The Leveraging of Silicon Valley

Transcription

1 Te Leveraging of Silicon Valley Jesse Davis, Adair Morse, Xinxin Wang Marc 2018 Abstract Venture debt is now observed in 28-40% of venture financings. We model and document ow tis early-stage leveraging can affect firm outcomes. In our model, a venture capitalist maximizes firm value troug financing. An equity-olding entrepreneur cooses ow muc risk to take, trading off te financial benefit against is preference for continuation. By extending te runway, utilizing venture debt can reduce dilution, tereby aligning te entrepreneur s incentives wit te firm s. Te resultant risk-taking increases firm value, but te leverage puts te startup at greater risk of failure. Empirically, we sow tat early-stage ventures take on venture debt wen it is optimal to delay financing: suc firms face iger potential dilution and exibit lower pre-money valuations. Consistent wit tis notion, suc firms take eigty-two fewer days between financing events. Tis strategy induces iger failure rates: $125,000 more venture debt predicts 6% iger closures. However, conditional on survival, venture debt-backed firms ave 7-10% iger acquisition rates. Our study igligts te role of leverage in te risking-up of early-stage startup firms. Aggregation of tese tradeoffs is important for understanding venture debt s role in te real economy. University of Nort Carolina - Capel Hill (jesse davis@kenan-flagler.unc.edu), University of California - Berkeley (morse@aas.berkeley.edu), and University of Nort Carolina - Capel Hill (xinxin wang@kenanflagler.unc.edu) respectively.

2 1 Introduction Entrepreneurial ventures foster tecnological development, drive competition and create economic growt. However, entrepreneurs are usually liquidity-constrained, making te financing of entrepreneurial ventures troug external capital an essential question in economics and finance. Altoug economic teory would generally predict tat external debt is an unlikely veicle for te financing of early-stage startups, te venture debt market as grown rapidly in recent years. Ibraim (2010) estimates tat venture lenders, including leader Silicon Valley Bank and specialized non-bank lenders, supply $1 - $5 billion to startups annually. In more recent work, Tykvová (2017) finds tat around 28% of venture-backed companies in Dow Jones Venture Source utilize venture debt. In our large-sample analysis, we find tat venture debt is often a complement to equity financing, wit over 40% of all financing rounds including some amount of debt. 1 Venture debt is generally structured as a sort-term (tree-year) loan, wit warrants for company stock. Its role differs from te now-ubiquitous convertible note contract (te standard early-stage seed financing contract), wose primary feature is its conversion to equity at a later stage. It also does not resemble traditional debt loans in tat it is a debt instrument for venture equity-backed companies tat lack collaterizable assets or cas flows. Instead, venture debt is secured (wit uncertainty) by future rounds of equity finance. Proponents of venture debt and te nascent, important literature on venture debt (e.g., de Rassenfosse and Fiscer (2016), Hocberg et al. ((fortcoming), González-Uribe and Mann (2017)) convincingly argue tat it provides growt capital to extend te runway of a startup, allowing tem to acieve te next milestone wile minimizing equity dilution for bot te founders and equity investors. Tese studies overlook te impact to startups and te real economy from te fact tat venture debt is still a debt product, wic carries te traditional implications wic arise wen leveraging a firm. In tis paper, we provide teoretical foundations, supported by empirical evidence, on te use of venture debt. In te model, an entrepreneur trades off te financial benefits of risk-taking wit te utility e forfeits if te firm fails. If te entrepeneur s equity is too diluted, e favors a low-risk (low-value) strategy. We sow tat venture debt can reduce dilution by delaying equity financing until a milestone is met and incents te entrepreneur to coose a ig-risk (ig-value) strategy. Empirically, we sow tat venture debt is utilized wen expected dilution is ig and wen it is optimal to delay financing so tat te next milestone may be reaced. Furtermore, startups tat take on venture debt ave sorter time between financing events, iger failure 1 See Figure 1 for a breakdown of financing round by types. Ibraim (2010) estimates tat te venture debt market is approximately 10-20% of aggregate venture capital. Te difference in magnitude is te syndication of rounds by bot debt and equity investors. 1

3 rates, and iger acquisition rates conditional on survival. Te optimal use of early-stage leverage suggests several major canges in our perception of startups. First, if venture debt incents entrepreneurs and firms to risk up, te innovation economy may be facing greater uncertainty (bot financial and strategic) tan in previous decades. Second, if venture debt increases expected firm value, more startups may be able to receive funding (ex-ante and interim) tan would oterwise. Tird, te use of venture debt may be canging te allocation of bot uman capital and startup finance capital toward te continuation of riskier endeavors and away from te alternative use of suc resources. To establis our teoretical predictions, we consider a tree-date model. At date zero, a firm owns a risky asset of uncertain quality. At date one, te asset s quality is revealed after wic te firm s strategy is cosen. At date two, te cas flow is realized. 2 te firm must raise capital to avoid closure, e.g., to pay employees. Before eac date, Te firm is owned by an entrepreneur and a venture capitalist; bot are risk-neutral. 3 Te venture capitalist cooses ow and wen to raise capital to maximize te expected value of te firm. 4 In particular, at date zero, se as (1) te option of raising some portion of te required financing after te asset s quality is revealed and (2) access to bot equity and venture debt investors. At date one, te entrepreneur implements te firm s going-to-market strategy, wic is unobservable. Specifically, te firm s strategy determines te riskiness of te distribution of te terminal cas flow. Te entrepreneur cooses ow muc risk to take, accounting for te value of is equity claim as well as te non-pecuniary utility e derives from continuation, i.e., te firm avoiding sutdown. Tis non-pecuniary utility creates a wedge between te venture capitalist s and entrepreneur s incentives. 5 Unsurprisingly, wen te entrepreneur s stake is excessively diluted, e cooses te low-risk (low-value) strategy. Preferring te ig-risk strategy, te venture capitalist makes er financing decisions to minimize te likeliood tis occurs. We sow tat if te firm s unconditional quality is sufficiently ig, te firm can raise te required capital ceaply in one round te entrepreneur cooses te ig-risk strategy and firm value is maximized. As unconditional quality falls, te entrepreneur s dilution increases; if it falls sufficiently, te entrepreneur cooses to scale back risk. In tis case, te venture capitalist cooses to raise some portion of te needed funds after firm quality is known. We sow tat tis is beneficial if te firm s asset is revealed to be ig-quality: at tat point, equity can be raised less expensively, reducing 2 Under te assumptions of te model, tis terminal cas flow need not be realized and is equivalent to an expectation of te firm s value as a going concern. 3 Te venture capitalist is an equity investor from an earlier round. 4 Tis is consistent wit bot te survey evidence from Ibraim (2010) and Sage (2010). 5 Tis wedge utilizes te well-documented fact tat wile bot venture capitalists and entrepreneurs seek to maximize firm value, venture capitalists often prefer iger volatility in teir investments relative to entrepreneurs (wo also value continuation of teir startups). 2

