Real-world financial contracts are sometimes so complex that it can be difficult to understand their exact payoff consequences. We develop and test a theoretical model of a venture capitalist (VC) negotiating with an entrepreneur who may overweigh or underweigh the payoff consequences of contractual downside protection (DP). A lawyer with expertise in venture capital can inform the entrepreneur about these consequences, but less expert (but otherwise high quality) lawyers cannot. We determine how a VC's decision to include DP is affected by the expected quality of the entrepreneur's project, the entrepreneur's experience, and the VC expertise of his/her legal counsel. We show that the VC's incentive to include unnecessary DP declines in expected project quality. Indeed, for inexperienced entrepreneurs involved with high-quality projects, VCs prefer that the entrepreneur's counsel has VC expertise. This implies that, when negotiating with inexperienced entrepreneurs, VCs who invest in high-quality companies should be more likely to negotiate with entrepreneurs who employ lawyers with VC expertise. We document broad empirical support for the model, and provide evidence against competing explanations. (JEL L24, G24, K12, LI4, L24)

I. INTRODUCTION

Downside protection (DP), which provides one contracting party proportionately larger payoffs in bad states of the world, is a prevalent feature of real-world financial contracts. A large literature motivates various forms of DP by deriving them as optimal solutions to moral hazard and adverse selection problems. These frictions are clearly present in many important contracting settings. Yet, there could be other reasons why contracts so often include DP.

Traditionally, economists presume that, although the contracting parties may not know each other's unobserved type (e.g., ability), they do understand the payoff implications of negotiated contract terms. However, in many settings, the premise that all contracting parties understand the payoff implications of all contract terms is implausible. Tirole (2009, 265) observes that "thinking about contingencies, designing covenants, and seeing through their implications is costly." He then explores the implications of cognition costs in contracting between parties who are unaware, and know that they are unaware.

We build on a related observation. We recognize that in some contexts, there are vast differences in how accurately the contracting parties can forecast the payoff consequences of complex contract terms. We study venture capital (VC) investment, where comprehension differences between the contracting parties--the venture capitalist (VC) and entrepreneur--are acute. VC contracts include a myriad of esoteric and intricate cash flow contingencies (Bartlett 1995; Hart 2001; Kaplan and Stromberg 2003; Metrick and Yasuda 2010): VCs are well-versed about these contingencies, as they repeatedly negotiate contracts and observe their outcomes. In contrast, entrepreneurs, whose attention is focused on innovation, often lack a detailed understanding of these terms when they negotiate their first VC contracts. As a result, otherwise identical entrepreneurs can come to very different assessments of the payoff implications of a cash flow contingency. In turn, those different assessments can lead them to reach very different conclusions about the attractiveness of a proposed contract.

We investigate the consequences of these observations both theoretically and empirically. In our theoretical model, a VC and entrepreneur negotiate a financial contract for which "extra" DP is sometimes, but not always, necessary to mitigate the relevant moral hazard. VCs understand the payoff implications of DP, but entrepreneurs may not unless they are represented by legal counsel with VC expertise. Expert (as opposed to high quality) counsels understand the implications of contract terms, and can level the playing field for entrepreneurs by explaining, evaluating and benchmarking contract terms. (1) A VC can offer a contract with unnecessary DP, hoping to extract surplus from a mis-informed entrepreneur. A VC can also choose to inform the entrepreneur by introducing him/her to an expert legal counsel. VCs know an entrepreneur's experience and his/her counsel's VC expertise. Hence, a VC can identify a subset of entrepreneurs who correctly evaluate the payoff implications of DP. In particular, when a VC negotiates a contract with a novice entrepreneur represented by an inexpert legal counsel, the VC understands that the entrepreneur may misprice the payoff consequences of this added DP. However, the VC does not know whether the novice entrepreneur would underweigh, overweigh, or correctly weigh these consequences.

It follows that a VC should internalize the possibility that an entrepreneur may misprice DP. If a VC believes an entrepreneur is likely to under-weigh the adverse payoff consequences of DP, he/she may introduce such contingencies, even when those contingencies can, themselves, subsequently distort incentives. The VC's offer may combine this DP with a slightly smaller equity position, which entrepreneurs can correctly price since its payoff consequences are clear. With this contract, a VC can increase both his/her own payoff from a given contract at the expense of an underweighting entrepreneur, and raise the likelihood that the entrepreneur accepts the offer. Conversely, offering a contract with DP to an overweighting entrepreneur may lead to rejection of a contract that is in everyone's interest to accept. Conversations with VCs indicate that contract negotiations frequently break down even when the VC believes there is positive surplus for both parties. This points to the economic costs introduced by asymmetries in the comprehension of the VC and entrepreneur of the consequences of contract terms, and to the real economic value that lawyers with VC expertise can provide.

We derive the conditions that need to hold for a VC to introduce unnecessary DP. Consistent with our empirical results pertaining to legal counsel expertise and VC contract design, the model predicts that VCs propose unnecessary DP only to novice entrepreneurs whose legal counsel lacks VC expertise. The model also demonstrates how the quality of an entrepreneur's company affects a VC's trade-off for proposing contracts with more DP. We show that when the expected surplus at stake from successfully contracting is not too high, a VC may have an incentive to design a contract with added DP that only appeals to entrepreneurs who underweigh the payoff consequences of DP. Indeed, a VC may be able to extract profits from a project that has a negative net present value.

However, when the potential surplus is higher, the opportunity cost to a VC of having an entrepreneur who overweighs the payoff consequences of DP reject an offer is higher. As a result, a VC may endogenously include less DP for more promising projects. For especially valuable projects, the opportunity costs of failed negotiations can be such that when extra DP is necessary, VCs want to deal with entrepreneurs who understand the payoff consequences. That is, sometimes unusual levels of DP are necessary to alleviate moral hazard, but to design a contract with the requisite protection that would appeal to most overweighting entrepreneurs, a VC would have to cede a large equity stake.

In this situation, a VC would like to credibly inform the entrepreneur about the payoff implications of DP in order to increase the likelihood that he/she accepts the contract. Credibly recommending legal counsel with expertise in VC contracting to entrepreneurs is one way to do this. In this way, differences in information about DP emerge endogenously. That is, we provide a theoretical rationale for why high-quality companies backed by inexperienced entrepreneurs should be more likely to be represented by law firms with VC expertise in their negotiations (as opposed to being represented by higher-quality law firms that lack VC-specific expertise).

We then test the predictions of the model. We analyze 908 U.S. VC contracts. We first document the wide variety of complicated cash flow contingencies in these contracts. (2) We then uncover several novel findings relating the expertise of the entrepreneur's legal counsel to contract design. We first show that entrepreneurs represented by law firms with more VC-specific expertise negotiate contracts with fewer VC-friendly cash flow contingencies. This result is robust and survives a large battery of controls. Importantly, we find no difference in DP for entrepreneurs advised by better (rather than more expert) law firms. Those "better," but nonexpert, law firms have more general expertise, as measured by law firm revenues, number of lawyers, gross profit margin, revenue per lawyer, or entry salary for new associates, and charge more for their services. Such law firms have good negotiation skills, but limited understanding of the cash flow contingencies in VC contracts. This indicates that (a) our finding that inexperienced entrepreneurs with inexpert lawyers negotiate contracts with more VC-friendly terms is not due to expert law firms being more skilled at negotiations; and (b) the correlation with lawyer VC expertise rather than lawyer quality strongly suggests rent extraction from unknowing entrepreneurs rather than the outcome of efficient contracting.

An analysis of each separate contingency reveals that a lawyer's expertise only matters for those terms that have particularly complicated payoff consequences. Concretely, we find no association between lawyer expertise and the inclusion of liquidation preference, which is a relatively straightforward contract term. This emphasizes that it is the entrepreneur's limited understanding of contract terms that drives our results. In particular, alternative rationales would have to account for why an expert lawyer matters only for complicated forms of DP.

We further find that the VC-specific expertise of the law firm matters less for entrepreneurs who previously negotiated VC contracts. For these experienced entrepreneurs, their law firm's VC expertise has no impact on the extent of DP. This suggests that learning-by-doing can substitute for the VC expertise of legal counsel: entrepreneurs can come to understand the payoff implications of DP via their legal counsel or via their own past experiences. (3)

We then document that when negotiating with inexperienced entrepreneurs, VCs who invest in high-quality companies are more likely to negotiate with entrepreneurs who employ lawyers with VC expertise. In sharp contrast, we find no evidence of assortive matching between VCs who invest in high-quality companies and entrepreneurs who employ lawyers with more general expertise. Because lawyers with more general expertise charge more for their services, this finding is inconsistent with the competing rationale for matching that high-quality entrepreneurs exhibit stronger willingness to pay for expert counsel. Rather, as predicted by our model, the assortive matching arises because VCs want to help match high-quality entrepreneurs with expert counsels to avoid negotiation breakdowns with otherwise-uninformed entrepreneurs. Our finding that DP is virtually unrelated to lawyer general expertise is evidence that matching is not based on willingness to pay or on complementarities in inputs. This should not be surprising: overall legal fees represent a small transaction cost for almost any entrepreneur. Interviews with lawyers and VCs also contradict the premise that willingness to pay is an important real-world consideration.

Our paper contributes to the theoretical work that investigates the limitations under which financial contracts are written. In particular, our paper adds to the incomplete contracting literature that investigates the influence of cognition costs on contract design (Anderlini and Felli 1994, 1999, 2004; Dye 1985; Tirole 2009). We also contribute to the literatures that study how intermediaries--such as banks, VCs, and initial public offering (IPO) underwriters--overcome informational frictions. Our paper is one of the few to provide theoretical and empirical evidence on how intermediaries mitigate information asymmetries related to the contract design, and adds to the small set of legal studies that have large samples and examine the economic role of lawyers (Ashenfelter, Bloom, and Dahl 2013; Block and Stiebler 1987; Coates 2001; Harcourt 2000; Vronsky 2012). Finally, we contribute directly to the growing literature that studies the determinants of VC contracts (Bengtsson and Ravid 2009; Bengtsson and Sensoy 2011; Bottazzi, Hellmann, and Da Rin 2009; Cornelli and Yosha 2003; Gompers 1998; Kaplan, Martel, and Stromberg 2007; Kaplan and Stromberg 2003, 2004; Lemer and Schoar 2005; Sahlman 1990; Suchman 1994). While previous studies have investigated how company characteristics, VC characteristics, or institutional factors affect contract design, our paper shows how contracts are influenced by the expertise of an entrepreneur's legal counsel.

The paper is organized as follows. Section II describes the design of VC contracts, discusses reasons why entrepreneurs may find it difficult to understand their implications, and explains the roles of legal counsels. Section III presents our theoretical model. Sections IV and V describe the data and present our empirical findings. Section VI concludes. II.

II. BACKGROUND CHARACTERISTICS

A. Design of VC Contracts

VC contracts are relational agreements that include a myriad of complicated contractual terms that determine the allocation of control and cash flow rights in different states of the world. Suchman (1994) finds that contracts have evolved toward more standardized templates, so that contracts differ largely in which terms are included, rather than in how specific terms are structured. Sahlman (1990), Gompers (1998), and Kaplan and Stromberg (2003) characterize the typical VC contract template. VCs typically receive convertible preferred stock with attached cash flow contingencies such as liquidation preferences, cumulative dividends, redemption rights, and antidilution protection. Cumulative dividends and liquidation preference jointly specify the fixed amount that VCs have the right to receive before distributions to other shareholders. Participation rights, which are attached to about half of contracts that use preferred equity, increase VC payoffs by giving the VC both a repayment of invested capital and a fraction of the remaining common equity (double dipping). Redemption rights give VCs the right to sell shares back to the company following a predetermined schedule. Antidilution protection gives VCs rights to additional shares if the company secures follow-up investment when the company valuation has declined (down round). Contracts may also include pay-to-play provisions, according to which VCs retain certain contractual rights only if they invest in follow-up financing rounds.

