Contracts and Returns in Private Equity Investments Stefano Caselli Bocconi University [email protected]

Emilia Garcia-Appendini Bocconi University [email protected]

Filippo Ippolito Bocconi University …[email protected]

June 14, 2011 Abstract We consider two contracting dimensions in a private equity (PE) deal: covenants and the appointment of directors in the board of target …rms. We examine how returns vary along these two dimensions, by employing several measures of returns such as the fund’s IRR and the growth in ROE, ROA, and sales of the target …rm over the investment period. We …nd that covenant-heavy contracts and the appointment of outside directors are associated with higher returns, independently of the measure of returns adopted. We examine the possible self-selection e¤ects in the use of covenants and show that …rms with higher pro…tability choose more covenants. Finally, we show that covenants positively a¤ect returns, and that their e¤ect is stronger on sales, ROA and ROE, than on IRR.

JEL Classi…cations: G11, G23, G24 Keywords: private equity funds, venture capital, IRR, covenants, board directors

We would like to thank MPS Venture SGR for kindly providing the data; Francesco Corielli, Douglas Cumming, John Doukas, Giuliano Iannotta, Josh Lerner, Marco Da Rin, Antoinette Schoar, Albrecht Glitz, two anonymous referees, and the editor Manju Puri for helpful comments. We also thank seminar participants at the EFA in Frankfurt, EFMA in Milan, EFMA Symposium on Private Equity in Montreal, and FMA in Reno. We acknowledge …nancial support from CAREFIN Centre for Applied Research in Finance. All remaining errors are our own. An earlier version of this paper was circulated under the title "Explaining Returns in Private Equity Investments". All authors are a¢ liated with the Department of Finance at Bocconi University, Via Roentgen 1, 20136 Milan, Italy. Corresponding author: Filippo Ippolito, Department of Finance, Università Commerciale Luigi Bocconi, Grafton Building, Second Floor - Room 2-D2-02, Via G. Röntgen n. 1 - 20136 - Milano, Tel. +39 02 58365918, Fax. +39 02 58365920.

Contracts and Returns in Private Equity Investments Abstract We consider two contracting dimensions in a private equity (PE) deal: covenants and the appointment of directors in the board of target …rms. We examine how returns vary along these two dimensions, by employing several measures of returns such as the fund’s IRR and the growth in ROE, ROA, and sales of the target …rm over the investment period. We …nd that covenant-heavy contracts and the appointment of outside directors are associated with higher returns, independently of the measure of returns adopted. We examine the possible self-selection e¤ects in the use of covenants and show that …rms with higher pro…tability choose more covenants. Finally, we show that covenants positively a¤ect returns, and that their e¤ect is stronger on sales, ROA and ROE, than on IRR.

JEL Classi…cations: G11, G23, G24 Keywords: private equity funds, venture capital, IRR, covenants, board directors

1

Introduction

In recent years, the uniqueness and complexity of private equity (PE) contracts has attracted much interest among academics and the wider public. On the one hand, this has been the result of the increasing importance of PE in the economy. On the other, PE contracts have been regarded by academics as a primary exploratory …eld for testing theories of optimal contracting. In this paper, we take up on this line of research and examine how contracts and returns are related in PE deals. Following Cumming (2008), Kaplan and Strömberg (2002) and Lerner (1995), we consider two contracting dimensions in a deal: covenants and the appointment of directors in the board of target …rms. We examine how returns vary along these two dimensions. We begin our analysis by observing that covenant-heavy contracts are generally associated with higher returns, regardless of the whether the measure of performance employed is the IRR of the fund, the change in ROE and ROA, or the increase in sales over the investment period. Starting from multiple correspondence analysis, we identify three main “contracting styles”. The …rst style includes contracts that carry a lockup clause, a right of …rst refusal or an exit ratchet. This style is observed in only 18.19% of cases and is associated with a yearly IRR of 20.2%, which is 13.2% higher than the yearly IRR of deals without these clauses. The second contracting style includes tag- and drag-along rights, it is observed in 90.48% of cases and is associated with higher returns (9.7% vs. 7.0%). The third style includes redemption rights and permitted-transfer rights, it occurs in 31.56% of cases: the yearly IRR of deals with these covenants is 10.7%, compared with a yearly IRR of 8.8% for deals without them. We label the three styles performance, default and protection. We then explore what may cause these …ndings. We envisage two possible explanations: unobserved heterogeneity (self-selection) in …rm pro…tability and causal e¤ects of covenants on returns (treatment). We examine …rst the issue of self-selection: …rms with speci…c (possibly unobserved) characteristics choose some covenants rather than others. More precisely, observed covenant choices are the results of public and private information held by the contracting parties at the time of investing. The unexplained part in a regression of covenants on public information re‡ects private information, i.e. the unobserved characteristics that drive the selection. As suggested by Li and

1

Prabhala (2007), a correct procedure to test for private information is a two-stage self-selection model. Speci…cally, we borrow the methodology of Puri (1996), itself based on the selection adjustment method of Heckman (1979).1 In the …rst stage we regress the inclusion of covenants on …rm, fund and market variables and compute the inverse Mills ratios. In the second stage we regress returns on these Mills ratios and estimate the information value of covenants. We …nd that Mills ratios explain a signi…cant part of return variability. More precisely, we …nd that the inverse Mills ratio is strongly and positively signi…cant for …rms that employ a number of covenants that is larger than average, and for …rms that employ performance covenants. We do not observe a selection e¤ect for default and protection covenants. The positive sign of the Mills ratio means that the covariance between the unexplained factors that a¤ect returns and those that a¤ect the choice of covenants is positive. To interpret these results we rely on existing theories of optimal contracts. The three components of performance covenants, exit ratchets, lockups and rights of …rst refusal, help reduce moral hazard of managers, as they tie performance to compensation. Lockups and rights of …rst refusal also signal high …rm pro…tability and a likely pro…table exit via a trade sale or an IPO. We then compare the signi…cance of the inverse Mills ratio across di¤erent return measures and …nd that performance and protection covenants are more strongly associated with IRR than with changes in sales and ROA. These …ndings suggest that performance and protection covenants increase the returns to VC investors more than they increase sales and ROA. The asymmetry may be due to performance covenants in the upside, and to protection covenants in the downside. There may also be a redistribution e¤ect from debtholders and other …rm stakeholders towards VCs, as well as the e¤ect of strong bargaining power of VCs versus other stakeholders. Having controlled for the selection issues associated with the choice of covenants, we can examine the relationship between the appointment of outside versus inside board directors. We refer to outsiders directors as having weak ties with the VC. We develop various measures for outsiders, based on the cumulative number of appointments held by any given director on behalf of the VC, on whether the appointee is a dependent of the VC, and on whether there is synchronism of 1 Dunbar (1995), Song (2004), Fang (2005) and Goyal (2005) all provide examples of how self-selection can be employed to study the role of private information in relation to the choices of economic agents.

2

mandates between the board of the target …rm and the appointment of the director. Independently of the measure employed we …nd that outside directors are associated with better performing …rms. Importantly, the relationships between covenants, directors and returns are strongly signi…cant after controlling for a number of …rm, fund and market-wide characteristics. These results are consistent with the hypothesis that VCs employ inside directors only in …rms that require intensive monitoring (Lerner (1995)). They are also consistent with the hypothesis that “busier” directors exert less monitoring (Fich and Shivdasani (2006)). Finally, we turn our attention to the causal e¤ects that covenants may have on returns. We adapt the model of Goyal (2005) because it allows for the separation between selection and treatment e¤ect, which is not possible in a straight self-selection model. The trade-o¤ of Goyal’s model is that the sign of the selection coe¢ cient has no clear economic interpretation. We …nd that employing more covenants than average leads to higher changes in sales, ROA and ROE, but not to higher IRR. A similar result holds for the use of performance covenants, while default and protection covenants do not seem to cause higher returns. These …ndings suggest that covenants change the behavior of managers and promote the e¢ ciency of the …rm. The e¤ect of performance covenants, such as exit ratchets, clearly emerges here: as managers are rewarded with shares of the …rms, they have strong incentives to increase the value of the …rm in general and of equity in particular. To our knowledge this is the …rst paper that addresses empirically the relationship between contracts and returns in PE investments. It establishes a link between the strand of literature on the returns of PE investments (Bygrave and Timmons (1992), Gompers and Lerner (1997), Groh and Gottschalg (2006), Kaplan and Schoar (2005), Lerner et al. (2007), Ljungqvist and Richardson (2003)) and that on venture capital contracts (Casamatta (2003) and Cornelli and Yosha (2003), Hellmann(1998, 2006), Gompers (1999), Kaplan and Strömberg (2002, 2004), Lerner and Merges (1998), Metrick and Yasuda (2010), Sahlman (1990), Schmidt (2003)). Our results complement the …ndings of Cumming (2008), according to which stronger venture capital control rights increase the likelihood that an entrepreneurial …rm will exit by an acquisition, rather than through a write-o¤ or an IPO. We regard our …ndings important for academics because they shed light on optimal

3

contracting in the presence of agency costs and asymmetric information. To practitioners our …ndings provide a guideline for the signalling implications of speci…c contracting strategies. The rest of the paper is organized as follows. Section 2 describes the data collection process and provides some descriptive statistics. Section 3 discusses covenants. Section 4 and 5 examine the relationship between covenants and returns, accounting for unobserved heterogeneity and selfselection. Section 6 examines the relationship between the choice of board directors and returns. Section 7 explores the e¤ect of covenants on returns. Finally, Section 8 concludes.

