Journal of Banking & Finance 35 (2011) 2099–2110

Contents lists available at ScienceDirect

Journal of Banking & Finance
journal homepage: www.elsevier.com/locate/jbf

The effect of leverage on the cost of capital of US buyouts
Alexander Peter Groh a,⇑, Oliver Gottschalg b
a
b

EMLYON Business School, 23 Avenue Guy de Collongue, 69134 Ecully, France
HEC School of Management, 1 Rue de la Libération, 78351 Jouy en Josas, France

a r t i c l e

i n f o

Article history:
Received 1 June 2010
Accepted 14 January 2011
Available online 23 January 2011
JEL classification:
G11
G24
G32
Keywords:
Private equity
Cost of capital
Buyout
Benchmarking alternative assets

a b s t r a c t
This paper addresses the problem to assess the effect of leverage on the cost of capital for buyout performance analyses. It draws on a unique and proprietary set of data on 133 US buyouts between 1984 and
2004. For each of them, we determine a public market equivalent that matches it with respect to its timing and its systematic risk. We show that under realistic mimicking conditions, the average cost of capital
is below the commonly used benchmark S&P 500. Thereby, we control for two important aspects: for the
risks taken by lenders in the buyout transactions (which affects the sponsors’ risks), and for the corresponding cost of debt (which lowers the return of the public market equivalent). Only with borrowing
and lending at the risk-free rate is the average cost of capital close to the average index return. This finding is particularly important as existing literature on that topic tends to rely on benchmarks without a
proper risk-adjustment.
Ó 2011 Elsevier B.V. All rights reserved.

1. Introduction
Since the late 1970s buyouts1 have become an important asset
class with significant economic impact. Yet relatively little is known
about the risk and return characteristics of this type of investment.
This is largely due to two factors. First, buyout investments differ
substantially from public market investments in several important
characteristics, especially regarding liquidity and information
symmetry. This implies theoretical challenges with respect to the
assessment of their risk and return. Second, buyout investments
are a sub-category of the private equity (PE) asset class for which
general disclosure requirements do not exist. In the absence of
detailed information on the underlying transactions and their investment characteristics, risk-adjusted returns are impossible to calculate. To prove that is the major contribution of our paper, we
assess the cost of capital of buyout transactions by public market
investments with an equal risk-profile. For this assessment, we draw
on a unique and proprietary set of data on target-company and
industry characteristics, as well as the applied financial structures
⇑ Corresponding author. Tel.: +33 (0) 4 78 33 78 00; fax: +33 (0) 4 78 33 61 69.
E-mail addresses: groh@em-lyon.com (A.P. Groh), gottschalg@hec.fr (O.
Gottschalg).
1
In the literature, buyout transactions are variously labelled (e.g., leveraged
buyout, management buyout, institutional buyout, management buy-in, etc.) and
these terms are often used synonymously. In this paper the term ‘‘buyout’’ is
preferred as being the broadest covering the different facets of this transaction type.
0378-4266/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jbankfin.2011.01.012

of 133 US buyouts. Based on this information, we construct a mimicking portfolio of investments in the S&P 500 Index, with additionally borrowed or lent funds to adjust for the applied degrees of
leverage. These mimicking investments match the buyouts in terms
of the timing of their cash flows and their systematic risk pattern.
The systematic risk of buyout transactions usually changes during
the holding period. Being initially high due to the amount of debt
used for financing the transaction, the risk decreases in the following
periods as debt is repaid. Our mimicking investments exactly replicate this evolution of the buyout risk patterns over time.
The chosen public market equivalent is not intended to present
a ‘‘buyout pricing model’’ nor shall it imply that we can replicate
buyouts correctly with traded securities. We simply propose an approach to benchmarking them in what we regard as at least necessary, and likewise the best possible way. In fact, we discuss the
rationale and the necessary framework for buyout benchmarking
and refer to standard asset pricing models to control for the
systematic risks of buyouts. We adopt the perspective of a welldiversified investor, such as a fund of fund investor, pension fund
or an endowment. This is a reasonable assumption as these investors are the primary sources of capital for buyout transactions.
Consequently, we do not consider idiosyncratic risks in our analyses because these investors are not affected by idiosyncratic
shocks. Our stylized model follows the alternative decision to
either invest in buyouts or in quoted assets. Thereby, we control
for the systematic risk involved, but not for other factors.

They presume that PE . we show in numerous sensitivity analyses and robustness checks that buyout transactions are more successful relative to their public market benchmark if the buyout fund managers are able to transfer substantial parts of the leverage risk to the lenders or if they are able to reduce operating risks by superior monitoring. this comes at some cost. reorganization. Kaplan and Schoar (2005). He compares his results with the returns of the S&P 500 Index and with several portfolios taken from the NASDAQ Index and detects alphas ranging from 22% to 45%. the average buyout cost of capital is 9. Kaplan and Stein (1990. inflows and management fees for investments in 73 different PE-funds.08 and an average annual internal rate of return of 21. leads to average (median) cost of capital of 12. The authors argue that. under what we regard as the most realistic mimicking approach. Taking into account these operating risk reductions and the risks taken by lenders very well explains lower than S&P 500 returns.83%. However. we claim that an appropriate replication of buyouts with public market securities should consider two aspects: First. Groh. without data on the leverage of the target companies.95%). they also find that funds exposed to more idiosyncratic risk earn higher returns than more diversified portfolios. and potential changes of the operating risk-profile of the buyout targets. As a conclusion. cost of borrowing and lending. a risk-adjusted premium exists for the PE transactions. which greatly exceeds the S&P 500 Index performance during the same period of 14. given the perception of buyouts as high risk transactions. we see no other reason why our results should not be generalized. It seems surprising that the return distribution of the mimicking transactions shifts towards lower values compared to the index. The key papers are discussed below before we highlight what our paper adds. different approaches to correct for sample selection biases and to adjust for risk may be responsible for a large part of these inconsistencies. they are unable to correct for different degrees of leverage. Despite these arguments and the actual willingness of lenders in various countries to participate in buyout transactions. Second. arguing that they play an important role that must be priced. Related literature Strikingly. the median return is slightly smaller. Due to limited data availability for other countries. such as Phalippou and Gottschalg (2009). Muscarella and Vetsuypens (1990) prove that these operational performance improvements result from cost cutting efforts and efficiency gains. lenders take a substantial part of the leverage risks in buyout financing. The first aspect lowers the risks borne by the buyout sponsors. These important findings are supported in the buyout literature.P. As we will show in this section. Kaplan and Stein (1990) calculate an average decrease of systematic operating risks by even 25% for corporations undergoing leveraged recapitalizations. research on the risk and return of private equity has lead to contradictory findings. With his reweighing procedure. receive new financing or are acquired by third parties. However. provided the degrees of leverage were no higher than about twice the industry average. We show that. Palepu (1990). We illustrate that it is not appropriate to assess the performance of buyout transactions without thoroughly determining leverage ratios. However. O. Legislations on corporate governance and investor protection affect the role and scope of active investors and the design of debt contracts. the transaction structures and vehicles are equivalent in most of the other countries. Relaxing our approach. They find that investors in PE-funds do not earn positive alphas. 2. Therefore. Furthermore. Cochrane (2005) calculates an arithmetic mean return of 59% and underlines the high idiosyncratic risks of the particular transactions. They obtain an average beta factor of all the different PE-fund portfolios of 1. Bongaerts and Charlier (2009). and risk sharing with lenders. They find high Sharpe-ratios of the buyouts but face a severe selection bias in their sample of transactions. and second. and assuming perfect market conditions with riskless borrowing and lending. the detected deviation from the S&P 500 returns results from controlling for two components: for the risks taken by lenders in the buyout transactions (which determines the sponsors’ risks).59%). Kaplan (1989a).16%. They argue that this risk reduction is caused by cutting fixed cost. and Ljungqvist and Richardson (2003). Groh et al. Cumming and Zambelli (2010) or with respect to taxation can favor or compromise the use of debt and private equity. as data limitations depending on their data sources make our proposed way of risk assessment impossible in these studies. In addition. These events are more likely to occur when good returns have already been experienced.g. risks borne by lenders. they acknowledge that their PE-fund sample may not be a random draw from the general population. and for the corresponding leverage cost (which lowers the return of the public market equivalent). They determine risk-adjusted returns according to Fama and French (1997).89% (12. it is important to note that many studies do not sufficiently differentiate between the different risk-characteristics of the venture capital and the buyout asset class. and a median of 14.79%) below what would have been earned on average (medial) if the investors had invested in the time-matching S&P 500 portfolio (yielding an average return of 12. International differences with respect to regulation. Kaplan and Schoar (2005) employ a public market equivalent approach to benchmark PE transactions using a large sample from the Thomson Venture Economics database. we apply this benchmarking approach in an empirical analysis on a unique data set and reveal the importance of a correct specification of risks taken by lenders. Ljungqvist and Richardson (2003) use extensive data obtained from a fund of fund investor on cash outflows. described in Cumming and Johan (2007). Smith (1990) report strong operating improvements after the closing of buyout transactions without laying-off employees or reducing R&D costs. First. and therefore implicitly assume average industry debt/equity ratios within their analysis. (2008) determine idiosyncratic buyout risks with a contingent claim approach based on the Ho and Singer (1982) model to price risky debt with an amortization payment. our paper also contributes to the discussion about the compensation of leverage risk by active monitoring. controlling for the systematic risk involved. our study focuses on US buyouts. This finding is important for the interpretation of results from literature. However.36%. This average return is slightly higher than the average return earned if the correct mimicking portfolio is indeed the S&P 500. as e. in a global financial market. but not from acquisitions or divestitures. Surprisingly. Jones and Rhodes-Kropf (2003) investigate the idiosyncratic risks of PE transactions. However. Cotter and Peck (2001) confirm that the controlling majority of buyout specialists motivates the target-company managers and decreases the financial risk of the transactions. while the median is 12.2100 A.1% per annum. 1993) focus on the high risks that buyout lenders take. Gottschalg / Journal of Banking & Finance 35 (2011) 2099–2110 The contributions of this paper are several. and this results in a sample selection bias that the author overcomes via a maximum likelihood estimate..07%. and controlling for the systematic risks carried by the sponsors.29% (2. Cochrane (2005) points out that empirical VC research usually only observes valuations if target companies go public. we discuss the rationale and model assumptions for benchmarking the buyout asset class with the public market. the second increases the borrowing cost for the equivalent public market investment. This is significantly 3.

