Acharya Et Al (2012) PDF
Acharya Et Al (2012) PDF
Acharya Et Al (2012) PDF
Viral Acharya completed this work in part at the London Business School and while visiting the Stanford Graduate
School of Business. We are especially grateful to the PE firms that assembled and gave us access to sensitive deal
data, with a special mention to the invaluable support of Golding Capital Partners. Viral Acharya was supported
during this study by the London Business School Governance Center and Private Equity Institute, the Leverhulme
Foundation, INQUIRE Europe, and London Business Schools Research and Materials Development (RAMD)
grant. Oliver Gottschalg was supported by the HEC Foundation and the HEC PE Observatory. McKinsey &
Company also devoted significant human resources to the fieldwork and analysis, not for any client but on its
own account. We are grateful to excellent assistance and management of data collection and interviews by Amith
Karan, Ricardo Martinelli, Prashanth Reddy, and David Wood of McKinsey & Co. Ramin Baghai-Wadji, Ann
Iveson, Hanh Le, Chao Wang, and Yili Zhang provided valuable research assistance. The study has benefited
from comments of two anonymous referees, Laura Starks (editor), Yael Hochberg (discussant), Steve Kaplan
(discussant), Tim Kelly (discussant), and from seminar participants at the European Central Bank and Centre
for Financial Studies Conference (2008) in Frankfurt, the Inaugural Symposium (2008) at the London Business
Schools Coller Institute of Private Equity, the joint conference of the University of Chicago and University of
Illinois at Chicago, the London Business School, McKinsey & Co., the Western Finance Association Meetings
(2008), and participating PE firms. All errors remain our own. Send correspondence to Oliver Gottschalg,
HEC School of Management, 78351 Jour en Josas, Paris, France; telephone: +33 (0) 670017664. E-mail:
[email protected].
The Author 2012. Published by Oxford University Press on behalf of The Society for Financial Studies.
All rights reserved. For Permissions, please e-mail: [email protected].
doi:10.1093/rfs/hhs117
Advance Access publication December 6, 2012
Page: 368
368402
1. Introduction
In a seminal piece on private equity, Jensen (1989) argued that leveraged
buyouts (LBOs) create value through high leverage and powerful incentives.
He proposed that public corporations are often characterized by entrenched
management that is prone to cash-flow diversion and averse to taking on
efficient levels of risk. Consistent with Jensens view, Kaplan (1989), Smith
(1990), Lichtenberg and Siegel (1990), and others provide evidence that LBOs
create value by significantly improving the operating performance of acquired
companies and by distributing cash in the form of high debt payments.
By contrast, the recent literature has focused on the returns that private
equity (PE) fundswhich usually initiate the LBO and own, or more precisely,
manage, at least a majority of the resulting private entitygenerate for their
end investors, such as pension funds. In particular, Kaplan and Schoar (2005)
studied internal rates of return (IRRs) net of management fees for 746 funds in
19852001 and found that the median fund generated only 80% of the S&P 500
return, and that the mean funds return was only slightly higher, at around 90%.1
However, the evidence suggests that returns are better for the largest and most
mature PE housesthose operating for at least five years. Kaplan and Schoar
note that, for funds in this subset of houses, the median performance is 150%
of S&P 500 return and the mean is even higher at 170%. Furthermore, this
performance is persistent, a characteristic that is generally associated with the
potential existence of skill in a fund manager. It is interesting to note that
such long-run persistence has rarely been found in mutual funds, and when
found has generally been in the worst performers (Carhart 1997).
Our article is an attempt to bridge these two strands of literature concerning
PE, the first of which analyzes the operating performance of acquired
companies, and the second that analyzes fund IRRs. In addition, we investigate
how human capital factors are associated with value creation in PE deals. We
focus on the following questions: (i) Are the returns to equity investments by
large, mature PE houses simply due to financial leverage and luck or market
timing from investing in well-performing sectors, or do these returns represent
the value created at the enterprise level in the so-called portfolio companies,
over and above the value created by the quoted sector peers? (ii) What is the
effect of ownership by large, mature PE houses on the operating performance of
portfolio companies relative to that of quoted peers, and how does this operating
performance relate to the financial value created (if any) by these houses?
(iii) Are there any distinguishing characteristics based on the background and
1 This evidence has been replicated by studies in Europe (Phalippou and Gottschalg 2009; Phalippou 2007), though
they raise the issue of certain survivorship biases in data employed that might imply no median outperformance
relative to the market even for large and mature PE houses. This by itself does not necessarily refute Jensens
original claim; it could simply be that PE funds keep the value they create through fees. The puzzle that the
evidence on median return of PE funds raises is thus more about why their investors (the limited partners) choose
to invest in this asset class as a whole, an issue investigated by Lerner and Schoar (2004) and Lerner, Schoar,
and Wong (2007).
369
Page: 369
368402
2 We believe this time period is particularly well suited for studying value creation through operational engineering.
Kaplan and Stromberg (2008) note that operational engineering became a key private equity input to portfolio
companies primarily in the past decade.
3 In the cross-section of deals, abnormal performance has a positive, albeit imperfect, correlation to IRR,
the public-market equivalent (PME) measures employed by Kaplan and Schoar (2005), and measures
of performance based on the cash-in/cash-out multiple (ratio of outflows to inflows). These alternative
performance measures are described in the Appendix and discussed in the text.
4 For example, in theory, what we label as operational improvements could in fact be the reversion of the acquired
deals performance to the mean. However, we find evidence against the mean-reversion argument: although the
370
Page: 370
368402
sample size of deals with more than two years of pre-acquisition data is small, they show no difference to
their respective sector companies in performance trends pre-acquisition: both targeted and sector companies
show nearly the same increase in nominal sales and constant profitability. Yet, PE might still be able to identify
companies that will be subject to a positive future shock. This is something we cannot rule out. However, a
systematic relationship between PE-ownership and anticipated future performance shocks can induce abnormal
performance in PE deals. To financially exploit individual shocks on a company, a PE house must have a
systematic informational advantage in forecasting the future in comparison to the seller and other bidding PE
houses. This systematic informational advantage appears questionable in a competitive buyout market, such as
that for the large firms in Western Europe.
371
Page: 371
368402
within the funds we sampled for our deals, we find no statistically significant
bias between the performance of deals sampled and those not sampled.5
In Section 2, we review the related literature. In Section 3, we provide
a description of the data we collected and some summary statistics. In
Section 4, we describe the methodology for calculating abnormal performance.
In Section 5, we discuss operating performance. In Section 6, we link abnormal
performance and operating performance. Section 7 discusses the role of deal
partner background. Section 8 concludes.
2. Related Literature
Following the seminal work of Jensen (1989) on LBOs, the early empirical
contributions verified the impact of PE ownership on firms (Kaplan 1989;
Smith 1990; Lichtenberg and Siegel 1990).6 However, the most recent wave of
PE transactions (2001 to mid-2007) has prompted researchers to reexamine
whether buyouts are still creating value in this new era. Guo, Hotchkiss,
and Song (2011) answer this question with a sample of ninety-four U.S.
public-to-private transactions between 1990 and 2006. They find that gains
in operating performance are not statistically different from those observed
for benchmark firms. Leslie and Oyer (2008) also find weak or no evidence
of greater profitability or operating efficiency for 144 LBOs in the United
States between 1996 and 2004, relative to public companies. However, Lerner,
Sorensen, and Stromberg (2008) find evidence in a sample of 495 buyouts
that, in contrast to the oft-cited claim that PE has short-term incentives, buyout
deals in fact lead to significant increases in long-term innovation. They find
that patents applied for by firms in PE transactions are more frequently cited (a
proxy for economic importance), show no significant shifts in the fundamental
nature of the research, and are more concentrated in the most important and
prominent areas of companies innovative portfolios.
This literature has focused mainly on U.S. data, while our data are based on
PE deals in the United Kingdom and Europe.7 Several studies have examined
LBOs in the United Kingdom, which also experienced a tremendous increase in
buyout activity prior to the crisis of 200709. Nikoskelainen and Wright (2007)
5 Moreover, in contrast to the extant literature that focuses mainly on public-to-private deals, our data set also
covers deals where only part of a company is acquired (e.g., carve-out deals) and private-to-private deals, where a
non-listed business is acquired. Using carve-out and private-to-private deals is important, because they compose
74% of PE deals in Western Europe over the past decade, and they are different in size (enterprise value) and
profitability (EBITDA margin) from public-to-private deals (according to a simple, nonstatistical analysis of data
provided by Private Equity Insight).
