The Levered Returns of Leveraged Buyouts: The Impact of Competition*

Transcription

1 The Levered Returns of Leveraged Buyouts: The Impact of Competition* Reiner Braun Technische Universität München (TUM) Center for Entrepreneurial and Financial Studies Nicholas Crain Vanderbilt University Owen Graduate School of Management Anna Gerl Technische Universität München (TUM) Center for Entrepreneurial and Financial Studies Current version, November 2015 First version, March 2015 ABSTRACT This paper investigates the relationship between leverage and returns in private equity buyout transactions. In contrast to the predictions of traditional capital structure theory, we find that transactions financed with large amounts of debt are associated higher transaction prices and lower returns to private equity sponsors. Consistent with the view that easy credit amplifies the intensity of bidding for deals, these relationships hold only when private equity buyers face competition from other funds, such as in deals sourced from investment bank auctions. Our results are distinct from changes in deal prices driven by private equity fundraising and the results are robust to alternative, plausibly exogenous, proxies for the competitiveness of deals. Finally, we show that the choice to pursue auction deals in particularly loose credit markets, when expected returns are low, is positively related to proxies for agency conflicts between fund managers and fund investors. Keywords: Leverage, pricing, returns, competition, agency conflicts, leveraged buyouts * Contact information: Braun: reiner.braun@tum.de, Crain: nicholas.crain@owen.vanderbilt.edu, Gerl: anna.gerl@tum.de. We would like to thank Andres Almazan, Jean-Noël Barrot, Sheridan Titman, Robert Parrino and seminar participants at both the 2015 Coller Private Equity Findings Symposium and the University of Texas for valuable comments.

2 JEL classification: G32, G34 1 Introduction In 1989, the private equity firm KKR raised nearly $21 billion of debt to finance the public-toprivate buyout of RJR Nabisco. The debt, which amounted to over six times the company s trailing EBITDA, provided several possible benefits. The tax savings would be large and immediate. Servicing the debt would require steep cuts in the somewhat infamous perquisites enjoyed by the firm s incumbent management. Further, KKR believed that it would be difficult for other bidders to find intermediaries with the experience and capacity to raise the capital required to compete for the deal. However, with RJR Nabisco s stable cash flows and the strength of the late-80 s high yield bond market, several bidders competed for the deal, backed by investment banks hoping to burnish their reputations and rise to the top of the leveraged buyout (LBO) financing league table. 1 To win the deal, KKR was forced to raise its bid by nearly $4 billion. Eventually the fund would record a loss of $730 million on the investment (Norris, 2004). In this paper, we investigate how the performance of private equity buyout deals are related to the amount of debt used to finance their purchase. The literature on private equity has largely focused on how leverage may affect the value of the target firm. 2 However, a private equity buyout involves both a recapitalization of the target firm and a transfer of ownership. As the anecdote about KKR s experience with RJR Nabisco illustrates, private equity deal returns are driven by both the value added by private equity ownership (including potential benefits from additional leverage) and the price that can be negotiated with the sellers. Axelson et al. (2013) find the price paid by private equity firms to acquire a portfolio company is positively related to the amount of debt used to financing the purchase. This suggests sellers of the target firm are better off when more debt is available to finance the deal, but tells us little about the value captured by private equity. 3 It could be that high levels of debt correspond to 1 Deal values were taken from ThomsonOne M&A database. Burrough and Helyar (2008) provide a journalist account of the bidding process for RJR Nabisco. 2 Jensen (1989) suggests that portfolio company leverage helps mitigate managerial agency problems. Kaplan and Strömberg (2008) suggest private equity funds may have an ability to time mispricing in debt markets. In addition, many papers have commented on the potential tax benefits of the additional debt used in buyout transactions. 3 In addition, Axelson et al. (2013) find that fund-level returns are negatively related to the average Debt / EBITDA among a fund s deals. However, the identification of this effect relies on a small minority of deals from each fund that are captured in commercial deal leverage data. Our focus on the returns to individual deals offers two advantages. First, it gives us power to conduct tests in the cross-section, particularly with respect to the competitive environment in which the deal takes place. Second, conditional on a fund entering our sample, we have a nearly complete record of leverage and performance for each of its deals. 1

