Essays in Private Equity

Johan Cassel
Green Templeton College
University of Oxford

A thesis submitted for the degree of

Doctor of Philosophy

Trinity Term 2020

 To Amanda,
for always being there.
Acknowledgements
At the end of a long, productive, doctoral studies journey, there are many people that have been
instrumental in ensuring that I arrived and which are deserving of my gratitude. The deepest debt
I have incurred during these years are to my advisors, Ludovic Phalippou and Tim Jenkinson. The
methodological rigour they required from my work, and the many hours they have spent guiding me, has
left me deeply humbled and will have a lasting impression on my research work and philosophy going
forward. Enrique Schroth has spent countless hours teaching and challenging me in our joint work, as well
as provided rigorous feedback and support on my solo-authored research. I am grateful for the generosity
with which he has given me time. I also owe much to my other research collaborator and co-author,
Rui Albuquerque, from whom I have learned a great deal. While all faculty at the Saïd Business School
deserves to be thanked for providing feedback to me at some point, I wish to extend a special thanks to
Thomas Hellmann and Pedro Bordalo. They have generously shared their time, giving feedback but also
expanded my research interests.
I acknowledge the research assistance provided by Dan Zaitsev and Kresimir Krasovic, which have
assisted me in the data gathering process. I cannot thank enough the Saïd Business School, Oxford-Man
Institute, Green Templeton College, and the Private Equity Institute for providing the quintessential
doctoral studies requirements: financial support, access to computing infrastructure, and unlimited tea.
I also gratefully acknowledge Vimal Balasubramaniam, who has been a mentor and all-weather friend
to me during these years. I also thank my other partners in crime in this quest at Oxford, Beata Gafka,
Neroli Austin, Alexander Montag, Francesca Brusa, Dunhong Jin, and Ilona Mostipan, who have helped
me grow as a researcher and, more importantly, made Oxford feel as home. I thank Oxford University
Floorball Club for providing me with a welcome break every week.
None of this would have been possible without the lifelong support and encouragement to pursue
further education provided by my family. I am particularly grateful for the support of my parents Margareta
Cassel and Torsten Horneij. Finally, I thank my wife, Amanda Cassel Pegelow, for pushing me to improve,
for keeping me sane, and for always being there when I have needed her. This work is dedicated to her.
Abstract
This dissertation presents three empirical analyses on private equity (PE) investments. The first chapter,
The accuracy of net asset values in private equity: Evidence from the secondary market, examines pricing
in the secondary market for PE fund stakes. The average transaction occurs at a discount, which is
commonly attributed to compensation for liquidity provision. As a quarter of bids come in at a par or
a premium, this cannot be the full explanation. I document that cross-sectional variation in discounts
predicts future fund returns. The second chapter explores The dynamics of pay-for-performance sensitivity
in private equity funds arising from carried interest. As carried interest is conditional on beating a hurdle
rate, pay-for-performance sensitivity is increasing in fund performance, and increasingly so as the fund
matures. I document that within-fund variation in pay-for-performance sensitivity is positively related
to growth in portfolio companies acquired in buyouts. The effect is stronger when the importance of
the direct component is high relative to the indirect component of pay for performance. To rule out
confounding factors, I utilize public market movements to generate exogenous variation in the measure. My
findings provide evidence of the importance of incentives for value creation. The third chapter, Managerial
ownership and operational improvements in buyouts, investigates the relevance of CEO ownership stakes
in companies acquired in buyouts. I find that post-buyout CEO equity ownership stakes are positively
associated with profitability improvements. Consistent with an effect of incentives, the results are stronger
when the ownership stake increases, and are driven entirely by the subsample in which the CEO is
retained. Taken together, these three studies contribute to the growing body of work on private equity
investments and the impact of private equity ownership on companies acquired in buyouts. They open
the door to further theoretical and empirical research in the field.
Contents

1 Introduction 1

References 20

2 Literature Review 25
1 Investing in Private Equity . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2 The Structure of Private Equity Funds . . . . . . . . . . . . . . . . . . . . 31
3 The Economic Impact of Buyouts . . . . . . . . . . . . . . . . . . . . . . . 33

References 40

3 The Accuracy of Net Asset Values in Private Equity: Evidence from

the Secondary Market 46
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2 Institutional Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.1 Evolution of the Secondary Market . . . . . . . . . . . . . . . . . . 52
3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.1 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4 Research Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.1 Determinants of Bid Levels . . . . . . . . . . . . . . . . . . . . . . 57
4.2 Informativeness of Bids . . . . . . . . . . . . . . . . . . . . . . . . 58
4.3 Bids, not Transactions . . . . . . . . . . . . . . . . . . . . . . . . . 60
5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.1 Determinants of Bids . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.2 Returns Post Bidding . . . . . . . . . . . . . . . . . . . . . . . . . 65
6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
A Appendix Figures and Tables . . . . . . . . . . . . . . . . . . . . . 83
B Variable Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . 87

References 90

v
4 The Dynamics of Pay-for-Performance Sensitivity in Private Equity
Funds 92
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
2 Theoretical Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
2.1 Institutional Setting . . . . . . . . . . . . . . . . . . . . . . . . . . 97
2.1.1 Hypotheses development . . . . . . . . . . . . . . . . . . 98
2.2 Estimating Explicit Incentives in Private Equity Funds . . . . . . 100
2.2.1 Modelling Expected Carried Interest . . . . . . . . . . . . 100
2.2.2 Estimating the Pay-for-Performance Sensitivity . . . . . . 103
3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.1 Sample Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.2 Sample Composition . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.3 Accounting Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.4 Fund cash flow data . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.1 Operational Performance and Pay-for-Performance Sensitivity . . . 113
4.1.1 Impact of reputation . . . . . . . . . . . . . . . . . . . . . 117
4.1.2 Deals made close to fundraising events . . . . . . . . . . 119
4.1.3 Public market movements as exogenous shocks . . . . . . 120
4.1.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
A Appendix Figures and Tables . . . . . . . . . . . . . . . . . . . . . 139
B Additional Data Description . . . . . . . . . . . . . . . . . . . . . . 154
B.1 Sample Construction . . . . . . . . . . . . . . . . . . . . 154
B.2 Additional Details on the Accounting Data . . . . . . . . 156
C Details on Monte Carlo Procedure . . . . . . . . . . . . . . . . . . 158
D Variable Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . 163

References 166

5 Managerial Ownership and Operational Improvements in Buyouts 170

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
2 Institutional Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
3.1 Sample Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
3.1.1 Accounting Data . . . . . . . . . . . . . . . . . . . . . . . 179
3.1.2 Managerial Ownership Data . . . . . . . . . . . . . . . . 181
4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
4.1 Cross-Sectional Variation in Ownership . . . . . . . . . . . . . . . 184

vi
 4.1.1 Controlling for industry effects . . . . . . . . . . . . . . . 186
4.2 Interpretation of the Results . . . . . . . . . . . . . . . . . . . . . 188
4.2.1 Changes in ownership stakes . . . . . . . . . . . . . . . . 189
4.2.2 The parallel trends assumption . . . . . . . . . . . . . . . 190
4.2.3 Retained or replaced CEO . . . . . . . . . . . . . . . . . 192
4.3 Sources of Profitability Improvements . . . . . . . . . . . . . . . . 195
4.4 Determinants of Post-Buyout CEO Ownership . . . . . . . . . . . 196
5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
A Appendix Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
B Data Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
B.1 Sample Construction . . . . . . . . . . . . . . . . . . . . 217
B.2 Additional Details on the Accounting Data . . . . . . . . 219
C Comparison PE targets and Matched Peers . . . . . . . . . . . . . 222
D Variables in the data set . . . . . . . . . . . . . . . . . . . . . . . . 225

References 227

6 Conclusion 230

References 237

vii
1 | Introduction

Despite the fact that the global economy and individual investors are increas-
ingly dependent on private capital, much remains poorly understood about
these investments. (Jeng and Lerner (2016))

The private equity industry has experienced a tremendous growth since the 1980s. Private
equity funds raised approximately $0.8 billion in 1980, $11 billion in 1990, $222 billion in
2000, $450 billion in 2007 prior to the onset of the financial crisis, and $860 billion in 2018,
having $4.11 trillion assets under management as of June 2019.1 The industry growth
has been accompanied by increased attention from both academics and policy makers.
Private equity firms act as financial intermediaries, raising private equity funds
with capital provided by investors. These funds make investments directly in portfolio
companies that are privately held. This differentiates investments by private equity firms
from angel investors that invest their own capital.
Private equity encompasses a range of investment strategies, differentiated by the
stage of a portfolio company’s development at which investments are made. Venture
capital (VC) and buyout funds are the most common types of private equity strategies.
VC investments are typically staged over multiple rounds in young, innovative companies
that have significant growth potential but tend to not yet be profitable. The venture
capital investment injects capital in the company, typically in exchange for minority
stakes, which is used to support company growth. By contrast, buyout investments
are often one-off investments, where the buyout fund acquires a controlling stake from
existing shareholders using a mix of equity and debt. Target companies tend to be
at a more mature stage, with some level of profitability to support interest payments.
This thesis focuses on the buyout segment of private equity, and the terms buyout and
private equity are used interchangeably.2
Buyouts became an important phenomenon during the 1980s, and Jensen (1989)
famously predicted that they would eclipse the public market as the dominant corporate
1
The numbers for 1980, 1990, 2000, and 2007 are estimated from Preqin data by summing up fund
sizes of the relevant vintage years in Preqin. The assets under management figure for 2019 are from the
2020 Preqin Global Private Equity and Venture Capital Report.
2
For a summary of research on venture capital, see Gompers and Lerner (2001); Metrick and Yasuda
(2011); Da Rin et al. (2013), and Tykvová (2018)

1
organizational form. In his view, the structure of buyouts alleviated several issues present
in the public equity market. With private equity firms taking large ownership stakes
they have incentives to engage with and monitor the management team, and sufficient
control to implement changes deemed necessary. A buyout typically employs significant
leverage, which act as a disciplinary mechanism to focus management on generating cash
flows (Jensen, 1986). While buyouts have not yet replaced the public equity market as
the dominant ownership form, it has grown in prominence over time.
A private equity fund is typically organized as a limited partnership between a group
of investors, called limited partners (LPs), and a private equity firm that manages the
fund, referred to as the general partner (GP). Investors in private equity funds tend to
be institutional investors such as pension funds, endowments, sovereign wealth funds
and family offices. When a fund is raised, LPs commit capital to the fund but do
not provide it upfront. Instead, the GP has an “investment period” during which the
committed capital should be invested. Once an investment is exited, the resulting cash
is generally not allowed to be recycled into new investments, but is distributed to LPs.
The timing and size of capital calls and distributions are ex ante uncertain, with the
GP provided significant discretion about the timing of investments and the exit. The
GP is constrained only by contractual terms set out in a limited partnership agreement,
stipulating the investment period as well as when the fund should be wounded up; private
equity funds have finite lifespans, usually 10–12 years.3
The compensation to GPs consist of three parts: a management fee, a carried interest
provision which gives GPs a fraction of fund profits, and various fees charged directly to
portfolio companies. The management fee can be viewed as the fixed part of compensation.
It is usually 2%, with the basis being committed capital during the investment period,
and invested capital in the later part of a fund’s life. Carried interest typically gives GPs
20% of fund profits, conditional on fund performance exceeding a hurdle rate (normally
8%). There is more variation in other types of fees, and various rebates offered to
LPs. Gompers and Lerner (1999) and Metrick and Yasuda (2010) provide detailed
descriptions and empirical investigations of common fee arrangements, and how each
part contributes to GP revenue, while Phalippou et al. (2018) focus on fees charged to
portfolio companies. The fee structure with carried interest and a fixed management
fees resembles the fee structure in hedge funds, but there are distinct differences such
as the use of high watermarks in hedge funds.
The purpose of carried interest is to provide GPs with a strong pay-for-performance
component of pay, aligning incentives of LPs and GPs. Chung et al. (2012) note that in

3
See Sahlman (1990) for a discussion of the common organizational structure of private equity funds,
and Metrick and Yasuda (2010) for a recent overview.

2
addition to the direct pay for performance provided by carried interest, the finite fund
life provides a natural indirect component as GPs regularly have to raise new funds in
the market. As documented by Chung et al. (2012), Barber and Yasuda (2017), and
Brown et al. (2019), high current fund performance increases the likelihood that a GP
raises a new fund and, conditional on successful fundraising, the fund is larger. Typically,
new funds are raised between the third and sixth year of the existing fund (Barber and
Yasuda, 2017). This ensures that the private equity firm has capital to deploy when
promising investment opportunities are found.
A major difference between private equity funds and other actively managed fund
structures is the role of the fund manager once an investment is made. While all actively
managed funds are responsible for selecting investments, and are evaluated on their
performance, not all are as actively involved in running the business. Metrick and Yasuda
(2011) view this as the raison d’être of private equity. By taking large, often controlling,
stakes, private equity firms have the means to implement any changes deemed necessary
to generate value in portfolio companies.
As private equity firms are financial intermediaries, generating value should be
interpreted as value for the private equity fund. Private equity is often associated
with negative headlines, commonly due to perceptions that private equity extract value
from other stakeholders.4 A theoretical channel is suggested in Shleifer and Summers
(1988), where the acquirer is able to transfer wealth by breaking implicit contracts.
Much of the criticism towards private equity are based on specific cases rather than
large-scale analysis of the impact of private equity. Analysing large, representative,
samples is difficult due to the lack of readily available data, as discussed in Jeng and
Lerner (2016). Investments made by private equity funds are privately held, and are
consequently often exempted from disclosure requirements facing publicly listed companies.
The private equity firms themselves are under no obligation to make their data available.
Even information that is basic in other areas, such as fund performance data, is typically
only available for subsets of private equity funds. Researchers tend to rely on commercial
databases such as Burgiss, Cambridge Associates, Pitchbook and Preqin.5 An alternative
is to use confidential datasets provided by individual investors (e.g. Robinson and Sensoy,
2013) or government agencies (e.g. Cohn et al., 2014).

4
A recent example of this is a letter sent to Blackstone, KKR, New Mountain Capital and Frazier
Healthcare Partners on the 15th of April 2020 by Elizabeth Warren, the US Democratic senator, and
Katie Porter, the US Democratic congresswoman. In the letter they criticize that portfolio companies of
these PE firm cut doctors’ pay and benefits during the ongoing Covid-19 pandemic. In the letter they
acknowledge that the healthcare industry is experiencing financial difficulties, while still criticizing the
approach taken.
5
Brown et al. (2015) compare these four commercial databases in terms of coverage and what they
say about fund performance. The authors also describe and discuss differences in how the data is sourced.

3
 This thesis combines a variety of databases to advance our knowledge in two areas
of the private equity field. The first paper utilizes a confidential dataset containing bids
on the secondary market for private equity fund stakes to examine whether bids in that
market are informative of the accuracy of net asset values (NAVs). This secondary market
was once marginal due to contractual restrictions on transfers (see Lerner and Schoar,
2004), but took off after the financial crisis. The second and third papers contribute to the
growing literature on the impact of private equity ownership on the portfolio companies
owned. For these papers I assemble a new, representative, dataset with financials on
portfolio companies acquired in buyouts. I do this by utilizing the unique UK setting,
where all companies are required to submit annual filings to Companies House, the
national registrar. The second paper examines the role of direct pay for performance
at the fund level, provided to GPs by the carried interest provision. I document how
this pay-for-performance sensitivity varies over a fund’s life, and establish a positive
relationship between portfolio company growth and the sensitivity of pay to performance
of the acquiring GP. The third paper shifts focus to the role of managerial ownership
stakes in portfolio companies. A salient feature of private equity acquisitions is changes to
the ownership stakes of CEOs in the acquired companies (e.g. Cronqvist and Fahlenbrach,
2013; Kaplan and Stein, 1993). I hand-collect data on managerial ownership stakes prior
to and following the buyout, and document a positive relationship between increases in
CEO ownership stakes and profitability improvements in the acquired company.
Given the distinct nature of each analysis, I review the major findings of each paper
separately, highlighting its relevance and contribution to the literature. A detailed survey
of the existing body of research related to these topics is presented in Chapter 2.

Paper 1: The accuracy of net asset values in private equity: Evidence from
the secondary market
For a long time, one of the distinct characteristics of private equity was the inability
of LPs to liquidate their fund stakes prior to the fund itself being liquidated. In other
fund structures that restrict the ability of investors to redeem invested capital from the
fund, such as closed-end funds and hedge funds, investors are typically able to liquidate
their stakes on a secondary market.6 For a long time, private equity was an exception to
this, commonly employing restrictions on the transferability of fund stakes. For example,
GPs are typically able to block the sale of a fund stake.
As a consequence, the secondary market for private equity fund (PEF) stakes remained
marginal for a long time, having only $2 billion of annual transactions in the early 2000s
(Nadauld et al., 2019). These features make investments in private equity illiquid from
6
Investments in hedge funds can be redeemable, but they often impose lock-up and redemption notice
periods (Ramadorai, 2012).

4
the point of view of LPs. Lerner and Schoar (2004) argue that this illiquidity is an
optimal design feature, chosen by GPs to screen for investors with “deep pockets” and
long investment horizons. However, the financial crisis exposed a need among investors to
be able to liquidate their stakes early, and the secondary market has grown in response,
reaching a high of $88 billion in transaction volume in 2019.7 While there are many
reasons for investors to want to liquidate a stake early, two points illustrate why the
market was bound to take off eventually.
First, restrictions on transferability of fund stakes may have been feasible when private
equity represented small fractions of investors’ portfolio. However, the growth of the
industry have followed an increased allocation to private equity by individual institutions.
The Yale endowment is an extreme example, having an asset allocation target of 38% to
buyout funds and venture capital for the fiscal year of 2020, the highest to date.8
Second, to reach any desired asset allocation, LPs have to account for the fact that
they do not have discretion over the entry and exit decisions of investments. To maintain
any desired allocation, commitments have to be made today to replace expected future
distributions. To achieve an increase in the allocation, investors face a lag between
committing today and achieving the desired increase in invested capital. As distributions
are not guaranteed, and capital calls are uncertain ex ante, the total potential exposure
to private equity is therefore larger than the value of current investments. Robinson
and Sensoy (2016) document that distributions are more cyclical than capital calls,
suggesting that an LP’s exposure is likely increased in bad times. If adverse shocks reduce
the value of the other parts of the portfolio and the availability of liquidity, uncalled
commitments may become a burden on liquidity due to the strict contractual obligations
to meet capital calls (Metrick and Yasuda, 2010).
The Harvard endowment’s exposure during the financial crisis exemplifies this. As of
June 2008, the endowment had $11 billion of uncalled capital commitments to private
equity and real estate funds, compared to their combined reported value of $15.3 billion,
and a total endowment value of $43.0 billion. The endowment shrunk in value during
the next fiscal year, and reported a value of $30.9 billion in June 2009. The uncalled
committed capital therefore represented a much larger fraction of the endowment. The
endowment reduced their uncalled commitments by $3 billion during that year.9 One
way of reducing the amount of committed capital is to sell a fund stake on the secondary
market, as the buyer takes over the responsibility to meet future capital calls.
7
https://www.greenhill.com/en/content/greenhill%E2%80%99s-secondary-market-analysis-record-
volume-third-consecutive-year, accessed 2020-05-18.
8
https://news.yale.edu/2019/09/27/investment-return-57-brings-yale-endowment-value-303-billion, ac-
cessed 2020-05-18.
9
The figures on endowment valuations and uncalled commitments are taken from the financial reports
of the Harvard Endowment for the fiscal years 2009 and 2010.

5
 While we might have expected 2009 to be a year with many secondary market
transactions, transaction volumes dropped to half of what it was in the years before and
after (see Nadauld et al., 2019). One of the reasons for this, as documented by Hege and
Nuti (2011), is that sellers and buyers disagreed on the fair price of fund stakes. Sellers
anchored their valuations around reported NAVs, while buyers argued that valuations
were stale and therefore demanded large discounts.
As the market has grown, substantial capital is now earmarked to purchasing PEF
stakes through the secondary market. For instance, secondary funds, focused on such
transactions, raised over $160 billion between 2011 and 2016.10 Even with this influx of
capital, Nadauld et al. (2019) show that, on average, transactions occur at a discount
to a fund’s NAV.11

A common explanation for the discounts is one of imperfect liquidity provision: when
investors are forced to liquidate their PEF stakes, buyers providing liquidity are able
to acquire assets for less than the underlying value of the asset (Hege and Nuti, 2011;
Kleymenova et al., 2012; Nadauld et al., 2019). This explanation alone cannot account
for all variation in discounts, as we observe a quarter of bids that arrive at par or at
a premium to NAVs. If NAVs reflect the underlying value, it is puzzling that most
of these bids do not lead to a transaction.
An alternative explanation is that variation in discounts reflect expected future
performance, as in the model of closed-end fund discounts by Berk and Stanton (2007).
Expected future performance could reflect beliefs about managerial ability. It could also
reflect a discrepancy between NAVs and fundamental value, owing to the lack of market
valuation of portfolio companies in private equity funds and an aspect of subjectivity
when estimating value (Crain and Law, 2018). Secondary market prices may represent
more accurate estimates of fund value. Boyer et al. (2019) construct a private equity index
based on secondary market transactions, which can be used to examine factor loadings of
private equity returns if secondary market prices reflect the market value of fund stakes.12
This paper asks whether bids are informative of the accuracy of NAVs. The main
findings of the paper is twofold. First, cross-sectional variation in bids predict the future
performance of funds, measured from the first NAV after the bid is posted. Funds receiving
high bids have high post-bid performance, measured as the public market equivalent
(PME) of Kaplan and Schoar (2005). This result does not hold true for variation in
10
Other buyers include funds-of-funds, which invest in both the primary and secondary market for PEF
stakes; and various asset owners (investment banks, hedge funds, endowments, pension funds, sovereign
funds), which have their own teams to manage purchases of PEF stakes.
11
Nadauld et al. (2019) report average discounts ranging from an average of 46% in 2009 to 7% in 2014.
12
The index of Boyer et al. (2019) is constructed to represent the returns of an investor that both buys
and sells through the secondary market, but the extent to which their index is relevant for evaluating
factors relevant for returns in private equity overall depend on what discounts reflect.

6
discounts over time, only when comparing cross-sectional variation in discounts between
funds within a given time period. To assess the economic magnitude, we sort funds into
quartiles based on the highest bid received during a quarter. Over a three-year period
following the bid, funds in the quartile receiving the highest bids report a PME of at least
0.103 above funds receiving the lowest bids. This difference is large when put in relation to
an average PME of 1.18 during the entire life of buyout funds, as reported in Harris et al.
(2016). These results refer to an ex post measure of performance, measured from the first
NAV report after the bid date and three years into the future. Additionally, we find that
bids are positively related to performance of funds during the period the bid is posted:
bids are higher for funds that report a higher NAV-to-NAV return during the quarter in
which the bid is placed. Note that this information is not available at the time of the bid.
Second, we find that the relationship between post-bid fund performance and the
bid placed is sufficiently strong to offset the difference in price paid. Investors posting
low bids would not achieve higher returns than those posting high bids, had their bids
been accepted. The reported gain for those posting low bids is large over short horizons,
while it is offset by the reduced fund performance over longer horizons. Again, this result
only holds within a given time period, not for variation in discounts over time. Together,
these results indicate that investors in the secondary market are able to identify the
relative accuracy of funds’ NAVs. Time-series variation in discounts may reflect varying
liquidity compensation, as suggested by Nadauld et al. (2019).
It may seem as if the results of the paper are obvious. However, the opaque nature
of private equity makes it hard to access any information on the portfolio companies
that make up the value of a fund. While investors in the fund receive regular updates,
potential outside investors do not, complicating any valuation exercise.13 Despite a rapid
growth, the secondary market remains small relative to assets under management of $4.11
trillion (Preqin 2020). In Boyer et al. (2019), the average fund transacts 2.7 times over
the period 2006 through 2018, conditional on trading at all. This market is therefore
characterised by limited price discovery. The results of the paper therefore suggests a
high level of sophistication among participants in the secondary market, echoing findings
in Brown et al. (2019) that investors are not tricked by misreported NAVs.
The secondary market is an over-the-counter market intermediated by specialized
organizations that match buyers and sellers. The data in this study is a proprietary
dataset provided by a London-based sell-side agent for investors wishing to exit via the
secondary market. The data provider gave us access to its entire bid book, which is
13
To highlight the inaccessible nature of the data, it is instructive to consider how our data provider
describes the process of selling a fund stake. After an initial interest is found for a given stake, the general
partner of the fund has to agree to a set of potential buyers. Bidders then sign non-disclosure agreements
and are only allowed access to data in “data rooms” set up to provide access for a limited period of time.

7
comprehensive and representative of the global market: the average bid in our sample each
year is close to the average annual transaction price reported by Greenhill-Cogent. Most
of the bids are made at a discount to NAV, and the discounts vary over time as well as
cross-sectionally. Our dataset consists of 4,242 bids on 488 buyout funds by 144 bidders.
It is important to understand how well the secondary market functions to evaluate
performance in private equity. While buyout funds tend to outperform the public equity
market (Harris et al., 2014, 2016; Robinson and Sensoy, 2016), it is not clear whether this
reflects an alpha or compensation for bearing the illiquidity of private equity investments.
Franzoni et al. (2012) find an alpha of zero when including a liquidity premium, and
the theoretical models in Sorensen et al. (2014) and Bollen and Sensoy (2016) suggest
that the reported performance may be just enough to break even.
A well-functioning secondary market reduces the illiquidity of investments in private
equity, lowering the return required to invest in the primary market. This is shown in
the models of Maurin et al. (2020) and Bollen and Sensoy (2016), where the discount
from the fair value of the fund directly impacts the required return from investing in a
private equity fund. Hence, to understand whether reported returns in private equity
are compensating investors for the risk they take, it is essential to know whether the
secondary market is efficient. This paper suggests that merely looking at the discount
from NAV to gauge the discount from the fair value is simplistic, as cross-sectional
variation reflects differences in underlying value.
This is one of few empirical studies of the secondary market for PEF stakes. Like us,
Kleymenova et al. (2012) utilize bid-level data. They explore the determinants of bids
in auctions of PEF stakes, finding that bids are lower when fewer bidders participate in
an auction, and higher if the stake is more liquid. Nadauld et al. (2019) is the first to
use transaction data. They document an average discount in the market which increases
in bad economic times. They also find that buyers of PEF stakes earn higher return
than sellers, a difference they attribute to compensation for liquidity provision. Using
the same dataset as employed in this paper, Albuquerque et al. (2018) study liquidity
provision in the secondary market by modelling demand for private equity fund stakes.
They find evidence of compensation for liquidity provision, as demand moves to where
selling pressure is highest. This paper complements existing studies by studying the
informativeness of bid levels, providing evidence that cross-sectional variation predicts
fund performance. In doing so it also provides empirical evidence in favour of the model
of Berk and Stanton (2007), from a setting outside the closed-end fund setting it was
initially developed for. In that it contributes to earlier evidence from secondary market
for hedge funds provided by Ramadorai (2012).

8
 The paper also contributes to the literature examining the accuracy of NAVs by
documenting that secondary market participants are able to spot which NAVs are
misvalued, and price it appropriately. Barber and Yasuda (2017) and Brown et al.
(2019) show that NAVs are not always accurately priced, particularly around fundraising
events. Ang et al. (2018) and Boyer et al. (2019) document significant autocorrelation
in private equity return indices, suggesting that NAVs do not fully incorporate recent
returns. To the extent that secondary markets become more liquid, it provides a potential
avenue to overcome the lack of market valuations.
A limitation of our data is that we work with bids and not transactions. While this
enable us to examine valuations of funds that LPs are not willing to sell, a natural concern
is that some bids may not be serious, and even if bids are serious they are lower than the
price which would lead to a transaction. We take several steps to mitigate this concern.
First, we note that our descriptive statistics closely match those published by Greenhill
Cogent and those in Nadauld et al. (2019) that are based on transaction prices from a
US-based financial intermediary. This similarity is not surprising given that i) bidders
submit the same bid to several intermediaries to maximize the probability of matching a
selling interest; and ii) submitting bids is costly both in due diligence and in terms of
reputation. A bidder attempting to manipulate the perception of demand, submitting
unrealistic bids, or reneging on a submitted bid, would be quickly spotted and excluded
from this market by the financial intermediary. Hence, we believe that bids are informative
of the bidder’s true valuation.
Second, we take steps to mitigate concerns arising from bids being systematically
lower than the price a fund would transact at. Let ∆ denote this difference. Evidence in
Hege and Nuti (2011) suggests that this difference varies over time. All of our analysis
therefore include specifications with time fixed effects to soak up this effect. Such time-
variation in ∆ may explain why we find weaker relationship between discounts and future
performance when not controlling for time fixed effects.
Third, while our data is bid-focused it does include a few transactions. For 35 such
transactions we observe earlier bids from the investor that eventually buys the stake. The
final bid placed prior to the completion of a transaction is on average the same price as
the completed transaction. However, we do observe that bids generally increase in the
submitted bid sequence, presumably as the bidder is attempting to convince the seller
by posting bids closer and closer to the bidder’s reservation price. This suggests that
the bid sequence correlates negatively with ∆, and we can test the whether our results
are sensitive to including this proxy from our specification. The results of the paper are
qualitatively unchanged whether we include and exclude this proxy.

9
 Finally, we note that the presence of ∆ is likely to work against us finding the results
of the paper. When we measure the hypothetical return to bidders, we use the bid placed
as an initial investment. An increase in ∆ equals a lower bid, leading to a higher return
to the investor. But what we find is that higher bids do not perform any worse than
lower bids.14 There is no such mechanical relationship between the bid and funds’ future
performance, but the presence of ∆ is likely to add noise, making it more difficult to
find any relationship. We conclude that it is unlikely that our results are driven by the
fact that we work with bids rather than transaction prices.

Paper 2: The Dynamics of Pay-for-Performance Sensitivity in Private Equity

Funds
The carried interest part of a GP’s compensation is intended to provide the GP with
a strong pay-for-performance component of pay. It is credited by some as one of the
drivers of strong returns in PE (e.g. Jensen, 1989; Kaplan and Strömberg, 2009). A
GP’s payoff from carried interest functions similarly to a call option, with the hurdle
rate acting as its strike price. The sensitivity of pay to performance corresponds to
the delta of this option. The hurdle rate is fixed and set at fund inception, leading
the delta to vary with fund performance. As with call options, the delta is low (high)
when a fund is far out of (in) the money.
A PE fund does not invest all the committed capital at once. Instead, it makes
staggered investments during an investment period that normally spans five years.
Consequently, when making investments late in the investment period, the delta depends
on the success of earlier deals. Proponents of PE claim that they add value to their
portfolio companies. Presumably, this takes effort on behalf of the GP, and agency
theory predicts that we expect higher effort when the sensitivity of pay to performance
is higher (e.g. Holmström, 1979). In this paper, I ask how variation in a GP’s pay-for-
performance sensitivity relates to operational improvements at the portfolio companies
acquired in buyouts.
The main finding of the paper is that there is a positive relationship between the growth
rate of companies acquired in buyouts and the delta of the acquiring PE fund. The results
suggest that a one standard deviation increase in delta is associated with between 2.7% and
3.5% higher annual revenue growth, measured from the pre-buyout period and up to four
years post-buyout. Delta is not related to improved profitability. While estimates suggest
a positive relationship, it is statistically insignificant and the economic magnitude is small.

14
If ∆ is an increasing function of the bid placed it would help us find our results. However, this seems
unlikely given that the mechanical effect of an increase in ∆ is to lower the bid. Note also that we may
have several bids targeting a given fund in specific time period, in which case the ones with the highest ∆
are the lowest bid and it is certain to work against us.

10
 Earlier studies document that one way PE facilitates growth in portfolio companies is
through relaxing credit constraints, allowing for increased investment (e.g. Boucly et al.,
2011; Cohn et al., 2020). Interestingly, the observed positive relationship between company
growth and delta does not reflect capital spending. There is not only a lack of significant
relationship between delta and asset growth in the target company, the point estimates
are consistently negative. This suggests that the results likely reflect organic growth.
This study utilizes a dataset of UK companies acquired in a buyout to measure
improvements in operational performance. All companies in the UK, public and private,
are required to submit filings to Companies House, the UK national registrar, which makes
these filings publicly available. This setting provides two distinct advantages for the study
at hand. First, a representative dataset containing both pre- and post-buyout financials
can be assembled.15 Second, UK companies submit annual ownership lists, allowing a
company to be linked to the acquiring fund. This is crucial as the delta is estimated at
the fund level, while transaction databases commonly only list the acquiring PE firm.
To estimate the delta of GP pay I build on the framework developed by Metrick
and Yasuda (2010). They provide a model for estimating the expected revenue to a GP,
of which carried interest is one part. A PE fund consists of investments in portfolio
companies with simulated return paths. Using a set of contractual features, the resulting
stream of cash flows is allocated to the fund’s investors (limited partners, or LPs) and
the GP. In equilibrium, GPs must add value to portfolio companies to compensate LPs
for the fees charged, and hence I calibrate the model so that LPs break even. The
delta of a GP’s compensation is measured as the expected increase in the present value
of a GP’s payoff from adding $1 extra of value to a portfolio company. To match an
additional moment of the data I add leverage to their framework, and calibrate the model
to match the empirical default rate in the sample.
To evaluate the dynamics of delta the model is estimated at intermediate points in
funds’ life, using reported cash flows and NAV valuations from Preqin. I find that the
delta is positively related to fund performance, and increasingly so as the fund matures
and performance becomes more informative.
The delta captures only the direct pay for performance sensitivity. As documented
by Chung et al. (2012), the indirect pay for performance — stemming from improved
fundraising prospects — is of at least an equal magnitude to the direct component.

15
Availability of data on privately held companies is often scarce. Consequently, researchers often
utilizes samples containing companies with publicly available financials (e.g. Guo et al., 2011), or samples
made available to them by individual investors (e.g. Acharya et al., 2013). Such samples are not necessarily
representative. Cohn et al. (2014) document that among public-to-private buyouts, those with publicly
available financials perform better than those without. I am not the first to utilize the UK setting for
these reasons. Bernstein et al. (2019) examine how PE-backed companies fared during the financial crisis,
while Chung (2011) focuses on the impact of PE ownership in private-to-private buyouts.