6 Note that Kaplan (1989), Smith (1990), and Lichtenberg and Siegel (1990) also investigate whether LBOs
improved operating performance at the expense of workers. They find that the wealth gains from LBOs were
not a result of significant employee layoffs or wage reductions (see Palepu 1990 for a detailed survey of these
articles and also Kaplan and Stromberg 2008 for a comprehensive survey).
7 One distinguishing feature of the U.K. and European PE deals relative to those in the United States is that there
are more deals in the United Kingdom and Europe that are buyouts of family-owned firms, whereas there are
more public-to-private deals in the United States.
372
Page: 372
368402
study 321 exited buyouts in the United Kingdom in the period 1995 to 2004. On
average, these deals generated a 22% return to enterprise value and a 71% return
to equity, after adjusting for market return. In a related article, Renneboog,
Simons, and Wright (2007) examine both the magnitude and sources of the
expected shareholder gains in 177 public-to-private transactions in the United
Kingdom in 19972003. They find that pre-transaction shareholders receive
a premium of 40%. They also find that the main sources of post-transaction
gains in shareholder wealth are attributable to an undervaluation of the pretransaction target firm, increased interest tax shields, and the realignment of
incentives. Harris, Siegel, and Wright (2005) study the productivity of U.K.
manufacturing plants subject to management buyouts (MBO). Such plants
experienced substantial increases in productivity, the post-MBO magnitudes
of which are substantially higher than those reported in the United States by,
for example, Lichtenberg and Siegel (1990).
Among the limited evidence on the impact of human capital or skill in
PE investments, Kaplan et al. (2008) analyze the relationship between 316
portfolio company managers (CEOs) and the success of buyouts. They find that
execution skills appear to be more strongly related to success than interpersonal
skills. To our knowledge, with the exception of this study, there has been
no systematic analysis of the link between the financial returns of LBOs and
human capital factors. As Cumming, Siegel, and Wright (2007) state, There
is a need to understand the human capital expertise that successful PE firms
require. There appears to be a need to broaden the traditional financial skills
base of private equity executives to include more product and operations
expertise.
Our evidence regarding the relevance of human capital factors and, in
particular, the importance of task-specific deal partner skills (operational
or financial) fills this important gap in the literature for PE investments.
We posit that task-specific skills attributable to deal partner background are
one significant part of the persistent abnormal financial return generated by
large, mature PE houses for their investors (Kaplan and Schoar 2005).8 PE
partners seem to add value to portfolio companies by applying skills they have
accumulated over time.
In contrast, for venture capital (VC) investments, it has been established
that VC funds have an impact on the development of new companies, and
that fund expertise matters for performance. Hellmann and Puri (2002), for
example, show that VC funds add professional structure and rigor to the startups
in which they invest: startups backed by venture capital (VC) funds replace
their CEOs more frequently; introduce stock-option plans; and hire a vice
president of sales and marketing. Dimov and Shepherd (2005) find a positive
8 We note that our article is silent about the conflicts of interest between private equity houses and their investors.
Axelson, Stromberg, and Weisbach (2009), Ljunqvist, Richardson, and Wolfenzon (2007), and Metrick and
Yasuda (2007) provide good coverage of theoretical as well as empirical issues on this front.
373
Page: 373
368402
374
Page: 374
368402
The LP sample has the additional advantage of being audited rather than selfreported. It is drawn from hand-collected track records of private equity firms as
reported in fund-raising documents (Private Placement Memorandums [PPMs])
created by the PE funds. In the PPMs, we observe the list of investments made
by each PE firm. From the pool of PPMs collected by the LP, we selected
thirty-four from large, well-established European multi-fund houses active in
either mid-cap or large-cap markets. From these thirty-four PPMs, we
collected all 285 realized transactions for which both cash flows and operating
performance were available. We checked each group of deals pertaining to a
specific fund to ensure the performance, or average IRR, of each group was
representative of the fund from which it was drawn. Importantly, buyout firms
PPMs generally disclose all investments, including ones that did not perform
well. As a consequence, this sample contains numerous poorly performing
investments.
The final, combined data set comprises 395 deals from forty-eight funds
covering nearly two decades. For each deal, we have the exact structure of cash
flows, from and to the parent fund, and detailed data on financial and operating
measures. We do not have all enterprise-level cash flows, which would include,
for example, the taxes or interest and principal paid on debt.9
To compare our sample to a set of publicly listed peers, we collected data,
supplied by Datastream, for approximately 7,000 publicly listed corporations
(PLCs) in Europe, from which we constructed sector indices based on ICB
(Industry Classification Benchmark) sector codes. These codes group publicly
listed peers into ten industries and thirty-nine sectors. We also collected data
on TRS (Total Returns to Shareholders): enterprise value, net debt, equity,
sales, and EBITDA(E). The latter removes the effects of exceptional items,
which are not included in EBITDA figures provided by the firms in our
sample.
Table 1, panel A, shows that, within our sample period (19912008), our
deals are well distributed over time, although there is some concentration in
19982003 in terms of deal entry or acquisition. Table 1, panel B, provides
additional summary statistics for the deals. Deals in our sample have a high
mean, gross IRR (56.1%) and cash multiple (4.4), with large values on the right
tail, even after winsorizing our sample (we replace the lowest 5% of values with
the fifth percentile value, and the highest 5% with the ninety-fifth percentile
value). However, a high value for average IRR is to be expected from a sample
of deals from large, mature PE houses (Kaplan and Schoar 2005). We also
report an average duration of 3.9 years.
9 We also do not have all cash flows for six un-exited deals in the McKinsey data set because there is no exit cash
flow from sale, nor can it be deemed to be zero, as in the case of bankruptcies. Therefore, the end enterprise-value
cash flow was simulated using the EV/EBITDA multiple at the start of the deal and applying that number to
2006 or 2007 year-end EBITDA. Our results are robust to alternative assumptions, including one assumption
that they produced no terminal cash flow whatsoever. However, we have verified that such a pessimistic scenario
is unlikely to be appropriate for these deals.
375
Page: 375
368402
Table 1
Data overview
Panel A: Distribution of deals by entry and exit years
Years
1991
92
93
94
95
96
97
98
99
Entry
Exit
3
2
23
2
34
3
42
8
44
12
2000
2008
SUM
38
14
46
21
54
19
45
36
28
54
20
71
8
64
2
70a
19
395
395
Entry
Exit
The table shows the years in which the PE houses bought (entry) or sold (exit) the portfolio companies (deals)
in the McKinsey and LP data set (our sample).a
a Including six deals for which exit is simulated.
Panel B: Summary statistics
Variable
deal IRR (gross) %a
cash in/cash out multiple
duration (years)b
deal size (entry)b
deal size (exit)b
debt/equity (entry)a,b
debt/equity (exit)a,b
mult (entry)c
mult (exit)c
debt/EBITDA (entry)c
debt/EBITDA (exit)c
395
385
385
381
354
354
mean (median)
56.1 (43.2)
4.4 (3)
3.9 (3.4)
430.2 (141.2)
749.6 (244)
2.2 (1.9)
0.8 (0.5)
6.8 (6.5)
8.6 (7.9)
4.1 (4.1)
3.0 (2.5)
std. dev.
46.6
9.5
2
756.8
1311.9
1.5
0.9
11.1
7.5
6.9
3.1
min
0.0
0.0
0.4
0.0
0.0
0.3
0.1
173.3
67.4
104.7
13.2
max
179.0
130.0
10.8
6485.6
10290.0
6.2
3.8
42.9
54.5
35.9
21.7
t -stat of diff.
exit and entry
8.48
15.18
2.91
3.37
The table shows various financial measures for the deals in the McKinsey and LP data set. The first part reports
the financial performance and the duration. We calculate the deal IRRs (internal rate of return) using the entire
time pattern of cash inflows and outflows for each deal (portfolio company), as experienced by the PE house
(before fees). The cash-in/cash-out multiple measures the absolute value of all positive cash flows divided by
all negative cash flows minus 1. The duration captures the length of the deals in years, using the entry and exit
months and years as reported by the PE house.
The second part of the table compares the enterprise value (deal size) and several financial ratios between entry
and exit date. The number of observations is smaller than in the first part, since we include only deals that the
PE funds sold, and have to exclude deals with an equity value of zero. In addition, information on EBITDA at
entry and exit is not available for all deals.
In the last column we test for differences between exit and entry values.
Note: In Mio EUR; significance level p < 0.1, p < 0.05, p < 0.01.
a Due to a few but large outliers, we winsorize the distribution of the variable by replacing the 5% highest and
5% lowest values by the next value counting inward from the extremes.
b Only exited deals.
c Only exited deals and if EBITDA for exit and entry date is available.
376
Page: 376
368402
Groh and Gottschalg (forthcoming), it appears that the debt-to-equity ratio falls
for PE deals during their life only partly due to improvements in coverage ratio
(Debt/EBITDA), and mainly due to improvements in equity value over deal
life.