Groh. have a beta equal to one. Similarly. we can only collect 152 transactions for which the following data is available. the financial sponsors owned 65%.A.49) at closing. they calculate unlevered and levered beta factors with a method similar to the one we will apply to perform a risk-adjustment. and 90% majorities. leverage risk. They focus on the correction of unrealized. Similarly.7 million). The minorities in a few transactions are. due to the redemption of debt. O. this underlines the variety of the risk patterns in buyout financing. Our data providers are among the world’s largest limited partners and collectively manage more than US$40 billion in the PE asset class.05. risk-adjusted mimicking of buyout investments with public market securities requires information on the business of the target companies. the average (median) is $313. This finally leaves us with these 133 transactions executed by 41 different funds. As no standard format exists for reporting transactions in PPM. Using precise information on the valuations of individual target companies. In the following. the necessary data to perform the proposed risk-adjustments could be extracted only for a small subset of transactions. as well as on the capital structure of the acquiring investment vehicles at least at entry and exit. and leverage cost. They find that gross of fees their sample funds outperform the S&P 500 by 3% p. From 122 PPM describing 2264 realized buyout transactions (1001 of which were in the US). in these two transactions. company valuation. the availability of public peers and adequate benchmark portfolios. . From the 152 target companies 133 are located in the United States.0 million). As we show in this paper. rather. for closing: the date.9 million ($135. with the remainder based in the United Kingdom.5 million to almost $9000 million. initial financial leverage. The equity stakes range from 8% to 100% ownership. respectively. it can only be gathered directly from institutions investing in buyouts as either general or limited partners. On average (median) the amount of equity invested is $53. it leads to potential selection biases. Phalippou and Gottschalg (2009) build on the Kaplan and Schoar (2005) article. Because our objective is to assess the cost of capital for ‘‘buyouts’’. Subsequent rounds usually have a lower risk than the initial payment given some debt redemption until the moment of the add-on payment.P. and 1.64) at exit. the valuations range from $0 (a write-off) to almost $13. company valuation. some of the buyouts are very highly levered. respectively. The average and the median holding period are below four years. Phalippou (2010) analyzes if the performance of individual funds is persistent. Gottschalg / Journal of Banking & Finance 35 (2011) 2099–2110 investments are as risky as the S&P 500 portfolio and hence. 31 in the early 1990s and 74 in the late 1990s. premature disbursements or large dividends financed by additional debt usually come at a higher risk level than the final exit. their competitors in their industry sector and on the capital structures of the investment vehicles at the closing date and at exit. continental Europe and Japan. In several transactions. The highest debt to equity ratio gets as high as 17. the documents are very heterogeneous in terms of the level 2101 of detail provided for each transaction – both within one fund and across general partners. the authors are unable to correct for different degrees of leverage in their sample transactions. we describe the data sources and sample characteristics of the data used in this study and discuss a potential bias. However. and subsequent redemption of debt. meaningful additional ‘‘add-on payments’’ occurred in subsequent financing rounds as well as premature disbursements. This data is neither publicly available. However. Rather. At exit. it constitutes the first large-scale analysis on the cost of capital for buyouts that fully corrects for operating risk. While at closing. equally risky (in terms of systematic risk) mimicking investments that allow determining buyout cost of capital and show that a thorough consideration of leverage risk and cost is of great importance. and consecutively decreases. Our dataset is compiled from information on buyout funds made available to us anonymously in Private Placement Memoranda – PPM. Our sample transactions show the characteristics as exhibited in Table 1: The first transaction was closed in November 1984 and the last was exited by June 2004: 16 transactions were closed in the 1980s. general partners describe their track record and their strategy in order to raise a new fund. it becomes possible for us to attribute a systematic risk measure to every transaction. amount paid for the equity.94 (2. Net of fees the performance is 3% below the benchmark. First and most importantly. This reflects the desire to secure majorities for the sponsors to control the buyout targets effectively. The authors conclude that average buyout fund returns are. We are able to control for these patterns by constructing well-defined.5 million ($88.4 million ($18. without data on the target companies’ leverage. There is. The average and median degree of financial leverage found in our sample emphasizes the need to control for the effect of leverage risk in buyout performance assessments. The holding periods range from 3 months to 15 years plus 1 month. This pattern typically starts at higher risks due to initially higher degrees of leverage. where the average (median) is 76% (86%). He finds no performance predictability for funds that are backed by sophisticated investors. As a result. For each transaction we are able to create the exact individual financial risk-profile considering operating risk. while the remaining were closed in the new millennium.2 acquired equity stake. On the other extreme. we only consider the PPM of funds that explicitly name themselves a ‘‘buyout fund’’. Second. but reported net asset values and run several analyses applying various approaches to partially or fully write them off. and we need to accurately track the risks of these cash flows. With respect to the degrees of financial leverage. While this approach has advantages regarding the depth and quality of available data. They calculate similar public market equivalents using the time-matching returns of the S&P 500 as benchmark. hence. Two transactions did not include any debt. Our study differs from and aims to extend prior work in several ways. The redemption capabilities of the target corporations vary to a great extent and. target-company industry and a short product and market description. As the non-US results would lack statistical weight for any individual country while also distorting the US results and raising questions about cross-currency returns.2 million in a small and syndicated transaction to almost $1150 million signaling the large exposure in certain transactions. nor is it recorded in any of the commonly used databases. smaller than those of the S&P 500.a. Gross of fees the returns marginally exceed the chosen benchmark. The final payoffs range be2 It is a common understanding among players in the buyout industry that the term ‘‘company valuation’’ comprises the value of a company’s equity plus the value of its net debt.28 (0.500 million with an average (median) of $547. the amount of equity invested ranges from $0. This risk measure is not an average over several transactions and not constant over time. First.0 million). the average (median) debt to equity ratio is 2. or description of competitors (in order to determine the SIC code). the values of the target companies range from $3. In a robustness check. Data collection and sample description The availability of data of sufficient breadth and depth has been one of the key constraints in addressing buyout cost of capital. for the exit: the date. 3. equity stake and amount returned to the buyout fund. syndicated equity stakes. in fact. the timing and size of underlying cash flows. an individual risk pattern for every single transaction. In these documents. after fees. we decided to exclude all non-US transactions.