4 dilution (and potentially incenting te entrepreneur to take te ig-risk strategy once more). On te oter and, it also creates te possibility of failure if te firm s asset is revealed to be low-quality. Venture debt amplifies tis effect. By borrowing today, te firm raises less equity at a low, unconditional value. Tis increases te required equity issued in te future, but tis is done at a potentially ig conditional value. Toug it comes wit increased risk of failure, we sow tat, in some cases, venture debt is strictly preferable from te venture capitalist s perspective. Te model generates several empirical predictions consistent wit features of te venture debt market. First, all else equal, venture debt is more likely to be optimal wen te entrepreneur faces ig potential dilution - for instance, wen te firm requires significant investments of capital. Second, we expect to see more venture debt wen te benefits of risk-taking are low; suc debt is necessary to incent te entrepreneur to coose te value-maximizing strategy. Tird, we expect to see venture capital utilized by mid-value firms: tose firm tat firms can raise capital, but do so at great cost. Finally, we sow tat wile te use of venture debt increases te sort-term probability of firm closure it also increases te value of te firm, conditional on survival. Wit tese teoretical predictions in mind, we offer five, novel empirical contributions. We begin by identifying wic startups coose debt in teir financing and ow it. First, we sow tat potential dilution is a strong predictor of te decision to raise venture debt instead of venture equity. Indeed, startups wit a standard deviation iger dilution from te current round are five percent more likely to issue suc debt. Bot entrepreneurs and investors value skin-in-te-game and te additional capital provided by a venture loan allows startups to acieve more progress before raising additional equity. Furter, if te firm is able to reac its milestone (i.e., is ig quality in te parlance of te model), tis approac minimizes te dilution tat occurs relative to securing suc external capital at an earlier time. We ten provide evidence consistent wit tis intuition of venture debt as extending te runway. Our second contribution sows tat firm quality realizations are a driver of venture capitalist preference for venture debt. We find tat in early rounds, low pre-money valuations, wic are indicative of missing milestones or targets, lead to an increase in te likeliood of raising debt. 6 Our tird contribution finds tat after early-stage startups coose venture debt, tey return to te venture investor market in eigty-two fewer days, even after controlling for te amount of capital raised. Tis suggests tat suc firms are using venture debt as an extension (aving failed to reac a needed milestone) and tat tey return to te market after more information is revealed about te firm s future prospects. 6 In later rounds, ig pre-money valuations, wic are indicative of stable returns, lead to an increase in traditional debt financing. 3

5 Turning to firm level outcomes, our fourt contribution sows tat leverage makes te company more risky, at least until te next milestone is met. Specifically, debt increases te probability of startup closure in te first tree years. An increase in early-stage financing to include $125,000 in venture debt is associated wit a 6% iger likeliood of firm closing. As expected, firms wic survive te risk generated by venture debt benefit. An early debt round increases te likeliood of exiting via acquisition, conditional on not closing, by 7-10%. Tis fiftn contribution is consistent wit te intuition tat firms utilize venture debt not simply to prevent dilution but to improve firm value as well. Our researc adds to te current finance literature in several areas. First, tis paper contributes to te growing literature on venture lending. Te existing literature as focused on determinants of te lending decision. Hocberg et al. ((fortcoming) empirically tests te collateriability of patents as a driver of venture lending lending wile de Rassenfosse and Fiscer (2016) finds tat backing from venture capitalists substitute for startups cas flow in te lending decision. González-Uribe and Mann (2017) provides contract-level data on venture loans and finds tat intellectual capital and warrants are important features. Tese results corroborate te earlier market survey work by Ibraim (2010) wo finds tat venture debt provides additional runway between early-stage rounds and are repaid troug future equity raises. Similarly, is researc also points to te importance of intellectual property as collateral for te loan. Missing from tis, owever, is a consideration of te risk implications of te leveraging of venture capital funded startups. Our paper instead studies te effects of te growt of te venture debt market on startup outcomes. Secondly, our paper contributes to te broader literature on te financing of growt startups. Empirically, Kortum and Lerner (2000), Hirukawa and Ueda (2011), Nanda and Rodes-Kropf (2013), and Kerr et al. (2014), sow te effect of different types of equity-based venture capital on firm level outcomes. Tis paper, on te oter and, documents a different mecanism for accessing financial markets and tus, a different set of incentives for investors and entrepreneurs. On te teoretical side, our paper igligligts a new cannel troug wic staged financing, and in particular, venture debt, can be optimal. In contrast to te large literature wic provides a role for staged financing (e.g., Bergemann and Hege (1998), Neer (1999), Casamatta (2003)), our model sows tat firms may prefer staged financing in order to reduce dilution, aligning te entrepreneur s incentives wit te firm. Te remainder of te paper is organized as follows. We present te model and develop testable empirical predictions in Section 2. Section 3 describes data sources and sample construction, wile section 4 presents te main empirical results. Section 5 concludes te paper. 4

6 2 Model 2.1 Model setup Tere are tree dates, t {0, 1, 2}. A firm owns a risky asset wic pays a cas flow γy at te end of date two. At te start of eac date, te firm must invest X t ; if it fails to do so for any t, te firm suts down (i.e., Y = 0). Oterwise, Y > 0. For instance, suc investments may be required to pay employees or produce for orders. Te asset is initially of unknown quality. If te initial investment (X 0 ) is made, te quality of te asset is revealed. If te intermediate investment (X 1 ) is made, te firm can coose its going-to-market strategy, wic determines te distribution of γ. Prior to te terminal investment (X 2 ),γ is realized. Te firm is initially owned by (i) a risk-neutral venture capitalist and (ii) a risk-neutral entrepreneur. Te firm as no debt outstanding and te venture capitalist owns a fraction θ of te firm s equity. Te entrepreneur as no wealt (outside of is equity stake in te firm) and no labor income. As a result, te capital required to make eac investment must be raised from (outside) risk-neutral investors. 7 Te price of eac claim is set suc tat outside investors breakeven in expectation, conditional on te information available on tat date. 8 Tis initial venture capitalist is responsible for all financing decisions. At eac date, er objective is to maximize te expected payoff from er equity claim, V t. Any equity issued by te firm is dilutive (of all existing owners) and we denote te fraction of te firm sold at eac date by α t. To igligt te potential role of venture debt, we allow te firm to issue one-period straigt debt (wit face value F ) at date zero. 9 Wile te firm generates no cas flows, tis venture debt is backed by te promise of equity issuance in te next period. If te firm is unable to repay te debt owed at date one, te asset value goes to zero. 10 At dates one and two te venture capitalist raises te required capital as long as it is less tan te expected value of te ongoing concern. At date zero, owever, se cooses bot (i) ow muc capital to raise and (ii) ow to raise it in order to maximize [ 2 ] V 0 θ (1 α 0 ) E (1 α 2 ) γy p 0. (1) j=1 We assume tat te venture capitalist as two options: se can raise X 0 + X 1 (wic we will 7 Future equity capital could also come from te inside venture capitalist, but for ease of exposition, we focus on tis setting. 8 Tis is equivalent to assuming (i) competitive capital markets and (ii) a perfectly elastic supply of te risk-free asset. 9 It is witout loss of generality to assume tat any capital raised at date two is via equity. 10 Tis assumption is not necessary but is made for tractability - te intuition for our results olds as long as tere is some liquidation cost in bankruptcy. 5

