship (see footnote 25 in section 8. This contrasts with, for instance, Lazear (2003), who argues that entrepreneurial talent requires a broad of skills while managerial talent requires perhaps more outstanding, but also more specialized skills. the beginning of the life-cycle than its main business counterpart, and that managerial e¤ort has a larger impact at the beginning of a product's life-cycle than at its end.
The main business and the new venture are linked in that the new venture yields a spillover, S 2 R, on the main business. The spillover could be positive and represent superior access to a new technology for the main line of business, for example, or may capture a reduced-form complementarity between the products developed in the new venture and those developed in the main business. Alternatively, S may be negative and capture some form of cannibalization of the main venture products by the new venture products. The spillovers enable us to take into account the idea, highlighted by Hellmann (2002), that the venturing …rms'objective is not restricted to maximizing the value of the stand-alone ventures. Here, the broader objective is the value created by the project as a stand-alone project plus the spillovers to the main line of business. In our setting, conditional on hiring the star, the …rm's gross (i.e. excluding the compensation cost) expected return is M when organized as a corporation, and + e + M + S when organized for corporate venturing.
The …rm may fail to recruit the star, who may turn to another institution, say a venture capitalist (VC), to …nance her new venture idea. 8 If the star chooses to develop the new venture with the VC, the …rm, regardless of its organizational form (C or CV ), has no access to the star's talent, but may still receive a spillover (S ) from the new venture to its main business. In other words, the spillover from the new venture to the …rm's main business is lower, be it positive or negative, if the new venture is developed by the venture capitalist than if it is developed under corporate venturing, and 0 represents this spillover di¤erential. This is simply a reduced-form way of capturing the idea that, unlike the …rm, the VC does not internalize the impact of spillovers on the …rm's main line of business.
Thus, if the star chooses to go with the VC, the …rm's expected payo¤ is M + S regardless of its organizational form. The VC expects a gross payo¤ + e from the project if he can recruit the star and …nance the venture; and zero otherwise.
Finally, the following two regularity conditions on the parameters will simplify our analysis (our results still obtain without these conditions, unless otherwise speci…ed):

Competition for Talent and Incentive Contracts
Competition a¤ects the o¤ers made by the …rm and the VC to the star manager. If the …rm is organized as a corporation and focuses on its main business, the value of recruiting the star does not depend on her non-contractible e¤ort. Hence, the corporation's contractual o¤er is a constant wage W C , and the star receives a net payo¤ U C = W C .
In contrast, if the …rm engages in corporate venturing, its expected payo¤ depends on the e¤ort level e exerted by the star manager, who must thus be incentivized. Since e¤ort cannot be contracted upon, the star's compensation package is made contingent on the new venture's realized payo¤. Speci…cally, the star manager receives a …xed wage CV with certainty and a fraction CV of incremental payo¤ in case of a home run. The fraction CV can be interpreted as a fraction of call options allocated to the star manager, and will a¤ect her e¤ort choice, e CV = e ( CV ). 9 The star's expected compensation can thus be expressed as W CV = CV + e CV CV , and her net expected payo¤ is Similarly, the competing VC also o¤ers a …xed wage, V C , and a fraction of call options, V C , to the star, which imply an expected compensation W V C = V C + e V C V C , and a net payo¤

Timing of the Game
The timing is as follows.
At date 0, Nature reveals parameters , , M , M , S, k and , and the …rm chooses its organizational form F 2 fC; CV g.
At date 1, the …rm and the VC both make a contractual o¤er to the star manager. Speci…cally, if F = C, the …rm o¤ers W C to the star. If F = CV , the …rm o¤ers f CV ; CV g, which implies a net expected payo¤ U CV to the star. The competitor o¤ers f V C ; V C g, which yields U V C to the star.
At date 2, the star chooses one of the two o¤ers.
At date 3, the star exerts e¤ort.
At date 4, the returns are generated and the payo¤s are distributed. 9 As documented in Chemla, Habib, and Ljungqvist (2005) and Schmidt (2003), such clauses appear to be frequently used in venture capital (and in particular corporate venture capital) contracts. Our results hold if the compensation package can also be made contingent on the spillover or on the payo¤ to the main line of business. Such contracts are not common practice.

The Corporation
In this section and the next, we analyze competition for the star manager at date 1, taking the …rm's organizational choice as given. We start by analyzing the case of the corporation, and then move on to corporate venturing in section 4.

Contract O¤ered by the Corporation, Taking U V C as Given
The …rm determines the optimal contract and its resulting expected payo¤ if it recruits the star, and if it does not, and then chooses among these two strategies the one that maximizes its payo¤.
If it intends to recruit the star manager, the …rm maximizes the following program: subject to the star's individual rationality (IR) constraint: where U V C represents the star manager's (net) reservation payo¤ (henceforth RP), i.e. her net expected payo¤ if she chooses the VC's o¤er to …nance the new venture. We assume that if indi¤erent, the star will choose the …rm's o¤er, hence the weak inequality in (4). The solution is simple: The corporation o¤ers W r C (U V C ) = U V C to the star, recruiting her at minimum cost, and generating a net 10 In contrast, if it does not intend to recruit the star, the …rm o¤ers any W nr C (U V C ) < U V C , and obtains total expected payo¤ P nr C = M + S .
We de…ne U V C as the threshold RP to the star (from the VC) such that the corporation is indi¤erent between recruiting her or not: (1) implies is strictly decreasing in U V C the corporation's best response is to recruit the star with an o¤er W r C (U V C ) if U V C U V C , and to not recruit her and o¤er any W nr 1 0 Throughout the paper, superscript r (resp. nr) stands for "recruiting" (resp. "not recruiting").