If the buyout’s beta is lower than 1.75 1. 4. and hence. (2007). This seems high and points to a potential selection bias in our sample towards successful transactions: As detailed information about the entire population of buyout investments is unavailable. we can establish the mimicking investments as follows: invest for every buyout transaction the equal amount of equity in a market proxy-portfolio which is levered up with borrowed funds until it matches the equity beta factor of the buyout at closing. Therefore. we have to consider a possible selection bias arising from the general partners’ (GP) reporting policy.a..08 100% 100% 17. Publicly available datasets on buyouts (such as from Thomson Venture Economics) usually provide the timing and the amount of cash flows.90 135. The reported internal rates of return gross of fees range from 100% (several total write-offs) to as high as 472% p. we unfortunately cannot quantify the extent to which our sample is biased. GPs have an incentive to provide detailed information only for their successful transactions in the PPM. However. tracking the risk of the buyout.g. after serving debt.61 0. transaction structures. which is primarily a marketing instrument for fundraising. the treatment of add-on investments.70% 0. and Axelson et al. the risk of tax shields. and changes in ownership in detail in Appendix A. 5.99 870. Min Max Average Median Std. leverage ratios. the mean average IRR of all the transactions and the median are 50. In general. Muscarella and Vetsuypens (1990). Closing date Exit date Holding period (years) Equity stake at closing Equity stake at exit Initial debt/equity Exit debt/equity Company valuations at closing ($m) November 84 February 88 0. Groh. Given the source of our data. the position is liquidated annually.82 Equity investment ($m) 0.2 million and $1150 million.92 Note: The Table presents the characteristics of our 133 sample transactions.a.94 1.00% March 03 June 04 15. Subsequently. premature payoffs.39 18.4 million and a median of $63. Since the derivation of the mimicking investment is essential but likewise exhaustive.25 8% 8% 0 0 3. We repeat this recalibration procedure until the exit date. the ratios and dimensions of our sample transactions are very similar to those from other researchers. we can compare the leverage pattern of buyouts with that of their publicly quoted peers (see Table 1). (c) unlevering of these beta factors to derive their operating or unlevered betas.. there are good reasons for suspecting that there are too many successful buyouts in our sample.70% p. We address this issue in a subsequent section and detect that a bias towards successful buyouts has no impact on our general conclusions.a. such a bias towards successful transactions needs not to affect any finding on cost of capital. or preferred industry segments at the time. (e) the re-levering of these peer group operating betas on the level of the buyout transactions at closing. never write a PPM that reports on their track record.50 November 95 July 99 3.38 63. The risk of the public market transaction is then adjusted every year.08% 35. Gottschalg / Journal of Banking & Finance 35 (2011) 2099–2110 Table 1 Descriptive statistics of sample data. With this cash flow. 1993).28 313. we can easily calculate the annualized return on the initial investment.08% p.09 Almost* 9000 Almost* 13. e.70 115. such as De Angelo and De Angelo (1987). as discussed above. with holding periods ranging from 3 months to 15 years or equity exposure between $0. Analyses and results First.25 0.a.00 1366.20 53. the position is closed and. The approach follows commonly accepted models to control for systematic investment risk and enables the analyses described in the following section. which is borne by the lenders. Additionally.1 million. we receive a residual final payoff to the investor. and exit. interest is paid. our sample can be regarded as ‘‘typical’’ compared to other samples. debt is redeemed and the residual equity is levered up again with borrowed funds (respectively funds are lent) to the prevailing systematic risk of the buyout. (d) the determination of a market-weighted average of these operating betas for every peer group.64 88.75 76% 74% 2.52 July 96 December 99 3.17 Company valuations at exit ($m) 0 547. sourcing or exit channels.) 100. * Due to non-disclosure agreements with our data providers we are not allowed to provide the maximal values as the transactions could be identified then. At least. With respect to transaction size and the observed financial structures.79 Final payoff ($m) 0 160.00 2. and 35.08 86% 86% 2.P. With respect to lever- . since our data is kept anonymous. funds can be lent. First. dev.500 Almost* 1150 Almost* 1800 472. The buyout mimicking investments We create a portfolio of public market mimicking investments to assess the buyouts’ cost of capital. The unlevering and re-levering procedures also require the specification of the risk. O. and. it is impossible for us to trace our sample transactions in any one of these (and similar) databases. The descriptive statistics further reveal the diversity of the transactions made by the buyout funds.49 0. Kaplan and Stein (1990. However. as well as an applicable corporate tax rate. we have to expect a survivorship bias based on the mechanism that unsuccessful GPs might not to raise another fund. Second. we can only speculate whether our sample represents typical transaction sizes.10 299. These investments replicate the timing and the systematic risk-profile of the buyouts and require for each transaction: (a) the identification of a peer group of publicly traded companies with the assumed same operating risk. Therefore. tween $0 (a write-off) and almost $1800 million with an average of $160.05 IRR (p.00% 50. (b) the calculation of the equity betas for each of these ‘public peers’.27 2. but do not include much more economically valuable information other than that. respectively. which represents the time and risk matched cost of capital of the buyout. we describe the individual steps. Then.05 14. The timing of the mimicking investments corresponds to the closing date.2102 A. our data gathering process is not determined by any economic variable but only by the fact that the PPM must provide sufficient information on the transaction.

slope of the regression. This highly skewed distribution signals the need for adjustments when benchmarking buyout performance.88 1 4.80 1 1.67 is different from zero with a p-value of 0. We test the appropriateness of our different mimicking approaches and regress the actual returns of the buyout transactions (as given in the PPM) on the mimicking returns that we calculate for each transaction according to the assumptions of the various sensitivity analyses and robustness checks: ~rBO ¼ a0 þ a1~r Mimicking þ ~e ð1Þ where ~r BO .83). Since we do not aim to control for additional factors that affect buyout returns than systematic risk. and comment on its potential effect on our results. which are independent of systematic risk.4% is low.P. risk-free borrowing and lending is possible in this case.95 percentile). we take a look at the operating risk and find that the resulting unlevered beta factors range between 0. signaling the need to add several other regressors to improve the assessment of buyout returns in further research.05 pct.95 percentile) at closing.40 1 1.195. Smith (1990). These assumptions are stressed in the subsequent sensitivity analyses and robustness checks..07%. Mean Median 0–2 3 4 Base cases Kaplan/Schoar (2005) Phalippou/Gottschalg (2009) 0.95 percentile). ~rMimicking .73 Note: This table presents the most important descriptive statistics of the equity betas at closing and at exit in our base scenario(s) and in different sensitivity analyses.49. We thank Ludovic Phalippou for raising this argument.3 Third. we relax this assumption and observe the consequences if lending is possible at the same credit spread. O. Hence. a corporate tax rate of 35%.4 Table 3 presents the descriptive statistics of the returns of the different mimicking approaches and the hypotheses tests for their appropriateness.05 percentile) and 2.82 0. assuming for the unlevering/ levering process initial debt/equity ratios of 3 and final debt/equity ratios at average industry levels. the regression slope 0. As a result.70 1.25 for the buyouts and 0 for the public market.88 (0. and no differentiation between the risks of debt tax shields in the quoted and unquoted market segment.28 (mean average). The next sensitivity analysis replicates Kaplan and Schoar (2005).2103 A. or De Angelo and De Angelo (1987).05 percentile) and 1. we consider the proposed model as adequate for benchmarking buyout systematic risk patterns. At exit. Jensen (1989a).95 pct.05 percentile) and 3. However. The standard deviation of the returns remains almost unchanged 4 We are grateful to Andrew Metrick and Luigi Zingales for suggesting this procedure. Some of the low asset betas could also result from the selection of infrequently traded peers.01 1 0.40 and a median of 0. the mimicking investments time-match the S&P 500 portfolio. a0.64 (median) respectively. When exited. lending is possible at the risk-free rate. white noise error term. and therefore. age risk at closing.94 1 1.80 (0. Groh.83.g.94. we find that the average debt/equity ratio of the buyout investments is 2. the target companies have even lower leverage ratios than their public peers.32 (0. 0. the only available collateral for every individual mimicking investment is the portfolio of the S&P 500 shares which have to be bought on the margin.30% (13.32 1 0. In the first sensitivity analysis (#1).50% and 38.40 (0. and 38. do not focus on a0. However. the resulting systematic risk of the transactions ranges between 0.95 pct. The mean average of the unlevered beta factors is 0. Each benchmark model that appropriately matches the systematic risk patterns of buyouts should have a slope a1 equal to one.92%. This is not surprising. That means that on average our sample transactions are initially levered more than twice as much as their public peers. The whole distribution of the mimicking returns shifts towards positive results as would be expected. Ninety percent of the returns are between 29.56.71 1 0. Gottschalg / Journal of Banking & Finance 35 (2011) 2099–2110 Table 2 Equity betas for the baseline case(s) and two more scenarios. even if we attempted to exclude these peers from our selection.e. we assume that lending is possible at the risk-free rate and borrowing at the cost of high yield debt to adequately replicate buyout non-recourse financing. while the standard deviation is 25. At exit those ratios are 1.25%. and one of its important features is the replication of non-recourse financing structured in buyout transactions. As argued above. the mean average leverage ratio of all quoted peers over the five years is 1.94 and their median is 2. Hence.32 2.05 and the 0. 3 See e. Subsequently. Mean Median 0. the equity betas vary between 0.01 and a median of 0. . The 0.95 percentile are 30. ~e.. we abandon the establishment of non-recourse financing. This leads to an increase of the mean (median) return of the mimicking investments to 10. as buyout fund managers typically choose low volatile businesses for their investments and. levering up the investments comes at the cost of a credit spread. We control for this issue in our sensitivity analyses. and we refer to Altman and Pasternack (2006) for the relevant spreads.21 (0. As we explain in detail in Appendix A. risk-free borrowing and lending. the important hypotheses for us to test are H0 ? a1 = 1. 0. we do not include any other regressors. The intercept 0 will reflect the assumed sample selection bias towards successful transactions and other potentially constant return determining factors. The last sensitivity analysis replicates a robustness check by Phalippou and Gottschalg (2009). we cannot propose a correction method for the sample selection bias. We further calculate an average (median) beta factor over the holding periods of all transactions of 1.32 (0.16%. and not significantly different from 1 with a p-value of 0.32 1 0.32 (0. Closing Exit # Scenario 0.05 pct.06%. we apply equal creditspreads for lending and borrowing.25% respectively. returns of the mimicking investments (our own calculations according to the assumptions of the various sensitivity analyses and robustness checks).50 0. i. In our first sensitivity analysis. hence. and the median is 12.71. In our base case. with a mean of 1. we should expect their asset betas to reflect this. a debt beta of 0. As a consequence. while the median is 0. Fourth. a1. Our base case yields a mean average return of the mimicking investments of 9.38.32 3.009.67 and their median is 0. but this is not surprising at the given high variance in the data and due to controlling for only one factor. and 0. and allow both. with a mean of 1.03 0. Second. we build the portfolio of mimicking investments. The beta factors resulting from our base case and the proposed sensitivity analyses are presented in Table 2. the construction of the adequate set of mimicking investments depends on several assumptions. We start with our ‘‘base case’’ approach. In comparison. The coefficient of determination will remain small for all of the subsequent regressions. H1 ? a1 – 1 for each of the proposed benchmark models. The explained variance of the model of 3. Additionally.36 1. intercept of the regression. returns of the buyout transactions (as provided in the PPM).91% respectively).