2

Data Collection and Sample Description

We use a proprietary database provided by MPS Venture SGR, a currently active Italian management company. The database covers all deals that satisfy the following requirements: 1) investment occurred between January, 1st 1999 and December, 31st 2005; 2) target company is incorporated in Italy; 3) investment company is registered as an Italian PE management company. Data were collected by MPS Venture SGR from several sources: 1) the Bank of Italy provided the list of Italian PE management companies and the aggregate number of deals that these companies have made;2 2) the Italian Private Equity and Venture Capital Association (AIFI) provided information for single deals, including: target …rm name, type and size of investment, percentage of shares acquired by the fund, entry and exit dates, exit, leverage, IRR, and covenants; 3) private interviews held by MPS Venture SGR with other fund managers completed the information of AIFI; 4) Bureau Van Dijk’s AIDA/Amadeus Database, and Italian Balance Sheet Central Database (Centrale dei Bilanci), provided balance sheet data; 5) the Italian Credit Bureau (Centrale dei Rischi) provided information about credit relationships; and 6) the Trade Ministry (Camera di Commercio) and the Italian Security and Exchange Commission (Consob) provided information on board members. Importantly, due to privacy restrictions, MPS Venture SGR has not disclosed the names 2 According to the European 1998 Financial Services Directive directive, PE management companies are regarded as regulated …nancial companies; as such, they must register with the Central Bank of their country of incorporation, and must disclose the aggregate number of deals made by each fund.

4

of any of the entities involved. Therefore, 1) we are unable to merge our data back into publicly available databases such as Amadeus/AIDA to complement the balance sheet information that may still be missing; 2) we cannot distinguish between …rst, second or higher rounds of …nancing to the same …rm by di¤erent VCs; and 3) we do not know which …rms are public or private at the time of investment. Finally, we complement the database of MPS Venture with aggregate macroeconomic data at the time of investment, obtained from Datastream and AIDA/Amadeus: 1) the returns of the Italian stock index, 2) the ratio of IPOs over newly created …rms, 3) industry leverage, and 4) industry ROE.

2.1

Sample Characteristics

The sample includes 834 investments made by 87 PE funds, which are respectively owned by a total of 58 management companies. A large fraction of the target …rms operates in the consumer goods sector (34%), the general industrial sector (25%), and the services sector (20%). Consistent with the structure of the Italian corporate sector, most target …rms in the sample are majority owned by individuals or families (69.28%). Banks, VCs and managers respectively controlled 1.87%, 17.04% and 11.82% of the target …rms before the investment. All investments were …nanced between 1999 and 2005, with 2000, 2001 and 2004 being peak years. The median investment size is 4.1 million euros. The median annualized IRR ranges from 4:33% to 49:31%, depending on the year of exit, with the greatest returns observed during the technology bubble of 1999-2001, and the lowest observed after the …nancial crisis in 2011. The majority of our deals consist of expansion …nancing (50.7%), followed by buyouts (27.1%), early stage (16.7%) and turnaround (5.39%). Early stage and expansions are much smaller than buyouts and turnarounds. Buyouts include most of the larger deals in our sample, which however look small by international standards, particularly if compared to the large buyouts recently witnessed in the US and the UK (see Axelson et al. (2011) and Colla et al. (2011)). All but 11 deals were exited at the time of data collection (May 2011). Trade sale is the most common form of exit (86.9%). IPOs and write-o¤s are relatively rarer, respectively 5.1% and 6.5%. 5

IPOs are more correlated with buyouts and expansions than with early-stage and turnarounds (untabulated). Table 1 provides descriptive statistics of our returns measures, which include yearly IRR, and the yearly changes in sales, ROA, and ROE over the investment period. IRR is simply computed as the di¤erence in VC’s equity stake at the time of investment and exit. Annualizations are based on yearly compounding.3 These measures respectively proxy for the returns to the VC (IRR), to all stakeholders in the …rm (sales, ROA) and to equityholders (ROE). The averages (medians) of yearly IRR, growth in sales, ROA and ROE are 9.6% (10.7%), 6.7% (3.9%), 6.0% (3.2%), and 17.6% (7.4%). As a term of comparison Kaplan and Schoar (2005) …nd that equal-weighted median and average IRRs reported by Venture Economics over the period 1980-2001 are 12% and 17%, respectively. Cochrane (2005) …nds that venture capital investments generate average log returns of 15% per year. Bygrave and Timmons (1992) …nd an average IRR of 13.5% for the period 1974– 1989. Gompers and Lerner (1997) report an arithmetic average yearly IRR of 30.5% gross of fees over the years 1972–1997. Ljungqvist and Richardson (2003) produce an estimated IRR of 19.8%. Finally, we observe that buyouts and early-stage are respectively the most and the least pro…table type of investments, irrespective of the measure employed. Kaplan and Schoar also …nd that returns to buyout funds are slightly higher than the returns to venture funds. As can be expected, we …nd that IPOs are the most pro…table type of exit.

3

Covenants

In this section, we provide some background on the relationship between contracts and returns. This will help in the interpretation of the empirical results that we present in the next sections. In our sample we observe seven di¤erent covenants: lockups, permitted-transfer rights, redemption rights, tag-along rights, drag-along rights, rights of …rst-refusal, and exit ratchets. A de…nition of these covenants is provided in Panel A of Table 2. We …nd that in our sample there is more variety in the choice of covenants than in previous studies. Cumming (2008) reports four 3 For the 11 un-exited deals in the sample, IRR is estimated on the basis of the change in net asset value during the investment period.

6

covenants: right to replace the CEO, redemption rights, drag-along rights, and antidilution rights. Kaplan and Strömberg (2002) report the use of redemption rights, anti-dilution provisions, and automatic-conversion provisions. In both papers VCs employ a mix of securities which includes various types of preferred stock and convertible debt, while in our sample VCs always hold common equity. Table 3 provides descriptive statistics for our set of covenants. Tag-along rights are the most common type of covenant (86.93% of the deals), followed by drag-along rights, permitted transfer and redemption rights. There is relatively little variation in the use of covenants across di¤erent investment types, but there is large variation across di¤erent types of exit. Lockups, permitted transfers, redemption rights, rights of …rst-refusal and exit ratchets are more likely to be included when exit occurs via an IPO. In untabulated results we …nd that over time there has been a tendency towards “covenant-lite” contracts, with particularly clear reductions in the use of tag-along, drag-along and redemption rights. We also …nd a positive correlation in the use of lockups, permitted transfers, redemption rights and rights of …rst-refusal. This suggests that certain types of covenants might be bundled together to form a “contracting style”. To explore this point in more detail and identify di¤erent contracting styles, we perform a multiple correspondence analysis (MCA). As in Principal Component Analysis, the MCA o¤ers a low-dimensional representation of the data, in a way that best preserves the original variance of the data. The …rst dimension of variation explains 84.2% of covenant variability. This component is strongly related to lockups and rights of …rst-refusal, and to a smaller extent also to permitted transfers, redemption rights, and exit ratchets. The second component explains only 0.03% of the total data variability and is strongly related tag-along rights and to drag-along rights. We observe three relatively well de…ned clusters, plus a singleton. Lockups and rights of …rst-refusal represent the …rst cluster. The second cluster is composed by tag-along and drag-along rights. We label this cluster default because tag- and drag-along rights are present in the majority of deals. Redemption rights and permitted-transfer rights form the third cluster. Since both give VCs an exit option, we refer to this cluster as protection. Exit ratchets represent a singleton which cannot be perfectly mapped into any of the previous clusters. From Table 4 exit ratchet appear

7

strongly related to performance which clearly matches the role of this type of covenants in providing incentives to managers. We therefore group exit ratchets together with lockups and rights of …rst refusal to form the third group of covenants that we de…ne performance.

3.1

The Information Content of Contracts

The existing theories on the role of contracts in the …nancing of an entrepreneurial …rm (Aghion and Bolton (1992), Bascha and Walz (2001), Casamatta (2003), Cornelli and Yosha (2003), Dessi (2005), Hellmann(1998, 2006), Inderst and Mueller (2004), Renucci (2000), Repullo and Suarez (2004), Schmidt (2003) among others) identify adverse-selection and double-sided moral hazard as the two main sources of information problems in venture capital contracts. We examine from a theoretical standpoint the interplay between covenants, information problems and returns. We start by looking at performance covenants. These covenants provide incentives to managers towards the maximization of …rm value, and also appear to be associated with high quality …rms. Aghion et al. (2004), Brav and Gompers (2003), Casares-Field and Hanka (2001), and in a broader setting Holmström (1979), suggest that lockups help align the incentives of managers and VCs with the maximization of equity value. Also, lockups signal …rm quality because they are generally associated with IPOs (see Table 3) which is the most pro…table type of exit (Table 2). Rights of …rst-refusal (a.k.a as preemption rights) and exit ratchets are also meant to preserve the incentives of managers towards value maximization, respectively by preventing a dilution of the managers’interest in the …rm, and by rewarding managers with new shares in case of high returns.4 Overall, we should expect performance covenants to be positively correlated with returns because they signal high …rm quality and strengthen incentives. With respect to protection covenants, permitted-transfer rights give VCs the option to sell their stake in the …rm without requesting the permission of other shareholders (Yates and Hinchli¤e (2010)), while redemption rights allow VCs to sell their shares back to the company (long put), typically in the event an IPO or a public merger becomes unlikely. Both types of covenants signal 4 Two special types of exit ratchets employed in the U.S. are returns and time vesting (Kaplan and Strömberg (2002)).

8

strong bargaining power of VCs, but also indicate uncertainty about the …rm’s future prospects. Finally, the two default covenants, tag- and drag-along rights, maximize the likelihood of a pro…table exit and should then correlate positively with returns. Empirically it is di¢ cult to estimate their relationship with returns, precisely because the vast majority of …rms in the sample includes these covenants, thus reducing the power of our tests.

4

The Correlation between Covenants and Returns

In this section we analyze the correlation between the use of covenants and investment returns. We postpone a discussion of the relationship between choice of directors and performance until Section 6. We start our analysis by observing the unconditional pairwise correlation between our returns measures and covenants (untabulated). We also aggregate the covenants into an index, covenant index, which contains the total number of covenants, and is meant to proxy for restrictiveness of the contract. We …nd that lockups, permitted transfers, …rst refusal, exit ratchet and the covenant index are positively and (almost always) signi…cantly correlated with returns. Redemption rights are positively and signi…cantly correlated with IRR and IPO, while they are not signi…cantly correlated with the other measures. Tag-along and drag-along rights are generally uncorrelated with returns. Next, in Table 4 we compare the univariate di¤erences on average returns measures between deals with and without each type of covenant, using t-tests and non-parametric Wilcoxon tests for di¤erences on means. Due to space restrictions, we choose yearly IRR and yearly change in sales as reference returns measures. Panel A shows that yearly IRR is practically always higher for deals containing any type of covenant (X = 1) than for those without (X = 0). Similar but weaker results hold for yearly change in sales as reported in Panel B. In both panels we also include our three contracting styles: performance, protection and default. Performance is a dummy for the inclusion of either a lockup, right of …rst refusal or exit ratchet. Protection is a dummy for presence of either a permitted transfer or a redemption right. Default is a dummy for presence of either tagor drag-along rights. Performance covenants yield high and signi…cant di¤erence; while neither protection nor default covenants yield a signi…cant di¤erence.