Contract O¤ered by the Competing VC, Taking U C as Given
The same type of (backward induction) process can be used to determine the VC's best response correspondence. The characterization of the optimal contract is detailed in the appendix at the end of the paper. If it intends to recruit the star manager, the VC maximizes the following program: subject to the star's incentive compatibility (IC) constraint (6), individual rationality (IR) constraint (7) and limited liability constraint (8): 0, 0; where U C is the star's RP (o¤ered by the corporation). The optimal contract can be described as follows: It is easy to check that in the …rst-best (FB) environment with a contractible e¤ort, one would obtain e F B = k . As is well-known, in our setting with non-veri…able e¤ort and limited liability for the star, the (second-best) e¤ort level is weakly lower than the …rst-best level: e r V C (U C ) e F B . 11 More interestingly for our purpose below, the following result emerges from this characterization: The power of incentives r V C o¤ ered by the VC, and the star's e¤ ort e r V C , are both weakly increasing in the star's reservation payo¤ U C .
) the contract o¤ered by the VC to hire the star would leave rents to the star in order to elicit the desired e¤ort level. An increase in U C would simply reduce the rent, but 1 1 The star's equilibrium e¤ort depends on her marginal bene…t from e¤ort, which is her expected share of incremental pro…ts in the good state, (see IC constraint (6)). The unconstrained program would generate …rst-best e¤ort by giving the star full residual control over incremental pro…ts, i.e. = 1, with the VC extracting all expected rents through a …xed payment < 0. However, this payment would violate the star's LL constraint 0. Hence any share of pro…ts given to the star is forfeited for good. This implies 1, and e r V C (UC ) e F B .
would have no impact on the contract f0; 1 2 g o¤ered, nor on the resulting e¤ort e r V C = 2k . However, , IR constraint (7) is binding: The VC can no longer o¤er f0; 1 2 g, because the star would turn down the o¤er and take the superior rival o¤er U C . Instead, the VC o¤ers a contract such that the star's net expected payo¤ is higher than U C by a small amount ". As U C rises, the VC must increase his bid, which he does by increasing r V C , thus "killing two birds with one stone." First, he increases the star's expected compensation and ensures her participation. Second, he increases the star's marginal bene…t from e¤ort, , and these higher-powered incentives lead to a higher e¤ort level. This is the key intuition behind Lemma 1. Finally, as U C becomes "high" and enters h 2 2k ; 1 , o¤ering the same contract as in would lead to an excessively high e¤ort.
Instead, the VC o¤ers a contract that elicits the …rst-best e¤ort and ensures participation.
Using (9), one can easily derive the expected net payo¤ for the star, if the star is hired by the VC: The VC will recruit the star provided his recruitment payo¤ P r V C > P nr V C = 0. We show that for each value of min there exists a unique threshold reservation payo¤ U C 2 8k de…ned implicitly as the solution to: such that, if U C < U C the VC recruits the star with a contract f r V C (U C ) ; r V C (U C )g to the star; and if U C U C , the VC does not recruit the star (and o¤ers any contract f nr V C ; nr V C g such that

Nash Equilibria of the Subgame
In the foregoing two subsections, we characterized the corporation's best-response for any U V C 0, and the VC's best response for any given U C 0. The determination of Nash equilibria (NE) follows directly by intersecting these two correspondences.

Proposition 1
The Nash Equilibria in this subgame are characterized as follows: which is also the pay-o¤ to the star. The VC o¤ ers one of many possible contracts The star takes the corporation's o¤ er, and this leads to payo¤ s P C1 U C = M W C1 U C to the corporation and P V C1 = 0 to the VC.
There exists a unique NE where the corporation o¤ ers W C2 = U V C and the with " ! 0; and the star opts for the VC's o¤ er. The VC and the corporation receive Consider the case where U C U V C . Clearly, the star cannot receive less than U C , otherwise at least one of the rivals would make a higher o¤er to attract the star. By de…nition of U C , the VC will not outbid an o¤er from the corporation that equals (or is higher than) U C . Hence, the corporation recruits the star at the lowest possible cost, i.e. U C . Note that there exist many NE, all Importantly, all of these NE yield the same outcome, and the same payo¤s to the di¤erent players. In the case where U C > U V C , the intuition is similar, with the VC o¤ering U V C + " and hiring the star at minimum cost.
The base return to venturing is one factor that will a¤ect the relative threshold RPs and hence whether the star is hired by the corporation or by the VC, and in turn the corporation net expected payo¤. 12 When increases, the VC's net expected payo¤ after hiring the star also increases. Hence threshold RP U C goes up as well. In contrast, the corporation, which focuses on its main line of business, remains una¤ected by an increase in , so its threshold RP, U V C , does not vary with .

