A pecking order of venture capital exits What determines the optimal exit channel for venture capital backed ventures? Carsten Bienza,b This version: March 1, 2004 Abstract I develop a model of exits from venture capital backed companies based on moral hazard. The focus is on two alternatives: a trade sale and an IPO. The alternatives differ in their governance structure. The model shows that highly profitable companies that need few oversight will go public, while less profitable companies that require more control will be sold. This suggests that the common notion that IPOs are generally more profitable than sales is wrong and observed returns suffer from a measurement bias. Finally, I present empirical evidence that IPOs have indeed higher rates of return than trade sales. Keywords: venture capital, IPO, corporate governance, moral hazard, disinvestment, exit strategy. JEL classification: G24, G32, G34.
a
Address: Graduate Program ”Finance and Monetary Economics”, Mertonstraße 17-21, Uni Postfach 77, 60054 Goethe-University Frankfurt, Germany. Tel : +49 (0) 69 79828943. Email :
[email protected]
b
The paper was presented at the CFS Workshop on Venture Capital and New Markets and at seminars in Frankfurt. I would like to thank Douglas Cumming, Christian Laux, Dima Leshinskii, Jan Pieter Krahnen, Ralf Oesterle, Mark Wahrenburg and Uwe Walz for valuable comments. All errors remain mine. Financial support by German Research Foundation (DFG) is gratefully acknowledged.
1
Introduction
Private equity investments by venture capitalists (VC) in young companies have been proposed as a method for start-up financing. Due to the specifics of venture capital, all these investments are of a temporary nature. Venture Capitalist are specialized intermediaries, who provide money, advice and monitoring. Once companies have reached a certain maturity, VCs loose their comparative advantages from specialization. Also, VCs raise money via closed-end funds that are dissolved after ten years. The exit decision is therefore important and demands our attention. There exist several different channels. By far the most important are IPOs and trade sales. Since each channel is characterized by a different allocation of issuing proceeds and control rights, the choice of the exit channel is important. In a trade sale, normally the whole firm is sold to an outside investor, namely a large cooperation. The new owner usually replaces the incumbent management. Frequently the founders are part of this incumbent management team. In an IPO, the founders do not completely sell their shares and are not replaced as managers. The VC sells off his complete holdings, although not all is sold immediately after the IPO. Empirically, there exists a pecking order of exit channels: IPOs yield higher returns than sales. See Bygrave et al. [3] and for such a result. I provide an explanation for this pecking order: Highly profitable companies go public whereas less profitable companies are sold off via trade sales. This result suggests that the common notion, for example stated by Smith et al. [12], that an initial public offering (IPO) is always more profitable than a sale may be misleading, as it is based upon a selection bias. An additional result is that the right to decide about a specific exit-channel is valuable for both the venture capitalist and the entrepreneur. Finally, I present empirical evidence that indeed IPOs have higher rates of return than trade sales. The main contribution of this paper is to provide a rationale for this pecking order. The explanation relies on the following mechanism: In very successful firms the entrepreneur’s equity stake in the venture assures incentive compatibility. In firms that are less successful, 1
private benefits matter more in relation to profits than in more successful firms and thus corporate governance becomes more important. Therefore firm characteristics determine for every type of firm a different optimal exit channel. The model then predicts a unique cut-off value such that firms whose profitability lies above this value go public while less successful ones are sold. To investigate this issue, I develop a simple moral hazard model of the going public process, where the moral hazard occurs after the IPO. Therefore, post-IPO control of the agent matters. However, the various exit channels result in different allocations of control and cash-flow rights to the principal. These differences affect the principal’s incentives to actually monitor the agent. The problem at hand is different from standard models of the going public process, as there are two parties involved, with differing interests, not one. My basic model is based upon Grossman and Hart’s [6] and Innes’s [8], its main difference from Innes is the the assumption of double limited liability. The model combines two different aspects of the corporate finance literature. The first strain are several studies from the venture capital literature that look at exit decisions. Aghion et al. [1], Bascha and Walz [2], Hellmann [7], Schwienbacher [10], and Tykvov´a [11] are among these papers. They look at various aspects of the VC’s exit decision, however, they do not look at control issues related to the exit decision. The paper closest to mine is Schwienbacher who studies the relation between innovation, product market competition and the optimal exit channel. However, Schwienbacher does not look at control issues that might affect the optimal decision to go public. In that sense the papers should be seen as complements that analyze different aspects of the optimal exit choice. Aghion et al. look at the interaction of the liquidity and the optimal amount of monitoring. Tykvov´a looks at the optimal exit timing while both Bascha/Walz and Hellmann look at the effect of different securities that the VC holds on the exit decision. The second strain are papers that look at the trade-of between the exit decision and control. However, these papers do not model the interaction between more than one owner. Among these are Chemmanur et al. [4], Maug [9], and Zingales [13]. Zingales looks at an owner-manager who wants to decide whether to go public or not. However, he does 2
not consider a possible loss of control by the manager, and as Chemmanur et al., who look at a similar problem, is not concerned with the strategic interaction between different owners. To my knowledge, this paper is the first one to look at the interaction between several parties that own a firm, the exit decision of at least one party and the possible loss of control from that decision. This paper is organized as follows: The next section outlines the model and the third section characterizes the decisions for one of the two exit channels taken in equilibrium. Comparative statics are derived in the fourth section, while section five presents extensions, the sixth empirical evidence, and the last section concludes.