11
They show that the relative importance of the direct component increases with GP
experience: GPs with a proven track record are less reliant on current fund performance
to successfully raise new funds. Their insight provides a cross-sectional prediction: the
observed relationship between delta and operational improvements at target companies
should be stronger when the direct component of pay is more important to a GP. Consistent
with this prediction, the documented positive relationship is driven by highly reputable
funds, with estimates suggesting a larger economic effect. A similar argument can be made
for deals completed around successful fundraising events. If investors care mostly about
the performance of the most recent fund when deciding to participate in future funds, the
indirect pay for performance will drop following successful fundraising. Consistent with
this prediction, the positive relationship between delta and operational improvements
is stronger for deals completed near fundraising events.
So far, I have documented a positive correlation between a GP’s pay-for-performance
sensitivity and post-buyout growth in portfolio companies. This may reflect an effect of
incentives provided by the pay-for-performance compensation. However, factors related to
the delta may be directly related to the operational improvements at portfolio companies.
For example, an obvious alternative is that the results reflect managerial ability: GPs
skilled at improving operations of portfolio companies likely translate that ability into
high fund performance, leading to a high delta. Additionally, there is a time-varying
mechanical relationship as PE performance closely follow public markets (e.g. Ang et
al., 2018): virtually all funds are in the money when equity markets are doing well.
To the extent that expected operational improvements are state-dependent, this leads
to a correlation unrelated to incentives.
To overcome these issues, the empirical specification includes fund fixed effects and
time fixed effects. The inclusion of fund fixed effects absorbs time-invariant differences in
GP ability, and implies that the results of the paper correspond to within-fund estimates.
The staggered nature of PE investments provides for natural within-fund comparisons.
With the amount of investment capital available to a PE fund fixed at inception, there
is no concern that capital flows, and diseconomies of scale, offset increases in delta due
to good performance as in Berk and Green (2004).
To further strengthen the interpretation I exploit that the valuation of PE funds
move with the public market (Ang et al., 2018), and that the delta increases with fund
performance. A public market valuation shock is exogenous and common to all PE funds.
While the shock is common, the resulting impact on delta varies between funds. The
degree to which a market shock impacts a fund’s delta largely depends on two factors:
the fund’s current performance and the degree to which the fund is invested. To illustrate,
the value of a fund that has yet to make its first investment is not impacted at all by

12
market movements, while one that is fully invested will see the value of all its investments
impacted. Similarly, a fund far in or out of the money will not be much impacted, while
one at the money will experience a big impact. This implies that market shocks can
be used to generate exogenous variation in delta.16
Extracting the part of delta due to a market shock is done in two steps. First, I
estimate what the NAV would have been absent any market movement, using a simple
market model. Second, using this counterfactual NAV and the implied counterfactual
fund performance, I estimate a delta absent the market shock. The difference between
the normal delta and the counterfactual delta is attributable to the market movement.
I confirm the main result of the paper using only the part of delta attributable to market
movements. The results suggest that a one standard deviation increase in the market-
driven component of delta translates to an increased revenue growth rate of between 1.8%
and 3.9%, depending on the model specification. Again, this growth is neither attributable
to increased capital expenditure nor does it come at the expense of worsened profitability.
The paper provides insights to the wider question of how pay-for-performance compen-
sation of managers impacts performance. This question has received considerable attention
to in the context of CEOs, but is notoriously difficult to give a convincing answer to.
Compensation packages are the outcome of a matching and bargaining process, so manager
quality is likely related to both the magnitude and structure of compensation. Managers
may influence the process in which compensation is set (Bebchuk et al., 2002) or time the
delivery of news to the awarding of stock options (Aboody and Kasznik, 2000; Yermack,
1997); compensation packages can be renegotiated ex post (Brenner et al., 2000); and
firms design compensation packages to exploit managerial characteristics such as optimism
and overconfidence (Humphery-Jenner et al., 2016; Otto, 2014).17 The private equity
setting overcomes many of these issues as compensation terms are fixed at fund inception.
This is similar to Agarwal et al. (2009), that examine how the managerial delta in hedge
funds relates to fund performance. The PE setting provides additional benefits as it is a
natural setting to perform within-fund estimations allowing the econometrician to control
for time-invariant managerial ability. As PE funds are always long the equity market,
unlike hedge funds, market shocks can be utilized to generate exogenous variation in delta.

16
This strategy is conceptually similar to that used in Shue and Townsend (2017), where they utilize
differences in the sensitivity to industry shocks of the value of options awarded to CEOs under fixed-value
versus fixed-number option plans.
17
Frydman and Jenter (2010) provides a survey of this literature. Notable papers on the efficacy of
compensation packages are Jensen and Murphy (1990), showing that CEO compensation is weakly related
to the dollar increase in firm value; Hall and Liebman (1998), showing that CEOs compensation is strongly
related to the percentage increase in firm value; Yermack (1995), who does not find that stock options
appear to be awarded in order to reduce agency costs; and Datta et al. (2001), showing that stronger
incentives lead to better acquisitions.

13
 The results of the paper are important for the private equity literature. Carried
interest is intended to align the incentives of managers with that of investors, but we
lack empirical evidence on the effect of variation in the sensitivity of pay to performance.
Hüther et al. (2020) document that VC partnerships with more “GP-friendly” contracts,
with carry on a deal-by-deal basis, are associated with higher returns. Robinson and
Sensoy (2013) find that funds with higher fees do not deliver lower net-of-fee performance.
However, these findings are subject to many of the same endogeneity concerns as the
CEO compensation literature, in particular the matching of contract terms and manager
quality. In that respect, this study provides additional insights by exploiting within-fund
variation in the strength of incentives as well as plausibly exogenous variation.
By examining how the pay-for-performance sensitivity of GPs relate to operational
improvements at target companies, the paper contributes to an extensive and growing
literature exploring the impact of private equity ownership on the performance of
portfolio companies (e.g. Acharya et al., 2013; Bernstein et al., 2019; Bernstein and
Sheen, 2016; Boucly et al., 2011; Cohn et al., 2014; Davis et al., 2014, 2019; Guo et
al., 2011; Kaplan, 1989).

Paper 3: Managerial Ownership and Operational Improvements in Buyouts

Given the large role that private equity plays in the economy, it is important to
understand if and how private equity impacts the performance of portfolio companies.
Although the existing literature suggests that private equity generates improvements
to portfolio companies’ operations, we know little of which actions taken generate
these improvements.18
Two recent papers give us some insight into what GPs do. Biesinger et al. (2020)
examine confidential value creation plans for specific emerging market deals, relating these
plans to subsequent success in the deals. Interestingly, they do not find that the ex ante
selection of strategy predicts deal success. Instead, they find that successful deals are
those in which GPs disclose that their proposed strategy was successfully implemented,
which the authors interpret as evidence of the importance of execution of these plans. A
complementary paper is that of Gompers et al. (2016), asking 79 private equity firms
how they operate. Their survey reveals that private equity firms themselves believe that
the two most important sources of value creation, both in expectation and ex post, are
increasing revenue and improving incentives of the management team.
Starting with Jensen and Murphy (1990), there has been a long debate over whether the
equity ownership of CEOs provides sufficient incentives to generate value for shareholders.
A salient feature of private equity acquisitions is that managerial equity ownership increases
18
Examples of studies documenting operational improvements are Acharya et al. (2013); Bernstein and
Sheen (2016); Boucly et al. (2011); Davis et al. (2014, 2019); Guo et al. (2011); and Kaplan (1989).

14
(e.g. Kaplan and Stein, 1993; Cronqvist and Fahlenbrach, 2013). Increasing the equity
ownership stake increases the pay-for-performance sensitivity of the management team,
providing stronger incentives to generate value. This paper examines how the change in
managerial ownership induced by a buyout is related to portfolio company performance.
This paper utilizes the dataset of portfolio company financials assembled for the
second paper of the thesis. Adding to the company-level financials, I utilize an additional
filing required to be submitted annually in the UK: a breakdown of the precise ownership
of every individual shareholder in the company. I compile these ownership filings by
hand to measure managerial ownership before and after the buyout. This data is used
to provide new insights into the impact of managerial ownership on the performance
of companies acquired in a buyout.
The main finding of the paper is that the post-buyout CEO ownership share is
positively correlated with improvements in profitability (EBITDA/Revenue). The results
suggest that companies with a 10 percentage points higher CEO equity ownership stake
(one standard deviation) improve their profitability margin by 3.1%–3.6% from the
unconditional average profitability. The relationship is more pronounced for changes in
CEO ownership stakes, where a 10 percentage points increase in the ownership translates
to a 4.3%–5.0% increase in profitability.
Clearly, the interpretation of any observed relationship between managerial ownership
and operational performance depends on whether the CEO is retained or replaced. Cornelli
and Karakas (2015) document that the CEO of a target company is often replaced by
the acquiring private equity firm, which is the case in 39% of the buyouts in my sample.
Interestingly, the positive relationship between managerial ownership and profitability
improvements found in this study is driven entirely by the subsample of buyouts where
the CEO is retained. By contrast, the coefficient is statistically insignificant and close
to zero when the CEO is replaced. The economic magnitude is up to 60% higher when
the CEO is retained compared to the full sample.
The positive relationship between operational performance and increased managerial
equity ownership may reflect matching of talented CEOs to companies as in Gabaix and
Landier (2008). Higher ability CEOs should be able to bargain for a bigger share, as
well as accept compensation packages that are more dependent on performance. The
results could also correspond to the outcome of efficient bargaining from managers
with inside information (Blanchflower et al., 1996). Additionally, they may capture
an effect of incentives provided through the ownership stake. The arrival of private
equity almost invariably induces a shock to the managerial ownership share. To the
extent managers have limited say over the post-buyout share, the change in ownership
level represents a shock to CEO incentives.

15
 The different sample splits help differentiating between these potential explanations. If
the positive relationship between managerial ownership and performance reflects attracting
talented CEOs, we would expect the results to be stronger when the CEO is replaced as
the pool of potential CEOs is larger. Instead, the positive relationship is only observed
when the CEO is retained, which casts doubt on this explanation. When a CEO is
retained, a change in the ownership stake implies a change in incentives, while CEO ability
is held constant. Furthermore, the incentive story suggests that changes in ownership are
more relevant than the level of ownership, which is what I find. In contrast, the bargaining
story suggests that the relevant metric is the ownership share of the management team.
To further sharpen the interpretation of the results I examine the determinants of
CEO equity ownership post-buyout. These results are largely consistent with a desire to
provide incentives. There is a negative relationship between ownership stake and initial
profitability. Taken together with the main results of the paper, this is consistent with
private equity firms setting contracts to maximize value. Providing management with
higher equity stakes comes at a cost as it limits the upside of the acquirer. Therefore,
we expect higher management stakes in companies where profitability improvements are
particularly valuable: those with low initial profitability. Furthermore, the pre-buyout
ownership stake is positively related to the post-buyout stake, but only when the CEO is
retained. This is consistent with an attempt to increase incentives from pre-buyout levels.
Taken together, the results of the paper support the view that improving managerial
incentives plays an important part in value creation by private equity firms.
A natural question to ask is how profitability improvements come about. Improved
profitability may reflect reduced costs or increased revenues with the existing cost base.
The results of the paper indicate that the profitability improvement related to increases
in managerial ownership stems from reducing costs rather than increasing productivity.
This could indicate that private equity firms provide CEOs with larger stakes when cost
cutting is desired. If cost cutting is associated with a high private cost to the CEO,
this is consistent with an optimal contracting view.
Any study on private equity portfolio companies has to deal with the fact that
companies are not randomly assigned to be buyout targets. Thus, simply comparing
buyout targets with non-targeted companies is tricky. Therefore the analysis relies on
cross-sectional variation in managerial ownership among buyout targets, rather than
comparing buyout targets and other companies. Following Boucly et al. (2011), the
analysis includes matched control companies to account for industry-time fixed effects,
but the variation in managerial ownership is only among the targets of a buyout.
However, even among buyout targets there is a concern that the optimal compensation
package varies with unobserved company characteristics. But this explanation cannot

16
explain why it is only in the retained CEO subsample that we see positive relationships
between managerial ownership and operational performance, and between pre- and
post-buyout CEO ownership stakes. Presumably, a high pre-buyout ownership stake
would indicate a higher CEO ownership stake in the optimal contract whether the
CEO is retained or not.
Finally, distinguishing between the efficient bargaining and the incentive story is
problematic as they often give similar predictions. For example, imagine a manager
of a division in a conglomerate. Suppose she has private knowledge of an unpleasant
action that could improve the division, but does not feel she is currently sufficiently
compensated for the cost of taking the action. She may approach a private equity firm
with the intention of increasing her ownership stake prior to implementing the necessary
improvements. The larger stake provides incentives to take action, but presumably the
manager’s inside information allows her to bargain for a larger stake. Absent exogenous
shocks to the ownership stake provided to CEOs, this paper cannot claim causality,
merely offering robust supportive evidence of the idea that higher CEO incentives lead
to improved operational performance.
This paper speaks directly to the literature on operational improvements provided
by private equity. Early studies by Kaplan (1989), Lichtenberg and Siegel (1990), and
Smith (1990), showed that management buyouts are associated with improved operations,
and that this generates value for the private equity sponsors. More recently, Guo et al.
(2011), and Acharya et al. (2013) document the importance of operational improvements
for generating value in buyouts. Boucly et al. (2011), using a sample of French portfolio
companies, suggest one important channel through which private equity can help their
portfolio companies: relaxing credit constraints to facilitate growth. Davis et al. (2014) and
Davis et al. (2019) provide evidence that private equity facilitates reallocation of capital to
more productive establishments, generating an overall increase in total factor productivity.
This paper adds to this literature by documenting how a specific change induced by a
buyout — changes in managerial ownership — relates to operational improvements.
As discussed in Jeng and Lerner (2016), data limitations is always an issue in private
equity research, and certainly when studying privately held portfolio companies with
limited disclosure requirements. Cohn et al. (2014) directly show how this may impact the
interpretation of results, as companies with publicly available financials perform better
than those without. Davis et al. (2019) highlight that not all buyouts are the same, with
public-to-private buyouts differing markedly from other types of buyouts. This speaks
to one of the strengths of this study, as the UK setting allows the inclusion of all types
of buyouts, and all companies are required to submit annual reports.

17
 The UK setting also allows for collecting a large-scale sample of managerial ownership
both pre- and post-buyout, and linking it to operational performance. This is a significant
improvement over the existing large-scale study focusing on managerial ownership of
private equity targets by Leslie and Oyer (2009).19 They utilize a set of public-to-private
buyouts to obtain pre-buyout ownership data, while using a different set of companies
going public through an IPO to measure managerial ownership levels after a period of
private equity ownership. Since these sets are different, the authors are unable to measure
changes induced at the time of the buyout. Additionally, the authors lack company
financials during the period of private equity ownership.
Beginning with Kaplan and Strömberg (2003), private equity and venture capital
have been used to explore optimal governance, examining the choices taken by value-
maximizing principals. Cornelli and Karakas (2015) study the role of boards and find
that CEO turnover decreases and becomes less contingent on performance when private
equity sponsors have higher involvement in the company. Edgerton (2012) shows that
private equity sponsors curb excessive usage of corporate jets. Cronqvist and Fahlenbrach
(2013) provide a detailed overview how CEO contracts are changed for a set of 20 private
equity deals. This paper contributes to this stream of papers by linking how changes
in managerial ownership is related to operational performance.
The question of how managerial ownership impacts the value of a company is an
important question in the agency literature. Agency theory suggests that if managers
can exert costly effort to increase performance, then we expect better performance
when the pay of the manager is more tightly linked to performance.20 Existing studies
on managerial ownership in publicly held companies have to overcome multiple issues:
compensation packages are the outcome of a matching and bargaining process; managers
may influence the process in which compensation is set (Bebchuk et al., 2002) or time
the delivery of news to the awarding of stock options (Yermack, 1997; Aboody and
Kasznik, 2000); compensation packages can be renegotiated ex post (Brenner et al.,
2000); and companies design compensation packages to exploit managerial characteristics
such as optimism and overconfidence (Otto, 2014; Humphery-Jenner et al., 2016). To

19
Cronqvist and Fahlenbrach (2013) also examines CEOs of companies acquired in public-to-private
buyouts. They obtain detailed information about the contract offered to CEOs in 20 large deals, offering
insights into optimal contracting. While this paper lacks that level of detail, it covers more transactions
and combines the ownership data with operational performance data.
20
Jensen and Meckling (1976), Fama (1980), Fama and Jensen (1983b,a) are examples of this theoretical
agency literature. Holmström (1979) and Grossman and Hart (1983) provide models of optimal contracts
when effort is unobservable and impacts contractable output with noise. There is some empirical evidence
that company value is positively impacted from increased managerial ownership and when compensation
is more sensitive to company performance (e.g. Abowd, 1990; McConnell and Servaes, 1990; Mehran,
1995). However, the evidence is mixed. Agrawal and Knoeber (1996) and Demsetz and Villalonga (2001)
do not find evidence of a relationship between managerial ownership and company performance.

18
the extent that a buyout shocks the ownership level of managers, this study provides
evidence of the importance of managerial ownership.
The remainder of the thesis is organised as follows. Chapter 2 presents an ex-
tensive literature review spanning the three topics above. Chapter 3, Chapter 4 and
Chapter 5 correspond to my first paper (titled The accuracy of net asset values in
private equity: Evidence from the secondary market), second paper (titled The Dynamics
of Pay-for-Performance Sensitivity in Private Equity Funds), and third paper (titled
Managerial Ownership and Operational Improvements in Buyouts), respectively. Chapter
6 concludes, highlighting the general implications of this three-paper dissertation and
directions for future work.

19
References

Aboody, David and Ron Kasznik, “CEO stock option awards and the timing of
corporate voluntary disclosures,” Journal of Accounting and Economics, 2000, 29 (1),
73 – 100.
Abowd, John M., “Does Performance-Based Managerial Compensation Affect Corporate
Performance?,” ILR Review, 1990, 43 (3), 52–S–73–S.
Acharya, Viral V., Oliver F. Gottschalg, Moritz Hahn, and Conor Kehoe,
“Corporate Governance and Value Creation: Evidence from Private Equity,” The Review
of Financial Studies, 2013, 26 (2), 368–402.
Agarwal, Vikas, Naveen D. Daniel, and Narayan Y. Naik, “Role of Managerial
Incentives and Discretion in Hedge Fund Performance,” The Journal of Finance, 2009,
64 (5), 2221–2256.
Agrawal, Anup and Charles R. Knoeber, “Firm Performance and Mechanisms
to Control Agency Problems between Managers and Shareholders,” The Journal of
Financial and Quantitative Analysis, 1996, 31 (3), 377–397.
Albuquerque, Rui, Johan Cassel, Ludovic Phalippou, and Enrique Schroth,
“Liquidity Provision in the Secondary Market for Private Equity Fund Stakes,” 2018.
Working Paper.
Ang, Andrew, Bingxu Chen, William N. Goetzmann, and Ludovic Phalippou,
“Estimating Private Equity Returns from Limited Partner Cash Flows,” The Journal
of Finance, 2018, 73 (4), 1751–1783.
Barber, Brad M and Ayako Yasuda, “Interim Fund Performance and Fundraising in
Private Equity,” Journal of Financial Economics, 2017, 124 (1), 172–194.
Bebchuk, Lucian Arye, Jesse M. Fried, and David I. Walker, “Managerial Power
and Rent Extraction in the Design of Executive Compensation,” The University of
Chicago Law Review, 2002, 69 (3), 751–846.
Berk, Jonathan B. and Richard C. Green, “Mutual Fund Flows and Performance
in Rational Markets,” Journal of Political Economy, 2004, 112 (6), 1269–1295.
Berk, Jonathan B and Richard Stanton, “Managerial Ability, Compensation, and
the Closed-End Fund Discount,” Journal of Finance, 2007, 62 (2), 529–556.
Bernstein, Shai and Albert Sheen, “The Operational Consequences of Private Equity
Buyouts: Evidence from the Restaurant Industry,” The Review of Financial Studies,
2016, 29 (9), 2387–2418.
, Josh Lerner, and Filippo Mezzanotti, “Private Equity and Financial Fragility
during the Crisis,” The Review of Financial Studies, 2019, 32 (4), 1309–1373.
Biesinger, Markus, Çağatay Bircan, and Alexander Ljungqvist, “Value Creation
in Private Equity,” 2020. Working Paper.
Blanchflower, David G., Andrew J. Oswald, and Peter Sanfey, “Wages, Profits,
and Rent-Sharing,” The Quarterly Journal of Economics, 1996, 111 (1), 227–251.

20
Bollen, Nicolas P B and Berk A Sensoy, “How much for a haircut? Illiquidity,
secondary markets, and the value of private equity,” 2016. Fisher College of Business
Working Paper.
Boucly, Quentin, David Sraer, and David Thesmar, “Growth LBOs,” Journal of
Financial Economics, 2011, 102 (2), 432 – 453.
Boyer, Brian, Taylor D. Nadauld, Keith Vorkink, and Michael S. Weisbach,
“Private Equity Indices Based on Secondary Market Transactions,” 2019. Dice Center
WP 2018-21.
Brenner, Menachem, Rangarajan K Sundaram, and David Yermack, “Altering
the terms of executive stock options,” Journal of Financial Economics, 2000, 57 (1),
103 – 128.
Brown, Gregory W., Oleg R. Gredil, and Steven N. Kaplan, “Do private equity
funds manipulate reported returns?,” Journal of Financial Economics, 2019, 132 (2),
267 – 297.
, Robert S. Harris, Tim Jenkinson, Steven N. Kaplan, and David Robinson,
“What Do Different Commercial Data Sets Tell Us About Private Equity Performance?,”
2015. Working paper.
Chung, Ji-Woong, “Leveraged Buyouts of Private Companies,” 2011. Working paper.
, Berk A. Sensoy, Léa Stern, and Michael S. Weisbach, “Pay for Performance
from Future Fund Flows: The Case of Private Equity,” The Review of Financial Studies,
2012, 25 (11), 3259–3304.
Cohn, Jonathan B., Edith S. Hotchkiss, and Erin M. Towery, “The Motives for
Private Equity Buyouts of Private Firms: Evidence from U.S. Corporate Tax Returns,”
2020. Working Paper.
, Lillian F. Mills, and Erin M. Towery, “The evolution of capital structure and
operating performance after leveraged buyouts: Evidence from U.S. corporate tax
returns,” Journal of Financial Economics, 2014, 111 (2), 469 – 494.
Cornelli, Francesca and Oğuzhan Karakas, “CEO Turnover in LBOs: The Role of
Boards,” 2015. Working paper.
Crain, Nicholas G and Kelvin K F Law, “The Bright Side of Fair Value Accounting:
Evidence from Private Company Valuation,” 2018. Working Paper.
Cronqvist, Henrik and Rüdiger Fahlenbrach, “CEO contract design: How do strong
principals do it?,” Journal of Financial Economics, 2013, 108 (3), 659 – 674.
Da Rin, Marco, Thomas Hellmann, and Manju Puri, “A Survey of Venture
Capital Research,” in George M. Constantinides, Milton Harris, and Rene M. Stulz,
eds., Handbook of the Economics of Finance, 2nd ed., Elsevier, 2013, pp. 573 – 648.
Datta, Sudip, Mai Iskandar-Datta, and Kartik Raman, “Executive Compensation
and Corporate Acquisition Decisions,” The Journal of Finance, 2001, 56 (6), 2299–2336.
Davis, Steven J., John Haltiwanger, Kyle Handley, Ben Lipsius, Josh Lerner,
and Javier Miranda, “The Economic Effects of Private Equity Buyouts,” 2019.
NBER Working Paper 26371.
, , , Ron Jarmin, Josh Lerner, and Javier Miranda, “Private Equity, Jobs,
and Productivity,” American Economic Review, 2014, 104 (12), 3956–90.
Demsetz, Harold and Belén Villalonga, “Ownership structure and corporate
performance,” Journal of Corporate Finance, 2001, 7 (3), 209 – 233.
Edgerton, Jesse, “Agency Problems in Public Firms: Evidence from Corporate Jets in
Leveraged Buyouts,” The Journal of Finance, 2012, 67 (6), 2187–2213.

21
Fama, Eugene F., “Agency Problems and the Theory of the Firm,” Journal of Political
Economy, 1980, 88 (2), 288–307.
and Michael C. Jensen, “Agency Problems and Residual Claims,” The Journal of
Law and Economics, 1983, 26 (2), 327–349.
and , “Separation of Ownership and Control,” The Journal of Law and Economics,
1983, 26 (2), 301–325.
Franzoni, Francesco, Eric Nowak, and Ludovic Phalippou, “Private Equity
Performance and Liquidity Risk,” Journal of Finance, 2012, 67 (6), 2341–2373.
Frydman, Carola and Dirk Jenter, “CEO Compensation,” Annual Review of Financial
Economics, 2010, 2 (1), 75–102.
Gabaix, Xavier and Augustin Landier, “Why has CEO Pay Increased So Much?,”
The Quarterly Journal of Economics, 2008, 123 (1), 49–100.
Gompers, Paul and Josh Lerner, “An analysis of compensation in the U.S. venture
capital partnership,” Journal of Financial Economics, 1999, 51, 3–44.
and , “The Venture Capital Revolution,” Journal of Economic Perspectives, 2001,
15 (2), 145–168.
, Steven N. Kaplan, and Vladimir Mukharlyamov, “What do private equity
firms say they do?,” Journal of Financial Economics, 2016, 121 (3), 449 – 476.
Grossman, Sanford J. and Oliver D. Hart, “An Analysis of the Principal-Agent
Problem,” Econometrica, 1983, 51 (1), 7–45.
Guo, Shourun, Edith S. Hotchkiss, and Weihong Song, “Do Buyouts (Still) Create
Value?,” The Journal of Finance, 2011, 66 (2), 479–517.
Hall, Brian J. and Jeffrey B. Liebman, “Are CEOs Really Paid Like Bureaucrats?*,”
The Quarterly Journal of Economics, 1998, 113 (3), 653–691.
Harris, Robert S., Tim Jenkinson, and Steven N. Kaplan, “Private equity
performance: What do we know?,” Journal of Finance, 2014, 69 (5), 1851–1882.
Harris, Robert S, Tim Jenkinson, and Steven N Kaplan, “How Do Private Equity
Investments Perform Compared to Public Equity?,” Journal of Investment Management,
2016, 14 (1), 1–24.
Hege, Ulrich and Alessandro Nuti, “The Private Equity Secondaries Market During
the Financial Crisis and the ”Valuation Gap”,” Journal of Private Equity, 2011, 14 (3),
42–54.
Holmström, Bengt, “Moral Hazard and Observability,” The Bell Journal of Economics,
1979, 10 (1), 74–91.
Humphery-Jenner, Mark, Ling Lei Lisic, Vikram Nanda, and Sabatino Dino
Silveri, “Executive overconfidence and compensation structure,” Journal of Financial
Economics, 2016, 119 (3), 533 – 558.
Hüther, Niklas, David T. Robinson, Sönke Sievers, and Thomas Hartmann-
Wendels, “Paying for Performance in Private Equity: Evidence from Venture Capital
Partnerships,” Management Science, 2020, 66 (4), 1756–1782.
Jeng, Leslie and Josh Lerner, “Making Private Data Accessible in an Opaque Industry:
The Experience of the Private Capital Research Institute,” American Economic Review,
2016, 106 (5), 157–160.
Jensen, Michael, “Eclipse of the Public Corporation,” Harvard Business Review, 1989,
67 (5), 61–74.
Jensen, Michael C., “Agency Costs of Free Cash Flow, Corporate Finance, and
Takeovers,” The American Economic Review, 1986, 76 (2), 323–329.

22
 and Kevin J. Murphy, “Performance Pay and Top-Management Incentives,” Journal
of Political Economy, 1990, 98 (2), 225–264.
and William H. Meckling, “Theory of the firm: Managerial behavior, agency costs
and ownership structure,” Journal of Financial Economics, 1976, 3 (4), 305 – 360.
Kaplan, Steven, “The effects of management buyouts on operating performance and
value,” Journal of Financial Economics, 1989, 24 (2), 217 – 254.
Kaplan, Steven N. and Antoinette Schoar, “Private Equity Performance: Returns,
Persistence, and Capital Flows,” The Journal of Finance, 2005, 60 (4), 1791–1824.
and Jeremy C. Stein, “The Evolution of Buyout Pricing and Financial Structure in
the 1980s*,” The Quarterly Journal of Economics, 1993, 108 (2), 313–357.
and Per Strömberg, “Financial Contracting Theory Meets the Real World: An
Empirical Analysis of Venture Capital Contracts,” The Review of Economic Studies,
2003, 70 (2), 281–315.
and Per Strömberg, “Leveraged Buyouts and Private Equity,” Journal of Economic
Perspectives, 2009, 23 (1), 121–146.
Kleymenova, Anya, Eli Talmor, and Florin P Vasvari, “Liquidity in the Secondaries
Private Equity Market,” 2012. Working paper, London Business School.
Lerner, Josh and Antoinette Schoar, “The illiquidity puzzle: theory and evidence
from private equity,” Journal of Financial Economics, 2004, 72 (1), 3–40.
Leslie, Phillip and Paul Oyer, “Managerial incentives and value creation: Evidence
from private equity,” 2009. Working paper.
Lichtenberg, Frank R. and Donald Siegel, “The effects of leveraged buyouts on
productivity and related aspects of firm behavior,” Journal of Financial Economics,
1990, 27 (1), 165 – 194.
Maurin, Vincent, David T. Robinson, and Per Strömberg, “A Theory of Liquidity
in Private Equity,” 2020. Working paper.
McConnell, John J. and Henri Servaes, “Additional evidence on equity ownership
and corporate value,” Journal of Financial Economics, 1990, 27 (2), 595 – 612.
Mehran, Hamid, “Executive compensation structure, ownership, and firm performance,”
Journal of Financial Economics, 1995, 38 (2), 163 – 184.
Metrick, Andrew and Ayako Yasuda, “The Economics of Private Equity Funds,”
Review of Financial Studies, 2010, 23 (6), 2303–2341.
and , “Venture Capital and Other Private Equity: a Survey,” European Financial
Management, 2011, 17 (4), 619–654.
Nadauld, Taylor D., Berk A. Sensoy, Keith Vorkink, and Michael S. Weisbach,
“The liquidity cost of private equity investments: Evidence from secondary market
transactions,” Journal of Financial Economics, 2019, 132 (3), 158 – 181.
Otto, Clemens A., “CEO optimism and incentive compensation,” Journal of Financial
Economics, 2014, 114 (2), 366 – 404.
Phalippou, Ludovic, Christian Rauch, and Marc Umber, “Private equity portfolio
company fees,” Journal of Financial Economics, 2018, 129 (3), 559 – 585.
Ramadorai, Tarun, “The Secondary Market for Hedge Funds and the Closed Hedge
Fund Premium.,” Journal of Finance, 2012, 67 (2), 479–512.
Robinson, David T. and Berk A. Sensoy, “Do Private Equity Fund Managers Earn
Their Fees? Compensation, Ownership, and Cash Flow Performance,” The Review of
Financial Studies, 2013, 26 (11), 2760–2797.
and , “Cyclicality, performance measurement, and cash flow liquidity in private
equity,” Journal of Financial Economics, 2016, 122 (3), 521–543.

23
Sahlman, William A, “The structure and governance of venture-capital organizations.,”
Journal of Financial Economics, 1990, 27 (2), 473–521.
Shleifer, Andrei and Lawrence H Summers, “Breach of trust in hostile takeovers,”
in Alan J Auerbach, ed., Corporate Takeovers: Causes and Consequences, University of
Chicago Press, 1988, pp. 33 – 56.
Shue, Kelly and Richard R. Townsend, “How Do Quasi-Random Option Grants
Affect CEO Risk-Taking?,” The Journal of Finance, 2017, 72 (6), 2551–2588.
Smith, Abbie J., “Corporate ownership structure and performance: The case of
management buyouts,” Journal of Financial Economics, 1990, 27 (1), 143 – 164.
Sorensen, Morten, Neng Wang, and Jinqiang Yang, “Valuing private equity,”
Review of Financial Studies, 2014, 27 (7), 1977–2021.
Tykvová, Tereza, “Venture capital and private equity financing: an overview of recent
literature and an agenda for future research,” Journal of Business Economics, 2018, 88
(3), 325–362.
Yermack, David, “Do corporations award CEO stock options effectively?,” Journal of
Financial Economics, 1995, 39 (2), 237 – 269.
, “Good Timing: CEO Stock Option Awards and Company News Announcements,”
The Journal of Finance, 1997, 52 (2), 449–476.