Lemma 2
If the …rm is organized as a corporation, there exists a threshold level of return to venturing b min -de…ned as the level of such that U C b = U V C -such that the …rm recruits the star i and the VC recruits the star when 2 A 2 = b ; 1 . The expected net payo¤ to the corporation can be written: 1 2 Our results would be qualitatively the same if rather than was allowed to vary.
When the return to venturing is small and close to min , the VC has little to gain from recruiting, and the maximum amount it can "bid" for the star, U C ( ), is low. The corporation can therefore recruit the star by o¤ering a salary W C1 U C ( ) ; which grants the star the same payo¤ as the VC's best bid. As the return to venturing increases, so does the VC's bid for the star, since U 0 C ( ) > 0. The corporation must increase its o¤er to the star accordingly, and this negative "compensation cost"e¤ ect (of an increase in ) decreases its net payo¤. As the return to venturing becomes large and crosses the threshold b , it becomes too expensive for the corporation to match the VC's o¤er. Therefore, beyond b the corporation does not match the VC's o¤er, and the VC recruits the star.
Proposition 2 An increase in retuns to venturing increases the VC's valuation for the star. This prompts the corporation to make a higher bid for the star, thereby reducing its net expected payo¤ .
Due to this "compensation cost" e¤ ect, the relative attractiveness of recruiting the star decreases with , and it is negative when > b : the IC and LL constraints are the same as in (6) and (8), respectively; and the IR constraint, di¤ers from (7) only in that the star's RP is the VC's o¤er U V Cs rather than U C , and in that the inequality is weak. 13 Hence, the equilibrium e r CV , r CV , r CV , and W r CV = r CV + e r CV r CV are as described in (9), but with " = 0 and U V Cs instead of U C . As in Subsection 3.2, for any min there exists a unique threshold RP to the star U V Cs such that if U V Cs U V Cs the CV …rm recruits the star with an o¤er f r CV (U V Cs ) ; r CV (U V Cs )g; and if U V Cs > U V Cs the CV …rm o¤ers any contract f nr CV ; nr CV g such that U CV ( nr CV ; nr CV ) < U V Cs and does not hire the star. Here, U V C2 is de…ned implicitly as the solution to P r CV U V Cs = P nr CV , that is: where the …rst square-bracketed term represents the CV …rm's expected payo¤ from the new venture, the second square-bracketed term represents its payo¤ from the main business, and the right-hand side represents its payo¤ if it does not recruit.

Contract O¤ered by the Competing VC, Taking U CV as Given
The program is the same as in Subsection 3.2, and hence so is the contractual o¤er, simply replacing

Nash Equilibria of the Subgame
As noted above, the CV …rm's program and the VC's program generate the same equilibrium effort functions (e r CV (:) = e r V Cs (:)), and very similar managerial compensation functions (W r CV (:) W r V Cs (:)). Comparing (11) and (15), since 0; the surplus generated by the star for the CV …rm is always higher than for the VC.
If the VC recruits the star, the e¤ects of spillovers on the …rm's main line of business will not be internalized, and hence the spillovers will be smaller -by an amount -than if the CV …rm recruits the star and internalizes the impact of spillovers. This bene…t from recruiting does not accrue to the VC, who has only one project and is not a¤ected by spillovers. 14 As a result, the CV …rm's threshold RP is higher than the VCs, and the CV …rm successfully recruits the star. The highest bid the VC can o¤er the star is U CV ; but since U V Cs U CV , the CV …rm matches that bid and hires the star.
Formally, the equilibria can be characterized as follows: 1 4 Spillovers to the VC's other projects or lines of business are examined in Subsection 7.2.

Proposition 3
In equilibrium the CV …rm recruits the star by o¤ ering contract f CV ; CV g U CV = f r CV ; r CV g U CV . This generates a payo¤ U CV U CV = U CV to the star, a compensation cost W CV U CV = W r CV U CV to the …rm, and an equilibrium e¤ ort e CV U CV = e r CV U CV . The VC o¤ ers one of many possible contracts f V Cs ; V Cs g such that U V Cs ( V Cs ; V Cs ) = U CV . The net expected payo¤ s to the CV …rm and to the VC are P CV U CV = + e CV U CV W CV U CV + M + S and P V Cs = 0, respectively. 15 In this subgame, returns to venturing have no impact on recruitment. The CV …rm recruits the star for all min because an increase in a¤ects the threshold RP levels for both the CV …rm and the VC, U V Cs and U CV , similarly, such that U V C2 remains higher than U CV for all .
Returns to venturing do a¤ect the expected net payo¤ from corporate venturing in several ways, however. Note that the CV …rm's compensation cost can be expressed as the sum of the star's net payo¤ and her cost of e¤ort, W CV U CV = U CV + c e CV U CV ; and recall from the previous subgame that U CV ( ) = U C ( ) is a strictly increasing function of for all min . 16 We can write the CV …rm's payo¤ as: where is the base return to venturing, U CV ( ) is the net payo¤ captured by the star, the …rst square-bracketed term represents the marginal expected return from a "home run", and the second one represent rents from the the main line of business.
Intuitively, an increase in has: A positive "gravy train" e¤ ect that takes place through , and that increases the payo¤ to …rms involved in new ventures, including the CV …rm.
A negative "compensation cost" e¤ ect that takes place through U CV ( ) = U C ( ), and that is identical to the one faced by the corporation. An increase in increases the VC's ability to "bid" for star managers; and this prompts the corporation to increase its own bid for the star, thereby reducing its net expected payo¤. 1 5 As in the previous subgame, there exist many NE, all with contractual o¤er f CV ; CV g by the CV …rm, but each with a di¤erent combination of f V Cs ; V Cs g. All of these NE yield the same outcome, and the same payo¤ to the di¤erent players. 1 6 Clearly, U CV ( ) = U C ( ) since these two thresholds make the same VC indi¤erent between recruiting or not. As discussed in the previous subgame, U A positive "managerial incentives"e¤ ect that takes place through e CV U CV ( ) c e CV U CV ( ) .
As increases the CV …rm must increase the star's compensation, but (as highlighted in Lemma 1) this is achieved by (weakly) increasing the power of incentives CV . This has a positive e¤ect on e¤ort e CV U CV ( ) and in turn on the marginal expected "home run" return.
Proposition 4 An increase in returns to venturing has three e¤ ects on the expected net payo¤ from engaging in corporate venturing: A positive "gravy train" e¤ ect, the same negative "compensation cost" e¤ ect as in the corporation, and a positive "managerial incentives" e¤ ect. In equilibrium the compensation cost e¤ ect exactly o¤ sets the direct and managerial incentives e¤ ects, and P CV is a constant function of :