3 the largest increases in the value of the target firm (through tax savings, for example). Sellers may capture a portion of this increase through higher deal prices, but the return to the private equity sponsor may be higher as well. Alternatively, the increase in deal price may come at the expense of private equity sponsors - in which case, we would expect the returns to private equity sponsors to decline with leverage. Our evidence comes from a large sample of deal-level data provided by several private equity fund-of-funds (FoFs). The data was collected during the FoFs due diligence investigations of buyout fund managers attempting to raise a new fund. It contains all the previous investments of fund managers, whether or not the manager successfully raised financing from the FoF. We are able to identify a large sample of 3,198 deals for which we have information on both deal performance and leverage; the latter is taken directly from the description of sources and uses of capital in the transaction. We begin by considering the relationship between the realized returns to individual buyout investments and the amount of debt used to finance the deal. We focus primarily on the ratio of Debt to the target firm s EBITDA, a measure of leverage-relative-to-firm fundamentals that PE industry participants often use as a metric of the debt available to do a deal. We focus on Debt / EBITDA rather than Debt / Equity, which is more common in the capital structure literature, because the amount of equity contributed to the deal is itself an endogenous outcome of the bargaining between buyer and seller. We find a strong negative relationship between returns and Debt / EBITDA. One additional turn of Debt / EBITDA corresponds to a nearly 2% decrease in the expected Internal Rate of Return (IRR) to the deal s private equity sponsor. This relationship holds after controlling for deal characteristics (industry, region, etc.) that may be related to systematic risk. The relationship also holds in the cross-section, which suggests that it is not likely to be driven by time-varying changes in discount rates or other macro factors that may drive investment returns. Rather the causing lower returns, it seems more plausible that the negative relationship between Debt / EBITDA and returns is an equilibrium outcome of the deal process, particularly the competitive environment surrounding the deal. A portion of our sample, 947 deals, contains information on whether the target firm was purchased through an investment bank auction or sourced from proprietary deal flow. We find that the negative relationship between Debt / EBITDA and returns only holds for auction deals. For deals sourced through auctions, one 2

4 additional turn of Debt / EBITDA corresponds to a nearly 5% lower IRR. We find no evidence of this relationship in proprietary deals. Two concerns limit our confidence in the evidence on deal source alone: First, the portion of our sample that contains information on deal source is relatively small. A larger sample would provide a more powerful test. Second, given that owners of the target may choose the method by which they sell the firm, it is natural to worry that the process by which deals select into an investment bank auction or proprietary deal is not independent of the expected price sellers hope to receive. We address these concerns by splitting the sample based on alternative proxies for the competitive environment surrounding a deal. The first is the Enterprise Value (EV) of the deal. 4 For both investment banks and PE funds, small deals produce less benefit (fees and potential gains) while requiring similar costs (marketing and due diligence) to larger deals. As a result, small deals are less likely to be sold through an auction and less likely to receive interest from a large number of potential investors. Consistent with our intuition about the effects of competition, we estimate that the relation between Debt / EBITDA and returns in large deals is twice as strong as for small deals. Second, we split the sample on the amount of capital recently raised by buyout funds specializing in the industry and region of the target firm. Gompers and Lerner (2000) show that the ratio between the amounts of capital committed to the PE industry and available investment opportunities varies widely, leading to changes in the competition for deals. 5 We find evidence of a much stronger negative relation between Debt / EBITDA and deal returns following periods of high fundraising. We then examine how both debt and deal price change with credit market conditions. As in Axelson et al. (2013) we interpret credit market conditions as an exogenous determinants of the amount of leverage lenders are likely to provide. For the full sample and high competition subsamples good credit conditions (low spreads in the high yield bond market) are related to both larger amounts of debt and higher prices. However, in low competition subsamples we find no evidence that credit market conditions are related to leverage or price. This suggests that the leverage agreed to by lenders and private equity borrowers itself may be affected by the competition surrounding the deal - for example, if private equity funds forced to pay higher prices because of competition, do so by seeking additional debt. Returns would be low and 4 The potential endogeneity of EV is addressed in the methodology section. 5 In the appendix, we show that this money-chasing deals channel is separate from the effect of leverage on returns. Intuitively, the primary channel for the effect of recent fundraising on deal price (and thus lower returns) should be more equity in the deal, not more debt. 3

5 debt levels high, particularly for deals where the competitive advantage of the winning fund is small. Finally, we examine which funds pursue competitive deals. Malenko and Malenko (2015) suggest that low reputation fund managers will have difficulty obtaining debt when credit market conditions are poor because of concerns that they will expropriate lenders. Axelson, Strömberg and Weisbach (2009) present an alternative model with similar implications. They point out that the time limit on a PE fund investment period may cause fund managers with unspent capital to pursue bad deals rather than let capital commitments expire. Lenders willingness to provide debt capital to these agency-conflicted investors (some of whom may have good opportunities) is determined by credit market conditions. The implication of both models is that managers with reputation or agency concerns are likely to win more deals when credit market conditions are good and their access to debt capital is less constrained. Accordingly, we examine how a fund s flow of new deals varies with credit market conditions and proxies for firm reputation. We find that the deal flow of funds whose interim performance is lower than that of their peers is particularly sensitive to credit spreads. This also holds for the sub-sample of auction deals. Poorly performing funds win more auctions, particularly when credit spreads (and expected returns) are low. In contrast, we find little evidence that interim performance is related to the rate that funds complete proprietary deals. This suggests that credit conditions predominantly effect poorly performing funds when they face competition from other buyers who may have better access to capital. Aside from the papers cited above, our results build on the literature related to the financing and performance of buyout PE deals. Demiroglu and James (2010) and Ivashina and Kovner (2011) find that fund manager characteristics (reputation and bank relationships, respectively) are related to a fund manager s ability to raise debt on favorable terms. Our results suggest that any of the rents created by preferential access to debt capital are likely to decrease as credit conditions improve and the competition for deals increases. Jenkinson and Stucke (2011) find that the estimated tax savings from debt in public-to-private LBOs are positively related to acquisition premiums. Our results show a corresponding effect on the returns to private equity sponsors of large, competitive deals like the ones in their sample. However, our results suggest that private equity sponsors may still earn rents when leverage is harder to obtain and for deals that are less competitive, as the expected reduced taxing savings from debt are compensated by the benefits of lower deal prices. 4