2
The model
In this chapter I will describe the model and it’s basic assumptions. Consider a company that has two current owners, an entrepreneur and an investor. Both hold a claim to some fraction of the value of the firm. The investor, whom I will call VC, has decided to exit the firm. The question to answer is, which of two possible exit channels maximizes firm value. The first exit channel is an IPO. I define an IPO to be a situation where the entrepreneur retains an equity stake in the firm after the exit. A second possible channel is a sale. I define this to be a situation where, after the exit, a new owner will own the whole firm and both of the original owners sell out. Figure 1: Time line of the basic model
Time: There are three periods in the model. In the first period all players become perfectly informed about the quality of the company, including the possible revenues of the 3
company and the probabilities that these revenues occur. After this information has been received, an exit channel is chosen. Following the exit, the new owner, whom I will call the principal, designs an optimal wage contract for the agent and decides whether to replace management or not. In the second period the new principal must supervise the agent. After the supervision, the agent decides about his optimal action while in the third returns are realized. The timing of the model is shown in figure one. Exit choice: The choice of the exit channel is of interest here. I assume that before the game started, the VC and E write a contract that allocates the right to choose the exit channel to one of the two. The allocation depends upon (verifiable) company quality. This information becomes known in period zero. The contract is assumed to maximize the total surplus to both parties. A value maximizing contract is defined to be one that looks at the combined monetary and nonmonetary revenues from the exit. The right to decide about the exit channel is separated from the question of who will act as principal in the new firm. I assume that after the exit, the new owner will act as principal, not the VC or the entrepreneur. Information: The information structure in the model is the following. In the beginning all parties have symmetric information. All parties observe the principal’s control effort, while the agent’s action in period 2.5 is unobservable. The final revenues of the project are drawn from a probability distribution determined by the agent’s action. However, ex-ante, all possible distributions that the agent may choose from are common knowledge. Players: Players in the game are the VC, the entrepreneur, owners and possibly an external manager. All parties are risk-neutral and are protected by limited liability. I assume that the entrepreneur is completely invested in the firm, so he has zero additional wealth, while this is not necessarily true for the other parties. Nature: In order to keep the model as simple as possible, there are only two states of nature, low and high. Each state is associated with a nonnegative outcome for the company, xl , and xh , where the outcome in the high state is larger than the return in the low state xh > xl . X, where xn ∈ X and n = l, h, is the set of all possible outcomes. In each state of nature xn can be interpreted as the revenues the company receives from its operations. 4
I set xl = 0 for simplicity. Each state of nature has different probabilities of occurring p(xn |a) given the action of the manager a. Agent’s actions: The agent’s actions are unobservable. There are two possible actions a: He can work hard/not shirk a = h, or he can not work/shirk a = l. If he works hard, he has disutility from working, so his private benefits of control, b(a, i), are lower than if he is lazy: b(l, i) > b(h, i), where i denotes the level of monitoring exerted by the principal. The reason why the agent should work hard is that hard work leads to higher expected returns for the firm than being lazy. I use the following construct: I assume that the agent derives some private benefits of control from being the boss in the firm. From these private benefits, disutility from working hard must be subtracted. This leads to lower, but still positive levels of private benefits of control from working hard. However, being lazy would be preferred by the agent in terms of control benefits. Principal’s action and possible agent