Next, we come to the important sample-selection issues. Table 2, in panels
AC, provides several relevant comparisons between our sample and the PE
universe. Overall, we conclude that our sample covers mainly large funds, but
seems to be representative in terms of performance, and includes all different
vendor types; that is, not just public-to-private deals but also the frequent
private-to-private deals.
First, Table 2, panelA, shows that the sampled funds are a good representation
of funds in Western Europe, especially when we take into account the fact that
we focus on funds whose sizes are above $300 million. All 146 funds in Western
Europe with a vintage year in 19912005 had a simple average net IRR of 23.5%
(based on 146 funds for which Preqin reports IRRs), which is not different to
the net IRR of our funds (t-statistic = 0.41 for the difference). Also, large
funds in the PE universe (again, for which Preqin reports IRRs) had returns
similar to our sample funds (t = 0.44 for the difference).
In Table 2, panels B and C provide evidence that we have not cherry-picked
the deals from the sampled funds. Table 2, panel B, compares the average
performance of the deals in the McKinsey sample, per fund (in terms of net
IRR), to the performance of the same thirty-two funds from which our deals
were drawn; average fund IRRs are based on Preqin figures. We show that the
funds in the McKinsey sample do not appear to have cherry-picked the deals
that they reported: the difference between the average, publicly reported net
fund IRR of 28.1% and the average net IRR of our deals, per fund, of 26.3% is
not statistically significant (t = 0.43). In terms of deal performance, therefore,
we have an unbiased representation of deals within the funds we sampled. For
the comparison, because publicly available data on fund performance are based
on net IRRs, we had to convert our gross deal-level IRRs, prior to fees and carry
paid to the PE funds by investors, to net IRRsor IRRs from the viewpoint
of fund investors.10 We deduct from the gross IRR a 2% annual fee and 20%
carry for IRR above the typical benchmark market return of 8%.11
10 To perform this conversion, we also construct an artificial fund of our sample deals and calculate its IRR. The
pseudo-fund starts in 1995 and lasts for thirteen years, until 2007. Investments or cash inflows take place in years
19 (with small investments in years 10 and 11 as well). The bulk of the investments occur in years 39. Cash
payouts start in year 5; in the last three years, the fund has only cash payouts. Using this pattern of cash inflows
and outflows, we calculate the gross IRR of the pseudo-fund.
11 More specifically, if (i) gross IRR 10%, then LPs keep all return except 2% fees, so that net IRR = gross IRR
2% fees; (ii) 10% < gross IRR <12.5%, then LPs keep all return up to 10% except for 2% fees and GPs keep
all return from 10% to 12.5%, so that net IRR = gross IRR 2% fees (Gross IRR 10%) = 8%; and (iii) gross
IRR 12.5%, then LPs and GPs share in 80:20 ratio the return exceeding 12.5%, so that net IRR = gross IRR
2% fees 2.5% 20%*(gross IRR 12.5%). The pooled net IRR for our deals is 23.9%, which is close to the
average net deal IRR of 26.2%.
377
Page: 377
368402
Table 2, panel C, shows a similar result for the LP data set. That is, there is no
statistical difference between the LP deals we selected (n = 285) and unselected
deals made by the same PE houses (n = 892), in terms of both performance
(t = 0.42 for the difference) and size (t = 0.35 for the difference). The deals
we selected were chosen because they provided us with the requisite financial
data, and because the names of the deal partners were available, allowing us to
determine their backgrounds.
Table 2
Data representativeness
Panel A: Benchmarking of our funds vs. PE universe funds
Net IRR, in %
mean (median)
48
146
78
24.88 (21.75)
23.53 (20.45)
22.28 (19.75)
t -stat of diff.
with (1)
0.41
0.44
mean (median)
1567.05 (920)
700.76 (248.29)
1402.4 (787.96)
t -stat of diff.
with (1)
4.43
0.89
This table compares the returns and size of the funds in the McKinsey and LP data set (our sample) with fund
returns and size of the PE universe as reported in Preqin. First, row (1) reports the net IRR and fund size for
48 (out of 84)a funds that participate in our sample and for which Preqin reports net IRRs. Second, we provide
net IRRs and fund size for the PE universe in Western Europe (with the same vintage years as the funds in our
sample) as reported in Preqin. Row (2) shows the net IRRs for all funds in Western Europe and (3) for large
funds only. We further test if the PE funds in our sample are different in terms of net IRR and fund size from (2)
the Western European universe and (3) the universe with similar fund size (above USD 300 Mio fund size, since
only five of the funds in our sample are smaller).
Note: Vintage year 19912005, significance level p < 0.1, p < 0.05, p < 0.01.
a In five cases, more than one fund of a PE house is involved; in these cases we take the simple average fund net
IRR of the funds involved and treat the funds as one fund.
b Funds larger than USD 300 Mio fund size.
Panel B: Benchmarking of deals vs. fund returns in McKinsey data set
Net IRR,a in %
(1) Our fundsb
(2) Our deals
(3) Our deals pooled in one pseudo fundd
mean (median)
t -stat of difference
32
93
1
28.13 (26.38)
26.25 (25.02)
23.91%
0.43c
This table compares the deals with the funds in the McKinsey sample by net IRRs. Row (1) provides the net
IRR for 32 out of 36 funds that participate in our sample and for which Preqin reports the net IRRs by end of
2007. In row (2) we show the simple average net IRRs of all deals in the McKinsey sample for which we have
publicly available fund return data (for 93 out of 110 deals in the McKinsey sample). In row (3) we pool these
deals artificially in one pseudo-fund.
Since the data on the European universe are primarily in the form of net IRRs, we convert our gross deal-level
IRRs (before fees charged by PE houses to fund investors) to net IRRs (after fees, or in other words, IRRs from
the viewpoint of fund investors). In the last column we test if the PE houses cherry-picked the deals in (2) out of
their funds (1) in terms of performance.
Note: Vintage year 19932003, significance level p < 0.1, p < 0.05, p < 0.01.
a Net IRR, estimated for our deals in the following way: If (i) gross IRR ! 10%, then LPs keep all return except
2% fees, so that net IRR = gross IRR 2% fees; (ii) 10% < gross IRR < 12.5%, then LPs keep all return up to
10% except for 2% fees and GPs keep all return from 10% to 12.5%, so that net IRR = gross IRR 2% fees
(gross IRR 10)% = 8%; and (iii) gross IRR " 12.5%, then LPs and GPs share in 80:20 ratio the return exceeding
12.5%, so that net IRR = gross IRR 2% fees 2.5% 20%*(gross IRR 12.5%).
b In five cases, more than one fund of a PE house is involved; in these cases we take the simple average fund
net IRR of the funds involved and treat the funds as one fund. For one deal the fund name is unknown; for three
funds we cannot find fund returns.
c Assuming unequal variance for (1) and (2).
d Pooled by calendar period using quarterly cash flows.
(continued)
378
Page: 378
368402
Table 2
Continued
Panel C: Benchmarking of deals used from LP data set
IRR, in %
n
0. 42
mean (median)
t -stat of diff.a
372.49 (106.505)
350.74 (85.0)
0.35
This table compares net IRRs and deal size of (1) the deals from the LP data set used in the analyses with
(2) the deals from the LP data set not used in the analysis (but which are from the same PE houses and for
which information on IRR and size was available). We exclude deals in (2) from the analyses, due to missing
information, e.g., on leverage or deal partner names.
In the first column we test if (1) and (2) are different by IRR and in the last column if (1) and (2) are different by
deal size.
Note: Significance level p < 0.1, p < 0.05, p < 0.01.
a Assuming unequal variance for (1) and (2).
b Deals with information on IRR (n = 671) or deal value available (n = 342).
(1)
379
Page: 379
368402
(2)
In essence, applying (1) and (2) allows us to make the following decomposition
or performance attribution of each deal IRR:
(i) Deal-level abnormal performance: i
(ii) Unlevered sector performance: RSU,i
(iii) Total leverage effect: RL,i RU,i .
12 Dealogic provides information on the base rate and the interest margin spread for only 67 deals (out of 110)
in our sample. For 19 deals we can find only the base rate (Libor vs. Euribor), and for the remaining 24 deals
we find no information. If the margin spread is unknown, we use the median spread of all PE deals in Western
Europe in the same year. If the base rate is unknown, we use LIBOR for the U.K. deals and Euribor for all other
deals.