7 call upfront financing) or se can raise only X 0 and delay te financing of X 1 until after te firm s quality is revealed (wic we will call staged financing). Tis revelation of quality is consistent wit te notion of reacing (or failing to reac) certain milestones, common in start-up financing objectives. If te venture capitalist cooses staged financing, se must also coose wat fraction of te initial capital to raise from equity and venture debt investors. 11 Te entrepreneur is responsible for coosing te firm s strategy, wic is unobservable. In wat follows, we focus on ow te firm s strategy can affect te riskiness of its cas flows. For tractability, we assume γ + δ wit probability τ γ = γ wit probability p 1 2τ. (2) γ δ wit probability (1 p 1 ) + τ Te parameter γ can be interpreted in many ways in our model, including a pricing multiple (e.g. price-to-sales), te fraction of te market obtained by te firm, even te likeliood te firm is able to successfully exit. We refer to p 1 as te quality of te firm - as p 1 increases, te expected value of te asset increases. On te oter and, an increase in τ, were τ [0, τ ], captures te riskiness of te firm s strategy: extreme realizations of γ (bot good and bad) are more likely. 12 Wile risk-neutral, we assume tat te entrepreneur receives some non-pecuniary utility over continuation, i.e., if Y > 0. We model tis simply, so tat te entrepreneur cooses τ to maximize A 1 E [(1 α 2 ) γy p 1, τ] bp [Y > 0 p 1, τ]. (3) were A 1 (1 θ) 1 j=0 (1 α j) is te entrepreneur s current stake in te firm and b > 0 parameterizes te level of continuation utility relative to is financial gains. Tis non-pecuniary utility is a source of potential misalignment between te entrepreneur and te venture capitalist s incentives. Furter, we note tat because te entrepreneur s distribution coice is unobservable it is not contractible. 13 As a result, te venture capitalist must use is financing decision, and its impact on te entrepreneur s stake in te firm, to influence te action taken by te entrepreneur. Finally, we assume tat firm quality is binary: wit probability q te asset is ig-quality 11 We assume tat te firm cannot repurcase equity at date zero (F X 0 ). 12 An increase in τ is a mean-preserving spread wit respect to te distribution of γ. On te oter and, as we detail below, suc risk-taking (weakly) increases te expected value of te firm. 13 We take as given tat te entrpreneur cannot be relieved of er role for instance, se may possess unique uman capital, specific to te firm s asset. 6

8 (p 1 = p ), oterwise it is low-quality (p 1 = p l < p ). 14 We let p 0 qp + (1 q) p l be te expected quality of te asset. To close te model, we note tat, excepting γ and p 1, all exogenous variables are known before date zero. 2.2 Optimal Issuance Policy In wat follows, we work recursively troug te optimal issuance policy. At eac date, we assume tat te firm was able to successfully finance te previous investments; oterwise, no actions would be necessary Date Two In order to raise sufficient capital for investment, te firm must sell a fraction, α 2 = X 2 γy, (4) of te firm s equity. In order for tis to be feasible (α 2 1), it must be te case tat γ γ X 2 Y. (5) In our setting, γ denotes te final milestone te firm needs to acieve in order to successfully raise capital and realize te asset s terminal value Date One Optimal Strategy Knowing tat is claim is wortless unless γ γ, te entrepreneur cooses te firm s strategy to maximize A 1 E [(1 α 2 ) γy p 1, τ] bp [ γ γ p 1, τ ]. (6) To igligt ow venture debt (troug its impact on dilution) affects te firm s strategy, we assume tat γ δ < γ < γ. 15 Wile te probability of successful exit is decreasing in te level of risk P [ γ γ p 1, τ ] = p 1 τ te expected value of is stake is actually increasing in τ, because 14 To ensure tat all probabilities are non-negative, let τ < p l If γ δ γ, ten te manager is indifferent wit respect to te coice of τ: te expected value of er claim is constant and se faces no risk of failure. Similary, if γ < γ, ten te manager always cooses to maximize te firm s riskiness: bot firm value and te probability of success are strictly increasing in τ. 7

9 E [(1 α 2 ) γy p 1, τ] = P [ γ γ p 1, τ ] E [(1 α 2 ) γy p 1, τ, γ γ] (7) = p 1 [ γy X 2 ] + τ δy ( γy X 2 ). (8) }{{} >0 Risk-taking increases te expected value of equity because te firm faces a tresold for financing in te next period. Toug te probability of itting tat tresold falls wit risk, it is outweiged by te increase in te value of equity, conditional on success. 16 Tus, te entrepreneur faces a tradeoff. If te entrepreneur eld no equity, e would always coose te lowest risk strategy. If e derived no utility from te firm s survival (i.e., b = 0), ten e would always coose te riskiest strategy. Tis intuition is generalized in te following lemma. 17 Lemma 1. Te entrepreneur optimally cooses te riskiest strategy (τ (A 1 ) = τ ) if and only if A 1 Oterwise, se optimally sets τ (A 1 ) = 0. b δy ( γy X 2 ) b. (9) Since all oter parameters in (9) are primitives of te model, we abuse notation and let τ (A 1 ) denote te entrepreneur s optimal coice of risk. Note tat te entrepreneur s cutoff for coosing wic strategy to take ( b) does not depend upon weter te firm is ig (p 1) or low (p l 1) quality. On te oter and, te realization of tis information can impact te size of te entrepreneur s stake, A 1, if te firm cooses to stage its financing. For instance, investors wo learn tat te firm is low-quality will demand a iger stake in te firm (in excange for teir investment), wic increases te entrepreneur s dilution. From te venture capitalist s perspective, risk-taking is always valuable, as se suffers no disutility if te firm fails to survive. In order to incent te entrepreneur to coose te risky strategy, owever, requires tat is stake not be too diluted. As a result, we turn our focus to te impact of te firm s financing coices on te entrepreneur s stake in te firm. Tere are two cases to consider: upfront financing and staged financing. 16 Te value of equity is convex in γ - as a result, a mean-preserving spread over te distribution of γ increases te expected value of equity. 17 We assume tat if te entrepreneur is indifferent between two levels of risk, e cooses te level wic maximizes te value of te firm. 8