68 3.44 0.01 0.P.39 2.92 1.025) 0.57 0.92 1. the buyout leverage ratios do not change over time. on average. a debt beta of 0. They also do not control for different cost of borrowing and lending. The regression confirms that this mimicking approach is also appropriate.60%.19 1.03.23 0. If we apply their approach the betas range between 0.95 percentile].70 1. Kaplan and Schoar (2005) suggest time-matching the buyout investments with S&P 500 exposure.05 percentile at 22.65 0. assume perfect market conditions. For the first three scenarios.95) [13. 0.88 to 31. and set all beta factors equal to 1. buyouts usually happen in less risky industries.011) 0.034 0.287 4 Phalippou/Gottschalg (2009) 12. Since all beta factors are equal to one in this case. 0. we allow for risk-free borrowing and lending.72 0.195 1 Base case.93 0.31 1. it does also not need any determination of the cost of/return on debt to establish the public market equivalent.95 2. Table 4 Equity betas for the robustness checks. leverage cost. and by the cost of/returns on debt to establish the mimicking transactions. they assume a corporate tax rate of 35%. but borrowing and lending at risk-free rate 12.65 (0.16) [30.055 0. These results.46 (0.029 0.95 pct. Groh.99%.99] 14. the approach needs no assumption about the capital structure of the individual buyouts.60) [26.70 1.65 3.46] 22.70 (0.25 for the buyouts and 0 for the public market and do not differentiate the risks of debt tax shields in the quoted and unquoted market segment.92] 25.91) [29. # Scenario Mean (Median) return of the mimicking investments (%) [0.89 (12.59%) and the return distribution shifts further towards positive returns with the 0.59) [22. standard deviation (%) Regression slope (two-tailed p-value) R2 Two-tailed p-value of test if regression slope is equal to 1 0 Base case 9.015) 0.60 to 38.001) 0.36% (14.46 4. and this correspondingly narrows 90% of the return observations between 13. In their unlevering and re-levering approach.95 percentile) at closing with a mean average of 1. Mean Median 1 2 3 4 5 6 7 Industry mix Leverage only Increased operating risk Risk-free debt Reduced operating risk Increased risk of debt Outliers dropped 0. The standard deviation of returns decreases to 14.70 Note: This table presents the most important descriptive statistics of the equity beta factors at closing and at exit of our robustness checks.50 and a median of 1.32 0.032 0. On average. the systematic risk of the Kaplan and Schoar (2005) mimicking portfolio is lower than the systematic risk in our base case.53 0.22 2. The scenarios confirm that buyout cost of capital is largely determined by the assumptions about the risk sharing between sponsors and lenders.32 (0. They calculate equity beta factors similar to Ljungqvist and Richardson (2003) with initial debt/equity ratios of 3 and final debt/equity ratios at average industry levels. Closing Exit # Robustness check 0. and deleverage .30 (13. buyout companies correspond to the average of the companies that comprise the S&P 500 index with respect to their business risk.71 (0.95 percentile is 26.87 0.09 1. their leverage. Gottschalg / Journal of Banking & Finance 35 (2011) 2099–2110 Table 3 Descriptive statistics of the mimicking portfolio returns and regression results of the scenarios.55 (12.40 1.50 1. or more precisely.83 1.91 0.05 pct.025 0. on average.31 1.322 Note: This table presents the descriptive statistics of the annualized rates of return of the mimicking investments.36 (14. If these preconditions are accepted.43 0. They decrease until exit to a range between 0.70 1.32 (0.50 to 38.87 6.35 3. We find that equity betas are slightly larger at closing and slightly smaller at exit compared to our base case approach.55% and a median of 12.95 percentile) with a mean of 0.31 to 43.95 percentile at 45.009) 0. follow the pattern of the S&P 500 companies.23%. The regression confirms that matching the buyouts with the S&P 500 directly is appropriate. In the next sensitivity analysis (#2).78 1.35 0.31% to 43.05 pct.05–0.16 5. the approach implies that. and the regression analysis signals appropriateness of this approach. O.89 0. there is no borrowing and lending involved.90 3.78 2.05–0. we replicate Kaplan and Schoars’ (2005) findings (scenario #3).82 and a median of 0.45 0.182 3 Kaplan/Schoar (2005) 12.89% (12. The 0.63 (0.25 0. Assuming risk-free borrowing and lending to adjust the mimicking investments leads to a mean return of 12. The standard deviation of returns slightly reduces to 23.2104 A.173 2 Base case. at 25. we only change assumptions that affect the mimicking portfolio.07 (12. The mean (median) mimicking return increases to 12.24 0. We achieve a mean (median) 12.67 (0.46%. Next.08 0.11 2.47 0. However. hence.25] 25. Further. However.75%.75 1. and also the standard deviation of the returns of 22.73.88% and 31.13 1.45 0.36 (0. the beta factors to match in the public market do not change because we do not change any assumption regarding the unlevering/re-levering process.35 0.23 0. the approach implies that. but Borrowing and Lending at Altman Spreads 10.66 0. and differ with regard to leverage ratios.32 0.07 0.95 pct.32 0.75%.05 percentile) and 1. This is in strong contrast to the betas in our baseline cases that vary substantially over time and across transactions.93 to 45.75] 23. and hence.46 2.24 0. Hence.69 0.12 0.93% and the 0. regression results and the test whether the mimicking model is applicable for our base case analysis and four different scenarios. Mean Median 0.95%) return of the mimicking transactions. we replicate one of the robustness checks from Phalippou and Gottschalg (2009) and use their approach for the unlevering/levering process.05 percentile) and 4. are close to scenario #2.99 0.08%. and their cost of debt. In the final scenario (#4).88 1.36 3. time-matching S&P 500 investments represent the correct cost of capital for buyout transactions.23% and the regression results signal applicability of that approach.