B. Complexity of VC Contracts

The joint payoff implications of different cash flow contingencies in VC contracts hinge sensitively on their finely interrelated details. To illustrate, consider a VC investment of $5 million in which the VC holds a 30% equity stake. If the company sold for $20 million after 3 years, the VC would receive $6 million if the contract gave VCs "plain" convertible preferred stock. The VC's payoff would be $13.9 million, or 70% of the selling price, if the contract gave the VC preferred stock with participation rights, a 2X liquidation preference and 8% compounding cumulative dividends: these additional cash flow contingencies raise the VC's payoff by $8 million. If, however, the VC instead received a 60% equity stake, this difference would shrink to $4.5 million ($12 million for convertible preferred versus $16.5 million for participating preferred). Moreover, the VC only receives payoffs from participation rights if the selling price is below the automatic conversion enterprise value. This example illustrates one way in which the payoff implications of DP are interlinked in truly complex ways.

C. Entrepreneur's Understanding

Inexperienced entrepreneurs face steep hurdles when trying to understand contract implications. An entrepreneur must first decipher the financial meanings of cash flow contingencies such as "narrow-based weighted-average anti-dilution." Although there are books and web-pages devoted to explaining how VC contracts work, these sources provide imprecise and incomplete explanations that do not reveal the exact payoff consequences. These consequences are hard to compute because expected payoffs depend on the probability distribution over company outcomes, investment amounts and share prices in different rounds, type of preferred stock, cumulative dividends, and other deal features.

Entrepreneurs must also understand how contracts evolve. Most companies receive multiple rounds of financing, and terms are often renegotiated if new investors set this as a precondition (Bengtsson and Sensoy, forthcoming) or companies are sold (Broughman and Fried 2010).

Finally, an entrepreneur must forecast what should go into a contract, and what other VCs might offer. The broad variation in the structure of negotiated contracts (Bengtsson and Ravid 2009; Bengtsson and Sensoy 2011; Gompers 1998; Kaplan and Stromberg 2003, 2004) makes it harder for entrepreneurs to assess whether a proposed contract is reasonable.

D. VCs' Understanding

Unlike entrepreneurs, VCs are well-versed with the consequences of cash flow contingencies. Zarutskie (2010), and Bengtsson and Hsu (2010) find that most VC partners hold MBA degrees from top universities and that many have prior work experience in finance. VC partners have seen how different cash flow contingencies affect realized payoffs at company exit events: the average VC in our sample makes a dozen investments per year, also participates in failed negotiations, and has about two decades of experience.

E. How Legal Counsels Help Entrepreneurs

As we formally model, the inability of some entrepreneurs to understand the payoff consequences of DP may entice VCs to propose excessive investor-friendly cash flow contingencies (i.e., unnecessary DP). An entrepreneur would not trust a VC's advice on contract terms because VCs have incentives to mislead. This opens up a role for external advisers--legal counsels with expertise in VC contracts. Such legal counsels can describe the exact meanings and consequences of cash flow contingencies. One lawyer describes this process: "Invariably, there are provisions ... that the entrepreneur either did not understand completely or simply did not devote enough attention to. As the cliche goes, the devil is always in the details and the reality is that this is precisely where and when a good attorney will add value." (Aspatore Books Staff 2008, 114).

A legal counsel can also give his/her opinion about what represents a reasonable "market" contract. The counsel can convey which contingencies are non-standard and are particularly favorable to a VC. Even if the counsel does not know the exact payoff consequences, but only whether a contingency is necessary, he/she can shape an entrepreneur's assessment. In turn, this can discourage a VC from trying to extract rents via unnecessary contingencies. One lawyer explains this; "Simply put, if someone on the other side of the table argues that a certain deal term we are asking for is not market or standard and I disagree, I am going to call them on it ... my experience provides some currency in the negotiation process" (Aspatore Books Staff 2008, 120). Finally, counsels who have observed the evolution and renegotiation of contracts can convey this knowledge to clients.

F. Importance of Counsel Expertise

A legal counsel's role of informing an entrepreneur about the payoff implications of DP differs from the routine task of translating negotiated agreements into the required legal documents. A counsel can only provide this information if he/she has personal experience with the specific contracts used in VC investments, or if he/she works at a law firm with that accumulated institutional knowledge.

It is not always easy for an uninformed entrepreneur to identify (a) a counsel's expertise and (b) the importance of that expertise for negotiating contracts. Concretely, an entrepreneur may know that a well-trained lawyer, for example, one from a top legal firm, may help translate terms into legal definitions that will hold in court. However, an entrepreneur may not recognize that a lawyer with VC-specific expertise may help him/her evaluate the payoff consequences of negotiable terms. As a result, some uninformed entrepreneurs select inexpert lawyers.

III. THE MODEL

In this section, we model the strategic calculus of a VC negotiating contracts with entrepreneurs, some of whom do not fully understand the implications of contractual DP. We derive how entrepreneur's characteristics--the quality of his/her project, his/her experience, and the expertise of his/her legal counsel--affects the VC's proposed DP.

In the model, entrepreneurs employ legal counsels who help them interpret and negotiate cash flow contingencies. Legal counsels with VC expertise correctly interpret and value DP clauses. Such clauses are sometimes necessary to mitigate moral hazard. However, while equity shares are easily understood and valued by all parties, DP is not. Uninformed, inexperienced entrepreneurs must rely on their legal counsels to value DP. Expert counsels always value DP correctly, while inexpert ones do not.

Formally, our model distinguishes three types of inexperienced entrepreneurs with inexpert legal counsels: a type p entrepreneur properly values DP, a type u entrepreneur under-weighs the payoff consequences of DP, and a type o entrepreneur overweighs the payoff consequences. We determine how the presence of these three types of entrepreneurs influences the DP the VC offers, and how the ex ante expected quality of an entrepreneur's project affects the VC's decisions.

We consider an entrepreneur with a project that requires one unit of external venture capital finance to go forward. The opportunity cost of the capital to the VC is R. The project is indexed by its ex ante expected quality, q, and possible moral hazard, [zeta] [member of] {0,1}. Here, [zeta] = 1 indicates the presence of a moral hazard issue that must be addressed, and [zeta] = 0 indicates that there is no moral hazard. Let d [member of] {0,1} index whether the contract features DP that mitigates the moral hazard problem. Thus, d = 1 indicates that the contract includes DP, and d = 0 indicates that the contract is a pure equity contract. If funded, the project pays [pi] = qg[([zeta], d).sub.[epsilon]], where [epsilon] is distributed according to a continuous density f with connected support, where [epsilon] [greater than or equal to] 0 and E[[epsilon]] = 1. The VC knows q and [zeta]. For simplicity, we assume that g(0,0) = 1 (a normalization), g(1,0) = 0 and g(1,1) = g(0,1) = 1 - [delta], where [delta] is positive, but small. These payoffs indicate that "uncontrolled" moral hazard destroys project payoffs, but that moral hazard can largely be mitigated by DP. We focus on the case where [delta] [right arrow] 0, so that DP only slightly distorts actions in other ways. Hence, introducing unnecessary DP--choosing d = 1 when [zeta] = 0--has only small costs. Finally, when [zeta] = 1, the potential consequences of moral hazard are severe enough that it is optimal to introduce DP. Thus, DP is necessary when [zeta] = 1, but unnecessary when [zeta] = 0.

A. Interpretation

In practice, VCs always have some moral hazard concerns. Indeed, our data reveal that contracts always include some DP. For example, contracts always have a convertible equity component with at least 1X liquidation preference. In terms of our model, one should interpret d = 0 as a contract that includes "standard" DP, and d = 1 as a contract that includes "extra" DP (e.g., 2X liquidation preference), that is only sometimes required to mitigate "extra" moral hazard. The other structure imposed just ensures that (1) when there is an "extra" moral hazard problem, it is optimal for a VC to add DP and (2) "unnecessary" DP has some cost, so that it is not socially optimal to include.

If the entrepreneur rejects the VC's offer, he/she receives an outside alternative that pays qz. We assume that z > 0 is distributed according to the cdf H(*) with associated density h(*) on (0,[infinity]). That is, the entrepreneur's outside alternative depends on his/her "ability" or project quality, reflecting either that the entrepreneur may find a second source of funding, or that a more able entrepreneur who can create better projects, is generally more productive, with higher opportunity costs. The realization of the value of the entrepreneur's outside alternative, z, is private information to him/her.

A contract is a pair {s,d}, where 5 is the share of project payoffs that the VC receives when the project payoffs, [pi], are sufficiently high. When d = 0, the VC just receives his/her equity share s of project payoffs. When d= 1, the VC's share is [rho]([pi]|s, d) = max {[??]([pi]),s}, where [rho]{[pi]|s,d)[pi] is non-decreasing in [pi], and [??]([pi]) is a weakly decreasing continuous function of [pi], strictly decreasing and differentiable once profits exceed some minimum level [[pi].bar]. That is, DP gives the VC a higher share when project payoffs are lower, but it does not drop off so quickly that the VC prefers lower realizations of profits to higher ones. Hence, the interests of the VC and entrepreneur are aligned in that they both prefer higher profit realizations. The entrepreneur receives the residual share of project payoffs, 1 - [rho]([pi]|s, d). Without loss of generality, we focus on entrepreneurs of sufficient quality, q, that the VC can identify a contract that provides him/her an expected payoff that covers his/her capital investment. We also assume that the expected payoffs of DP do not cover the VC's investment--the VC will always choose s > 0.

Entrepreneurs are either informed or not. Entrepreneurs who have past experience or who employ expert legal counsels understand the payoff consequences of DP. In contrast, inexperienced entrepreneurs who employ inexpert counsels may not. In this way, the knowledge of an expert counsel can substitute for an entrepreneur's lack of experience. In essence, we assume that an entrepreneur may have trouble in distinguishing a counsel who is an expert in VC contracting from one who is not. The VC can distinguish whether the entrepreneur is experienced, and whether his/her legal counsel is an expert. Without loss of generality, we focus on informed and uninformed entrepreneurs, and ignore the reasons underlying why the entrepreneur is informed. We let i [member of] {0,1} index whether the entrepreneur is informed. Here, i = 1 indicates that the entrepreneur is informed, and i = 0 indicates that he/she is uninformed.

We contemplate three types of uninformed entrepreneurs: those who properly weight the consequences of DP, believing [[??].sub.p]([pi]) = [??]([pi]); those who optimistically underweigh the consequences, [[??].sub.u]([pi]) [less than or equal to] [??]([pi]); and those who pessimistically overweigh the consequences with [[??].sub.0]([pi]) [greater than or equal to] [??]([pi]). These inequalities can be weak for [pi] [less than or equal to] [[pi].bar], but are strict for [pi] > [[pi].bar]. We index these types of entrepreneurs by t [member of] {p,u,o}, with associated probabilities of [p.sub.p], [p.sub.u], and [p.sub.o]. Consistent with DP, we assume that [[??].sub.t]([pi]) is weakly decreasing in [pi], differentiable for [pi] > [[pi].bar].

The timing of the game is as follows: given q, [zeta], and i, the VC offers a contract {s,d}. Then, given {s,d}, q, [zeta], and the realization of his/her outside alternative, z, the type t entrepreneur chooses whether to accept (a = 1) or reject (a = 0) the contract. If accepted, the project pays off, and the contract determines payments to the VC and entrepreneur. If rejected, the VC receives zero, and the entrepreneur receives his/her outside alternative, qz.

We will also contemplate the possibility that given q, [zeta], and i = 0, the VC can introduce the entrepreneur to expert legal counsel. This captures the idea that while an entrepreneur may not trust a VC's assertions about DP, an entrepreneur would accept the assessment of a legal counsel, who has no direct conflict of interest, and who has reputational capital at stake. In this way, we endogenize the information status of the entrepreneur. This allows us to make sense of the observation that some VCs dislike the uncertainty created when entrepreneurs are represented by inexpert legal counsels, and will try to guide them to expert counsel.