6 Our study also contributes to the literature on competition and its influence on the private equity industry. Gompers and Lerner (2000) introduce the money chasing deals phenomenon, arguing that increasing levels of fundraising lead to more intense competition among VC firms for a finite amount of attractive investment opportunities. Our results contrast with those of Guo, Hotchkiss and Song (2011) who propose that club deals, in which PE firms form bidder syndicates, may help to reduce competition among PE funds. Finally, our results relate to the empirical literature on macroeconomic conditions and buyout fund returns. Robinson and Sensoy (2013) show that when funds invest their capital during economic expansions (when high levels of leverage are also available), the performance is poor. Similarly, Kaplan and Strömberg (2009) provide further evidence for the countercyclicality in fundraising and performance for buyout funds. An alternative explanation is that private equity investments may require a risk or liquidity premium that is particularly high when economic conditions are poor (e.g., Franzoni, Nowak and Phalippou (2012) and Haddad, Loualiche and Plosser (2015)). While we find that leverage (and thus returns) are driven by credit conditions that are clearly related to broader economic conditions, we also find that the negative relationship between leverage and returns is present in the cross-section. The remainder of the paper is organized as follows: Section 2 describes the construction of the sample, while Section 3 provides empirical evidence concerning the effect of the competitive environment on leverage, pricing and returns. Section 4 examines the relation between agency conflicts, deal leverage and returns, and Section 5 concludes. 2 Data 2.1 Data Sources The primary basis for our analysis of buyout investments is a large, proprietary database compiled by three European fund-of-fund managers. The database includes fund-level and investment-level information on venture capital, private equity and other forms of alternative assets. The subsample of buyout investments contains more than 13,500 portfolio companies from around the globe, sponsored by 1,016 funds over a period from 1974 to One unique feature of this database is that it contains detailed information at the deal-level, including monthly gross cash flows between the fund and the portfolio company. The data with rich information about fund manager, fund and investment come from the FoFs due diligence process. PE FoFs are intermediaries that pool capital, typically from 5

7 institutions, and invest in PE funds. In exchange for fees, investors are able to allocate a portion on their portfolio to PE, while delegating the process of evaluating and performing due diligence on a large number of potential PE fund investments to the FoFs. PE fund managers seeking an investment from the FoF are asked early in the process to provide the full track record of historic and current funds, and respective deals, since inception. The data in our sample includes all the fund managers and funds the FoFs performed due diligence on, including those in which the FoF decided not to invest. This mitigates some of the concerns regarding selection bias, which are discussed in detail below. Every time a fund manager approaches the FoF to commit money to a new fund (and a due diligence process is started), the record on past funds and investments is updated. The most recent updates on some fund managers are from 2007, while some entries were brought up-to-date as late as mid We merge the three individual due diligence databases and eliminate duplicate funds and deals. Drawing on the full record of timed deal-level cash flows we are able to calculate deal-level performance gross of fees for all investments. For the small number of past investments that were not realized (9%) or were partially realized (22%) at the time due diligence was performed, the database reports the Net Asset Value (NAV). For these investments, we use the corresponding NAV as a proxy for cash flows to the fund and compute deal-level performance using this information. So far, few studies have had access to such a rich multi-level data set that includes investment-level performance information. Braun, Jenkinson and Stoff (2015) also use some information from the same database as a starting point to study performance persistence among buyout fund managers; Lopez-de-Silanes, Phalippou and Gottschalg (2013) investigate the performance dispersion and determinants of PE investments, drawing on a sample of PE firms providing their respective private placement memorandum. In this paper we take advantage of another unique feature of the FoF database. For a subset of more than 3,000 buyout investments, the data contains financial details of the transaction. We observe the enterprise value determined in the transaction, as well as the amounts of debt and equity used. Further, our sample contains the earnings before interest, taxes, depreciation and appreciation (EBITDA), as well as Industry Classification Benchmark (ICB) code and country. Finally, for another subset of these buyout investments the database indicates whether the fund manager directly acquired the portfolio company from the seller, or whether the purchase was made through an investment-bank run auction. To our knowledge, this large-scale private equity study is the first one that links performance with such transaction details. 6