24
2 | Literature Review

Private equity funds today are organized as limited partnerships where investors, as
limited partners (LPs), provide capital to a private equity fund that is managed by a
private equity firm, acting as a general partner (GP). This structure has its origins in
1958, when the first venture capital limited partnership - Draper, Gaither, and Anderson
- was formed, while the first venture capital firm - American Research and Development -
was established 12 years earlier (Gompers and Lerner, 2001). Buyout funds, the focus of
this dissertation, is a more recent phenomenon. Some of the earliest firms, such as KKR
and Thomas H Lee Partners, were formed in the 1970s. In 1978, an important regulatory
change increased the potential pool of investors: the clarification of the “prudent man
rule” by the US Department of Labor, which allowed pension funds to invest in high-risk
assets such as private equity funds. Following that change, leveraged buyouts grew
rapidly during the 1980s, and Jensen (1989) famously predicted that the ownership form
provided by buyouts would eclipse the public market.
As it turned out, Jensen’s prediction came at the end of the first, though not the
last, buyout boom. Kaplan and Strömberg (2009) observe that private equity appears
to go through recurring boom and bust periods, both in terms of transaction volumes
and capital raised in new private equity funds.
Buyouts utilize significant amounts of leverage, and each buyout wave has been
accompanied with developments in the debt market. The 1980s wave tapped the junk bond
market (Kaplan and Stein, 1993), the 2000s wave was fuelled by growth in collateralized
debt obligations (Shivdasani and Wang, 2011), and the growth of the most recent decade
has been accompanied by an increase in the usage of covenant-light leveraged loans, loans
with weaker enforcement features (Becker and Ivashina, 2016).
Becker and Ivashina (2016) document that 70% of newly-issued leveraged loans were
covenant-light in 2015, three times the level in the earlier peak in 2007. The authors
provide evidence that this increase is attributable to an increased involvement by non-
bank institutions, which increases the cost of coordination among lenders that is required
when enforcing covenants. However, the changed investor base cannot explain changes
in the strength of restrictions on indebtedness, liens, payments, or asset sales. Loan

25
terms also depends on the sponsoring GP: Demiroglu and James (2010) and Ivashina
and Kovner (2011) show that reputable GPs with good banking relationships are able
to obtain more favourable borrowing terms.
Kaplan and Strömberg (2009) hypothesize that private equity firms are able to take
advantage of mispricings in the debt and equity markets, by borrowing when interest
rates are low. They suggest that one criteria for a private equity boom to occur is
that earnings yields (EBITDA/Enterprise Value) exceed the interest rate charged on
high-yield rates. Consistent with this hypothesis, Ljungqvist et al. (2020) provide evidence
that the investment pace of private equity funds increases when credit spreads are low,
and Axelson et al. (2013) document that debt levels in buyouts are largely driven by
credit market conditions, while being unrelated to cross-sectional factors suggested by
traditional capital structure theories.1
Jensen (1989) views leverage as one of the advantages of buyouts. Leverage disciplines
management by alleviating the free cash flow problem articulated in Jensen (1986). He
further argues that buyouts provide greater incentives to improve value compared to
public corporations. The management team of companies acquired in buyouts typically
have higher equity ownership stakes than management teams of publicly listed companies
(Cronqvist and Fahlenbrach, 2013; Kaplan and Stein, 1993; Leslie and Oyer, 2009).
Additionally, the concentrated ownership ensures that the private equity firm has sufficient
incentives to monitor and get actively involved in the company when required. Finally,
private equity firms’ compensation is conditional on performance, both directly through the
carried interest provision and indirectly as high performance supports future fundraising
(e.g Chung et al., 2012).
The governance model of private equity attracts criticism as well. Investing in private
equity funds comes at a cost of illiquidity and loss of discretion over investment timing.
Evaluating investments is much less straight-forward than for funds investing in assets with
market valuations, and the lack of transparency complicates performance evaluation (e.g.
Kaplan and Sensoy, 2015). The private equity firm is an agent relative to the investors in
the private equity fund, adding an additional layer of potential agency conflicts (Phalippou,
2009). While proponents point to value gains in portfolio companies, a common critique
is that at least part of the value is a gain for the private equity fund at the expense
of other stakeholders (e.g. Morris and Phalippou, 2020).

1
An opposing view is offered by Haddad et al. (2017). They argue that the cyclicality in buyout
activity is driven primarily by variations in the equity risk premium rather than credit conditions. Their
explanation is based on a simple present value argument: the present value of future performance gains is
higher when discount rates are low. Their model framework captures why we would expect high deal
leverage at times of high buyout activity, as well as low expected returns of those deals, consistent with
the empirical evidence in Axelson et al. (2013).

26
 We can broadly categorize these three areas as issues around investing in private
equity; issues regarding the LP-GP interaction and the structure of private equity funds;
and the impact of private equity ownership on portfolio companies. The next three
sections will review the literature in these areas.
The papers in the dissertation contribute to each of these topics. The first paper
examines the secondary market for private equity fund stakes, a market that can alleviate
the illiquidity of investing in private equity funds. The second paper focuses on carried
interest, the part of GP compensation that is performance-based and is intended to align
incentives of GPs and LPs. By linking how variation in pay-for-performance sensitivity
relates to operational improvements of portfolio companies, the paper contributes to the
literature covered in the final two sections of this literature review. The third paper focuses
solely on the impact of private equity ownership on portfolio companies, by examining
how changes in CEO ownership stakes relates to operational improvements.

1 Investing in Private Equity

Since private equity funds invest in private companies there are no readily available market
prices. Thus, the usual framework to evaluate fund performance is not easily applicable to
private equity funds. Instead, once a fund is fully liquidated, the timing and magnitude
of cash flows can be used to assess performance. If an investor (or researcher) wishes
to evaluate performance prior to liquidation, one has to rely on net asset value (NAV)
estimations provided by the GP. As GPs have no requirement to disclose these valuations
publicly, many databases relies on voluntary reporting from GPs, source information from
LPs, or a mix of both (see Brown et al., 2015, for a comparison of databases).
Performance has historically been evaluated, and still commonly is by practitioners,
using internal rate of returns (IRR) and money multiples (TVPI). There are multiple
issues with using IRR when comparing investments in general in corporate finance. Since
GPs control the decision of when to exit investments, the suitability of IRR is even
worse in a private equity context (Phalippou, 2008). Despite these issues, Larocque et
al. (2019) find that investors care more about IRRs than TVPIs when deciding whether
to invest in follow-up funds. While the TVPI has fewer issues than IRR, it does not
account for exposure to movements in the public market.
Kaplan and Schoar (2005) provide an alternative measure which accounts for market
movements: the public market equivalent (PME). The PME is defined as the ratio of
the present value of distributions to the present value of capital calls. The discount rate
used is the realized return on the public market, providing a simple interpretation of the
ratio: a value above (below) one indicates that an investor in the private equity fund did

27
better (worse) than she would have done by replicating the cash flows via investments in
the public market. The PME has become the preferred measure to evaluate performance
in academic research, and the measure has received theoretical support by Korteweg
and Nagel (2016) and Sorensen and Jagannathan (2015).
Early studies using the PME found that, net-of-fees, returns in private equity were
equal to or worse than those of the public market (Kaplan and Schoar, 2005; Phalippou
and Gottschalg, 2009). However, both studies utilized databases that were later discovered
to not provide an accurate picture of the performance of private equity funds. More recent
studies by Harris et al. (2014a), Harris et al. (2016), and Robinson and Sensoy (2016)
find that buyout funds have outperformed the public market by roughly 3% per year,
although performance have not been as impressive for the more recent vintages.
Phalippou (2014) and Stafford (2017) suggest that this outperformance may be an
artefact of the type of companies invested in. Compared to investments that more
closely resembles typical buyout targets, private equity funds have not generated any
additional value to their investors. As noted in Kaplan and Sensoy (2015), the investable
market of listed companies with such characteristics is only a fraction of the buyout
market. Hence, private equity may offer an efficient way for large institutional investors
to get exposure to such companies.
There is some evidence of performance persistence among private equity funds,
contrasting evidence among mutual funds (e.g. Carhart, 1997; Fama and French, 2010).
Kaplan and Schoar (2005) found evidence of performance persistence among GPs, and
Lerner et al. (2007) of heterogeneity in performance among LPs. Persistence in performance
may arise due to asymmetric information between investors in the fund and outsiders
(Hochberg et al., 2014), or due to heterogeneity among LPs in their ability to absorb
liquidity shocks (Maurin et al., 2020). Empirically, persistence appears to have declined
over time among GPs (Braun et al., 2017; Harris et al., 2014b) as well as LPs (Sensoy
et al., 2014), though it has not disappeared (Korteweg and Sorensen, 2017; Cavagnaro
et al., 2019). Evidence of skill is not only found at the institutional level, as Ewens and
Rhodes-Kropf (2015) document skill at individual partners in the venture capital context.
Overall, the literature suggests that there is an element of persistent ability in private
equity, both among GPs and LPs, and that the performance of private equity funds
exceed that of publicly listed markets. The trend over time has been a decline in both,
suggesting a maturing of the industry with increased competition.
Even if private equity funds outperform the publicly listed market, it is not immediately
obvious that it represents a risk-adjusted outperformance for investors. In particular,
private equity investments are much more illiquid than public markets. Theoretical
models by Sorensen et al. (2014) and Bollen and Sensoy (2016) suggest that the observed

28
outperformance is just enough to compensate LPs for the illiquidity of their investments.
Ang et al. (2018) and Franzoni et al. (2012) provide empirical evidence that outperformance
disappears when controlling for the exposure of private equity returns to a liquidity factor.
Lerner and Schoar (2004) argue that the illiquidity of private equity investments is an
optimal design feature, chosen by GPs to screen for investors with “deep pockets” and
long investment horizons. The intuition of their model is that investors in the current
fund possess information about the quality of the GP. If current investors are unable to
provide capital for the next fund due to liquidity constraints, outsiders cannot distinguish
this from an active choice to not invest with a GP of low skill. Having investors with
deep pockets therefore decrease the future cost of capital for the GP, at the expense of a
higher cost of capital in early funds to compensate LPs for the illiquidity.
Lerner and Schoar (2004) model illiquidity as a choice variable for GPs since they
observe severe restrictions on the transferability of LPs’ fund stakes. For example, GPs
are typically able to block the sale of a fund stake. Consequently, the secondary market
for private equity fund stakes, where LPs sell their stakes to other investors, was marginal
for a long time. But the market has grown rapidly from an annual volume of $2 billion
in the early 2000s to a record $88 billion in 2019, presumably because the cost of not
allowing LPs the ability to liquidate early became too costly.
The development of a secondary market can potentially spur further growth in
fundraising in the primary market, as the cost of illiquidity is reduced. The impact of
the secondary market on required returns in the primary market is modelled in Bollen
and Sensoy (2016). In their model, the degree to which the secondary market reduces
the illiquidity cost of private equity investments depends on the discount from a fund’s
fair value at which fund stakes transact in the market.
Maurin et al. (2020) further build on the importance of illiquidity in private equity.
LPs in their model differ in their resilience to liquidity shocks. With heterogeneity in
tolerance for illiquidity among LPs, their model provide an explanation for rigidity in
contract terms, return persistence at both the GP and LP level, and why some LPs
have preferential access to private equity funds. Their model further suggests that GP
performance persistence disappear when a secondary market exist, which could explain
why performance persistence has declined among GPs in recent years.
Despite the rapid growth of the secondary market, Nadauld et al. (2019) document
that the average transaction occurs at a discount to net asset value (NAV), on average
15% for buyouts though varying over time. They also document that the performance
of buyers exceed that of the sellers. They interpret these findings as consistent with
buyers being compensated for providing liquidity. Their interpretation of the difference in
buyer and seller performance is based on an assumption of constant returns throughout

29
a private equity fund’s life. However, rational buyers should pay at most a price that
ensures they earn their required rate of return in expectation, whereas the performance
of sellers depends on the fund’s performance to date in addition to the transaction
price. If sellers opt to sell stakes in poorly performing funds, as evidence in their paper
suggests, we would expect buyers to perform better.
Kleymenova et al. (2012) examine how factors related to the liquidity of a fund stake
impact the price investors are willing to pay. They define liquidity as having more bidders
in an auction, more money bidded, and a smaller dispersion in the bids placed. They
find that large buyout funds are the most liquid, and that more liquid funds are sold at
smaller discounts. In a paper using the same dataset as employed in the first paper of
the thesis, Albuquerque et al. (2018) study liquidity provision in the secondary market
by modelling demand for private equity fund stakes. They find that aggregate demand
is pro-cyclical, and that demand flows to funds where selling pressure is expected to be
high, particularly in bad times. As bids placed for these funds are lower, they interpret
this as evidence that bidders require compensation for providing liquidity.
Overall, these papers suggest that liquidity compensation is an important reason for
discounts in the secondary market. While one might expect secondary market transactions
to be high during distressed times, it is somewhat surprising that volumes where markedly
lower in 2009 after the financial crisis (see Nadauld et al., 2019). One potential explanation,
provided by Hege and Nuti (2011), is that sellers and buyers disagreed on the value of the
fund stakes. Sellers anchored their valuations at reported NAVs, while buyers believed
that those valuations were stale and demanded large discounts.
As prices are quoted as a percentage of NAVs, our interpretation of discounts depends
on whether reported NAVs are fair estimations of the underlying funds value. Though
Jenkinson et al. (2019) find that NAVs are fair on average, evidence in Barber and Yasuda
(2017), Bienz et al. (2017), and Brown et al. (2019) suggests that NAVs are sometimes
misreported. Ang et al. (2018) and Boyer et al. (2019) document autocorrelation in
private equity indices, suggesting that NAVs do not fully incorporate recent returns.
If NAVs are overvalued (undervalued), expected future performance is low (high). In
addition to liquidity compensation, we may therefore expect variation in discounts due
to expectation of future fund performance, similarly to the model of closed-end fund
discounts by Berk and Stanton (2007). The first paper of the thesis contributes to this
literature by examining if bids in the secondary market for private equity fund stakes are
informative of the future performance of funds for which the bids are placed.

30
2 The Structure of Private Equity Funds

Private equity funds are typically organized in similar ways, as limited partnerships
(Sahlman, 1990). LPs provide ex ante equity financing, providing the GP with significant
leeway over when and what to invest in, while deals employ significant amount of leverage.
The profit-sharing arrangement from carried interest is typically based on fund profits,
but GPs could be compensated on a deal-by-deal basis.
Axelson et al. (2009) provide a model that justifies why buyouts are financed in
this way. Their model suggests that providing ex ante financing and compensating
GPs on a fund level reduce the incentive of GPs to invest in bad deals. Since carried
interest payoffs are similar to call options, deal-by-deal level compensation incentivizes
gambling, making even poor investments attractive from the GP’s point of view. However,
there is an incentive for GPs to spend all the capital towards the end of the investment
period, even if no good investments are found. Complementing the ex ante financing
with deal-level debt financing provides additional discipline that reduce the amount
of bad deals pursued in bad times. The financing structure optimally reduces agency
conflicts, although investment still deviates from its first-best level, with overinvestment
(underinvestment) in good (bad) times.
Even contracts that are set up optimally ex ante may be incomplete and therefore
not ex post optimal in all possible states. For example, GPs may invest in bad deals
when they have large amount of unspent committed capital, as the basis of management
fees often changes from committed capital to net invested capital after the investment
period ends (Metrick and Yasuda, 2010). Consistent with this idea, Arcot et al. (2015)
and Degeorge et al. (2016) provide evidence that secondary buyouts undertaken under
pressure underperform. Arcot et al. (2015) show that pressured buyers pay higher
multiples, syndicate less, and use less leverage, consistent with a desire to spend equity.
Restrictions on GPs are set out in the limited partnership agreement (LPA). Investors
cannot get directly involved in running the fund if they wish to retain the status as limited
partners. Gompers and Lerner (1996) examine variation in the usage of covenants in
LPAs in venture capital funds. They find that covenants are more widely used when
more severe agency problems are expected. They also find a supply and demand effect:
when there is an abundance of capital available to be invested in funds, GPs are able
to negotiate contracts with fewer restrictions.
An important part of the LPA is the fee structure of private equity funds, which
constitutes the economics of a private equity fund. While headline fees for buyouts are
typically 2% of committed capital and a 20% carried interest, there is more heterogeneity
in the details. The basis for management fees is usually changed after the investment

31
period, but not always, and the fee level itself may change. Similarly, the basis for carried
interest could be committed capital or invested capital, determining whether investors
receive their fees back before the GP starts earning carry.
Metrick and Yasuda (2010) discuss these and other contract terms in detail, and
estimate how these terms impact GPs’ expected fee revenue.2 Litvak (2009) provides a
legal perspective to compensation terms in venture capital (VC), extending the analysis
of Gompers and Lerner (1996) by considering additional fund terms. In particular, Litvak
(2009) emphasizes the importance of the timing of carry payments for its value to GPs.
One of the major findings in Metrick and Yasuda (2010) is that about two-thirds of
the expected revenue in private equity funds is attributable to the fixed component of
pay that is independent on performance. Their analysis considers the economics of a
given fund, and consequently only captures the direct pay from performance. Chung et
al. (2012) complements their analysis by considering the indirect pay from performance:
the increased ability of a GP to raise additional funds that generate future streams of
revenue. They estimate that the magnitude of indirect pay is at least as big as the
direct component. Consistent with their learning model, they find that fundraising
success is increasingly dependent on performance for younger GPs that lack a strong
track record (see also Barber and Yasuda, 2017).
How much this pay-for-performance sensitivity of GP pay incentivizes effort of the
partners involved in specific deals is unclear; Ivashina and Lerner (2019) find that the
allocation of carry is primarily related to the status of founder of a PE firm, rather than
past success. However, there may still be an implicit career concern, as they find that
partners depart if they do not get a fair share of profits. Such departures negatively
impact the ability of GPs to raise more capital.
GPs additionally charge monitoring and transaction fees directly to portfolio companies,
which Metrick and Yasuda (2010) find puzzling as management fees are supposed to cover
such costs. Phalippou et al. (2018) collect information about portfolio company fees from
SEC filings. They find that these fees amount to about 6% of equity invested by GPs.
The authors consider several potential justifications for these fees, such as tax-efficient
distributions and ex post adjustments necessary due to incomplete contracting in the
LPA, but find it hard to rationalize the fees. They suggest that these fees may reflect
GPs extracting value from inattentive LPs. They document that GPs backed by more
skilled LPs charge lower fees, and present evidence that LPs learn over time: the amount

2
Other contract terms they cover are: catch-up clauses in funds with hurdle rates, allowing GPs to earn
a higher fraction of profits after beating the hurdle rate to “catch up” to the pre-specified profit-sharing
level; early carry, where the GP may receive carry based on unrealized (estimated) profits; early carry is
typically associated with a clawback provision, allowing LPs to claw back early carry if the ultimate fund
performance does not warrant the amount of carry paid out.

32
of fees charged by GPs prior to the financial crisis is negatively related to fundraising
success in the period after the financial crisis.
Other contract terms discussed in Metrick and Yasuda (2010) can give rise to
misalignment in incentives between LPs and GPs. Robinson and Sensoy (2013) document
that GPs accelerate distributions when funds enter their “catch-up” region, where GPs
earn 100% of profits until reaching the profit-sharing level. The authors interpret the
acceleration of distributions in order to enjoy this “waterfall” as likely to lead some
investments to be exited too early. They also find evidence consistent with GPs holding
onto investments for too long, when doing so allows the GP to keep earning management
fees on remaining invested capital.
Two papers examine how variation in fee terms relate to fund performance. Robinson
and Sensoy (2013) do not find evidence of lower net-of-fee performance in funds with
higher fees or lower managerial ownership. Their results are akin to the equilibrium
model of Berk and Green (2004), where high ability GPs extract rents in the form of
improved contract terms. Hüther et al. (2020) find that VC funds with “GP-friendly”
carry provisions, where carry is paid on a deal-by-deal basis, are associated with higher
returns both gross and net of fees.
The second paper of the thesis contributes to this literature by examining the dynamics
of the pay-for-performance sensitivity of carried interest within a given fund. The paper
relates the resulting within-fund variation in pay-for-performance sensitivity to operational
improvements at portfolio companies acquired in buyouts.

3 The Economic Impact of Buyouts

While there is debate on whether private equity funds perform well enough to justify
the fees charged, the gross returns of private equity deals are substantially above public
market returns (e.g. Braun et al., 2017; Brown et al., 2020). An important question
is how private equity generate these gains.
Acharya et al. (2013) provide a method that helps us broadly think about the
components of private equity returns. Returns are partly driven by exposure to market
movements, including multiple expansion (Guo et al., 2011; Jenkinson et al., 2018),
partly a result of leverage - providing a tax shield and amplifying returns to equity
- and partly through a residual effect. This residual effect could arise due to M&A
expertise that allow private equity firms to identify promising targets or successfully
grow portfolio companies with add-on acquisitions (Bansraj et al., 2019), or an ability
to improve operations of target companies. Given the amount of companies controlled
by private equity, the impact of private equity ownership on portfolio companies is an
area that has received considerable interest.

33
 Jensen (1989) views buyouts as a way to improve how companies are run by improving
governance, incentive schemes, and providing an optimized financial structure. Early
studies documented operational improvements, and their importance for value generation,
in buyouts (Kaplan, 1989; Lichtenberg and Siegel, 1990; Muscarella and Vetsuypens, 1990;
Smith, 1990). Guo et al. (2011) largely confirm the importance of operating gains for
deal performance, though on average they find, at best, small improvements in operating
performance compared to benchmark companies. On corporate governance, buyouts
typically lead to increased managerial equity ownership stakes (Kaplan and Stein, 1993;
Cronqvist and Fahlenbrach, 2013), venture capitalists support professionalization of
companies (Hellmann and Puri, 2002), and portfolio companies backed by private equity
have better management practices (Bloom et al., 2015).
In the absence of regular market valuations of companies held private, some studies
use company financials to measure improvements in operational performance. However,
private companies in the US are typically not required to make financial statements
publicly available. As a consequence, US studies often rely on subsamples of companies
with publicly available financials (e.g. Guo et al., 2011), or data made available to
them by industry practitioners (e.g. Acharya et al., 2013). There is a concern that
such samples are not necessarily representative of a typical buyout.3 Cohn et al. (2014)
use tax data to show a marked difference in operational gains among companies with
and without publicly available financials, with the latter displaying significantly worse
operational performance. An alternative is to examine countries outside of the US,
where data access is more readily available.
Public-to-private buyouts have received the largest amount of attention, presumably
due to their size, visibility, and data availability. However, they represent only a small
fraction of all buyouts. In Strömberg (2008), 6.7% of all deals are public-to-private
buyouts, though they represent 28.2% of enterprise value. Recent studies by Davis et
al. (2019) and Cohn et al. (2020) highlight stark differences between different type of
buyouts in the impact private equity ownership has on target companies. Davis et al.
(2019) find that employment at target companies shrinks in public-to-private buyouts, but
expands for privately held companies. Similarly, Cohn et al. (2020) find rapid growth of
already private companies, which they attribute to alleviating financing constraints. This
interpretation is similar to Boucly et al. (2011) that, examining small French companies
acquired in buyouts, provide evidence consistent with the notion that private equity
ownership can support growth by relaxing credit constraints. Similar evidence is provided
by Chung (2011) for small, private, companies acquired in buyouts in the UK.
3
Financials are publicly available only if the company has public debt that requires disclosure of
financials, or if the company subsequently lists on the public market and in that process provides historical
financial statements.

34
 An important change brought about by buyouts is the increase in leverage, a practice
that has received considerable criticism as it can lead to an increased risk of bankruptcy.
For example, Kaplan and Stein (1993) argue that PE transactions in the second half
of the 1980s used excessive amounts of leverage, which they attribute to a “demand
push” from the junk bond market. Andrade and Kaplan (1998) estimate the cost of
financial distress to be 10 to 20 percent of company value.
As private equity firms are serial investors that commonly utilize high levels of leverage,
their experience may make them better able to manage companies in financial distress.
Hotchkiss et al. (2014) show that, after controlling for leverage, companies backed by
private equity are no more likely to default than other leveraged loan borrowers. When
default does occur, PE-backed companies restructure faster, more often restructure out
of court, and the company is more likely to remain an independent entity. They also
document that the likelihood of a successful restructuring increases when the PE owner
injects new capital in the company. Bernstein et al. (2019b) similarly find that having a
private equity owner can help companies withstand economic crises. Companies backed
by private equity during the financial crisis saw smaller declines in investment, and were
more likely to receive equity injections as well as take on new debt. The documented
effect is stronger when the PE fund has high amounts of dry powder, supporting an
interpretation that private equity owners can support struggling companies by capital
injections. More successful resolution of financial distress can be socially valuable, as
reorganizing a company result in higher asset utilization than liquidation (Bernstein et
al., 2019a). However, the main analysis in these papers compare companies backed by
private equity with comparable companies with similar leverage. We do not yet have a
complete understanding of the net effect of the initial increase in leverage.4
Private equity is often accused of creating job losses, cutting costs, and enriching
themselves by focusing on short-term profits at the expense of the long-term investments
or by extracting wealth from other stakeholders. The academic evidence is not entirely
positive of private equity, though it has refuted many of the accusations aimed at private
equity (see e.g. Morris and Phalippou, 2020). Looking at plant level data, Davis et al.
(2014) document that private equity firms are active in reallocating capital by exiting
less productive establishment while expanding in more productive ones. The net effect
of this is a small net job loss, combined with large gross job creation as well as job
destruction, and an increase in total factor productivity. In continuing plants, they find

4
The current economic downturn caused by lockdowns to stem the spread of Covid-19 may provide
for an additional stress test of the private equity model with high leverage. However, comparisons
between portfolio companies and those not backed by private equity may be problematic as many
portfolio companies risk not qualifying for relief packages, at least in Europe (see for example
https://www.ft.com/content/46641ba1-fd7d-4718-8f17-262881bf31ef, accessed 2020-05-21).

35
no evidence of improved productivity relative to peers, similarly to what Bharath et al.
(2014) document for public companies transitioning to private ownership.
Examining the claim that short-term profits come at the expense of long-term
investments, Lerner et al. (2011) find no evidence that investment in innovation is
harmed. Instead, after undergoing a buyout, patents are more cited and become more
concentrated areas of high importance for the company. In a similar spirit, Harford and
Kolasinski (2014) do not find evidence of short-terminism or the extraction of wealth from
debt holders, except in some special situations. Looking at spillover effects on competitors,
Bernstein et al. (2017) show that industries in which private equity invests grow more
quickly in terms of both employment and total production.
Some studies focus on the impact of private equity ownership on customers. Matsa
(2011) finds that companies with high financial leverage degrade their product quality
by reducing inventory to preserve cash to service debt, potentially harming consumer
loyalty. Fracassi et al. (2019) examine manufacturing companies of consumer goods that
are acquired in buyouts. Companies in their sample increase revenue substantially, but
they do so via expanding geographically and by introducing new products rather than
raising price. Prices on existing products only increase by 1% relative to competitors’
products. Biesinger et al. (2020) find similar results, with revenue growth accompanied
by declines in price markups. Studying buyouts of supermarket chains in the 1980s,
Chevalier (1995) documents price increases (decreases) when competitors are highly
leveraged (have low leverage).
Eaton et al. (2019) suggest that private equity ownership is not equally likely to be
beneficial for customers in all types of industries. They argue that sectors with intense
competition are likely to experience gains for customers, while industries characterized
by opaque product quality and substantial government subsidies are likely areas where
private equity value comes at the expense of customers. Their study focus on one
such sector: higher education. Consistent with their argument, they find that buyouts
are associated with an increase in enrolment, tuition costs and student debt, while
quality drops. They observe lower graduation rates, lower earnings, and reduced loan
repayment rates, while at the same time spending on teaching is reduced and the ratio of
faculty to students decreases. Gupta et al. (2019) similarly find that quality decreases
in healthcare homes acquired in buyouts.
The theoretical model of Shleifer and Summers (1988) suggests that investor-led
acquisitions may transfer value from employees to the investors by breaking implicit
contracts. Several papers examine the impact of buyouts on employees. Davis et al.
(2019) document a modest net loss of jobs in buyout targets, particularly for public-
to-private buyouts. Faccio and Hsu (2017) show that buyouts by politically connected

36
private equity firms are associated with higher employee growth than other buyouts. The
authors interpret this as a quid pro quo arrangement, as this effect is more pronounced
around election years and target companies are more likely to receive government
contracts and grants.
Focusing on wage growth, Amess and Wright (2007) document that UK buyouts are
associated with lower employee wage-growth. Lichtenberg and Siegel (1990) find that
employment and compensation of white-collar workers decline, while that of blue-collar
workers are unchanged. That not all workers are equally at risk are found in other
countries as well. Antoni et al. (2019) show that average earnings decline in German
buyouts, with employees that are least likely to find new jobs experiencing the most
negative impact. Olsson and Tåg (2017) examine the type of job that is most at risk in
Swedish buyouts. They find that workers doing routine tasks, or have jobs that can be
offshored, are more likely to be laid off. This result is particularly strong for companies
with low pre-buyout productivity relative to peers, suggesting that buyouts may accelerate
necessary but unpleasant actions to improve competitiveness.
Agrawal and Tambe (2016) document that buyouts pushing to modernize portfolio
companies can yield positive spillover effects on employees. In companies that receive
significant IT investment, employees experience improved long-run employability, and
higher wages, by gaining valuable human capital in the form of IT skills. Workers may
also benefit from other initiatives undertaken under private equity ownership. Cohn et
al. (2019) find that both establishment-level workplace injury rates and safety inspection
violations decline after public-to-private buyouts.
One obstacle facing research attempting to evaluate the impact of private equity
ownership is that targets are not randomly chosen, making it difficult to attribute any
observed changes to a causal impact of private equity ownership. Ofek (1994) examines
this hypothesis by evaluating the performance of management buyouts that failed despite
management supporting it, and find no improvements in operational performance for
targets of failed management buyouts. Bernstein and Sheen (2016) provide a detailed
study of restaurant chains, utilizing health inspection records. They show that, following
a buyout, restaurants become cleaner, safer, and better maintained. Importantly, their
setting allows for a natural control group: franchised restaurants. Consistent with a
causal interpretation, improvements are larger at chain-owned stores than in franchised
ones over which the PE firm has limited control. The value of involvement by venture
capitalists is examined by Bernstein et al. (2016). They utilize exogenous variation in the
ability of VCs to visit portfolio companies, using newly opened airline routes that reduce
travel time. They find that the introduction of a new airline route results in portfolio
companies becoming more innovative, and are more likely to be successfully exited.

37
 The second paper of the thesis contributes to the wider literature on operational
improvements by examining how variation in the pay-for-performance sensitivity of a
GP’s payoff relates to operational improvements at portfolio companies. The paper
exploits within-fund variation arising due to staggered investments, as well as exoge-
nous, heterogeneous, variation in the pay-for-performance sensitivity arising from public
equity market movements.
Most of the evidence discussed above, including the second paper of the thesis, do
not link specific actions taken by private equity firms to improvements in operational
performance of portfolio companies. Instead, they document an overall impact of private
equity ownership.5 For example, workplace injury rates could decline in Cohn et al.
(2019) as a result of reduced short-term pressure from public markets, because the new
private equity owners redirects investments to reduce injuries, or for another reason
completely. Similarly, low productivity companies in Olsson and Tåg (2017) may increase
offshoring because management receives sufficient incentives to take unpleasant actions
and lay off workers, the old management team may have been replaced, or another
reason unrelated to the management team.
Two recent papers offer insights into what GPs say they do. Biesinger et al. (2020)
examine value creation plans prepared for individual deals. They document a significant
heterogeneity in strategies, but find that ex ante selection of strategy is not related to deal
returns. Instead they document that, somewhat less surprisingly, deals in which the GP
claims they successfully executed their strategies have high returns. The authors interpret
this as execution being the key to achieve high returns for investors. A complementary
analysis is provided by Gompers et al. (2016) that survey GPs and ask what they believe
generate value in their deals. The top five expected sources of value creation are, in order,
“increase revenue or improve demand”, “improve incentives”, “follow-on acquisitions”,
“facilitate a high-value exit”, and “improve corporate governance”. Clearly, these sources
are not mutually exclusive as, for example, follow-on acquisitions increase revenue while
improving incentives is likely part of improving governance. The belief that GPs can
help increase revenue is consistent with evidence in buyouts of already private companies
(e.g. Boucly et al., 2011; Chung, 2011; Cohn et al., 2020).
Private equity firms are often seen as strong principals whose actions can inform
us about optimal governance. In this spirit, Kaplan and Strömberg (2003); Kaplan
and Strömberg (2004) examine how contracts are designed when VCs provide financing

5
Some exceptions are Davis et al. (2014) that document the importance of reallocating capital to
improve productivity, though again the exact actions that lead to these decisions are not well understood.
There is also evidence on the importance of relaxing credit constraints to spur growth and investment
provided by Bernstein et al. (2019b), Boucly et al. (2011), Chung (2011), and Cohn et al. (2020).

38
to entrepreneurs. They document that contracts are generally designed in line with
what optimal contracting theory suggests.
Similarly, Cronqvist and Fahlenbrach (2013) study changes in CEO contracts when
companies are taken private, looking in detail at the contracts of 20 such transactions.
They find an increase in CEO equity ownership stakes, and CEO pay becomes more
sensitive to performance. Leslie and Oyer (2009) look at changes in CEO ownership
stakes. However, they rely on different sets of companies for pre-buyout and post-buyout
stakes, and their post-buyout stakes are after private equity exits through an IPO. They
do not observe CEO ownership stakes during the period of private equity ownership,
and cannot observe changes in any given company.
Cornelli and Karakas (2015) study the role of boards in companies owned by private
equity. While CEOs are often replaced in the acquisition, CEO turnover during private
equity ownership is reduced when the private equity firm sits on the board, and is less
sensitive to performance. These results suggest that monitoring allows private equity firms
to rely less on verifiable performance data. The private equity setting has also been used to
examine agency issues in the form of corporate jets as an excessive perk. Edgerton (2012)
documents a sharp reduction in the ownership of corporate jets of companies taken private
by private equity funds. This is driven by the top 30% of the distribution of corporate jet
ownership, suggesting that there is a subset of companies where perks are excessive.
The third paper of thesis contributes to this literature by examining a salient feature of
private equity acquisitions: the increase in managerial equity ownership stakes. The paper
examines how changes managerial ownership stakes relate to operational improvements
at companies acquired in buyouts.
The next three chapters contain the three empirical papers in this dissertation,
and Chapter 6 concludes with a summary of what this dissertation contributes to
the field of private equity.