B. Analysis

Optimization by a type t entrepreneur with a project of quality q implies that he/she accepts contract {s,0} if and only if [[integral].sub.[epsilon]](1-s) q[member of]f([epsilon])d[epsilon] [greater than or equal to] qz; and that he/she accepts contract {s,1} if and only if [[integral].sub.[epsilon]](1-[[rho].sub.t](q[epsilon]))(1 - [delta])q[epsilon]f([epsilon]) d[epsilon][greater than or equal to]qz. Let [[alpha].sub.t](s,d;q) be the probability of acceptance:

[[alpha].sub.t] (s, 0; q) = H ([[integral].sub.[epsilon] (1 - s) [epsilon]f ([epsilon]) d[epsilon])

and [[alpha].sub.t] (s, 1; q)

= H([[integral].sub.[epsilon]] (1 - [[rho].sub.t](g[epsilon])) (1-[delta])[epsilon]f([epsilon])d[epsilon]).

Because all entrepreneurs correctly evaluate the payoff consequences of pure equity contracts, each type of entrepreneur is equally likely to accept a contract that does not have DP. Moreover, the entrepreneur's quality does not affect the probability that he/she accepts a given equity contract, as quality affects equally both project payoffs and the value of his/her outside alternative. In sharp contrast, for a contract with DP and equity share s, the likelihood that an entrepreneur accepts rises with his/her quality q. That is, the marginal impact of increasing q on the probability of acceptance by a type t entrepreneur is

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where [[pi].sub.t](s) is the highest profit realization for which a type t entrepreneur believes DP is relevant, that is, [[??].sub.t]([[pi].sub.t](s)) = s. Intuitively, DP affects high-quality entrepreneurs by less because they are less likely to have low project payoffs. (4)

The VC's expected payoff from a contract without DP (when DP is not necessary) is then

[[alpha].sub.p] (s, 0; q) [sq - R].

The VC's expected payoff from a contract with DP is

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

We begin the analysis with the observation that the VC offers a contract with DP to an informed entrepreneur if and only if it is required to mitigate moral hazard.

PROPOSITION 1. VCs offer contracts with d=1 to informed entrepreneurs if and only if [zeta]=1

Intuitively, unnecessary DP destroys surplus, and since the entrepreneur is informed, the VC would be the party that incurs the costs associated with reduced surplus.

When the social costs of unnecessary DP are small, that is, when [delta] [right arrow] 0 we can provide necessary and sufficient conditions for the VC to offer unnecessary DP to an uninformed entrepreneur with a quality q project. In the notation below, [s.sup.*.sub.1] is the optimal equity share chosen by the VC when he/she offers DP.

PROPOSITION 2. Suppose that i = 0 and [zeta] = 0. Then as [delta] [right arrow] 0, the VC offers unnecessary DP, d = 1, if and only if

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

The VC loses when he/she offers DP to entrepreneurs who overweigh the payoff consequences of DP, but gains when he/she offers it to those who optimistically underweigh the payoff consequences. The result says that for unnecessary DP to be optimal, the proportion of uninformed entrepreneurs who optimistically underweigh the adverse payoff consequences of DP must be sufficiently high relative to the fraction of entrepreneurs who pessimistically overweigh the consequences, and hence are less likely to accept a contract with DP.

The flip side of Proposition 2 is that sometimes there will be moral hazard that must be addressed by DP for the project to have value. Proposition 2 then provides necessary and sufficient conditions for a VC to have an incentive to introduce an uninformed entrepreneur to expert legal counsel who can correctly evaluate the DP:

COROLLARY 1. Suppose that DP is necessary, that is, [zeta] = 1. Then as [delta] [right arrow] 0, the VC's expected payoff is higher if the entrepreneur is informed than uninformed if and only if inequality (1) is reversed, that is, if and only if

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

When the outside opportunity z is uniformly distributed, we can relate differences in the likelihood of acceptance directly to the differences in beliefs about the payoff consequences of DP as

(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

and

(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

We now show that for plausible parameterizations, the right-hand side of equation (3) exceeds that of (2). That is, DP reduces the willingness of pessimistic overweighters to accept a contract by more than it raises the willingness of optimistic underweighters, that is,

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

In particular, when the outside alternative is uniformly distributed, and the extent of overweighting equals the extent of underweighting for a given profit level (the difference can vary with [pi]), that is,

(4) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

But then when this is so, the condition in Equation (1) of Proposition 2 for the VC to offer unnecessary DP only holds when [p.sub.u] > [p.sub.o]: for the VC to want to offer unnecessary DP there must be significantly more optimistic type u entrepreneurs who underweigh the adverse payoff consequences of DP than pessimistic type o entrepreneurs who overweigh the consequences. The following is immediate.

PROPOSITION 3. Suppose the outside alternative z is uniformly distributed and that [delta] [right arrow] 0. Then there exists a [mu] > 0 such that if

E0 [[??].sub.o] ([pi]) - [[rho].sub.p] ([pi]) [greater than or equal to] [[??].sub.p] ([pi]) - [[??].sub.u] ([pi]) - [mu], [for all][pi],

the VC only offers unnecessary DP when more entrepreneurs optimistically underweigh their payoff consequences than overweigh them, that is, [p.sub.u] > [p.sub.o]

Intuitively, the perceived impact of DP persists for more and higher values of profits for a pessimistic overweighter, with a corresponding greater impact on the probability of acceptance. Hence, not pricing DP correctly has a bigger (negative) impact on the probability of acceptance by a pessimistic overweighter than the positive impact on the probability of acceptance by an optimistic underweighter. Consequently, there must be substantially more underweighting entrepreneurs than overweighting ones for a VC to want to introduce unnecessary DP.

A similar result would obtain if entrepreneurs, rather than being uncertain about the payoff consequences of DP, were overly optimistic or overly pessimistic about the probability of low payoff states in which DP bites, with optimism predominating. Puri and Robinson (2007) find that many entrepreneurs are overly optimistic about their companies.

Rewriting Equation (1), it is optimal for a VC to offer unnecessary DP if and only if

(5) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

The attractiveness of unnecessary DP falls with quality as long as equity terms [s.sub.1] do not decline too sharply with q. To see this, observe that if we augment condition E0 with [[??]'.sub.t] = [??]', for t = o, p, u, then we can describe when the incentives to offer unnecessary DP fall as the ex ante quality of the entrepreneur's project rises. Differentiating the left-hand side of Equation (5) with respect to q, and observing that the Leibnitz terms drop out, reveals that this incentive falls with q as long as the equity share that the VC demands is not too sensitive to q and [p.sub.u] >> [p.sub.0]:

(6) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

This expression reveals that some minimal structure is required for the optimality of unnecessary DP to fall with quality. In particular, with no DP, we have s' < 0; the VC demands smaller equity shares from better entrepreneurs because he/she is more concerned that better entrepreneurs will reject an offer (the opportunity cost of rejection is higher). Hence, if the payoff consequences of DP are arbitrarily small, then [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] is positive.

However, in the more plausible setting in which DP has significant payoff consequences, that is, in which [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] is sufficiently negative, the attractiveness of unnecessary DP falls with quality as long as there are enough underweighters of DP relative to overweighters. We believe that this scenario in which DP has significant payoff consequences is the relevant one; and in this case, Equation (6) becomes negative, implying that the attractiveness of unnecessary DP falls with quality.

We now provide an intuitive argument for why a VC's desire to offer unnecessary DP falls with the ex ante quality of the entrepreneur's project, one that hinges only on an elasticity-based comparison. As such, it is robust to concerns that DP takes many different forms and, in practice, explicit functional forms are unclear. Using prime to denote the derivative with respect to q, consider the following elasticity condition on acceptance probability differences:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Condition E1 says that an increase in q has a greater percentage increase on the acceptance probability difference for underweighters than for overweighters. Under the conditions in Proposition 3, the denominators in this elasticity comparison are ordered,

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

so E1 holds as long the derivative differences in the numerators are close.

PROPOSITION 4. Suppose that condition E1 holds and z is uniformly distributed. Then for [delta] [right arrow] 0 if the VC offers unnecessary DP to some entrepreneur qualities but not others, his/her decisions reflect a cutoff rule. That is, there exists a [q.sup.*] such that a VC optimally offers unnecessary DP to entrepreneurs of quality q < [q.sup.*], but not to higher-quality entrepreneurs with q > [q.sup.*]. Further, for q < [q.sup.*], the VC prefers to deal with inexperienced entrepreneurs who are represented by inexpert legal counsels, whereas the preference is reversed for q > [q.sup.*].

C. Other Role of Expert Legal Counsel

Finally, we argue that the qualitative result that VCs will not offer unnecessary DP to entrepreneurs represented by expert legal counsels extends when counsels only know whether the DP is unnecessary, but do not know the exact payoff implications. As long as unnecessary DP introduces any distortions in behavior, the legal counsel and entrepreneur can understand that if it is offered, the VC must be gaining at their expense (total surplus is reduced, and VC optimization implies that he/she would only introduce unnecessary DP if he/she gained). As a result, the legal counsel should guide the entrepreneur to hold pessimistic views about the payoff consequences; and a VC does not want to offer unnecessary DP to pessimistic entrepreneurs.

D. Empirical Predictions

We now describe three central predictions of our model that we test. The first prediction follows from Propositions 1-3: entrepreneurs agree to less DP if they have expert legal counsel. The second prediction is that the negative association between DP and legal counsel expertise should be less pronounced for entrepreneurs with more experience at VC contract negotiations because those entrepreneurs learn about the payoff consequences via past experience. The third prediction is that for inexperienced entrepreneurs, higher-quality companies should match with expert legal counsels. This positive assortive matching follows because as entrepreneur quality rises, the opportunity cost that a VC incurs when an overweighting entrepreneur rejects a proposed contract rises. As a result, when project quality is high enough, the VC wants to introduce the entrepreneur to expert counsel.

IV. DATA

A. Sample Construction

Data on VC contracts are difficult to obtain. To address the questions posed in our paper we collaborated with VCExperts, a company that analyzes VC data, and use mandatory legal filings to collect a large and representative sample of VC contracts. A limitation of the data is that it lacks comprehensive information on control rights. Although cost and time considerations preclude extracting legal filings from all recent venture-backed companies, our sample is random in the sense that we do not systematically extract contracts for certain types of entrepreneurs, companies, VCs, or legal counsels. Our final sample includes contracts from 908 unique companies. The sample size is far larger than those considered by other researchers (Broughman and Fried 2010; Gompers 1998; Kaplan and Stromberg 2003, 2004; Suchman 1994). (5)

We match each contract with its corresponding financing round in Venture Economics to obtain variables that capture company, VC, and round characteristics. We complement this data with hand-collected information on backgrounds of company founders. As the summary statistics in Table 1 illustrate, the sample is a good representation of the VC industry, with about half of all companies being headquartered in California and two-thirds operating in either the life science or high-technology industries. One company in four has a founder who previously started a venture-backed company, and 7% have a founder who previously started a venture-backed company that realized an IPO. The average company is 4 years old, has raised 1.8 prior rounds of financing and receives $12.3 million in the contract round. About half of the lead VCs are headquartered in California, and the average number of VCs in the round is 4.3. The average lead VC has 21 years of experience and has invested in 237 portfolio companies prior to the contract negotiation: most VCs have had ample experience at contract negotiations and have observed the ultimate payoff implications of many combinations of cash flow contingencies.