We made sure that the assumption on the spread does not have a large impact on our results. First, the spread
does not vary much in the cross-section. In our sample period and for all deals covered in Dealogic, the standard
deviation of the weighted (by risk tranches) average spread is 1.1%, with an average (median) spread of 2.6(2.3)%
(n = 984). Second, the sensitivity of the abnormal performance of a deal to different interest rate assumptions
is less than 1. It varies according to the un-levering formula by (D/E)/ (1+D/E) * "i. For example, with a D/E
ratio of 2, a "i of 1 bp increase of the interest rate only changes the abnormal performance by 2/3 bp.
380
Page: 380
368402
The leverage effect measures the total effect of leverage on deal return.
However, we are more interested in measuring the effect of the additional
leverage that companies take on after their acquisition by PE. To arrive at the
incremental effect of increased leverage, we first rewrite (1) in terms of RL,i
as follows:
RL,i = RU,i (1+D/Ei )RD,i (1t)(D/Ei ).
Then, we substitute for RU,i based on (2):
RL,i = (i +RSU,i )(1+D/Ei )RD,i (1t)(D/Ei )
= i (1+D/Ei )+RSU,i (1+D/Ei )RD,i (1t)(D/Ei ).
And finally, we substitute for D/Ei in terms of incremental leverage:
D/Ei = D/ES,i +(D/Ei D/ES,i ).
To arrive at the following decomposition of deal IRR:
!
"
RL,i = + RSU,i (1+D/ES,i )RD,i (1t)(D/ES,i )
!
"
+ (RSU,i RD,i (1t))(D/Ei D/ES,i )+(D/Ei ) .
381
Page: 381
368402
13 The precise way in which we calculated the public market equivalent metric (PME) was as follows: (i) We
defined the deal exit date as the date when the last CF occurred. (ii) We calculated for each CF of the deal a
sector reinvestment rate from the date when the CF occurred to the exit date, based on each sector companys
total return series (TRS). For example, if a deal CF was in Q1/2000 and the exit date in Q1/2001, then we would
use a sector companys Q1/2000 TRS (e.g., 100) and Q1/2001 TRS (e.g., 150), and derive a reinvestment rate
for this CF (e.g., 50%). For a deal CF in a given sector, we used the median of all companies reinvestment rate
in its sector. (iii) We grew (effectively reinvested) each CF with this median sector interest rate over the time
(from realization of the CF until the deal exit) and did the same for the negative CFs. (iv) We summed up all
such reinvested positive CF and also all negative CFs per deal. (v) We divided the sum of all positive CFs by all
negative CFs to obtain the deals PME ratio.
382
Page: 382
368402
Table 3
Financial performance analyses
Panel A: IRR decomposition
n
(i) deal-level
abnormal
performance
(ii) levered
sector return
(iii) return
from
incremental
leverage
total IRRc
Abnormal
performance
per IRR
395
19.8 (15.4)
8.5 (8.4)
27.9 (19.1)
56.1 (43.2)
33.93%
(2) 199193
(3) 199497
(4) 19982000
(5) 200102
(6) 200307
7
61
124
100
103
32.2 (28.5)
20.5 (13.8)
16.5 (12)
17.5 (14.6)
24.7 (21.8)
9.8 (9.1)
8.5 (7)
0.7 (1.5)
11.2 (12.1)
15 (16.1)
29.1 (26.4)
30.4 (19.7)
18 (12)
21.9 (17.9)
43.9 (36.6)
71.1 (70.9)
59.4 (41.9)
35.2 (25)
50.6 (42.9)
83.5 (73.3)
45.07%
33.90%
45.71%
34.00%
28.92%
67
11 (10.1)
11.2 (10.1)
20.7 (15.2)
42.8 (40)
26.19%
Scenario
na
395
mean (median)
std. dev.
min
max
19.8 (15.37)
56.1 (43.18)
1.88 (1.37)
0.85 (0.70)
0.32 (0.30)
0.53 (0.37)
22.88
46.62
1.87
0.67
0.14
0.62
22.87
0
0.16
0.10
0.03
0.20
145.63
179.05
7.17
4.69
1.11
3.99
The table reports for all deals in the McKinsey and LP data set (our sample) summary statistics on (1) abnormal
performance, (2) IRR, and (3) the public market equivalent (PME) for each deal in the spirit of Kaplan and
Schoar (2005). In the PME calculation, we discount all cash flows with the total sector return and then calculate
the ratio of discounted cash flows to the largest cash inflow for the deal.
a Due to a few but large outliers, we winsorize the IRR, PME, and cash-in/cash-out multiple distribution by
replacing the 5% highest and 5% lowest values by the next value counting inward from the extremes.
(continued)
383
Page: 383
368402
Table 3
Continued
Panel C: Correlation matrix: Abnormal performance, IRR, PME, and sector returns
Abnormal IRR gross
performance
Abnormal
performance
IRR gross
PME sector
Levered sector
return
Cash-in/cash-out
multiple
Cash-in/cash-out
multiple sector
Abnormal
cash-in/cash-out
multiple
PME
Sector
Levered
sector
return
CashCashAbnormal
in/cash-out in/cash-out
cashmultiple
multiple in/cash-out
sector
multiple
1
0.73
0.53
0.28
1
0.53
1
0.17 0.32
0.55
0.84
0.70
0.10
0.08
0.51
0.15
0.76
0.44
0.57
0.78
0.79
0.07
0.97
1
0.18
The table shows for all deals in the McKinsey and LP sample (n = 395)a the correlation between abnormal
performance (scenario 1), IRR, the public market equivalent (PME), levered median sector returns, and different
measures of a cash-in/cash-out multiple.
Note: significance level (of the pairwise correlation coefficient) p < 0.1, p < 0.05, p < 0.01.
a Due to a few but large outliers, we winsorize the IRR, PME, and cash-in/cash-out multiple distribution by
replacing the 5% highest and 5% lowest values by the next value counting inward from the extremes.
deals in the McKinsey sample for which we were able to find the exact cost of
deal debt in Dealogic. The key findings are as follows:
Out of the average IRR of 56.1% for all 395 deals, sector risk and leverage
amplification accounts for 8.5%. In other words, less than one-fifth of the total
return is attributable either to the sector-picking ability of PE houses or simply
to pure luck. The incremental leverage effect of 27.9% is due to high deal
leverage relative to a comparatively low sector leverage. The average abnormal
performance of 19.8% is statistically significant (at the 1% level), confirming
that large, mature PE houses do in fact generate higher (enterprise-level) returns
compared with benchmarks, and that not all of these are attributable to sector
exposure and financial leverage. The medians tell a similar story.
Abnormal performance stays statistically significant (at the 1% level) when
we cluster the deals by entry date. Even in years with very low sector returns,
as in row (4), PE was able to generate abnormal performance. The high return
from incremental leverage in row (6) might correspond to the availability of
cheap debt financing, a phenomenon believed to be at work especially for PE
deals struck between 2003 and mid-2007, and that is likely responsible for the
somewhat high valuation multiples paid by PE houses in that period (Acharya,
Franks, and Servaes 2007; Kaplan and Stromberg 2008).
Abnormal performance is also positive and statistically significant for the
sixty-seven deals for which we have data on the cost of debt (via Dealogic),
although the abnormal performance of 11.0% for these deals is lower than that
for all deals (19.8%). This is partly explained by trends in the availability of
384
Page: 384
368402
data; that is, the cost of debt is typically available only for later deals, and, on
the whole, abnormal performance declines over our sample period.
Overall, the evidence points to outperformance of PE deals in our sample
in a manner that is robust to alternative measures. In Table 3, panel B, we
provide IRR, PME, cash-in/cash-out multiple, and abnormal cash-in/cash-out
multiple as alternative measures of deal performance. Consistent with our
earlier resultswhich showed that PE firms in our sample generate positive
abnormal performanceour deals also display high IRRs, PME, and cashin/cash-out multiples. For example, the average PME (relative to the sector)
is 188%, the median being 137%, and abnormal cash-in/cash-out multiple
(also relative to the sector) is 0.53, the median being 0.37. In Table 3,
panel C, we see that the abnormal performance is positively correlated with
IRR, PME, cash-in/cash-out multiple, and abnormal cash-in/cash-out multiple,
albeit imperfectly, with correlation coefficients of 0.73, 0.53, 0.55, and 0.57,
respectively. Interestingly though, abnormal performance measures are weakly
or even negatively correlated with sector returns.
5. Operating Performance
5.1 Operating measures
The next step in our analysis is determining if, at the enterprise level, abnormal
financial performance is related to abnormal operating performance. Abnormal
operating performance can be displayed in two ways: as a larger EBITDA
growth rate of the portfolio company during PE ownership compared with
pre-acquisition, or as a larger EBITDA growth rate after PE ownership
compared with the sector. To disentangle the PE impact on EBITDA during PE
ownership, we focus on the effects on (i) sales and (ii) profitability (margin =
EBITDA/Sales). We capture the impact on the company after the PE ownership
period by analyzing (iii) the EBITDA multiple (Enterprise Value/EBITDA) at
time of exit from the deal. Here, we have to rely on the assumption that market
expectations are rational at exit, since we do not have operational figures after
the PE phase for many of the deals (trade sales, for example).