39
References

Acharya, Viral V., Oliver F. Gottschalg, Moritz Hahn, and Conor Kehoe,
“Corporate Governance and Value Creation: Evidence from Private Equity,” The Review
of Financial Studies, 2013, 26 (2), 368–402.
Agrawal, Ashwini and Prasanna Tambe, “Private Equity and Workers’ Career Paths:
The Role of Technological Change,” The Review of Financial Studies, 2016, 29 (9),
2455–2489.
Albuquerque, Rui, Johan Cassel, Ludovic Phalippou, and Enrique Schroth,
“Liquidity Provision in the Secondary Market for Private Equity Fund Stakes,” 2018.
Working Paper.
Amess, Kevin and Mike Wright, “The Wage and Employment Effects of Leveraged
Buyouts in the UK,” International Journal of the Economics of Business, 2007, 14 (2),
179–195.
Andrade, Gregor and Steven N. Kaplan, “How Costly is Financial (Not Economic)
Distress? Evidence from Highly Leveraged Transactions that Became Distressed,” The
Journal of Finance, 1998, 53 (5), 1443–1493.
Ang, Andrew, Bingxu Chen, William N. Goetzmann, and Ludovic Phalippou,
“Estimating Private Equity Returns from Limited Partner Cash Flows,” The Journal
of Finance, 2018, 73 (4), 1751–1783.
Antoni, Manfred, Ernst Maug, and Stefan Obernberger, “Private equity and
human capital risk,” Journal of Financial Economics, 2019, 133 (3), 634 – 657.
Arcot, Sridhar, Zsuzsanna Fluck, José-Miguel Gaspar, and Ulrich Hege, “Fund
managers under pressure: Rationale and determinants of secondary buyouts,” Journal
of Financial Economics, 2015, 115 (1), 102 – 135.
Axelson, Ulf, Per Strömberg, and Michael S. Weisbach, “Why Are Buyouts
Levered?,” The Journal of Finance, 2009, 64 (4), 1549–1582.
, Tim Jenkinson, Per Strömberg, and Michael S Weisbach, “Borrow Cheap,
Buy High? The Determinants of Leverage and Pricing in Buyouts,” Journal of Finance,
2013, 68 (6), 2223–2267.
Bansraj, Dyaran, Han T.J. Smit, and Vadym Volosovych, “Can Private Equity
Funds Act as Strategic Buyers? Evidence from Buy-and-Build Strategies,” 2019.
Working Paper.
Barber, Brad M and Ayako Yasuda, “Interim Fund Performance and Fundraising in
Private Equity,” Journal of Financial Economics, 2017, 124 (1), 172–194.
Becker, Bo and Victoria Ivashina, “Covenant-Light Contract and Creditor Coordina-
tion,” 2016. Sveriges Riksbank Working Paper Series No. 325.
Berk, Jonathan B. and Richard C. Green, “Mutual Fund Flows and Performance
in Rational Markets,” Journal of Political Economy, 2004, 112 (6), 1269–1295.

40
Berk, Jonathan B and Richard Stanton, “Managerial Ability, Compensation, and
the Closed-End Fund Discount,” Journal of Finance, 2007, 62 (2), 529–556.
Bernstein, Shai and Albert Sheen, “The Operational Consequences of Private Equity
Buyouts: Evidence from the Restaurant Industry,” The Review of Financial Studies,
2016, 29 (9), 2387–2418.
, Emanuele Colonnelli, and Benjamin Iverson, “Asset Allocation in Bankruptcy,”
The Journal of Finance, 2019, 74 (1), 5–53.
, Josh Lerner, and Filippo Mezzanotti, “Private Equity and Financial Fragility
during the Crisis,” The Review of Financial Studies, 2019, 32 (4), 1309–1373.
, , Morten Sorensen, and Per Strömberg, “Private Equity and Industry
Performance,” Management Science, 2017, 63 (4), 1198–1213.
, Xavier Giroud, and Richard R. Townsend, “The Impact of Venture Capital
Monitoring,” The Journal of Finance, 2016, 71 (4), 1591–1622.
Bharath, Sreedhar, Amy Dittmar, and Jagadeesh Sivadasan, “Do Going-Private
Transactions Affect Plant Efficiency and Investment?,” The Review of Financial Studies,
2014, 27 (7), 1929–1976.
Bienz, Carsten, Charlotte Wallem Rakner, and Anne Therese Rasmussen,
“Strategic Valuation Management in Private Equity Fundraising,” 2017. Working Paper.
Biesinger, Markus, Çağatay Bircan, and Alexander Ljungqvist, “Value Creation
in Private Equity,” 2020. Working Paper.
Bloom, Nicholas, Raffaella Sadun, and John Van Reenen, “Do Private Equity
Owned Firms Have Better Management Practices?,” American Economic Review, 2015,
105 (5), 442–46.
Bollen, Nicolas P B and Berk A Sensoy, “How much for a haircut? Illiquidity,
secondary markets, and the value of private equity,” 2016. Fisher College of Business
Working Paper.
Boucly, Quentin, David Sraer, and David Thesmar, “Growth LBOs,” Journal of
Financial Economics, 2011, 102 (2), 432 – 453.
Boyer, Brian, Taylor D. Nadauld, Keith Vorkink, and Michael S. Weisbach,
“Private Equity Indices Based on Secondary Market Transactions,” 2019. Dice Center
WP 2018-21.
Braun, Reiner, Tim Jenkinson, and Ingo Stoff, “How persistent is private equity
performance? Evidence from deal-level data,” Journal of Financial Economics, 2017,
123 (2), 273 – 291.
Brown, Gregory, Robert S. Harris, Wendy Hu, Tim Jenkinson, Steven N.
Kaplan, and David Robinson, “Private equity portfolio companies: A first look at
Burgiss holdings data,” 2020. Working Paper.
Brown, Gregory W., Oleg R. Gredil, and Steven N. Kaplan, “Do private equity
funds manipulate reported returns?,” Journal of Financial Economics, 2019, 132 (2),
267 – 297.
, Robert S. Harris, Tim Jenkinson, Steven N. Kaplan, and David Robinson,
“What Do Different Commercial Data Sets Tell Us About Private Equity Performance?,”
2015. Working paper.
Carhart, Mark M., “On Persistence in Mutual Fund Performance,” The Journal of
Finance, 1997, 52 (1), 57–82.
Cavagnaro, Daniel R., Berk A. Sensoy, Yingdi Wang, and Michael S. Weisbach,
“Measuring Institutional Investors’ Skill at Making Private Equity Investments,” The
Journal of Finance, 2019, 74 (6), 3089–3134.

41
Chevalier, Judith A., “Do LBO Supermarkets Charge More? An Empirical Analysis
of the Effects of LBOs on Supermarket Pricing,” The Journal of Finance, 1995, 50 (4),
1095–1112.
Chung, Ji-Woong, “Leveraged Buyouts of Private Companies,” 2011. Working paper.
, Berk A. Sensoy, Léa Stern, and Michael S. Weisbach, “Pay for Performance
from Future Fund Flows: The Case of Private Equity,” The Review of Financial Studies,
2012, 25 (11), 3259–3304.
Cohn, Jonathan B., Edith S. Hotchkiss, and Erin M. Towery, “The Motives for
Private Equity Buyouts of Private Firms: Evidence from U.S. Corporate Tax Returns,”
2020. Working Paper.
, Lillian F. Mills, and Erin M. Towery, “The evolution of capital structure and
operating performance after leveraged buyouts: Evidence from U.S. corporate tax
returns,” Journal of Financial Economics, 2014, 111 (2), 469 – 494.
, Nicole Nestoriak, and Malcolm Wardlaw, “Private equity buyouts and workplace
safety,” 2019. Working Paper.
Cornelli, Francesca and Oğuzhan Karakas, “CEO Turnover in LBOs: The Role of
Boards,” 2015. Working paper.
Cronqvist, Henrik and Rüdiger Fahlenbrach, “CEO contract design: How do strong
principals do it?,” Journal of Financial Economics, 2013, 108 (3), 659 – 674.
Davis, Steven J., John Haltiwanger, Kyle Handley, Ben Lipsius, Josh Lerner,
and Javier Miranda, “The Economic Effects of Private Equity Buyouts,” 2019.
NBER Working Paper 26371.
, , , Ron Jarmin, Josh Lerner, and Javier Miranda, “Private Equity, Jobs,
and Productivity,” American Economic Review, 2014, 104 (12), 3956–90.
Degeorge, Francois, Jens Martin, and Ludovic Phalippou, “On secondary buyouts,”
Journal of Financial Economics, 2016, 120 (1), 124 – 145.
Demiroglu, Cem and Christopher M. James, “The role of private equity group
reputation in LBO financing,” Journal of Financial Economics, 2010, 96 (2), 306 – 330.
Eaton, Charlie, Sabrina T Howell, and Constantine Yannelis, “When Investor
Incentives and Consumer Interests Diverge: Private Equity in Higher Education,” The
Review of Financial Studies, 2019. Forthcoming.
Edgerton, Jesse, “Agency Problems in Public Firms: Evidence from Corporate Jets in
Leveraged Buyouts,” The Journal of Finance, 2012, 67 (6), 2187–2213.
Ewens, Michael and Matthew Rhodes-Kropf, “Is a VC Partnership Greater Than
the Sum of Its Partners?,” The Journal of Finance, 2015, 70 (3), 1081–1113.
Faccio, Mara and Hung-Chia Hsu, “Politically Connected Private Equity and
Employment,” The Journal of Finance, 2017, 72 (2), 539–574.
Fama, Eugene F. and Kenneth R. French, “Luck versus Skill in the Cross-Section
of Mutual Fund Returns,” The Journal of Finance, 2010, 65 (5), 1915–1947.
Fracassi, Cesare, Alessandro Previtero, and Albert Sheen, “Barbarians at the
Store? Private Equity, Products, and Consumers,” 2019. Working Paper.
Franzoni, Francesco, Eric Nowak, and Ludovic Phalippou, “Private Equity
Performance and Liquidity Risk,” Journal of Finance, 2012, 67 (6), 2341–2373.
Gompers, Paul and Josh Lerner, “The Use of Covenants: an Empirical Analysis of
Venture Partnership Agreements,” Journal of Law & Economics, 1996, 39 (2), 463–498.
and , “The Venture Capital Revolution,” Journal of Economic Perspectives, 2001,
15 (2), 145–168.

42
 , Steven N. Kaplan, and Vladimir Mukharlyamov, “What do private equity
firms say they do?,” Journal of Financial Economics, 2016, 121 (3), 449 – 476.
Guo, Shourun, Edith S. Hotchkiss, and Weihong Song, “Do Buyouts (Still) Create
Value?,” The Journal of Finance, 2011, 66 (2), 479–517.
Gupta, Atul, Sabrina Howell, Constantine Yannelis, and Abhinav Gupta,
“When Investor Incentives and Consumer Interests Diverge: Private Equity in Higher
Education,” 2019. NYU Stern School of Business Research Paper Series.
Haddad, Valentin, Erik Loualiche, and Matthew Plosser, “Buyout Activity: The
Impact of Aggregate Discount Rates,” The Journal of Finance, 2017, 72 (1), 371–414.
Harford, Jarrad and Adam Kolasinski, “Do Private Equity Returns Result from
Wealth Transfers and Short-Termism? Evidence from a Comprehensive Sample of Large
Buyouts,” Management Science, 2014, 60 (4), 888–902.
Harris, Robert S., Tim Jenkinson, and Steven N. Kaplan, “Private equity
performance: What do we know?,” Journal of Finance, 2014, 69 (5), 1851–1882.
Harris, Robert S, Tim Jenkinson, and Steven N Kaplan, “How Do Private Equity
Investments Perform Compared to Public Equity?,” Journal of Investment Management,
2016, 14 (1), 1–24.
Harris, Robert S., Tim Jenkinson, Steven N Kaplan, and Rudiger Stucke, “Has
persistence persisted in Private Equity? Evidence from buyout and venture capital
funds,” 2014. Darden Business School Working Paper No. 2304808.
Hege, Ulrich and Alessandro Nuti, “The Private Equity Secondaries Market During
the Financial Crisis and the ”Valuation Gap”,” Journal of Private Equity, 2011, 14 (3),
42–54.
Hellmann, Thomas and Manju Puri, “Venture Capital and the Professionalization of
Start-Up Firms: Empirical Evidence,” The Journal of Finance, 2002, 57 (1), 169–197.
Hochberg, Yael V., Alexander Ljungqvist, and Annette Vissing-Jørgensen,
“Informational Holdup and Performance Persistence in Venture Capital,” The Review
of Financial Studies, 2014, 27 (1), 102–152.
Hotchkiss, Edith S., David C. Smith, and Per Strömberg, “Private Equity and
the Resolution of Financial Distress,” 2014. Working Paper.
Hüther, Niklas, David T. Robinson, Sönke Sievers, and Thomas Hartmann-
Wendels, “Paying for Performance in Private Equity: Evidence from Venture Capital
Partnerships,” Management Science, 2020, 66 (4), 1756–1782.
Ivashina, Victoria and Anna Kovner, “The Private Equity Advantage: Leveraged
Buyout Firms and Relationship Banking,” The Review of Financial Studies, 2011, 24
(7), 2462–2498.
and Josh Lerner, “Pay now or pay later? The economics within the private equity
partnership,” Journal of Financial Economics, 2019, 131 (1), 61 – 87.
Jenkinson, Tim, Stefan Morkoetter, and Thomas Wetzer, “Buy Low, Sell High?
Do Private Equity Fund Managers Have Market Abilities?,” 2018. University of
St.Gallen, Working Papers on Finance No. 2018/13.
, Wayne R. Landsman, Brian R. Rountree, and Kazbi Soonawalla, “Private
Equity Net Asset Values and Future Cash Flows,” The Accounting Review, 2019, 95
(1), 191–210.
Jensen, Michael, “Eclipse of the Public Corporation,” Harvard Business Review, 1989,
67 (5), 61–74.
Jensen, Michael C., “Agency Costs of Free Cash Flow, Corporate Finance, and
Takeovers,” The American Economic Review, 1986, 76 (2), 323–329.

43
Kaplan, Steven, “The effects of management buyouts on operating performance and
value,” Journal of Financial Economics, 1989, 24 (2), 217 – 254.
Kaplan, Steven N. and Antoinette Schoar, “Private Equity Performance: Returns,
Persistence, and Capital Flows,” The Journal of Finance, 2005, 60 (4), 1791–1824.
and Berk A. Sensoy, “Private Equity Performance: A Survey,” Annual Review of
Financial Economics, 2015, 7 (1), 597–614.
and Jeremy C. Stein, “The Evolution of Buyout Pricing and Financial Structure in
the 1980s*,” The Quarterly Journal of Economics, 1993, 108 (2), 313–357.
and Per Strömberg, “Financial Contracting Theory Meets the Real World: An
Empirical Analysis of Venture Capital Contracts,” The Review of Economic Studies,
2003, 70 (2), 281–315.
and Per Strömberg, “Characteristics, Contracts, and Actions: Evidence from
Venture Capitalist Analyses,” The Journal of Finance, 2004, 59 (5), 2177–2210.
and , “Leveraged Buyouts and Private Equity,” Journal of Economic Perspectives,
2009, 23 (1), 121–146.
Kleymenova, Anya, Eli Talmor, and Florin P Vasvari, “Liquidity in the Secondaries
Private Equity Market,” 2012. Working paper, London Business School.
Korteweg, Arthur and Morten Sorensen, “Skill and luck in private equity
performance,” Journal of Financial Economics, 2017, 124 (3), 535 – 562.
Korteweg, Arthur G and Stefan Nagel, “Risk-Adjusting the Returns to Venture
Capital,” Journal of Finance, 2016, 71 (3), 1437–1470.
Larocque, Stephannie A., Sophie Shive, and Jennifer Sustersic Stevens, “Private
Equity Returns, Cash Flow Timing, and Investor Choices,” 2019. Working paper.
Lerner, Josh and Antoinette Schoar, “The illiquidity puzzle: theory and evidence
from private equity,” Journal of Financial Economics, 2004, 72 (1), 3–40.
, , and Wang Wongsunwai, “Smart Institutions, Foolish Choices: The Limited
Partner Performance Puzzle,” Journal of Finance, 2007, 62 (2), 731–764.
, Morten Sorensen, and Per Strömberg, “Private Equity and Long-Run Investment:
The Case of Innovation,” The Journal of Finance, 2011, 66 (2), 445–477.
Leslie, Phillip and Paul Oyer, “Managerial incentives and value creation: Evidence
from private equity,” 2009. Working paper.
Lichtenberg, Frank R. and Donald Siegel, “The effects of leveraged buyouts on
productivity and related aspects of firm behavior,” Journal of Financial Economics,
1990, 27 (1), 165 – 194.
Litvak, Kate, “Venture capital limited partnership agreements: understanding compen-
sation arrangements,” University of Chicago Law Review, 2009, 76 (1), 161–218.
Ljungqvist, Alexander, Matthew Richardson, and Daniel Wolfenzon, “The
investment behavior of buyout funds: Theory and evidence,” Financial Management,
2020, 49 (1), 3–32.
Matsa, David A., “Running on Empty? Financial Leverage and Product Quality in
the Supermarket Industry,” American Economic Journal: Microeconomics, 2011, 3 (1),
137–73.
Maurin, Vincent, David T. Robinson, and Per Strömberg, “A Theory of Liquidity
in Private Equity,” 2020. Working paper.
Metrick, Andrew and Ayako Yasuda, “The Economics of Private Equity Funds,”
Review of Financial Studies, 2010, 23 (6), 2303–2341.

44
Morris, Peter and Ludovic Phalippou, “Thirty years after Jensen’s prediction: Is
private equity a superior form of ownership?,” Oxford Review of Economic Policy, 2020,
36 (2), 291–313.
Muscarella, Chris J. and Michael R. Vetsuypens, “Efficiency and Organizational
Structure: A Study of Reverse LBOs,” The Journal of Finance, 1990, 45 (5), 1389–1413.
Nadauld, Taylor D., Berk A. Sensoy, Keith Vorkink, and Michael S. Weisbach,
“The liquidity cost of private equity investments: Evidence from secondary market
transactions,” Journal of Financial Economics, 2019, 132 (3), 158 – 181.
Ofek, Eli, “Efficiency Gains in Unsuccessful Management Buyouts,” The Journal of
Finance, 1994, 49 (2), 637–654.
Olsson, Martin and Joacim Tåg, “Private Equity, Layoffs, and Job Polarization,”
Journal of Labor Economics, 2017, 35 (3), 697–754.
Phalippou, Ludovic, “The hazards of using IRR to measure performance: The case of
private equity,” Journal of Performance Measurement, 2008, 4, 1–23.
, “Beware of Venturing into Private Equity,” Journal of Economic Perspectives, March
2009, 23 (1), 147–66.
, “Performance of Buyout Funds Revisited?,” Review of Finance, 2014, 18 (1), 189–218.
and Oliver Gottschalg, “The Performance of Private Equity Funds,” Review of
Financial Studies, 2009, 22 (4), 1747–1776.
, Christian Rauch, and Marc Umber, “Private equity portfolio company fees,”
Journal of Financial Economics, 2018, 129 (3), 559 – 585.
Robinson, David T. and Berk A. Sensoy, “Do Private Equity Fund Managers Earn
Their Fees? Compensation, Ownership, and Cash Flow Performance,” The Review of
Financial Studies, 2013, 26 (11), 2760–2797.
and , “Cyclicality, performance measurement, and cash flow liquidity in private
equity,” Journal of Financial Economics, 2016, 122 (3), 521–543.
Sahlman, William A, “The structure and governance of venture-capital organizations.,”
Journal of Financial Economics, 1990, 27 (2), 473–521.
Sensoy, Berk A, Yingdi Wang, and Michael S Weisbach, “Limited Partner
Performance and the Maturing of the Private Equity Industry,” Journal of Financial
Economics, 2014, 112 (3), 320–343.
Shivdasani, Anil and Yihui Wang, “Did Structured Credit Fuel the LBO Boom?,”
The Journal of Finance, 2011, 66 (4), 1291–1328.
Shleifer, Andrei and Lawrence H Summers, “Breach of trust in hostile takeovers,”
in Alan J Auerbach, ed., Corporate Takeovers: Causes and Consequences, University of
Chicago Press, 1988, pp. 33 – 56.
Smith, Abbie J., “Corporate ownership structure and performance: The case of
management buyouts,” Journal of Financial Economics, 1990, 27 (1), 143 – 164.
Sorensen, Morten and Ravi Jagannathan, “The Public Market Equivalent and
Private Equity Performance,” Financial Analysts Journal, 2015, 71 (4), 43–50.
, Neng Wang, and Jinqiang Yang, “Valuing private equity,” Review of Financial
Studies, 2014, 27 (7), 1977–2021.
Stafford, Erik, “Replicating Private Equity with Value Investing, Homemade Leverage,
and Hold-to-Maturity Accounting,” 2017. Working Paper.
Strömberg, Per, “The new demography of private equity,” in Lerner J Gurung A, ed.,
Globalization of Alternative Investments Working Papers Volume 1: Global Economic
Impact of Private Equity, (World Economic Forum USA, New York), 2008, pp. 3–26.

45
3 | The Accuracy of Net Asset
Values in Private Equity
Evidence from the Secondary Market
∗

We study the demand side of the secondary market for private equity interests using
proprietary data from an intermediary in London. Despite a quarter of the bids coming in
at a par or a premium to the net asset values (NAVs), few lead to a transaction. If NAVs are
reflective of the market value this is puzzling. We find that cross-sectional variation in bids
reflects expectations of future fund returns. Not only are bids informative of future returns,
the bid levels adjust to the point that the return the bidder would experience if their bid
was accepted is unrelated to the bid placed, suggesting a high level of sophistication among
bidders.

1 Introduction
Investors in private equity funds commit to locking their investments into the fund, ceding
control of cash flow timing to the fund manager. The only option to liquidate their stake
early is through the secondary market for private equity fund (PEF) stakes. This market
remained marginal for a long time due to contractual restrictions on transfers (see Lerner
and Schoar, 2004). The increased allocation to private equity, combined with the financial
crisis, exposed an increasing need for liquidity solutions, and the market has grown in
response from $2 billion in the early 2000s to $88 billion in 2019.
Substantial capital is now earmarked to purchasing PEF stakes through the secondary
market. For instance, secondary funds, focused on such transactions, raised over $160
billion between 2011 and 2016.1 Even with this influx of capital, Nadauld et al. (2019)
show that, on average, transactions occur at a discount to a fund’s NAV.2
∗
I wish to express my sincere gratitude to Ludovic Phalippou for countless hours of supervision in
this project. I also wish to thank Tim Jenkinson, Joel Shapiro, and Per Stromberg, as well as seminar
participants at Saïd Business School, Oxford-Man Institute for Quantitative Finance, and the 2nd Private
Markets Conference (Lausanne) for many helpful comments and constructive suggestions. We are grateful
to Elm Capital, and in particular Alberto Badino, for providing us with the data and assisting us in
interpreting it.
1
Other buyers include funds-of-funds, which invest in both the primary and secondary market for PEF
stakes; and various asset owners (investment banks, hedge funds, endowments, pension funds, sovereign
funds), which have their own teams to manage purchases of PEF stakes.
2
Nadauld et al. (2019) report average discounts ranging from an average of 46% in 2009 to 7% in 2014.

46
 A common explanation for the discounts is one of imperfect liquidity provision: when
investors are forced to liquidate their PEF stakes, buyers providing liquidity are able
to acquire assets for less than the underlying value of the asset (Hege and Nuti, 2011;
Kleymenova et al., 2012; Nadauld et al., 2019). This explanation alone cannot account
for all variation in discounts, as we observe a quarter of bids that arrive at par or at
a premium to NAVs. If NAVs reflect the underlying value, it is puzzling that most
of these bids do not lead to a transaction.
An alternative explanation is that variation in discounts reflect expected future
performance, as in the model of closed-end fund discounts by Berk and Stanton (2007).
Expected future performance could reflect beliefs about managerial ability. It could also
reflect a discrepancy between NAVs and fundamental value, owing to the lack of market
valuation of portfolio companies in private equity funds and an aspect of subjectivity
when estimating value (Crain and Law, 2018). Secondary market prices may represent
more accurate estimates of fund value. Boyer et al. (2019) construct a private equity index
based on secondary market transactions, which can be used to examine factor loadings of
private equity returns if secondary market prices reflect the market value of fund stakes.3
This paper asks whether bids are informative of the accuracy of NAVs. The main
findings of the paper is twofold. First, cross-sectional variation in bids predict the future
performance of funds, measured from the first NAV after the bid is posted. Funds receiving
high bids have high post-bid performance, measured as the public market equivalent
(PME) of Kaplan and Schoar (2005). This result does not hold true for variation in
discounts over time, only when comparing cross-sectional variation in discounts between
funds within a given time period. To assess the economic magnitude, we sort funds into
quartiles based on the highest bid received during a quarter. Over a three-year period
following the bid, funds in the quartile receiving the highest bids report a PME of at least
0.103 above funds receiving the lowest bids. This difference is large when put in relation to
an average PME of 1.18 during the entire life of buyout funds, as reported in Harris et al.
(2016). These results refer to an ex post measure of performance, measured from the first
NAV report after the bid date and three years into the future. Additionally, we find that
bids are positively related to performance of funds during the period the bid is posted:
bids are higher for funds that report a higher NAV-to-NAV return during the quarter in
which the bid is placed. Note that this information is not available at the time of the bid.
Second, we find that the relationship between post-bid fund performance and the
bid placed is sufficiently strong to offset the difference in price paid. Investors posting
low bids would not achieve higher returns than those posting high bids, had their bids
3
The index of Boyer et al. (2019) is constructed to represent the returns of an investor that both buys
and sells through the secondary market, but the extent to which their index is relevant for evaluating
factors relevant for returns in private equity overall depend on what discounts reflect.

47
been accepted. The reported gain for those posting low bids is large over short horizons,
while it is offset by the reduced fund performance over longer horizons. Again, this result
only holds within a given time period, not for variation in discounts over time. Together,
these results indicate that investors in the secondary market are able to identify the
relative accuracy of funds’ NAVs. Time-series variation in discounts may reflect varying
liquidity compensation, as suggested by Nadauld et al. (2019).
It may seem as if the results of the paper are obvious. However, the opaque nature
of private equity makes it hard to access any information on the portfolio companies
that make up the value of a fund. While investors in the fund receive regular updates,
potential outside investors do not, complicating any valuation exercise.4 Despite a rapid
growth, the secondary market remains small relative to assets under management of $4.11
trillion (Preqin 2020). In Boyer et al. (2019), the average fund transacts 2.7 times over
the period 2006 through 2018, conditional on trading at all. This market is therefore
characterised by limited price discovery. The results of the paper therefore suggests a
high level of sophistication among participants in the secondary market, echoing findings
in Brown et al. (2019) that investors are not tricked by misreported NAVs.
The secondary market is an over-the-counter market intermediated by specialized
organizations that match buyers and sellers. The data in this study is a proprietary
dataset provided by a London-based sell-side agent for investors wishing to exit via the
secondary market. The data provider gave us access to its entire bid book, which is
comprehensive and representative of the global market: the average bid in our sample each
year is close to the average annual transaction price reported by Greenhill-Cogent. Most
of the bids are made at a discount to NAV, and the discounts vary over time as well as
cross-sectionally. Our dataset consists of 4,242 bids on 488 buyout funds by 144 bidders.
It is important to understand how well the secondary market functions to evaluate
performance in private equity. While buyout funds tend to outperform the public equity
market (Harris et al., 2014, 2016; Robinson and Sensoy, 2016), it is not clear whether this
reflects an alpha or compensation for bearing the illiquidity of private equity investments.
Franzoni et al. (2012) find an alpha of zero when including a liquidity premium, and
the theoretical models in Sorensen et al. (2014) and Bollen and Sensoy (2016) suggest
that the reported performance may be just enough to break even.
A well-functioning secondary market reduces the illiquidity of investments in private
equity, lowering the return required to invest in the primary market. This is shown in
the models of Maurin et al. (2020) and Bollen and Sensoy (2016), where the discount
4
To highlight the inaccessible nature of the data, it is instructive to consider how our data provider
describes the process of selling a fund stake. After an initial interest is found for a given stake, the general
partner of the fund has to agree to a set of potential buyers. Bidders then sign non-disclosure agreements
and are only allowed access to data in “data rooms” set up to provide access for a limited period of time.

48
from the fair value of the fund directly impacts the required return from investing in a
private equity fund. Hence, to understand whether reported returns in private equity
are compensating investors for the risk they take, it is essential to know whether the
secondary market is efficient. This paper suggests that merely looking at the discount
from an NAV to gauge the discount from the fair value is simplistic, as cross-sectional
variation reflects differences in underlying value.
This is one of few empirical studies of the secondary market for PEF stakes. Like us,
Kleymenova et al. (2012) utilize bid-level data. They explore the determinants of bids
in auctions of PEF stakes, finding that bids are lower when fewer bidders participate in
an auction, and higher if the stake is more liquid. Nadauld et al. (2019) is the first to
use transaction data. They document an average discount in the market which increases
in bad economic times. They also find that buyers of PEF stakes earn higher return
than sellers, a difference they attribute to compensation for liquidity provision. Using
the same dataset as employed in this paper, Albuquerque et al. (2018) study liquidity
provision in the secondary market by modelling demand for private equity fund stakes.
They find evidence of compensation for liquidity provision, as demand moves to where
selling pressure is highest. This paper complements existing studies by studying the
informativeness of bid levels, providing evidence that cross-sectional variation predicts
fund performance. In doing so it also provides empirical evidence in favour of the model
of Berk and Stanton (2007), from a setting outside the closed-end fund setting it was
initially developed for. In that it contributes to earlier evidence from secondary market
for hedge funds provided by Ramadorai (2012).
The paper also contributes to the literature examining the accuracy of NAVs by
documenting that secondary market participants are able to spot which NAVs are
misvalued, and price it appropriately. Barber and Yasuda (2017) and Brown et al.
(2019) show that NAVs are not always accurately priced, particularly around fundraising
events. Ang et al. (2018) and Boyer et al. (2019) document significant autocorrelation
in private equity return indices, suggesting that NAVs do not fully incorporate recent
returns. To the extent that secondary markets become more liquid, it provides a potential
avenue to overcome the lack of market valuations.
A limitation of our data is that we work with bids and not transactions. While this
enable us to examine valuations of funds that LPs are not willing to sell, a natural concern
is that some bids may not be serious, and even if bids are serious they are lower than the
price which would lead to a transaction. We take several steps to mitigate this concern.
First, we note that our descriptive statistics closely match those published by Greenhill
Cogent and those in Nadauld et al. (2019) that are based on transaction prices from a
US-based financial intermediary. This similarity is not surprising given that i) bidders

49
submit the same bid to several intermediaries to maximize the probability of matching a
selling interest; and ii) submitting bids is costly both in due diligence and in terms of
reputation. A bidder attempting to manipulate the perception of demand, submitting
unrealistic bids, or reneging on a submitted bid, would be quickly spotted and excluded
from this market by the financial intermediary. Hence, we believe that bids are informative
of the bidder’s true valuation.
Second, we take steps to mitigate concerns arising from bids being systematically
lower than the price a fund would transact at. Let ∆ denote this difference. Evidence in
Hege and Nuti (2011) suggests that this difference varies over time. All of our analysis
therefore include specifications with time fixed effects to soak up this effect. Such time-
variation in ∆ may explain why we find weaker relationship between discounts and future
performance when not controlling for time fixed effects.
Third, while our data is bid-focused it does include a few transactions. For 35 such
transactions we observe earlier bids from the investor that eventually buys the stake. The
final bid placed prior to the completion of a transaction is on average the same price as
the completed transaction. However, we do observe that bids generally increase in the
submitted bid sequence, presumably as the bidder is attempting to convince the seller
by posting bids closer and closer to the bidder’s reservation price. This suggests that
the bid sequence correlates negatively with ∆, and we can test the whether our results
are sensitive to including this proxy from our specification. The results of the paper are
qualitatively unchanged whether we include and exclude this proxy.
Finally, we note that the presence of ∆ is likely to work against us finding the results
of the paper. When we measure the hypothetical return to bidder we use the bid placed
as an initial investment. An increase in ∆ equals a lower bid, leading to a higher return
to the investor. But what we find is that higher bids do not perform any worse than
lower bids.5 There is no such mechanical relationship between the bid and funds’ future
performance, but the presence of ∆ is likely to add noise, making it less likely that we
find the relationships we do find. We conclude that it is unlikely that our results are
driven by the fact that we work with bids rather than transaction prices.
The remainder of this paper proceeds as follows. Section 2 discusses the institutional
features and provides a brief history of the secondary market. Section 3 describes the
dataset and provides descriptive statistics. Section 4 discusses our empirical approach
and the predictions of the paper. Section 5 presents our results. Section 6 concludes.

5
If ∆ is an increasing function of the bid placed it would help us find our results. However, this seems
unlikely given that the mechanical effect of an increase in ∆ is to lower the bid. Note also that we may
have several bids targeting a given fund in specific time period, in which case the ones with the highest ∆
are the lowest bid and it is certain to work against us.

50
2 Institutional Setup

A private equity fund is a private partnership between a group of asset owners (pension
funds, sovereign wealth funds, family offices) called limited partners (LPs), and a private
equity firm (referred to as the general partner, or GP). During the fundraising period,
a GP (e.g. KKR) seeks capital for a given private equity fund (e.g. KKR millennium
fund). Once a fund has reached its fundraising target, it has its “final close”, and the year
this occurs is the fund’s vintage year. After a fund’s final close it is not possible for new
investors to invest in the fund, except by purchasing a fund stake from one of the fund’s
existing LPs on the secondary market. Typically, a sale requires the approval of the GP.
When investors enter a limited partnership agreement, they commit to provide a
certain amount of capital to the fund. The aggregate commitments to a fund is the
fund’s size. No capital is paid other than in connection with fees and the fund making
specific investments, at which point the fund “calls” its investors for capital. Such capital
calls are ex ante uncertain in terms of quantity, amount and timing. Capital is called
project by project. When an investment is exited, the resulting cash is distributed to
the LPs, and cannot (usually) be recycled into new investments.
The difference between fund size and the sum of all capital calls at a given point in
time is called “dry powder” or “unfunded commitments”. The transfer of the limited
partnership interest allows LPs to be released from any remaining unfunded obligations
to the fund. Sellers on the secondary market, therefore, sell a combination of a set of
on-going non-traded investments and the remaining unfunded commitments.
Each quarter, funds self-report an estimate of the value of all on-going investments,
the fund’s net asset value (NAV). At fund inception, NAV is zero and dry powder is equal
to capital committed. As funds have a soft deadline of (usually) 10 years, funds reaching
their tenth anniversary typically have an NAV close to zero. Since the end of 2008, FAS
157 requires that NAVs are marked to market in the US, while such requirements were
in place from 2006 in Europe (Crain and Law, 2018).
Bids on private equity fund interests are denoted as a percentage of a reference NAV.
In general the latest known NAV is used, and the reporting date of that NAV is called
the “reference date”. Any cash flows occurring between the reference date and the date
of a purchase are accounted for when determining the final purchase price: the seller is
reimbursed for capital calls, whereas distributions are kept by the seller and reduce the
purchase price payable. Any interim valuation changes to the underlying fund interests
typically accrue to the benefit - or detriment - of the buyer.