8 We follow Axelson et al. (2013) and match our sample of PE buyouts against public market counterparts in the same year and, industry drawn from The Center for Research on Security Prices (CRSP) and Standard & Poor's COMPUSTAT (North America) database. 17 From Thomson Reuters Datastream/Worldscope, we obtain debt market and macroeconomic information to gauge the effect of the market environment for PE. The major variable to show the effects of debt markets on PE comes in form of the US high-yield spread according to the Merrill Lynch High-yield Master Index, tracking the performance of belowinvestment grade, US-dollar denominated corporate high-yield bonds publicly issued in the US domestic market, minus US LIBOR. To account for the size of the economy, we provide additional information on macroeconomic conditions, such as the Gross Domestic Product (GDP) for each country in each year. In the last part of our analysis, we investigate the factors influencing the number of auctions in which a fund manager participates. In order to obtain more information on the interim state of the fund, we benchmark against those with interim performance information in the Preqin PE database. As Axelson et al. (2013) find this commercial database a reliable source for PE sponsor characteristics, we are confident in using fund information, such as the percentage of investment amount already called at the time of the investment or fund performance variables. 2.2 Sample representativeness: Selection bias For any database in the notoriously opaque PE asset class, there is an inherit challenge to capture the investable universe to ensure representativeness. We are aware of potential sample selection issues in our dataset that might originate from the following major channels: First, our sample could be systematically flawed by omitting fund managers that did not seek capital commitments from one of the three FoFs due to unobserved characteristics. Second, some fund managers may avoid raising capital from FoFs, instead favoring direct relationships with the institutional investors. However, having data from due diligence performed over different years mitigates these selection issues. The private equity asset class is characterized by strong boom and bust cycles in terms of fundraising (e.g., Gompers and Lerner (2000)), and thus changing power dynamics in the GP-LP-relationship. As fund managers experienced fundraising difficulties especially in the latest financial crises in the 2000s, they were forced to 17 Public companies are drawn from COMPUSTAT North America. In a subsample of only North American PE buyouts, we find qualitatively comparable results to the full international dataset. A correspondence table between SIC codes and one-digit ICB codes was created using current firms in Thomson Reuters Datastream, which contains both codes. 7

9 extend their investor base and thus are likely to be part of the due diligence process of our FoFs. Third, as we take into account transactions with different realization status, we face a rather low probability of underestimating the poor performance of funds not yet fully divested. Since the FoFs force the fund managers to show all their past and current investments with complete information, both performing and underperforming, our data set is unlikely to suffer from any reporting bias unlike public commercial databases or data sourced by single LPs. Related to the data compilation, we cannot rule out another source of survivorship bias: When unsuccessful PE firms decided to quit the business and did not contact our FoFs for capital commitments (again), they are unavoidably not part of our data compilation. This might particularly apply to poorly performing first-time funds. However, Braun, Jenkinson and Stoff (2015) argue that this is a fairly infrequent phenomenon. In addition, our data sample contains funds of fund managers that failed to raise sufficient amounts of capital commitments and thus were never closed. Overall, we are optimistic that our dataset, derived from large institutional investors directly, is not biased towards a non-random sample in any significant way. For a further discussion of the overall sampling process and potential sample selection biases, please also refer to Braun, Jenkinson and Stoff (2015). Another source of bias in this study might be due to the fact that we rely on self-reported transaction details. While we do not know whether (and, if yes, why) the three FoFs selectively asked some fund managers to provide additional details on their transactions, such as information on EBITDA or debt, it is reasonable to assume that the likelihood to voluntarily report additional details increases with success. Unless forced to report details on all historic deals, fund managers may selectively report such details on their most successful deals to make a good impression to the FoFs as potential LP investors. Therefore, restricting the sample on buyouts for which these details are observable may introduce some additional positive selection bias. However, in the next section we will introduce main sample characteristics in terms of leverage, pricing and performance, and show that our final sample compares very well with existing studies. 2.3 Sample characteristics Table 1 provides a detailed overview over the composition and various characteristics of our data set at portfolio company-level. The overall sample, after restricting to buyout deals only being eligible for our research setting, includes 3,198 investments from 442 funds made in the investment period between 1986 and

10 The EV/EBITDA multiple 18 is an aptly used measure to analyze the transaction price of a portfolio company and is calculated as the ratio of enterprise value (EV) to earnings before interest, taxes, depreciation and amortization (EBITDA) at investment entry. The median EV/EBITDA pricing multiple in our total sample is 6.7x. Axelson et al. (2013) report a higher median value of 7.6x for a sample of 1,009 buyouts. The major reason for this discrepancy is the difference in average transaction size between the two samples. Their sample, obtained by combing several commercial databases, contains relatively large transactions with a median EV of $677 million (mean: $1,514 million). They report that this is higher than the median EV of $63 million (mean: $330 million) in the entire Capital IQ sample. Panel A in Table 1 displays that EVs in our sample are much closer to these values. The median EV in our sample is $87 million (mean: $299 million). However, the subsample of the 25% largest transactions in our sample is fairly comparable with Axelson et al. (2013). The median EV of these 798 transactions is $674 million (mean: $967 million). The median EV/EBITDA pricing multiple in this subsample is 8.01x (mean: 8.51x) and even slightly higher than in Axelson et al. (2013). As expected, the buyout transactions in our sample is highly levered. The median Debt / Equity ratio in our sample is 1.49 (mean: 2.05) and indicates that on average about 60% (mean: 66%) of the deal value is financed with debt. However, likely because buyout funds focus on firms with strong cash flows, the amount of debt used in the transactions seems much more modest when measured relative to the cash flow being produced by the firm. We find a median Debt / EBITDA of 3.93 (mean: 4.05). Both these measures are smaller than the leverage reported in Axelson et. al., but this appears to be predominantly related to the size of transactions captured in the two samples. The mean Debt / Equity ratio of 2.4 for the largest EV quartile in our sample indicates a debt share of approximately 71%, which is close to the 69% reported in Axelson et al. (2013). Similarly, the mean Debt / EBITDA value in the subsample of largest transactions in our sample is 5.06 and very close to the value of 5.2 in their study. In addition, Table 1 displays descriptive statistics by time categories (Panel B) and by regions (Panel C). Since this paper deals with the effect of competition, we also distinguish how the fund manager has acquired the asset for a subsample of 947 portfolio companies. We differentiate between portfolio companies sold via a competitive investment bank auction and those that were acquired via a proprietary sales process. 19 In such an auction, the owners of the 18 We winsorize the variables used in our sample at a 3% level to exclude extreme values and ensure comparability. 19 An investment bank s auction process closely matches the individual value, English auctions used to model competition between funds in Malenko and Malenko (2015). 9