B. Contractual Cash Flow Contingencies

We code each contract along six important cash flow contingencies: cumulative dividends, liquidation preference, participation, antidilution rights, redemption, and pay-to-play. These six contingencies jointly determine the extent of the DP: these contingencies are especially favorable to VCs when company performance is bad (see Kaplan and Stromberg 2003). We use the coding scheme in Bengtsson and Sensoy (2011): we code redemption rights as 0 or 1 and the other cash flow contingencies as 0, 1, or 2 based on how much DP they provide a VC. Table 2 outlines the coding scheme and presents statistics on the frequency of each contingency. A higher-value means that a contract provides more DP. We then aggregate the six cash flow contingency variables into a downside protection index (DPI). DPI can take any value between 0 and 11, where 0 is a contract that has a minimum of VC-friendly contingencies (i.e., the weakest DP) and 11 is a contract that includes all possible contingencies (i.e., the strongest DP). The average DPI is 4.64 and the standard deviation is 1.54.

This simple summation index has the virtue of being more transparent than other aggregations. One can contemplate two concerns with it. First, summing cash flow contingencies may not account for the possibility that the relative weighting of contingencies depends on many company- and contract-specific factors, including the probability distribution over outcomes. Obviously, the data do not exist to account for such factors. Importantly, our findings also hold when we disaggregate and look at individual cash flow contingencies. Second, one might believe that pay-to-play should not be incorporated into the index because it primarily governs relationships between VCs investing in different rounds. We include pay-to-play in DPI because it determines when VCs retain other contingencies (e.g., anti-dilution and cumulative dividends) that are important for DP. Our robustness analysis shows that excluding pay-to-play from DPI does not alter the relationship between DP and legal counsel expertise. We also show that this relationship is robust to excluding redemption rights from DPI.

C. Legal Counsel Expertise

We identify the law firm or self-employed lawyer who represents each sample company using company web-pages, law firm web-pages, data from VCExperts, and legal filings. The sample only includes companies for which we can identify an outside legal counsel. Because large established law firms are more likely to publish data on the companies they represent than self-employed lawyers, our sample is tilted toward venture-backed companies counseled by large law firms. This, however, is unlikely to affect our cross-sectional analysis of how contract design varies with a lawyer's expertise.

Because we cannot identify the names of individual lawyers, we conduct our analysis at the law firm level. The data on the expertise and reputation of legal counsel are supplemented with hand-coded law firm characteristics.

We use four different proxies to measure a counsel's expertise on VC transactions. The first proxy, and in our view the best, is the number of companies in the sample that are advised by the legal counsel. (6) The two most represented law firms are Wilson, Sonsini, Goodrich, and Rosati, and Cooley, Godward, Kronish LLP, which together counsel almost a quarter of all sample companies. Our second proxy for VC expertise is whether the counsel is one of the 19 law firms ranked as top counsel for U.S. VC transactions by two publishers of law firm intelligence: Chambers and Partners and Legal 500. Most companies in the sample (63%) are represented by top-ranked counsel. The third proxy is whether the company has a branch office in or around Silicon Valley (51% of all companies), and the fourth proxy is whether the company has a self-reported practice area specialized on either VC or emerging innovative companies (87% of all companies). The data on law firms are collected in 2007, save for our first proxy which we compute using the full sample period. The final sample includes 189 unique legal counsels. We find that all four proxies for lawyer expertise are positively correlated with each other, confirming that they capture a similar dimension of expertise.

While our focus is on a counsel's specialized expertise on VC investments, we also collect data on a counsel's general expertise. This allows us to compare contracts associated with counsels who are high-quality lawyers (in general terms) but lack VC expertise. Our expectation is that such counsels should not be associated with less DP. We use several general expertise measures to rule out other possible explanations for our results. The ALM Research database provides information on law firm revenues, number of lawyers, gross profit margin, revenue per lawyer, and entry salary for new associates. Because ALM Research only tracks the 200 largest U.S. law firms, ranked by revenues, we can only measure the legal counsel's expertise in detail for 76% of the companies in the sample. Although three-quarters of all venture-backed companies use a top 200 law firm, only 5% are represented by a top 10 firm, and 17% by a top 25 firm.

We do not have data on which law firm represents a VC in our sample. However, as discussed earlier, VCs have extensive experience with the payoff consequences of negotiable terms, so the expertise of their legal counsel should not affect their choices of contract terms.

V. EMPIRICAL RESULTS

A. Univariate Analysis

We begin with a univariate comparison of how a contract's DP relates to the expertise of the company's counsel. Table 3 reports the mean value of DP for subsamples formed on counsel expertise. Differences in DP between subsamples are tested with a Wilcoxon rank-sum test.

Consistent with our prediction that expert legal counsels should reduce DP for inexperienced entrepreneurs, Panel A shows that average DPI is lower for counsels with more VC expertise. This impact is statistically significant and economically meaningful for each proxy used to measure counsel VC expertise. To illustrate, DPI is 4.36 for companies represented by the "most expert" legal counsel based on a combination of our four proxies (i.e., a counsel who advises more than the sample median number of companies, is top ranked, and has both a VC practice area and a branch office in Silicon Valley). DPI is 5.17 for companies represented by the "least expert" legal counsel (who advise fewer than the sample median number of companies, is not top ranked, and has neither a VC practice area nor a branch office in Silicon Valley). The difference, 0.81 DPI units, represents about one-half of the standard deviation of DPI. To highlight that the association between counsel VC expertise and DPI does not depend on how we construct DPI, Panel B presents results for each cash flow contingency. For each contingency, contracts negotiated by the "most expert" legal counsel have less DP than those negotiated by the "least expert" counsel. These differences are statistically significant except for liquidation preference and pay-to-play. We posit that the coefficient on liquidation preference is insignificant because liquidation preference has relatively straightforward payoff consequences. As a result, most entrepreneurs understand its implications. That the impact of counsel expertise only matters for hard-to-understand DP is hard for alternative explanations to reconcile. A likely reason for why the coefficient on pay-to-play is insignificant is that this term affects the payoff split between VCs (in different rounds) and not the split between the entrepreneur and VCs. The association between counsel VC expertise and DPI does not hinge on our aggregation method. In particular, our qualitative findings are reinforced when pay-to-play is excluded.

Panel C compares DPI across subsamples formed using proxies for a counsel's general expertise/quality. In important contrast to our results on VC expertise, the level of general expertise is not associated with significantly less DPI. The sole exception is "Counsel General Entry Salary": law firms that pay new associates more have less DP. However, this difference ceases to be significant in multivariate tests that include a counsel's VC expertise and control variables. The fact that the legal counsel's general expertise does not correlate with contract design supports our thesis that it is the specific VC expertise of counsels that affect the inclusion of cash flow contingencies in a contract. That is, more general counsel knowledge about contracts, or higher general reputation or general expertise of a counsel does not improve contract terms.

B. Multivariate Analysis

To address the possibility that our univariate findings may be biased due to correlation of DPI with company or VC characteristics, we run multivariate ordinary least squares (OLS) regressions with DPI as the dependent variable. (7) Table 4 presents results of multivariate regressions that include different proxies for counsel VC expertise as independent variable. The regressors include as controls key variables such as company age, amount invested in round, number of VCs investing in round, whether the founder is a serial founder, whether the founder is a serial founder whose previous company went public in an IPO, and whether the founder is a serial founder whose previous company was acquired. We also control for VC experience (measured by its number of portfolio company investments), whether the VC firm and company are located in the same state, and whether the VC was organized as an independent private partnership (versus a corporate, financial, etc. VC). (8) The regressions also include as fixed effects: VC location (California, Massachusetts, Texas, New York, and other states), company location (state), company industry (Venture Economics 10-grouping classification), year of investment round, and round number. To adjust for potential cross-correlation within a counsel, all regressions cluster residuals by counsel. (9)

Specifications 1-4 and 5-8 are identical save that specifications 5-8 include fixed effect controls. The negative and significant coefficient estimates reveal that the univariate relationship between counsel VC expertise and DPI holds even after controlling for variables that have empirically been shown to influence the investor-friendliness of VC contracts. In unreported regressions, we find that DPI monotonically declines between quintiles of counsel VC expertise.

The coefficient estimates on the lawyer expertise proxies vary from 0.16 to 0.47 across the regression models in Table 4, which correspond to between 10% and 31% of one standard deviation of DPI. We also find a large increase in predictive power from using the lawyer expertise variables. For example, in the regression with our full battery of controls, the adjusted R-squared rises from 21.94% without the lawyer variable to 23.67% with the lawyer variable. This increase in R-squared far exceeds the 0.12% increase from including round amount and is of similar magnitude to the 1.99% increase from including VC experience, both of which are significant determinants of DPI. These comparisons indicate that our results are both economically and statistically significant.

Table 5 presents results of multivariate regressions that, in addition to the proxy for counsel VC expertise ("Counsel VC # of Deals"), include different proxies for counsel general expertise. These regressions have the same control variables and fixed effects as Table 4, but their coefficients are not reported for brevity. Confirming our finding from the univariate analysis, the coefficient on counsel VC expertise remains negative and statistically significant, while coefficients on all proxies for counsel general expertise are statistically insignificant.

C. Robustness

In unreported tests, we validate the robustness of the association between counsel VC expertise and DP. These regressions include "Counsel VC # of Deals," "Counsel General AmLaw Top 25," with the same controls and fixed effects as in Tables 4 and 5. We also find that the coefficient on our main proxy for counsel VC expertise "Counsel VC # of Deals" remains negative and significant after controlling for "Counsel VC Office in Silicon Valley." This shows that the effect we estimate for lawyer expertise is distinct from the regional effects documented by Bengtsson and Ravid (2009).

Another series of unreported tests included additional controls that capture omitted VC or company characteristics. The coefficient on "Counsel VC # of Deals" remains negative and significant when we include VC fixed effects, which absorbs any time-invariant VC characteristic. The results are also qualitatively unchanged when we include the pre-money valuation of the round (which is reported for about half of our sample). Although this variable absorbs much unobserved heterogeneity in company quality and risk, it is also intrinsically correlated with DPI, because terms and pricing are determined jointly during contract negotiations. For this reason, we do not include it in our reported specifications. Third, the results are qualitatively unchanged when we include dummies that capture the outcome of the company (IPO, Merger/Acquisition, Other). The outcome is a proxy for a company's ex ante quality, but it may also be endogenously correlated with DPI.

Our qualitative findings do not depend on how we aggregate individual cash flow contingencies to DPI. When we exclude pay-to-play or redemption rights, we find significantly less DP for legal counsels with more VC expertise. We also corroborate the univariate analyses for each separate cash flow contingency, running a probit regression for each contingency, recoded as a binary measure of DP. The impact of legal counsel expertise is unaffected save for the regression using redemption rights, where the coefficient remains negative, but not significant.

We also explore different subsamples based on the counsel's VC expertise. DPI decreases in counsel VC expertise even if we exclude the largest counsel, Wilson, Sonsini, Goodrich, and Rosati. The results also hold if we include either counsels above or counsels below the sample median level of counsel VC expertise. So, too, the results hold if we restrict the sample to large counsels (e.g., the largest 200 U.S. law firms) or to smaller counsels. Thus, the association between counsel VC expertise and downside protection is not due to the behavior of a few selected counsels.

We split the sample by geographical area and year of financing round. Within each subsample, DPI falls with counsel VC expertise, but not with counsel general expertise. Thus, the association between counsel VC expertise and DP holds broadly throughout the VC industry.

We split the sample based on the riskiness of the project. The idea behind this split is to test the expectation that lawyers matter more for risky projects, because the contracts for such projects have greater payoff consequences. We classify a project as risky if it is in the internet-specific, or computer software or services industries. Our analysis of historical outcomes, using Venture Economics data, shows that these segments are riskier in the sense that low fractions of funded companies achieve a successful IPO outcome. We find that the coefficient estimate on "Counsel VC # of Deals" is greater for risky than for safer projects. However, the difference is not statistically significant, probably because our proxy for risk is so imperfect.