10 Upfront Issuance If te firm as already raised te capital necessary for investment (X 1 ) as part of te initial round of financing, tere is no issuance decision to be made and no furter dilution occurs (α 1 = 0). Let A 0 1 denote te entrepreneur s stake wen tere is upfront financing. We note tat learning te quality of te firm at date one does not alter te entrepreneur s stake (A 0 1). Ten, by Lemma 1, if te firm is financed upfront, te entrpreneur eiter (i) always takes risk (A 0 1 b) or (ii) never takes risk (A 0 1 < b). We will return to tis point wen we consider te initial financing decision. 18 Staged Finance If te venture capitalist initially cose to raise only X 0, ten se must now raise enoug capital to make (i) te additional investment (X 1 ) as well as (ii) sufficient funds to repay te venture debt (if any) issued at date zero. 19 Tus, α 1 (p 1 ) = X 1 + F E [(1 α 2 ) γy p 1, τ (A 1 (p 1 ))] Investors, knowing tat te size of te entrepreneur s stake affects ow muc risk se takes, account for tis wen valuing teir investment in firm. For example, as te required financing needs (X 1 + F ) grow, so must te fraction of te firm sold to new investors. Tis decreases te entrepreneur s stake (A 1 ). If A 1 falls sufficiently, te entrepreneur opts for te low-risk (low-value) strategy. Since te size of te entrepreneur s stake (A 1 ) is increasing in te quality of te firm, p 1, tis intuition creates a clear link between firm quality and firm strategy, as summarized in te lemma below. Lemma 2. Wit staged financing, tere exists a tresold p e suc tat if p 1 p e, te entrepreneur picks te ig-risk strategy, i.e. τ (A 1 (p 1 )) = τ ; oterwise, e opts for te low-risk 18 Te realization of te firm s quality (p 1 ) does impact weter or not it is optimal to (i) invest X 1 and move on to te next stage or (ii) distribute X 1 to te firm s equityolders.te venture capitalist closes down te firm and returns X 1 if se fails to breaks even, i.e., if (12) E [ (1 α 2 ) γy p 1, τ ( A 0 1)] X1 < 0. (10) Te entrepreneur would coose to invest (and keep te firm as a going concern) as long as ( [ b p1 τ ( ) A1)] 0 E [ (1 α 2 ) γy p 1, τ ( A 0 1)] X1 A 0 1. (11) Equation (11) says tat te entrepreneur is willing to invest in a negative NPV project as long as te nonpecuniary benefit of survival is sufficiently ig. Tere is a literature wic explores ow tis impacts start-up financing. We abstract away from tis conflict to igligt a new cannel troug wic staged finance (and particularly, venture debt) is optimal. 19 Recall tat we are restricting te firm to equity issuance at date one. 9

11 strategy, i.e., τ (A 1 (p 1 )) = 0. Tus, unlike wit upfront financing, wen te firm cooses to stage its capital raising, te realized quality of te firm can alter te firm s strategy (and terefore, its expected value). In order for te firm to successfully raise capital at date one, it must be te case tat α 1 (p 1 ) 1. If tis tresold were reaced, te entrepreneur would be fully diluted and so cooses te low-risk strategy, implying tat 20 p 1 X 1 + F [ γy X 2 ] p (13) As at date two, tis sould be interpreted as a milestone te firm must reac in order to successfully issue equity at date one. Equation (13) igligts one potential cost of debt - its issuance at date zero may preclude te entrepreneur from receiving financing at date one. Finally, as we sow in te proof of Lemma 2, it is not necessarily te case tat p e p: engaging in te risky strategy may be necessary to secure financing (due to te increase in expected value te risky strategy generates) Date Zero If te venture capitalist cooses upfront financing, ten investors breakeven in expectation if α 0 = X 0 + X 1 E [(1 α 2 ) γy α 0 ]. (14) Ten, wit upfront financing, te entrepreneur cooses te risky strategy (regardless of asset quality) as long as A 1 0 = (1 θ) (1 α 0 ) b, i.e., X 0 + X 1 p 0 [ γy X 2 ] + τ [δy ( γy X 2 )] 1 b (1 θ) [δy ( γy X 2 )]. (15) and te firm is able to obtain upfront financing as long as α 0 1, i.e. X 0 + X 1 p 0 [ γy X 2 ] 1 (16) To make clear our teoretical predictions, we will make use of tis observation and utilize te following definitions: 20 In order for te entrepreneur to coose te risky strategy e must still own some fraction of te firm s equity. 10

12 A low-value firm cannot obtain upfront financing. A mid-value firm can obtain upfront financing but pursues te low-risk strategy. A ig-value firm can obtain upfront financing and pursues te ig-risk strategy. Moreover, as te following proposition makes clear, upfront financing is always preferable to staged financing wen (15) olds. Proposition 1. A ig-value firm raises te capital required to reac te next stage in one round, i.e., utilizes upfront financing. If staged financing induces te entrepreneur to coose te ig-risk strategy, regardless of asset quality, ten te venture capitalist is indifferent between te two types of financing. 21 example, if te initial investment in te firm (X 0 ) is low, staged financing performs as well as upfront financing. However, tis will not always be te case. By delaying some portion of te capital raise until date one, te venture capitalist runs te risk tat te asset is revealed to be low-quality. In tat state of te world, te entrepreneur is more diluted tan if te capital ad simply been raised upfront. If te required investment increases sufficiently, te entrepreneur will be so diluted in te low-quality state tat e will opt for te low-risk strategy. lowers te venture capitalist s expected return wit staged financing making upfront financing preferable. Proposition 2. A low-value firm always prefers staged financing. A mid-value firm prefers to utilize staged financing as long as (1) capital can be raised wen te asset is revealed to be low-quality (p l p) or (2) te ig-quality asset is sufficiently valuable (p p ) and te low-quality asset is not too valuable (p l p l ), were p, p l are defined in te proof. Suppose te entrepreneur cooses te low-risk strategy wit upfront financing te firm is mid-value. Ten, as argued above, te entrepreneur wose asset is revealed to be low-quality will do te same wit staged financing. But wat appens if te asset is revealed to be igquality? In tis case, staged financing reduces dilution relative to upfront financing investors are willing to pay more for any equity issued wen tey know te asset is ig-quality. If te information revealed about te asset is good (i.e., if p p e ), te entrepreneur will coose te ig-risk strategy, making staged financing strictly preferable. Moreover, if te information revealed about te asset is sufficiently good (i.e., if p 21 In bot cases, te venture capitalist earns θ [qv τ matter (i.e., Modigliani-Miller olds). + (1 q) V τ l For Tis p ) and te value of financing a X 0 X 1 ]: capital structure does not 11