Organizational Choice
The last step in the characterization of the subgame-perfect equilibrium is the determination of the …rm's organizational choice at date 0. The …rm will organize for corporate venturing rather than as a corporation if and only if the relative "attractiveness" of the former organizational form vis-à-vis the latter, P = P CV P C , is positive. In the following subsections we characterize the …rm's optimal organizational form, and how it is a¤ected by returns to venturing, competition for talent, and spillover e¤ects.

Impact of Returns to Venturing
Region Using (12) and (16), the relative attractiveness of corporate venturing in region A 1 , P 1 = P CV P C1 , can be written: The relative attractiveness of corporate venturing in region A 2 , P 2 = P CV P C2 , can be written: There are two key di¤erences between P 1 and P 2 . First, in region A 2 , the payo¤ to the star under corporate venturing is no longer o¤set by an identical payo¤ if she is hired by the corporation, because in the latter case there is no recruitment. Hence U CV ( ) appears in P 2 and it has a negative impact on the relative attractiveness of corporate venturing. Second, the fact that the corporation does not recruit in region A 2 a¤ects the rents on the main line of business through spillovers from the new venture. With corporate venturing, the new venture is developed under the umbrella of the …rm, which yields spillovers S on its main line of business. With the corporation, it is the VC who recruits the star and develops the new venture, leading to smaller spillovers S .
The rents on the main line of business are therefore larger under corporate venturing than with the corporation, by an amount , the spillover di¤erential; and this has a positive impact on P 2 .
We know from Subsection 4.3 that the rents extracted by the star under corporate venturing exactly o¤set the expected bene…t from the new venture. This has two consequences. First, the expression for the relative attractiveness of corporate venturing simpli…es to P 2 = > 0: corporate venturing is the optimal organizational form in region A 2 , because it allows the …rm to successfully recruit the star and internalize the spillover di¤erential. Second, returns to venturing have no impact on P 2 . This is because the compensation cost e¤ect associated with corporate venturing (which cancels out in P 1 but not in P 2 ), exactly o¤sets the direct e¤ect and the managerial e¤ect (which are present in both P 1 and P 2 ).

Lemma 4
In region A 2 , P 2 = 0. An increase in returns to venturing has no impact on the relative attractiveness of corporate venturing, and corporate venturing is the optimal organizational form.
Lemma 3 and 4 imply that the relative attractiveness of corporate venturing weakly increases with returns to venturing over ( min ; 1). Together with regularity conditions (1) and (2) which ensure P ( min ) < 0, they yield the following Proposition:

Impact of Competition for Talent
Competition for talent and its interaction with returns to venturing in the organizational choice are central features of our analysis. As a benchmark, suppose that the …rm does not face competition for the star manager (e.g. there is no VC interested in her), and can recruit her as long as it pays her at least some reservation payo¤ U 0 = 0. 17 The expected net payo¤ to the …rm (which successfully recruits the star for all min ), depends on whether it is organized as a corporation or for corporate venturing, and is written P C0 = M U 0 = M or P CV 0 = + 2 4k + M + S, respectively. 18 Clearly the relative attractiveness of corporate venturing in the absence of competition, P 0 = + Competition from the VC prompts the …rm to react in several ways. In region A 1 , it increases the star's compensation whether the …rm is organized as a corporation or for corporate venturing, with zero net e¤ect on P 1 ; but under corporate venturing it has the additional e¤ect of increasing the power of incentives. This leads to higher e¤ort, yielding an incentives bene…t like the one mentioned above. In region A 2 , competition for talent creates a recruiting problem for the corporation, and corporate venturing is the solution to that problem: It enables the …rm to successfully recruit a star who would otherwise accept the VC's o¤er, and to internalize the di¤erential impact of new venture spillovers on its main business.
Thus whether e 0 is located in region A 1 or A 2 , the marginal …rm, which is indi¤erent between the two organizational forms absent competition at e 0 , would strictly prefer corporate venturing in a competitive market for talent, in an attempt to capture either incentives bene…ts, or recruitment/retention bene…ts. The implication, then is that e < e 0 : Competition for talent strictly increases the prevalence of corporate venturing over h e ; e 0 .