types: The variable i denotes the intensity of monitoring by the principal. For simplicity, i can take two values, monitoring or no monitoring. I denote these by i = i∗ and i = 0. Monitoring causes costs for the agent. I assume that the principal’s cost function to be strictly greater than zero c(i|t) > 0 if he monitors the agent, but zero monitoring causes zero costs (c(0) = 0), where t stands for the agent’s type. I assume that there are two types of agents t, entrepreneurs e and professional managers m. Their only difference lies in the cost of monitoring them. Monitoring the entrepreneur is very expensive, while monitoring the manager is much cheaper. In fact, I assume that controlling the entrepreneur is not feasible. Since the entrepreneur is an insider in the firm, he has a far better knowledge of the firm than any outside director. This constitutes an enormous informational advantage. Therefore, (1)
p(xh |h)xh − (b(l, i∗ ) − b(h, i∗ )) < c(i∗ |e)
is assumed to hold for the entrepreneur, while it does not hold for the manager. The effects of monitoring are the following: the agent’s disutility from working hard falls close to zero, so his private benefits from working hard are nearly the same as those from being
5
lazy. Then the difference in private benefits has the following form: (2)
∆b(0) = b(l, 0) − b(h, 0) > ∆b(i∗ ) = b(l, i∗ ) − b(h, i∗ )
Control is indivisible and needs to be exerted by at least one shareholder to be effective. As mentioned above, the reason why the agent should work hard is, that hard work leads to higher expected returns for the firm than being lazy. Since revenues are fixed, probabilities have to differ. Formally, p(xh |h) > p(xh |l), so the probability for a high return is greater for a hard working agent than for a lazy agent. Again for simplicity I assume that p(xh |l) = 01 . Also, a frequently used shortcut is ph = p(xh |h), which is the probability that the good state of nature results, given that the agent works hard. IPO costs: I assume that an IPO causes fees f > 0 that are paid for at the IPO. A sale does not cause costs, so f may be understood as the difference in costs. For the model to be interesting, the following parameter restriction on IPO fees has to hold, otherwise IPOs would never occur in equilibrium: (3)
f < ph (whsale (m, i∗ ) − whIP O (e, 0)) + c(i∗ )
where whsale (i∗ |m) is the wage in case of a high return, given that the exit channel is a sale, the agent type is a manager and the control intensity is i∗ . Utility functions: The percentage of the company held by the principal after the exit is described by the fraction α, while the agent’s percentage is βt where t again is the agents type. Together α + β ≤ 1, where in a sale alpha = 1. The wage payed to the agent given that revenues xn are realized is wn (xn ). The principal’s utility function can then be written as: " # X (4) UP = α p(xn |a)(xn − wn (xn )) − c(i), xn ∈X
while the agent’s utility function is 1
Since there are only two states of nature, this implies first order stochastic dominance (FOSD) and the monotone likelihood ratio property: F (xn |l) ≤ F (xn |h), where F (x|a) denotes the cumulative distribution function of x given the agent’s actions a.
6
(5)
U A = βt
"
X
xn ∈X
#
p(xn |a)(xn − wn (xn )) +
X
xn ∈X
p(xn |a)wn (xn ) + b(a, i).
Note that the agent’s utility function depends on the type of the agent. Given an IPO, βe > 0, as the entrepreneur retains part of his shares in the company. The outside manager will not hold shares of the venture, so βm = 0.
3
Solution of the model
Since the action space and the return space are discrete, one can use a modification of Grossman’s and Hart’s [6] approach to derive the optimal contract. Thus the game is solved backwards. I will use the following algorithm: First compute the optimal wage scheme that implements action a given the agent’s type and the level of control he is exposed to. Then determine for each type of the agent the optimal level of control i. Thus the optimal exit channel is found. Since there are only two states of nature, a = l can always be implemented by paying no wages. Thus, the interesting question is: How can a = h be induced? Are the revenues generated by a = h sufficient to cover wages and control costs? In the rest of this section I will first state the principal’s2 program. Then I will solve this program using the above algorithm.