9

4.1

Multivariate Evidence

We next explore the relationship between covenants and returns in a multivariate setting controlling for …rm and deal characteristics, and monitoring. We consider the following …rm characteristics: pro…tability, size, leverage, and industry (nine dummies for the sector in which the target …rm operates). The investment characteristics include: the percentage of shares acquired, investment type (dummies for early, expansion, buyout and turnaround), vintage (dummies for the year of the investment), investment duration (length of the investment expressed in months), bank owned (a dummy that takes value of one if the fund is majority owned by a bank). We include a dummy for un-exited deals. We also control for the cumulative number of board seats of the fund appointee on other boards during the investment. These variables are de…ned and summarized in the appendix in Table A.2. We run a set of 66 regressions, one for each return measure (plus IPO and Write-o¤) on each covenant (plus the covenant index and the three contracting styles), controlling for the above …rm and deal characteristics and for directors’board seats. We use OLS to estimate all equations involving a continuous measure of returns, and probit models to estimate the probability of exit. We cluster the standard errors at the VC level.5 The results of all the regressions performed above are summarized in Table 5. Panel A contains the complete model speci…cations for the regressions of each returns measure on the covenant index. In Panel B we run the same speci…cations as in Panel A with each single covenant separately. We also run three speci…cations, one for each contracting style (performance, protection, default). To be concise, in Panel B we report only the estimated coe¢ cients of the relevant covenant for each regression. The estimated (untabulated) coe¢ cients of the control variables in Panel B are qualitatively very similar to their counterparts reported in Panel A. The main …ndings of Table 5 are: 1) the relationship between the covenant index and returns is strong and positive also in a multivariate setting; 2) this relationship appears to be driven mainly by lockups, permitted transfers, rights of …rst-refusal, and exit ratchets, and to a lesser extent by 5 In untabulated results we carry out a robustness check by inlcuding VC …xed e¤ects (a dummy for each management company) in the regressions. Results on covenants are una¤ected.

10

redemption rights; 3) tag- and drag-along appear generally uncorrelated to returns, thus con…rming their status of default covenants.

4.2

Selection and Treatment E¤ects

The empirical …ndings of the previous section raise non-trivial theoretical and empirical issues. From a theoretical standpoint, in a model of optimum contracting in which covenants are chosen to maximize returns, a reduced form equation of returns should depend solely on the parameters of the model. The same should hold for covenants.6 If we take this line of reasoning literally, then empirically we should not expect to observe a relationship between covenants and returns. However, in practice it is di¢ cult to correctly identify and estimate empirically a fully reduced equation of investment returns. Insofar as in the estimated equation of returns there are omitted variables for which covenants are proxying, a correlation between covenants and returns may arise empirically. This may explain the evidence provided in Table 5. Furthermore, as suggested by Li and Prabhala (2007), we may regard this as a problem of self-selection in which covenants capture some of the information privately held by the contracting parties. The selection of covenants gives us information about some unobserved heterogeneity in …rm quality. Self-selection, however, may not be the only reason why covenants are interesting from an empirical point of view. Covenants are also likely to change the behavior of managers by improving their incentives to maximize …rm value. Stringent covenants may induce managers to run the …rm more e¢ ciently and increase returns for investors. This would imply a positive relationship between covenants and returns, and provide an alternative explanation to the empirical facts observed in Table 5. We then have two competing explanations, one based on selection and the other on treatment e¤ ects. Ideally, we would like to separate between the two e¤ects and distinguish the information 6 From a theoretical point of view returns and covenants should not be related at the optimum. To illustrate why, consider a simple theoretical model in which y(C; X) is a measure of …rm returns, which depends on the choice of covenants to be included in the contract, C, and a set of exogenous parameters, X. If we assume that contracts are chosen optimally to maximize returns and that y is concave, then the optimal contract must satisfy the …rst order condition @y=@C = 0. This implies that the optimal choice function C (X) depends only on X. As a consequence, the optimal value of y is y (C (X); X) = y (X). In other words, at the optimum returns should not depend on C. X determines both the choice of covenants C and returns y . We thank an anonymous referee for suggesting this argument.

11

content of covenants from their e¤ects on managerial behavior. Empirically this is not an easy task. In the next section, we start by examining selection e¤ects using a model originally developed by Puri (1996). In Section 7, we employ a di¤erent model, developed by Lee (1978) and proposed by Goyal (2005) in a …nance context, to estimate the treatment e¤ects of covenants on returns.

5

Unobserved Heterogeneity and Self-Selection

We start by examining the choice of covenants in the context of self-selection. Empirically, we deal with self-selection by setting up a two-step procedure à la Heckman (1979) in its speci…c implementation provided by Puri (1996). Puri’s model was originally designed to examine the role of bank underwriting. Markets’ expectations depend on whether a bank chooses to underwrite a security or not, and underwriting reveals information about issuer quality. In our context, by observing a speci…c choice of covenants we can obtain information about unobserved …rm characteristics, such as …rm pro…tability. More formally, returns yi depend on a vector Xi of deal and …rm characteristics: yi = Xi0 + "i :

(1)

The self-selection mechanism can be formalized as in the following system:

wi Ci

= Zi0 + i 8 > < 1; if w 0 i = ; > : 0; if wi < 0

(2)

where Zi is a vector of observable …rm and deal characteristics that a¤ect the choice of introducing a covenant or not, wi is a scalar representing the net bene…t of introducing the covenant, and Ci is the observable counterpart of wi , i.e., a dummy variable equal to one if the covenant is included, and zero otherwise.

i

is the part of wi not explained by the publicly observable variables Zi , and

hence it represents private information about …rm quality. We can incorporate this information

12

into the returns equation by augmenting Equation 1 with a term expectation of private information

i,

0 C (Zi

), which is the conditional

given the choice of including covenant Ci . Using the standard

assumption that the error terms of Equation 2, "i ; and Equation 1,

i;

are bivariate normal,

)

is the inverse Mills ratio. This ratio equals the covenant is not included, where ( ) and

( ) are the standard normal density and the standard

when the covenant is included and

() 1 ()

C(

() ()

when

normal cumulative distribution functions, respectively. To estimate the parameters

and , we follow a two-step procedure. In the …rst step, we

estimate Equation 2 and obtain estimates of , b. We use these estimates to construct the inverse

Mills ratio

0 C (Zi b),

and augment Equation 1 with this variable. In the second step, we estimate

the augmented equation using OLS. The selection-adjustment estimation procedure sketched here is useful for the following reasons: 1) by adjusting for self-selection we obtain consistent estimates for ; 2) the statistical signi…cance of the coe¢ cient for the inverse Mills ratio captures the degree of self-selection in the inclusion of covenants; 3) the …rst-stage regression (Equation 2) helps us predict the use of covenants. Vector Xi of Equation 1 are the same as in the previous sections (pro…tability, size, leverage, industry, shares acquired, investment type, vintage, investment duration, bank ownership, board seats during the investment). In addition to the variables that are in Xi ; vector Zi also contains the following three instruments: stock market returns over the previous six months, de…ned as the returns on Italian equity market (S&P Mib) over the six months preceding the investment start date; the average industry ROE, based on a two-digit SIC classi…cation of Italian …rms at the investment starting date; the average ratio of IPOs to newly created …rms in Italy over the six months preceding the investment start date. These variables may represent useful instruments because they are exogenously given at the time of the investment and capture market and investment conditions at the time of contracting. However, we acknowledge that it is di¢ cult to identify instruments that correlate with covenants but not with performance, and thus do not violate the exclusion condition. In particular, it is likely that pre-deal market conditions may directly a¤ect the IRR of an investment. As shown by Gompers and Lerner (1999, 2000) entry prices are in‡uenced by market conditions, with prices

13

being higher in heated markets than in low markets. Therefore, we can think of IRR as a function of the share price at the time of contracting and at the time of exit. Changes in sales, ROA and ROE appear to be less subject to this problem.

5.1

Selectivity-bias Adjusted Regressions for the Use of Covenants

Table 6 contains the …rst stage of the Heckman procedure. The table provides a set of speci…cations of Eq. 2 in which each contracting style is regressed on …rm and VC characteristics, directors, and market conditions. In column 1 we also include a dummy for the covenant index being greater than its median. The table shows that observable …rm characteristics are poor predictors for the inclusion of covenants. The only exceptions are the relationship of size with protection covenants, and the negative relationship of shares acquired with default covenants. Bank ownership of the fund and pro…tability also appear to play a small role. The second result of the table is the signi…cance of market conditions. All sets of covenants are positively and signi…cantly associated with either prevailing market or industry conditions at the time of contracting. Using the results of Table 6, we estimate the inverse Mills ratios associated with the choice of covenants. We use these ratios to augment the regression speci…cations provided in Table 5. More precisely, the inverse Mills ratios estimated in the …rst stage are used as an extra variable in the estimation of Equation 1. The results of this estimation are presented in Table 7 which contains the second stage of the Heckman procedure and provides our main results for the role of self-selection in the use of covenants. The table reports the coe¢ cients on the inverse Mills ratios estimated in the …rst stage for di¤erent regressions using all our measures of returns yi as dependent variables. The …rst stage regressions (Eq. 2) vary across columns, with the covenant index a di¤erent set of covenants for each column. We are interested in the sign and signi…cance of the coe¢ cient of the inverse Mills ratio

C(

)

because it reveals how important private information is in our model: “The omitted variable used to correct for self-selection,

C(

), is an estimate of the private information underlying a …rm’s choice

and testing its signi…cance is a test of whether private information possessed by a …rm explains ex-post outcomes” (Li and Prabhala (2007)). 14