The regression signals appropriateness of the approach. but not for the impact of different operating risks in the chosen industries. As several studies have found. As expected. Their mean average is 0. we claim that it is. (2004). Hence. Hence. The resulting beta factors in the robustness checks.11 and a median of 1. or the applicable tax rate in our unlevering/ re-levering approach (as described in Appendix A). the equivalent approach is to purchase stocks on the margin. As a result.21%. not appropriate to use the unadjusted index portfolio as a benchmark.70. we introduce risky debt to lever-up the investments into the S&P 500 index.45 (0. we have to specify the cost and the risk of debt. means of 1978–2002 data column. with a mean of 2.05–0. 1977.59%) which is.05 percentile) to 5.13.84 (which is the unlevered beta factor of the S&P 500 Index). which is scenario #2.30% (median 13.95 percentile) with a mean of 1. In practice. Here. However. The standard deviation of returns becomes as high as 26. on average. it is in fact doubtful that banks allow borrowed share purchases up to the risk-levels that we detect at closing of some buyout transactions. the mean (median) return of the mimicking investments increases to 15. of debt tax shields.97%). At exit they are between 0.16 (0. In other words. the mimicking returns.55% (median 12. This decrease is not surprising. Pogue and Solnik. and this is mirrored in a narrowed 0.95 percentile. Second.A. we claim that corrections are necessary to control for these differences and propose these corrections in Appendix A.05 percentile) and 1. which is 11. The approach leads to equity betas between 0. as described in Appendix A.46 (0. The S&P 500 portfolio must be purchased on the margin. one could argue that our approach inherently leads to a lower risk boundary for the buyout transactions. This results in a mean (median) return of the mimicking portfolio of 10. and additionally. than in our baseline case.P. .1. This robustness check underlines an important implication: the average of the peer group betas is smaller than one.and time-matched investment in a hypothetically unlevered public market index. Consequently. As a result. This underlines the necessity to choose an appropriate mimicking approach that controls for this issue. Table 1.86%.83% to 56. Groh.95 percentile) with a mean average of 2. and observe a strong and significant (twotailed p-value = 0.35 (0. 1977. we examine the impact of leverage alone on returns.65 (0. This highlights once more the proposition that buyout fund managers search within low-risk industries. the target corporations borrow on their own assets.60%) which is very close to the result of scenario #2.5 We then lever-up each mimicking investment with the actual leverage of the corresponding buyout. the risk and the cost of debt will rise.53% and the 0.93 (0. the business risks of the target companies could be downward biased. The third robustness check (#3) controls for the sensitivity of our results with respect to the calculation of the operating betas for the peer group companies. the buyouts are benchmarked with their public peers directly.07% (median 12. we control for our assumptions regarding the risk-adjustment process. as the returns of the index portfolio are eaten up by the cost of debt.05 percentile) to 4. proper adjustments are not required.40 and a median of 1.72% (19. we alter the assumptions for the unlevering/re-levering approach that determine the beta factors which have to be tracked in the public market. panel C. At exit. Thus the betas are larger. even if the large dispersion of returns suggests that every single buyout should be benchmarked with ‘‘its specific’’ cost of capital.92. which determines unlevered beta factors for the Fama and French (1997) industry classification. first. The standard deviation of re- 5 See Bernado et al. signaling that buyout funds search for targets in less risky industries.013) decrease of the average return of the mimicking portfolio to 9. The described procedure leads to a set of levered and unlevered purchases of the S&P 500 index portfolio.89% (median 12. Using this approach.36% (median 14. Along the same lines. One could conclude that the mentioned differences with respect to operating risks.95 percentile is 24. This leads to partial risk-adjustment as such a 2105 mimicking portfolio replicates the industry mix of our buyouts but fails to capture the effect of (additional or less) leverage. Schwert.87 (0. The resulting betas range at closing from 0. analyze the impact of outliers in robustness checks. Another reason to perform this check is that we might have miss-specified the risk of debt. an appropriate replication of buyouts with public market securities should consider the limited liability of the transactions that come at some cost: Buyout sponsors do not borrow on the buyout fund’s other assets. If we assume that the non-recourse financing of buyouts is a valuable and important characteristic that also needs to be replicated.95 percentile) that remain constant over the holding period. O. In a next step.50 and a median of 1. the betas are lower than both the market beta and that of our baseline case. Robustness checks For all of our robustness checks.60%). infrequently traded assets do not closely follow market movements (Fisher. The first robustness check (#1) controls for the industry mix.90 (0.12% (11. we assume perfect market conditions with riskless borrowing and lending when setting up the mimicking portfolio to facilitate comparison. Adopting the approach of one of the robustness checks of Phalippou and Gottschalg (2009) leads to 12.09. and the median is 0.05–0.95% to 27. Hence.27%). We apply the average equity beta factors of our peer groups directly to the mimicking investments without considering different degrees of leverage. As a consequence. net out on average. we achieve an average cost of capital of 12. 1974. the average return of the mimicking transactions becomes 10. We set all the investments of the mimicking portfolio to have an unlevered beta of 0. as we use comparables transferred from the public market to the unquoted segment. only slightly different from the mean average 12. we increase the operating risk of each of the investments in the mimicking portfolio by an arbitrary factor that corrects for a suspected 25% understatement of the operating betas in our calculations.05 percentile) to 3. Dimson.05 percentile) and 2. in fact. Hence. However. Buyout transactions often take place in niche markets in which shares might be traded infrequently. Third.95 percentile) with a mean of 1. This scenario adjusts for differences in leverage risk. 1966.70% (16. the resulting equity betas increase (in reference to our baseline case) within the range of 0. and the regression analyses are presented in Tables 4 and 5. Gottschalg / Journal of Banking & Finance 35 (2011) 2099–2110 patterns from the ‘‘average’’ S&P 500 company. and the S&P 500 may serve as a benchmark for buyouts. the mimicking portfolio yields a mean (median) return of 17. If we also allow for lending at the same rates. leverage risks and cost. 5. In the second robustness check (#2).70. we assume in the simplest case that borrowing and lending is possible at the risk-free rate when setting up the mimicking investments.78. Corresponding with the higher beta factors. Scholes and Williams. The regression proves the applicability of that approach. 1979).95 percentile) at closing. For the establishment of the mimicking portfolio. (2004). we draw on data provided by Bernado et al.22 and a median of 1.95%) return in the Kaplan and Schoar (2005) case.43 (0. This leads to a comparison of the buyouts with a leverage.91%). results change if we include a credit spread for the establishment of the mimicking transactions as we do in our base line case and scenario #1.53% is the lowest of all of our different analyses. The standard deviation of 12.16%). they range from 0. adjusting for operating and leverage risks requires numerous specifications. Thus.87 (0. As a result.

Hence.68%. Palepu (1990).95 (0. we use a debt beta of 0.25%. we instead replicate our calculations using risk-free debt. where debt layers are often provided against little or no collateral. with debt redemption abilities (q = 0.95 interval ranges from 22. The regression slope of 0.95 percentile). As more risk is transferred to the lenders.93 and hence close to one. At exit. The mimicking portfolio yields a mean (median) return of 12. and an underestimation of the corporate tax rate while initially calculating the operating beta factors.004.90%. Despite the papers mentioned above.a.68 (0. At exit. as our sample is biased towards higher IRRs it might likewise be biased towards highly levered transactions. However. this also leads to a lower truncation level for the risk of debt and hence lower levered mimicking investments (see Appendix A).95 percentile).05 percentile) to 3. Lichtenberg and Siegel (1990) as a particular characteristic of buyout transactions. One can conclude that if buyout sponsors are not able to transfer risks to the lenders. or long holding periods. and this yields to an overestimation of their cost of capital.53 and a median of 1. the regression slope of 0.13% and 72. we exclude transactions from our sample with a debt/equity ratio below 0. we control for the impact of the risk of debt. Even with full data regarding the transactions. This reduction of operating beta factors could also be justified by an overestimation of the risk of investment-grade debt. In the final robustness check we exclude those transactions from our sample that could be considered ‘‘atypical’’ or outliers.01%. Sample selection bias As mentioned above. The next robustness check (#5) directly follows Kaplan and Stein (1990).2106 A. The standard deviation of returns is 21. Groh. Unfortunately.19 and a median of 0.31 (0.89 (0.95 percentile).25 and where the institutional investors’ equity ownership is below 50%. Hence.32 (0.41 in our base case analysis to lever-up the buyout betas. When no risk can be transferred to the lenders. improved governance by the active investor is also debated in Jensen (1989a. Due to the resulting low beta factors the mean (median) return of the mimicking investments is only 10. we search for correlations between the IRRs achieved in the buyout transactions.37%.95 percentile to 40.36 (0. 1993).70 is not significantly different from one. In our sixth robustness check. we analyze the impact if lenders take on an even higher proportion of risk than that assumed in our baseline case. the mean (median) return of the mimicking investments rises to peaking 17.01%) which is slightly lower than in our comparable baseline case #2. and 90% of the returns are between 42. The standard deviation is also low at 14. dropping ‘‘outliers’’ from our sample of transactions does not meaningfully change the results. the mean (median) return of the mimicking investments decreases (compared to our base case analysis) to 11.69 (0. They range at closing from 0. This confirms the applicability of this approach. This bias can affect our finding on their cost of capital if actually achieved returns are correlated with any parameters determining the mimicking investments. with a mean (median) of only 0.50 to lever-up the mimicking investments.47% to 41. the appropriate cost of capital becomes very large and dispersed. The standard deviation is 35. corporations with strong free cash flows. these improvements should be accounted for in buyout performance analyses.07% and 31.34%. some are characterized by relatively low or zero leverage ratios.335). Thus. and all determinants necessary to set up the mimicking portfolio. the betas range from 0.07 respectively.86% (11. O. (2005). the regression results suggest that assuming risk-free buyout debt does not appropriately match the systematic buyout risks: The slope of 0. At exit they range from 0.66 (0.05 percentile) and 3.211). we arbitrarily increase our debt beta to 0.99.95 percentile) at closing.05 percentile) to 3. the beta factors for this scenario shrink to between 0. This leads us to conclude that the cost of capital of buyout transactions becomes smaller if we assume that GPs are able to structure buyout transactions transferring a substantial part of the transaction risk to the lenders. The standard deviation of returns is 22.47 (0. This results in a reduced sample of 108 transactions.05 percentile) to 3.96% to 45.24 (0.01 (0. The regression slope of 0. risky peers.72. However. These peaks are mirrored by the dispersion of returns. with the peer group’s average beta factor (q = 0. Hence. The equity betas then range between 0. Additionally.57 and a median of 1.12 (0.11% (10.95 percentile) with a mean (median) of 1.51%. In fact. The resulting equity betas range from 0.44). with a mean of 1.21%. However. Accordingly. They decrease to a range between 0. Finally. Kaplan and Stein (1990. We conclude that our sample might be biased towards highly levered transactions.70).95 percentile). Hence.39 (0.65%).35%.05–0. or either very short.556). as increased leverage risk is directly compensated by active monitoring and majority ownership.05 percentile) to 2.91. with a mean of 1. Hence. we also verify if these parameters drive the returns of the mimicking portfolio in parallel. Smith (1990). and short holding periods.06% and the 0.240) positively affects buyout cost of capital. while 90% of the return distribution is between 16.95 interval is from 21.32 and a median of 0.53 is significantly different from one at the 0. and negatively with the holding period (q = 0.27% (12.95 percentile).41% to 67. our sample of buyouts might be biased towards successful transactions. they likewise decrease their cost of capital. Gottschalg / Journal of Banking & Finance 35 (2011) 2099–2110 turns becomes as high as 35. where we also assume borrowing and lending at the risk-free rate. the assumption that lenders in buyout transactions do not take risks is rather hypothetical. Muscarella and Vetsuypens .05 percentile) and 2. this is a common observation for buyout transactions. As explained in detail in Appendix A. 1991). In this check. with a mean of 0. However. with a mean (median) at 0. We find that only the initial degree of leverage (q = 0.46 is significantly different from 1 at a p-value of 0. which expands the 0. and therefore. the equity betas for our mimicking transactions increase substantially. our observations are in line with the samples used by other researchers. with a mean of 1. this seems not to be an appropriate benchmarking approach.35 (0.204).05–0.P.05–0.31 (0.67 is not significantly different from one.57% (18. all the leverage risk is borne by the equity sponsors. and we assume that operating improvements have lowered the operating risk of the buyout targets.05 percentile) to 1. Hence. In the following robustness check (#4). we cannot correct for this bias because initial leverage ratios are unknown for the population of buyouts. minority ownership of the active investor.24 (0. Accordingly.45) at closing. 5. and turns to its discussion: If buyout sponsors are indeed able to lower the operating risks of their portfolio companies by enhanced governance or restructuring activities.2. A reduction of operating risks might be caused by improved governance by the active investors as highlighted in Kaplan (1989a).009 level. Kaplan (1989b.b). This assumption is also reasonable as high yield bonds and mezzanine financing are often extensively used to structure buyout transactions. while the 0. We find four significant correlations at the two-tailed 0.65.07%). we discard the most successful buyouts where the achieved internal rates of return are above 200% p. the betas range from 0. Therefore.92 and a median of 0.83 (0.95 percentile) at closing. However. we (arbitrarily) reduce all the operating beta factors by 25%. It is also consistent with prior research that found even higher debt betas for buyouts (such as Kaplan and Stein (1990)). Cotter and Peck (2001). The regression slope is 0.32 (0.05 percentile) to 6. such as De Angelo and De Angelo (1987). or Cumming et al.13%). We further exclude transactions with holding periods below a year and above nine years.88 (0.05 significance level: Buyout IRRs correlate positively with the degree of leverage at closing (q = 0. with a mean of 2.