Impact of Spillovers
Spillovers also have an impact on the …rm's optimal organizational form. Indeed, at the switching threshold level of returns to venturing e 2 A 1 the surplus created with corporate venturing increases with S, while the payo¤ when organizing as a corporation is not a¤ected.
Proposition 7 Spillovers S increase the relative attractiveness of corporate venturing and, hence, decrease the threshold level of returns to venturing e beyond which corporate venturing is the optimal organizational form. Interestingly, in contrast to his model, here corporate venturing may still be optimal even if S is negative. Important for this result is the fact that even when the new venture has a very negative impact on the main business, corporate venturing may still be optimal if it prevents the VC from recruiting the star and starting the venture, in which case the impact on the main business would be even worse, by an amount . (Another di¤erence with his model is that here spillovers are not restricted to product-market complementarity or substitutability, and may take a more general form. See discussion in section 8.)

Control Rights Allocation and the Di¤erent Types of Corporate
Venturing Corporate venturing is a general expression that is typically used to describe all types of investments made by corporations into new ventures that are distinct from their core business. In practice, however, corporate venturing investments can take many forms. At one end of the spectrum is the "internal To make this framework comparable to our base complete contracting model, we assume that the "home run" marginal payo¤ is not contractible, while the base payo¤ is. The initial contract at date 1 therefore speci…es only a payo¤ to be paid to star at date 4, and the remaining payo¤ to the …rm. The contract may also specify the relative allocation of control/property rights over the idea, . The variable may for example represent the fraction of the realized payo¤ that the star can obtain (or the probability that she would successfully replicating the venture elsewhere) if bargaining breaks down, while the …rm gets (1 ). The marginal payo¤ is bargained over at date 4 if it is realized. Assuming Nash-like bargaining, the fraction f of extracted by the star will depend on both negotiating parties' relative threat points at that time, which in turn depend on .
In the example just given, Nash bargaining implies f ( ) = . 19 Then the problem with incomplete contracting becomes exactly the same as the one in the base case, simply replacing by and by . all rents in renegotiation at date 4. In the …rm/VC-friendly regime, the …rm or the VC (depending on whom the star chose) owns the idea, and extracts all rents at date 4. 21 In the …rm/VC-friendly regime, the intuition is very similar to that of our main model, but with zero e¤ort from the the star, since she anticipates that all rents will be extracted by her employer.
In that case, the gravy train and recruitment/retention e¤ects are still present and there still exists a threshold level of returns to venturing such that corporate venturing is optimal beyond that threshold.
In the entrepreneur-friendly regime, if the venture were …nanced, the star would have …rst-best incentives, since she anticipates extracting all rents. However, for the same reason the VC expects a

Competition, E¢ ciency, and Economic Activity
Corporate venturing is socially optimal if and only if the total surplus generated in the …rst-best (net of e¤ort cost), + 2 2k + M + S, is higher than both i) the total surplus that could be created by allocating the star manager to the competitor, i.e. + 2 2k from the venture plus M + S from the main line of business, and ii) the total surplus that could be generated when organizing the …rm as a corporation, M . and Condition (20) reduces very simply to 0, which always holds in our model; and inequality . Thus, corporate venturing is socially optimal for all F B . The following result then obtains: In equilibrium there is less corporate venturing than is socially optimal: e > F B .
Competition for talent simultaneously increases the prevalence of corporate venturing and improves economic e¢ ciency.
This ine¢ ciency is due to agency costs in corporate venturing. When e¤ort is not directly contractible, the …rm has to leave rents to the star to induce her to exert a high e¤ort level, and this usually leads to a second-best e¤ort level from the star. This agency cost constrains the rents captured by the …rm if it engages in corporate venturing, and induces …rms to switch to corporate venturing "too late"relative to the social optimum: for all 2 h F B ; e , the …rm chooses to remain organized as a corporation even though from a social point of view it ought to organize for corporate venturing.
This ine¢ ciency is even larger absent competition for talent, since as shown in Proposition 6, in that case the …rm switches to corporate venturing "even later,"at e 0 > e . As discussed in subsection 5.2, competition for talent increases equilibrium compensation for the star, and under corporate venturing leads to higher-powered incentives and higher e¤ort exertion. This simultaneously improves the relative attractiveness of corporate venturing and its prevalence in the h e 0 ; e region, and economic e¢ ciency.

Spillovers
So far we have assumed that spillovers a¤ect only the …rm's main line of business, and that the VC receives no spillovers on other ventures in its portfolio. This assumption was made both for simplicity, and to yield starker results. In this Subsection we generalize the model to allow the VC's other projects to bene…t from spillovers from the new venture. 22 The following results can readily be shown:

Proposition 10
In the more general model where the VC also enjoys a spillover S V C if it recruits the star, and S V C V C if she is not, all results remains unchanged as long as V C , i.e. as long as the spillover di¤ erential is higher for the …rm than for the VC. If V C > , then for all 2 A 1 the …rm chooses to organize as a corporation rather than for corporate venturing, and for all 2 A 2 , the …rm is indi¤ erent between the two organizational forms.

As long as V C
, if organized for corporate venturing the …rm has a higher incentive to recruit the star than the VC; and therefore, as before, it recruits the star under this organizational form. On the other hand, if V C > , the VC bene…ts more from recruiting the star, and it recruits the star whenever the …rm is organized for corporate venturing. Hence, organizing as a corporation is optimal when it yields a higher payo¤ than not recruiting, i.e. in region A 1 . In region A 2 , the …rm cannot recruit the star regardless of its organizational form, and is therefore indi¤erent between the two. 23 8 Empirical Implications Our model yields a number of empirical predictions that can improve our understanding of corporate venturing in general, and of its structure in particular. Beyond corporate venturing, it also points to a potential link between labor market competition and …rm productivity. In what follows we discuss these predictions and their relation to the empirical literature.