The three measures we analyze in detail are:14
(i) Sales, equal to operating revenues earned in the course of ordinary
operating activities.
(ii) Margin (EBITDA/Sales). EBITDA (Earnings before Interest, Taxes,
Depreciation, and Amortization) is equal to Operating revenues COGS
(cost of goods sold) SG&A (selling, general, and administrative
expenses) Other (e.g., R&D) = Operating income. Note that EBITDA
is commonly used as it shows a companys fundamental operational
that year.
385
Page: 385
368402
15 The reason for the exclusion of Non-operating income is that this measure contains income derived from a
source other than a companys regular activities and is by definition nonrecurring. For example, a company may
record as non-operating income the profit gained from the sale of an asset other than inventory (which can be
large in relation to the operating income). From a practitioners perspective, an EBITDA multiple including
Non-operating income would not be a helpful measure to understand the price paid in relation to the current
performance capability. From our perspective, the operational performance indicator EBITDA would then be
subject to a measurement error.
386
Page: 386
368402
0.08 (0.60)
0.23 (0.17)
3.29
6.71
zero
0.34
0.65
sector
1.44 (1.5)
0.47 (0.3)
1.33 (1.02)
0.15 (0.32)
15.89 (16.9)
1.93 (4.66)
5.86
5.79
6.88
1.45
3.99
7.24
1.35
(n = 242)a,b
6.39
4.43
0.75
sector
15.29 (20.16)
4.49 (8.6)
1.46 (1.37)
0.27 (0.59)
1.43 (1.65)
0.59 (0.39)
6.46 (4.78)
6.95 (6.32)
5.11
2.43
5.94
1.71
(n = 146)a,b
4.75
4.28
9.59
18.42
zero
3.28
4.18
2.64
0.67
sector
16.79 (13.74)
1.89 (5.13)
1.12 (0.63)
0.04 (0.27)
1.46 (1.42)
0.3 (0.17)
9.95 (7.59)
8.21 (7.11)
4.61
1.07
4.13
0.3
(n = 96)a,b
4.27
1.79
12.05
17.46
zero
5.48
4.38
3.06
2.06
sector
The table reports for various operating measures (x ) the difference ("xi = xiT xit ) from the last pre-PE-ownership year (t = 0) to last PE-ownership year (T ) for all exited deals. We also
report the same for deal corresponding sector companies ("xs = xsT xst ). First, we report the changes for all deals (i). Second, we show in column (ii) changes only for deals that had no
M&A event (organic deals) during PE ownership. Third, we report in column (iii) changes only for deals that had M&A events (inorganic deals).
We divide the difference for log sales between t and T by the number of PE ownership years (T t) to get annual nominal sales growth. We use median sector changes, given that there are
often less than 100 companies in each sector. For each column (i)(iii) we test if the changes are different from zero and also for differences between deal and median sector changes, in the
spirit of a difference-in-difference (DiD).
Note: All values in percent, except change in multiples; significance level p < 0.1, p < 0.05, p < 0.01.
a Due to a few but large outliers, we winsorize the distribution of the variable by replacing the 5% highest and 5% lowest values by the next value counting inward from the extremes.
b Including deals only with entry and exit EBITDA multiple available and for log mult excluding observations with negative EBITDA.
7.89 (5.96)
7.47 (6.65)
14.82
25.18
zero
Variable
The table provides performance trends in the last year PRE PE ownership (t = 0) for all deals in our sample, for which we have at least two years of PRE PE ownership data (n = 69). More
specifically, the table reports for log sales and margin (x ) the difference ("xi,pre = xit xit1 ) from two years pre-PE-ownership (t 1) to last pre-PE-ownership year (t = 0). We also report
the same for deal corresponding sector companies ("xs,pre = xst xst1 ).
Note: All values in percent, median sector changes.
5.87 (7.18)
6.54 (5.52)
mean (median)
Variables
Table 4
Operating performance analyses
Corporate Governance and Value Creation: Evidence from Private Equity
387
Page: 387
368402
Table 4, panel B, captures the change during PE ownership and shows the
difference ("x i = xiT xit ) from the last pre-acquisition year (t) to the last PEownership year (T ). We analyze the annual change by dividing the difference,
"x i , by the number of years of PE ownership (T t).
In the first set of columns, we report the changes for all deals with sufficient
operational data available (n = 255). In the second and third columns, we
separate deals with organic strategies (n = 151) from those with inorganic
strategies (n = 104)the latter of which include major, follow-on M&A events
during the private phaseso that we can analyze "x i by strategy. This also
allows us to control for the effects of M&Aevents, which cause sales to increase,
and EBITDA margins and multiples to either increase or decrease, depending
on whether the ratios of the acquired entities are higher or lower than those of
the acquirer. It is important to note that the first M&A event tends to happen
after one year, on average, for deals in the McKinsey sample (for which detailed
information on M&A events is available). This tends to suggest that, given the
planning required to execute an acquisition, M&Aevents are planned rather than
a reaction to performance that is better or worse than anticipated. We also test if
the changes are different from zero and, in the spirit of a difference-in-difference
(DiD), test for differences between deal and median sector changes.
Overall, PE ownership tends to have a positive impact on the operating
performance in our sample.As shown in Table 4, panel B, column (i) on average,
all deals show a margin improvement of 1% and a multiple improvement of 1.1
during PE ownership, relative to their sector peers. Interestingly, the margin
improvements of the deals in our sample are slightly lower than the 1.4% to
3.8% reported in Kaplan (1989). In contrast, deals do not outperform the sector
in terms of annual sales growth, although portfolio company sales do grow by
a statistically significant 7.9% during PE ownership.
Another important result shown in Table 4, panel B, is that large M&A
transactions do not cause our findings on underlying operating improvements
to change: separately, both organic and inorganic deals, columns (ii) and (iii)
respectively, display improvements in both EBITDA margin and multiple,
relative to their sector peers.16 Only inorganic deals seem to increase sales above
sector, but this isnt surprising due to the large, post-deal M&A transactions
involved, which naturally boost revenue.
6. Abnormal Performance and Operating Performance
Having separately identified abnormal financial and operating performance of
PE deals relative to their quoted peers, we can now investigate the relationship
16 In a robustness check in the McKinsey data set, we also include deals with M&A events after two years of PE
ownership in the organic deal set and find our results qualitatively unchanged. We include those deals, since late
M&A events might be endogenously determined by the observed performance of the deal. M&A events in the
first two years can be treated as exogenous, if we assume that it takes at least one year to find out that a deal
is underperforming and at least another year to identify and buy another company.
388
Page: 388
368402
Table 5
Financial and operating performance
Panel A: Abnormal performance and operating performance changes
Independent variables
(1)a
(3)a
0.44
(2.40)
1.94
(5.82)
2.23
(4.55)
duration
log value
0.14
(0.20)
(2)a,b
0.37
(2.12)
1.49
(4.34)
2.02
(4.10)
3.64
(4.66)
0.34
(0.50)
0.84
(3.61)
2.09
(4.00)
2.47
(4.05)
0.63
(0.68)
(4)a,b
0.70
0.02
(3.16)
(0.04)
1.78
1.83
(3.30)
(2.96)
2.28
1.40
(3.82)
(1.17)
4.00
(3.88)
0.21
1.33
(0.27)
(1.01)
(6)a,b
0.13
(0.41)
0.79
(1.49)
0.50
(0.37)
4.69
(3.07)
1.27
(1.04)
PE house
entry period
intercept
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
no. of deals
R 2 adjusted
234
0.25
234
0.34
140
0.34
140
0.43
94
0.13
94
0.29
The table relates cross-sectional financial abnormal performance to changes in operating measures with OLS
regressions for all exited deals in the McKinsey and LP data set. For the operational changes of a deal we calculate
for EBITDA margin, log sales, and log EBITDA multiple the average difference ("xi = xiT xit ) from the last
pre-PE-ownership year (t = 0) to last PE-ownership year (T ). We divide the difference for log sales by the number
of PE ownership years (T t ) to get annual nominal sales growth. In the same way we calculate median changes
in the deal corresponding sector companies ("xs = xsT xst ). We add to the regressions the difference between
deal changes and sector changes "xi "xs .
First, in regressions (1)(2) we use organic and inorganic deals. Second, in regressions (3)(4) we show
regressions for only organic (without major M&A events) deals, and in regressions (5)(6) for only inorganic
(with major M&A events) deals.