004) 0.35] 35.90 0. our calculations might overstate buyout cost of capital.13) [42. limited partners themselves do not borrow to lever-up their transactions.21] 21. standard deviation (%) Regression slope (two tailed p-value) R2 Two tailed p-value of test if regression slope is equal to 1 1 Industry mix 10.05–0. it becomes possible to attribute operating and leverage risk measures to each individual transaction.186 Note: This table shows the descriptive statistics of the annualized rates of return of the mimicking investments. Third. Groh.72 (19.041 0.002) 0.97) [11. whether operating risks might decrease due to improved governance by active investors. our study exploits the detailed information available on a large sample of individual buyouts to illustrate the importance of an appropriate risk-adjustment for the assessment of their performance. and what credit-spreads have to be paid when levering up the equivalent public market investment.036) 0. As a result.67 (0.86] 12. Second. as well as with respect to credit-spreads and operational risks.53 0.01) [21.95 to 27. Hence.06 0. Additionally.12 (11. Contrarily.P. we show that using the returns of the S&P 500 as cost of capital for buyout performance analyses is misleading because the risk pattern in buyout transactions differs substantially from that of the public market benchmark index.53 (0. the average return of the mimicking investments is also upward biased. regression results and the test whether the mimicking model is applicable for the robustness checks. a comprehensive discussion of the inherent risks caused by the operations of the target firms and the applied leverage ratios is necessary. Thus.022 0.93 (0. With our rich data set. Our paper shows that these effects significantly impact buyout cost of capital.29% (2.029) 0. if we include too many highly levered transactions in our analyses. A potential bias in our sample towards highly levered transactions does not affect this result because high leverage ratios influence the mimicking returns positively. again comparing our approach and the time-matching S&P 500 returns. This is directly relevant for interpreting findings from prior research where this kind of risk separation has not been possible.57 (18. Our study builds on and extends existing work on the comparison of public and private equity performance in several respects: First. O.57 (0. we contribute to the discussion of how to determine cost of capital for buyout transactions. If they use financial leverage where favorable and if they can transfer an important portion of the leverage risks to the lenders.47 to 41.25] 35.34 0.41 to 67.21] 26. active involvement and by transferring risks to lenders.002) 0.2107 A.77 (0.01 0. # Robustness check Mean (median) return of the mimicking investments (%) [0. The discussion includes the question of how much risk is taken by lenders in these transactions.07 to 31.86 (11. lenders take a substantial part of the risks in buyout transactions. they are levered up at closing to an extent that exceeds the risk of the market proxy. For a set of 133 US buyouts completed between 1984 and 2004. Much research on buyouts has focused on the ability of general partners to lower their investment risk by monitoring.13 to 72. In any case. buyout transactions happen in less than average risky industries.27 (12. our study provides detailed insights into the importance of different assumptions regarding the risk-profile of debt.68 0.37] 22. Therefore. Conclusion In this paper.285 7 Outliers dropped 12.60) [24. 6.96 to 45. controlling for the systematic risk involved.876 6 Increased risk of debt 11.83 to 56. However.11 (10.046 0.36% points below the S&P 500 benchmark. The analyses confirm the conjecture that buyout investors choose industries with low operating risk. their competitors.70 (0.049 0.004 4 Risk-free debt 17. we discuss the rationale and model framework for benchmarking the buyout asset class with the public market. then the detected differences among the return distributions of our scenarios and robustness checks. The sensitivity analyses highlight the importance of a comprehensive risk-adjustment that considers both operating and leverage risks for an accurate assessment of buyout performance.07) [16. Using precise information on the valuations of individual target companies.262 2 Leverage only 17. The risk-adjustment process follows commonly accepted asset pricing principles. the mean buyout cost of capital becomes close to the average of the time-matching S&P 500 returns. Therefore. over the holding period this risk is reduced by amortization payments.65) [22. The fund’s liability is limited to its equity exposure. and debt tax shields. we claim that for buyout performance analyses.70 (16. We show that taking into account these real constraints leads to average (median) cost of capital which is 3.27) [40. As a result. researchers should thoroughly specify their presumptions about the discussed determinants when assessing buyout performance. An approach like ours had not previously been used due to the non-availability of required data for the determination of the buyouts’ systematic risks. and Axelson et al. . the median cost of capital is still 2. However.056 0.79%) below the mean average (median) of the time-matching S&P 500 returns.009 5 Reduced operating risk 10.95 percentile]. our calculations of the average cost of capital for buyout transactions indicate upper limits and unbiased cost might even be lower. The standard deviation of the returns is much higher. we can comprehensively control for the systematic transaction risks in constructing a well-defined mimicking portfolio with an equal risk-profile. (1990). as well as their industry sector and on the capital structures of the investment vehicles at the closing date and at exit.031 0.327 3 Increased operating risk 15.51] 14. we propose a portfolio of levered mimicking investments in the S&P 500 Index that match the buyouts with respect to the timing of their cash flows and their systematic risks.46 (0.002) 0. Only if we assume that borrowing and lending is possible at the risk-free rate when setting up the public market investments. (2007). and the required debt is borrowed by the target corporations.53 1. Usually. Gottschalg / Journal of Banking & Finance 35 (2011) 2099–2110 Table 5 Mean IRRs of the mimicking portfolios and regression results of robustness checks.012) 0.024 0. and with respect to the commonly used benchmark S&P 500 become large.