We also split the sample based on whether the VC continued to finance the company or not (see Cumming and Dai 2013 for an analysis of VC switching). One might expect to find a stronger effect on lawyer expertise for the subsample of continued VC financing, as the VC could also benefit from unnecessary DP in future rounds due to cross-round inertia in VC contracting (Bengtsson and Sensoy, forthcoming). In contrast, discontinued financing (i.e., switching) means that the benefit of DP only comes from the current round contract, and any distortions in actions from unnecessary DP persist. Consistent with this expectation, we find that the coefficient on lawyer expertise is more negative for continued VC financing. However, the difference is not statistically significant.

D. Multivariate Analysis of Entrepreneur Experience

In our model, an experienced entrepreneur's own understanding of DP can substitute for legal counsel expertise. This suggests that the association between counsel VC expertise and DP should be less pronounced if an entrepreneur has past experience negotiating VC contracts. The entrepreneur has such experience if he/she raised VC financing before, either for his/her current company or for a venture-backed company that he/she, by virtue of being a serial entrepreneur, previously founded.

To investigate the substitution between counsel VC expertise and entrepreneur experience, we interact counsel VC expertise with the dummy "Follow-up Round," which equals 1 if the contract is from a follow-up round and is 0 if from a first round. Specifications 1-4 in Table 6 present results for this interaction variable, formed for each proxy for counsel VC expertise. Specifications 5-8 present results for the interaction between counsel VC expertise and "Serial Founder." All specifications include the same controls and fixed effects as Specification 1 of Table 5. The interaction variables are positive and generally significant: the negative association between counsel VC expertise and DPI is less pronounced for experienced entrepreneurs. Indeed, for experienced entrepreneurs, the VC expertise of counsel has no impact on the extent of DP: a counsel's VC expertise and the entrepreneur's own understanding are indeed substitutes. (10)

E. Multivariate Analysis of Counsel Matching

Our model also predicts endogenous matching between legal counsels and entrepreneurs. Specifically, for inexperienced entrepreneurs, higher-quality companies should match with legal counsels with more VC expertise. To investigate this, we run a multivariate regression with lawyer expertise as dependent variable and a measure of company quality as an independent variable. To do this, we must confront the challenge of obtaining a reliable measure of company quality prior to VC funding. Accounting or detailed performance data for private venture-backed firms do not exist. Further, company valuation should not be used because it is a negotiated term that is endogenously related to the other negotiated contract terms.

We use two proxies for company quality. The first proxy, "Round Amount," captures how much money the company raises from VCs in the round. Analyses of historical Venture Economics data show that companies that raise more aggregate financing are far more likely to have successful exits in the form of IPO or acquisition--better firms can raise more money. As Table 4 shows, "Round Amount" correlates negatively with DPI, consistent with this variable measuring company quality. The second set of proxies is the reputation of the lead VC, as measured by "(log) VC Experience," and "VC IPO Ratio" (we also use "(log) VC Age" and "VC Fund Sequence" in untabulated tests). Sorensen (2007) finds that companies financed by more reputable VCs are more likely to have successful exits, and he shows that this is partly due to their better pre-financing quality. (11) We extend his finding of positive assortive matching between company quality and VC reputation, by testing whether there is also positive assortive matching between counsel expertise and VC reputation.

Table 7 presents the results on matching. We control for company age, and include fixed effects for VC firm location, company location, company industry, year of investment round, and round number. All regressions cluster residuals by counsel. Specifications 1-3 are OLS regressions with "Counsel VC # of Deals" as the dependent variable. We document a strong positive relationship between legal counsel VC expertise and company quality. Qualitatively similar results obtain in specifications 4-12, which use our other proxies for legal counsel VC expertise as the dependent variable. Thus, the data are consistent with endogenous assortive matching because VCs want higher-quality entrepreneurs to be represented by knowledgeable counsel in order to reduce the likelihood that negotiations break down due to misunderstandings over payoff implications.

F. Competing Explanation of Counsel Matching

A possible competing explanation--distinct from our model--is that the results reflect operational complementarities between entrepreneur quality and counsel VC expertise. Better entrepreneurs may gain more from having better lawyers because they anticipate situations (other than contract negotiations with VCs) where high-quality legal help matters. For example, better entrepreneurs may need to use more complex patenting/licensing agreements, and their companies may be more likely to go public or be acquired. (12)

Our model suggests that it may not be necessary for experienced entrepreneurs to match with legal counsels with more VC expertise, as via their experience the entrepreneurs can come to learn the payoff consequences of VC contract terms. In contrast, operational complementarities would suggest that experienced entrepreneurs should be more likely to employ counsel with VC expertise, since experienced entrepreneurs are of higher quality than first-time entrepreneurs. Also, our model demonstrates that there should be no matching between higher-quality companies and counsels with more general expertise, because such lawyers do not inform entrepreneurs about the payoff consequences of cash flow contingencies. In contrast, operational complementarities would suggest assortive matching should emerge for measures of legal quality other than VC expertise, because better entrepreneurs can gain more from having better lawyers help them with various legal tasks.

We run a series of unreported regressions that test the competing theory of operational complementarities regressing the legal counsel's VC expertise on "Serial Founder," "Serial Founder with IPO," and "Serial Founder with Merger," all of which capture the entrepreneur's experience level. We control for company age, and include fixed effects for VC location, company location, industry, round year, and round number. In all specifications the relationship between the counsel's VC expertise and the entrepreneur's experience is insignificant. We also investigate whether there is matching between a legal counsel's general expertise and company quality, using the same methodology as in Table 7. Uniformly, we find no significant relationship between general expertise and company quality. We conclude that the pattern of matching is consistent with our model, but difficult to reconcile with theories based on operational complementarities.

We do not test why some entrepreneurs match with inexpert lawyers who cannot give them good advice on contract terms. However, our findings on contracts and matching do shed light on this question. We have shown that some entrepreneurs fail to understand the consequences of DP. This suggests that some entrepreneurs may fail to understand that the choice between an expert and inexpert lawyer can affect the negotiated contract. Moreover, identifying an expert lawyer need not be easy. The idiosyncratic nature of different entrepreneurial projects makes it difficult for an entrepreneur to evaluate whether a contract negotiated by a lawyer was a good one. Further, indirect indicators of a lawyer's expertise are imperfect. In particular, the quality of a law firm (and the price charged for legal advice) is minimally correlated with a lawyer's VC expertise, and a good, but inexpert, lawyer is not associated with reduced DP. That is, it is not enough for an inexperienced entrepreneur to identify a "good law firm."

Entrepreneurs may also choose lawyers for other reasons. DP does not matter if a company goes public because VC contracts include an "Automatic conversion" provision that cancels most protections given to a VC in this event. As a result, entrepreneurs who believe their companies are likely to have IPOs may choose inexpert lawyers because they care more about other factors (see Cumming 2008).

G. Other Observations

Our model predicts that DP decreases with company quality. Accordingly, it can provide a complementary explanation for the empirical findings of Gompers (1998) and Kaplan and Stromberg (2003, 2004) that VC contracts include less DP for high-quality companies. These papers posit that their findings reflect agency problems that vary with company characteristics. Our model shows that VCs may strategically choose not to include "extra" DP for higher-quality companies, because the cost of an overweighting entrepreneur rejecting the investment is higher. Further, VCs may strategically suggest expert lawyers to better entrepreneurs.

Finally, one prediction of the model, which cannot be tested with our data, is that entrepreneurs with higher-quality companies should be less likely to experience break-downs in contract negotiations. This prediction follows both because VCs endogenously are less likely for higher-quality entrepreneurs to include contract terms that the entrepreneurs do not understand, and because VCs optimally choose to introduce higher-quality entrepreneurs to expert legal counsels who can explain the payoff consequences of contract terms to them. VCs do this for better entrepreneurs in order to avoid incurring the high opportunity cost of rejection. In contrast, VCs do not make such introductions to entrepreneurs with lower-quality companies. This increases the likelihood that negotiations break down. Rather, VCs offer these entrepreneurs less attractive contracts with more DP that may appeal to those who underweigh the adverse payoff consequences, but are more likely to be rejected by less optimistic entrepreneurs. This mechanism adds a friction to the VC market that is conceptually distinct from agency and adverse selection problems.

VI. CONCLUSION

Our paper begins with the observation that not all contracting parties may understand the payoff implications of all financial contract terms. We develop and test a theoretical model of this in the venture capital setting, where contracts are particularly complicated. We show that the presence of entrepreneurs who may not fully understand DP can induce a VC to propose unnecessary DP. By doing so, the VC can expropriate an "optimistic" entrepreneur, who underweighs the consequences of DP, even though they may sometimes negotiate with a "pessimistic" entrepreneur, who overweighs those consequences, and hence would be excessively likely to reject a contract with DP. We explore this trade-off, and observe that this trade-off vanishes when the entrepreneur is represented by an expert legal counsel. That is, we predict that expert legal counsel should be associated with less DP, unless the entrepreneur is experienced and therefore need not rely on expert advice. We also predict that the cost to the VC of having a pessimistic entrepreneur reject a contract rises with the entrepreneur's quality. As a result, the VC offers better entrepreneurs contracts with less DP. Indeed, if the quality of the entrepreneur's company is high, we predict that the VC has an endogenous incentive to recommend expert counsel to the entrepreneur, giving rise to assortive matching between entrepreneurial quality and the VC expertise of their counsel.

We confirm these predictions for a novel sample of 908 VC contracts. Because these associations are derived based on the counsel's VC-specific expertise and not on the counsel's general expertise, we can rule out alternative explanations that are based on a lawyer's bargaining power or negotiation skills. One could think of ways other than counsel expertise to identify informed entrepreneurs. For example, entrepreneurs whose companies are older may have more knowledge about complicated contracts. However, older companies likely differ in many other ways that affect contracting: the conceptual advantage of our identification strategy is that expert counsels can clearly provide information about DP, but such expertise seems not to matter along other dimensions. By documenting how the use of DP varies systematically with an entrepreneur's legal adviser, and the assortive matching of entrepreneur quality along VC but not general expertise dimensions, we add to a small but growing literature on the determinants of VC contracts. These insights may be more broadly relevant in contracting settings that feature complicated contract terms.

The key feature of VC contracting that makes cognition costs matter is that each entrepreneurial project is idiosyncratic, with its own distinct set of issues and incentives. This means that the optimal contract will hinge sensitively on the project in ways that an inexperienced entrepreneur is unlikely to understand. This means that the optimal contract must be tailored to the particular circumstances, making it difficult to determine whether a contract is appropriate.

To see the importance of this feature, observe that in a franchising setting, an inexperienced franchisee may also not understand some incentives. But a potential franchisee does understand that each franchise is in a similar boat. Hence, to determine whether he/she is being exploited, a franchisee only has to verify whether he/she is receiving the same contract as others. As a result, a franchisee can draw the correct inferences about a contract, so a lack of understanding does not matter. Cognition issues would also not be relevant in repeated intermediary supply contracting between sophisticated firms. There, the initial stakes are likely low, and so the gain from exploitation is small, and repetition speeds learning. In contrast, cognition issues can affect contracting over strategic alliances or negotiating book and movie rights. Like entrepreneurship, these settings are complex, with idiosyncratic high stakes and substantial uncertainty about what should be done and about what payoffs will be, where, for example, Google may know far more than its "partner."

ABBREVIATIONS

DP: Downside Protection

DPI: Downside Protection Index

IPO: Initial Public Offering

OLS: Ordinary Least Squares

VC: Venture Capitalist

doi: 10.1111/ecin.12088

APPENDIX

Proof of Proposition 1

Suppose that [zeta] = 0. Note that for any contract {[s.sub.1], 1} (with DP, there exists a contract, {[s.sub.0], 0} without DP that is payoff equivalent for the VC given that the contracts are accepted by :he entrepreneur:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

implying that the probability the entrepreneur accepts contract {[s.sub.1], 1} is strictly less,

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Hence, if [zeta] = 0, and i = 1, the VC chooses d = 0. That d = 1 if [zeta] = 1 is immediate.

Proof of Proposition 2.