In the lower part of the table we control for deal duration and size. We also add entry time period (199193,
199497, 19982000, 200102, 200307) and PE house dummies for time period and fund fixed effects.
Note: OLS regressions, t -stats in parentheses with robust standard errors; significance level p < 0.1, p < 0.05,
p < 0.01.
a Due to a few but large outliers, we winsorize the IRR distribution (before deriving abnormal performance) by
replacing the 5% highest and 5% lowest values by the next value counting inward from the extremes (and for
the operational measures by replacing the 10% highest and 10% lowest values).
b Including deals with entry and exit EBITDA multiple available only.
(continued)
17 However, in unreported results, we find that the significance and size of the estimates on operating improvements
is minimally affected by omitting time dummies for entry years. We do not include dummies for the size of the
deals since size does not show up as significant and does not increase the explanatory power. This is potentially
due to a lack of variation in size in our sample that consists solely of large deals.
389
Page: 389
368402
Table 5
Continued
Panel B: IRR and operating performance changes
Independent variables
(1)a
(3)a
(4)a,b
1.78
1.40
0.71
(2)a,b
(6)a,b
0.62
(3.22)
(0.85)
2.40
3.30
(2.51)
(2.88)
1.78
3.62
(1.51)
(1.78)
10.95
(4.91)
0.48
5.09
(0.33)
(1.83)
0.35
(0.70)
0.81
(0.75)
1.47
(0.76)
11.25
(4.15)
4.94
(1.90)
1.83
(1.32)
0.51
(1.54)
1.92
(3.12)
1.75
(1.97)
10.51
(6.48)
2.42
(1.73)
PE house
entry period
intercept
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
no. of deals
R 2 adjusted
234
0.29
234
0.45
140
0.33
140
0.48
94
0.31
94
0.51
(1.85)
3.24
(5.25)
2.35
(2.38)
duration
log value
(3.47)
3.27
(3.33)
2.29
(1.84)
0.68
(0.42)
The table relates IRR (in %) to operational changes ("xi and "xs ) with OLS regressions for all exited deals in the
McKinsey and LP data set. We calculate the PME in the spirit of Kaplan and Schoar (2005). We use as independent
variables the difference between deal changes and sector changes "xi "xs . For the operational changes of a deal
we calculate for EBITDA margin, log sales, and log EBITDA multiple the average difference ("xi = xiT xit )
from the last pre-PE-ownership year (t = 0) to last PE-ownership year (T ). We divide the difference for log sales
by the number of PE ownership years (T t ) to get annual nominal sales growth. In the same way we calculate
median changes in the deal corresponding sector companies ("xs = xsT xst ).
First, in regressions (1)(2) we use all organic and inorganic deals. Second, in regressions (3)(4) we show
regressions for only organic (without major M&A events) deals, and in the regressions (5)(6) for only inorganic
(with major M&A events) deals.
In the lower part of the table we control for deal duration and size. We also add entry time period (199193,
199497, 19982000, 200102, 200307) and PE house dummies for time period and fund fixed effects.
Note: OLS regressions, t -stats in parentheses with robust standard errors; significance level p < 0.1, p < 0.05,
p < 0.01.
a Due to a few but large outliers, we winsorize the distribution of the dependent variable by replacing the 5%
highest and 5% lowest values by the next value counting inward from the extremes (and for the operational
measures by replacing the 10% highest and 10% lowest values).
b Including deals with entry and exit EBITDA multiple available only.
(continued)
been lucky on some deals simply because they bought them at the right time
when the margins or multiples in the sector were growing. We therefore include
the change above sector for all three operating measures.
Of the three measures of operating performance, the two that we identified as
being improved during PE ownershipEBITDA margin and multiplealso
appear to be significant determinants of abnormal performance: a change in
either measure has a positive and economically meaningful impact on abnormal
performance (columns (1) and (2)). Sales growth also relates to abnormal
performance, even though it is not generally improved above sector by PE
(as shown previously in Table 4, panel B).
For organic deals, as described in Table 5, panel A, regression (3), abnormal
performance is driven by changes in all three operating measures: a 1%
improvement in EBITDA margin above sector increases abnormal performance
390
Page: 390
368402
Table 5
Continued
Panel C: PME and operating performance changes
Independent variables
(1)a
(2)a,b
(3)a
(4)a,b
0.05
0.08
0.07
0.05
(6)a,b
0.06
0.06
(3.83)
(1.64)
(1.71)
0.16
0.25
0.24
(3.31)
(3.78)
(3.27)
0.14
0.30
0.30
(2.38)
(2.71)
(2.48)
0.18
0.02
(1.88)
(0.10)
0.09
0.20
0.20
(1.00)
(1.42)
(1.41)
0.13
(1.87)
0.08
(0.87)
PE house
entry period
intercept
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
no. of deals
R 2 adjusted
234
0.31
234
0.31
140
0.33
140
0.35
94
0.2
94
0.19
duration
log value
(3.94)
0.18
(3.70)
0.15
(2.51)
(5)a
(3.40)
0.17
(4.86)
0.21
(4.39)
0.10
(1.19)
0.14
(1.85)
(3.55)
0.18
(5.65)
0.22
(4.58)
The table relates public market equivalent (PME) to operational changes ("xi and "xs ) with OLS regressions for
all exited deals in the McKinsey and LP data set. We calculate the PME in the spirit of Kaplan and Schoar (2005).
We use as independent variables the difference between deal changes and sector changes "xi "xs . For the
operational changes of a deal we calculate for EBITDA margin, log sales, and log EBITDA multiple the average
difference ("xi = xiT xit ) from the last pre-PE-ownership year (t = 0) to last PE-ownership year (T ). We divide
the difference for log sales by the number of PE ownership years (T t ) to get annual nominal sales growth. In
the same way we calculate median changes in the deal corresponding sector companies ("xs = xsT xst ).
First, in regression (1)(2) we use all organic and inorganic deals. Second, in regression (3)(4) we show
regressions for only organic (without major M&A events) deals, and in the regression (5)(6) for only inorganic
(with major M&A events) deals.
In the lower part of the table we control for deal duration and size. We also add entry time period (199193,
199497, 19982000, 200102, 200307) and PE house dummies for time period and fund fixed effects.
Note: OLS regressions, t -stats in parentheses with robust standard errors; significance level p < 0.1, p < 0.05,
p < 0.01.
a Due to a few but large outliers, we winsorize the distribution of the dependent variable by replacing the 5%
highest and 5% lowest values by the next value counting inward from the extremes (and for the operational
measures by replacing the 10% highest and 10% lowest values).
b Including deals with entry and exit EBITDA multiple available only.
by roughly 2.1%; when the multiple from entry to exit grows by 1, abnormal performance increases by roughly 2.5%; and a 1% sales growth above sector alters
abnormal performance by 0.8%. Our results are qualitatively unchanged when
we include duration in the regression in column (4). For inorganic deals, however, there is little evidence to suggest that changes in operational performance
drive abnormal performance: only margin improvements seem to contribute
to abnormal performance (although the size and significance is lower than for
organic deals), but the result is inconclusive, as deals with significant, follow-on
M&A are subject to an error term due to distortion by the acquired entity.
The economic contribution of these operating improvements is substantial
for explaining abnormal performance. In the previous sections we identified a
median abnormal performance of 15.4% for all deals (Table 3, panel A) and,
for organic deals, a typical EBITDA margin increase of approximately 1%
391
Page: 391
368402
IQ database, the PE house website, or press articles. We had to change the background for only a few deals, since
another single leading deal partner with a different background from the interviewee was mentioned.
392
Page: 392
368402
393
Page: 393
368402
17.1
18.6
24.3
16.5
18.8
23.7
17.7
18.5
25.7
16.7
18.2
26.1
18.9
18.3
20
(1)
abnorm.
perform.
53
47.3
60.2
41.8
53.8
66
50.9
51.6
69.1
57
48.9
67.1
54.8
50.4
56.3
(2)
IRR
1.7
1.7
2.2
1.2
1.9
2.2
1.7
1.7
2.3
1.5
1.8
2.2
1.9
1.7
1.8
(3)
PME
121
50
27
43
147
8
63
126
9
48
131
19
60
117
21
16.8
19.2
25
17.4
18.7
19.9
19
17.8
24
18.9
17.2
26.1
22.4
16.4
19.1
(1)
abnorm.
perform.
56.2
49.6
63.4
43.5
58.1
73.7
56.7
53.3
78.8
61.1
50.5
75.8
65.5
49.3
61.7
(2)
IRR
1.6
1.5
2.4
1.2
1.7
2.6
1.5
1.6
2.9
1.4
1.6
2.4
2
1.4
2
(3)
PME
58
51
22
14
109
8
40
83
8
22
97
12
35
83
13
17.9
18
23.3
13.7
18.9
27.5
15.6
19.6
27.5
11.7
19.6
26.2
12.9
20.9
21.6
(1)
abnorm.
perform.