. funds are lent at the risk-free rate. if the mimicking investment includes unlevering the index portfolio. and Luigi Zingales. LBOs. Gottschalg / Journal of Banking & Finance 35 (2011) 2099–2110 Acknowledgements We are grateful to Andrew Metrick (NBER conference discussant). and the participants of the Conference on Private Equity. D. in order to be meaningful. We net total debt of each period (which includes shortand long-term interest bearing debt) by cash positions and divide it by the year-end market capitalizations (of straight and preferred equity). Next. we stress these assumptions to investigate the effect of credit-spreads. A. which is the cost of debt rd. (2010). We use the total return calculations for the S&P 500 Index. The risk of not being able to fully exploit the advantage from debt finance is then as high as the risk of obtaining the income itself and hence. Donald Siegel. The approach is based on several cited seminal papers from corporate finance and asset pricing literature. For a few buyouts. Tim Jenkinson. a peer group has to consist of at least three companies. Mandelker and Rhee (1984) present how operating company risk is borne by equity investors and risky debt providers according to the applied leverage ratio: bu ¼ be þ bd ð1  sÞ DE 1 þ ð1  sÞ DE ð2Þ where bd. Mathew RhodesKropf. However. This assumption merits discussion in general. the appropriate discount rate for the tax benefits equals the unlevered cost of capital.1. Groh. during buyout transactions. and stress. Arzac (2005) or Koller et al. efforts are often made to reduce operating risks. However. marginal tax rate. we determine the companies’ leverage ratios during that time from balance sheet and market data obtained by DataStream. systematic risk borne by equity investors (levered equity beta). the participants of the LBS Private Equity Symposium. Framework For the theoretical background for our mimicking strategies we refer to Modigliani and Miller (1958). In these cases. O. assuming that every company is exposed to some unavoidable and constant economic risk by its business.6 but especially with respect to buyouts.e. we calculate the (levered) beta factors of every single peer. which accurately reflects the fact that. Annette Vissing-Jørgensen. Ball and Brown (1967). In our base case. systematic operating risk (unlevered beta). We would also like to thank Douglas Cumming. we assume that lenders take risk and receive appropriate compensation. Especially. Some of the transactions were simply made in the same business segment. provided by DataStream as the performance benchmark. Joseph Zechner. and Governance. However. This leads to a final sample of 1207 peers to be incorporated into our analysis.g. we cannot correct for any kind of risk class transition. and follow Graham (2000)’s suggestion to use 35% as tax rate. we narrow the search by including an appropriate company size in terms of market capitalization and eliminate those companies from the peer group that are out of the range of 50–200% of the equity value of the target. our sample transactions. we determine the arithmetic average over the periods. Levering up the individual transactions Formula (2) reflects the assumption that uncertainty regarding the company’s ability to gain the tax benefits from debt financing is best measured by the rate at which its creditors lend the money. we require the peer companies’ shares to be traded frequently. and vary the model prerequisites. Ruback (2002) proposes the following relation: . Mike Wright. As additional filter. The formula requires the specification a debt beta factor bd and the marginal tax rate. sponsored by Caisse d’Epargne Rhone Alpes. Steve Kaplan. We describe our assessment of the debt beta subsequently. we use a beta transformation formula to derive the unlevered beta factor for each company. Rebecca Zarutskie (WFA conference discussant). As long as leverage ratios are moderate. We discuss the framework to apply the existing models on buyout transactions. companies might be unable to realize the full tax benefits. We construct an unlevered/levered investment in an index portfolio to track the individual transactions. bu. the assumptions regarding the risks taken by lenders and different rates for borrowing and lending largely determine the returns of the mimicking investments. Therefore. be. We aim to be as precise as possible assigning peer groups to our 133 sample companies and identify their 116 different industry sectors.P. Finally. Morten Sörensen. systematic risk borne by debt providers (debt beta). e. less volatile) business strategies. the principles are also described in Damodaran (2002). by focusing on safer (i. E. using the S&P 500 Index as a benchmark and weekly returns from January 1999 to December 2003. if leverage ratios increase. and the participants of the NBER New World of Private Equity conference Ulf Axelson. we determine the market capitalization weighted average of the unlevered beta factors of all the companies of a particular peer group and refer to this as our measure for the systematic operating risk of the target companies. Gonedes (1969).g. A. we do not want to rely on broad industry definitions to classify 6 E. and the participants of the Gutman Center Symposium on Real Assets and Portfolio Management for their support and helpful comments to our paper. Engelbert Dockner. Unlevering the peer groups’ business class risks Since buyout transactions often occur in niche markets. A peer group is defined by an equal four-digit SIC code and by company headquarters in the United States. Ludovic Phalippou. hence to sufficiently track the stock market movements.. Constructing the mimicking portfolio We take the perspective of a well-diversified investor not exposed to idiosyncratic risks of the particular buyout transactions.. but provide two robustness checks where we control for a potential reduction and also for an increase of the business risk during the buyout holding period. Alexander Groh appreciates the support by the Chair ‘‘Financial Practice of SMEs’’ at EMLYON Business School. market value of equity (common and preferred). There. We decided that.1. dividends are not usually paid but free cash flows are used for debt redemption. and also change the risk borne by lenders. We measure the business class risks for our transactions by a market-weighted average of the unlevered beta factors of their peers.2108 A. Jeremy Stein. Josh Lerner. and Sharpe and Cooper (1972). To unlever these betas. s. For these industry sectors we determine peer groups of quoted companies.1. subordinated and mezzanine debt). We attribute a risk class to every target-company defined by the operating risk of its public peers. market value of debt (all tax-deductible sources of capital such as senior. In a final step. Per Strömberg. Appendix A. A. In the course of robustness checks. Antoinette Schoar.2. we find more than 20 peer group members. this is the appropriate method. This index assumes that dividends are reinvested. In this case.1. Contrarily.

The remaining equity following the disbursement is retained in the mimicking portfolio. This leads us to consider that the leverage ratios are finally unaffected by the add-on investments in ‘‘write-off’’ companies. we differentiate again the exited and the ‘‘write-off’’ transactions. debt is redeemed and the residual equity is invested in the S&P 500 Index portfolio being levered to the prevailing systematic risk. However. using the S&P 500 as benchmark over a 2-year horizon. During the holding period the leverage ratio might decrease. Deriving debt betas We need to specify the systematic risk of debt in order to be able to lever and unlever the systematic equity risk according to Formulas (2) and (3). We calculate each funds beta factor. and at closing. 7 We thank Steve Kaplan for this suggestion. The exited transactions are closed and divested at certain leverage ratios. The systematic risk of the mimicking strategy is adjusted each year until exit. A.P. providing us with the company valuation. Add-on investments are usually made to prevent the debt providers from claiming bankruptcy. respectively. A. 23 in Altman and Pasternack (2006). interest on debt is paid. and the corresponding cost to determine our model’s sensitivity. for the less risky transactions. The ‘‘write-offs’’ are closed at a given degree of leverage and. there should be no other assets serving as collateral than the equity investment in the index portfolio itself. and follow Cornell and Green (1991) who calculate average debt beta factors from the price data of open-end bond funds. our beta calculation for high yield debt. the investment was written-off (the ‘‘write-offs’’). we refer to the cause of bankruptcy and assume that the investment was written-off because covenants were breached and debt providers claimed their rights. but. Hence.4. This leads us to keep leverage risk constant in the mimicking position over the total holding period of the ‘‘write-off’’ buyout transactions.296 and 0. the buyout was exited. the cash flows can then be duplicated by a single payment at closing and a single payoff at exit. funds are lent. Again. we control for different risk of debt. For a few buyout transactions. grow. We assume that the buyouts are settled on the last trading day of the proposed month. the index is bought on the margin at some spread above the risk-free rate. A. and hence the degree of leverage at exit. the lenders would not have demanded for additional equity if the company’s prospects were good. if the prevailing beta factor is lower than one. we assume that lending is possible at the 1-year US treasury-bill rate. The mimicking strategy is structured by investing the same amount of equity in the S&P 500 Index portfolio and levering it up according to Formula (3) with borrowed funds to achieve an equal systematic risk. by definition. Changes in ownership In some transactions the ownership structure changes within the holding period. we follow Kaplan and Ruback’s (1995) argumentation regarding buyout transactions and capitalize the tax benefits by the operating cost of capital with Formula (3) to determine the betas of the buyouts at closing and exit. Since equity claims (as residual claims) must be at least as risky as debt claims.2. the mimicking portfolio is liquidated every year. to secure parity with the buyout. The debt beta in our base case is 0. the equity payoff. we refer to Fig. Early disbursements lower the sponsors’ exposure and therefore. In the robustness checks. that debt and equity investors bear the same (low) risk. or stay constant. we deduct them at the relevant month from the prevailing equity. O. Treatment of the individual transactions Each transaction is analyzed thoroughly in terms of the timing and the character of the underlying cash flows. we only have to consider add-ons. who provide credit-spreads for high yield securities during the relevant period. As there are no premature disbursements in our ‘‘write-offs’’. Second. written-off at an infinitely large leverage ratio (the collateral preserves some debt value while the equity value approaches zero). 2109 In the simplest case without add-on investments and premature disbursements. more highly levered buyout transactions. This implies for the mimicking investment the unrealistic need to lever the benchmark portfolio to an infinite exposure.41. For the corresponding cost of debt. Ownership changes are generally denoted in the PPM but not in sufficient detail to permit further investigation. we account for these changes in the proportion . The initial payment takes place at a certain systematic risk level characterized by the operating risk and the additional leverage risk. For the ‘‘write-off’’ transactions the approach is also straightforward. we assume that leverage changes linearly over time and assign different assumptions with respect to the both outcomes. To rule this scenario out. This view is supported by the fact that these investments finally default. However. We assume that all buyout fund investments are equity investments and all remaining layers other than common or preferred equity provided by third parties are treated as debt. To appropriately mimick the transactions. the degrees of leverage are moderate and. If the equity beta of the buyout is lower than one. For the exited buyouts.A. For simplification. we extrapolate the equity beta at the time of either the add-on investments or the early disbursements. Then we determine the market capitalization weighted average for the investment-grade and for the high yield samples which are 0. Gottschalg / Journal of Banking & Finance 35 (2011) 2099–2110 bu ¼ be þ bd DE 1 þ DE ð3Þ We assume that. We distinguish between the moderately levered publicly traded companies and the in general. In order to determine a transaction’s risk structure we must differentiate between two general outcomes.410. Groh. we always truncate the risks of debt at the levels of the operating betas. meaning that leverage ratios could not be decreased by the add-on. Therefore. they immediately affect the leverage ratios and then follow the same risk pattern as the initial investments. First. The rationale for keeping lending risk-free but borrowing risky is to replicate the limited liability of the buyout sponsor.3. The add-on payments would lower the leverage ratio.3. In our stylized approach. This is either caused by non-proportional add-on investments/distributions or by exercising any kind of contingent claims. therefore.7 This implies. In our base case.1. for the beta unlevering process of the publicly quoted companies. One can argue that the required debt service could not be maintained by the buyout target. funds are lent. borrowing is associated with higher borrowing cost. the tax benefits are discounted by the cost of debt. Add-on investments and premature payoffs To track add-on investments and premature payoffs. We retrieve (from Bloomberg) weekly gross-returns and 2004 year-end market capitalizations for 314 open-end funds investing in investment-grade corporate debt and the same data for 101 open-end bond funds investing in lowgrade debt securities. Provided that add-on payments are not accompanied by changes in debt. A. our estimated debt betas are larger than the operating risks of the target companies. The systematic risk level at closing is determined by the initial equity beta of the corresponding buyout.