13 low-quality asset isn t too ig (i.e., if p l p l ), ten te venture capitalist will coose staged financing even if te firm must sut down once te asset is revealed to be low-quality. Finally, wit a low-value firm, te venture capitalist cannot obtain upfront financing te firm s unconditional value is negative. Of course, wit staged financing, te firm will also surely sut down at date one if te asset is revealed to be low-quality. On te oter and, investing in a firm revealed to be ig-quality can be profitable at date one. Knowing tis, investors at date zero may finance te firm in te opes tat tis comes to pass. Even wit a low-value firm, it is possible tat te entrepreneur will coose te ig-risk strategy wen date one financing is extended; in fact, suc financing may only be feasible wen tis coice is made (wen p p e, as discussed above). Proposition 3. If te firm can obtain financing at date one, a mid-value firm prefers venture debt, sometimes strictly. Te value of staged financing is tat it can reduce dilution wen te asset is revealed to be ig-quality. In our setting, venture debt amplifies tis effect. By borrowing at date zero, te firm raises (1) less equity wen te firm is valued unconditionally (date zero) and (2) more equity wen te firm is revealed to be ig quality (date one). As te proposition makes clear, in some cases tis amplification is necessary: relying on equity only can leave te entrepreneur wit too little incentive to take risk. As te proof of proposition 3 argues, venture debt is more likely to be necessary at suc a mid-value firm wen 1. required investment (X 0, X 1 ) and initial dilution (1 θ) increase, and 2. gains from risk-taking (δ, τ ) and unconditional asset quality (p 0 ) decrease. All else equal, suc canges make it more likely tat te entrepreneur will coose te low-risk strategy, making venture debt a valuable antidote. We end tis section by summarizing te implications of venture debt for firm outcomes. Corollary 1. Te optimal use of venture debt increases te expected value of te firm, (1) increases te probability of sort-term failure, (2) increases te firm s expected value, conditional on survival, and (3) decreases te firm s dilution if te asset is revealed to be ig-quality. Wit tese predictions in mind, we turn now to a description of te data analyzed. 12

14 3 Data and Descriptive Findings Our data is collected from CruncBase, a crowd-sourced database tat tracks start-ups. 22 CruncBase, wic investors and analysts alike consider te most compreensive dataset of early-stage start-up activity, describes itself as te leading platform to discover innovative companies and te people beind tem. CruncBase was founded in 2005 but include backfill data from te mid-1900s. To address concerns of backfill bias, we limit te sample from 2000 onwards. Te start-up firm caracteristics of interest from CruncBase include: te entrepreneur(s), ig-leveled employees, founding date, current status (ongoing, inactive), and exit outcomes (IPO, acquired, closed). We also ave round level data on eac financing event. Te round level caracteristics include: date of closing, investors name and type (debt, equity, angel, etc.), investment amount, and stage of financing (Series A, B, C, D). For a subset of te rounds, we also ave data on pre-money valuations. CruncBase as many advantages over traditional finance databases suc as VentureOne. One distinct benefit necessary in our context is tat CruncBase collects and aggregates all relevant startup data from te greater Web. If a startup receives Bloomberg press coverage regarding a C-suite employee cange, CruncBase will incorporate tis information automatically. Additionally, CruncBase will timestamp te event. Given tat many startups rarely (and potentially endogenously) self-report closures, tis provides us wit a way to distinguis inactive firms from ongoing firms. We classify any firm tat as no updates witin te last two years as inactive. Te second benefit tat is useful for our analysis is te availability of detailed investor information. Many financing rounds are syndicated, meaning te round as more tan one investor. Wile CruncBase classifies tese syndicated rounds as venture, tis greatly understates te use of venture debt in early-stage financing. Instead of classifying rounds as fully debt or equity, we look at te type of investors and sort investors into debt or equity investors based on teir past portfolio investments. We call any round tat as a known debt investor to be a syndicated debt round. We ceck tat tis is accurate troug qualitative assessments and google searces. Te limitation of our data is tat we do not ave contract-level data on te loans meaning we don t ave information on te interest rates or associated warrants. However, we take comfort in knowing tat te contracts of venture loans are relatively standard across firms. 23 Te main dataset includes 61,667 firms and 135,069 financing rounds during te period 22 ttps:// and ttp://teccrunc.com/ CruncBase. For more information on te use of tis dataset, refer to Wang (2017). 23 Cite needed. 13

15 Table 1 presents te company-level summary statistics. 24 A startup in our sample as on average two rounds of financing, wit te first round occurring approximately tree year after startup founding and 40% of all rounds involving some debt financing. Te total amount of investment received during a startups lifetime is $16.6 million of wic $2 million is from early debt rounds. Consistent wit industry-level estimates of exit rates, 1.7% of startups go troug an initial public offering (IPO), 12.4% are acquired, and 62.9% of te firms are closed/inactive. Table 2 presents te round-level summary statistics broken down by Series. Te Series sow in te different panels is te actual round (for equity rounds or equity-debt syndicate rounds) or te would-be round for debt financing ad te firm issued a equity round. Dilution Proxy is Current Investment divided by te sum of current + te immediate prior investment round. Te pre-money valuation, wic is sparsely reported in CruncBase, is te valuation accruing to founders and prior investors as implied by te valuation of te current investment. Burn Rate Duration is te number of days forward until te next financing. 4 Empirical Analysis First, we examine te decision of a startup to take on venture debt. Proposition 3 of te model states tat venture debt is more likely to be necessary wen 1. te required investment and initial dilution increases, and 2. te gains from risk-taking and te unconditional quality decreases. In table 3, we present te results of a logit regression and te marginal effects of te round-level caracteristics on te coice of debt versus equity. Eac column subsamples only to estimate rounds for Series A, Series B, or Series C/D in order to bot control for a startup s milestones and to sow ow te coefficients cange across a startup s lifecycle. In columns 1-3, we find tat te decision to take on debt does increases as dilution increases just as te model predicts. Furtermore, te coefficient increases in magnitude and statistical significance as te startup moves furter along in financing rounds. In column 4, we find tat a lower Series A pre-money valuation leads to an increase in te probability tat a startup takes on debt instead of equity. However, te coefficient flips in te later Series C round (column 6). Tese results suggest tat in earlier rounds, wen uncertainty is iger, te inability to reac certain milestones leads to venture debt. On te oter and, venture debt is functioning like it s traditional counterpart in later rounds wen te startup as more consistent cas flows. 24 All tables are found in te Appendix. 14