51
2.1 Evolution of the Secondary Market

The Venture Capital Fund of America was the first fund specializing in buying private
equity fund interests on the secondary market; it raised $6 million in 1984. In 2016, Paris-
based Ardian raised the largest secondary fund to date with $10.8 billion of committed
capital for its ASF VII. We note a strong concentration of capital among secondary
funds. From 2011 to mid-2016, the largest 30 fund families raised more than 90% of
a total of $164 billion (the top 10 players have raised more than $127 billion between
them according to Private Equity International).
Global transaction volumes are estimated by Greenhill Cogent (Global Secondary
Market Trends & Outlook 2020). Volumes reported in several of their secondary market
reports are gathered in Figure 1. We see an exponential growth in the early 2000s.
From 2007 to 2013 volumes were comparable at about $25 billion per year (except for
2009, which stands out, with only $10 billion of transactions). There is an increase in
2014 to about $40 billion, with similar numbers achieved in 2015 and 2016. The last
three years are characterised by rapid growth, with transaction volumes reaching a
record $88 billion in 2019.
Buyers of private equity fund interests on the secondary market used to be mainly
private equity funds specializing in buying through the secondary market, called secondary
funds. Over the last ten years, fund-of-funds and asset owners have increased their
participation in the market. Setter Capital, in a 2015 report, estimates that there are
more than 1,000 potential buyers including “opportunistic” secondary buyers (including
pension funds, sovereign wealth funds, asset managers).6

3 Data

The secondary market for private equity fund stakes is intermediated by specialized
organizations that matches potential buyers and sellers. One such financial intermediary -
based in London and operating on that market since September 2009 (its other operations
are older) - gave us access to its entire database. Most of the requests it receives are

6
The first $1 billion secondary transaction occurred in 2000 when Abbey National plc sold $1.3 billion
of stakes in 41 private equity funds (and 16 interests in private companies) to Coller Capital, a secondary
fund. Bank One sold $1 billion of fund interests to Landmark Partners (2004), Dresdner Bank sold $1.4
billion of fund interests to AIG (2005), Mellon Financial Corporation, following the announcement of
Mellon’s merger with Bank of New York selling $1.4 billion of fund interests to Goldman Sachs Vintage
Funds (2006), Ardian’s 2010 acquisition of a $1.9bn portfolio of fund interests from Bank of America
comprising 60 fund interests. In the UK, Lloyds Banking Group plc sold 33 fund interests for $730m to
Lexington Partners and a $2 billion portfolio to Coller. Even pension funds with a long horizon have sold
large portfolios: State of Wisconsin Investment Board sold 12 fund interests managed by six firms for
$1 billion. California Public Employees’ Retirement System, sold $6 billion of fund interests between
2007 and 2012, while later announcing it is moving to the other side of the trade, with an objective of
deploying $600 million a year to purchase fund interests.

52
from LPs offering to buy fund stakes at an indicative price (the bid). In essence, what we
observe is the demand side of the market. Bids are quoted as a fraction of a reference
NAV, and often given as a range (e.g. 80%-85%). We use the midpoint of a range as
the bid level, though all results presented below are qualitatively unchanged to using
the high or low points of the range instead.
Our data provider is viewed as a market leader for individual fund stake transactions
(in contrast to transactions on portfolios of fund stakes.) This is an attractive feature
because it is problematic to infer the valuation of individual fund stakes from the pricing
of a portfolio of fund stakes. Having access to valuations of individual funds allows us
to examine whether bid levels are informative of future fund returns.7
For each bid received, we observe the name of the fund, but not that of the bidders
(only their type and country are provided to us).8 One third of the bids come from the
UK – where our data provider is based. Switzerland is the second largest country in
the sample, followed by the US (10% of the bids), France, Germany, Netherlands and
Norway. The other countries count for less than 1% of the bids each.
We match the funds in our dataset to two databases provided by Preqin. The first
database contains fund characteristics such as size, vintage year, type and geographic
focus. It covers 22412 funds and we have matched funds in our dataset using fuzzy lookup
algorithms and, for funds we could not match this way, manually searching the database.
A manual search is often required due to databases differing in how they record fund names.
It is for example common that we observe an abbreviation in one database and the full
name in the other. The second Preqin database contains cash flows and NAV valuations up
until June 2018. The cash flow database contains significantly fewer funds, so whenever we
rely on it our sample size decreases. The upside of using this database is that having access
to cash flows and intermediate NAV valuations allow us to compute fund performance at
the time a bid is placed, as well as fund performance in the period after a bid is posted.

3.1 Descriptive Statistics

Our initial sample contains 5576 bids made on buyout, venture and some other PE funds
(e.g. infrastructure, fund of funds). Figure 2 shows the number of bids on different
types of PE funds over time, measured at a monthly frequency. We observe that our
sample really starts in October 2009 when we count 35 bids, and that the sample is
mainly made up of buyout funds.

7
The data employed in Nadauld et al. (2019) includes single fund exchanges as well as portfolio
transactions.
8
The funds most represented in our dataset are well-known funds. Those with the highest numbers of
bids are: Apax Europe VII, Bain Capital IX, Blackstone V, and Thomas H Lee VI. They received more
than 30 bids each.

53
 Buyouts represent the largest sub-type of private equity funds in terms of dollar
invested. They also tend to invest in larger companies and we suspect that their reported
NAVs is less noisy, making it perhaps easier to trade on a secondary market. 94% of
the buyout funds in the sample focus on Western Europe (including Scandinavia and
UK) and North America. In order to work with a homogeneous sample, we include in
our sample only buyout funds focusing on these geographies.9
Our working sample contains 4242 bids from October 2009 to October 2016. Whenever
we rely on the Preqin cash flow database, the sample decreases to 2905 bids.10
Panel A of Figure 3 plots the average bid per month. We notice that average bid is
particularly low in the first few months of the sample (reaching 72% in January 2010).
From February 2010 on, the monthly average bid stays in between 76% and 97%.11
We expect the bids received by our financial intermediary to represent serious interests
and valuations of a given fund stake at that given point in time. Although we can never
rule out strategic behavior, it is important to note that submitting bids is costly both in
terms of due diligence and in terms of reputation. A bidder attempting to manipulate
the perception of demand, submitting unrealistic bids, or reneging on a submitted bid,
would be quickly spotted and excluded from this market, as financial intermediaries would
ignore future demand or supply emanating from that organization.
In addition, we observe a number of completed transactions and do not find any
significant deviation between initial bids and eventual transaction prices. Moreover we
compare the time-series of average bids to the time-series of average transaction prices
as reported by the US market leader on the secondary market, Greenhill Cogent. Panel
B of Figure 3 displays these dynamics, and the correlation between the two time-series
is 0.94, with the two averages differing by 1.7 percentage points.12 The average bid
in both samples is low in the early part of the period and displays a slight upward
trend at the end of the sample. Similarly, we compare the time-series of yearly averages
9
Assets of mature businesses should be easier to value than those of venture capital backed companies.
In our sample, venture capital funds experience a much larger discount on average. In conversations
with practitioners a common explanation is that this reflects the higher volatility in venture returns.
However, the significant price difference between buyout and venture prices on the secondary market may
disincentivize investors to sell venture funds. In a market where most buyout funds can be sold at small
discounts or better, LPs that care about accounting losses would prefer to sell buyout funds and hold on
to their venture capital funds.
10
While the drop in bids from 5576 to 4242 is quite dramatic, the reduction is much smaller (we lose
15% of bids) when we merge with the Preqin cash flow database. Venture capital and other PE funds, as
well as funds operating in developing markets, tend to be smaller and are less likely to be included in the
Preqin cash flow database.
11
The large drop in March 2016 is due to our provider only receiving 3 bids on US or Europe buyout
funds for that month, and they were unusually low. They have provided no explanation to why that
particular month is different, and there does not appear to be any specific news at the time which explain
such a drop in activity.
12
Greenhill Cogent, in its Secondary Market Trends & Outlook, reports semi-annual statistics from 2010
to 2013 and annual statistics for 2009, 2014, 2015. The figure averages our bids at the same frequency for
ease of comparison.

54
in our sample to the transaction prices in Nadauld et al. (2019). The correlation is
0.99, and the average discount is the same.
Figure 4 presents the cross-sectional distribution of bids. 11% of bids are made at
par (i.e. 100% of NAV) and about 13% are made at a premium to NAV.13 Additional
descriptives are provided in Figure A.1, which summarizes the distribution of bids by
fund age, fund value, fraction of committed capital called, and the fraction of fund
value distributed. We see that most bids are concentrated on funds in their mid-life,
with very few bids targeting funds in their early years or funds more than ten years
old. Bidders target funds which have called most of committed capital and with most
fund value yet to be distributed.
Table 1 presents key descriptive statistics, with variable definitions provided in Table
B.1. In Panel A, the first three columns present statistics when treating all bids as unique.
There are 4242 bid observations, with an average (median) bid of 88.2% (90.0%). An
alternative approach is to consider all bids received by a fund in a quarter as a single
observation. Columns 4-6 presents descriptives taking this approach instead, summarising
2213 fund-quarter observations. Each fund-quarter observation is defined as the average
bid the fund has received during the quarter. The average (median) bid of these fund-
quarter observations is 89.4% (92.5%). The average time at which a fund receives a
bid is 6.6 years after inception. The last three columns of statistics are for the 488
unique funds that receive at least one bid in our sample, taking averages of all bids
targeting each fund. The average (median) bid is 88.9% and 91.0%. Noteworthy is that
the average fund size is much larger when weighting fund size by the total number of
bids, suggesting that larger funds receive far more bids. The sample is weighted towards
Europe, with 69% of bids aimed at European funds and 31% at US funds. This reflect
the fact that our data provider is based in Europe.
Panel B compare differences in the two samples used in the paper. 488 unique funds
receive bids, which we split up by funds that are covered by the Preqin cash flow dataset
(255 funds) and those that are not (233 funds). The average bid is very similar at 88.6%
of NAV for funds with Preqin cash flow data and 89.2% of NAV for those that are not in
the Preqin cash flow data. Funds covered by Preqin are far more likely to be US-based
and are much larger, corresponding to the overall bias in Preqin towards large, US-based
funds, that arises as Preqin partially sources its data from US pension funds.
For the funds that are covered by Preqin we also provide a comparison between our
funds and comparable funds in the Preqin cash flow database. Funds receiving bids are
larger on average. It is also more common for bids to target funds of well-established

13
As some bids are extreme and may drive results we winsorize at 1% and 99%, which correspond to
bids at 30% and 120% of NAV.

55
GPs, those that have raised more funds and whose funds are generally performing better.
This likely reflect a desire of potential buyers to gain access to reputable funds.14
For this subsample we can formally examine systematic differences between funds
receiving bids in our sample against those that do not, as well as when in a fund’s life bids
are likely to arrive. We define an indicator variable taking the value of 1 if the fund received
a bid in a given quarter, and 0 if it did not. The funds that are included in this analysis
are the 255 funds in our sample which are covered in Preqin, as well as 287 comparable
funds in the Preqin database which do not receive a bid. Table 2 presents results from
estimating a logit model that characterizes the likelihood of receiving a bid within a given
quarter. The three models we run differ only from the set of fixed effects that are included.
The first result of the table is that better performing funds are more likely to be
targeted by bidders, though it is only statistically significant at the 10% level in the
model with no fixed effects. The relevance of past performance increases when controlling
for quarter fixed effects. As past performance varies from one quarter to the other,
this indicates that it is the better performing funds at the time that are more likely
to be targeted by bids. We note that this differs from the results in Nadauld et al.
(2019), where poorly performing funds are the most likely to be sold. Presumably, this
difference arises as their data is seller-initiated whereas ours is buyer-focused. It may
reflect a preference of LPs wanting to sell their worst performing funds, while buyers
are primarily interested in the better performing funds.
In untabulated results, we find that the effect of performance is driven by a reluctance
to target bottom quartile funds in any given quarter, while there is no statistical difference
between the other three quartiles.
Consistent with the descriptive statistics and results in Nadauld et al. (2019), larger
funds are more likely to receive bids than smaller funds. The effect of fund age is non-
linear: it has an inverted U-shape where young and old funds are less likely to receive
a bid, again corresponding to the descriptive statistics of fund age distribution. The
number of funds available by a GP is negatively related to the likelihood of receiving a
bid, though only significant at the 10% level. If investors bid in the secondary market
to gain access to a GP, this result could reflect the increased amount potential funds
through which a bidder could get access to this GP. As the intermediary providing the
dataset is base in London, we find that US funds are less likely to be targeted.

14
The comparison Preqin funds are restricted to be to be buyout funds focusing on either the US or
Europe. Furthermore we require that they have at least one NAV reported at 10% of capital committed
during our sample period, and we only include vintages of 2000 or later. This filtering ensures that we do
not capture funds that would not have been receiving bids during the sample period.

56
4 Research Design

This section outlines the methodology employed in the paper and discusses what we
expect to find from our analysis.

4.1 Determinants of Bid Levels

We begin with a descriptive analysis of the determinants of the bid level. To that end we
will look at correlations between the bid level and fund characteristics, the state of the
economy, and characteristics about the bidders. We estimate the following model,

Bidf,t,i = α + βXf + λXt + θXi + φXf,t + ψXf,i + γXt,i + ξXf,t,i + εi,t , (1)

where Bidf,t,i is the bid placed by investor i at time t for fund f . Bidf,t,i is expressed as a
percentage of the reference NAV. Xf contains fund characteristics that are constant over
time, such as fund size and the geographical focus of the fund. Xt contains time-varying
variables shared by all funds. Those examined here are the Price/Earnings Ratio (following
Nadauld et al., 2019), the Listed Private Equity Price/NAV and Rm (tN AV − 1 year to
tN AV ).15 Xf,t contains time-varying fund variables such as fund performance, dry powder
and fund age. The vectors Xi , Xf,i , Xt,i and Xf,t,i similarly contains variables relating
to the bidder placing the bid such as type of bidder, number of funds bidded for, which
bid in a sequence it is, and the geographical proximity of a bidder to the GP.
The existing literature help us form predictions of what we expect to find. While
there is no direct theoretical model on discounts in the secondary market for private
equity fund stakes, the model of Berk and Stanton (2007) on closed-end fund discounts is
relevant here, as closed-end fund quote prices with discounts from NAVs. Their model
predicts that the price is an increasing function of managerial ability. To the extent that
the current fund performance is informative of managerial ability, we expect a positive
relationship between a fund’s current performance and the price a bidder is willing to pay.
There is plenty of evidence that the values of private equity fund investments move
together with the stock market (e.g. Jegadeesh et al., 2015; Franzoni et al., 2012; Ang et
al., 2018). If private equity funds are slow at incorporating recent movements in their
NAV valuations — as the positive autocorrelation in private equity indices documented in
Ang et al. (2018) and Boyer et al. (2019) suggests — we expect recent market movements
to be positively related to the price a bidder is willing to pay.16 It is possible that
15
The Listed Private Equity Price/NAV is measured in the same way as the bid is, using market prices
of listed private equity funds. We thank the LPX group for providing us with this data. For the public
market return we use regional indices based on the fund location. See Table B.1 for details.
16
If the public market went up (down) recently and NAVs only slowly incorporates this information,
then the NAVs referred to is undervalued (overvalued) and hence the price, as a fraction of NAV, should
be higher (lower).

57
such an effect is picked up by the price/earnings ratio or by the discount on listed
private equity funds. Nadauld et al. (2019) document a positive relationship with the
price/earnings ratio and the price paid, which they interpreted as transaction costs
being higher in worse economic times.
The results of Table 2 suggest that large funds and funds in their mid-life are more likely
to receive bids. To the extent that these funds are generally more desirable, bidders may
be willing to pay more for them. That would be consistent with the findings of Nadauld et
al. (2019). They also note that one reason why there might be a discount for young funds
is the the fact that the buyer takes over the responsibility for future capital calls. As
capital calls have to be met at par, the effective discount is smaller than the quoted one.

4.2 Informativeness of Bids

The theory of Berk and Stanton (2007) suggests that discounts are a function of managerial
ability, as investors are willing to pay a high price when expected future performance
is high. Future (reported) performance also depends on the extent to which NAVs are
accurate estimations of funds value. Boyer et al. (2019) create a private equity index
based on transaction prices on the secondary market, implicitly assuming that market
prices in the secondary market correspond to the correct valuation of funds.
Jenkinson et al. (2019) show that NAVs of buyout funds on average reflect the expected
discounted future cash flows. Although their evidence suggest that NAVs of buyout funds
are correct on average, there may be cross-sectional variation in the accuracy of any given
fund’s NAV. Similarly, there may be points in time when NAVs are more or less accurate,
which would be the case if NAVs are slow in incorporating recent market movements.
Evidence in Barber and Yasuda (2017), Bienz et al. (2017), and Brown et al. (2019) suggest
that NAVs do not always provide a fair representations of the underlying fund value.
We begin by considering a simple model of the accuracy of NAVs to examine if there
are any systematic bias in the accuracy of NAVs, as a function of the state of the economy
or fund-characteristics. We use a methodology similar to that employed in Jenkinson
et al. (2019), which compares discounted cash flows (DCFs) to reported NAVs. The
intuition is that if NAVs are unbiased estimates of fund value, then they should on average
equal the future DCF given an appropriate discount rate. For ease of interpretation,
we measure discrepancies by dividing the DCF with the reported NAV. We estimate
the DCF in the spirit of PME calculations (Kaplan and Schoar, 2005), by discounting
future cash flows with realized public market returns.17
17
Jenkinson et al. (2019) use an ex ante discount rate of 11% when estimating DCFs. As private equity
cash flows are highly sensitive to overall market movements (e.g. Ang et al., 2018), a PME-type measure
takes out some of that expected co-movements. In essence, this captures whether the reported NAV is
over- or under-valued relative to the expected cash flows conditional on the realized market movements.
If such patterns exist related to for example specific macro-economic conditions, we expect secondary
market participants to adjust the prices they are willing to pay relative to reported NAVs.

58
 We specify the following econometric model18
DCFf,t
= α + βXf + λXt + γXf,t + εf,t (2)
N AVf,t

where DCFf,t is the discounted future cash flows for fund f in quarter t. The independent
variables correspond to those used in Table 3, though bidder-specific variables are
not part of this model.
We estimate this model for all fund-quarter observations for buyout funds in the Preqin
database that focus on either Europe or North America. Fund-quarters are included from
the third quarter of 2009. These filter ensure we estimate the model for a comparative
sample to those in our bid dataset. Many of the funds reporting in this time period will
not have been liquidated at the end of our Preqin sample (June 2018), and for those we
include the final NAV as a liquidating distribution when calculating the DCFs. This is a
limitation as we wish to assess how the reported NAVs relate to future cash flows. To
mitigate it we require that we have cash flow data until at least the fund’s sixth year.
A particular NAV is included only if we have at least four quarters of cash flow data
after that NAV report; the NAV is at least 10% of capital committed; and the fund is
not older than 10 years at the time of the NAV report.19
While the previous analysis could potentially help us understand some of the variation
in discounts, it is likely to remain a significant amount of unexplained variation. Partici-
pants in the secondary market may be able to better assess idiosyncratic variation in the
accuracy of NAVs, or the ability of GPs. To the extent they expect competition when
bidding for stakes, their information should feed into what they are willing to pay for a
fund stake. If that is the case, we expect bid levels to predict future fund performance.
To test this prediction we will estimate a model of the future performance of a fund
as a function of the discount in bids placed. The following model is estimated

Fund Performancef,t,i = α + βBidf,t,i + θXf,t,i + εf,t,i (3)

where f, t, i as previously denotes fund f , time t and the bidding investor i. Xf,t,i captures
the same set controls that was included in the analysis of determinants of the bid price.
It is not obvious over which distance one should measure Fund Performancef,t,i . In the
analysis this will be calculated for several different horizons, from one to three years.
18
Our specification uses the ratio of the DCF to the reported NAV, while Jenkinson et al. (2019)
instead use a log specification, with ln DCF on the left-hand side and ln N AV on the right-hand side.
We prefer the ratio as it simplifies the interpretation of the coefficients for other variables. We use the
alternative log specification in an appendix table to show the robustness of the results.
19
The choices of minimum NAV and maximum fund age are motivated by Panel A and Panel C in
Figure A.1 as very few bids are made for funds older than 10 years or with an NAV of less than 10% of
capital committed. In untabulated results we vary these cutoffs on the restrictions and find qualitatively
similar results.

59
Fund performance is defined as the PME of the fund, measured from the closest NAV
report following the date of the bid, until a point in time defined by the horizon. For
the PME calculation, the first post-bid NAV is accounted for as an initial investment,
while the NAV at the end point is accounted for as a liquidating distribution. It therefore
reflects a measure of the fund’s reported performance, measure in a period strictly after
the period in which the bid is placed.
While it is hypothesised that fund performance is positively related to the bid placed,
a natural follow-on question is whether the pricing is rational, in the sense that bidders
are indifferent to buying a fund at a premium to NAV versus a discount. To evaluate
this we can measure a bidder’s performance similarly to the fund performance measure,
but using the submitted bid value as an initial investment in place of the fund’s NAV.
As any inaccuracies in NAVs or managerial ability is only likely to be resolved over time,
we expect the bid amount to be negatively related to a bidder’s performance in short
horizons due to (paper) valuation gains. However, if the bids are sufficiently informative
of future fund performance, we expect no differences in performance over longer horizons
if bids are rational, akin to efficient pricing in the market.
To that end, we estimate the following model,

Bidder Performancef,t,i = α + βBidf,t,i + θXf,t,i + εf,t,i (4)

where Bidder Performancef,t,i is computed analogously to Fund Performancef,t,i , but

using the bidded amount as an initial investment rather than the closest NAV following
the bid.20 In this way it corresponds to the hypothetical return that the bidder would
report, had the bid been accepted. We hypothesize that the bid placed will be unrelated
to the bidder’s performance over longer horizons.
Several bids may be placed for a given fund around the same time, and bids are not
evenly distributed over time but arrive in large numbers in certain months. We therefore
going to double-cluster the standard errors by the month in which the bid is made, as
well as by fund ID. To account for overall time trends in both bids and fund performance
we include time fixed effects and vintage fixed effects in some specifications.

4.3 Bids, not Transactions

It is important to discuss one limitation of our dataset: it primarily consists of bids rather
than completed transactions. While this enable us to evaluate valuations of funds that
LPs are unwilling to sell, a natural concern is that bids are lower than transaction prices

20
The bidded amount is the reference NAV times the bidded percentage, accounting for any cash flows
in between. The seller is reimbursed for capital calls, whereas distributions are kept by the seller and
reduce the price payable.

60
and that this may drive results. While we cannot change the nature of the data, we can
think about potential issues that may arise and take steps to mitigate them.
Let ∆f,t denote the difference between the value at which fund f would transact at
time t to the bid placed for the fund at that time. Our analysis is largely focused on
cross-sectional variation in bids, which means that the problem is limited if ∆f,t = ∆,
i.e. is constant between funds and over time. This assumption appears rather strict and
unlikely to hold. We know from Hege and Nuti (2011) that the divergence between the
amount sellers asked and bidders offered increased substantially around the financial
crisis compared to both prior to and following the crisis.
If instead ∆f,t = ∆t , i.e. varies over time but is a constant across funds, we can
account for this difference with the inclusion of time fixed effects. All tables presented
therefore include specifications with and without time fixed effects, to examine whether
the results are materially different.
Of course, we cannot rule out that it varies not only over time, but between funds and
perhaps between bidders as well, if some investors place bids closer to their reservation price
than others. When analysing whether bids are informative of funds’ future performance,
an increase in the difference ∆f,t , all else equal, introduces noise which will work against
us finding any statistically significant results. In the case of the analysis on bidders’
performance, it would bias our estimates downwards. We provide a more detailed
discussion when we present the results of the paper.
We note that our average bid closely matches the average transaction price, and is
much higher than the average bid, in Nadauld et al. (2019) for the time period over
which the two studies both have data. Panel B of Figure 3 suggests that our descriptive
statistics also matches those reported by Greenhill Cogent.
There are 35 transactions for which we observe an earlier bid from an LP that later
acquires the fund stake, allowing us to follow the negotiation process. For those cases we
identify the last bid prior to the transaction. In 26 cases the transaction price is identical
to the amount bidded, in 4 cases the transaction price is higher (on average less than 1
percentage point) and in 5 cases the transaction occurs at a lower value than the prior
bid (on average less than 1 percentage point). For cases where we observe multiple prior
bids we find that bids generally increase during a bid sequence, which we interpret as the
result of a negotiation process. This increase is on average 1 percentage point.
One lesson from this exercise is that it may be worthwhile to explicitly control for where
in a bid sequence a bid is placed, as bidders may be increasing their bids in an attempt to
convince sellers. We therefore construct a “bid sequence” variable, defined as the number
of bids placed by a given investor for a specific fund in a given sequence. As this is expected
to capture part of variation in ∆f,t , it acts as a proxy for ∆f,t , allowing us to examine
whether our results are significantly altered by including it. Though imperfect, we find
that our results remain qualitatively similar whether this proxy is included or excluded.

61
5 Results
5.1 Determinants of Bids

We begin by examining the determinants of the bid (quoted as a fraction of NAV). Results
are presented in Table 3. To make the coefficients easier to interpret, the dependent
variable takes the value of 70 to represent a bid of 70% of NAV.
The first column includes fund characteristics and bidder-specific variables. The
fund characteristics included are fund age and fund size. Age is represented by three
indicator variables denoting if a fund is a young fund (up to 3 years old), middle-aged
(between 4 and 7 years old, omitted from the table), or an old fund (8 years or older).
Size is split into four categories: small, medium (omitted), large, and very large. The
size dummies are suppressed from the table for presentation purposes, as none of the
dummies are statistically significant. Larger funds receive more bids, but the price
investors are willing to pay does not appear be related to fund size. Young funds
command a higher discount compared to middle-aged funds, with the discount being
6 to 8 percentage points higher, depending on specification.21 Old funds receive bids
with a discount of about 2.5 percentage points above middle-aged funds, though the
difference is not statistically significant in all models.
We include four bidder-specific variables. Two indicator variables tests for differences
in price paid by the type of bidder, classified as being a specialist secondary fund, a
fund-of-funds, or an asset owner category that group together pension funds, endowments
and other investors. Asset owners tend to place higher bids than the other two investor
types, possibly in an attempt to gain access to a particular GP they wish to invest with.
To capture negotiation rounds, or attempts by bidders to convince an LP to sell by
offering prices closer and closer to their reservation price, we include the log of a bidder’s
bid sequence, defined as the number of bids placed by a bidder on a specific fund in the
past 180 days. As expected, the variable is positively related to the price submitted by
the bidder.22 Finally, to examine whether there is a potential home bias we include an
indicator variable for whether the bidding investor and the GP have are based in the

21
The discount on young funds partly reflects that uncalled capital has to be met fully by the buyer, as
noted in Nadauld et al. (2019). In untabulated analysis we calculate the effective discount as the offered
dollar discount as a fraction of quoted NAV and uncalled capital. In this specification, young funds are
not associated with a significantly different discount. However, note that simply adding dry powder in the
model, as in column 5, does not drive out the result on young funds.
22
As discussed in section 4.3, the variable is expected to be related to the discrepancy between the
price at which a transaction would take place and the bidded amount. As such, comparing coefficient
estimates of other variables when it is included and excluded can inform us whether having bids rather
than transaction prices appear to introduce a bias in the estimates. In untabulated results we find
qualitatively similar results for all other variables when omitting the bid sequence variable.

62
same country. Bids for funds managed by local GPs are approximately one percentage
point higher, though the coefficient is not always statistically significant.23
In the second column we add monthly fixed effects but keep the same set of explanatory
variables. This has little impact on the estimated coefficients, but does increase R2 from
0.30 to 0.40, suggesting that there is significant time-variation in bid levels. In the
third column we replace time fixed effects with macro-variables to examine what this
time-variation is driven by. The return on the stock-market over the twelve months
running up to the NAV report is strongly positively related to bid. This may be due to a
belief that NAVs have not fully incorporated recent market movements. In untabulated
results we find that the effect is there for recent market movements over shorter horizons
as well. The effect of market movements is not absorbed by the price-earnings ratio of
the stock market or the discount on listed private equity, perhaps the closest substitute
to secondary market prices. The discount on listed private equity is highly correlated
with the price-earnings ratio.24 All three of these variables are positively related to the
bid price, though the listed private equity discount is only statistically significant at
the 10% level, and only for the final two specifications.
The specifications in the fourth and fifth columns are estimated on the sample for
which we have cash flow data from Preqin available. Column four confirms that the
previously documented relationships hold in this subsample. In the fifth column we add
fund characteristics gathered in Preqin. We find that bids are higher for funds that have
done well, as measured by the to-date PME, consistent with the Berk and Stanton (2007)
theory. However, the bid is lower when a higher fraction of that performance is realized
(i.e. already distributed), perhaps reflecting a belief that well-performing funds have
conservative valuations, as found by Brown et al. (2019), providing the buyer with an
upside when investments are exited. Note that dry powder is unexpectedly positively
related, even though we expected a negative relationship. Dry powder is highly related to
the (young) age dummies, and in untabulated results we find that when age dummies
are excluded the coefficient on dry powder takes the expected negative sign.

23
Home bias among investors is documented in for example French and Poterba (2016) and Coval
and Moskowitz (1999). A preference for investing in GPs that are geographically closer could reflect an
information advantage. Evidence of informational advantages in investments is provided by Hau (2001),
Ivkovic and Weisbenner (2005) and Bae et al. (2008). In the private equity setting, Hochberg and Rauh
(2013) document home-state bias of LPs in private equity, which is particularly pronounced for pension
funds that underperform with local investments.
24
In untabulated results we collect a large set of macroeconomic variables and test various permutations
of them. The set of macrovariables is the one used in Albuquerque et al. (2018), with some additions.
The set includes credit spreads, interest rate measures, various measures of liquidity proposed in the
literature, GDP growth, measures of global economic policy uncertainty, and the VIX. In addition to the
three presented in the table we find that only the VIX is significantly related to the discount. However,
the VIX does not survive a horse-race against the price-earning ratio or the discount of listed private
equity funds, and is consequently not presented in the specification.

63
 Having examined factors that drive the discount we turn to the analysis of whether
these factors are associated with expected over- and undervalued NAVs, measured as
the ratio of the discounted future cashflow to the reported NAV. Table 4 presents the
results of this exercise. The variables included are those used in Table 3, excluding the
bidder-specific variables. We demean all explanatory variables and subtract 1 from the
dependent variable for ease of interpretation of the coefficients. A positive (negative)
intercept indicates that NAVs are below (above) the DCF on average. The first column
includes only fund-characteristics, while the second to fourth examines the macro-variables
in some more detail. The final column includes fund performance, measures of dry powder,
and the realized part of performance.
The results on fund characteristics resemble those in Jenkinson et al. (2019), with
NAVs being conservative (aggressive) when the fund is young (old), and the largest funds
being the most conservative. Large funds tend to be raised by the most successful and
reputable GPs, which are shown by Barber and Yasuda (2017) and Brown et al. (2019)
to have more conservative valuations. The large coefficient on young funds reflects low
NAVs as not all capital has yet been invested, while the average PME of funds in the
analysis is positive, resulting in the DCF being much higher than the reported NAV that
only reflects investments undertaken to-date. When controlling for dry powder in column
(5), the coefficient on young funds decreases though it does not disappear.
The economic variable that best captures divergence of NAVs and future DCF is the
listed private equity premium. The listed private equity premium is defined in the same
fashion as the bids in our data: price of listed private equity divided by the reported NAV.
The positive coefficient suggests that when the price (discount) on listed private equity is
high (low), NAVs in private equity are more (less) conservatively valued. This echoes the
finding in Jegadeesh et al. (2015), that returns on listed private equity forecasts returns on
indices constructed using reported NAVs and cashflows. The return on the public market
in the previous year is also positively related to the NAV, albeit statistical significance is
diminished when controlling for the listed private equity premium. I therefore provide
some weak evidence that NAVs lag in incorporating market returns.25
To alleviate concerns that our results are due to the specific modelling choice of
the ratio DCF/NAV, we perform a robustness check by using a log-log specification,
as in Jenkinson et al. (2019). The results are presented in Table A.1, which confirms
the patterns observed in the main specification.
25
The model is estimated for the same sample period over which the bid data is available, from Q3
2009. This implies that fair value requirements in FAS 157 have been imposed on NAVs (see Crain and
Law, 2018). In a previous version of the paper the model was estimated for a longer horizon, comparing
the periods prior to and following the implementation of fair value requirements. While the results on
fund characteristics were qualitatively similar in that model, NAVs were more conservatively valued prior
to the implementation of FAS 157.

64
 Comparing the results of Table 4 to those in Table 3, there are at least three noteworthy
observations. First, similarly to Nadauld et al. (2019) we find that bids are higher (lower)
in good (bad) times. They interpret this as transaction costs being high in bad times.
The results of this paper suggest that there may be a complementary reason: NAVs may
not fully incorporate changes in the valuations in a timely fashion, and the secondary
market participants adjust their bids accordingly. Second, while better performing
funds attract higher bids, there is no evidence here that this is motivated by identifying
undervalued NAVs. Third, while young funds receive lower bids, they also have the
most conservatively valued NAVs. This largely reflect the effect of uncalled capital:
the NAVs are low by construction for young funds as most investments are yet to be
made, and as noted by Nadauld et al. (2019), the effective discount is higher than the
quoted as capital calls have to be met in full.