11 target firm employ an investment bank that solicits initial bids from a large number of potential buyers. 20 Out of the respondents to the initial round of solicitation, the investment bank helps to select a portion of the respondents to participate in the future rounds. In each round, bidders are granted more access to proprietary firm information with which to perform due diligence and asked to submit an updated bid. Eventually the process settles on a winning bidder. For buyout fund managers, seeking deal flow through auctions has relatively low expected search costs as the amount of resources required to perform due diligence grows in each round with the probability of winning the deal. The marketing efforts of the investment bank and the relatively low entry costs ensure the participation of many bidders. 21 For a buyout fund manager, the alternative to building a portfolio via winning investment bank auctions is to generate proprietary deals. In this case, the portfolio company is sold in a first chance acquisition and the private equity fund manager, as the buyer, is the first one to purchase the portfolio company. In general, these deals involve high search costs to identify potential targets that are not marketing themselves for sale. Practitioners we have spoken with describe extensive networking and even cold-calling large numbers of firms that meet a particular investment thesis. Potential targets may directly approach a private equity fund that has developed a reputation for expertise in a particular industry or geographic area, but evaluating these deals remains costly because of the due diligence involved. While proprietary deal flow is costly to attract, private equity funds face less competition. When raising a new fund, buyout fund managers often market the share of their portfolio that was obtained from proprietary deals. Panel D in Table 1 shows that in our sample performance is quite similar for both groups. However, the average deal sold via an auction is significantly larger. The median EV for auction deals is $136 million and more than twice the size than the median proprietary transaction with $48 million. When rescaled by EBITDA, this difference in pricing is much less pronounced. Nevertheless, the median EV/EBITDA pricing multiple for auction transactions is with 6.85x still higher than the 6.43x median value for proprietary deals. Auctioned deals in our sample are substantially more levered. The median Debt / EBITDA value for auctions of 20 Bankers typically approach both financial and strategic buyers. Our sample and corresponding analysis consists of only deals won by private equity (financial) buyers. Previous evidence suggests that the level of competition from both types of buyers are correlated, though good credit environments may favor financial buyers. See Gorbenko and Malenko (2014) and Martos-Vila, Rhodes-Kropf and Harford (2013). 21 Gorbenko and Malenko (2014) find an average of 16.5 participants in investment bank auctions of public targets that were eventually purchased by private equity buyout funds from 2000 to Our discussion with practitioners suggests that auction participation is somewhat lower for sales of private firms, but remains very competitive. 10

12 4.31 is much higher than the 3.65 for proprietary deals. With a 1.71 to 1.17 difference, the pattern is the same for the Debt / Equity ratio. Unlike commercial databases and most previous literature, we have all monthly deallevel cash flow information gross of fees, i.e. before management and performance fees, to ensure comparability among different limited partnerships, between the portfolio company and the general partner, as well as valuations (NAV) for unrealized portfolio company investments. Consequently, we are able to compute deal gross Internal Rate of Returns (IRR). The top line in Table 1 shows that the median deal gross IRR in our sample of 3,198 buyouts is 27.7%. This value is comparable with Lopez-de-Silanes, Phalippou and Gottschalg (2013) who report a median IRR of 21.0% for their data set of 7,452 buyout deals. 3 Investment performance, pricing, leverage and competition 3.1 Leverage and returns We begin by establishing some stylized facts about the correlations between the leverage used to finance a deal and the returns generated on the equity contributed by the private equity fund. We focus on the gross IRR of the deal, but obtain similar results using the return multiple and the public market equivalent of Kaplan and Schoar (2005). Previous research has shown that deals where the buyer is able to obtain high leverage are also deals where the buyer pays a higher price (Axelson, et al. (2013)). By looking at the returns to the private equity sponsor, rather than the purchase price, we provide more direct evidence of whether PE funds are capturing a higher NPV. Figure 1 presents graphical evidence that the expected returns of private equity investments are increasing in Debt / Equity. We sort deals into quintiles according to Debt / Equity, then compute the median gross IRR in each quintile. The figure s bottom panel displays the range of leverage used to fund deals in each quintile. Returns are monotonically increasing in leverage measured as Debt / Equity. The lowest leverage quintile, with Debt / Equity ranging from 0 to 0.63, has the lowest return with a median IRR of 21%. The highest leverage quintile, with a Debt / Equity ratio ranging from 2.8 to 568.3, has a median return of 35%. Figure 2 presents the equivalent pattern of returns with deals sorted on Debt / EBITDA. Quintile 1 represents deals with the lowest leverage; Debt / EBITDA among these deals ranges from 0 to Quintile 5, the highest leverage quintile, has Debt / EBITDA ranging from 5.68 to The top panel presents the mean gross IRR to deals in each quintile. Over very modest amounts of leverage, the relationship matches the prediction from traditional corporate finance 11