Determinants of Corporate Venturing Investments
A key contribution of our model is to shed light on the reasons why …rms engage in corporate venturing activities, and we identify three key explanations. The gravy train rationale simply states that …rms engage in corporate venturing in an attempt to capture some the high returns returns they observe in new ventures, and therefore suggests that higher venturing returns should increase CV investments.
The incentives rationale captures the idea that as the star's reservation payo¤ (the compensation she expects to obtain in the labor market) increases, …rms respond by o¤ering higher-powered incentives; 24 this induces stars to exert more e¤ort, and increases the relative attractiveness of venturing activities where the star's marginal product of e¤ort is higher. This increase in the star's market compensation may come from higher returns to venturing, which prompt employers to "bid" higher to recruit star managers. Alternatively, it may come from a rise in the degree of competition for talent, for a given level of venturing returns. For example, the shortage of talent may vary across industries/sectors and across time, generating variation in the intensity of competition to recruit stars.
Thus the incentives rationale suggests that both higher venturing returns and stronger competition for talent should increase CV investments.
Finally, the recruitment/retention suggests that …rms may engage in corporate venturing in an attempt to successfully recruit stars who would otherwise take employment elsewere. 25  Firms'concern about failing to recruit or retain star managers is stronger when these stars have access to higher levels of market compensation, which again could result from high levels of venturing returns or from strong competition for talent.
Corporate venturing investments should therefore increase in both of these situations.
Thus, our three rationales point to two primary factors that will likely a¤ect corporate venturing investments: Prediction 1: Higher returns to venturing should increase corporate venturing investments. the …rm for the recruitment of a star manager -with measures of labor mobility for example. Indeed, Fallick, Fleischmann and Rebitzer (2006) …nd higher job mobility in the Silicon Valley's computer industry than in computer clusters elsewhere, and a natural application of our model would be to relate variation in job mobility to variation in corporate venturing investments. Another possible proxy for labor market competition would be the geographic proximity of …rms to centres of venture capital activity: …rms closely located to the Silicon Valley, for instance, would likely face stronger competition for talent from local venture capitalists and startups. Finally, temporal and industry variations in the university wage premium (i.e. the di¤erence in wage between university graduates and high-school graduates) may also adequately capture variations in the degree of competition for talent. Using these measure of competition for talent to examine its impact on corporate venturing investment would -it seems -be fruitful.
Our analysis also suggests that corporate venturing investments may be a¤ected by two other factors. One such factor is spillovers, broadly de…ned as any (positive or negative) impact of the new venture on the main business. Spillovers are sometimes grouped under the more generic name of "strategic factors" in the corporate venturing literature. Block and MacMillan (1993), and Chesbrough (2002) report that these strategic factors play an important role in the decision to pursue corporate venturing activities. In the US for example, 76% of corporate venturing …rms pursue these activities for strategic purposes (Block and MacMillan, 1993).
Spillovers could be interpreted in several ways. For example, they could be technology spillovers, allowing the …rm to gain access to a new technology for the main business; knowledge spillovers, enabling the …rm to improve its expertise; or product-market spillovers such as complementarity/substitutability between the new venture product and main business product. 26 All three types of spillovers have been identi…ed in the empirical literature as having an impact on CV. Siegel et al. (1988) identify "exposure to new technologies and markets" as a key objective for CV …rms, while Winters and Mur…n (1988) underline how corporate venturing may enable …rms to gain a "window on new technology/business." Dushnitsky and Lenox (2005b) document the impact of knowledge spillovers, reporting that "the more closely aligned the domains of expertise of the …rm and a particular sector, the greater the likelihood that the …rm will invest in [corporate venturing] in that sector."Finally, Dushnitsky and Shaver (2006) …nd that the greater the complementarity between the products of the corporate parent and those of the entrepreneur, the greater the likelihood of a corporate venturing relationship to form.
Our model also predicts that the strength of intellectual property (IP) protection for the inventor should have an impact on CV. Corporate venturing should be more prevalent when IP protection is weaker, because in such cases investors (in the form of …rms or VCs) anticipate reaping rewards from the new venture, and are more willing to …nance ventures in the …rst place.

Prediction 4: Weaker IP protection for the inventor should increase corporate venturing investments.
This prediction is consistent with recent empirical work on corporate venturing and its connection to IP protection. Dushnitsky and Lenox (2005b), for instance, derive a measure of the e¤ectiveness of patents in protecting inventors'pro…ts, and …nd that the weaker the IP protection for the inventor (i.e. the lower the e¤ectiveness of these patents), the higher the prevalence of corporate venturing. They also show that weaker IP protection may improve the innovation rate in …rms engaged in corporate venturing (Dushnitsky and Lenox, 2005a).