5.2 Returns Post Bidding

We now turn to a more direct analysis of how well the bidders separate the wheat from
the chaff. To this end we will examine whether bids predict funds’ performance post
bidding. To ensure that we truly measure performance after the bid is placed we start
measuring it from the first NAV report after the date the bid was placed. The results
are presented in Table 5. We measure the performance post-bidding over three different
horizons: 1, 2, and 3 years. For each horizon, the first column includes only the bid, the
second adds vintage and monthly fixed effects, and the third add the control variables
used in earlier tables, including bidder characteristics. For ease of presentation, only
the coefficients on the bid is displayed.
Panel A displays the result using the bid placed measured as a continuous variable.
The bid variable takes the value of 0.90 for a bid that is 90% of NAV. Every unique bid
is linked to the post-bid performance of the fund bidded for, conditional on that fund
having cash flow data available in Preqin. In Panel B, the unit of observation changes to a
fund-quarter observation. Fund are sorted into quartiles each quarter based on the highest
bid received during the quarter. We repeat the estimation in Panel A, replacing the
continuously measured bid level with indicator variables for the quartile a fund belongs to.
The omitted category is the quartile of funds receiving the lowest bid within that quarter.26
The results suggest that the bid level is informative of a fund’s future performance,
in line with the prediction of the Berk and Stanton (2007) model of closed-end fund
discounts. Interestingly, in Panel A this effect is largely present only when controlling for
time fixed effects, which compares the relative performance of funds receiving lower or

26
The average bid for funds in the bottom quartile is 74.0% of NAV, for the 2nd lowest quartile it is
88.2%, for the 2nd highest quartile 95.0%, and for the highest quartile 99.9%.

65
higher bids within a quarter. This indicates that bidders are able to distinguish which
funds are worth more at a given point in time, but time-series variation in the discount
does not reflect the time-series variation in fund returns.
The results in Panel B suggests an almost monotonic effect for funds’ ranked higher
on bids, though there is essentially no difference in future performance between the
two quartiles containing funds receiving the highest bids. The coefficients allow for
a simple interpretation of the economic magnitude, corresponding to an increase in
post-bid fund performance for a given quartile of funds relative to the quartile of funds
receiving the lowest bid. The estimates suggest that funds receiving the highest bids
have PME of 0.057 to 0.058 higher on a one-year horizon, 0.068 to 0.072 higher on a
2-year horizon, and 0.103 to 0.125 higher at a three-year horizon. These effects are
sizeable compared to estimates in Harris et al. (2016) that buyout funds on average
have a PME of 1.18 over the entire fund life.
Having established that, within a given time period, the price bidders are willing to
pay predict future fund performance, we turn to the question of whether these two effects
cancel out. An investor should be happy to pick up a fund that performs relatively poor,
as long as the discount is sufficiently high. In the short horizon we may expect the paper
gain from buying at a discount to dominate even if stakes are fairly priced. For example,
if a GP consistently values NAVs conservatively, the undervaluation will only be resolved
at exits which may take place several years in the future. Similarly, if a premium is paid
for a higher ability GP, it likely takes time for that higher ability to translate to value
increases. Indeed, our analysis on funds’ performance suggests that higher bids predict
larger cumulative performance differences as the horizon increases.
Table 6 presents results of the relationship between bids and the hypothetical
performance of bidders, had their bid been accepted at the time of the bid. The price
paid by the bidder is assumed to be the bid price times the reference NAV, accounting
for any intermediate cash flows between the reference NAV date and the bid date. The
performance of the bidder is otherwise measured in the same way as the performance
of funds. Panel A presents the results using the continuously measured bid value, while
Panel B provides estimates using the quartile indicator variables that sort funds into
quartiles based on the highest bid received in a quarter.
There are two main takeaways from the table. First, the bid is negatively related
to bidders’ performance for short horizons, but this negative relationship disappears
almost entirely at longer horizons. The quartile ranking specification even suggests
that those willing to pay more would perform better at longer horizons, as the increase

66
in fund performance outweighs the higher price.27 Second, the negative relationship
between the bid placed and the performance of the bidder in Panel A only disappears
when including time fixed effects, and thus compare high and low bids within a given
time-period. The quartile ranking does this by construction, and there is a significant
difference between the coefficient estimates in Panel A between models with and without
time fixed effects. This suggests that cross-sectional variation in discounts predicts
cross-sectional variation in fund performance, whereas time-variation in discounts do
not correspond to time-variation in fund performance.
One concern may be that the patterns we observe are due to the changed sample when
observing different horizons, rather than the change in horizon itself. To alleviate such
concerns, Table A.2 repeats Table 6, using only observations for which the dependent
variable exists at the 3-year horizon. The results are qualitatively unchanged. Similar
results are found in untabulated results for funds’ future performance.
One may ask whether bids of all investor types are equally informative. Table 7
examines whether there are different sensitivities of bid prices to post-bid fund performance
for different types of investors. The results indicate that bids by asset owners are the
ones most sensitivity to funds’ future performance. This could reflect that secondary
funds and fund-of-funds build up large, diversified portfolios and are therefore interested
in being present for any liquidity-shocked LP, offering to buy a large amount of stakes.
Secondary funds are by far the most active in the market, with asset owners being the
least active. However, it should be cautioned that these results should be interpreted with
great caution, as the coefficient for secondary funds is not statistically significantly smaller
in any specification, and only at the 10% level for fund-of-funds in some specifications.
So far we have shown that bid levels are positively related to funds’ performance up
until the point of the NAV reference date, and that bids are informative of future fund
performance, measured from the first post-bid NAV. There is one additional window over
which fund performance may impact the cross-sectional variation in the discount: valuation
changes between the reference NAV date and the bid date. NAVs are typically reported
quarterly, while bids may arrive at any point. To the potential buyers can estimate the
likely fund returns in that period, it should reflect the price they are willing to pay.
We construct two measures to examine whether this is incorporated in the bid price.
The first is a linearly interpolated measure of the fund return between the reference NAV
27
At first glance it appears puzzling that the quartile ranking again suggests an almost monotonic
increase in bid bid ranking at the longest horizon, while there is still a negative (albeit small and
insignificant) coefficient on the bid placed even when including time fixed effects. The difference is
attributable to the difference in treating multiple observations for a given fund around the same time. In
Panel A, all unique bids are treated as individual observations. Thus, if a fund receives multiple bids
within the same quarter, the lower bids will by construction be measured as achieving higher performance.
Of course, in an auction process the lowest bid is not the one accepted. By contrast, in Panel B at most
one fund-quarter observation is included, using the highest bid received in the quarter.

67
date and the first NAV after the bid. We assume that the ex post realised return of
the fund accrued linearly over the two NAV reports, so if the period tN AV to tbid is one
third of the time between the two NAV reports, one third of that return had accrued
at tbid . Clearly, this measure uses information not available to the bidder at the time of
the bid, as it use information from a NAV valuation at a date that has not yet taken
place, and even less been reported. The second measure is the movement on public
markets between the reference NAV date and the bid date.28 Unlike the first measure,
this measure is available to the bidder at the time the bid is placed. As valuations of
private equity funds move with the public stock market we expect this to be positively
related to the price investors are willing to pay.
Table 8 adds these two variables to the final specification on bid determinants used in
Table 3. As the in-between fund return requires the availability of an NAV both prior
to and following the bid we lose some observations. Surprisingly, the results suggest
that movements in the public market since the reference NAV date are not positively
related to the bid. It is not only statistically insignificant, but the coefficient has a
negative sign. By contrast, the unobserved intermediate fund performance is strongly
positively related to the bid. The effect is not taken out by either including time fixed
effects or the in-between market movements. In fact, the in-between market movements
become significantly negatively related to the bid when both are included, though only
statistically significant at the 10% level.
Taken together, the results of the paper that bidders in the secondary market are
sophisticated, with cross-sectional variation in bids reflecting expected future fund returns.
They also appear to incorporate unreported intermediate fund returns. This suggests
that simply looking at a discount to evaluate the attractiveness of a stake is misleading,
as it may reflect an inaccurate NAV. Overall, this provide some evidence in favour of the
approach taken in Boyer et al. (2019). It should be noted that the results of this paper
are strongest for cross-sectional variation in discounts at any given time. By contrast,
time-series variation in discounts does not appear to be strongly related to realized future
fund returns, and may reflect variation in required compensation for providing liquidity.
As a final point of the analysis we return to whether our results may be driven by
the dataset consisting of bids rather than transaction prices. As before, we use ∆f,t to
denote the difference between the price at which a stake would transact and the bid price.
Regarding the results for bidders’ performance, an increase in ∆f,t implies a lower bid
and hence a lower initial investment which, all else equal, implies higher performance
for the bidder. Thus, to the extent that the ∆f,t differs among funds it would bias the
28
As in other instances where we use public market returns, we tailor the return to the focus of the
fund. The S&P500 is used for funds focusing on North America, Eurostoxx 600 for funds focused on
Europe, and FTSE 250 for UK funds.

68
relationship between the bid and the bidders’ performance downwards. As the main
result is that there the negative relationship almost disappears over longer horizons, ∆f,t
works against that finding and suggests that if anything, the estimate is likely biased
downwards.29 For the funds’ performance, variation in ∆f,t has no mechanical impact,
but adds noise which would work against us finding any relationship between the bid
and future fund performance. We conclude that it is unlikely that our results are driven
by the fact that we examine bids rather than transaction prices.

6 Conclusions
This paper examines the informativeness of bids placed in the secondary market for
private equity fund stakes. The secondary market has grown rapidly over the last two
decades, and capital earmarked to be deployed in the market has increased significantly.
Despite this influx of capital, the average transaction occurs at a discount to the net
asset value of the fund (Nadauld et al., 2019).
A common explanation for the discount is that of compensation to the buyers for
providing liquidity (e.g. Maurin et al., 2020; Nadauld et al., 2019). While liquidity provision
is likely playing a role (Albuquerque et al., 2018), it cannot be the full story: a quarter of
bids arrive at par or at a premium, but still rarely lead to a transaction. In this paper we
explore an alternative explanation motivated by the theory of closed-end fund discounts by
Berk and Stanton (2007): discounts are informative of future expected fund performance.
The paper finds that cross-sectional variation in bids predicts cross-sectional variation
in future fund performance. Sorting funds into quartiles based on the highest bid received
during a quarter shows that the quartile of funds receiving the highest bids outperform the
quartile of funds receiving the lowest bid by at least 0.102 over three years, measured as
a public market equivalent (PME). This is a substantial difference given that the average
buyout fund generates a PME of 1.18 over the entire life of a fund. We further find that bids
adjust to the point where the hypothetical return to a bidder if a bid would be accepted is
not increasing with the size of the discount, except for paper gains at very short horizons.
The results of the paper suggests a high level of sophistication among market
participants, despite the scarcity of data available to evaluate private equity funds.
This echoes findings in Brown et al. (2019) that investors in private equity funds are
not duped by manipulated net asset values. Recent work by Boyer et al. (2019) explore
transactions to construct a private equity index, an effort that is complicated by the
low frequency of fund transactions. An interesting avenue for future research is whether
29
This is not true if ∆f,t is an increasing function of the bid placed. While we cannot guarantee that
this is not the case, it seems unlikely as the mechanical impact of shrinking ∆f,t is an increase in the bid.
We also note that a fund may receive multiple bids in a given time period, and among those the ∆ will be
negatively related to the bid.

69
information in bids, essentially making up a limit order book, could be used to improve
the accuracy of such index.

70
 Figure 1: Global Secondary Transaction Volume

The figure displays annual volumes in the secondary market for private equity fund stakes, as estimated by Greenhill Cogent. Numbers compiled from annual reports
of Greenhill Cogent’s “Secondary Market Trends & Outlook” published between 2012 and 2020.
71
 Figure 2: Number of Bids per Month

The figure plots the number of bids per month for buyout funds, venture capital funds and other funds. The series begins in March 2009 and ends in October 2016.
72
 Figure 3: Average Bid over Time (% of NAV)

Panel A displays the monthly average bid on buyout funds from October 2009 until October 2016, together
with the total number of bids received in the month. They share a joint scale on the y-axis, with bids
displayed as a percentage of NAV, while the number of bids are a total count. Panel B displays the
average high bid in our sample compared to the average high bid as published by Greenhill Cogent 2016.
We define the ’high bid’ as the highest bid a fund receives within a month. The frequency of observations
from Greenhill Cogent varies over time. For 2010 through 2013 we have values semi-annually while in
2009, 2014 and 2015 we only observe it once per year. The average numbers presented for our sample is
calculated over the matching time intervals. The last point in the figure corresponds to the the first half
of 2015.
Panel A

Panel B

73
 Figure 4: Dispersion of Bids

The histogram displays the fraction of bids that are made within a given bid range in our sample. Each bid range should be read as being ≥ low number and < the
higher number. The exceptions for this are the ’100’ category, which contains only bids that are exactly 100%, the ’100-105’ category, which contains bids that are
strictly above 100 and below or equal to 105, and the < 40 and > 105 bins which contains all bids that are strictly below and above those numbers, respectively.
74
 Table 1: Descriptive Statistics

This table reports summary statistics on the funds in the secondary market as well as the average bid. Panel A presents the results weighted by unique bids, by
fund-quarter observations or by unique funds. A fund-quarter observation is defined by taking the average of all bids on a given fund within a quarter and treating
that as one fund-quarter observation. A fund observation is defined by taking the average of all bids for a fund as one observation. Panel B presents a comparison of
funds in the Preqin dataset and funds in our sample. The first column uses all funds in our sample for which we have a match in the Preqin cash flow database. The
second column contains funds in our sample which do not exist in the Preqin database. The third column presents corresponding characteristics for funds in the
Preqin database which are not part of our sample. The comparison funds from the Preqin database are buyout funds focusing on either US or Europe, they are
required to have an NAV reported of at least 10% of capital committed during our sample period and they have a vintage of 2000 or later.

Panel A
All Bid Observations Fund-Quarter Observations Fund Observations
Mean Median Std Mean Median Std Mean Median Std
Bid 88.2 90.0 12.6 89.4 92.5 11.7 88.9 91.0 11.3
Fraction US Funds 0.31 0 0.46 0.36 0 0.48 0.39 0 0.49
75

Fund Size ($ billion) 4.8 3.5 4.87 4.3 2.8 4.59 2.3 0.9 3.25
Fund Age (years) 6.6 7 2.7 6.6 7 2.93 n/a n/a n/a

Number of Observations 4242 4242 4242 2213 2213 2213 488 488 488

Panel B
Funds in Preqin Dataset Yes Excluded Yes
Funds in our Sample Yes Yes Excluded
Average Bid 88.6 89.2 n/a
Fraction US Focused Funds 0.56 0.21 0.78
Fund Size ($ billion) 3.3 0.9 0.9
Number of Funds in Fund Family 8.1 n/a 5.0
Fraction Low Reputation Fund Family 0.06 n/a 0.09
Fund Performance (PME) 1.16 n/a 1.09

Number of Funds 255 233 287

 Table 2: Probability of Receiving a Bid

This table presents the estimates of a logit model of the probability that a private equity fund receives a
bid in a quarter. The dependent variable takes the value of 1 in any quarter that a fund receives at least
one bid, and zero otherwise. The coefficients show the change in this probability for an infinitesimal
(discrete) change in each continuous (binary) variable. The marginal effects are evaluated at the variables’
sample means. The standard errors used to calculate z-statistics are reported in parenthesis below each
estimate. The model is estimated including a constant and the independent variables are winsorized at
the 1% level. The estimation is performed on fund-quarters in Preqin between the third quarter of 2009
and the third quarter of 2016. Only vintages of 2000 or later are included. For a specific fund-quarter
combination to be included we require that the fund is no older than 12 years old and that the NAV is at
least 10 % of committed capital. All variables are defined in Table B.1. The ’Fraction in Sample’ is the
fraction of fund-quarter observations in which we observe a bid. ***, **, and * indicate significance at the
1%, 5% and 10% level, respectively.

Fund Past Performance (PME) 0.012∗ 0.026∗∗∗ 0.019∗∗∗

(0.0068) (0.0061) (0.0062)
Fund Size (log) 0.070∗∗∗ 0.061∗∗∗ 0.059∗∗∗
(0.0023) (0.0020) (0.0019)
Fund Age 0.037∗∗∗ 0.035∗∗∗ 0.041∗∗∗
(0.0034) (0.0030) (0.0073)
Fund Age Squared −0.003∗∗∗ −0.003∗∗∗ −0.003∗∗∗
(0.0003) (0.0003) (0.0003)
Number of Funds in Fund Family −0.001∗ −0.001∗ −0.001∗
(0.0005) (0.0005) (0.0005)
Low Reputation Fund Family −0.009 −0.008 −0.006
(0.0081) (0.0075) (0.0067)
US-focused (1/0) −0.136∗∗∗ −0.118∗∗∗ −0.120∗∗∗
(0.0061) (0.0054) (0.0055)

Quarter Fixed Effects No Yes Yes

Vintage Fixed Effects No No Yes
Fraction in Sample 0.127 0.127 0.127
Number of Observations 10752 10752 10752
Pseudo R2 0.301 0.354 0.359

76
 Table 3: Determinants of the Bid

The dependent variable is the bid (% of latest reported NAV). A bid of 70% takes the value 70. “Bid
Sequence” is the number of bids placed by an investor on the fund in the last 180 days, including this bid.
“Same HQ Region” is an indicator variable taking the value of 1 if the head quarter of the bidder and
the fund are in the same country, and 0 otherwise. Suppressed from the table are indicator variables for
fund size, classifying funds as small, medium (omitted), large, or very large. All variables are defined
in Table B.1. Vintage fixed effects are included for all models. Column 3 includes month fixed effects,
which is replaced by time-varying variables in columns 4 through 6. All variables are winsorized at the 1%
level. ***, **, and * indicate significance at the 1%, 5% and 10% level, respectively. Standard errors are
clustered by fund ID and by month, and are reported in parenthesis below each estimate.

(1) (2) (3) (4) (5)

Fund Age ≤ 3 −8.21 ∗∗∗
−6.02 ∗∗∗
−5.79 ∗∗∗
−7.28 ∗∗∗
−7.91∗∗∗
(1.44) (1.08) (1.31) (1.30) (1.88)
Fund Age ≥ 8 0.11 −1.29 −2.37∗∗ −2.50∗∗ −2.74∗∗
(1.12) (0.99) (1.03) (1.00) (1.14)
(log) Bid Sequence 2.14∗∗∗ 2.25∗∗∗ 2.03∗∗∗ 1.83∗∗∗ 1.79∗∗∗
(0.59) (0.52) (0.60) (0.68) (0.66)
Same HQ Region 1.16∗∗ 0.73 1.07∗∗ 0.95∗ 0.88
(0.53) (0.56) (0.51) (0.53) (0.55)
Bidder is a Fund-of-Funds −7.12∗∗∗ −6.23∗∗∗ −6.90∗∗∗ −5.11∗∗∗ −5.07∗∗∗
(0.91) (1.11) (0.97) (1.00) (0.99)
Bidder is a Secondary Fund −5.45∗∗∗ −4.67∗∗∗ −5.44∗∗∗ −4.45∗∗∗ −4.17∗∗∗
(0.68) (1.06) (0.91) (0.89) (0.83)
Rm (tN AV − 1 year to tN AV ) 8.54∗∗∗ 8.38∗∗ 7.53∗∗
(1.99) (3.49) (3.42)
Price/Earnings Ratio 0.73∗∗ 0.81∗∗ 0.88∗∗
(0.33) (0.36) (0.39)
Listed Private Equity Discount 22.96 23.62∗ 23.64∗
(14.35) (13.89) (13.91)
Distributions to Paid-in Capital −4.22∗∗
(1.72)
Dry Powder (%) 4.96∗
(2.92)
Fund Past Performance (PME) 6.51∗∗∗
(1.96)
Adjusted R2 0.30 0.40 0.34 0.34 0.35
Number of Observations 4242 4242 4242 2905 2905
Fund Size Dummies Yes Yes Yes Yes Yes
Vintage Fixed Effects Yes Yes Yes Yes Yes
Month Fixed Effects No Yes No No No

77
 Table 4: Funds’ Discounted Cash Flows to Net Asset Value ratio

The dependent variable is the ratio of the fund’s discounted cash flow to the reported Net Asset Value,
minus one. All explanatory variables are demeaned. To be included in the estimation, the fund is required
to have an NAV reported of at least 10% of committed capital, the fund should not be older than 10
years, we require that there is data on the fund until at least the fund’s sixth year and that there is
at least four quarters of cash flow data after the NAV date. Only European and US buyout funds are
included in the estimation. We only include NAVs reported in Q3 2009 or later, matching our bid sample.
All variables are defined in Table B.1. All variables are winsorized at the 1% level. ***, **, and * indicate
significance at the 1%, 5% and 10% level, respectively. Standard errors are clustered by fund ID and by
quarter, and are reported in parenthesis below each estimate.

(1) (2) (3) (4) (5)

Intercept 0.11 ∗∗∗
0.11 ∗∗∗
0.11 ∗∗∗
0.11 ∗∗∗
0.11∗∗∗
(0.03) (0.02) (0.02) (0.02) (0.02)
Fund Age ≤ 3 0.35∗∗∗ 0.36∗∗∗ 0.36∗∗∗ 0.36∗∗∗ 0.13∗∗∗
(0.04) (0.04) (0.04) (0.04) (0.04)
Fund Age ≥ 8 −0.06∗∗ −0.07∗∗ −0.07∗∗∗ −0.07∗∗∗ −0.05∗
(0.03) (0.03) (0.03) (0.03) (0.03)
Small Fund −0.04 −0.04 −0.04 −0.04 −0.04
(0.04) (0.04) (0.04) (0.04) (0.04)
Large Fund −0.06 −0.06 −0.06 −0.06 −0.05
(0.05) (0.05) (0.05) (0.05) (0.06)
Very Large Fund 0.11∗∗ 0.11∗∗ 0.11∗∗ 0.11∗∗ 0.12∗∗
(0.05) (0.05) (0.05) (0.05) (0.05)
Rm (tN AV − 1year to tN AV ) 0.30∗∗∗ 0.20 0.18∗ 0.19∗
(0.10) (0.13) (0.10) (0.11)
Equity Market P/E Ratio 0.01 −0.00 −0.00
(0.01) (0.02) (0.02)
Listed PE Price/NAV 0.59∗∗∗ 0.75∗∗ 0.78∗∗
(0.17) (0.32) (0.33)
Distributions/Paid-in Capital 0.10∗
(0.05)
Dry Powder 0.55∗∗∗
(0.11)
Fund Performance (PME) −0.08
(0.07)

Adj R2 0.16 0.17 0.18 0.19 0.21

N 6412 6412 6412 6412 6412

78
 Table 5: Funds’ Performance Post Bidding

The dependent variable is the Funds’ Performance Post Bidding, defined as the fund’s reported PME between the NAV closest following the bid date and 1, 2 or 3
years later. The NAV closest following the bid date is accounted for as an initial investment and the final NAV as a liquidating distribution. The public market index
used depends on whether it is a European, US or UK fund. Control variables included are those used in Table 3, though time-varying variables are excluded to allow
for time fixed effects to be included. Definitions of all variables can be found in Table B.1. Panel A presents result using the “Bid Placed” variable as a continuously
measured bid level, while Panel B replaces it with dummies sorting funds into quartiles based on the highest bid received by a fund within a quarter. The dependent
variable in Panel A is a fund’s post-bidding performance for each unique bid, while Panel B it is restricted to one fund-observation per quarter. All variables are
winsorized at the 1% level. ***, **, and * indicate significance at the 1%, 5% and 10% level, respectively. Standard errors are clustered by fund ID and by month, and
are reported in parenthesis below each estimate.
Panel A: Bid measured as a continuous variable
1 Year Post Bidding 2 Years Post Bidding 3 Years Post Bidding
Bid Placed 0.166∗ 0.143∗∗∗ 0.138∗∗ 0.031 0.283∗∗∗ 0.245∗∗∗ 0.023 0.339∗∗∗ 0.324∗∗∗
(0.091) (0.052) (0.064) (0.122) (0.078) (0.073) (0.101) (0.103) (0.102)
Adj R2 0.01 0.33 0.40 0.00 0.35 0.43 0.00 0.03 0.41
N 2149 2149 2149 1521 1521 1521 1133 1133 1133
79

Vintage Fixed Effects No Yes Yes No Yes Yes No Yes Yes

Time Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Controls No No Yes No No Yes No No Yes

Panel B: Funds sorted into quartiles based on bids received in a quarter

1 Year Post Bidding 2 Years Post Bidding 3 Years Post Bidding
2nd Lowest Bid Quartile 0.015 0.019 0.015 0.023 0.029 ∗
0.023 0.037 ∗
0.045∗∗ 0.038∗
(0.011) (0.012) (0.012) (0.015) (0.015) (0.016) (0.021) (0.021) (0.021)
2nd Highest Bid Quartile 0.055∗∗ 0.057∗∗ 0.055∗∗ 0.069∗∗ 0.073∗∗ 0.069∗∗ 0.100∗∗∗ 0.121∗∗∗ 0.103∗∗
(0.027) (0.027) (0.028) (0.031) (0.031) (0.033) (0.038) (0.039) (0.041)
Highest Bid Quartile 0.057∗∗ 0.058∗∗ 0.057∗∗ 0.068∗∗ 0.072∗∗ 0.069∗∗ 0.105∗∗∗ 0.125∗∗∗ 0.103∗∗∗
(0.025) (0.025) (0.027) (0.030) (0.029) (0.031) (0.034) (0.034) (0.036)
Adj R2 0.01 0.03 0.12 0.01 0.04 0.15 0.02 0.04 0.16
N 1022 1022 1022 723 723 723 509 509 509
Vintage Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Time Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Controls No No Yes No No Yes No No Yes
 Table 6: Bidders’ Performance

The dependent variable is the “Bidder’s Performance”, defined as the return the bidder would have experienced had the bid been accepted. It is calculated as a
PME of the cash flows from the bid date up until the NAV reported 1, 2 or 3 years later, counting from the closest NAV following the bid date. The final NAV is
accounted for as a liquidating distribution and the bidded amount, adjusted for cash flows in between the NAV date referred to by the bid and the bid date, as an
initial investment. The public market index used depends on whether it is a European, US or UK fund. Control variables included are those used in Table 3, though
time-varying variables are excluded to allow for time fixed effects to be included. Definitions of all variables can be found in Table B.1. Panel A presents result using
the “Bid Placed” variable as a continuously measured bid level, while Panel B replaces it with dummies sorting funds into quartiles based on the highest bid received
by a fund within a quarter. The dependent variable in Panel A is the bidder’s performance post-bidding for each unique bid, while Panel B it is restricted to one
fund-observation per quarter. All variables are winsorized at the 1% level. ***, **, and * indicate significance at the 1%, 5% and 10% level, respectively. Standard
errors are clustered by fund ID and by month, and are reported in parenthesis below each estimate.

1 Year Post Bidding 2 Years Post Bidding 3 Years Post Bidding

Bid Placed −0.317∗∗∗ −0.312∗∗∗ −0.331∗∗∗ −0.446∗∗∗ −0.117 −0.151∗ −0.409∗∗ −0.013 −0.017
(0.109) (0.073) (0.078) (0.165) (0.113) (0.094) (0.202) (0.204) (0.127)
Adj R2 0.03 0.33 0.40 0.04 0.37 0.44 0.30 0.30 0.41
80

N 2149 2149 2149 1521 1521 1521 1133 1133 1133

Vintage Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Time Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Controls No No Yes No No Yes No No Yes

Panel B: Funds sorted into quartiles based on bids received in a quarter

1 Year Post Bidding 2 Years Post Bidding 3 Years Post Bidding
2nd Lowest Bid Quartile −0.004 −0.002 −0.002 −0.003 0.019 0.017 −0.002 0.031 0.027
(0.020) (0.026) (0.028) (0.031) (0.036) (0.037) (0.047) (0.052) (0.051)
2nd Highest Bid Quartile −0.029 −0.021 −0.021 0.002 0.023 0.021 0.021 0.061 0.059
(0.026) (0.029) (0.030) (0.056) (0.061) (0.061) (0.054) (0.058) (0.058)
Highest Bid Quartile −0.032 −0.025 −0.026 0.002 0.020 0.019 0.020 0.059 0.056
(0.025) (0.031) (0.032) (0.052) (0.054) (0.056) (0.052) (0.063) (0.062)
Adj R2 0.01 0.03 0.12 0.01 0.04 0.15 0.02 0.04 0.16
N 1022 1022 1022 723 723 723 509 509 509
Vintage Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Time Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Controls No No Yes No No Yes No No Yes
 Table 7: Funds’ Performance Post Bidding by Investor Types

The dependent variable is the “Funds’ Performance Post Bidding”, defined as the fund’s reported PME between the NAV closest following the bid date and 1, 2 or 3
years later. The NAV closest following the bid date is accounted for as an initial investment and the final NAV as a liquidating distribution. The public market index
used depends on whether it is a European, US or UK fund. Control variables included are those used in Table 3, though time-varying variables are excluded to allow
for time fixed effects to be included. Definitions of all variables can be found in Table B.1. The bid placed variable is interacted with the type of investor, divided into
secondary funds, other fund-of-funds, and asset owners (reference group) that group together all other investors such as pension funds, endowments and similarly. All
variables are winsorized at the 1% level. ***, **, and * indicate significance at the 1%, 5% and 10% level, respectively. Standard errors are clustered by fund ID and
by month, and are reported in parenthesis below each estimate.

1 Year Post Bidding 2 Years Post Bidding 3 Years Post Bidding

Bid Placed 0.301 ∗∗∗
0.286 ∗∗∗
0.281 ∗∗∗
0.074 0.411 ∗∗∗
0.369 ∗∗∗
0.023 0.452∗∗∗ 0.443∗∗∗
(0.103) (0.087) (0.094) (0.134) (0.123) (0.127) (0.151) (0.160) (0.158)
Bid Placed × Secondary Fund −0.149 −0.146 −0.138 −0.101 −0.167 −0.154 −0.076 −0.175 −0.166
(0.096) (0.089) (0.092) (0.104) (0.134) (0.126) (0.156) (0.147) (0.145)
Bid Placed × Fund-of-Funds −0.142 −0.141∗ −0.143∗ −0.116 −0.158∗ −0.151∗ −0.081 −0.186∗ −0.185∗
81

(0.091) (0.078) (0.078) (0.108) (0.086) (0.081) (0.120) (0.104) (0.105)

Adj R2 0.03 0.33 0.40 0.04 0.37 0.44 0.30 0.30 0.41
N 2149 2149 2149 1521 1521 1521 1133 1133 1133
Vintage Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Time Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Controls No No Yes No No Yes No No Yes
 Table 8: Bid Levels and Intermediate Fund Returns

The dependent variable is the bid (% of latest reported NAV). A bid of 70% takes the value 70. “Fund
Return (tN AV to tBid )” is a measure of the fund’s (unreported) return between the date of the reference
NAV (the most recent NAV) and the date of the bid. The return is a linearly interpolated return measure,
where the return is the ex post realized return between the reference NAV and the first NAV following the
bid date. The return is calculated as the increase in NAV between the two dates, net of any intermediate
cash flows and as a percentage increase from the reference NAV. Rm (tN AV to tBid ) is measured as
the market return between the reference NAV date and the bid date, using a regionally specified index
depending on the fund’s geographical focus. See Table B.1 for definitions of index used. Control variables
are included in all models, and include the set of variables included in model (5) of Table 3, except the
fourth column which drops the time-varying economic variables and instead includes month fixed effects.
Vintage fixed effects are included for all models. All variables are winsorized at the 1% level. ***, **, and
* indicate significance at the 1%, 5% and 10% level, respectively. Standard errors are clustered by fund ID
and by month, and are reported in parenthesis below each estimate.

(1) (2) (3) (4)

Fund Return (tN AV to tBid ) 20.33∗∗∗ 22.61∗∗∗ 21.68∗∗∗

(5.67) (5.53) (5.74)
Rm (tN AV to tBid ) −4.78 −11.22∗
(4.82) (6.34)

Adjusted R2 0.38 0.35 0.38 0.42

Number of Observations 2788 2788 2788 2788
Controls Yes Yes Yes Yes
Vintage Fixed Effects Yes Yes Yes Yes
Month Fixed Effects No No No Yes

82
Appendix
A Appendix Figures and Tables

83
 Figure A.1: Characteristics of Funds Targeted by Bidders

The histograms display fund characteristics of funds that receive a bid. Panel A documents the distribution of bids targeting funds of a specific age. Panel B
documents the distribution of the ratio of NAV to committed capital at the time of the bid. Panel C presents the fraction of committed capital that has been called at
the time of the bid, while Panel D presents the fraction of total fund value, including distributions, that has been distributed at the time of the bid.

Panel A Panel B
84

Panel C Panel D
 Table A.1: Funds’ Discounted Cash Flows to Net Asset Value - Alternative Specification

The dependent variable is the log of the fund’s discounted future cash flow (DCF). Statistical significance
for the coefficient on (log) NAV is testing whether the coefficient is significantly different from 1. All
models include an intercept (not displayed in the table). To be included in the estimation, the fund is
required to have an NAV reported of at least 10% of committed capital, the fund should not be older
than 10 years, we require that there is data on the fund until at least the fund’s sixth year and that there
is at least four quarters of cash flow data after the NAV date. Only European and US buyout funds are
included in the estimation. We only include NAVs reported in Q3 2009 or later, matching our secondary
sample. All variables are defined in Table B.1. All variables are winsorized at the 1% level. ***, **, and *
indicate significance at the 1%, 5% and 10% level, respectively. Standard errors are clustered by fund ID
and by quarter the NAV is reported in, and are reported in parenthesis below each estimate.