13 theory. From Quintile 1 (the deals with the lowest leverage) to Quintile 2, we observe a substantial increase in average returns of about 8%-points. After this initial increase, returns are monotonically decreasing with leverage over the remaining quintiles. The drop between Quintiles 2 and 5 is over 13%-points in IRR. Thus, unconditionally, over 80% of the sample returns are strongly decreasing in Debt / EBITDA. We extend the analysis of Figures 1 and 2 through ordinary least squares (OLS) regressions where gross IRR for each deal is the dependent variable and leverage (as Debt / Equity or Debt / EBITDA) is the main explanatory variable. The regressions control for deal characteristics, such as portfolio company industry, that may drive both returns and leverage. Dummy variables representing each quartile of deal enterprise value are included to account for risks that may be correlated with target firm size. In all specifications, we control for the realization status of the investment and ten ICB industries to account for industryspecific risk. Furthermore, all specifications include fund fixed effects to account for different approaches, e.g., investment styles, and different fund manager abilities. Standard errors in these (and all following) regressions are clustered at the LBO deal-year level. The corresponding results reported in Table 2 are consistent with the patterns evident graphically in Figures 1 and 2. In Column 1, when leverage is measured as Debt / Equity at entry, we find a strongly significant positive effect on deal gross IRR. One unit increase in the Debt / Equity ratio corresponds to a 3.6% change in IRR. However, interpreting the coefficient is difficult because of endogeneity concerns about the measurement of the firm s equity. Our measure of Equity comes from the sources of capital used to finance the purchase price. It represents the amount of equity capital contributed to the deal, rather than the value of equity immediately following the transaction. If the transfer to private equity ownership increases firm value, then our measure of equity is likely to be biased low and the magnitude of this bias is tied to the outcome of bargaining between the buyer and seller. Thus, for the remainder of the paper we predominantly focus on Debt / EBITDA. In Column 2 we find a negative and significant relation between buyout deal Debt / EBITDA at entry and gross deal IRR. Increasing the amount of debt used to finance the deal by one turn corresponds with a decrease in expected IRR of 1.7%. This suggests that private equity firms perform poorly in deals that are highly levered relative to the firm s earnings. One concern about this interpretation is that the Debt / EBITDA available to buyout funds may be correlated with macro factors that drive expected returns. For example, Haddad, Loualiche and Plosser (2015) argue that aggregate changes in risk premia may drive buyout 12

14 returns and leverage. Controlling for investment year fixed effects in Column 3 shows that the negative association between Debt / EBITDA and returns continues hold in the cross-section, such that the relationship is unlikely to be driven by time series macro-factors. In Column 4 we include the Price / Dividend ratio of the S&P 500 as a proxy for time varying aggregate risk premia. Consistent with the argument that time varying risk premia affects private equity returns, the coefficient on the S&P 500 P/D ratio is negative and statistically significant. However, the negative coefficient on Debt / EBITDA remains is slightly larger, and remains statistically significant. 3.2 Leverage, returns and competition We next examine how the relationship between leverage and returns varies with the competitive environment in which a deal takes place. If the negative relation between Debt / EBITDA and returns documented in Table 2 is an equilibrium outcome of the competition between private equity funds, then we would expect the relationship to be strongest for deals that are heavily marketed and receive interest from many funds or strategic acquirers. We repeat the regressions of gross deal IRR on leverage over subsamples that differ in characteristics that are likely to affect competition. In Columns 1 and 2 of Table 3, we split the sample according to the source of the deal. Panel A presents results from the pooled sample, while Panel B includes investment year fixed effects. Column 1 exhibits results for a subsample of 387 LBOs that are classified as proprietary deals. These deals were directly negotiated between the seller and private equity fund manager without the seller widely soliciting other interest. In both the pooled sample, and the sample with investment year fixed effects, we find that Debt / EBITDA is unrelated to deal returns in proprietary deals. In Column 2 of each panel we report results from identical regressions on a subsample of 560 deals that were sold by an investment bank auction with multiple bidders. In each case the coefficient on Debt / EBITDA is statistically significant and negative. If the portfolio company is acquired through a competitive auction, one additional turn of leverage is associated with a 5.1%-points lower deal gross IRR. Including investment-year fixed effects in the analysis results in a negligible decrease in the size of the coefficient. In the remaining columns of Table 3, we present additional regressions over subsamples that are split based on alternative proxies for the competitive environment surrounding each deal. The goal of this analysis is to address two concerns regarding the subsamples based on 13