Structure of Corporate Venturing
The literature on corporate venturing suggests that it may be structured in many ways, di¤ering mainly in the degree of autonomy of the new venture from the corporate parent (Rind, 1981;Roberts and Berry, 1985;Bleicher and Paul, 1987;Gompers, 2002). Our model captures this idea explicitly: The structure of corporate venturing is de…ned in terms of relative control rights allocation and hence in terms of the star autonomy from the corporate parent. Moreover, prior work suggests that the key advantage of autonomy in the structure of CV is improved innovative activity (Fast, 1981, Sykes, 1986Russell, 1995;Thornhill and Amit, 2000). This is also consistent with our model where more autonomy in the form of control rights increases the star's ex post bargaining power and hence her ex ante incentives to exert e¤ort. On the other hand, a tighter relationship between the new venture and the corporate parent may enable the new venture to take advantage of the parent's core competencies (Dougherty, 1995). Dushnitsky and Shaver (2006) argue that the degree of autonomy of the new venture might facilitate the formation of corporate venturing activities, by alleviating the star/entrepreneur's concerns about imitation/expropriation by the corporate parent; and …nd empirical support for this hypothesis. But what exogenous factors might a¤ect the optimal choice of corporate venturing structure? Our model suggests that returns to venturing may play a role in that regard. It predicts that as returns to venturing increase, more autonomy (i.e. control rights) should be allocated to the star, and the organization of corporate venturing should gradually switch from a structure closer to internal ventures, to one that resemble more corporate venture capital: Prediction 5: Higher returns to venturing should a¤ ect the organization of corporate venturing activities, leading to structures where the new venture gradually becomes more independent from the corporate parent.

Beyond Corporate Venturing: Competition for Talent and Productivity
Finally, the model allows us to make a more general prediction about the connection between labor market competition and …rm productivity: Prediction 6: Competition for talent should increase …rm productivity.
Competition for talent, through its positive impact on equilibrium compensation and the power of incentives, may lead to increased managerial e¤ort towards the …rst-best, thus generating productivity gains. Investigating this link empirically would be the logical next step, and in our opinion a promising avenue for future research. 27

Conclusion
A natural explanation for corporate venturing and its pro-cyclicality with returns to entrepreneurship is that …rms engage in corporate venturing in an attempt to capture a share of these returns when they are high. This "gravy train" e¤ect may explain the positive impact of venturing returns on corporate venturing. This seems unlikely to be the only explanation, however. The lower returns associated with CV investments, relative to independent VC investments returns (Fast, 1981;Zahra, 1996;Gompers andLerner, 1998, Gompers, 2002), suggest that investments through traditional venture funds would likely be a better way to scoop the "gravy." In this paper we propose two novel explanations for corporate venturing and its pro-cyclical ‡uctuations with entrepreneurial activity. We argue that …rms may engage in corporate venturing activities also to bene…t from the superior managerial incentives in new ventures, and to recruit or retain key talent. A important insight to emerge from our analysis is that competition for talent is a key factor in determining corporate venturing investments. This and the other testable predictions of the model suggest several fruitful avenues for future empirical research.

A Appendix: Characterization of the VC' s Optimal Contract
The IC and the IR constraints respectively reduce to: Replacing in the objective function (5), we can re-write the VC's program as follows: subject to (23) and 0: Solving (24) for e, we obtain e r V C (U C ) = 2k which implies r V C (U C ) = ke = 1 2 . Hence: (23) and (8) hold with e = e r V C (U C ) = 2k and = 0. Hence fe r V C ; r V C ; r V C g (U C ) = f 2k ; 0; 1 2 g. When U C 2 h 2 8k ; 2 2k , (23) no longer holds with fe r V C ; r V C ; r V C g (U C ) = f 2k ; 0; 1 2 g. This is solved by increasing e¤ort along with U C , so that k 2 e 2 = U C or e r V C (U C ) = q 2 k U C ; and by setting When U C 2 h 2 2k ; 1 , keeping e = q 2 k (U C ") would generate an e¤ort level that would be (ine¢ ciently) higher than e F B . Instead the VC elicits e r V C (U C ) = e F A 2 = k by setting r V C (U C ) = U C + " 2 2k and r V C (U C ) = 1, so that the IC, IR, and LL constraints all hold.
The star's and the VC's expected payo¤s when the VC recruits the star, c (e r V C (U C )) and P r V C (U C ) = + e r V C (U C ) W r V C (U C ) are described in (10). If the VC does not intend to recruit the star it o¤ers any f nr V C (U C ) ; nr V C (U C )g such that U V C ( nr V C ; nr V C ) 2 [0; U C ). As long as U nr V C ( nr V C ; nr V C ) < U C , the star chooses the corporation's o¤er, and the VC receives a payo¤ P nr V C = 0.
We de…ne U C as the threshold RP to the star (from the corporation) such that the VC is indi¤erent between recruiting or not: P r V C U C = P nr V C , which is expressed more precisely in (11). Regularity condition (2) ensures that, for any given is continuous over R + and strictly decreasing in U C over h 2 8k ; 1 , and since P r V C (1) = 1, there must exist a unique U C 2 8k de…ned as in (11), such that it is optimal for the VC to recruit if and only if U C < U C . 28  It follows directly from the characterization of the optimal contract, and from expression (9) in the text, that the power of incentives r V C (U C ) and equilibrium e¤ort e r V C (U C ) are both weakly increasing functions of U C .

Proof of Proposition 1
Follows directly from the text.