We …nd that the inverse Mills ratio is always positive and signi…cant at the one percent level for the covenant index and for performance covenants, while it is not signi…cant for the other covenants. The sign of the signi…cant coe¢ cients is positive thus indicating that the inclusion of these covenants represents a signal of high returns. In particular, the sign of the inverse Mills ratio tells us that the covariance between the errors in the selection and outcome equations is positive. This means that there is a positive covariance between the unexplained factors that a¤ect returns and those that a¤ect the choice of covenants. A possible interpretation of this result is that covenants are introduced when their e¤ect on returns is more pronounced. This would lead to a treatment e¤ect of covenants on returns, in line with our …ndings of Section 7 below. We also …nd the dummy of bank ownership (bank owned) and the number of board appointments (“board seats”) to be negative and statistically signi…cant. In untabulated results, we replace bank owned with previous bank relationship, a dummy equal to one when the fund is majority owned by a bank and the target …rm has had a previous credit relationship with the bank. The coe¢ cient for this variable is still negative and signi…cative. This result is in line with Hellmann et al. (2008) who suggest that banks as limited partners might diverge from maximizing returns on investments in order to maximize future banking income from the portfolio of …rms in which they invested. We postpone the discussion of the signi…cance of board seats to Section 6. 5.1.1

Comparison of IRR Versus Other Measures of Returns

Next we compare the inverse Mills ratios of IRR with those of the other measures of returns. To this purpose, in the last three rows of Table 7 we compute the inverse Mills ratios for the di¤erence between IRR and sales growth, ROA growth, and ROE growth, respectively. Interestingly, we observe a positive and signi…cant coe¢ cient for the di¤erence of IRR versus sales and ROA both for the covenant index and for performance covenants. To a smaller extent we observe this patter also for protection covenants. These …ndings suggest that performance and protection covenants increase the returns to VC investors more than proportionally than the increase in …rms sales and ROA. This mechanism is likely to operate in the upside for performance covenants and in the downside for protection covenants. It suggests the existence of a redistribution e¤ ect from debtholders and

15

other …rm stakeholders towards VCs. This …nding may also re‡ect strong bargaining power of VCs versus other stakeholders. At the same, however, we …nd a negative coe¢ cient for the inverse Mills ratio of (IRR ROE) for performance covenants, which indicates that the e¤ect of these covenants on ROE is stronger than on IRR. A possible interpretation of this di¤erence is that performance covenants have a mechanical redistribution e¤ect in favour of managers via the dilution of the VC’s stake. This clearly is the case for exit ratchets. In sum, the main …ndings of Table 7 are, …rst, that the inclusion of a number of covenants higher than average and the inclusion of performance covenants provides information about (unobserved) expected returns. Second, the information revealed by the inclusion of covenants varies signi…cantly across di¤erent measures of returns, in a way that re‡ects (at least partially) the redistribution e¤ects of cash ‡ows associated with the various classes of covenants, and the relative bargaining power of the contracting parties.

6

The Choice of Board Directors

As shown by Barry et al. (1990), Lerner (1995) and Sahlman (1990), it is common practice in the U.S. for VCs to elect one or more directors in the board of the …rms in their portfolio. The role of these directors is to represent the interests of the VC by exerting direct oversight on the managers of the …rm. In our sample deals we also …nd that the appointment of directors is a common phenomenon. Almost invariably VCs appoint only one director, regardless of the percentage of shares acquired. We classify appointed directors as “outsiders” and “insiders”, depending on whether the appointee is or has been an employee of the VC, or has strong ties with the VC. Our de…nition of “outsider”is di¤erent from that employed by Lerner (1995), which classi…es directors into outsiders and insiders with respect to their relationship with the …rm, rather than with respect to the VC. More precisely, we provide four di¤erent measures of outside directors. Our …rst measure is based on the cumulative number of appointments held by any given director on behalf of the VC during the time of investment. A higher number means stronger ties with the VC. Our second

16

measure is the cumulative number of appointments held by a director before the time of investment. This measure has the same meaning of the …rst one, but is backward looking. Third, we look at whether the director is or has been an employee of the VC. Fourth, we look at whether there is synchronism of maturity between the appointment of the director and the board of the target …rm. Synchronism is typically observed when a director is an outsider. These measures are summarized in Panel B of Table 1. Table 8 provides descriptive statistics for our four measures of outside directors. The …rst two measures are highly correlated between each other and both have a mean of approximately 6.9. We observe little variation of these measures across di¤erent investment types. However, both are lower for IPOs. In 46.3% of cases the appointed director is or has been an employee of the fund. This percentage is 65.1% in case of IPOs. The synchronism of mandate is observed in 38% of cases, with a low mean of 28.9% in case of turnarounds, and a mean of 46.5% in case of IPOs.

6.1

The Relationship Between Board Directors and Returns

We now examine the relationship between the choice of directors and returns. We employ our four measures of outside directors illustrated in Panel B of Table 1: contemporaneous board seats, previous board seats, internally appointed director (employee), and no synchronism. The …rst two are similar to each other except that for the latter the number of appointments is computed over the period that precedes the investment, rather than during the investment. Internally appointed director is a dummy that takes value of one if the VC’s appointee is either an employee of the VC or has entertained a long-term relationship with the VC. No synchronism is a dummy that takes value of one if the mandate of the VC’s appointee does not coincide with that of the target board. The interpretation of these variables is as follows: a higher number of board seats, contemporaneous or previous, proxies for a stronger relationship between VC and appointed director. Inside appointment and no synchronism proxy for a stronger relationship between VC and appointed director. As suggested by Lerner (1995), “if venture capitalists are intensive monitors of managers, their involvement as directors should be more intense when the need for oversight is greater.” Therefore, we expect outsiders to be appointed in …rms where less monitoring is required. These 17

…rms are likely to be the less risky and more pro…table ones, which implies that we should observe a positive correlation between the appointment of outsiders and returns. We …rst explore the univariate correlation between the choice of director and returns. We …nd a positive relationship between the choice of an outside director and sales, ROA, and ROE, and a mild correlation between the choice of an external director and an IPO. There is not a signi…cant unconditional correlation between IRR and the choice of an external director. We also …nd that contemporaneous board seats are conditionally correlated to all performance measures except yearly IRR after controlling for other …rm, fund, and market characteristics and for the endogeneity of covenant choices. We then analyze whether the signi…cance observed between returns and the presence of an externally appointed director is robust to all other measures of outside directors. In Table 9 we report the coe¢ cients for all four measures of outside directors, as computed in 16 separate second-stage regressions that are similar to those reported in Tables 13. For conciseness, we report only the coe¢ cients on the director seats rather than the whole regression. In the …rst stage we group covenants into the four groups obtained from previous analyses (covenant index, performance, default, protection). For each of these four groups, we run four …rst-stage (as in Eq. 2) and four second-stage regressions (as in Eq. 1), one for each measure of outside directors. Our main …ndings are: …rst, the number of contemporaneous and previous seats in boards is negatively related with performance. Second, internally appointed directors and no synchronism are negative and signi…cantly related to returns. The former result is consistent with the …ndings of Fich and Shivdasani (2006) who show that if a director is sitting simultaneously in many boards he/she may be too “busy”to e¢ ciently monitor the …rm. However, this explanation does not apply as well to the number of previous board seats, as this variable does not proxy for how busy a director is during the time of investment. Overall, our results for the appointed directors indicate that …rms in which there is a weak relationship between the fund and the appointees perform better than …rms in which this relationship is strong. This …nding is consistent with the explanation provided by Lerner (1995) that internal directors are appointed where there is a need for more intensive monitoring. Our …ndings

18

do not reject the possible alternative explanation that inside board members create less value than outside board members. However, we regard this explanation as less likely.

7

The E¤ect of Covenants on Returns

We next examine the e¤ect that covenants have on returns. The model discussed in the previous section is entirely focussed on the information content of covenants in a context of self-selection. A major strength of that model is that it allows for a clear interpretation of the sign of the inverse Mills ratio. The Mills ratio tells us the direction of the selection. On the other hand, the model that we discuss in this section does not allow for a interpretation of the selection variable. However, it has the main strength of allowing for an identi…cation of the treatment e¤ects that covenants have on returns. We borrow our modelling strategy from Goyal (2005), which itself is based on Lee (1978). Adapting Goyal’s model to our context, the decision to include a covenant in the PE contract can be written as the following probit model:

Pr(Ci = 1) = f ( (y1i

y2i ) + Zi0 )

(3)

where Ci is a dummy variable for the inclusion of the covenant in the contract, y1i is the expected return on an investment with covenants, and y2i is the expected return on an investment without covenants; in this way, (y1i

y2i ) is the expected increase in performance when the covenant is

included in the contract, relative to where it is not included. Zi contains other variables that may a¤ect the choice of introducing a covenant apart from the expectations on performance. In the empirical speci…cation, Zi contains …rm and fund-speci…c characteristics, as well as market conditions at the time of the investment. Expected performance is di¤erent when the covenant is included and when it is not, so we model these two separately as a function of vector XiF which we de…ned in the previous section: y1i = XiF 0 19

1F

+ u1i ;

(4)

y2i = XiF 0

2F

+ u2i :

(5)

Equation (4) is the performance regression for investments with covenants and Eq. (5) is for investments with covenants. We cannot estimate Equations (3)–(5) separately because expected performance is conditioned on the inclusion of the covenant or not. We therefore implement a procedure based closely on Goyal (2005) to correct for the bias. The procedure consists in substituting the performance equations (4) and (5) into the covenant choice equation (3). This allows us to obtain a reduced-form model in which covenants are solely a function of …rm-deal characteristics, X F as well as a vector of market condition controls Z M :

Pr(Ci = 1) = f XiF 0

F

+ ZiM 0

:

(6)

Equation (6) is estimated using a probit model with maximum likelihood, and the linear d F0 predictions, ci = X i ( b )=(1

As usual,

F

M 0 are used to calculate the inverse Mills ratio, which is de…ned as + Zd i

( b )) when covenants are not included and

( b )= ( b ) when covenants are included.

is the standard normal density function and

is the standard normal cumulative

distribution function. As Lee (1978) and Maddala (1983) show, consistent estimates of

1F

and

2F

are obtained

by augmenting Eqs. (4) and (5) with the inverse Mills ratio as the right-hand side variables, and estimating the equation with OLS. Finally, by substituting the di¤erence in expected performance for the whole sample, yb1i yb2i , into the structural probit equation (3), we obtain consistent estimates

of the structural probit model parameters

and .