3.1
The principal’s program
The program for the optimal contract is standard. The principal’s expected utility is described by his utility function: 2
Note that the principal is the new owner of the firm or the largest shareholder. Neither the VC nor the entrepreneur can act as principal.
7
(6)
max UP = α
wn ,a,i,t
"
X
xn ∈X
#
p(xn |a)(xn − wn (xn )) − c(i),
and the following constraints: First, the participation constraint of the agent and the principal (the new owner): (PC)
UA ≥ 0,
(LLP)
UP ≥ 0
the incentive compatibility constraint of the agent: UA (a) ≥ UA (a0 ) ∀ a0 ,
(IC)
and the limited liability constraints of the firm and the agent: (LLF)
x n − wn ≥ 0
(LLA)
wn ≥ 0.
Note that the participation and incentive compatibility constraint depend upon the agent’s type. The principal’s limited liability constraint differs from the firm’s as control costs are born by the principal himself, not by the company.
3.2
Solution
In this subsection I solve for the optimal contract. The first result that follows from the setup is: The agent’s participation constraint is slack and can be eliminated from the program. This is obvious since B(a, i) > 0 and wages are nonnegative Before I start with the derivation of the optimal wage scheme, some definitions will be useful: I define a wage scheme in the following way: A wage scheme is a non negative vector w(t, i) = (w1 , w2 ) that specifies a wage for each outcome x1 , x2 , respectively xl and xh , given manager type t and optimal control intensity i. To avoid notational clutter, I will write w instead of w(t, i). 8
Proposition 1 (Wage Scheme) If a solution exists, the trade sale wage scheme is a contract of the following form: (7)
w
sale
(t, i) =
0,
∆b(i) ph
while the IPO wage scheme is a contract of the following form: h xh (8) wIP O (t, i) = 0, max {0, ∆b(i)−βp }. (1−β)ph
Proof 1.1 First, note that in the first step of the Grossman & Hart algorithm, the maximization of the principal’s revenues is equivalent to a minimization of the agent’s wage. Then the principal’s objective function can be rewritten as: X (9) min p n wn wn
xn ∈X
subject to the incentive compatibility and limited liability constraints. Second, the firm’s limited liability constraint forces wl = 0 as xl = 0 and LLP has to hold. The two inequalities hold only if wl = 0. This allows one to concentrate on the wage in the high return state wh . Again, the lower bound for wh is given by the agent’s limited liability constraint. But, in case of a sale, setting wages equal to zero would violate the IC constraint. Therefore IC holds with equality. Since wl = 0, the IC can be rewritten as: (10)
whsale =
∆b(i) ph
In the case of an IPO, the agent’s wage can be determined in the same fashion. The wage is then given by (11)
wh∗ =
∆b(i) − βph xh . (1 − β)ph
However, the limited liability constraint for the agent may bind3 , depending on the size of the expected revenues of the firm, so (12)
whIP O = max{0, wh∗ }.
2 3
This implies that E could sell some of his shares in the firm ex-ante without any effects on firm value. However, this does not imply that there is room for a hold-up by the VC before the exit, as the VC will hold the right to decide about the exit channel only if he would maximize social surplus
9
Limited liability: Note that there may be cases when the optimal w2 violates the firm’s limited liability constraint. In this case there exists only a = l is a s solution with w = 0 I will call a wage scheme that w2 does not violate the firm’s limited liability constraint feasible. If w2 violates the principal’s (not the firm’s) participation constraint for α, then an IPO is not possible, but a sale may still be feasible. If the wage scheme is feasible, the principal’s participation constraint can only be violated if he exerts control, so c(i ∗ ) > 0. If no control is exerted, then the firm’s and the principal’s limited liability constraints are identical Control intensity Going back one step in the game tree, the next issue in the solution is to determine the principal’s optimal control intensity i. This intensity depends upon the type of agent: Since the entrepreneur is not affected by control, exerting control when the agent is E is a waste of money. Since the manager is affected by control, control may constitute an efficient decision. Then, for each feasible wage scheme w, either i = 0 and/or i = i∗ implements action a = ah . Both alternatives have to be checked. If i = i∗ only implements a = ah , then there is no choice left. Optimal agent type and management replacement: Once the optimal w, i combination is found the principal can determine the agent’s optimal type t∗ . Formally: (13)