(1) (2) (3) (4) (5)

(log) NAV 0.95 0.94 ∗
0.94 ∗
0.94 ∗
1.06
(0.03) (0.03) (0.03) (0.03) (0.05)
Fund Age ≤ 3 0.19∗∗∗ 0.22∗∗∗ 0.20∗∗∗ 0.21∗∗∗ 0.07∗
(0.04) (0.04) (0.04) (0.04) (0.04)
Fund Age ≥ 8 −0.09∗∗∗ −0.11∗∗∗ −0.11∗∗∗ −0.11∗∗∗ −0.05
(0.03) (0.03) (0.03) (0.03) (0.03)
Small Fund −0.06 −0.06 −0.06 −0.06 −0.07
(0.05) (0.05) (0.05) (0.05) (0.06)
Large Fund −0.01 −0.01 −0.01 −0.01 0.00
(0.04) (0.04) (0.04) (0.04) (0.04)
Very Large Fund 0.13∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.14∗∗∗
(0.05) (0.05) (0.05) (0.05) (0.05)
Rm (tN AV − 1year to tN AV ) 0.24∗∗∗ 0.15∗ 0.15 0.15
(0.09) (0.09) (0.10) (0.11)
Equity Market P/E Ratio 0.02∗∗∗ 0.01 0.01
(0.01) (0.03) (0.03)
Listed PE Price/NAV 0.77∗∗∗ 0.73∗∗ 0.72∗∗
(0.15) (0.34) (0.34)
Distributions/Paid-in Capital 0.09
(0.06)
Dry Powder 0.54∗∗∗
(0.13)
Fund Performance (PME) −0.05
(0.08)

Adj R2 0.59 0.60 0.60 0.60 0.61

N 6412 6412 6412 6412 6412

85
 Table A.2: Bidders’ Performance - Common Sample for all Horizons

The dependent variable is the “Bidder’s Performance”, defined as the return the bidder would have experienced had the bid been accepted. It is calculated as a
PME of the cash flows from the bid date up until the NAV reported 1, 2 or 3 years later, counting from the closest NAV following the bid date. The final NAV is
accounted for as a liquidating distribution and the bidded amount, adjusted for cash flows in between the NAV date referred to by the bid and the bid date, as an
initial investment. The public market index used depends on whether it is a European, US or UK fund. Control variables included are those used in Table 3, though
time-varying variables are excluded to allow for time fixed effects to be included. Definitions of all variables can be found in Table B.1. Panel A presents result
using the “Bid Placed” variable as a continuously measured bid level, while Panel B replaces it with dummies sorting funds into quartiles based on the highest bid
received by a fund within a quarter. The dependent variable in Panel A is a fund performance post-bidding for each unique bid, while Panel B it is restricted to one
fund-observation per quarter. All variables are winsorized at the 1% level. ***, **, and * indicate significance at the 1%, 5% and 10% level, respectively. Standard
errors are clustered by fund ID and by month, and are reported in parenthesis below each estimate.

1 Year Post Bidding 2 Years Post Bidding 3 Years Post Bidding

Bid Placed −0.467 ∗∗∗
−0.308 ∗∗∗
−0.311 ∗∗∗
−0.543 ∗∗
−0.112 −0.117 −0.409 ∗∗
0.013 −0.017
(0.168) (0.085) (0.078) (0.217) (0.111) (0.102) (0.202) (0.204) (0.127)
Adj R2 0.07 0.43 0.50 0.05 0.40 0.49 0.03 0.30 0.41
86

N 1133 1133 1133 1133 1133 1133 1133 1133 1133

Vintage Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Time Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Controls No No Yes No No Yes No No Yes

Panel B: Funds sorted into quartiles based on bids received in a quarter

1 Year Post Bidding 2 Years Post Bidding 3 Years Post Bidding
2nd Lowest Bid Quartile −0.005 −0.002 −0.002 −0.004 0.020 0.019 −0.002 0.031 0.027
(0.029) (0.033) (0.034) (0.037) (0.040) (0.041) (0.047) (0.052) (0.051)
2nd Highest Bid Quartile −0.032 −0.024 −0.023 0.001 0.021 0.019 0.021 0.061 0.059
(0.029) (0.033) (0.035) (0.059) (0.062) (0.061) (0.054) (0.058) (0.058)
Highest Bid Quartile −0.046∗ −0.029 −0.030 0.002 0.021 0.020 0.020 0.059 0.056
(0.027) (0.034) (0.035) (0.054) (0.056) (0.056) (0.052) (0.063) (0.062)
Adj R2 0.01 0.03 0.12 0.01 0.04 0.15 0.02 0.04 0.16
N 509 509 509 509 509 509 509 509 509
Vintage Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Time Fixed Effects No Yes Yes No Yes Yes No Yes Yes
Controls No No Yes No No Yes No No Yes
B Variable Definitions

Table B.1: Variable Definitions

Variable Description

Dependent Variables
Bid (% of NAV) The bid placed for a particular fund, expressed as a percentage
of the latest reported NAV. The variable will take the value 80
for a bid of 80%. The variable is used in Table 3.

Funds’ Discounted Cash The ratio is defined as DCFt /N AVt , where N AVt is the reported
Flows to NAV ratio NAV at time t and DCFt is the discounted future cash flows,
discounted to time t. DCFt is calculated as the net present
value of all cash flows of the fund that take place after t. If
the fund is not fully liquidated at the end of our sample (June
2016), the final NAV has been incorporated in the calculation
as a distribution. The discount rate used is the public market
return between t and the cash flow. The index used is S&P500
for North American funds, Eurostoxx 600 for European funds,
and FTSE 250 for UK funds.

Funds’ Performance Post The fund’s performance post bidding is defined as the public
Bidding (X Years) market equivalent of the fund between the reported NAV closest
following the bid date, and the reported NAV X years later. The
first NAV is accounted for as an initial investment and the final
NAV as a distribution. The calculation of the public market
equivalent is explained in the definition of PME.

Bidders’ Performance (X The bidder’s performance is defined as the public market equiva-
Years) lent of all cash flows between the bid date and ending X years
later. In the case the date X years later does not coincide with
a date at which the NAV is reported, the closest NAV following
that date is used as the final time in the calculations. The price
paid by the bidder is used as an initial investment. The price
paid is the bid % times the NAV referred to by the bid, adding
the any capital calls and subtracting any distributions made in
between the NAV date and the bid date. The NAV at the final
date is accounted for as a distribution. The calculation of the
public market equivalent is explained in the definition of PME.

Bid and LP Variables

Bidder is a Fund-of-Funds Takes the value of 1 if the bid was made by a fund-of-funds, and
0 if it was made by a secondary fund or a asset owner.

Bidder is a Secondary Fund Takes the value of 1 if the bid was made by a secondary fund,
and 0 if it was made by a fund-of-funds or a asset owner.

Bidder is an Asset Owner Takes the value of 1 if the bid was made by a asset owner, and
0 if it was made by a fund-of-funds or a secondary fund. The
types of LPs included in the asset owner category are: insurance
companies, banks, asset managers, government agencies, pension
funds, foundations, endowments and others.

(Continued)

87
 Table B.1−Continued

Variable Description

Bid Placed The bid placed for a particular fund, expressed as a fraction of
the latest reported NAV.

Bid Sequence on Fund by Bidder (log) The logarithm of the number of bids the bidder has placed on
this particular fund in the last 180 days.

Fund HQ Region is the same Takes the value of 1 if the LP and GP have head quarters in the
as that of the Bidder same region, and zero otherwise. The region is defined as either
US, Continental Europe, UK, or the rest of the world.

Fund Variables
Age Number of years since the fund’s first capital call.

Fund Size (log) The logarithm of the fund size as reported in Preqin.

Large Fund Takes the value of 1 if the fund size is more than $1.5 billion but
less than $5 billion.

Medium Fund Takes the value of 1 if the fund size is more than $0.5 billion but
less than $1.5 billion.

Number of Funds in Fund Family The logarithm of the number of funds raised by the GP previously.

PME The fund’s to-date public market equivalent (PME). It is the

ratio of the discounted value of all distributions to date (including
the latest available NAV in case the fund is not liquidated in the
end of our sample) to the discounted value of all capital calls to
date. The discount rate used is the public market return, where
the index used is S&P500 for North American funds, Eurostoxx
600 for European funds that are not UK funds, and FTSE 250
for UK funds.

Small Fund Takes the value of 1 if the fund size is less than $0.5 billion.

US Fund Takes the value of 1 if the fund is a US fund, and 0 otherwise.

Very Large Fund Takes the value of 1 if the fund size is more than $5 billion,
corresponding to the 90th percentile of funds in the Preqin
database over the period of our study.

Economy Variables
Equity Market Price/Earnings Ratio The aggregate market price/earnings ratio in
a given month, provided by Robert Shiller at
http://www.econ.yale.edu/ shiller/data.htm. The measure used
is the cyclically adjusted price/earnings ratio (CAPE).

(Continued)

88
 Table B.1−Continued

Variable Description

Listed Private Equity Price/NAV A monthly index of the price/NAV, minus 1, of the listed private
equity index LPX50 which is provided by the LPX group. The
measure corresponds to the percentage premium, or discount, of
listed private equity. The constituents of the index for which no
NAV is available is excluded from the calculation of this index.

Rm (tN AV to tBid ) The public market return between the NAV date and the Bid
date. This variable use a regionally specified index depending
on the fund’s geographical focus. The index used is S&P500 for
North American funds, Eurostoxx 600 for European funds, and
FTSE 250 for UK funds.

Rm (tN AV − 1 year to tN AV ) The public market return over the year prior to the NAV date.
For the secondary data set the ’NAV date’ is the NAV date
referred to in the bid. The index used is S&P500 for North
American funds, Eurostoxx 600 for European funds, and FTSE
250 for UK funds.

89
References

Albuquerque, Rui, Johan Cassel, Ludovic Phalippou, and Enrique Schroth,

“Liquidity Provision in the Secondary Market for Private Equity Fund Stakes,” 2018.
Working Paper.
Ang, Andrew, Bingxu Chen, William N. Goetzmann, and Ludovic Phalippou,
“Estimating Private Equity Returns from Limited Partner Cash Flows,” The Journal
of Finance, 2018, 73 (4), 1751–1783.
Bae, Kee-Hong, René M. Stulz, and Hongping Tan, “Do local analysts know
more? A cross-country study of the performance of local analysts and foreign analysts,”
Journal of Financial Economics, 2008, 88, 581–606.
Barber, Brad M and Ayako Yasuda, “Interim Fund Performance and Fundraising in
Private Equity,” Journal of Financial Economics, 2017, 124 (1), 172–194.
Berk, Jonathan B and Richard Stanton, “Managerial Ability, Compensation, and
the Closed-End Fund Discount,” Journal of Finance, 2007, 62 (2), 529–556.
Bienz, Carsten, Charlotte Wallem Rakner, and Anne Therese Rasmussen,
“Strategic Valuation Management in Private Equity Fundraising,” 2017. Working Paper.
Bollen, Nicolas P B and Berk A Sensoy, “How much for a haircut? Illiquidity,
secondary markets, and the value of private equity,” 2016. Fisher College of Business
Working Paper.
Boyer, Brian, Taylor D. Nadauld, Keith Vorkink, and Michael S. Weisbach,
“Private Equity Indices Based on Secondary Market Transactions,” 2019. Dice Center
WP 2018-21.
Brown, Gregory W., Oleg R. Gredil, and Steven N. Kaplan, “Do private equity
funds manipulate reported returns?,” Journal of Financial Economics, 2019, 132 (2),
267 – 297.
Coval, Joshua D. and Tobias J. Moskowitz, “Home Bias at Home: Local Equity
Preference in Domestic Portfolios,” The Journal of Finance, 1999, 54, 2045–2073.
Crain, Nicholas G and Kelvin K F Law, “The Bright Side of Fair Value Accounting:
Evidence from Private Company Valuation,” 2018. Working Paper.
Franzoni, Francesco, Eric Nowak, and Ludovic Phalippou, “Private Equity
Performance and Liquidity Risk,” Journal of Finance, 2012, 67 (6), 2341–2373.
French, Kenneth R . and James M . Poterba, “American Economic Association
Investor Diversification and International Equity Markets,” The American Economic
Review, 2016, 81, 222–226.
Harris, Robert S., Tim Jenkinson, and Steven N. Kaplan, “Private equity
performance: What do we know?,” Journal of Finance, 2014, 69 (5), 1851–1882.
Harris, Robert S, Tim Jenkinson, and Steven N Kaplan, “How Do Private Equity
Investments Perform Compared to Public Equity?,” Journal of Investment Management,
2016, 14 (1), 1–24.

90
Hau, Harald, “Location Matters: An Examination of Trading Profits,” The Journal of
Finance, 2001, 56, 1959–1983.
Hege, Ulrich and Alessandro Nuti, “The Private Equity Secondaries Market During
the Financial Crisis and the ”Valuation Gap”,” Journal of Private Equity, 2011, 14 (3),
42–54.
Hochberg, Yael V and Joshua D Rauh, “Local Overweighting and Underperformance:
Evidence from Limited Partner Private Equity Investments,” Review of Financial
Studies, 2013, 26 (2), 403–451.
Ivkovic, Zoran and Scott Weisbenner, “Local Does as Local Is: Information Content
of the Geography of Individual Investors’ Common Stock Investments,” The Journal of
Finance, 2005, 60, 267–306.
Jegadeesh, Narasimhan, Roman Kräussl, and Joshua M Pollet, “Risk and
Expected Returns of Private Equity Investments: Evidence Based on Market Prices,”
Review of Financial Studies, 2015, 28, 3269–3302.
Jenkinson, Tim, Wayne R. Landsman, Brian R. Rountree, and Kazbi
Soonawalla, “Private Equity Net Asset Values and Future Cash Flows,” The Accounting
Review, 2019, 95 (1), 191–210.
Kaplan, Steven N. and Antoinette Schoar, “Private Equity Performance: Returns,
Persistence, and Capital Flows,” The Journal of Finance, 2005, 60 (4), 1791–1824.
Kleymenova, Anya, Eli Talmor, and Florin P Vasvari, “Liquidity in the Secondaries
Private Equity Market,” 2012. Working paper, London Business School.
Lerner, Josh and Antoinette Schoar, “The illiquidity puzzle: theory and evidence
from private equity,” Journal of Financial Economics, 2004, 72 (1), 3–40.
Maurin, Vincent, David T. Robinson, and Per Strömberg, “A Theory of Liquidity
in Private Equity,” 2020. Working paper.
Nadauld, Taylor D., Berk A. Sensoy, Keith Vorkink, and Michael S. Weisbach,
“The liquidity cost of private equity investments: Evidence from secondary market
transactions,” Journal of Financial Economics, 2019, 132 (3), 158 – 181.
Ramadorai, Tarun, “The Secondary Market for Hedge Funds and the Closed Hedge
Fund Premium.,” Journal of Finance, 2012, 67 (2), 479–512.
Robinson, David T. and Berk A. Sensoy, “Cyclicality, performance measurement,
and cash flow liquidity in private equity,” Journal of Financial Economics, 2016, 122
(3), 521–543.
Sorensen, Morten, Neng Wang, and Jinqiang Yang, “Valuing private equity,”
Review of Financial Studies, 2014, 27 (7), 1977–2021.

91
4 | The Dynamics of Pay-for-
Performance Sensitivity in
Private Equity Funds∗

I explore the dynamics of pay-for-performance sensitivity in private equity funds arising

from carried interest. Carry pay is conditional on beating a hurdle rate, making pay-for-
performance sensitivity increase in fund performance. I show that within-fund variation
in pay-for-performance sensitivity is positively related to company growth in portfolio
companies acquired in buyouts. The effect is stronger when the importance of the direct
component relative to the indirect component of pay for performance is high. Exogenous
variation is generated by utilizing public market shocks, which impact the NAV of all private
equity funds but have a heterogeneous impact on pay-for-performance sensitivity due to
funds’ varying degrees of capital invested and the non-linear payoff structure of carried
interest. My findings provide evidence of the importance of incentives for value creation in
private equity.

1 Introduction

Resolving agency conflicts between agents and principals is an important topic in economics
and finance. One tool for alleviating such conflicts is to align incentives through the use
of performance-sensitive compensation packages. Examples include equity grants and
stock options to CEOs, performance-based fulcrum fees in mutual funds, and carried
interest in hedge funds and private equity (PE) funds.
Private equity takes an active role in the companies they are invested in, and with $4.1
trillion of assets under management in 2019 (Preqin 2020) the industry is an important
actor in the economy. It is therefore important to understand how well carried interest
achieves its intended purpose. Compensation to managers of PE funds (general partners,
or GPs) usually includes a mix of fixed and performance-based fees (Metrick and Yasuda,
2010). The fixed fee stems primarily from management fees (typically 1.5% to 2.5%
of committed capital), while the performance-based fees stems mainly from carried
∗
I thank Thomas Hellmann, Tim Jenkinson, Kasper Meisner Nielsen, Thomas Noe, Ludovic Phalippou,
and Martin Schmalz, as well as seminar participants at the 2019 FMA European Conference for many
helpful comments and constructive suggestions. I am grateful to Brad Barber and Ayako Yasuda for
generously sharing their classification of private equity funds into high- and low-reputation GPs.

92
interest, where the GP earns a fraction (usually 20%) of fund profits if fund performance
exceeds a hurdle rate. Some argue that it is one of the drivers of strong PE returns
(e.g. Jensen, 1989; Kaplan and Strömberg, 2009).
A GP’s payoff from carried interest is similar to a call option, with the hurdle rate
as its strike price. The sensitivity of pay to performance corresponds to the delta of
this option. The hurdle rate is fixed and set at fund inception, leading the delta to
vary with fund performance. As with call options, the delta is low (high) when a
fund is far out of (in) the money.
A PE fund does not invest all the committed capital at once. Instead, it makes
staggered investments during an investment period that normally spans five years.
Consequently, when making investments late in the investment period, the delta depends
on the success of earlier deals. Proponents of PE claim that they add value to their
portfolio companies. Presumably, this takes effort on behalf of the GP, and agency theory
predicts that we expect higher effort when the sensitivity of pay to performance is higher.
In this paper, I ask how variation in a GP’s pay-for-performance sensitivity relates to
operational improvements at the portfolio companies acquired in buyouts.
The main finding of the paper is that there is a positive relationship between the growth
rate of companies acquired in buyouts and the delta of the acquiring PE fund. The results
suggest that a one standard deviation increase in delta is associated with between 2.7% and
3.5% higher annual revenue growth, measured from the pre-buyout period and up to four
years post-buyout. Delta is not related to improved profitability. While estimates suggest
a positive relationship, it is statistically insignificant and the economic magnitude is small.
Earlier studies document that one way PE facilitates growth in portfolio companies is
through relaxing credit constraints, allowing for increased investment (Boucly et al., 2011;
Cohn et al., 2020, e.g.). Interestingly, the observed positive relationship between company
growth and delta does not reflect capital spending. There is not only a lack of significant
relationship between delta and asset growth in the target company, the point estimates
are consistently negative. This suggests that the results likely reflect organic growth.
This study utilizes a dataset of UK companies acquired in a buyout to measure
improvements in operational performance. All companies in the UK, public and private,
are required to submit filings to Companies House, the UK national registrar, which makes
these filings publicly available. This setting provides two distinct advantages for the study
at hand. First, a representative dataset containing both pre- and post-buyout financials
can be assembled.1 Second, UK companies submit annual ownership lists, allowing a
1
Availability of data on privately held companies is often scarce. Consequently, researchers often
utilizes samples containing companies with publicly available financials (e.g. Guo et al., 2011), or samples
made available to them by individual investors (e.g. Acharya et al., 2013). Such samples are not necessarily
representative. Cohn et al. (2014) document that among public-to-private buyouts, those with publicly
available financials perform better than those without. I am not the first to utilize the UK setting for
these reasons. Bernstein et al. (2019) examine how PE-backed companies fared during the financial crisis,
while Chung (2011) focuses on the impact of PE ownership in private-to-private buyouts.

93
company to be linked to the acquiring fund. This is crucial as the delta is estimated at
the fund level, while transaction databases commonly only list the acquiring PE firm.
To estimate the delta of GP pay I build on the framework developed by Metrick
and Yasuda (2010). They provide a model for estimating the expected revenue to a GP,
of which carried interest is one part. A PE fund consists of investments in portfolio
companies with simulated return paths. Using a set of contractual features, the resulting
stream of cash flows is allocated to the fund’s investors (limited partners, or LPs) and
the GP. In equilibrium, GPs must add value to portfolio companies to compensate LPs
for the fees charged, and hence I calibrate the model so that LPs break even. The
delta of a GP’s compensation is measured as the expected increase in the present value
of a GP’s payoff from adding $1 extra of value to a portfolio company. To match an
additional moment of the data I add leverage to their framework, and calibrate the model
to match the empirical default rate in the sample.
To evaluate the dynamics of delta the model is estimated at intermediate points in
funds’ life, using reported cash flows and NAV valuations from Preqin.2 I find that the
delta is positively related to fund performance, and increasingly so as the fund matures
and performance becomes more informative.
The delta captures only the direct pay for performance sensitivity. As documented
by Chung et al. (2012), the indirect pay for performance — stemming from improved
fundraising prospects — is of at least an equal magnitude to the direct component.
They show that the relative importance of the direct component increases with GP
experience: GPs with a proven track record are less reliant on current fund performance
to successfully raise new funds. Their insight provides a cross-sectional prediction: the
observed relationship between delta and operational improvements at target companies
should be stronger when the direct component of pay is more important to a GP. Consistent
with this prediction, the documented positive relationship is driven by highly reputable
funds, with estimates suggesting a larger economic effect. A similar argument can be made
for deals completed around successful fundraising events. If investors care mostly about
the performance of the most recent fund when deciding to participate in future funds, the
indirect pay for performance will drop following successful fundraising. Consistent with
this prediction, the positive relationship between delta and operational improvements
is stronger for deals completed near fundraising events.
So far, I have documented a positive correlation between a GP’s pay-for-performance
sensitivity and post-buyout growth in portfolio companies. This may reflect an effect of
incentives provided by the pay-for-performance compensation. However, factors related to

2
A detailed discussion of the model and deviations from Metrick and Yasuda (2010) is deferred to
section 2.2 and appendix C.

94
the delta may be directly related to the operational improvements at portfolio companies.
For example, an obvious alternative is that the results reflect managerial ability: GPs
skilled at improving operations of portfolio companies likely translate that ability into
high fund performance, leading to a high delta. Additionally, there is a time-varying
mechanical relationship as PE performance closely follow public markets (e.g. Ang et
al., 2018): virtually all funds are in the money when equity markets are doing well.
To the extent that expected operational improvements are state-dependent, this leads
to a correlation unrelated to incentives.
To overcome these issues, the empirical specification includes fund fixed effects and
time fixed effects. The inclusion of fund fixed effects absorbs time-invariant differences in
GP ability, and implies that the results of the paper correspond to within-fund estimates.
The staggered nature of PE investments provides for natural within-fund comparisons.
With the amount of investment capital available to a PE fund fixed at inception, there
is no concern that capital flows, and diseconomies of scale, offset increases in delta due
to good performance as in Berk and Green (2004).
To further strengthen the interpretation I exploit that the valuation of PE funds
move with the public market (Ang et al., 2018), and that the delta increases with fund
performance. A public market valuation shock is exogenous and common to all PE funds.
While the shock is common, the resulting impact on delta varies between funds. The
degree to which a market shock impacts a fund’s delta largely depends on two factors:
the fund’s current performance and the degree to which the fund is invested. To illustrate,
the value of a fund that has yet to make its first investment is not impacted at all by
market movements, while one that is fully invested will see the value of all its investments
impacted. Similarly, a fund far in or out of the money will not be much impacted, while
one at the money will experience a big impact. This implies that market shocks can
be used to generate exogenous variation in delta.3
Extracting the part of delta due to a market shock is done in two steps. First, I
estimate what the NAV would have been absent any market movement, using a simple
market model. Second, using this counterfactual NAV and the implied counterfactual
fund performance, I estimate a delta absent the market shock. The difference between
the normal delta and the counterfactual delta is attributable to the market movement.
I confirm the main result of the paper using only the part of delta attributable to market
movements. The results suggest that a one standard deviation increase in the market-
driven component of delta translates to an increased revenue growth rate of between 1.8%

3
This strategy is conceptually similar to that used in Shue and Townsend (2017), where they utilize
differences in the sensitivity to industry shocks of the value of options awarded to CEOs under fixed-value
versus fixed-number option plans.

95
and 3.9%, depending on the model specification. Again, this growth is neither attributable
to increased capital expenditure nor does it come at the expense of worsened profitability.
The paper provides insights to the wider question of how pay-for-performance compen-
sation of managers impacts performance. This question has received considerable attention
to in the context of CEOs, but is notoriously difficult to give a convincing answer to.
Compensation packages are the outcome of a matching and bargaining process, so manager
quality is likely related to both the magnitude and structure of compensation. Managers
may influence the process in which compensation is set (Bebchuk et al., 2002) or time the
delivery of news to the awarding of stock options (Aboody and Kasznik, 2000; Yermack,
1997); compensation packages can be renegotiated ex post (Brenner et al., 2000); and
firms design compensation packages to exploit managerial characteristics such as optimism
and overconfidence (Humphery-Jenner et al., 2016; Otto, 2014).4 The private equity
setting overcomes many of these issues as compensation terms are fixed at fund inception.
This is similar to Agarwal et al. (2009), that examine how the managerial delta in hedge
funds relates to fund performance. The PE setting provides additional benefits as it is a
natural setting to perform within-fund estimations allowing the econometrician to control
for time-invariant managerial ability. As PE funds are always long the equity market,
unlike hedge funds, market shocks can be utilized to generate exogenous variation in delta.
The results of the paper are important for the private equity literature. Carried interest
is intended to align the incentives of managers with that of investors, but we lack empirical
evidence on the effect of variation in the sensitivity of pay to performance. Hüther et al.
(2020) show that VC partnerships with more “GP-friendly" contracts, with carry on a
deal-by-deal basis, are associated with higher returns. Robinson and Sensoy (2013) find
that funds with higher fees do not deliver lower net-of-fee performance. However, these
findings are subject to many of the same endogeneity concerns as the CEO compensation
literature, in particular the matching of contract terms and manager quality. In that
respect, this study provides additional insights by exploiting within-fund variation in
the strength of incentives as well as plausibly exogenous variation.
By examining how the pay-for-performance sensitivity of GPs relate to operational
improvements at target companies, the paper contributes to an extensive and growing
literature exploring the impact of private equity ownership on the performance of
portfolio companies (e.g. Acharya et al., 2013; Bernstein et al., 2019; Bernstein and

4
Frydman and Jenter (2010) provides a survey of this literature. Notable papers on the efficacy of
compensation packages are Jensen and Murphy (1990), showing that CEO compensation is weakly related
to the dollar increase in firm value; Hall and Liebman (1998), showing that CEOs compensation is strongly
related to the percentage increase in firm value; Yermack (1995), who does not find that stock options
appear to be awarded in order to reduce agency costs; and Datta et al. (2001), showing that stronger
incentives lead to better acquisitions.

96
Sheen, 2016; Boucly et al., 2011; Cohn et al., 2014; Davis et al., 2014, 2019; Guo et
al., 2011; Kaplan, 1989).
The next section presents the institutional setting, develops the hypotheses of the
paper, and outlines the framework for measuring pay-for-performance sensitivity in GPs
pay. Section 3 describes the dataset in the paper. Section 4 outlines the empirical
approach and presents the results of the paper. Section 5 concludes.

2 Theoretical Framework

This section outlines and discusses the model used to estimate the explicit incentives
from carried interest. It begins with a brief discussion of the institutional setting, which
is used to develop the hypotheses of the paper.

2.1 Institutional Setting

A private equity fund is a partnership between a group of investors (referred to as limited

partners, or LPs), and a private equity firm (referred to as the general partner, or GP).
When a fund is raised, LPs commit capital to the fund but do not provide it up front.
The GP is provided significant freedom over what to invest in and the timing of those
investments, although the fund has a maximum lifetime and a specified investment period.
Once an investment is identified (or fees are being collected), the GP “calls” the capital.
The amount and timing of capital calls is ex ante uncertain, but LPs are contractually
obligated to meet the calls. When investments are exited, the resulting cash is distributed
to LPs and cannot usually be recycled in new investments.
The terms of the partnership are defined in a limited partnership agreement which
stipulates the length of the investment period, the maximum allowed lifetime of the
fund, and the compensation structure to the GP. As documented in Metrick and Yasuda
(2010), the most common compensation arrangement for buyout funds includes a 2%
management fee and carried interest that gives the GP 20% of fund profits.5 The vast
majority of buyout funds utilizes a hurdle rate of 8% and of those that do, virtually all
have a catch-up rate provision. Once a fund’s performance reaches the hurdle rate, a
catch-up provision allows the GP to receive 100% of the returns until the LPs and the
GP are on the pre-specified profit-sharing (e.g. 80/20).
Axelson et al. (2009) provide a model demonstrating how the structuring of deal
financing reduces agency conflicts between GPs and LPs. Compared to providing financing

5
In addition to management fees and carried interest, Metrick and Yasuda (2010) report that buyout
funds commonly charges transaction fees and monitoring fees to their portfolio companies. Transaction
fees can be considered part of the fixed component of pay, while monitoring fees is often based on EBITDA
and hence performance-based.

97
and compensation on a deal-by-deal basis, compensating GPs with a fraction of the fund’s
profits reduces the incentive to invest in bad deals. Ex ante committed financing thereby
allows GPs to be patient and not rush to deploy capital. However, towards the end of an
investment period the GP might not have sufficient discipline to not invest, particularly as
the basis for the fixed management fee often shifts to capital deployed past the investment
period. By complementing the ex ante equity with deal-by-deal debt financing, the
debt providers add discipline which leads to further efficiency gains. While the limited
partnership agreement is optimally designed ex ante, it does not perfectly cover all future
states. In the model, underinvestment (overinvestment) takes place in bad (good) states
of the world, as financing is hard (easy) to come by.

2.1.1 Hypotheses development

In the model of Axelson et al. (2009), GPs are assumed to possess skill in identifying
as well as managing their investments. Presumably, the GP has to exert costly effort in
order to add value to investments. Agency theory suggests that if managers can exert
costly effort to increase performance, better performance is expected when the manager’s
pay is more sensitive to performance.6 In private equity, operational improvements are
an important way in which private equity funds generate value for their investors.7
Thus, the first hypothesis to be tested is

Hypothesis 1. All else equal, operational improvements observed at target companies

are higher when the pay-for-performance sensitivity of the GP’s payoff is higher.

The sensitivity of pay to performance provided by carried interest can be thought of as

the expected increase in carried interest for each additional dollar of value added by the
GP. This corresponds to the measure used by Jensen and Murphy (1990). The sensitivity
of pay to performance mainly depends on the probability that the fund performance
exceeds the hurdle rate, as no carry is paid out otherwise, and carry is a fixed fraction
of fund profits when it is paid out. The presence of a catch-up rate provision implies
a region of outcomes where gains from adding value is higher. Drivers of variation in
this measure are discussed in section 2.2, where the model framework used to estimate
the pay-for-performance sensitivity is presented.
6
Jensen and Meckling (1976), Fama (1980), Fama and Jensen (1983a,b) discuss agency issues relating
to the separation of ownership and control in firms. Holmström (1979) and Grossman and Hart (1983)
provide models of optimal contracts when effort is unobservable and impacts contractable output with
noise. There is some empirical evidence that firm value is positively related to managerial ownership and
compensation that is more sensitive to firm performance (e.g. Abowd, 1990; McConnell and Servaes, 1990;
Mehran, 1995). However, the evidence is mixed. Agrawal and Knoeber (1996) and Demsetz and Villalonga
(2001) do not find evidence of a relationship between managerial ownership and firm performance.
7
Early studies such as Kaplan (1989), Lichtenberg and Siegel (1990), Smith (1990), showed this for
management buyouts, and more recent studies such as Guo et al. (2011) and Acharya et al. (2013) confirm
the importance of operational improvements to generate returns in buyouts.

98
 The explicit incentive stemming from carried interest is not the only motivation for
fund managers to perform well. As demonstrated by Chung et al. (2012) and Barber and
Yasuda (2017), high performance increases the likelihood to raise a new fund as well as the
size of the fund conditional on one being raised. Thus, the implicit incentive is an important
additional driver for GPs. As in Gibbons and Murphy (1992), learning about the ability of
GPs is important; performance matters more for fundraising success for young GPs lacking
reputation of having successfully managed multiple funds (Chung et al., 2012). Similarly,
Barber and Yasuda (2017) document that performance is of much higher importance for
the fundraising success of low reputation GPs. This implies that the explicit incentive is
relatively more important for high reputation GPs. Indeed, Chung et al. (2012) show that
the relative importance of the explicit incentive is increasing in the experience of a GP.
Therefore, the second hypothesis is

Hypothesis 2. All else equal, the relationship between operational improvements observed
at target companies and the pay-for-performance sensitivity of the GP’s payoff should be
stronger for high-reputation GPs.

When a GP has successfully raised a new fund there are at least two reasons why attention
and effort may be shifted to the new fund. First, if investors care about IRRs when
deciding to invest in new funds by the GP, investments early in a fund’s life are far more
important than later deals, as IRRs are heavily influenced by early cash flows.8 Second,
to the extent that investors pay more attention to the most recent fund’s performance
when deciding to participate in new funds by a GP, the performance of the newly raised
fund becomes matters more for future fundraising success. Both these points suggest that,
following successful fundraising, the importance of the explicit incentive relative to the
implicit incentive increases in the old fund. Additionally, if the GP’s old fund is performing
poorly, the direct pay-for-performance sensitivity may be higher in the newly raised fund.
This makes the third hypothesis

Hypothesis 3. All else equal, the relationship between operational improvements observed
at target companies and the pay-for-performance sensitivity of the GP’s payoff should be
stronger post-fundraising.

Having established the three hypotheses to be tested against the data, we now turn to
the framework used to estimate the pay-for-performance sensitivity of a GP’s payoff.

8
Chung et al. (2012) document that higher IRRs increase the likelihood of fundraising and the size
conditional on raising a fund. Larocque et al. (2019) run a horse-race to test whether IRR or TVPI is
more important for investors when deciding whether to participate in follow-on funds, and find that IRR
is favoured.