Consider the pure equity contract, {[s.sub.0], 0} that is payoff equivalent for the VC to {[s.sup.*.sub.1[, 1} given that it is accepted by the entrepreneur. As [delta] [right arrow] 0, note that [[alpha].sub.t] ([s.sup.*.sub.1], 1; q) [right arrow] [[alpha].sub.t] ([s.sub.0]; 0; q). The VC prefers the contract with DP only if

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

As [delta] [right arrow] 0, the left-hand side goes to [[summation].sub.t[member of]{u,p,o}][p.sub.t][[alpha].sub.t] ([s.sup.*.sub.1], 1; q) [[integral].sub.[epsilon]][s.sub.0]q[epsilon]f([epsilon])d[epsilon], implying that the VC prefers the contract with DP only if

[[summation over (t[member of]{u,p,o}] [p.sub.t][[alpha].sub.t] ([s.sup.*.sub.1], 1; q) > [[alpha].sub.p]{[s.sub.0], 0; q) = [[alpha].sub.p] ([s.sup.*.sub.1], 1;q).

Subtracting [[alpha].sub.p] ([s.sup.*.sub.1], 1; q) from both sides delivers the result. To see that the implication goes the other way, consider the optimal pure equity contract {[s.sup.*.sub.0], 0}, and let ([s.sub.1], 1) be the payoff-equivalent contract. Then as [delta] [right arrow] 0, the VC prefers the contract without DP only if

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Hence, {[s.sup.*.sub.0], 0} is preferred only if

[summation over (t[member of]{u,p,o}] [p.sub.t][[alpha].sub.t]([s.sub.1], 1; q) < [[alpha].sub.i] ([s.sup.*.sub.0], 0;q).

Subtracting [[alpha].sub.i] ([s.sup.*.sub.0], 0; q) from both sides yields the result.

Proof of Proposition 4.

When [delta] [right arrow] 0, [q.sup.*] solves

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

To establish the single-crossing property, suffices to show that at [q.sup.*],

(A2) [p.sub.u] ([[alpha].sub.u] ([s.sup.*.sub.1] (q), 1; q) - [[alpha].sub.p] ([s.sup.*.sub.1](q), 1; q))' < [p.sub.o] ([[alpha].sub.p] ([s.sup.*.sub.1](q), 1; q) - [[alpha].sub.o] ([s.sup.*.sub.1](q), 1;q))'.

Substitute for--at using Equations (Al) into (A2) yields

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

which holds by elasticity condition El.

REFERENCES

Aspatore Books Staff. Analyzing VC Deal Terms: Leading Lawyers on Structuring Term Sheets, Developing Negotiation Strategies, and Assessing Risks. Boston: Aspatore Books, 2008.

Anderlini, L., and L. Felli. "Incomplete Written Contracts: Undescribable States of Nature." Quarterly Journal of Economics, 109, 1994, 1085-124.

--. "Incomplete Contracts and Complexity Costs." Research in Economics, 58, 1999, 3-30.

--. "Bounded Rationality and Incomplete Contracts." Theory and Decision, 46, 2004, 23-50.

Ashenfelter, 0., D. Bloom, and G. Dahl. "Lawyers as Agents of the Devil in a Prisoner's Dilemma Game: Evidence from Long Run Play." NBER Working Paper 18834, 2013.

Bartlett, J. Equity Finance: Venture Capital, Buyouts, Restructurings and Reorganizations. New York: John Wiley & Sons, 1995.

Bengtsson, O., and D. Hsu. "How Do Venture Capital Partners Match with Startup Founders?" SSRN Working Paper 1568131, 2010.

Bengtsson, O., and S. Ravid. "The Importance of Geographical Location and Distance on Venture Capital Contracts." SSRN Working Paper 1331574, 2009.

Bengtsson, O., and B. Sensoy. "Investor Abilities and Financial Contracting: Evidence from Venture Capital." Journal of Financial Intermediation, 20, 2011, 472-502.

--. Forthcoming. "Changing the Nexus: The Evolution and Renegotiation of Venture Capital Contracts." Journal of Financial and Quantitative Analysis.

Block, R., and J. Stiebler. "The Impact of Attorneys and Arbitrators on Arbitration Awards." Industrial and Labor Relations Review, 40, 1987, 543-55.

Bottazzi, L., T. Hellmann, and M. Da Rin. "What Is the Role of Legal Systems in Financial Intermediation? Theory and Evidence." Journal of Financial Intermediation, 18, 2009,559-98.

Broughman, B., and J. Fried. "Renegotiation of Cash Flow Rights in the Sale of VC-Backed Firms." Journal of Financial Economics, 95, 2010, 384-99.

Coates, J. "Explaining Variation in Takeover Defenses: Blame the Lawyers." California Law Review, 89, 2001, 1301-421.

Comelli, F., and O. Yosha. "Stage Financing and the Role of Convertible Securities." Review of Economic Studies, 70,2003,1-32.

Cumming, D. J. "Contracts and Exits in Venture Capital Finance." Review of Financial Studies, 21, 2008, 1947-82.

Cumming, D. J., and N. Dai. "Why Do Entrepreneurs Switch Lead Venture Capitalists?" Entrepreneurship Theory and Practice, 37, 2013, 2013-17.

Cumming, D. J., and S. A. Johan. "Preplanned Exit Strategies in Venture Capital." European Economic Review, 52, 2008, 1209-41.

Dye, R. "Costly Contract Contingencies." International Economic Review, 26, 1985, 233-50.

Gompers, P. "An Examination of Convertible Securities in Venture Capital Investments." Working Paper, 1998.

Harcourt, M. "How Attorney Representation and Adjudication Affects Canadian Arbitration and Labor Relations Board Decisions." Journal of Labor Research, 21, 2000, 149-59.

Hart, O. "Financial Contracting." Journal of Economic Literature, 39, 2001, 1079-100.

Kaplan, S., and P. Stromberg. "Financial Contracting Meets the Real World: An Empirical Analysis of Venture Capital Contracts." Review of Economic Studies, 70, 2003,281-316.

Kaplan, S., and P. Stromberg. "Characteristics, Contracts, and Actions: Evidence from Venture Capitalist Analyses." Journal of Finance, 59, 2004, 2177-210.

Kaplan, S., F. Martel, and P. Stromberg. "How Do Legal Differences and Experience Affect Financial Contracts?" Journal of Financial Intermediation, 16, 2007, 273-311.

Lerner, J., and A. Schoar. "Does Legal Enforcement Affect Financial Transactions? The Contractual Channel in Private Equity." Quarterly Journal of Economics, 120, 2005, 223-46.

Metrick, A., and A. Yasuda. Venture Capital and the Finance of Innovation. Hoboken, NJ: Wiley, 2010.

Puri, M., and D. Robinson. "Optimism and Economic Choice." Journal of Financial Economics, 86. 2007, 71-99.

Sahlman, W. "The Structure and Governance of Venture Capital Organizations." Journal of Financial Economics, 27, 1990,473-521.

Sorensen, M. "How Smart Is Smart Money? A Two-Sided Matching Model of Venture Capital." Journal of Finance, 62, 2007, 2725-62.

Suchman, M. "On Advice of Counsel: Law Firms and Venture Capital Funds as Information Intermediaries in the Structuration of Silicon Valley." Doctoral dissertation, Stanford University, 1994.

Tirole, J. "Cognition and Incomplete Contracts." American Economic Review, 99, 2009, 265-94.

Vronsky, P. "Venture Capital Law Firms: An Analysis of Equity Investments and Networks, " in Oxford Handbook of Venture Capital, Chapter 20, edited by D. J. Cumming. New York: Oxford University Press, 2012, 602-26.

Zarutskie, R. "The Role of Top Management Team Human Capital in Venture Capital Markets: Evidence from First-Time Funds." Journal of Business Venturing, 25, 2010, 155-72.

(1.) As one lawyer who specializes on VC investments observes, "in the majority of situations, the [entrepreneur] client will not understand most of the deal terms that are being proposed. Therefore, it is up to me to explain [...]. Indeed, there is generally a big difference in what VCs understand and what your typical entrepreneur understands going into a venture capital deal" (Aspatore Books Staff 2008, 45).

(2.) A limitation is that our data lack information about control rights.

(3.) Cumming and Johan (2008) also find that the experience of the entrepreneur affects contracting. In their sample of European VC investments, experienced entrepreneurs are more likely to get financed with common equity, which has no DP. Further, experienced entrepreneurs are not significantly more likely to employ lawyers with VC expertise, reinforcing that complementarities between entrepreneur quality and lawyer VC expertise do not drive our empirical findings.

(4.) This result indicates that our model can be distinguished from search-based explanations for the positive correlation between q and having a good legal counsel, in which an increase in q raises the benefit of an expert legal counsel, so that better entrepreneurs search harder for expert legal counsels. DP, whether necessary or not, has a smaller impact on better entrepreneurs. Thus, a search-based explanation would require that expert legal counsels provide services other than just evaluating DP, and better entrepreneurs would have to value those services by more.

(5.) Our 908 contracts represent about half of the 1,804 contracts in Bengtsson and Ravid (2009). Our sample is smaller for two reasons. First, we study only contracts for which we can identify legal counsels. Second, to avoid issues with cross-correlated standard errors, we restrict our sample to one contract per company--our results are robust to including multiple contracts for a company.

(6.) This measure of lawyer expertise persists over time. The correlation for number of advised companies in 2005-2008 (our sample period) and 2001-2004 is very high.

(7.) The results remain qualitatively similar if we instead use ordered logit regressions.

(8.) In untabulated robustness tests, we use other measures of VC experience, such as the age of the firm and the prior success rate of its investments. We also include separate dummies for corporate VCs, financial VCs, and government-related VCs. Results on lawyer experience remain unchanged.

(9.) The results remain qualitatively similar if we instead cluster on industry, year, or company location.

(10.) Cumming and Johan (2008) have a related finding: experienced entrepreneurs are more likely to be financed with common equity than preferred equity, which gives them less contractual DP.

(11.) Sorensen also finds that VC reputation correlates with successful exits because higher-quality VCs are better able to add value and monitor their portfolio companies. Arguably, if such post-investment treatment adds more to better firms, this may reinforce the importance of matching between entrepreneurs and their legal counsels.

(12.) The equity value of high-quality companies does not directly influence matching, because law firms advising venture-backed companies currently receive compensation in cash and not company stock. Prior to the "dot-com crash" of the early 2000s (which is before our sample period), equity compensation was more common.

OLA BENGTSSON ([dagger]) and DAN BERNHARDT *

* We thank Murillo Campello, Rafael da Matta, Dirk Hackbarth, Yelena Larkin, and Berk Sensoy, and seminar participants at DePaul/Federal Reserve Bank of Chicago, Cornell University. Indiana University, Lund University, Research Institute of Industrial Economics, Law and Society Meeting, and American Law and Economics Conference for helpful comments. We are grateful to VCExperts and Joseph Bartlett for access to the contract data, and to Neely Tang and Bonny Lee for valuable research assistance. All remaining errors are our own.

([dagger]) Shortly after this paper was accepted, Ola Bengtsson was diagnosed with leukemia. Ola passed away on January 5, 2014. Ola will be missed by everyone.

Bengtsson: Department of Economics, Lund University.