Table 10 reports our estimation of the e¤ects that covenants have on returns by providing an estimation of the structural and outcome equations, respectively equations (3), (4) and (5). The main …nding of Panel A is that the coe¢ cient for yb1i

yb2i is positive and signi…cant for

the covenant index with respect to change in sales, ROA and ROE, while for the performance

covenants the coe¢ cient is signi…cant for sales and ROA. The coe¢ cient of default and protection covenants are never signi…cant. This …nding suggests that the presence of certain types of covenants a¤ects the performance of the …rm, after accounting for the possible self-selection in the choice of 20

covenants. These …ndings corroborate those of the previous section and suggest how performance covenants are particularly e¤ective at increasing the performance of the …rm. We envisage that the mechanism that is at work here relies on the e¤ect that performance covenants have on the incentives of managers. For completeness, Panel B of Table 10 reports the inverse Mills ratios associated with the selection equation (6) of the covenant index.7 Importantly, as explained by Lee (1978) only the signi…cance but not the sign of the coe¢ cient of the inverse Mills ratio is relevant here. The results from this panel con…rm the presence of a selection e¤ect in the choice of covenants. The results on board seats are also con…rmed.

8

Conclusions

In this paper we examine the relationship between the returns of PE investments and contract characteristics. We identify covenants and the appointment of directors as the two main contract features that are relevant for the analysis. We …nd that there is a strong positive relationship between the stringency of contracts and investment returns. This relationship holds for measures of performance that capture the return to VC investors (IRR), to equityholders (ROE), and to all of the …rm’s stakeholders (growth in sales, ROA). And, it is robust after controlling for a number of …rm, industry and investment characteristics. We conjecture that the observed correlations between covenants and returns are caused by a self-selection process in which the involved parties choose the best set of covenants according to their private expectations about the pro…tability of the deal. We use a two-stage methodology that allows us to (i) adjust for potential biases on the returns equations that could result from this selection process, and (ii) test whether covenants are capturing unobservable factors such as private information about future returns, distortions caused by incentive alignment mechanisms, or higher bargaining power of some of the contracting parties. As a by-product of this procedure, we also are able to explore the observable determinants of contracts in PE deals. 7 Similar

panels for the other sets of covenants are reported in the appendix in table A.3.

21

Our central …nding is that contracts contain private information about agents’expectations of future returns, moral hazard concerns, and on how bargaining power is allocated among the contracting parties. We …nd that covenants are generally a good predictor of returns over and above other …rm and deal characteristics. We also …nd that the appointment of board director is informative about returns. More external directors are appointed in …rms with higher returns. Finally, we examine the e¤ect that covenants have on returns, after accounting for selfselection in their use. We show that lockups, rights of …rst refusal and exit ratchets drive higher growth in sales, ROA and ROE, but do not lead to higher IRR for VC investors. We regard this evidence as consistent with the idea that covenants change the behavior of managers and give the incentives to maximize …rm value. Other …ndings in this paper include (i) that funds owned by banks perform on average worse than others which con…rms the results of Hellmann et al. (2008), and (ii) that the choice of covenants is generally independent of observable …rm characteristics, but varies with the macroeconomic conditions at the time of investment.

22

References [1] Aghion P., Bolton, P., Tirole, J., 2004. Exit Options in Corporate Finance: Liquidity versus Incentives. Review of Finance 8, 327–353. [2] Aghion, P., Bolton, P., 1992. An Incomplete Contracts Approach to Financial Contracting. Review of Economic Studies 77, 338–10l. [3] Axelson, U., Jenkinson, T., Strömberg, P., Weisbach, M.S., 2011. Leverage and Pricing in Buyouts: An Empirical Analysis. Working Paper, SIFR. [4] Barry, C.B., Muscarella, C.J., Peavy, J.W., Vetsuypens, M.R., 1990. The Role of Venture Capital in the Creation of Public Companies: Evidence from the Going-Public Process. Journal of Financial Economics 27, 447-471. [5] Bascha, A., Walz, U., 2001. Convertible Securities and Optimal Exit Decisions in Venture Capital Finance. Journal of Corporate Finance 7, 285–306. [6] Brav, A., Gompers, P.A. , 2003. The Role of Lockups in Initial Public O¤erings. Review of Financial Studies 16(1), 1–29. [7] Bygrave, W., Timmons, J., 1992. Venture Capital at the Crossroads. Harvard Business School Press, Boston. [8] Casamatta, C. , 2003. Financing and Advising: Optimal Financial Contracts with Venture Capitalists. Journal of Finance 58(5), 2059–2085. [9] Casares-Field L., Hanka, G., 2001. The Expiration of IPO Share Lockups. Journal of Finance 56(2), 471–500. [10] Cochrane, J. H. , 2005. The Risk and Return of Venture Capital. Journal of Financial Economics 75(1), 3–52. [11] Colla, P., Ippolito, F., Wagner, H., 2011. Leverage and Pricing of Senior and Junior Debt in LBOs. Working Paper, Università Bocconi. [12] Cornelli, F., Yosha, O., 2003. Stage Financing and the Role of Convertible Securities. Review of Economic Studies 70(1), 1 –32. [13] Cumming, D., 2008. Contracts and Exits in Venture Capital Finance. Review of Financial Studies 21(5), 1947–1982. [14] Dessi, R., 2005. Start-up Finance, Monitoring and Collusion. RAND Journal of Economics 36 (2), 255–274. [15] Dunbar, C.G., 1995. The Use of Warrants as Underwriter Compensation in Initial Public O¤erings. Journal of Financial Economics 38, 59–78.

23

[16] Fang, L.H., 2005. Investment Bank Reputation and the Price and Quality of Underwriting Services. Journal of Finance 60 (6), 2729–2761. [17] Fich, E.M., Shivdasani, A. , 2006. Are Busy Boards E¤ective Monitors?. Journal of Finance 61 (2), 689–724. [18] Gompers, P., Lerner J., 1997. Risk and Reward in Private Equity Investments: the Challenge of Performance Assessment. Journal of Private Equity 1, 5–12. [19] Gompers, P., Lerner, J. , 1999. An Analysis of Compensation in the U.S. Venture Capital Partnership. Journal of Financial Economics 51(1), 3–44. [20] Gompers, P., Lerner, J. , 2000. Money Chasing Deals? The Impact of Fund In‡ows in Private Equity Valuations. Journal of Financial Economics 55, 281–325. [21] Gompers, P.A., 1999. Ownership and Control in Entrepreneurial Firms: An Examination of Convertible Securities in Venture Capital Investments. Working Paper, Harvard Business School. [22] Goyal, V.K., 2005. Market Discipline of Bank Risk: Evidence from Subordinated Debt Contracts. Journal of Financial Intermediation 14, 318–350. [23] Groh, A., Gottschalg, O., 2006. The Risk-Adjusted Returns of U.S. Buyouts. Working Paper. [24] Heckman, J.J., 1979. Sample Selection as a Speci…cation Error. Econometrica 47, 153–161. [25] Hellmann, T., 1998. The Allocation of Control Rights in Venture Capital Contracts. RAND Journal of Economics 29(1), 57–76. [26] Hellmann, T., 2006. IPOs, Acquisitions, and the Use of Convertible Securities in Venture Capital. Journal of Financial Economics 81, 649–79. [27] Hellmann, T. F., Lindsey, L., Puri, M., 2008. Building Relationships Early: Banks in Venture Capital. Review of Financial Studies 21(2), 513–541. [28] Holmström, B., 1979. Moral Hazard and Observability. Bell Journal of Economics 10, 74–9l. [29] Inderst, R., Mueller, H., 2004. The E¤ect of Capital Market Characteristics in the Value of Start-Up Firms. Journal of Financial Economics 72, 319–356. [30] Kaplan, S., Schoar, A., 2005. Private Equity Performance: Returns, Persistence and Capital Flows. Journal of Finance 60 (4), 1791–1823. [31] Kaplan, S. N., Strömberg, P., 2002. Financial Contracting Theory Meets the Real World: An Empirical Analysis of Venture Capital Contracts. Review of Economic Studies, 70(2), 281–315. [32] Kaplan, S. N., Strömberg, P., 2004. Characteristics, Contracts, and Actions: Evidence from Venture Capitalist Analyses. Journal of Finance 59(5), 2173–2206. [33] Lee, L.F., 1978. Unionism and Wage Rates: a Simultaneous Equations Model with Quantitative and Limited Dependent Variables. International Economic Review 19, 415–433.

24

[34] Lerner, J., 1995. Venture Capitalists and the Oversight of Private Firms. Journal of Finance 50, 301–318. [35] Lerner, J., Merges, R.P., 1998. The Control of Technology Alliances: An Empirical Analysis of the Biotechnology Industry. Journal of Industrial Economics 46(2), 125–56. [36] Lerner, J., Schoar, A., Wong, W., 2007. Smart Institutions, Foolish Choices?: The Limited Partner returns Puzzle. Journal of Finance 62, 731–64. [37] Li, K., Prabhala, N.R., 2007. Self-Selection Models in Corporate Finance. in: Eckbo, B.E., (Eds.), Handbook of Corporate Finance, Volume I, Elsevier. [38] Ljungqvist, A., Richardson, M., 2003. The Cash-Flow, Return and Risk Characteristics of Private Equity. NYU, Finance Working Paper, (No. 03-001). [39] Metrick, A., Yasuda, A., 2010. The Economics of Private Equity Funds. Review of Financial Studies 23, 2303–2341. [40] Phalippou, L., Gottschalg, O., 2009. The returns of Private Equity Funds. Review of Financial Studies 22(4), 1747–1776. [41] Puri, M., 1996. Commercial Banks in Investment Banking: Con‡ict of Interest or Certi…cation Role?. Journal of Financial Economics 40, 373–401. [42] Renucci, A., 2000. Optimal Contracts between Entrepreneurs and Value-enhancing Financiers. Working Paper, University of Toulouse. [43] Repullo, R., Suarez, J., 2004. Venture Capital Finance: A Security Design Approach. Review of Finance 8(1), 75–108. [44] Sahlman, W. A., 1990. The Structure and Governance of Venture Capital Organizations. Journal of Financial Economics 27, 473–521. [45] Schmidt, K.M., 2003. Convertible Securities and Venture Capital Finance.Journal of Finance 58 (5), 2059–86. [46] Song, W., 2004. Competition and Coalition among Underwriters: The Decision to Join a Syndicate. Journal of Finance 59, 2421–2444. [47] Yates, G., Hinchli¤e, M., 2010. A Practical Guide to Private Equity Transactions.Cambridge University Press.