46.2
45
56.3
36.4
48
58.3
41.7
49.1
58.3
47.9
46.6
53.2
36.4
52
47.7
(2)
IRR
2
1.8
2.1
1.2
2.1
1.8
2.1
1.9
1.8
1.7
2.1
1.8
1.7
2.2
1.4
(3)
PME
The table gives an overview of the background of the partners (who led the deals in our sample). The operation dummy indicates if a leading deal partner has at least once in his or her
professional career worked in an consulting or industry-related job. The science dummy indicates a university degree in science, and the mba dummy indicates a master of business
administration. The ca dummy stands for chartered accountants. In the last part of the table we report the tenure of the deal partner in the PE industry in years at entry date. Column (i)
reports the background variables for all organic and inorganic deals in the McKinsey and LP sample, and column (ii) reports for deals with organic and (iii) inorganic strategy separately. For
each category, we report the mean of (1) abnormal performance, (2) IRR, and (3) PME.
Note: Simple averages, medians in parentheses, abnormal performance, and IRR in percent.
a Due to a few but large outliers, we winsorize the distribution of the variable by replacing the 5% highest and 5% lowest values by the next value counting inward from the extremes.
179
101
49
57
256
16
ca = 1
ca = 0
n/a
> 5 yrs
up to 5 yrs
n/a
103
209
17
mba = 1
mba = 0
n/a
PE experience
70
228
31
science = 1
science = 0
n/a
education
95
200
34
operation = 1
operation = 0
n/a
professional background
Table 6
Performance overview by deal partner background and PE strategy
394
Page: 394
368402
Table 7
Abnormal performance by deal partner background and PE strategy
(1) Dependent variable: Abnormal performance in % a
Independent
variables
(1)
operation
organic
operation
organic
science
mba
ca
tenure
tenure
organic
duration
log value
0.46
(0.17)
(2)
0.46
(0.17)
0.15
(0.06)
(3)
(4)
(5)
5.51
(1.56)
2.64
(0.88)
8.84
(1.88)
5.20
0.05
(1.29)
(0.01)
2.32
(0.40)
11.15
(2.12)
1.47
(0.63)
3.51
(1.37)
1.34
(0.48)
0.14
(0.40)
0.38
(0.89)
5.86
5.86
5.73
5.85
(6.93)
(6.89)
(6.70)
(7.12)
1.46
1.46
1.10
1.22
0.87
(1.95)
(1.95)
(1.01)
(1.20)
(1.22)
(6)
(7)
(8)
0.01
(0.00)
1.28
(0.46)
9.05
(2.28)
3.23
(0.95)
14.41
(2.77)
11.61
(2.38)
5.43
(0.80)
20.66
(3.44)
1.75
(0.62)
6.10
(2.22)
1.01
(0.31)
0.07
(0.18)
0.26
(0.15)
0.75
(1.05)
0.98
(0.97)
0.88
(1.23)
PE house
entry period
intercept
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
no. of obs.
R 2 adjusted
295
0.26
295
0.25
295
0.26
251
0.35
295
0.06
295
0.05
295
0.08
251
0.14
The table relates cross-sectional financial abnormal performance to deal partner background with OLS regressions
for all exited deals in the McKinsey and LP sample.
The first eight independent variables capture the background of the partners (who led the deals in our sample). The
operation dummy indicates if a leading deal partner has at least once in his or her professional career worked
in a consulting or industry-related job. The science dummy indicates a university degree in science, and the
mba dummy indicates a master of business administration. The ca dummy stands for chartered accountants.
The tenure variable represents the tenure of a deal partner in the PE industry in years at entry date. We also add
interaction terms for (a) operation and organic and (b) tenure and organic.
In the lower part of the table we control for deal duration and size. We also add entry time period (199193,
199497, 19982000, 200102, 200307) and PE house dummies for time period and fund fixed effects.
Note: OLS regression, t -stat in parentheses with robust standard errors; significance level p < 0.1, p < 0.05,
p < 0.01.
a Due to a few but large outliers, we winsorize the distribution of the variable by replacing the 5% highest and
5% lowest values by the next value counting inward from the extremes.
395
Page: 395
368402
251.00
0.26
no. of obs.
R 2 adjusted
251.00
0.13
Yes
Yes
Yes
251.00
0.35
Yes
Yes
Yes
0.66
(0.30)
4.25
(2.39)
0.45
(0.15)
2.57
(1.75)
0.28
(0.19)
0.40
(0.22)
0.16
(1.33)
0.92
(1.93)
levered
sector
return in %
(3)
251.00
0.36
Yes
Yes
Yes
0.73
(0.38)
3.89
(2.50)
0.59
(0.23)
2.30
(1.86)
0.17
(0.13)
0.76
(0.47)
0.15
(1.44)
0.79
(1.89)
unlevered
sector
return in %
(4)
251.00
0.18
Yes
Yes
Yes
0.02
(0.10)
0.04
(0.20)
0.15
(0.51)
0.13
(0.65)
0.18
(0.85)
0.12
(0.63)
0.01
(0.77)
0.18
(2.94)
(5)
debt/equity
at entry date
Dependent variables a
201.00
0.22
Yes
Yes
Yes
5.35
(2.69)
5.46
(3.41)
3.77
(1.44)
1.24
(0.82)
1.22
(0.78)
0.94
(0.48)
0.07
(0.64)
0.56
(1.07)
(6)
"xi "xs
log sales
201.00
0.01
Yes
Yes
Yes
2.19
(1.85)
1.16
(1.44)
2.64
(1.74)
0.26
(0.33)
0.28
(0.37)
0.62
(0.62)
0.03
(0.48)
0.58
(2.75)
(7)
"xi "xs
margin
212.00
0.03
Yes
Yes
Yes
0.52
(0.62)
0.85
(1.62)
0.64
(0.62)
1.09
(1.83)
0.79
(1.53)
0.04
(0.05)
0.02
(0.32)
0.08
(0.55)
(8)
"xi "xs
mult
The table relates cross-sectional different financial and operational performance measures to deal partner background with OLS regressions for all exited deals in the McKinsey and LP sample.
The regressions (1)(4) relate IRR, public market equivalent (PME), levered and unlevered sector returns to deal partner background. In regressions (6)(9) we use the deal changes above
sector changes ("xi "xs .) during PE ownership in the operational measures as dependent variable.
The first eight independent variables capture the background of the leading partners (who led the deals in our sample). The operation dummy indicates if a leading deal partner has at least
once worked in a consulting or industry related job. The science dummy indicates a university degree in science, the mba dummy a master of business administration, and the ca dummy
for chartered accountants. The tenure variable represents the tenure of a deal partner in the PE industry in years at entry date.
In the lower part we also add entry time period (199193, 199497, 19982000, 200102, 200307) and PE house dummies for time period and fund fixed effects.
Note: OLS regression, t -stat in parentheses with robust standard errors, significance level p < 0.1, p < 0.05, p < 0.01.
a Due to a few but large outliers, we winsorize the distribution of the dependent variable by replacing the 5% highest and 5% lowest values by the next value counting inward from the extremes
(and for the operational measures by replacing the 10% highest and 10% lowest values).
Yes
Yes
Yes
(2)
0.39
(0.87)
0.62
(2.08)
1.32
(2.38)
0.01
(0.05)
0.29
(1.06)
0.16
(0.59)
0.03
(1.27)
0.11
(1.01)
(1)
PME
21.47
(3.07)
8.22
(1.24)
41.65
(4.12)
4.93
(0.77)
10.66
(1.92)
5.84
(0.94)
0.04
(0.08)
2.37
(1.25)
IRR
in %
PE house
entry period
intercept
log value
tenure
ca
mba
science
operation organic
organic
operation
Independent variables
Table 8
Different performance measures by deal partner background and PE strategy
396
Page: 396
368402
397
Page: 397
368402
confront issues similar to those faced after the boom and bust cycles of the
late 1980s and early 1990s. Significant policy interest has also been expressed
in understanding and quantifying the long-run impact of PE in terms of value
creation at the enterprise level, and in attributing this value creation to financial
engineering, luck or sector-picking ability, and operational engineering.19
This article is best viewed as an attempt to get at some of these issues
with three significant contributions. First, we rely only on returns and leverage
information at the level of a deals equity, and the returns and leverage of quoted
peer firms, in order to extract a measure of abnormal performance of the deal
at enterprise level. The methodology also quantifies the contributions to deal
return due to leverage and luck or sector-picking ability. Second, by using this
measure of abnormal performance, we show that, for deals of large, mature PE
houses in Western Europe, there is evidence consistent with value creation for
portfolio companies on average. Furthermore, abnormal performance correlates
well with operating outperformance of deals measured as improvement in
EBITDA margins and multiples relative to the quoted sector. Third, we provide
evidence that the superior performance of these PE houses is at least partly due
to differences in human capital or skill factors at level of the deal partner. In
particular, the match between the deal strategy (M&A-based or organic) and
partner background (financial/accounting or operating/consulting) is correlated
with deal performance.