15 deal source. First, only about one third of the deal observations in our sample contains information about the source of the deal. By relying on alternative proxies that are available for all deals, we mitigate concerns that previous results are driven by factors affecting the probability of observing deal source in the data. Second, it may be the case that the observed source of a deal is an endogenous outcome of strategic choices by the seller, who even when approached directly by a PE fund, may choose to initiate an investment bank auction when it would be likely to produce a higher price. Given these concerns, the ideal alternative proxy would be correlated with the competition surrounding a deal, available for all deals in the sample, and, in the spirit of an instrumental variable, would be relatively unaffected by strategic choices of the seller. We consider both the size of the deal and the magnitude of recent inflows into the private equity industry. Small deals (those where the enterprise value of the firm is low) are likely to receive less interest from rival PE funds. Private equity deals require similar levels of diligence regardless of size, and thus firms willing to invest in pursuing a small deal are only likely to do so when there are few competitors and the expected probability of winning the deal is high. 22 As a result the competition for smaller deals is less intense and that a small company is much more likely to be acquired in a proprietary sourcing process than larger transactions (Preqin, 2014). While we view deal size as a proxy for the general level of competition surrounding a deal rather than the specifics of the sale process, in unreported probit regressions we confirm that consistent with our intuition deal size is positively related to the likelihood that a deal is sourced from an investment bank auction. We construct subsamples of small and large deals by splitting the deals at the median EV of $87 million. Columns 3 and 4 of Table 3 present estimates of these regressions for a subset of 1,599 small and 1,599 large deals, respectively. We find that the relationship between Debt / EBITDA and deal gross IRR is stronger in larger deals, albeit with modest statistical significance. The results in Column 3 of Panel A suggest that for larger deals in our sample, an additional turn of Debt / EBITDA ratio is associated with a 2.4% lower return. This value amounts to 1.3% and is statistically insignificant for smaller companies given in Column 4. The point estimates for deal size subsamples in Panel B are nearly identical, but the coefficient 22 Although on the margin deal size is likely to be related to the ease of accessing credit, the assumption required is that very few deals have switched across the boundary set at median deal size from being a small to large deal as a result. 14

16 on Debt / EBITDA in large deals is not statistically significant at the 10% level (the p-value for the coefficient is 0.13). As an additional proxy for the level of competition surrounding a deal, we consider the total buyout funds raised in the same industry and region three years prior to the respective transaction, divided by current year s regional gross domestic product (GDP) (see Braun, Jenkinson and Stoff (2015) for more details on this variable). This proxy is motivated by the money chasing deals phenomenon documented by Gompers and Lerner (2000), who find evidence of increased competition following high levels of fundraising in the US venture capital industry. One concern about this proxy is that excessive capital flowing into the private equity industry may be responsible for higher deal prices. In Appendix A, we show evidence from our sample that flows into the private equity industry has an effect on deal prices, but that it s primarily related to an increase in the equity capital contributed to deals. Columns 5 and 6 of Table 3 show the corresponding regression results using a subsample of 1,594 deals in a low competitive and 1,604 deals in a high competitive PE fundraising environment, respectively. In line with the findings presented above, we observe that the coefficient on Debt / EBITDA is much stronger at a significant level, in economic and statistical terms, when a company acquired a portfolio company in a highly competitive environment. While the effect of one additional turn of leverage is 1.1% and statistically insignificant when there is few money chasing deals (Column 5), it is 3.1% and highly statistically significant when competition in the buyout market is high (Column 6). The negative correlation between Debt / EBITDA and deal returns has consistently higher economic magnitude and statistical significant for deals which are likely subject to competition between multiple private equity funds and strategic acquirers. One interpretation is that observed Debt / EBITDA is a proxy for the ease of obtaining leverage for the deal. Confidence in this interpretation suffers to the extent that Debt / EBITDA is endogenous. For example, Debt / EBITDA is likely to be correlated with unobservable differences between target firms, such as growth prospects. One remedy would be to find one or more instruments for ease of obtaining leverage. Finding relatively strong instruments for available leverage (such as changes in credit market conditions) is plausible. However, the second stage is likely to lack power in the face of both noise created by the first-stage estimation 15

17 of Debt / EBITDA and the noise associated with realized returns. 24 Instead, we argue a more powerful test is to apply an instrumental variables approach to the deal price. Price does not incorporate shocks to portfolio company value following the deal, and corresponds directly to our conjecture that leverage in competitive deals is associated with a transfer of value from the buyer to the seller. 3.3 Leverage, price and debt market conditions In this section we analyze how the leverage used in private equity buyouts and the price paid for portfolio companies respond to credit market conditions. The goal is to examine how deals with different levels of competition respond to plausibly exogenous changes in the leverage available to private equity bidders. For competitive deals we expect improving credit markets to be associated with more leverage and higher deal prices. For deals which are less competitive we expect to find no change in deal price with credit market conditions. This would suggest that differences across competition in returns documented in the previous section are driven by the value captured by the seller through higher deal prices. It s less clear how we should expect leverage in less competitive deals to respond to credit market conditions. If leverage increases as credit markets improve it would suggest that leverage is driven by similar factors regardless of the competitive environment. In contrast, if leverage in non-competitive deals is unrelated to credit market conditions then it suggests that competition itself helps determine the leverage of the deal. Our proxy for the credit market conditions in the highly levered debt market in which private equity funds raise financing is the spread between the Merrill Lynch High-Yield Master Bond and LIBOR (HY Spread). Axelson et. al (2013) find the same measure of HY spread to be a significant determinant of leverage in their sample of buyout transactions Deal Leverage In Table 4 we regress log (Debt / EBITDA) from our sample of buyout deals against HY spread. To control for differences in leverage which may be associated with size, we include dummy variables corresponding to the quartile of EV in which the deal falls. We control for other firm characteristics in two ways. In some specifications we include the median log (Debt / EBITDA) observed in a COMPUSTAT firms in the same ICB industry and fiscal 24 In unreported results we do not find statistically significant evidence that credit market conditions are related to returns. 16