UP (t∗ ) ≥ UP (t)
To avoid confusion, note that initially the entrepreneur is the agent. If the principal decides to replace the entrepreneur, he will hire an external manager who will then act as the agent. If the principal decides against replacement, the entrepreneur stays on as the agent. Lemma 2 If no control is exerted, (i = 0), it is never optimal to replace the entrepreneur with a manager. Proof 2.1 By assumption, if i = 0, both types of agent have the same private benefits. In a sale, the principal is indifferent between replacing or not, as wages do not differ. In case of an IPO, the entrepreneur’s wage is given by whIP O (e, 0) = max{0, wh∗ } where wh∗ 10
is: (14)
wh =
β ∆b(0) − xh (1 − β)ph (1 − β)
while the manager’s is simply wh (m, 0) =
∆b(0) . ph
Since the entrepreneur’s wage falls in xh ,
and in an IPO β > 0, the entrepreneur’s wage will always be lower than the manager’s. 2 What is the optimal agent type for i = i∗ ? By assumption, controlling E is too expensive. As control is costly, it is exerted only when it is worth to do so. Therefore, control will only be exerted when the agent is a manager. Lemma 3 If control is exerted, (i = i∗ ), the entrepreneur will be replaced by an outside manager. The proof follows directly from the above discussion. Exit channel and firm value: Once the optimal type of agent t for each possible path is found, the principal can choose the optimal exit channel. A comparison of total firm value yields the most valuable one. The firm value is the utility the principal can derive from the venture. For an IPO, total firm value will be (15)
UIP O =
X
xn ∈X
pi (xi |a)(xi − wi ) = ph (xh − wh (0, e)) − f.
Thus fraction α costs αU (IP O). Prices for all investors are identical. In a sale total firm value will be: (16)
Usale =
X
xn ∈X
pi (xi |a)(xi − wi ) − c(·) = ph (xh − wh (i, t)) − c(i)
First, what are valid parameter constellations for each exit channel? Corollary 4 If no control is exerted, (i = 0), an IPO is optimal iff ∆w(0) ≥ f , where ∆w = whsale (e, 0) − whIP O (e, 0) 11
Proof 4.1 If i = 0 is feasible, both a sale and an IPO are possible. Proposition 2 states that wages are given by w = whsale (e, 0) in a sale and by whIP O (e, 0) in an IPO. Firm value is then given by: UIP O (0) = ph (xh − whIP O (e, 0)) − f
(17) while (18)
Usale (0) = ph (xh − wh (e, 0)) − c(0)
Simplification yields than an IPO is better if and only if ph ∆w(0) ≥ f holds. Otherwise a sale would be better. 2 In the case of control the following holds: Corollary 5 For i = i∗ , a sale will always be better than an IPO. Proof 5.1 For i = i∗ the optimal agent type is a manager. This implies that wages in an IPO and a sale are identical. However, in an IPO additional fees f occur that do not occur in a sale. This makes a sale always more attractive than an IPO4 . 2 What is left to do now is to test for the following: Which strategy yields higher returns? Either, if for i = 0 a sale is better, no control is optimal if: Usale (0) ≥ Usale (i∗ ) ph (xh − whsale (e, 0)) ≥ ph (xh − whsale (m, i∗ )) − c(i∗ ) (19)
c(i∗ ) ≥ ph (whsale (e, 0) − whsale (m, i∗ ))
Or, if for i = 0 an IPO is better, then UIP O (0) ≥ Usale (i∗ ) ph (xh − whIP O (e, 0)) − f ≥ ph (xh − whsale (m, i∗ )) − c(i∗ ) (20)
ph (whsale (m, i∗ ) − whIP O (e, 0)) ≥ f − c(i∗ )
Such, the optimal exit channel can be found. 4
Additionally, in an IPO the prinicpal’s fraction α might be too low to induce the principal to exert control.