Overall, our results can be interpreted as providing a microscopic view of
the operational expertise employed by large, mature PE houses in improving
companies they acquire. Returns to this expertise are likely the reason behind
persistent and significant outperformance of funds run by these houses.
Appendix
Table A1
Variables
Panel A: Description of the independent variables used
Topic
Variable
Unit
Professional
background
Operation
binary
Description
Source
(continued)
19 Indeed, in some cases, such as in the United Kingdom, policy makers have undertaken independent
recommendations based on interactions with the PE industry to improve disclosure on such value attribution.
See the House of Commons Treasury Committees Tenth Report in the U.K. of Session 200607 and the Sir
David Walker Report on Disclosure and Transparency in Private Equity (2007).
398
Page: 398
368402
Table A1
Continued
Panel A: Description of the independent variables used
Topic
Variable
Unit
Finance
binary
Science
binary
Mba
binary
ca
binary
PE experience
Tenure
years
Deal strategy
Inorganic
binary
M&A activity as
reported by the PE
house, press, Capital
IQ, or Dealogic
Financial
measures
Log value
EUR mio
As reported by the PE
houses
Duration
years
IRR
percent
Log sales
EUR mio
Margin
Mult
Log mult
ratio
ratio
ratio
Education
Operational
measures
Description
Source
As reported by the PE
houses
The table describes the main independent variables used in the article. The first part of the table captures the
variables for the deal partner background, and the second part the main financial and operational variables.
a We assign for a topic n/a if the biography is not available or silent about a topic (topics are professional
background, education, or PE experience).
b Partner interviewed if not mentioned otherwise in Capital IQ database, PE house website, or press articles.
For deals without an interview available or if the interviewed partner is not the leading deal partner, we use the
leading deal partner if mentioned in LP database, Capital IQ database, PE house website, or press article.
c We assign n/a for deals with insufficient information on deal strategy in web/press and without significant
transactions in Capital IQ or Dealogic.
(continued)
399
Page: 399
368402
Table A1
Continued
Panel B: Description of the main dependent variables used
Topic
Sector
Deal
Deal outperformance
Variable
Unit
Description
Source
Datastream
As reported
by the PE
houses
Datastream
and PE
houses
Public market
equivalent
(PME)
ratio
Abnormal cashin/cash-out
multiple
ratio
Levered sector
return rates
Unlevered sector
return rates
Cash-in/cash-out
multiple sector
ratio
IRR
Unlevered IRR
Cash-in/cash-out
multiple
ratio
Abnormal
performance
The table describes the measures used to derive the main dependent variables used in the article. The first part
of the table captures sector variables; the second part, the deal-level variables derived from the data provided by
the PE houses; the last part, the difference between the two.
References
Acharya, V., J. Franks, and H. Servaes. 2007. Private equity: Boom and bust? Journal of Applied Corporate
Finance 19(4):4453.
Acharya, V., C. Kehoe, and M. Reyner. 2009. Private equity vs. PLC boards in the U.K.: A comparison of practices
and effectiveness. Journal of Applied Corporate Finance 21(1):4556.
Axelson, U., T. J. Jenkinson, P. Stromberg, and M. Weisbach. 2008. Leverage and pricing in buyouts: An empirical
analysis. Working Paper, Swedish Institute for Financial Research.
Axelson, U., P. Stromberg, and M. Weisbach. 2009. The financial structure of private equity funds. Journal of
Finance 64(4):154982.
400
Page: 400
368402
Bottazzi, L., M. Rin, and T. Hellmann. 2008. Who are the active investors? Evidence from venture capital.
Journal of Financial Economics 89:488512.
Carhart, M. 1997. On persistence in mutual fund performance. Journal of Finance 52(1):5782.
Chevalier, J., and G. Ellison. 1999. Are some mutual fund managers better than others? Cross-sectional patterns
in behavior and performance. Journal of Finance 54(1):87599.
Cumming, D., S. D. Siegel, and M. Wright. 2007. Private equity, leveraged buyouts, and governance. Journal of
Corporate Finance 13:43960.
Dimov, S., and D. Shepherd. 2005. Human capital theory and venture capital firms: Exploring home runs and
strike outs. Journal of Business Venturing 20:121.
Gibbons, R., and M. Waldman. 2004. Task-specific human capital. American Economic Review 94(2):2037.
Groh, A. P., and O. Gottschalg. Forthcoming. The effect of leverage on the cost of capital of U.S. buyouts. Journal
of Banking and Finance.
Guo, S., E. S. Hotchkiss, and W. Song. 2011. Do buyouts (still) create value? Journal of Finance 66:479517.
Harris, R., S. D. Siegel, and M. Wright. 2005. Assessing the impact of management buyouts on economic
efficiency: Plant-level evidence from the United Kingdom. Review of Economic and Statistics 87:14853.
Hellmann, T., and M. Puri. 2002. Venture capital and the professionalization of start-ups: Empirical evidence.
Journal of Finance 57:16997.
Jensen, M. 1989. Eclipse of the public corporation. Harvard Business Review, Sept.Oct.: 6174.
Jones, C. M., and M. Rhodes-Kropf. 2003. The price of diversifiable risk in venture capital and private equity.
Presented at American Finance Association meeting, Washington, DC.
Kaplan, S. 1989. The effects of management buyouts on operations and value. Journal of Financial Economics
24:21754.
Kaplan, S., N. Klebanov, M. Mark, and M. Sorensen. 2008. Which CEO characteristics and abilities matter?
Presented at American Finance Association meeting, New Orleans, LA, July.
Kaplan, S., and A. Schoar. 2005. Private equity performance: Returns, persistence, and capital flows. Journal of
Finance 60(4):1791823.
Kaplan, S., and J. Stein. 1993. The evolution of buyout pricing and financial structure in the 1980s. Quarterly
Journal of Economics 108:31358.
Kaplan, S., and P. Stromberg. 2008. Leveraged buyouts and private equity. Journal of Economic Perspectives
23(1):12146.
Lerner, J., and A. Schoar. 2004. The illiquidity puzzle: Theory and evidence from private equity. Journal of
Financial Economics 72(1):340.
Lerner, J., A. Schoar, and W. Wong. 2007. Smart institutions, foolish choices? The limited partner performance
puzzle. Journal of Finance 62(2):73164.
Lerner, J., M. Sorensen, and P. Stromberg. 2008. Private equity and long-run investment: The case of innovation.
Working Paper, Stockholm School of Economics.
Leslie, P., and P. Oyer. 2008. Managerial incentives and value creation: Evidence from private equity. Working
Paper, Stanford Graduate School of Business.
Lichtenberg, F., and D. S. Siegel. 1990. The effects of leveraged buyouts on productivity and related aspects of
firm behaviour. Journal of Financial Economics 27:16594.
Ljunqvist, A., M. Richardson, and D. Wolfenzon. 2007. The investment behavior of buyout funds: Theory and
evidence. Working Paper, Stern School of Business, New York University.
Metrick, A., and A. Yasuda. 2007. Economics of private equity funds. Working Paper, Wharton Business School.
401
Page: 401
368402
Nikoskelainen, E., and M. Wright. 2007. The impact of corporate governance mechanisms on value increase in
leveraged buyouts. Journal of Corporate Finance 13(4):51137.
Palepu, K. G. 1990. Consequences of leveraged buyouts. Journal of Financial Economics 27:24762.
Phalippou, L. 2007. Investing in private equity funds: A survey. Working Paper, University of Amsterdam.
Phalippou, L., and O. Gottschalg. 2009. The performance of private equity funds. Review of Financial Studies
22(4):174776.
Renneboog, L., T. Simons, and M. Wright. 2007. Why do firms go private in the U.K.? Journal of Corporate
Finance 13(4):591628.
Smith, A. 1990. Corporate ownership structure and performance: The case of management buyouts. Journal of
Financial Economics 27:14364.
Zarutskie, R. 2010. The role of top management team human capital in venture capital markets: Evidence from
first-time funds. Journal of Business Venturing 25:15572.
402
Page: 402
368402
Copyright of Review of Financial Studies is the property of Oxford University Press / USA and its content may
not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.