16 Next, we look at te effect of venture debt on firm outcomes. In table 4, we regress burn rate, defined as te realized time until te next round in days, on a dummy variable for weter te current round is financed via debt or equity. We control for te amount of current investment since a iger investment amount sould by definition provide a longer runway for te firm. We find tat in early rounds (Series A), taking on debt decreases te amount of time between financing events, consistent te idea tat debt repayment requires new financings sooner due to a iger burn rate. Put differently, debt is extending te runway of a firm by providing capital wen te burn rate is iger. In tables 5 and 6, we focus our attention on corollary 1 of te model, restated below. Corollary 2. Te optimal use of venture debt increases te expected value of te firm, (1) increases te probability of sort-term failure, (2) increases te firm s expected value, conditional on survival, and (3) decreases te firm s dilution if te asset is revealed to be ig-quality. In table 5, te dependent variable is an indicator for te startup closing. Te estimation is logit, reporting te marginal effects effect for a cange in te probability of closing. Te variables of interest are te total money raised by te startup (Log Total Investment), te total money raised in a debt or debt syndicated round, and te total money raised in a debt or debt syndicated round prior to a Series B equity round. Focusing on column 1, te coefficient on Log Total Investment is negative and statistically significant at te 1% level. Consistent wit our intuition, tis implies tat raising more capital leads to a lower probability of startup failure. Te coefficient on Log Debt Investment is also negative and statistically significant, suggesting tat debt extends te runway, tus delaying creative destruction in preference for risk. In column 2, we disaggregate debt investment into debt investment before and after Series B. Interestingly, te coefficient on Log Debt Investment remains negative and statistically significant, but te coefficient on Log Debt Investment Prior to Series B is positive and significant. A 10% increase in te amount of early-stage debt investment increases te probability of closure by 6%. Te results are consistent wit te model s predictions - wile te optimal use of debt increases te firms expected value and extends te runway, it also increases te probability of sort-term failure. Venture debt provides a lever for te VC to induce risk-taking. In table 6, we sow te effect of debt on oter exit outcomes (IPO, Acquisition, Ongoing), conditional on not closing. Te estimation is multinomial logit, and tus eac estimation as two columns, reporting te marginal effects effect for increasing te probability of exit for outcomes Acquisition or IPO relative to te probability of exit in te offset category of Ongoing. Te independent variable of interest, Debt Round, is te coice of venture debt versus venture equity for eac round. We also control for te current opportunity set by controlling for te 15

17 log of te money raised in te current investment round. Financing round year fixed effects are included. Eac numbered set of two columns subsamples only to estimate exit outcomes as of Series A rounds (column 1), Series B (column 2), or Series C/D (column 3). Across all series, a debt round increases te likeliood of exiting via acquisition relative to ongoing by 7-10%. Conversely, across all series, a debt round decreases te likeliood of an IPO exit relative to ongoing by 1-4%. Wile tese two results seem contradictory at first glance, it is easily acknowledged and reconciled in te model. Te igest-quality firms raises te capital required to reac te next milestone in one round, i.e., utilizes upfront financing, and as less use for debt financing. Tese are exactly te firms tat ave te igest expected value and te most likely to go troug an initial public offering. On te oter and, venture debt amplifies te value of staged financing for mid-value firms. Venture debt tus increases te probability of a positive outcome (acquisition) for tese firms. In sum, our empirical results indicate tat te startup landscape is fundamentally altered by te introduction of venture debt. Firms tat take on leverage experience more downside (closures) along wit more upside (acquisitions). 5 Conclusions Our results demonstrate tat te introduction of venture debt as potentially dramatic implications for early-stage firms. Wile suc issuance may increase firm value and allow firms to obtain oterwise unavailable financing, it can carry wit it significantly more risk, bot strategic and financial. We find empirical evidence consistent wit our teoretical predictions and, in particular, te role venture debt plays in extending te firm s runway. Given te recent growt in te venture debt market, and its prevalence across te innovation economy, we ope to build on tis researc to study its implications for te real economy. 16

18 References Bergemann, D., Hege, U., Venture capital financing, moral azard, and learning. Journal of Banking & Finance 22, Casamatta, C., Financing and advising: Optimal financial contracts wit venture capitalists. Journal of Finance 58 (5), Cornelli, F., Yosa, O., Stage financing and te role of convertible securities. Review of Economic Studies 70, Daiya, S., Ray, K., Staged investments in entrepreneurial financing. Journal of Corporate Finance 18, de Rassenfosse, G., Fiscer, T., Venture debt financing: Determinants of te lending decision. Strategic Entrepreneursip Journal 10, González-Uribe, J., Mann, W., New evidence on venture loans. Working Paper. 1 Hirukawa, M., Ueda, M., Venture capital and innovation: Wic is first? Pacific Economic Review 16 (4), Hocberg, Y., Serrano, C., Ziedonis, R., (fortcoming). Patent collateral, investor commitment, and te market for venture lending. Jounrnal of Financial Economics. 1 Ibraim, D., Debt as venture capital. Te University of Illinois Law Review, , 1, 4 Kerr, W., Lerner, J., Scoar, A., Te consequences of entrepreneurial finance. Review of Financial Studies 27, Kortum, S., Lerner, J., Assessing te impact of venture capital on innovation. Rand Journal of Economics 31, Modigliani, F., Miller, M., Te cost of capital, corporation finance and te teory of investment. American Economic Review 48 (3), Nanda, R., Rodes-Kropf, M., Investment cycles and startup innovation. Journal of Financial Economics 110 (2), Nanda, R., Salman, W., Keller, N., Western tecnology investment. HBS No

19 Neer, D., Staged financing: An agency perspective. Review of Economic Studies 66, Repullo, R., Suarez, J., Venture capital finance: A security design approac. Review of Finance 8, Sage, S., Te rise of venture debt in europe. Tec. rep., Winston & Strawn. 4 Tykvová, T., Wen and wy do venture capital-backed companies obtain venture lending? Journal of Financial and Quantitative Analysis 52 (3), Wang, S., Zou, H., Staged financing in venture capital: Moral azard and risks. Journal of Corporate Finance 10,

20 A Proofs Proof of Lemma 1 First, we confirm (7): [( ) ] X 2 E [(1 α 2 ) γy p 1, τ] = τ 1 ( γ + δ) Y ( γ + δ) Y [( + (p 1 2τ) 1 X ) ] 2 ( γ) Y ( γ) Y Note tat te last term in brackets is positive as long as (17) (18) = p 1 [ γy X 2 ] + τ [δy ( γy X 2 )] (19) δy > ( γy X 2 ) (20) X 2 Y > γ δ (21) wic is true by assumption - te firm cannot get financing at date two if γ = γ δ. Rewriting te entrepreneur s objective function yields A 1 (p 1 [ γy X 2 ] + τ [δy ( γy X 2 )]) + b (p 1 τ) (22) Tis is linear in τ, implying a corner solution: τ {0, τ }. Te entrepreneur s utility is weakly increasing in τ as long as wic completes our proof. A 1 [δy ( γy X 2 )] b 0 (23) Proof of Lemma 2 Rewriting (9), te entrepreneur cooses te risky strategy as long as [ ] b α 1 (p 1 ) 1 (24) (1 α 0 ) (1 θ) X 1 + F [ ] E [(1 α 2 ) γy p 1, τ (A 1 (p 1 ))] (25) 1 b (1 α 0 )(1 θ) 19