12
4
A pecking order theory
In the last section I showed how the optimal exit channel is derived for given parameters. This channel constitutes the equilibrium exit choice and is described by a wage contract w and decisions i, t. It differs with firm characteristics. It is interesting to derive the best exit channel for every possible firm configuration. Therefore, I characterize optimal contracts in their relation to firm revenues and costs. Firms can differ in their profitability in various ways. Either a company has higher revenues than its competitors while costs are equal, or costs differ but revenues are equal or combinations of both are possible. However, for comparative statics, I fix costs and vary revenues. The following proposition results then: Proposition 6 (Pecking order) For given parameters α, β, ph , b(a, i) there exists a critical x∗h , such that companies whose returns exceeds x∗h go public via an IPO, while those whose return does not exceed this critical value are exited via a trade sale. To simplify the proof, I will first derive the next proposition: Proposition 7 (IPO optimality) For given parameters α, β, ph , b(a, i) there exists a critical x∗h , such that, if xh ≥ x∗h , an IPO with no control where the entrepreneur acts as an agent is the most efficient choice. Proof 7.1 First, let’s differentiate equation 11 with respect to xh : Either (21)
∂whIP O β =− ∂xh 1−β
or the slope is zero, as the wage contract is differentiable at each point except at the x w h where E’s wage is zero for the first time. If one differentiates any other wage scheme with respect to wh , the partial derivative will zero, as all other wage schemes are independent of the profit of the company. The slope of the profit functions for the IPO wage scheme is thus one for all wage schemes except the IPO wage. The slope of the profit function for 13
the IPO is greater one up to xw h and falls then to one. The parameter restriction upon f ensures that, for all xsale and xhIP O such that USale = 0 and UIP O = 0, the IPO return is h larger than the sale return, starting at some xh ≥ xhIP O 5 . Figure 2: Exit decision: comparative statics
Figure 2 explains this: It draws revenues, expenditures and the different wage schemes in a two-dimensional xh , wh space, where xh is on the horizontal axis while wh is on the vertical axis. The profit from each exit channel is given by the difference between revenues and total costs for each channel. The 45 degree line draws the possible revenues from operations. From these revenues, control costs, going public fees and wages need to be subtracted. The costs are independent from firm value and are represented by parallel lines to the horizontal axis. Wages either fall in xh in the case of an IPO, or are, again, 5
If that were not the case, only sales were feasible.
14
parallel to the horizontal line. The difference between revenues, costs and wages represents profits and therefore firm value. The larger the difference, the higher the profit. As one can clearly see, for high values of xh , the IPO’s profit, which is represented by the kinked line, cannot be exceed by a sale, represented by the straight line. Since there is only one possible configuration for IPOs, this results in the postulated pecking order. From this follows that for low values of xh , sales have higher profits. Note the kink in the IPO profits. This kink marks the point where the entrepreneur’s wage falls to zero in the IPO case. Lemma 8 (Monotonicity) Once IPOs become suboptimal, they stay suboptimal. That is, for all xh < x˜h , a sale will be optimal. The proof follows directly from the graph. Since the profits of a sale grow slower than of an IPO, they also fall slower than IPO profits. Then, if for xh < x˜h sales are more efficient, they will stay more efficient. Therefore, there exists a monotonic relationship. This constitutes the above postulated pecking order.
5
Optimal Exit Choice
After the insight that the exit decision and therefore the venture’s governance structure affects firm value, the question at hand is whether the right to decide about the exit channel is valuable to the players. Is this a relevant question? Yes, as firm value alone is not the appropriate criterion to decide about efficient exit choice, and inefficiencies can not be avoided. To see this assume that (22)
UIP O < Usale .