18 year in which the deal takes place. In other specifications we include industry and region fixed effects. Panel A in Table 4 reports results from OLS regressions of LBO leverage on a set of explaining factors for the final sample of 3,198 deals. 25 For the total sample, we detect a statistically significant and negative relation between the high-yield credit spread and Debt / EBITDA leverage (Column 1). A one-unit higher high-yield credit spread is associated with a 1.9% lower Debt / EBITDA ratio. In this regression, we find the coefficient on median log of Debt / EBITDA of comparable public benchmark companies to be statistically insignificant. In Column 2, we find the relationship between high-yield spread and LBO Debt / EBITDA to be robust to excluding public matched leverage and including industry and region fixed effects instead. In line with our previous findings regarding the determinants of LBO pricing (and consistent with Axelson et al. (2013) for larger buyouts), Panel B of Table 4 reveals that highyield credit spread significantly drives both LBO leverage measures and that there is no strong link between industry and LBO leverage when competition is strong, but not if it is weak. In none of the regressions, using our proprietary deal subsample (Columns 1 and 2) the high-yield spread is economically or statistically relevant. However, in Column 1 the coefficient on matched public Debt / EBITDA is significant and positive. In turn, in Debt / EBITDA regressions on the auction subsample (Columns 3 and 4), coefficients on high-yield credit spread are both statistically significant and negative. If competition is high, time-series variation in debt market conditions determines LBO Debt / EBITDA leverage. Further, matched industry leverage is insignificant for auctions. These findings reinforce our interpretation of competition as a major channel in the usage of leverage in private equity and its impact on pricing and returns. Table 5 substantiates this picture using our alternative proxies for the competitive environment surrounding the deal. Credit conditions play a more relevant role in explaining the cross-sectional variation of LBO Debt / EBITDA leverage for large deals (Panel A) and those investments done in a high PE fundraising environment (Panel B). 25 In several of these estimations, we lose observations, as the median EBITDA value of the matched public benchmark sample is negative. Hence, Debt / EBITDA leverage is not numerically interpretable for these observations. 17

19 3.3.2 Deal Pricing and Competition We next examine the relation between the price at which deals take place and leverage used to finance the deal for evidence that sellers are capturing more of the value from leverage when competition is high. While similar in spirit to the analysis of the effects of leverage on deal price in Axelson et al. (2013), the analysis in this section extends their results in by examining subsamples with different levels of competition. Given the results in the previous section, we expect to find no connection between leverage and returns in low competition deals where changes in credit market conditions are not associated with additional leverage. Analyzing the relation between leverage and deal price presents an omitted variables problem which precludes simply running a regression with leverage as an explanatory variable. Our measures of relative price (EV / EBITDA) and leverage (Debt / EBITDA) are nearly certain to be correlated because of unobservable characteristics of the portfolio company (e.g., future expectations of growth). Similar to Axelson et al. (2013) we use the spreads of high-yield corporate bonds as a source of exogenous variation in credit market conditions at the time of the transaction. A low high-yield credit spread indicates low costs for levering up a LBO transaction and is therefore interpreted as loose credit market conditions. In this study, we obtain high-yield spread for each buyout by deducting the US LIBOR rate from the Merrill Lynch High-yield Master Index, both measured at investment entry. We include the high-yield spread directly in regressions with EV / EBITDA as the dependent variable. In addition, we estimate an instrumental variables model, where high-yield spread is used as an instrument in the first-stage to capture exogenous variation in Debt / EBITDA. We note that in most developed economies leverage increases with firm size. Potential reasons are that diversification increases and the probability of financial distress decreases with firm size. As a result, lower bankruptcy costs enable firms to take up more debt (Rajan and Zingales, 1995). Hence, changes in the general credit market conditions may well have a different marginal effect on the changes in LBO leverage, and ultimately prices, contingent on firm size. To control for these effects, we include dummy variables for the EV at investment entry in all specifications. Panel A in Table 6 displays results for our full sample of 3,198 buyout deals. In order to control for the economy-wide changes in discount rates or expectations of growth, we also include the log EV multiple of industry matched public firms in Column 1. We find high-yield spread to be significantly negatively related to LBO pricing. If debt is cheap, fund managers pay higher EV / EBITDA prices for a given firm. In economic terms, a 1% increase in HY 18