Solnik. Measuring idiosyncratic risks in leveraged buyout transactions. Cumming.. S.. The impact of the degrees of operating and financial leverage on systematic risk of common stock. Journal of Financial Economics 7. Journal of Banking and Finance 31. Columbia University. Rhodes-Kropf. 29–48. D. Beta and growth. Peck. Scholes. Groh. Financial Management 31. 2005. 2003. Journal of Applied Corporate Finance 2. 1990.. R. Stein. Journal of Financial Economics 59.. Gottschalg. 1791–1823. The evolution of buyout pricing and financial structure in the 1980s. Pogue. Harvard Business Review. M.. B. M. T. Stein. 1931–1967. S.. 1990. Jensen. 2007.. Williams. R. Working Paper.. 1982. 61– 74. Wessels... Siegel. Bongaerts. 165– 194. Cumming.. 1974. Bond indenture provisions and the risk of corporate debt.P. Hoboken. D. Kaplan. Kaplan. 2005. 1989b. Jensen. 3218–3250. Estimating betas from non-synchronous data. M. 5–23. E... Kaplan. 1995... The price of diversifiable risk in venture capital and private equity. Selected Studies.. 1958. 3–52. Vetsuypens. K. G.. Singer. Hoboken. Risk-return classes of New York Stock Exchange common stocks. L. LBOs. Journal of Finance 45. Fleming.. Ruback. 1993.. 568–577.. 47–54. Koller. 2005.... Wiley... Eclipse of the public corporation. Risk measurement when shares are subject to infrequent trading. 2010. Journal of Financial and Quantitative Analysis 9. The staying power of leveraged buyouts.. J. O. 217–254. P. Jones. E. 1989a. Journal of Financial Economics 75. E... Miller. Journal of Financial Economics 43... Wiley. G. and the privatization of bankruptcy. 1977. 2000. 2010. fundraising and drawdowns. 1990. 1977. Corporate ownership structure and performance: the case of management buyouts. corporation finance. 2009. L. 375–406. 2003.. Journal of Business 39. Defaults and returns in the high yield bond market. A test of the equivalent-risk class hypothesis.. 313–357. Richardson. 1984. Review of Financial Studies 22. Suchard.. Private equity and regulatory capital. S. R. How big are the tax benefits of debt? Journal of Finance 55. Journal of Financial Economics 4. J. M. 35–44. References Altman. 1966. 611–632.. Ruback. U. How risky is the debt in highly leveraged transactions? Journal of Financial Economics 27. R. Cochrane.. Smith. M. Bernado. Brown.. S. G. G. Lichtenberg. Pasternack. Journal of Finance 46. Phalippou. 2009. Journal of Financial Economics 27. Leverage and pricing in buyouts: an empirical analysis. Journal of Financial Economics 5. Journal of Financial Economics 10. S. A. Gonedes. 1987. A. Journal of Financial and Quantitative Analysis 4. 441–456. and the theory of investment. The risk and return of venture capital. Arzac. second ed. Quarterly Journal of Finance and Accounting 47.. . American Economic Review 48.. Green. J. 2007. A.. Cooper. 2001. S. 917– 944. Dimson. The structure of debt and active equity investors: the case of the buyout specialist.. Gottschalg / Journal of Banking & Finance 35 (2011) 2099–2110 of the equity stake at exit. F. Cornell. 1991... Axelson. Schoar. De Angelo. 3–29. fifth ed. The cash flow.. W. 1990.. D. Journal of Finance 60. persistence and capital flows. 1211–1220. 153–193. Wiley.2110 A. Investment Valuation: Tools and Techniques for Determining the Value of Any Asset. Journal of Financial and Quantitative Analysis 19. 1989a. Valuation for Mergers. Rhee.. 2010. Journal of Finance 44. Some preliminary findings on the association between the earnings of a firm. The investment performance of low-grade bond funds. C. Journal of Banking and Finance 29. Groh. Some new stock-market indexes. K. 1979. Chowdhry. 2002. S. L. available on SSRN. D. Journal of Banking and Finance 34. Journal of Finance 50... H. UCLA.. Empirical Research in Accounting. Zambelli. Phalippou. Ho. S. 1901– 1941. Journal of Financial Economics 29. French. J. O. Gottschalg. Weisbach. F. 159–177. 38–49.. Valuation: Measuring and Managing the Value of Companies. C. Palepu. G. Journal of Financial Economics 27. B. Kaplan.. S. N. Schwert. 1972. Quarterly Journal of Economics 108. 55–77. Journal of Financial Economics 24. 197–226. Baule. 1967.. K. 261–297. 2005. The performance of private equity funds.. 45–57. 1059–1093. Modigliani. B. J. The effects of management buyouts on operating performance and value. R.. Goedhart. 215–245. Fama. T. Regulatory harmonization and the development of private equity markets. Goyal. Efficiency and organizational structure: a study of reverse LBOs. 191–225. Industry costs of equity. The valuation of cash flow forecasts: an empirical analysis. Venture capital funds: flow-performance relationship and performance persistence. A. Venture capitalist value-added activities. 1989b. The effects of leveraged buyouts on productivity and related aspects of firm behavior.. Journal of Banking and Finance 34. T.. The cost of capital. Supplement to Journal of Accounting Research 5. Kaplan. Management buyouts: evidence on taxes as a source of value.. E. A.. 2008. M. 247–262. G. Johan.. Buyouts. Fisher. 295–331. 309–327.. M. 1969.. Cumming.... Jenkinson. 1991.. Private equity performance: returns. S. 1747–1776. Strömberg.... Damodaran. 2004.. O. 101–147. Graham.. NBER Working Paper. Capital cash flows: a simple approach to valuing risky cash flows. return and risk characteristics of private equity. Kaplan. Consequences of leveraged buyouts.. Sharpe.. Hoboken.. E. J. 1990. 287–313. Charlier... 2006... Financial Analysts Journal 28. B.. D. A. Illegal buyouts. Working paper. Cotter. Muscarella. Journal of Financial Economics 27.. 143–164. and Restructuring. Working Paper. A. De Angelo. Financial Analysts Journal 43. 51–78. Kaplan. its industry and the economy. M. M. Stock exchange seats as capital assets. P. 1997. The market model applied to European common stocks: some empirical results.. 1389–1413. Mandelker. Journal of Applied Research in Accounting and Finance 1. thus assuming that all changes in ownership develop linearly over the holding period. J. L.. Management buyouts of publicly traded corporations. D.. Journal of Banking and Finance 33. Active investors. Ljungqvist. 85–103. Ball. 2002.