Then the VC would prefer the sale, while the entrepreneur would also take into account his private benefits and might prefer an IPO. The private benefits may compensate the entrepreneur for the relative loss incurred by the IPO. For the VC this deal destroys value as his return from the company will be lower than it could be. This shows that the right to decide about the exit channel is valuable. 15
6
Testable Hypothesis & Empirical evidence
In this section I present additional empirical evidence for the fact that IPO’s internal rates of return (IRR)6 are higher than those of trade sales’. The data is from the CEPRES database. It is presented in table 6. In this table, the median, mean and the standard deviation of the IRRs are given for both sales and IPOs. The data presented is for all investments where data where available, in total 531 observations, where 423 observations belong to sales while 108 observations belong to IPOs. The most important observation is that both the median and mean IRR are considerably higher for IPOs than for sale. However, the standard deviation is quite high, although it is higher for sales than for IPOs. Table 1: Internal rates of return for IPOs and trade sales. Exit Choice
No. of Obs. Median IRR
Mean IRR
Standard Deviation
IPO
108
58,39
123,42
207,97
TS
423
18,32
75,32
408,27
Notes: Data is from CEPRES. Sample period 1971 to 2003. IRR is defined as the average annual return over the investment period. Mean and standard deviation are the unweighed sample moments. The t-test statistic for a difference in the means is t = 1, 706, where µIP O −µsale t= √ . The result is inside the 90 % confidence interval. V arIP O /nIP O +V arsale /nsale
A simple test for differences shows that the two means are different on the 90 % interval. This test underestimates the true differences as the means are driven by outliers in the data. While these results are in line with the results in Bygrave and Timmons [3], they are in contrast to results presented by Cumming [5]. Cumming reports that IPOs yield lower IRRs than sales. However, the number of observations is quite low, around 15 for each exit channel. Given the longer observation period and higher number of observations in the CEPRES data, this suggests that there might be a bias in the data presented by Cumming. However, while this result supports the model, more evidence in favor of the 6
The IRR is defined as the average yearly rate of return on invested capital during the investment period for each company.
16
pecking order is clearly necessary.
7
Conclusion
This paper shows that the exit choice of a venture depends upon the expected profitability of the venture. The value of the venture depends upon the ex-ante determined ownership structure of the venture and therefore the design of the ownership structure at the point where the VC leaves the venture influences firm value. The findings are in line with other papers such as Basha and Walz [2]. However, Bascha and Walz did not consider the influence of post-exit corporate governance on firm value. This paper’s contribution is to analyze this feature of exit decisions and to provide an explanation why observed returns for VC investments show higher returns for IPOs than for sales. The paper does however not consider some aspects of the going public process: On the IPO side it neglects the usual payoff uncertainty. On the sale side it assumes that the market for sales is competitive and that there are no strategic synergies associated with sales. This aspect of the going public process is analyzed by Schwienbacher [10] in detail. The exit decision is also a decision about the extent of the realignment of ownership and control. The better the company, the less control is needed and the less realignment needed. This may allow for an IPO with lots of passive shareholders and control shareholders that hold only relatively small stakes. Contrary, if the company needs lots of control this is not possible and a complete realignment of control is necessary. This requires the founder to give up his stake and to relinquish his control benefits. This implies that the right to decide which exit channel should be pursued is valuable and therefore VC contracts will include provisions that will assign this right to one of the two involved parties.
17
References [1] Philippe Aghion, Patrick Bolton, and Jean Tirole. Exit options in corporate finance: Liquidity versus incentives. Unpublished Working Paper, 2000. [2] Andreas Bascha and Uwe Walz. Convertible securities and optimal exit decisions in venture captital finance. Journal of Corporate Finance, 7:285–306, 2001. [3] William D. Bygrave and Jeffry A. Timmons. Venture Capital at the crossroads. Harvard Business School Press, Boston, Massachusetts, USA, 1992. [4] Thomas J. Chemmanur and Paolo Fulghieri. A theory of the going public decision. Review of Financial Studies, 12(2):249–279, 1999. [5] Douglas J. Cumming. Contracts and exits in venture capital finance. Unpublished Working Paper, 2002. [6] Sandy Grossman and Oliver Hart. An analysis of the principal-agent problem. Econometrica, 59:7–45, 1983. [7] Thomas Hellmann. Ipos, aquisitions and the use of convertible securities in venture capital. Unpublished Working Paper, 2002. [8] Robert D. Innes. Limited liability and incentive contracting with ex-ante action choices. 52:45–67, 1989. [9] Ernst Maug. Ownership structure and the life-cycle of the firm: A theory of the decision to go public. European Finance Review, 5:167–200, 2001. [10] Armin Schwienbacher. Innovation and venture capital exits. Unpublished Working Paper, 2002. [11] Tereza Tykvov´a. The decision of venture capital exits. Unpublished Working Paper, 2003.
18
[12] John Wall and Julian Smith. Better exits. Technical report, European Venture Capital Association, 1998. [13] Luigi Zingales. Insider ownership and the decision to go public. The Review of Economic Studies, 62:425–448, 1995.
19