Bernhardt: Department of Economics and Department of Finance, University of Illinois; Department of Economics, University of Warwick, U.K. Phone 217-244-5708, Fax 217-244-6678, E-mail danber@illinois.edu; M.D.Bemhardt@warwick.ac.uk

TABLE 1
Summary Statistics

                                    Number of               Standard
                                   Observations    Mean     Deviation

Counsel VC Expertise Counsel VC        908          39         40
  number of deals
Counsel VC top ranked                  908         0.63       0.48
Counsel VC practice area               908         0.87       0.33
Counsel VC office in Silicon           908         0.51       0.50
  Valley
Counsel General Expertise              908         0.05       0.21
  Counsel General AMLaw Top 10
Counsel General AMLaw Top 25           908         0.17       0.38
Counsel general revenue (000s)         694        554,000    338,000
Counsel general lawyers                701          706        401
Counsel general entry salary           653        151,714    12,136
Counsel general revenue/lawyer         694        761.437    93,697
Counsel general gross profit           694         0.36       0.07
  margin

Contract DP

Downside protection index (DPI)        908         4.64       1.54

Other Variables

Pre-money valuation (000s)             507        51,400     63.600
Round number                           908         2.81       1.58
Total round amount (000s)              908        12,200     13,300
Number of VCs in round                 908         4.31       2.64
Company in California                  908         0.49       0.50
High technology industry               908         0.39       0.49
Life science industry                  908         0.30       0.46
Serial founder                         908         0.24       0.43
Serial founder with IPO                908         0.07       0.25
Company age                            908         4.07       2.72
VC in California                       908         0.50       0.50
VC and company in same state           908         0.49       0.50
VC partnership                         908         0.81       0.39
VC experience                          908          136        164
VC age                                 908          15         11
VC IPO ratio                           908         0.15       0.08

                                   Minimum    Maximum

Counsel VC Expertise Counsel VC       1         114
  number of deals
Counsel VC top ranked               0.00       1.00
Counsel VC practice area            0.00       1.00
Counsel VC office in Silicon        0.00       1.00
  Valley
Counsel General Expertise           0.00       1.00
  Counsel General AMLaw Top 10
Counsel General AMLaw Top 25        0.00       1.00
Counsel general revenue (000s)     99,000    1,620,000
Counsel general lawyers              162       3,082
Counsel general entry salary       100,000    160,000
Counsel general revenue/lawyer     413,352   1,051,020
Counsel general gross profit        0.22       0.63
  margin

Contract DP

Downside protection index (DPI)     0.00       11.00

Other Variables

Pre-money valuation (000s)           910      493,000
Round number                        1.00       5.00
Total round amount (000s)            50       110,000
Number of VCs in round              1.00       17.00
Company in California               0.00       1.00
High technology industry            0.00       1.00
Life science industry               0.00       1.00
Serial founder                      0.00       1.00
Serial founder with IPO             0.00       1.00
Company age                         0.00       10.00
VC in California                    0.00       1.00
VC and company in same state        0.00       1.00
VC partnership                      0.00       1.00
VC experience                         1         797
VC age                                0         47
VC IPO ratio                        0.00       0.53

Notes: One observation is one contract between a venture-backed
company and its lead VC investor. Counsel General Expertise variables
are from the 2007 edition of American Lawyer 200. Each contract is
matched by company name and round date with an investment round
listed in Venture Economics. All VC variables are updated to match
the year of the contract. Table 2 details the coding of cash flow
contingencies that we aggregate to compute the downside protection
index (DPI). Lead VC is identified as the largest investor in the
round.

TABLE 2
Description of the Cash Flow Contingencies and their Coding

Cumulative Dividends: Dividends that the investor earns annually
until the company is sold or liquidated. Cumulative means that the
dividends are not paid out annually, but when the company is sold or
liquidated. Cumulative dividends are senior to common stock.

               Above 8% = 2   8% or Below = 1   Not Included = 0

Number of         42 (5%)        170 (19%)         696 (77%)
observations

Liquidation Preferences: The multiple of the investors investment
that is paid back to the investor when the company is sold or
liquidated. Liquidation preference is senior to common stock.

               Above 2X = 2   Above IX, Up to 2X = 1    IX = 0

Number of         7 (1%)              51 (6%)          850 (94%)
observations

Participation: With participation the investor receives both a
liquidation preference and a fraction of common stock when the
company is sold or liquidated. With "Capped" participation, the
investor only receives the liquidation preference if his/her
investment IRR is below a certain hurdle. With no participation, the
investor holds convertible preferred stock.

               Not Capped = 2   Capped = 1 Not   Included = 0

Number of        371 (41%)        237 (26%)        300 (33%)
observations

Antidilution-. The investor is issued additional shares if the
company raises a new financing round at a lower valuation than what
the investor paid (down round). "Full Ratchet" gives the investor
more additional shares than "Weighted Average," especially if the new
financing round is small.

                                   Weighted
               Full Ratchet = 2   Average = 1   Not Included =0

Number of           14 (2%)        832 (92%)        14 (2%)
observations

Redemption: The investor has the right to sell his/her shares back to
the company after a specified time period. A typical redemption right
provision gives the investor the right to sell back 1/3 of his/her
shares after 5 years, 1/3 after 6 years, and the remaining 1/3 after
7 years.

                         Included = 1   Not Included = 0

Number of observations     434 (48%)        474 (52%)

Pay-to-Play. Pay-to-Play provisions specify what contractual rights
the investor loses if he/she does not invest in a follow-up financing
round of the company. With "Convert to Preferred," the investor loses
some contractual rights (typically antidilution rights) that are
attached to his/her preferred stock. With "Convert to Common," the
investor loses all contractual rights that are attached to his/her
preferred stock.

                                    Convert
               Not Included = 2   Preferred = 1   Convert Common = 0

Number of          746 (82%)         36 (4%)           126 (14%)
observations

Notes: Each cash flow contingency contributes 0. 1. or 2 to the
downside protection index (DPI), where 2 is the most DP.

TABLE 3
DP and Legal Counsel Expertise, Univariate

Panel A: Downside Protection Index and Counsel VC Expertise,
Different Proxies

                       Less        More
                     Expertise   Expertise   Difference   Significance

Counsel VC #           4.86        4.39         0.47          0.00
  deals
Counsel VC top         4.92        4.48         0.44          0.00
  ranked
Counsel VC             5.11        4.58         0.53          0.00
  practice area
Counsel VC office      4.82        4.48         0.35          0.00
  in Silicon
  Valley
All four counsel       5.17        4.36         0.81          0.00
  proxies combined

Panel B: Separate Cash Flow Contingencies and Counsel VC Expertise

                       Less        More
                     Expertise   Expertise   Difference   Significance

Cumulative             0.40        0.15         0.25          0.00
  dividends
Liquidation            0.08        0.07         0.01          0.86
  preference
Participation          1.28        0.97         0.32          0.00
Antidilution           1.09        1.01         0.09          0.01
Redemption             0.49        0.39         0.11          0.07
Pay-to-play            1.83        1.75         0.08          0.33

Panel C: Downside Protection Index and Counsel General Expertise,
Different Proxies

                       Less        More
                     Expertise   Expertise   Difference   Significance

Counsel general        4.65        4.44         0.22          0.74
  AmLaw top 10
Counsel general        4.65        4.61         0.05          0.98
  AmLaw top 25
Counsel general        4.59        4.54         0.05          0.69
  revenue
Counsel general        4.54        4.57        -0.04          0.88
  lawyer
Counsel general        4.68        4.44         0.24          0.02
  revenue per
  lawyer
Counsel general        4.53        4.60        -0.07          0.50
  gross profit
  margin
Counsel general        4.50        4.57        -0.07          0.59
  entry salary

Notes: Panel A presents the mean downside protection index (DPI) for
different sorts on Counsel VC Expertise. All four Counsel Proxies
Combined takes the value 1 if each proxy of Counsel VC Expertise
takes the value 1, and 0 if each proxy takes the value 0. Panel B
presents the mean for separate cash flow contingencies for Counsel VC
Expertise, based on all four proxies combined. See Table 2 for a
description of separate contract terms and their codings. Panel C
presents the mean (DPI) for different sorts on counsel general
expertise. Low (High) Expertise for nonbinary variables includes
variables below (above) the sample median. Rank test is a Wilcoxon
rank-sum test of difference of means.

TABLE 4
Downside Protection Index and Legal Counsel VC Expertise

Specification                     1            2            3

Dependent variable
(log) Counsel VC# deals       -0.155 ***
                               [0.034]
Counsel VC top ranked                      -0.290 ***
                                            [0.104]
Counsel VC practice area                                -0.434 ***
                                                         [0.163]
Counsel VC office in SV

(log) Total round amount      -0.137 ***   -0.145 ***   -0.150 ***
                               [0.048]      [0.048]      [0.048]
(log) VC Experience           -0.229 ***   -0.240 ***   -0.248 ***
                               [0.055]      [0.055]      [0.055]
VC Partnership                -0.404 ***   -0.417 ***   -0.405 ***
                               [0.109]      [0.112]      [0.111]
VC, Company same state        -0.353 ***   -0.362 ***   -0.382 ***
                               [0.119]      [0.117]      [0.116]
(log) #VCs in round           0.349 ***    0.362 ***    0.382 ***
                               [0.115]      [0.113]      [0.112]
(log) Company age             0.303 ***    0.298 ***    0.306 ***
                               [0.085]      [0.088]      [0.089]
Serial founder                  -0.040       -0.037       -0.020
                               [0.166]      [0.168]      [0.169]
Serial founder with IPO         -0.188       -0.184       -0.175
                               [0.211]      [0.214]      [0.215]
Serial founder with merger      -0.020       -0.026       -0.057
                               [0.186]      [0.186]      [0.189]
Constant                      8.291 ***    8.211 ***    8.499 ***
                               [0.767]      [0.780]      [0.788]
Observations                     908          908          908
[R.sup.2]                        0.12         0.11         0.11
VC location FE                    No           No           No
Company location FE               No           No           No
Company industry FE               No           No           No
Contract year FE                  No           No           No
Round number FE                   No           No           No

Specification                     4            5            6

Dependent variable               DPI
(log) Counsel VC# deals                    -0.163 ***
                                            [0.032]
Counsel VC top ranked                                   -0.333 ***
                                                         [0.100]
Counsel VC practice area

Counsel VC office in SV        -0.210 *
                               [0.115]
(log) Total round amount      -0.155 ***     -0.063       -0.072
                               [0.050]      [0.044]      [0.044]
(log) VC Experience           -0.246 ***   -0.235 ***   -0.243 ***
                               [0.054]      [0.059]      [0.059]
VC Partnership                -0.404 ***   -0.382 ***   -0.389 ***
                               [0.111]      [0.112]      [0.113]
VC, Company same state        -0.382 ***   -0.132 ***   -0.122 ***
                               [0.119]      [0.118]      [0.120]
(log) #VCs in round           0.366 ***    0.198 ***    0.208 ***
                               [0.115]      [0.115]      [0.117]
(log) Company age             0.293 ***    0.122 ***    0.122 ***
                               [0.092]      [0.101]      [0.103]
Serial founder                  -0.031       -0.112       -0.110
                               [0.168]      [0.148]      [0.149]
Serial founder with IPO         -0.167       -0.093       -0.088
                               [0.218]      [0.189]      [0.190]
Serial founder with merger      -0.037       0.024        0.014
                               [0.188]      [0.204]      [0.206]
Constant                      8.318 ***    6.602 ***    6.518 ***
                               [0.782]      [0.935]      [0.914]
Observations                     908          908          908
[R.sup.2]                        0.10         0.24         0.23
VC location FE                    No          Yes          Yes
Company location FE               No          Yes          Yes
Company industry FE               No          Yes          Yes
Contract year FE                  No          Yes          Yes
Round number FE                   No          Yes          Yes

Specification                     7            8

Dependent variable
(log) Counsel VC# deals

Counsel VC top ranked

Counsel VC practice area      -0.469 ***
                               [0.162]
Counsel VC office in SV                    -0.239 **
                                            [0.116]
(log) Total round amount       -0.077 *     -0.081 *
                               [0.044]      [0.045]
(log) VC Experience           -0.250 ***   -0.246 ***
                               [0.058]      [0.059]
VC Partnership                -0.372 ***   -0.368 ***
                               [0.113]      [0.114]
VC, Company same state        -0.144 ***   -0.132 ***
                               [0.122]      [0.121]
(log) #VCs in round           0.217 ***    0.211 ***
                               [0.117]      [0.118]
(log) Company age             0.129 ***    0.130 ***
                               [0.105]      [0.105]
Serial founder                  -0.096       -0.114
                               [0.149]      [0.148]
Serial founder with IPO         -0.082       -0.055
                               [0.195]      [0.195]
Serial founder with merger      -0.013       0.012
                               [0.208]      [0.203]
Constant                      6.692 ***    6.574 ***
                               [0.900]      [0.907]
Observations                     908          908
[R.sup.2]                        0.23         0.22
VC location FE                   Yes          Yes
Company location FE              Yes          Yes
Company industry FE              Yes          Yes
Contract year FE                 Yes          Yes
Round number FE                  Yes          Yes

Notes: Table 2 details how we compute the Downside Protection
Index (DPI). A higher DP means that the contract includes more
investor- friendly cash flow rights. OLS regressions. Specifications
5-8 include fixed effects for VC firm location (California,
Massachussets, Texas, New York, and other), company location (state),
company industry (Venture Economics 10-group classification), round
year, and round number. Standard errors are clustered by counsel and
reported in brackets.