25

Table 1 Investment Returns IRR Mean Median

ΔSales Mean Median

ΔROA Mean Median

ΔROE Mean Median

0.019 0.086 0.113 0.108 0.126 0.122 0.033 0.128

0.038 0.025 0.068 0.039 0.081 0.051 0.077 0.042

0.031 0.020 0.062 0.031 0.073 0.043 0.069 0.039

0.073 0.054 0.182 0.079 0.222 0.093 0.209 0.082

Type Early Expansion Buyout Turnaround

Exit Trade Sale

0.133 0.071 0.063 0.191 0.109 0.041 0.034 0.080 0.333 0.137 0.117 0.272 IPO 0.293 0.099 0.080 0.179 -0.579 -0.029 -0.026 -0.081 Write-off -0.541 -0.006 -0.003 -0.005 -0.026 -0.001 0.005 -0.012 Un-exited -0.043 -0.036 0.004 0.001 All 0.096 0.067 0.060 0.176 0.107 0.039 0.032 0.074 Notes. This table provides descriptive statistics for investment returns across investment types and exits. We use the following measures of returns, all expressed in yearly terms: IRR, growth in sales, growth in ROA, growth in ROE and exit. Investment type includes: early, expansion, buyout, and turnaround. Exit takes the form of trade sale, IPO, write-off or un-exited.

26

Table 2 Definitions of Covenants and Directors Panel A – Covenants Lockup A provision in the underwriting agreement between some or all existing shareholders that prohibits the sale of shares before a predetermined date. Permitted Transfer The permission to a VC to make transfers of shares without pre-emption in favor of the remaining shareholders. The types of permitted transfers may vary according to the class of shares. There will usually be a permitted transfer provision allowing transfers between two or more separate funds managed by the same VC. Redemption right Rights to force the company to purchase shares (a "put"). A redemption right allows one shareholder to liquidate an investment in the event an IPO or a public merger becomes unlikely. One may also negotiate a redemption provision to become effective when the company defaults or fails to make payments upon a key employee’s death, etc. Tag-along Rights A minority shareholder protection affording the right to include their shares in any sale of control and at the offered price (“right of co-sale”) Drag-along Rights A majority shareholders right, obligating minority shareholders whose shares are bound into the agreement, to sell their shares into an offer the majority wishes to execute. Right of First A negotiated obligation of the company or existing investors to offer shares to the company Refusal or other existing investors at fair market value or a previously negotiated price, prior to selling shares to new investors (“pre-emption right”). Exit Ratchet An exit ratchet is used to adjust the respective shareholdings of the VCs and insiders depending on either the level of returns or on an exit. This clause is principally used to provide additional incentives/rewards to the managers for delivering high returns to investors. Panel B – Directors Cumulative Board Board Seats During Cumulative Board Board Seats Before Employee No Synchronism of Mandate

Cumulative number of appointments held by the VC’s appointee on behalf of the VC during the time of investment. Cumulative number of appointments held by the VC’s appointee on behalf of the VC before the time of investment. The VC’s appointee is or has been an employee of the VC. No synchronism of maturity between the appointment of the VC’s representative and the board of the target firm

27

Table 3 Covenants Type Covenant

Exit

Firms #

All (%)

Early (%)

Exp. (%)

Buy. (%)

Turn. (%)

Trade (%)

IPO (%)

W-Off (%)

Unexit. (%)

Lockup

55

6.59

11.43

5.20

6.64

4.44

2.48

79.07

1.82

18.18

Permitted transfer

157

18.82

13.57

19.62

20.80

17.78

15.17

76.74

20.00

27.27

Redemption right

156

18.71

14.29

17.02

24.78

17.78

15.03

72.09

18.18

54.55

Tag-along right

725

86.93

85.00

87.94

85.40

91.11

86.76

95.35

85.45

72.73

Drag-along right

172

20.62

20.00

21.04

20.35

20.00

20.55

16.28

18.18

54.55

First refusal

73

8.75

9.29

7.57

11.95

2.22

5.24

65.12

5.45

36.36

Exit ratchet

65

7.79

3.57

7.33

11.06

8.89

6.62

25.58

9.09

9.09

Performance

151

18.57

16.08

22.57

13.33

18.57

13.38

95.35

14.55

45.45

Protection

262

26.43

30.26

36.73

31.11

26.43

27.45

93.02

29.09

63.64

Default

751

89.29

89.60

90.27

95.56

89.29

89.93

95.35

89.09

81.82

Notes. This table provides descriptive statistics on covenants. Total sample size consists of 834 deals. Each column gives the % of firms that carries a specific covenant (or set of covenants) within a group defined according to exit (Trade, IPO, Write-off, Un-exited) and type of deal (Early, Expansion, Buyout, Turnaround). “Performance” covenants are: lockup, right of first refusal, or exit ratchets. “Protection” covenants are redemption rights or permitted transfers. “Default” covenants are tag or drag-along rights.

28

Table 4 Univariate Differences in Performance across Covenants Panel A – IRR across covenants IRR if X=0 (%)

IRR if X=1 (%)

t-test

p-value

Wilcoxon z

p-value

Lockup

8.5

22.9

-5.171

0.000

-4.345

0.000

Permitted transfer

8.9

11.7

-1.237

0.217

-1.615

0.053

Redemption right

9.0

11.2

-1.109

0.269

-0.502

0.308

Tag-along right

5.0

10.1

-1.732

0.086

-2.507

0.006

Drag-along right

9.5

8.9

0.35

0.726

2.27

0.012

First refusal

8.9

14.3

-1.79

0.077

-1.122

0.131

Exit ratchet

7.6

30.4

-5.378

0.000

-7.054

0.000

Performance

7.0

20.2

-5.668

0.000

-5.282

0.000

Protection

8.8

10.7

-1.057

0.291

-0.686

0.246

Default

7.0

9.7

-0.824

0.412

-0.659

0.255

ΔSales if X=0 (%)

ΔSales if X=1 (%)

t-test

p-value

Wilcoxon z

p-value

Lockup

6.3

12.7

-3.789

0.000

-5.296

0.000

Permitted transfer

6.5

7.4

-0.785

0.433

-0.713

0.238

Redemption right

6.9

5.9

1.053

0.293

0.55

0.291

Tag-along right

6.5

6.7

-0.18

0.857

0.128

0.449

Drag-along right

6.9

5.7

1.337

0.182

1.86

0.031

First refusal

6.6

7.6

-0.643

0.522

-1.671

0.047

Exit ratchet

5.4

21.5

-5.61

0.000

-6.636

0.000

Performance

5.2

13.5

-5.60

0.000

-6.496

0.000

Protection

6.8

6.4

0.526

0.599

0.578

0.282

Default

6.9

6.7

0.203

0.839

0.924

0.178

X

Contracting styles:

Panel B – Change in sales across covenants X

Contracting styles:

Notes. This table shows the difference in mean performance of investments when covenants are not present relative to when covenants are present. We use the following two measures of performance: IRR (Panel A) and growth in sales (Panel B). Covenants include: lockup, permitted transfer, redemption right, tag-along right, drag-along right, right of first refusal, exit ratchet. We also form groups of covenants according to the following contracting styles (derived from MCA): “Performance” (dummy for the inclusion of either a lockup, right of first refusal or exit ratchet), “Protection” (dummy for presence of either a permitted transfer or a redemption right), and “Default” (dummy for presence of either tag- or drag-along rights). To test for differences we employ a t-test and a Wilcoxon test.

29

Table 5 Regressions of Performance Measures on Covenants Panel A – Performance and Covenant Index (1) OLS ΔSales

(2) OLS IRR

(3) OLS ΔROA

(4) OLS ΔROE

(5) Probit IPO

(6) Probit Write-off

0.016*** (0.004) -0.042 (0.032) 0.027*** (0.009) -0.110 (0.122) -0.016 (0.012) 0.111 (0.069) -0.001 (0.001) -0.024 (0.110)

0.042*** (0.007) -0.040 (0.062) 0.075*** (0.022) -0.335 (0.227) -0.018 (0.017) 0.285* (0.167) 0.000 (0.003) -0.150 (0.161)

0.012*** (0.003) -0.037 (0.029) 0.027*** (0.008) -0.045 (0.089) -0.019 (0.011) 0.083 (0.064) -0.004*** (0.001) -0.028 (0.069)

0.028** (0.011) -0.221** (0.095) 0.069*** (0.024) -0.055 (0.329) -0.069 (0.043) 0.157 (0.245) -0.012** (0.005) -0.071 (0.233)

1.480*** (0.160) 0.772** (0.368) 0.568** (0.237) -6.972** (2.732) 0.010 (0.257) 6.936*** (2.080) -0.074 (0.049) -8.649*** (2.567)

-0.064 (0.065) 0.122 (0.341) -0.160 (0.137) 3.988** (2.033) -0.035 (0.131) -0.687 (1.138) 0.008 (0.021) -5.368*** (1.675)

834

834

834

834

789

823

0.118

0.095

0.102

0.061

(0.695)

(0.089)

Industry FE

YES

YES

YES

YES

YES

YES

Investment FE

YES

YES

YES

YES

YES

YES

Year FE

YES

YES

YES

YES

YES

YES

Investm. Duration

NO

NO

NO

NO

YES

YES

Covindex Profitability Size Leverage Bank owned % Shares Board Seats Dur. Constant

Observations Adj (Pse) R

2

30

Panel B – Performance and Single Covenants (1) (2) OLS OLS Covenants IRR ΔSales Lockup Adj. R2