21 [ X 1 +F b 1 (1 α 0 )(1 θ) ] τ [δy ( γy X 2 )] [ γy X 2 ] p e p 1 (26) Tis tresold exceeds p as long as X 1 + F [ γy X 2 ] τ [δy ( γy X 2 )] τ [δy ( γy X 2 )] [ [ ] b (1 α 0 )(1 θ) [ ] b δy ( γy X 2 ) (1 α 0 )(1 θ) Tis completes te proof. [ X 1 +F b 1 (1 α 0 )(1 θ) X 1 + F [ 1 b (1 α 0 )(1 θ) ] τ [δy ( γy X 2 )] [ γy X 2 ] (27) ] (X 1 + F ) (28) ] (X 1 + F ) (29) Proof of Proposition 1 First, we establis te following lemma regarding te impact of financing. Lemma 3. Holding fixed te entrepreneur s coice of strategy, te venture capitalist is indifferent between stage financing and upfront financing. Proof. To see tis, note tat wit upfront financing se earns in expectation, ( ) X 0 + X 1 θ 1 E [(1 α 2 ) γy α 0 ] = (30) E [(1 α 2 ) γy α 0 ] θ (ψ X 0 X 1 ) (31), were ψ p 0 [ γy X 2 ] + τ [δy ( γy X 2 )]. By te same logic, if te low-risk strategy is cosen, se earns θ (ψ 0 X 0 X 1 ), were ψ 0 p 0 [ γy X 2 ]. To simplify our notation, let V τ s E [(1 α 2 ) γy p s, τ] = p s [ γy X 2 ] + τ [δy ( γy X 2 )] denote te expected value of te (diluted) equity claim, conditional on te asset quality and te entrepreneur s coice of strategy. If stage financing incents te ig-risk strategy regardless of asset quality, te venture capitalist s expected earnings are ( ) X 0 F θ 1 E [(1 α 1 ) (1 α 2 ) γy α 0 ]were (32) E [(1 α 1 ) (1 α 2 ) γy α 0 ] 20

22 E [(1 α 1 ) (1 α 2 ) γy α 0 ] = (q (1 α 1 (p )) V τ + (1 q) (1 α 1(p l )) V τ l ) (33) = ψ (X 1 + F ) = (34) ( ) X 0 F θ 1 E [(1 α 1 ) (1 α 2 ) γy α 0 ] = θ (ψ X 0 X 1 ). (35) E [(1 α 1 ) (1 α 2 ) γy α 0 ] Again, by te same logic, if stage financing incents te low-risk strategy, er expected earningsfinancing are just θ (ψ 0 X 0 X 1 ). Tus, te only effect capital structure as on te expected value of te firm is troug its effect on te entrepreneur s coice of strategy. doesn t matter as long as te entrepreneur takes te same action. Wit tis establised, we can complete te proof. Suppose tat p e p. Ten te entrepreneur ( cooses te risky strategy, regardless of asset quality, as long as 25 1 X ) ( ) 1 + F X 0 F V τ 1 l q (V τ (X 1 + F )) + (1 q) (V τ b l (X 1 + F )) 1 θ. (36) If bot (15) and (36) old, te venture capitalist is indifferent between staged and upfront financing, by Lemma 3. To sow tat tis will not always be te case, we can rewrite te left-and side of (36) as 1 (X 1 + F ) [ψ (X 1 + F )] + (V τ l (X 1 + F )) (X 0 F ) [ψ (X 1 + F )] V τ. (37) Second, wit a little algebra it can be sown tat l (X 1 + F ) [ψ (X 1 + F )] + (V τ l (X 1 + F )) (X 0 F ) [ψ (X 1 + F )] V τ l [X 0 ] ((ψ V τ l > X 0 + X 1 ψ (38) ) (X 1 + F )) [X 1 ] [ψ (X 1 + F )] (ψ V τ l ) < ψf [ψ V τ l ] (39) X 0 + X 1 < ψ (40) were te last inequality obviously olds because te firm is able to obtain financing upfront. Note tat wen we move from te second to te tird inequality te sign stays te same because ψ > V τ l. On te oter and, As a result, tere exist parameters suc tat (15) olds but (36) does not. 26 Under tose conditions, if te firm uses staged financing and te asset is low quality, 25 Tere is less dilution at date one if te asset is revealed to be ig-quality and so we focus on te incentive to take risk in te low-quality state. 26 Using similar steps, it is straigtforward to sow tat, under staged financing, a ig-value firm always cooses te ig-risk strategy wen te asset is ig-quality. 21

23 te entrepreneur cooses te low-risk strategy, wic lowers te expected value of te venture capitalist s claim. Tus, se strictly prefers upfront financing under tese conditions by Lemma 3. To complete te proof, we consider te case wen p e < p. If (36) does not old, ten p l < p, and so te entrepreneur cannot even finance te investment if te asset is low-quality. Furter, by continuity of te diluted equity stake, parameter values exist suc te investment cannot be financed in te low-state (even toug (36) is not violated). Tus, by te same logic, upfront financing remains preferable wen p e < p, sometimes strictly. Proof of Proposition 2 Suppose te firm is mid-value. If te entrepreneur cooses staged financing (and can finance te firm wen it is revealed to be low-quality), ten e cooses te ig-risk strategy wen te asset is ( ig-quality as long as 1 X ) ( ) 1 + F X 0 F V τ 1 q (V τ (X 1 + F )) + (1 q) (Vl 0 (X 1 + F )) b 1 θ. (41) Let ψ 0 p 0 [ γy X 2 ] and ψ 1 p 0 [ γy X 2 ] + qτ [δy ( γy X 2 )]. Note tat ψ 1 is te unconditional expetation of te diulted cas flow at date one - by te above proof, it is easy to sow tat te entrepreneur cannot coose te ig-risk strategy if te asset is low-quality in tis setting. On te oter and, following steps similar to tose found in te proof of Proposition 1, we can sow tat (X 1 + F ) [ψ 1 (X 1 + F )] + (V τ (X 1 + F )) (X 0 F ) [ψ 1 (X 1 + F )] V τ < X 0 + X 1 ψ 1 (42) as long as X 0 + X 1 < ψ 1. But of course tis olds because te entrepreneur can successfully engage in upfront financing, i.e., X 0 + X 1 < ψ 0 < ψ 1. Tus, if te firm receives financing (even wen te asset is low-quality), staged financing creates te possibility of (41) olding, in wic case te entrepreneur cooses te ig-risk strategy wen te asset is ig-quality. Tus, te venture capitalist prefers staged financing, sometimes strictly, by Lemma 3. If te entrepreneur cooses staged financing and cannot finance te firm wen it revealed to be low-quality, e defaults some portion of te time. Knowing tis, e cooses te ig-risk strategy wen te asset ( is ig-quality as long as 1 X ) ( ) 1 + F X 0 qf V τ 1 q (V τ (X 1 + F )) b 1 θ. (43) If tis doesn t old, ten te venture capitalist strictly prefers upfront financing by Lemma 3. First, we sow tat it is feasible for (43) to old even toug (15) does not. 22