Significance at 10% level marked with *, 5% with **, and 1% with ***.

TABLE 5
Downside Protection Index and Legal Counsel VC and General Expertise

Specification                  1            2            3

Dependent variable
(log)Counsel VC #deals     -0.247 ***
                            [0.058]
Counsel VC top ranked                   -0.619 ***
                                         [0.175]
Counsel VC practice area                             -0.760 ***
                                                      [0.188]
Counsel VC office in SV

(log)Counsel VC #deals      0.124 *
  X Follow-up round         [0.071]
Counsel VC top rank                      0.405 *
  X Follow-up round                      [0.212]
Counsel VC practice area                              0.495 *
  X Follow-up round                                   [0.277]
Counsel VC office in SV
  X Follow-up round
(log)counsel VC # deals
  X Serial founder
Counsel VC top ranked
  X Serial founder
Counsel VC practice area
  X Serial founder
Counsel VC office in SV
  X Serial founder
Counsel gen AmLaw top 25     -0.017       0.064        -0.005
                            [0.104]      [0.115]      [0.113]
Observations                  908          908          908
[R.sup.2]                     0.24         0.23         0.23
Full set of company, VC,
 Round, founder controls      Yes          Yes          Yes
VC location FE                Yes          Yes          Yes
Company location FE           Yes          Yes          Yes
Company industry FE           Yes          Yes          Yes
Contract year FE              Yes          Yes          Yes
Round number FE               Yes          Yes          Yes

Specification                  4            5            6

Dependent variable            DPI
(log)Counsel VC #deals                  -0.197 ***
                                         [0.040]
Counsel VC top ranked                                -0.463 ***
                                                      [0.129]
Counsel VC practice area

Counsel VC office in SV    -0.508 ***
                            [0.161]
(log)Counsel VC #deals
  X Follow-up round
Counsel VC top rank
  X Follow-up round
Counsel VC practice area
  X Follow-up round
Counsel VC office in SV     0.391 **
  X Follow-up round         [0.189]
(log)counsel VC # deals                  0.155 *
  X Serial founder                       [0.086]
Counsel VC top ranked                                 0.502 *
  X Serial founder                                    [0.229]
Counsel VC practice area
  X Serial founder
Counsel VC office in SV
  X Serial founder
Counsel gen AmLaw top 25     0.001        -0.033       0.047
                            [0.118]      [0.105]      [0.118]
Observations                  908          908          908
[R.sup.2]                     0.23         0.24         0.23
Full set of company, VC,
 Round, founder controls      Yes          Yes          Yes
VC location FE                Yes          Yes          Yes
Company location FE           Yes          Yes          Yes
Company industry FE           Yes          Yes          Yes
Contract year FE              Yes          Yes          Yes
Round number FE               Yes          Yes          Yes

Specification                  7            8

Dependent variable
(log)Counsel VC #deals

Counsel VC top ranked

Counsel VC practice area   -0.619 ***
                            [0.169]
Counsel VC office in SV                  -0.281 *
                                         [0.151]
(log)Counsel VC #deals
  X Follow-up round
Counsel VC top rank
  X Follow-up round
Counsel VC practice area
  X Follow-up round
Counsel VC office in SV
  X Follow-up round
(log)counsel VC # deals
  X Serial founder
Counsel VC top ranked
  X Serial founder
Counsel VC practice area    0.678 **
  X Serial founder          [0.340]]
Counsel VC office in SV                   0.173
  X Serial founder                       [0.241]
Counsel gen AmLaw top 25     -0.014       0.001
                            [0.113]      [0.119]
Observations                  908          908
[R.sup.2]                     0.23         0.23
Full set of company, VC,
 Round, founder controls      Yes          Yes
VC location FE                Yes          Yes
Company location FE           Yes          Yes
Company industry FE           Yes          Yes
Contract year FE              Yes          Yes
Round number FE               Yes          Yes

Notes: Table 2 details how we compute the Downside Protection
Index (DPI). A higher DPI means that the contract includes more
investor-friendly cash flow rights. OLS regressions. All
specifications include fixed effects for VC firm location
(California, Massachussets, Texas, New York, and other), company
location (state), company industry (Venture Economics 10-group
classification), round year, and round number. All specifications
also include (log) Company Age, Serial Founder, Serial Founder with
IPO, Serial Founder with Merger, (log) Number of VCs in Round, (log)
Total Round Amount, VC and Company in Same State, (log) VC Experience
and VC Partnership dummy. Counsel General Expertise measured using
data from the 2007 issue of American Lawyer 200. Constant is
estimated, but not reported. Standard errors are clustered by counsel
and reported in brackets.

Significance at 10% level marked with *, 5% with **, and 1% with ***.

TABLE 6
Downside Protection Index, Interactions with Entrepreneur Experience

Specification                         1            2            3

Dependent variable
(log) Counsel VC# deals           -0.164 ***   -0.163 ***   -0.095 **
                                   [0.032]      [0.032]      [0.039]
Counsel general AmLaw top 25        -0.076
                                   [0.161]
Counsel general AmLaw top 10                     -0.026
                                                [0.104]
Counsel general (log) revenue                                 -0.093
                                                             [0.062]
Counsel general (log) counsel

Counsel general (log) revenue
  per counsel
Counsel general gross profit
  margin
Counsel general entry salary
  above median
Observations                         908          908          694
[R.sup.2]                            0.24         0.24         0.26
Full set of company, VC, round
Founder controls                     Yes          Yes          Yes
VC location FE                       Yes          Yes          Yes
Company location FE                  Yes          Yes          Yes
Company industry FE                  Yes          Yes          Yes
Contract year FE                     Yes          Yes          Yes
Round number FE                      Yes          Yes          Yes

Specification                         4            5

Dependent variable                   DPI
(log) Counsel VC# deals           -0.087 **    -0.093 **
                                   [0.038]      [0.042]
Counsel general AmLaw top 25

Counsel general AmLaw top 10

Counsel general (log) revenue

Counsel general (log) counsel       -0.076
                                   [0.068]
Counsel general (log) revenue                    -0.138
  per counsel                                   [0 477]
Counsel general gross profit
  margin
Counsel general entry salary
  above median
Observations                         701          653
[R.sup.2]                            0.25         0.26
Full set of company, VC, round
Founder controls                     Yes          Yes
VC location FE                       Yes          Yes
Company location FE                  Yes          Yes
Company industry FE                  Yes          Yes
Contract year FE                     Yes          Yes
Round number FE                      Yes          Yes

Specification                         6            7

Dependent variable
(log) Counsel VC# deals           -0.099 **    -0.159 ***
                                   [0.039]      [0.032]
Counsel general AmLaw top 25

Counsel general AmLaw top 10

Counsel general (log) revenue

Counsel general (log) counsel

Counsel general (log) revenue
  per counsel
Counsel general gross profit        -0.191
  margin                           [0.644]
Counsel general entry salary                     -0.066
  above median                                  [0.068]
Observations                         694          694
[R.sup.2]                            0.26         0.24
Full set of company, VC, round
Founder controls                     Yes          Yes
VC location FE                       Yes          Yes
Company location FE                  Yes          Yes
Company industry FE                  Yes          Yes
Contract year FE                     Yes          Yes
Round number FE                      Yes          Yes

Notes: Table 2 details how we compute the downside protection
index (DPI). A higher DPI means that the contract includes more
investor-friendly cash flow rights. OLS regressions. All
specifications include fixed effects for VC firm location
(California, Massachussets, Texas, New York, and other), company
location (state), company industry (Venture Economics 10-group
classification), round year, and round number. All specifications
also include (log) Company Age, Serial Founder, Serial Founder with
IPO, Serial Founder with Merger, (log) Number of VCs in Round, (log)
Total Round Amount, VC and Company in Same State, (log) VC Experience
and VC Partnership dummy. Counsel General Expertise measured using
data from the 2007 issue of American Lawyer 200. Constant is
estimated, but not reported. Standard errors are clustered by counsel
and reported in brackets.

Significance at 10% level marked with *, 5% with **, and 1% with ***.

TABLE 7
Legal Counsel VC Expertise and Matching Company Quality

Specification             1             2            3           4

Dependent variable                (log) Counsel              Counsel VC
                                   VC # deals                Top ranked

(log) Total round     0.131 ***                               0.032 *
  amount               [0.048]                                [0.019]
(log) VC experience                 0.175 **
                                     [0.069]
VC IPO ratio                                      3.077 **
                                                  [1.025]
Observations             693           693          693         693
[R.sup.2]               0.22          0.23          0.24        0.14
Comp, age control        Yes           Yes          Yes         Yes
VC location FE           Yes           Yes          Yes         Yes
Comp, industry FE        Yes           Yes          Yes         Yes
Comp, location FE        Yes           Yes          Yes         Yes
Contract year FE         Yes           Yes          Yes         Yes
Round number FE          Yes           Yes          Yes         Yes

Specification             5           6          7           8

Dependent variable                                      Counsel VC
                                                        Pract. area

(log) Total round                              0.003
  amount                                      [0.013]
(log) VC experience   0.052 ***                           0.021 *
                       [0.020]                            [0.012]
VC IPO ratio                      0.791 ***
                                   [0.293]
Observations             693         693        693         693
[R.sup.2]               0.14        0.14       0.21        0.22
Comp, age control        Yes         Yes        Yes         Yes
VC location FE           Yes         Yes        Yes         Yes
Comp, industry FE        Yes         Yes        Yes         Yes
Comp, location FE        Yes         Yes        Yes         Yes
Contract year FE         Yes         Yes        Yes         Yes
Round number FE          Yes         Yes        Yes         Yes

Specification            9        10           11           12

Dependent variable                         Counsel VC
                                             Office
                                          in Sil. Val.
(log) Total round                0.005
  amount                        [0.022]
(log) VC experience                        0.057 ***
                                            [0.021]
VC IPO ratio          0.307 *                            0.825 **
                      [0.159]                            [0.376]
Observations            693       693         693          693
[R.sup.2]              0.22      0.10         0.11         0.11
Comp, age control       Yes       Yes         Yes          Yes
VC location FE          Yes       Yes         Yes          Yes
Comp, industry FE       Yes       Yes         Yes          Yes
Comp, location FE       Yes       Yes         Yes          Yes
Contract year FE        Yes       Yes         Yes          Yes
Round number FE         Yes       Yes         Yes          Yes

Notes: The specifications use different proxies for counsel VC
expertise (higher number means a more expert law firm).
Specifications 1-3 are OLS regressions. Specifications 4-12 are
probit regressions. All specifications include fixed effects for VC
firm location (California, Massachussets, Texas, New York, and
other), company location (state), company industry (Venture Economics
10-group classification), round year, and round number. Constant is
estimated, but not reported. Standard errors are clustered by counsel
and reported in brackets.

Significance at 10% level marked with *, 5% with **, and 1% with ***.