Proof of Lemma 2
From (11), the derivative U 0 C ( ) can written using the implicit function theorem: (A1) We know that @P r V C @ = 1 > 0. We verify from (10) that U C ( min ) = 2 8k , and that In contrast, U V C = M (M + S ) is independent of . From regularity constraint (1)

Proof of Proposition 2
The payo¤ to the corporation is clearly expressed in (12), and the impact of returns to venturing on this payo¤ can be written: 1 If regularity conditions (1) and (2)

Proof of Proposition 3
Follows directly from the text.

Proof of Proposition 4
Di¤erentiating (16) with respect to we obtain: The direct e¤ect is represented by the number one on the right-hand side (RHS) of (A3). The "compensation cost" e¤ect is captured by the second term on the RHS of (A3). A proof identical to that of Lemma 2 shows that this e¤ect is negative. Finally, the third term on the RHS of (A3) re ‡ects the positive "managerial incentives" e¤ect. Using the proofs of Lemmas 1 and 2, this e¤ect is easily shown to be positive.
To see that P CV = M + S + " is independent of , recall that by de…nition, U CV satis…es: This can be rewritten as: Since the CV …rm and the VC make the same o¤er and elicit the same e¤ort, i.e. e r V Cs (:) = e r CV (:) = e CV U CV , we can substitute (A5) into (16) and obtain P CV = [M + S] + ", with " ! 0.
Accordingly, the negative compensation cost e¤ect of exactly o¤sets the positive direct e¤ect and managerial incentives e¤ect: 7 Proof of Lemma 3 The …rm recruits the star under both organizational forms. The star's payo¤ is the same under both organizational structures (U CV ( ) = U C ( )), and therefore cancels out of P 1 . Equation (18) then obtains from (10) and (16). Di¤erentiating (18) with respect to yields: 8 Proof of Lemma 4 Follows directly from the text.

Proof of Proposition 5
It is su¢ cient to prove the existence of e . Like P CV ( ) and P C ( ), P ( ) is a continuous function

Proof of Proposition 6
From the text we know that P CV 0 = + 2 4k + (M + S) and P C0 = M . Therefore the relative attractiveness of corporate venturing in the absence of competition can be written:  (1) and (2) do not hold, then for some parameter values, a special case may occur where P ( ) 0 for all values of 2 [ min; 1). In that less interesting case, corporate venturing is always optimal.
The e¤ect of competition on the relative attractiveness of corporate venturing can be written: From the proof of Lemma 2 we know that U CV ( min ) = U C ( min ) = 2 8k . Using (9), we obtain P 1 ( min ) P 0 ( min ) = , and weakly increasing in for all > min . Therefore so is P 1 P 0 .
Hence, P 1 P 0 > 0 for all 2 A 1 . In particular P 0 e < P 1 e = 0. This implies that P 0 switches from negative to positive at a threshold level e 0 > e .

Proof of Proposition 7
From (18), the relative attractiveness of corporate venturing in region A 1 at the threshold level e can be written: @ e > 0. Since @ P 1 @S > 0, this implies that d e dS < 0.

Proof of Lemma 5
Suppose that is not contractible, while is. The initial contract at date 1 therefore speci…es only a payo¤ to be paid to star at date 4, and the remaining payo¤ to the …rm. The contract may also specify the relative allocation of control/property rights over the idea, . The variable represents the fraction of the realized payo¤ that the star can obtain (or the probability that she would successfully replicating the venture elsewhere) if bargaining breaks down, while the …rm gets (1 ).
The marginal payo¤ is bargained over at date 4 if it is realized. Assuming Nash bargaining, the star will give a transfer t to the …rm, such that t 2 arg max ( t ) (t (1 ) ). This yields t = (1 ) leaving the star with t = , and the …rm with t = (1 ) .
Thus, the …rm chooses and to maximize + e ( + e ), in exactly the same way as it chose and , respectively, in our base model. Then the problem with incomplete contracting becomes exactly the same as the one in the base case, simply replacing by and by .

Proof of Proposition 8
Follows directly from the text.
14 Proof of Proposition 9 We have shown in the text that corporate venturing is optimal from a social point of view if  The proof that competition for talent simultaneously increases the prevalence of corporate venturing and improves economic e¢ ciency follows directly from the text.

Proof of Proposition 10
Assume that the VC enjoys spillovers from the new venture to its other portfolio ventures. These spillovers take on value S V C if the new venture is developed by the VC, and S V C V C if it is developed by the …rm. Then: 1. If the …rm is organized as a corporation, the threshold RP U C faced by the VC is now slightly higher, re ‡ecting the impact of the VC's spillover di¤erential V C on its surplus from recruitment.
Equation (11) must thus be replaced by: In turn the higher value of U C implies a lower value of b , the threshold level of returns to venturing beyond which the VC, rather than the corporation, recruits the star.
2a If the …rm is organized for corporate venturing, its threshold RP U V C remains unchanged, implicitly de…ned as in (15). It follows directly that, if V C , the …rm can always recruit the star if organized for corporate venturing. Our qualitative results are as above, even though the key thresholds e and b might be di¤erent.
2b If on the other hand, < V C , then the VC, rather than the …rm, recruits the star when the …rm is organized for corporate venturing. The …rm receives M + S if i) it organizes for corporate venturing, or ii)) it organizes as a corporation and > b . Therefore, the …rm organizes as a corporation for all 2 h min ; b i , and is indi¤erent between organizational forms C and CV for all 2 b ; 1 (in which case we have assumed it chooses corporate venturing).