(3) OLS ΔROA

(4) OLS ΔROE

(5) Probit IPO

(6) Probit Write-off

0.072*** 0.118

0.187*** 0.096

0.052*** 0.101

0.095*** 0.059

3.892*** 0.636

-0.700 0.095

Perm.Trans Adj. R2

0.015* 0.097

0.038** 0.066

0.010 0.090

-0.001 0.057

1.503*** 0.348

0.018 0.088

Redemption Adj. R2

-0.004 0.094

0.030* 0.064

-0.005 0.089

-0.036 0.058

1.612*** 0.313

0.012 0.087

Tag-along Adj. R2

0.006 0.094

0.042 0.065

0.002 0.088

-0.021 0.057

1.039*** 0.125

-0.111 0.088

Drag-along Adj. R2

-0.001 0.094

0.009 0.063

-0.004 0.089

-0.022 0.057

-0.127 0.0987

-0.113 0.088

First Refus. Adj. R2

0.016 0.095

0.079*** 0.070

0.005 0.089

-0.027 0.057

2.104*** 0.386

-0.237 0.089

Exit Ratchet Adj. R2

0.152*** 0.230

0.207*** 0.113

0.138*** 0.204

0.552*** 0.171

0.777*** 0.128

0.003 0.087

Performance Adj. R2

0.087*** 0.186

0.147*** 0.112

0.076*** 0.159

0.273*** 0.112

2.223*** 0.420

-0.139 0.0255

0.002 0.094

0.028 0.065

0.000 0.088

-0.023 0.057

1.871*** 0.348

-0.063 0.088

Protection Adj. R2 Default Adj. R2

0.003 0.026 0.001 -0.016 0.716** -0.162 0.094 0.063 0.088 0.057 0.110 0.088 Notes. This table examines the relationship between covenants and performance in a multivariate setting. In Panel A we regress each measure of performance on the covenant index, also controlling for firm and investment characteristics, and investment, year, and industry fixed effects. In Panel B each performance measure is regressed on each covenant and for each contracting style (performance, protection, default), also controlling for firm and investment characteristics, and investment, year, and industry fixed effects. For each regression in Panel B we report only the coefficient on the relevant covenant and the adjusted R-squared. Columns indicate the measures of performance used as the dependent variable. Covenants include: lockup, permitted transfer, redemption right, tag-along right, drag-along right, right of first refusal, exit ratchet. Pre-investment firm characteristics include: profitability, size, leverage, industry and bank-owned dummy. We also control for the cumulative number of board seats of appointee in other boards during the investment (board seats). Investment characteristics include: market to book ratio of equity as priced in the investment, investment type (early, expansion, buyout, turnaround), and investment date. We also include a dummy for un-exited deals. The covenant index is the sum of all covenants. All regressions have 782 observations. Standard errors are clustered at the management company level. Standard errors are in parenthesis. ***, **, and * indicate respectively significance at the 1, 5, and 10% levels.

31

Table 6 Predicting the Use of Covenants

Profitability Size Leverage Bank owned % Shares Board Seats Market Returns Prev. 6m Industry ROE IPO/New Observations Pseudo R2 Wald: Chi squared Wald: Prob. >Chi squared Industry FE Investment FE

(1) Probit Covenant Index > Median

(2) Probit Performance

(3) Probit Default

(4) Probit Protection

0.578 (0.378) -0.129 (0.096) -1.303 (1.171) -0.052* (0.029) 0.016 (0.094) -2.359*** (0.748) 0.930*** (0.148) 0.062*** (0.024) 1.489*** (0.380) 830 0.083 15.85 0.0012 YES YES

0.621* (0.354) -0.032 (0.100) -1.648 (1.237) 0.020 (0.032) -0.081 (0.108) -0.845 (0.784) -0.036 (0.110) 0.417*** (0.160) 0.045 (0.028) 830 0.0702 17.54 0.0005 YES YES

0.262 (0.501) -0.075 (0.120) -1.620 (1.466) -0.030 (0.039) -0.310** (0.137) -3.723*** (0.965) 0.286 (0.214) 0.088** (0.036) 1.859*** (0.596) 830 0.117 23.16 0.0000 YES YES

0.261 (0.340) -0.239** (0.095) 0.357 (1.195) -0.057* (0.030) 0.057 (0.098) -1.372* (0.759) 0.765*** (0.150) 0.064*** (0.024) 1.233*** (0.397) 830 0.072 45.28 0.0000 YES YES

Notes: This table reports the probit estimations of Eq. 2. The independent variable varies in each column, and is a dummy for the inclusion of the covenant indicated at the top row. RHS variables are: profitability, size, leverage, bank ownership, % shares, cumulative board seats of director in other boards, industry dummies, investment type dummies, and market conditions: market returns, average industry ROE, and ratio of IPOs to new firm creation during the 6 months previous to the investment. We also report the Wald test on the significance of the instruments in the first stage. ***, **, and * indicate respectively significance at the 1, 5, and 10% levels.

32

Table 7 Inverse Mills Ratios of Covenants across Different Return Measures Dependent variable in 1st stage (2) (3) Performance Default

(1) Covenant Index > Median

(4) Protection

Dependent variable in 2nd stage IRR ΔSales ΔROA ΔROE

IRR – ΔSales IRR – ΔROA IRR – ΔROE

0.042*** (0.011) 0.016*** (0.005) 0.016*** (0.005) 0.056*** (0.019)

0.078*** (0.013) 0.040*** (0.005) 0.037*** (0.005) 0.131*** (0.021)

0.017 (0.015) 0.003 (0.007) 0.001 (0.007) -0.007 (0.027)

0.016 (0.011) 0.001 (0.005) 0.000 (0.005) -0.014 (0.020)

0.026*** (0.009) 0.026*** (0.009) -0.013 (0.018)

0.034*** (0.010) 0.040*** (0.011) -0.072*** (0.020)

0.014 (0.012) 0.016 (0.013) 0.024 (0.025)

0.016* (0.009) 0.016* (0.009) 0.030* (0.018)

Notes. This table reports the coefficients of the inverse Mills ratios across different 2 nd stage regressions. Each column identifies a set of 2nd stage regressions (Eq. 1 augmented with the inverse Mills ratio), which share the same 1 st stage regression (Eq. 2). Each row identifies a set of regressions, which share the same dependent variable in the 2nd stage. For example, column 1 contains the coefficients on the inverse Mill’s ratios in Eq. 1, obtained by estimating Eq. 2 for the covenant index (>median) in the 1st stage. In the last three rows, the dependent variables of the 2 nd stage are respectively the differences between IRR and ΔSales, IRR and ΔROA, and IRR and ΔROE. ***, **, and * indicate significance at the 1, 5, and 10% levels.

33

Table 8 The Appointment of Board Directors Early 6.90 3.00

Exp. 6.79 2.81

Type Buy. 6.49 3.10

Board Seats Dur.

Mean SD

Board Seats Bef.

Mean SD

6.60 3.45

6.34 3.49

6.77 3.38

6.34 3.63

7.13 2.87

6.64 3.43

6.19 3.79

7.02 3.18

3.18 2.44

Employee

%

0.46

0.46

0.46

0.44

0.42

0.45

0.65

0.47

0.27

#

0.50

0.50

0.50

0.50

0.50

0.50

0.48

0.50

0.47

% #

0.37 0.48

0.39 0.49

0.37 0.48

0.38 0.49

0.29 0.46

0.37 0.48

0.47 0.50

0.31 0.47

0.45 0.52

Synchronism of Mandate

Turn. 6.67 2.77

Trade 6.80 2.85

IPO 6.26 3.47

Exit W-Off 6.65 2.78

All 6.72 2.92

Unexit 3.45 4.03

Notes. This table provides descriptive statistics for our four measures of outside. We calculate statistics on all the sample and within groups defined by type of investment (Early, Expansion, Buyout, Turnaround) and by type of exit (trade, IPO, write-off, un-exited). For board seats (in other boards) during and board seats (in other boards) before, we report mean and standard deviation. For employee and synchronism of maturity, we report percentage and number of deals for which the appointed director is or has been a fund employee, or whose mandate coincides with the maturity of the target board.

34

Table 9 Appointment of Directors and Returns

Covenants in 1st stage

(1)

(2)

(3)

(4)

IRR

ΔSales

ΔROA

ΔROE

Regr. #

Covenant Index > Median Performance Protection Default

A. Cumulative Board Seats During the Investment Period -0.003 -0.003** -0.005*** -0.016*** -0.002 -0.003** -0.005*** -0.015*** -0.003 -0.003** -0.005*** -0.016*** -0.002 -0.003** -0.005*** -0.016***

1 5 9 13

Covenant Index > Median Performance Protection Default

B. Cumulative -0.003 -0.003 -0.003 -0.003

2 6 10 14

Covenant Index > Median Performance Protection Default

-0.007 -0.006 -0.006 -0.006

Board Seats Before the Investment Period -0.003*** -0.005*** -0.016*** -0.003*** -0.005*** -0.017*** -0.003** -0.004*** -0.016*** -0.003** -0.004*** -0.016*** C. Employee -0.022*** -0.023*** -0.022*** -0.023*** -0.022*** -0.023*** -0.022*** -0.023***

-0.048* -0.047* -0.048* -0.048*

3 7 11 15

D. No Synchronism of Mandate Covenant Index > Median 4 -0.016 -0.014** -0.018*** -0.022 Performance 8 -0.015 -0.013** -0.018*** -0.020 Protection 12 -0.016 -0.014** -0.019*** -0.024 Default 16 -0.017 -0.014** -0.019*** -0.024 Notes. This table reports the estimated coefficients of the following regressors (Cumulative Board Seats During, Cumulative Board Seats Before, Employee, No Synchronism of Mandate), as computed in the 2nd stage of a selectivitybias adjusted regression also including the following controls: inverse Mills, profitability, size, leverage, % shares, industry, investment, year FE, and VC fixed effects. More precisely, we run sixteen 2nd stage regressions, as follows: A) covenants in the 1st stage is a dummy equal 1 if Covenant Index > Median. Regressors in 2 nd stage are controls plus: #1: Board Seats Before, #2: Board Seats During., #3: Employee, #4 No Synchronism. B) covenants in the 1st stage are “Performance” (lockup, first refusal, exit ratchet). Regressors in 2nd stage are controls plus: #1: Board Seats Before, #2: Board Seats During., #3: Employee, #4 No Synchronism. C) Covenants in the 1st stage are “Protection” (permitted transfer or redemption). Regressors in 2 nd stage are controls plus #5: Board Seats Before., #6: Board Seats During, #7: Employee, #8: No Synchronism, D) Covenants in the 1st stage are “Default” (tag- or drag-along). Regressors in 2nd stage are controls plus: #9: Board Seats Before., #10: Board Seats During, #11: Employee, #12: No Synchronism. ***, **, and * indicate significance at the 1, 5, and 10% levels respectively.