24 ( 1 X ) ( 1 + F V τ 1 X 0 qf q (V τ (X 1 + F )) ) = 1 qx 1 + X 0 qv τ (44) Ten it is possible for te entrepreneur to coose te ig-risk strategy (wit staged financing) as long as qx 1 + X 0 qv τ < X 0 + X 1 ψ (45) X 0 V τ l < qx 1 (p p L ) [ γy X 2 ] (46) It is clear tis olds if X 0 = 0, for example. Second, te venture capitalist would prefers staged financing over upfront financing as long as: θ (ψ 0 X 0 X 1 ) < θ (q [V τ (X 1 + F )] (X 0 qf )) (47) (1 q) ( V 0 l X 1 ) < q (τ [δy ( γy X 2 )]). (48) wic clearly olds if X 1 = Vl 0. Under tese assumptions, te venture capitalist can still raise capital because X 0 + X 1 = Vl 0 < (1 q)vl 0 + qv 0. Tus, conditions exist under wic te venture capitalist prefers staged financing to upfront financing, even toug se cannot raise capital wit a low-quality asset. Now, we formally establis te tresolds suc tat staged financing is preferable. First, we note tat 1 qx 1 + X 0 qv τ p b 1 θ (49) qx 1 +X 0 q(1 b 1 θ) τ [δy ( γy X 2 )] [ γy X 2 ] p (50) Second, we sow tat if tis olds, tere exists an upper bound on p l suc tat te venture capitalist prefers staged financing to upfront financing: θ (ψ 0 X 0 X 1 ) θ (q [V τ (X 1 + F )] (X 0 qf )) (51) p l X 1 (1 q) + qτ [δy ( γy X 2 )] (1 q) [ γy X 2 ] p l (52) Finally, we sow tat te low-value firm prefers staged financing, sometimes strictly. In tis 23

25 case, we need to establis tat it is possible for te firm to raise capital in te ig-quality state, even toug X 0 + X 1 > ψ 0. Te venture capitalist can raise capital at date one if te firm is revealed to be ig-quality as long as p [ γy X 2 ] X 1 + F and at date zero as long as α 0 1, i.e. X 0 qf E [(1 α 1 ) (1 α 2 ) γy α 0 ] (53) X 0 qf q [p [ γy X 2 ] (X 1 + F )] (54) X 0 q [p [ γy X 2 ] X 1 ] (55) Suppose tat X 0 = F = 0 and let X 1 = p [ γy X 2 ]. Ten te entrepreneur can raise te capital necessary if te asset turns out to be ig-quality. Moreover, it is still te case tat X 1 = p [ γy X 2 ] > p 0 [ γy X 2 ] = ψ 0. Proof of Proposition 3 We will start by focusing on te setting in wic te firm is mid-value and te firm can raise capital in te low-state. Let ( 1 X ) ( ) 1 + F X 0 F V τ 1 χ. (56) ψ 1 (X 1 + F ) Ten, we want to sow tat issuing venture debt can induce te entrepreneur to coose te ig-risk strategy wen staged equity financing was insufficient to get (9) to old. Specifically, we want to sow tat χ F > 0. ( χ F = 1 ( 1 = V τ ) ( ) ( X 0 F 1 ψ 1 (X 1 + F ) V τ ) [( 1 1 X 1 + F V τ ) ( ) (ψ1 (X 1 + F )) + X 0 F (ψ 1 (X 1 + F )) 2 ) ( )] X 0 F + (V τ ψ 1 (X 1 + F ) (X X1 + X 0 ψ F )) (ψ 1 (X 1 + F )) 2 We want to sow tat te term in brackets is less tan zero, tat is (57) (58) ( ) X 0 F 1 < (V τ ψ 1 (X 1 + F ) ( ) ψ1 (X 1 + X 0 ) ψ 1 (X 1 + F ) ( (X ψ1 (X 0 + X 1 ) 1 + F )) (ψ 1 (X 1 + F )) 2 < (V τ (X 1 + F )) ( ψ1 (X 0 + X 1 ) (ψ 1 (X 1 + F )) 2 ) ) (59) (60) 24

26 Note tat ψ 1 > ψ 0 > X 1 + X 0 (upfront financing is feasible) and ψ 1 > p l [ γy X 2 ] > X 1 + F (date one financing wit a low-quality asset is feasible). Tus, we can rewrite te inequality above, ( V τ 1 < wic of course olds because ψ 1 = qv τ (X 1 + F ) ψ 1 (X 1 + F ) + (1 q)v 0 l < V τ ), (61) and bot are greater tan X 1 + F (date one financing wit te ig-quality asset is feasible). Tus, χ F > 0. On te flip side, issuing venture debt makes it less likely tat te firm can obtain financing if it owns a low-quality asset. If financing fails wit a low-quality asset, ten we are in te second case of proposition 2; ere, venture debt does not slacken te incentive compatibility constraint and so if low-quality financing fails, no venture debt is utilized. 27 Finally, to establis under wat conditions we are more likely to observe venture debt, it is straigtforward to see tat χ χ χ X 0, X 1, X 2 < 0, wereas χ, χ, χ δ Y τ, χ, χ γ p 0 > 0. On te oter side, b is always decreasing in δ (consistent wit te partial effects on χ), but can increase in Y, γ, X 2. Proof of Corollary 1 Te optimal use of venture debt increases te expected value of te firm because it induces te entrepreneur to take risk if te asset is revealed to be ig-quality. At date two, tis (1) increases te likeliood of failure (unable to raise funds) and (2) increases te expected value of te firm, conditional on successfully raising capital. Te value of venture debt is tat it decreases dilution if te asset is revealed to be ig-quality. 27 Tis will not necessarily old under more general assumptions about te distribution of p 1. 25

27 Figure 1: Type of Financing Rounds by Funding Year Depicted are te frequency of financing rounds by type {venture debt, venture equity, angel financing} based on year of funding round..

28 Figure 2: Exits by Firm Founding Year Depicted are te firm exits {Ongoing, Acquisition, IPO, and Closing} as a percent of firms starting in te year on te x-axis.