35

Table 10 The Effect of Covenants on Returns Panel A – Estimation of the Structural Equation (1)

(A) Covenant Index > Median (B) Performance (C) Default (D) Protection Observations

Coefficient for the Difference in Returns (y1 – y2) (2) (3)

(4)

IRR

ΔSales

ΔROA

ΔROE

0.049 (0.122) -0.067 (0.090) -0.062 (0.063) -0.175 (0.179) 830

0.863*** (0.279) 0.395** (0.175) -0.036 (0.151) -0.373 (0.349) 830

0.806*** (0.289) 0.374** (0.174) -0.040 (0.184) -0.228 (0.615) 830

0.156** (0.072) 0.056 (0.043) -0.042 (0.039) -0.092 (0.140) 830

Panel B – Inverse Mills Ratios and Board Seats in the Outcome Equations Inverse Mills 1 Board Seats During Observations Inverse Mills 0 Board Seats During

-0.138* (0.081) -0.003 (0.022) 375

Covenant Index =1 if > Median -0.198*** -0.179*** (0.041) (0.042) -0.002 -0.026** (0.002) (0.011) 375 375

-0.097 -0.103*** (0.105) (0.036) -0.006 -0.003* (0.004) (0.002) 455 455 Chi2( 3) =18.95

-0.736*** (0.179) -0.006 (0.008) 375

-0.072** -0.191 (0.035) (0.132) -0.005*** -0.013** (0.001) (0.005) 455 455 Prob > Chi2 =0.0003

Observations Wald on Instruments in the first stage Note. This table contains estimated coefficients for selected variables in equations (3)-(5), as obtained through the twostage procedure described in Lee (1978) and Goyal (2005). Panel A contains estimated coefficients of sixteen different structural equations (3). The dependent variable in Panel A varies in each row and are dummy variables that take a value of one when: (A) the number of covenants included in the PE contract is larger than the median; (B) the contract contains at least one “Performance” covenant (lockup, exit ratchet, right of first refusal); (C) the contract contains at least one “Default” covenant (tag- or drag-along); (D) the contract contains at least one “Protection” covenant (redemption or permitted transfer). Independent variables are market returns, average industry ROE, and ratio of IPOs to new firm creation during the 6 months previous to the investment, plus the differences in: IRR (column 1), ΔSales (Column 2), ΔROA (Column 3), ΔROE (column 4) when sets of covenants A, B, C, or D are included relative to when they are not included.. In Panel B we report coefficients for the inverse mills ratio and the number of board seats during the investment for the selection-adjusted performance equations corresponding to contracts with more covenants than the median (375 observations) and contracts with less covenants than the median (455 obs), for each of the four

36

performance measures as indicated at the top of each column. These equations also control for , profitability, size, leverage, % shares, industry, investment, year FE, and VC fixed effects.

37

Appendix Table A.1 Characteristics of Investments Panel A –Investment size by vintage year and IRR by exit year Vintage year

Exit year

Investment size (million €)

Annualized IRR (%)

N

Mean

Median

N

Mean

Median

1999

64

6.46

3.68

1

48.12

48.12

2000

247

7.00

4.10

10

32.62

49.31

2001

212

6.11

3.80

32

35.07

30.00

2002

56

6.55

4.10

99

8.88

13.24

2003

68

8.23

5.23

157

12.51

12.41

2004

125

6.31

4.10

166

8.74

9.75

2005

62

6.87

4.38

92

9.39

10.28

2006

-

-

-

139

6.25

9.76

2007

-

-

-

78

8.89

8.60

2008

-

-

-

22

-11.25

2.75

2009

-

-

-

15

-8.40

0.79

2010

-

-

-

12

2.41

3.46

2011

-

-

-

11

-2.61

-4.33

Total

834

6.69

4.10

834

9.40

10.62

38

Panel B –Investment size by type and exit N

Mean

Investment size (millions of €) Median Min

Max

StDev

Type Early

140

1.05

1.00

0.04

6.15

0.70

Expansion

423

4.14

3.80

0.01

60.80

3.35

Buyout

226

14.66

15.40

0.01

30.40

5.36

Turnaround

45

8.10

8.25

1.80

25.00

3.46

725

6.72

4.10

0.04

60.80

6.32

43

7.50

4.20

0.50

24.50

6.99

55

6.28

4.00

0.01

21.50

6.11

11

3.70

2.18

0.30

20.74

5.82

834

6.69

4.10

0.01

60.80

6.34

Exit Trade Sale IPO Write-off Un-exited

All

Notes. This table provides descriptive statistics for the investments contained in our sample. Panel A of this table contains the distribution of investments by vintage year (number of deals and investment size) and by exit year (number of sample deals exited each year and IRR). Panel B reports summary statistics of the investment size by type of investment (early, expansion, buyout, turnaround) and by exit type (trade sale, IPO, write-off, un-exited).

39

Table A. 2 Summary Statistics of Firms, Funds and Market Conditions Mean Median Variable Definition

Firm and Fund Characteristics

IRR ρ

Δ Sales ρ

Covenant Index ρ (p-values)

(p-values)

(p-values)

Profitability

Pre-investment EBIDTA divided by pre-inv. book value of assets

0.169 0.114

0.141 (0.000)

-0.068 (0.046)

-0.079 (0.022)

Size

Log of pre-investment book value of assets

4.212 4.205

-0.023 (0.503)

0.198 (0.000)

0.184 (0.000)

Leverage

Pre-investment ratio of debt to Assets

0.791 0.799

-0.081 (0.018)

-0.012 (0.730)

-0.021 (0.539)

% shares

Percentage of shares acquired by VC

Duration

Log of duration of investment (months)

Bank Ownership

=1 if management company (MCO) is owned by a bank

Previous Bank Rel.

=1 if firm had previous relationship with bank owning MCO

0.227 0.250 3.516 3.569 0.607 1.000 0.488 0.000

-0.064 (0.063) 0.034 (0.315) -0.021 (0.546) 0.001 (0.982)

-0.020 (0.553) -0.394 (0.000) -0.034 (0.325) -0.033 (0.342)

-0.053 (0.123) -0.659 (0.000) -0.069 (0.042) -0.096 (0.005)

6.718

-0.007

0.047

-0.012

(0.844) 0.006 (0.865) 0.010 (0.773)

(0.174) 0.026 (0.455) 0.051 (0.133)

(0.738) 0.123 (0.000) 0.080 (0.020)

Board Directors Cumulative Board Board Seats During Employee

=1 if board director is an employee

Synchronism

=1 if synchronism between appointment of director and board of target firm

7.000 0.456 0.000 0.366 0.000

MktRet. 6m

Returns on Italian equity market (S&P MIB) over previous 6m.

0.165 0.153

0.160 (0.000)

0.114 (0.001)

0.089 (0.009)

Ind. ROE

Avg. Italian industry ROE at time of investment

8.071 7.805

0.092 (0.007)

-0.055 (0.107)

-0.043 (0.212)

IPO/New

Avg. ratio of IPOs to new firms created in Italy over prev. 6m

0.156 0.119

0.010 (0.773)

-0.100 (0.003)

-0.170 (0.000)

Market Conditions

Notes. This table provides definitions and summary statistics for firm and fund characteristics, directors and market conditions, as well as their pairwise correlation with IRR and growth in sales. Covenant index (Cov. Index) is defined as the sum of all covenants.

40

Table A.3 Inverse Mills Ratios and Board Seats in the Outcome Equation Panel A – Performance Covenants (1) IRR

(2) ΔSales

(3) ΔROA

(4) ΔROE

0.091

0.121

0.130

0.466

(0.171)

(0.117)

(0.124)

(0.583)

0.004

-0.003

-0.011**

-0.052**

(0.007)

(0.005)

(0.005)

(0.025)

148

148

148

148

0.102

0.083*

0.056

0.174

(0.146)

(0.047)

(0.045)

(0.158)

Performance Inverse Mills 1 Presences D. Observations Inverse Mills 0 Presences D. Observations Wald on Instruments in the first stage

-0.001

-0.003**

-0.004***

-0.009**

(0.003)

(0.001)

(0.001)

(0.004)

682

682

682

682

Chi2( 3) =

17.59

Prob > Chi2 =

0.0005

(1) IRR

(2) ΔSales

(3) ΔROA

(4) ΔROE

Panel B – Default Covenants

Default Inverse Mills 1 Presences D. Observations Inverse Mills 0 Presences D. Observations Wald on Instruments in the first stage

-0.523*** (0.126) -0.008** (0.003) 747

-0.307*** (0.053) -0.004** (0.001) 747

-0.253*** (0.052) 0.001 (0.001) 747

-0.952*** (0.206) 0.003 (0.006) 747

-0.132 (0.214) -0.008 (0.014)

-0.169** (0.071) -0.006 (0.005)

-0.136* (0.077) -0.006 (0.005)

-0.436 (0.368) -0.016 (0.024)

83 Chi2( 3) =

83 17.71

41

83 Prob > Chi2 =

83 0.0005

Panel C – Protection Covenants (1) IRR

(2) ΔSales

(3) ΔROA

(4) ΔROE

-0.031 (0.073) 0.007 (0.005)

-0.056** (0.028) 0.001 (0.002)

-0.092 (0.081) -0.002 (0.005)

-0.024 (0.023) -0.001 (0.002)

259

259

259

259

-0.171** (0.084) -0.004 (0.004)

-0.143*** (0.036) -0.003* (0.002)

-0.133*** (0.037) -0.005*** (0.002)

-0.558*** (0.151) -0.018*** (0.007)

Protection Inverse Mills 1 Presences D. Observations Inverse Mills 0 Presences D. Observations Wald on Instruments in the first stage

571 Chi2( 3) =

571 42.31

42

571 Prob > Chi2 =

571 0.0000