99
2.2 Estimating Explicit Incentives in Private Equity Funds

This section outlines the methodology used for estimating the pay-for-performance
sensitivity at various points in a fund’s life. Two inputs to the model are from the
assembled dataset of UK companies acquired in a buyout. The description of that
data is deferred to section 3.
The methodology of the paper builds on the framework developed by Metrick and
Yasuda (2010). They model the expected revenue from fees in private equity funds, of
which carried interest is one component and the focus of this study. Carried interest can
be viewed as a call option on a fraction of the fund’s profit, with the hurdle rate acting as
the strike price: GPs receive their share of profits only if the fund’s performance exceeds
the hurdle rate. However, there are several features of private equity funds that make
the valuation of carried interest less straightforward than plain call options. A private
equity fund is a portfolio of staggered investments; profits from individual investments
are aggregated to a fund level; entry and exit dates are ex ante uncertain; and the GP
may add value to the investment. This paper largely follows Metrick and Yasuda (2010)
in dealing with these issues, and deviations from their framework are clearly noted.
The purpose of this paper is not to evaluate the model of Metrick and Yasuda
(2010) but to use it to estimate the pay-for-performance sensitivity of carried interest.
Consequently, section 2.2.1 provides only a brief description of their model, while
highlighting any alterations made. Section 2.2.2 discusses how the paper estimates
the pay-for-performance sensitivity measure. For the interested reader, Metrick and
Yasuda (2010) provide an extensive analysis of the impact of altering model inputs,
and discuss the assumptions underpinning the model. The qualitative patterns of the
pay-for-performance measure in this study are robust to varying model inputs along
the lines done in Metrick and Yasuda (2010).

2.2.1 Modelling Expected Carried Interest

This paper follows Metrick and Yasuda (2010) in using Monte Carlo simulations to
estimate the expected value of carried interest and other cash flows in private equity
funds. A simulation consists of: generating a PE fund portfolio, which consists of
investments in portfolio companies; simulating a value path for each portfolio company;
simulating exit timings for each investment; generating fees from a set of pre-specified
contractual features; attribute cash flows to LPs and GPs; and calculate the present
value of these cash flows for a risk-neutral agent.
The value Xti of a portfolio company i at time t is assumed to follow a Gaussian
diffusion process of the form
dXti
q 
= rdt + σ 1 − ρ2 dW i + ρdW F ,
t t (1)
Xti

100
where Wti and WtF are two mutually independent standard Brownian motions. Wti is
specific to the portfolio company while WtF is common among all portfolio companies. It
is further assumed that the company-specific processes are independent of each other, so
that ρ captures the correlation between portfolio companies stemming from a market-
wide factor. As in Metrick and Yasuda (2010), ρ is assumed to be 0.2 while the risk-
free rate r is assumed to be 0.05.
The simulated Xti path captures the total value of the portfolio company. In a
deviation from Metrick and Yasuda (2010), this paper includes leverage at the portfolio
company level, allowing for the equity position to be wiped out. Interest on debt is
assumed to be rolled up for simplicity. If the value of the company at any point prior
to the exit date drops to the value of debt, control of the company is transferred to
creditors and the equity position is wiped out. This ensures that debt is risk-free, and
the interest rate r ensures that debt holders break even.
The inclusion of leverage allows σ to be calibrated to match a moment of the data on
portfolio companies: the fraction of buyouts where equity holders are wiped out. Without
this calibration, σ has to be assumed. A leverage ratio of 2:1 is used.9
As LPs are charged fees, GPs need to add value in equilibrium to ensure that LPs
break even. To account for value added, a fixed amount is added to the starting value
of each investment. Conceptually, this amounts to adding the discounted value added
by the GP to the initial value of the investment.
In practice, the amount of value added required is uncertain at the beginning of the
simulation process, as the amount of fees charged is itself a function of value added.
Additionally, the added value increases enterprise value which reduces leverage, and
therefore lowers the probability of default. Thus, value added feeds into the calibration of
σ, which in turn impacts expected carried interest, and hence influence the value added
required to ensure that LPs break even. Value added and σ are determined together in
an iterative process where one is calibrated after the other. The process is repeated until
it converges so that expected present value to LPs after fees equal the present value of
their capital calls, and the default rate matches the empirically observed one.
Fixed fees are charged on committed capital during the investment period, and on
the remaining investment base after that. The fixed fees plus investment capital equals
the committed capital in each simulation.10
In Metrick and Yasuda (2010), holding periods are simulated using an an exponential
model with instantaneous hazard rate of 0.20 is used, corresponding to an average
9
The leverage ratio corresponds to the one documented for buyouts in western Europe in Axelson et
al. (2013). Alternative leverage ratios lead to different σ, while σE remains virtually unchanged.
10
In Metrick and Yasuda (2010), fixed fees for buyout funds are on average $11.64 per $100 of committed
capital. The corresponding number in this paper amounts to $11.01.

101
duration of 5 years. In this paper the holding periods are drawn with replacement from
the empirical distribution of holding periods, described in section 3. In an earlier version,
this paper utilized the hazard model approach, with qualitatively similar results. Limited
partnership agreements typically preclude holding investments past the 12th year of
the fund without explicit LP agreement (Metrick and Yasuda, 2010). Therefore, any
remaining investments are liquidated at the end of the 12th year.
The number of investments per fund is set to 16, and the investment pace is set
deterministically to be 26%, 23%, 25%, 18%, and 8% over an investment period of the
fund’s first five year.11 Both the entry and exit assumptions ignore any timing effects
or impact of realized returns on exit decisions. While unlikely to be true, Metrick and
Yasuda (2010) document that it does not significantly alter the estimation of fees.
The fund terms assumed are the ones reported to be employed by virtually all buyout
funds in Metrick and Yasuda (2010) and Robinson and Sensoy (2013), with: carried
interest at 20%; a hurdle rate of 8%; a catch-up rate of 100%; the carry basis being
equal to committed capital on the whole fund, rather than measured at a deal-by-deal
level; and the management fee is assumed to be 2%, with the basis being committed
capital in the investment period and net invested capital thereafter.12 It is further
assumed that all committed capital, increased by the hurdle rate, has to be returned
to LPs before the GP gets any carried interest.13
Table C.1 gathers model inputs and the sources used to justify model choices. A
summary of the model outputs after calibrating parameters is provided in Table C.2,
where the are compared to model outputs for buyout funds with comparable fund terms
in Metrick and Yasuda (2010). All fees are lower in the model estimated in this paper.
Transaction and monitoring fees used here are from Phalippou et al. (2018), which
documents lower percentage fees, as well as a higher share rebated to LPs, compared to
those used in Metrick and Yasuda (2010). The larger amount of portfolio companies in
the fund (16 compared to 11) provides greater diversification, reducing expected carried
11
The number of investments per fund is taken from Braun et al. (2017), while the investment pace
follows Metrick and Yasuda (2010). In section 2.2.2, the estimations will be conducted from a point
during the investment period, with some investments already undertaken. In that section, the investment
pace is scaled to ensure that any uncalled capital is fully invested at the end of the investment period.
As the basis for management fees is often switched to net invested capital from committed capital, GPs
have an incentive to ensure that all committed capital is invested. Evidence in Arcot et al. (2015) and
Degeorge et al. (2016) is consistent with the idea that GPs accelerate their investment pace at the end of
the investment period when they have large amount unspent capital.
12
Preqin has data on fee structures for a subset of funds in the database. Numbers in Preqin largely
support the assumptions used here. Of buyout funds in Preqin with fee data: 96.0% (of 860 funds) have a
carried interest of 20%; 94.5% (of 474 funds) have a hurdle rate of 8%; 65.1% (of 573 funds) charge carry
on the whole fund; and 75.3% have a management fee of 2%.
13
There are provisions where the GP can earn carry early, for example when the valuation of a fund’s
ongoing investments indicates that the fund is in the carry. In such instances there is normally a clawback
provision where the GP has to pay back early carry in the event the final performance is not good enough.
Such early carry is valuable to the GP due to earlier payments.

102
interest. With fees lower, the required value added to make LPs break even is reduced,
leading to lower net invested capital and consequently reduced management fees, as well
as a further decrease in expected carried interest.
The calibrated value added by GPs is $9.23, while the calibrated σ = 0.174, cor-
responding to an equity σE = 0.521.14 With these calibrations, LPs break even in
expectation, and the number of buyouts in which equity holders are wiped out is 19.4%,
the fraction of fully exited buyouts where the equity is wiped out, documented in Table
2. These calibrated values are taken as granted when empirically estimating a GP’s
pay-for-performance sensitivity in the next section.
Appendix C provides further details on the Monte Carlo simulation and the procedure
followed to calibrate value added and σ.

2.2.2 Estimating the Pay-for-Performance Sensitivity

We wish to measure the sensitivity of a manager’s payoff to increased performance, here

interpreted as adding more value to a given investment. In the language of option pricing,
we wish to measure the delta of the GP’s payoff function. The delta is defined as
∂E [GP Payoff]
∆= . (2)
∂ Value Added
To operationalize this measure, the delta used in this study will be estimated as the
expected change in the GP payoff from adding $1 extra of value to the next transaction
undertaken. This definition allows the delta to be estimated at any point in the fund’s life.
It is instructive to examine what we expect will drive variation in ∆. To this end,
consider a simple model where a PE fund makes a single investment of $100 at t = 0 and
is liquidated at t = 2. In each period, the value of the investment increases by u > 1.01
with probability p, and decreases by d = 1
u with probability (1 − p). Suppose further that
there are no fees except for carried interest of 20%, and that there is no hurdle rate.
Figure A.1 draws all possible paths for this investment. At t = 2 there are three
potential nodes. At each node, C denotes the value of carried interest, with the subscripts
uu, ud, and dd denoting the path taken to a particular node. In this simple specification,
the GP only earns carry at the top node, with Cuu = 0.2 ∗ 100(u2 − 1). At the middle node,
profits in the fund is exactly 0, and as such the GP is just short of earning some carry.
Suppose now that the GP adds $1 of value at one of the middle nodes, say the down
node. This increases the value of the investment at that point to $100d + $1. If the
next draw of the value process is u, the added value ensures that the fund makes a
profit of u, and the GP’s share from carried interest is 0.2u. If instead the draw from

14
Metrick and Yasuda (2010) assume σ = 0.6.

103
the value process leads to a down step, the added value is not enough to ensure that
the fund makes a profit, and no carry is paid out.
Let C ∆ denote the change in carry payment at t = 2 due to increasing value by $1
at an earlier node, with the subscript denoting the path since adding value. Cd∆ thus
denote the change in carry if $1 of value was added at t = 1 and the draw reduced the
investment value. Similarly, Cud
∆ denotes the change in carry if the value was added at

t = 0, and we end in the middle node. With this terminology we can calculated the GP’s
pay-for-performance sensitivity ∆ as the present value of the expected C ∆ .
Figure A.2 summarizes the change in carry C ∆ at the end nodes, as well as summarizing
the pay-for-performance sensitivity ∆ at earlier nodes. The superscript of ∆ denotes
the time period in which value is added, while the subscript (when applicable) denotes
at which intermediate node value is added.
This simple model provide some useful insights into what we expect will drive variation
in pay-for-performance sensitivity. First, it is clear that ∆ is increasing in fund performance,
as illustrated by ∆1u > ∆1d . The reason this is the case is that higher performing
funds are more likely to be in the carry, and adding an additional dollar of value
therefore generates increased payoffs to the GP with higher probability. Second, if we
add additional time periods to the model, we find that nodes of comparable performance
generally have a different ∆. The intuition for this difference is similar to that of time
to maturity for options: to-date performance becomes increasingly informative of final
performance as we approach liquidation.
With the intuition clear, we now turn to empirically estimating ∆. The simulation
model described in the previous section described simulating a PE fund from inception.
By contrast, here we wish to examine how the dynamics of delta as a fund ages, and
how early success, or lack thereof, impacts the delta. Therefore, a major deviation from
Metrick and Yasuda (2010) arises from running the simulation at various points in a
fund’s life, accounting for intermediate cash flows and valuation changes in the fund up
until the point the simulation starts. The equilibrium value added, and the calibrated
σ, from the previous section are taken as granted in this section.
The following adjustments are made to the estimation of the expected carried interest:
the series of cash-flows to-date are used to establish the amount and timing of investments
already undertaken, net of estimated fixed fees; the holding periods of current investments
are drawn from the empirical distribution of holding periods, conditional on that holding
period exceeding the current time the investment has been held; the investment speed of
remaining investment capital is scaled so that the fund is fully invested at the end of the
investment period; the currently reported NAV is assumed to represent a fair valuation of

104
investments held to date;15 as a simulated investment gets the full value added immediately,
while in practice any such process would take place over time, a fraction of this amount
is assumed to have been incorporated, while the remainder is added at the start of the
simulation. The fraction assumed to have been incorporate represents the average duration
of existing investments divided by the average holding period of investments.
The model is estimated empirically for all quarters with NAV valuations and cash flow
data for buyout funds in Preqin, up until the fund’s 20th quarter. For each fund-quarter,
the simulation generates 10,000 paths for all portfolio investments. The simulation is
performed twice for each fund-quarter, once with and once without adding an additional
dollar of value.16 The delta is measured as the difference in the average present value
of GP pay. In total, delta is estimated for 13927 fund-quarters.
Figure 1 summarizes the empirical fit of delta as a function of fund performance and
fund age. The figure is generated as the best fit of a simple linear model of fund IRR,
fund age measured in quarters, and an interaction of the two. The figure empirically
confirms the intuition provided by the simple model earlier: ∆ is increasing in fund
performance, and increasingly so as the fund matures and fund performance becomes
more informative of final performance. The figure is complemented by Table A.1 which
summarizes the model fit. Tellingly, while fund performance is relatively informative
on its own, the adjusted R2 increases from 0.136 to 0.468 when interacting it with fund
age. While there is much variation not captured by these simple terms, they are clearly
important drivers of the variation in ∆. What these results further suggests is that any
shock to performance will not impact the delta of all funds equally, as the sensitivity
to performance depends on when in a fund’s life the shock takes place. We will return
to this in section 4.1.3 to generate exogenous variation in delta.

3 Data

This section describes the dataset utilized in this paper. The dataset consists of two
parts: i) a partly hand-collected dataset containing accounting data, deal outcomes, and
ownership information of UK-registered companies acquired in buyouts; and ii) a Preqin
dataset containing fund-level cash flows and valuations.

15
The number of investments assumed to be held is estimated as the fraction of investment capital that
has been spent less any distributions. As we cannot observe the precise timings of each investment nor
individual company valuations, the holding period of all investments is assumed be the cash-flow weighted
duration of investment capital. The value of each investment is assumed to represent an equal fraction of
NAV.
16
For each fund-quarter, the same seed is used at the start of the simulation process both times to
ensure that variation in expected pay is solely due to the increased value added. This significantly reduces
the standard error of the estimated delta, giving a high precision with only 10,000 runs. Different seeds
are used for each of the 10,000 runs, and for every fund-quarter that delta is estimated for.

105
 The portfolio company level dataset is used to generate inputs to the model described
in section 2.2. In particular, empirically observed holding periods are used for investment
durations, and the model is calibrated to match the level of deals in which equity holders
are wiped out. The deal-level financials are used to measure operational improvements
for the purpose of evaluating the hypotheses outlined in section 2.1.1.
To relate the estimated pay-for-performance sensitivities of GPs’ pay to operational
improvements, we need to know the acquiring fund in buyouts. Transaction announcements
usually only contain information about the acquiring PE firm. Ownership filings are
used to establish the PE fund providing the financing and, in syndicated deals, the
lead investor. Finally, estimating GPs’ pay-for-performance sensitivity at intermediate
points in a fund’s life requires fund-level cash-flow and valuation data, which is what
the Preqin dataset is used for.
Appendix B contains additional details of the data collection process.

3.1 Sample Selection

Assembling a representative dataset on portfolio companies owned by private equity is a

challenge in any private equity study. As portfolio companies are privately held they have
limited disclosure requirements in most countries. Consequently, researchers often rely on
a set of companies with publicly available financials (e.g. Guo et al., 2011) or on datasets
that have been made available to them by investors or GPs (e.g. Acharya et al., 2013).
In the largest buyout market in the world, the US, financials are generally only available
for private companies if they have publicly traded debt or subsequently undergoes an
IPO and make financials available for some years prior to the listing. Cohn et al. (2014)
document that companies with publicly available financials perform better than buyouts
lacking publicly available financials. Davis et al. (2019) document marked differences
between different types of buyouts, with employment shrinking in public-to-private
buyouts while increasing for other types of buyouts. These studies highlight the need
for representative samples containing various types of buyouts.
Cohn et al. (2014) and Cohn et al. (2020) utilize confidential corporate tax return data
to track buyouts while privately trading. While corporate tax return data allows a more
representative sample, it does not provide the same detailed information that consolidated
financial reports do. As an example, Cohn et al. (2014) points out that their analysis relies
on tax reporting definitions of revenues and expenses that are subject to efforts to avoid tax
payments. Furthermore, the lack of depreciation and amortisation information implies that
they are unable to work with the income measure of most interest in buyouts: EBITDA.
In this study I focus on buyouts of companies in the second largest buyout market
in the world: the United Kingdom. In the UK, both public and private companies are

106
required to submit filings to the UK national registrar, Companies House. These filings are
made publicly available, and include annual reports containing financials, “annual returns”
(more recently “confirmation statements”) containing precise holdings of all shareholders
in a company, and filings relating to insolvency procedures (called “administration” in
the UK). The UK therefore offers a unique setting where a representative dataset of
buyouts can be assembled.17
To generate a set of buyouts involving UK-registered companies I follow Strömberg
(2008) and use the screening tool in Capital IQ. Appendix B describes the screening criteria
used in detail. The transaction screening captures buyouts, as defined by Capital IQ, in
which a majority ownership stake is sold. Venture capital investments, growth investments
and deals for minority stake positions are excluded from the study to ensure a focus on
companies in which the acquiring buyout fund has sufficient control to implement changes.
Table 1 outlines the sample construction process. 1964 buyouts of UK companies
in the period 1999 to 2018 are identified from the Capital IQ screen. Of these, 850 are
randomly selected to be processed. A random selection is necessary due to a substantial
amount of manual work required in processing the data.
Of the 850 processed transactions, 59 are dropped for reasons outlined in the table
and discussed in Appendix B. The type of buyout is classified for each of the 791
fully processed deals, while the outcome of a deal, or deal exit, is determined in all
but 4 deals. This is the largest sample used in the paper, and it provides inputs to
the model described in section 2.2.
Company financials are required to measure operational improvements and evaluate
the hypotheses outlined in section 2.1.1. To that end, 230 buyouts are dropped that do
not have financials available in at least the year prior to the buyout and the year following
the buyout, required to measure changes in operating performance around the acquisition
event. To be consistent with earlier private equity studies (e.g. Bernstein et al., 2019), a
further 69 buyouts of companies operating in the financial (SICs 600-699) or utility (SICs
489-493) sectors are excluded, leaving us with 494 buyouts with financials available.18
There are two primary reasons why financials are unavailable: the company lack
independent accounts pre-buyout, or consolidated accounts move outside of the UK
post-buyout. The majority of cases where pre-buyout financials are missing are divisional
buyouts, as not all divisions submit independent annual reports. A typical buyout involves

17
I am not the first to utilize the UK setting to study private companies. Brav (2009) and Michaely
and Roberts (2012) study the UK to gain insights from privately trading companies. In the private equity
literature, Chung (2011) examines the effect of buyouts on privately held UK companies. More recently,
Bernstein et al. (2019) study how UK companies under private equity ownership fared during the financial
crisis.
18
The results of the paper are robust and remain qualitatively unchanged if we include these sectors.
The final sample increases from 246 to 285 when including these sectors.

107
the acquiring private equity firm setting up one or several shell companies in preparation of
the acquisition. In virtually all buyouts, one of these shell companies begins consolidating
the financials of the group. I lose access to the accounts, and consequently exclude the
deal from the sample, if the top consolidating entity is registered outside of the UK.19
Finally, 248 buyouts are excluded because the fund is only involved in a single
transaction, or because the ∆ of the acquiring fund cannot be estimated due to lack of
cash flow data in Preqin. The empirical strategy describe in section 4 utilizes within-fund
variation, requiring at least two deals per fund. This leaves the final sample at 246 buyouts.

3.2 Sample Composition

The type of buyout is determined based on the type of ownership prior to the buyout.
To establish the exit route I utilize a combination of sources. Capital IQ, Zephyr, and
news articles are scanned for information regarding sales. When a sale is identified I
corroborate the information through reading annual reports of the acquired company
and, when possible, the acquirer.
If no mention of a sale is identified I look at whether the company ended up in insolvency
procedures, called being placed in “administration” in the UK. When a company is placed
in administration, the administrator regularly submit filings to Companies House on the
progress.20 A company, or part of it, can be acquired out of administration. There are
cases in the sample where Capital IQ (and, less commonly, Zephyr) incorrectly attribute
the seller in these cases as the PE firm, even when none of the proceeds go to the PE
firm. This is why sales identified in the first step are corroborated by reading annual
report and searching for potential insolvency.21
In the remaining cases I utilize ownership filings to detect when a PE firm ceases to be
the ultimate owner. This process either confirms that the PE firm remains the ultimate
owner, or document when a change in ownership takes place. In the latter case I consult
annual reports to identify the reason behind the change in ownership. This exercise picks
up some sales not picked up by the other sources. Typically, these concern small companies
19
While unconsolidated accounts are sometimes still provided by companies lower in the group structure
which may be UK-registered, they are not reliable indications of operational performance as intra-group
transactions are not netted out, and various entities in the group may take on different costs. Only the
consolidated accounts at the top properly account for everything.
20
Administration is the UK equivalent of Chapter 11 in the US. The filings submitted by an administrator
include: a “Statement of administrator’s proposal”, which provides background information on the events
leading to administration, the strategy and progress to date of the administrator(s), and an assessment of
the likely outcome for the various classes of creditors; and semi-annual “Administrator’s progress report”
outlining the progress of the administrator.
21
Since the Enterprise Act 2002, administrators can be appointed out of court. This enabled pre-
packaged administration, where a sale is negotiated prior to the company being placed in administration,
with the sale taking place immediately after. Almost all misclassified sales by Capital IQ relate to
pre-administration packages. Even in pre-packaged administration cases the administrator submits filings
to Companies House.

108
that are sold to the management team or other individuals. Importantly, this exercise also
picks up cases where creditors seize control of a company by, for example, a debt-to-equity
swap, or other events in which the equity holders receive no compensation for their stakes.
Table 2 summarizes the sample in terms of buyout type, exit routes, and holding
periods. Panel A presents the composition of buyout types for the full sample as well as four
5-year sub-periods. Buyout types are defined in Appendix D.1. The most common type of
buyout in the sample is secondary buyouts with 34.1% of the sample. Secondary buyouts
are followed by private-to-private buyouts, with 26.3% of the sample; divisional buyouts,
with 23.1%; public-to-private buyouts, with 13.4%; and distressed buyouts, with 3.0%.
The time-trends are depicted on a year-by-year basis in Figure 2. The clearest trend
is the increased prevalence of secondary buyouts over time at the expense of the other
categories. It is evident in the figure that the sample is weighted towards the earlier
part of the sample period, with coverage particularly low for the most recent years of
2016-2018. This raises some concerns about representativeness, and should be kept in
mind in the following analysis. However, the time-distribution corresponds to the overall
distribution of the transactions identified by Capital IQ, as illustrated in Figure A.3.
Panel B of Table 2 presents the composition of exit routes. The most common exit
route, with 31.6% of all transactions, is through a sale to a financial institution. This
vast majority of these sales are to other private equity firms. The second most common
exit route, with 29.0%, is through a trade sale to another company or the management
team. The third most common route, representing 16.2% of all transactions, is losing
control either through administration processes or with creditors seizing control without
compensating equity holders. Almost as many transactions, 15.9%, remained under
private equity ownership as of 2019-09-30. 6.2% of all exits take place through an IPO.
The 16.2% figure (19.4% of exited) of buyouts that end in a exit where the private equity
owner receives nothing is high compared to the 8% bankruptcy rate of exited UK deals
reported in Strömberg (2008). The discrepancy partly reflect differences in sample period,
as the sample in this paper spans the outcome for portfolio companies owned through the
financial crisis. Another contributing factor is that, for UK companies, 12% of exits are
unknown in Strömberg (2008), compared to 0.5% in this paper. It is plausible that at least
part of the 12% unknown exits are less reported outcomes, for example when creditors
seize control. In the sample in the paper, 9.4% (11.1% of exited) of the transactions
go through formal administration processes while 6.8% (8.1%) are cases where creditors
seized control of the business, which are usually only mentioned in the annual reports.22
22
The numbers from Strömberg (2008) are from Table 4, Panel B. The numbers in this study are
comparable to the overall bankruptcy rate of 18.4% documented in Table 1 of Degeorge et al. (2016).
Note that deals classified as losing control in this paper do not necessarily correspond to a 100% write-off,
as there may be intermediate cash flows to the private equity fund.

109
 Figure 3 document the distribution of exit route over time, with transactions grouped
by the year of the buyout. The figure documents a clear cyclicality in deals that eventually
ends up with the private equity firm losing control. This figure is highest for transactions
taking place just before market crashes; 30% of deals completed in 2007 ended with the
private equity firm losing control. Figure A.4 depicts the composition of exit routes by
the year the exit takes place in, excluding those where the private equity firm remains in
control. Interestingly, while 2009 experienced the highest number of companies ending up
in financial distress, both in absolute numbers and as a fraction of all exits, the numbers
remained above the sample-average up until 2012. In annual reports and administrator
reports, it is often mentioned that companies experienced difficulties following the financial
crisis, though it took some time before the company entered formal insolvency procedures,
or creditors seized control. This may reflect the long maturities on loans documented
in Achleitner et al. (2011), or efforts by private equity firms to support their portfolio
companies through distressed times as documented in Bernstein et al. (2019).
To put these figures in perspective it is useful to contrast the number of exits in a
year with the total number of companies under private equity ownership during the year.
Figure A.5 depicts the outcome of all companies that were under private equity ownership
for at least part of a year. The figure makes clear that although the rate of losing control
is elevated in the period 2008 to 2012, no single year exceeded a failure rate of 7%, as
most companies remained under private equity control and overall exit activity was low.
Panel C of Table 2 presents the holding periods, defined as number of years from
the transaction date to the exit date, of all exited transactions. The average (median)
holding period is 4.99 (4.43) years. Figure A.6 complements the table and presents
the holding periods by year of the buyout. For buyouts towards the end of the sample
period there is a substantial amount of companies still under private equity ownership,
implying that those with holding periods were the ones to quickly exit. The average
(median) holding period increases to 5.09 (4.59) if we include only buyouts completed up
until 2013, while if we look at only buyouts completed up until 2008 the corresponding
numbers are 5.27 (4.88). Overall, the typical buyout in the sample lasts about 5 years,
compared to a median of 4.3 years for fully realized deals in Braun et al. (2017) and
an average 4.4 years in Degeorge et al. (2016).

3.3 Accounting Data

The accounting data in the study comes from Fame, a Bureau van Dijk (BvD) database.
BvD digitalizes the financials available in the scanned annual reports that are publicly
available at Companies House. The FAME database provides financials for up to 20

110
years worth of annual reports.23 Since my sample period includes transactions as far back
as 1999, I have manually complemented the Fame database with pre-buyout financials
available in annual reports, downloaded from Companies House, when necessary.
In buyouts, the acquiring private equity firm typically creates one or several ac-
quisition vehicles to facilitate the acquisition. These entities are shell companies with
no economic activity on their own prior to the acquisition. Figure A.7 presents a
simple example of the change in the group structure of MPM Products, following a
buyout by ECI Partners in 2016.
Although all private companies have to file annual reports with Companies House,
wholly owned subsidiaries may submit unconsolidated accounts; it is sufficient that one
(parent) company in a company group submits consolidated accounts. Unconsolidated
accounts can be misleading as they do not net out intra-group transactions, tend to under-
state costs, occasionally misrepresent turnover, and frequently lack financial information
altogether. One of the vehicles created for the acquisition virtually always consolidates
group accounts after a buyout. As part of the data collection I document the entity
reporting consolidated accounts for each company-year, and merge financials of the different
consolidating entities to create a coherent time-series of financials for each target.24
It may be instructive to take a concrete example to illustrate why tracking the
unconsolidated accounts of the underlying portfolio company can be misleading, even
when financials are available. In ECI’s buyout of MPM Products, the portfolio company
submits unconsolidated accounts post-buyout. For the fiscal year 2017 (2018), the
unconsolidated revenue is 7% (11%) lower than the consolidated revenue. While revenue is
lower, the unconsolidated accounts understate costs, which means that the unconsolidated
EBITDA/Revenue margin is overstated by 1.1 (1.7) percentage units, or 15% (17%)
above the consolidated EBITDA margin.
Table 3 presents summary statistics for the portfolio companies in the final sample,
measured at the most recent fiscal year-end prior to the buyout. Panel A present summary
statistics in levels. The average (median) company in the sample have revenues of £157.2
million (£69.3 million), an EBITDA/revenue margin of 15.9% (14.1%), and 1797 (614)
employees. Panels B and C report growth rates leading up to the buyout. At a 1-year
(2-year) horizon, the average company increases revenue by 17.7% (49.9%), the number of
employees by 11.2% (30.0%), and improve the EBITDA margin by 0.9 (2.0) percentage

23
Fame covers companies in the UK and Ireland, while Orbis and Amadeus, two other BvD products,
have a much wider geographical coverage. The main advantage of Fame is that it provides 20 years data
per company, compared to 10 years provided by Orbis and Amadeus.
24
Fame provides time-series of financials for individual companies, but one needs to manually track the
ultimate consolidating entity. Fame does indicate whether financials are unconsolidated or consolidated,
but within a group structure there may be multiple entities that consolidate financials at various levels.

111
points. Panel D presents industry composition. Services is by far the dominant sector
covering 45% of all companies, followed by manufacturing (23%) and retail (16%).
To assess representativeness of the subsample used in the analysis, Table A.2 compares
the summary statistics to the full sample of 494 companies with financials. Panels
A, B and C compares the descriptive financials available. The subsample in the paper
contains larger companies on average, having £28.8 million more in revenues and 369 more
employees. The companies in the sample have also grown faster prior to the buyout, and
are 0.9 percentage units more profitable. Panels D, E, and F document the differences in
buyout types, exit routes and industry composition. Consistent with the larger companies
in the sample, public-to-private buyouts are overrepresented in the subsample used in
the paper. Exit routes and industry composition are both comparable to that of the full
sample. It is not too surprising that companies in the sample are larger, as the acquiring
fund is required to have Preqin cash flow data available; the average US or European
buyout fund with Preqin cash flow data is a $1.7 billion fund, while the average buyout
fund in Preqin’s larger universe is a $0.9 billion fund. While expected, it is important to
keep in mind that the analysis may not be representative of smaller buyouts.25

3.4 Fund cash flow data

To estimate the pay-for-performance sensitivity measure of the buying fund we need

information on cash flows and NAV valuations. To this end, the buying funds are matched
with a Preqin database containing fund characteristics, cash flows, and NAV valuations for
a large amount of private equity funds. Fuzzy lookup is used to match funds by name and
other identifying information is used to corroborate and resolve any uncertain matches.26
Preqin relies on freedom of information requests, as well as voluntary disclosures
by GPs, and as such it is likely not a representative of the whole PE universe. As
noted in Brown et al. (2015), Burgiss is considered the most accurate and comprehensive
database. However, this paper requires access to fund names in order to match the

25
Compared to recent studies, the companies in the sample are relatively large, with the exception of
studies focusing exclusively on public-to-private buyouts. In studies of UK buyouts, the average company
has sales of £158.6 million in Renneboog et al. (2007) (public-to-private only), £10.4 million in Chung
et al. (2012) (private-to-private only), and £98 million in Bernstein et al. (2019). Comparing to studies
in other countries: the average French company has €32.6 million in sales in Boucly et al. (2011); the
average company acquired in a private-to-private buyout in the US has sales of $117.4 million in Cohn et
al. (2020), while the average company acquired in public-to-private buyouts has sales of $604.2 million in
Cohn et al. (2014).
26
To establish a match, a fund is required to have the same name and sequence number (e.g. EQT
IV). In practice, it is common with variations in the names of funds between various databases and the
ownership filings (e.g. EQT IV Fund, EQT IV LP, EQT IV LLC, etc.). Furthermore abbreviations are
often used in one place (e.g. GTCR Fund VII) while another spells out the full name (e.g. Golder Thoma
Cressey Rauner VII). To ensure that matches are accurate, a name match is corroborated through sanity
checks utilizing a fund’s vintage year together with the transaction date, and whenever available the fund
listed in Capital IQ as the acquiring fund is utilized.

112
buying fund to performance data. As Burgiss is provided on an anonymized basis to
researchers, it is not suitable for this study. The dataset with the best coverage that
allows matching on fund names is the Preqin database which is used in this study. Brown
et al. (2015) and Harris et al. (2014) find that the overall performance of buyout funds
is similar between the Preqin and Burgiss databases.

4 Results
This section develops the empirical approach taken to test the hypotheses in section
2.1.1 and present the results of the paper.

4.1 Operational Performance and Pay-for-Performance Sensitivity

We are interested in how operational improvements at portfolio companies are related to

the sensitivity of a GP’s pay-for-performance sensitivity, defined in section 2.2.1.
As there is not a single measure of operational performance, the analysis will utilize
several different dependent variables. To facilitate the current discussion I focus on
revenue growth. It is not clear at what horizon we might expect any improvements to
occur, or whether they will be persistent. Revenue growth will therefore be measured
from the pre-buyout period t − 1 to various post-buyout periods t + x. The following
equation is estimated,
5
Yi,j,t−1:t+x =α + λ · ∆j,t + φj + φt + βs 1{fund aget = s} (3)
X

s=1
+ Γ1 Xi,t−1 + Γ2 Xi,t + i,j,t ,

where Yi,j,t−1:t+x is the annualized revenue growth between t − 1 to t + x of portfolio

company i, acquired by the private equity fund j in year t. ∆j,t is the pay-for-performance
sensitivity of fund j measured at the most recent NAV prior to the transaction.
The coefficient of interest is λ, which captures the relationship between the pay-
for-performance sensitivity and revenue growth.
The set of controls included is rather strict, but is chosen to avoid omitted variable
bias. In section 2.2.2 we established that ∆j,t is increasing in fund performance. If there is
heterogeneity in ability among GPs, we expect high ability GPs to have higher ∆ as well
as to be associated with larger operational improvements in portfolio companies. If the
ability of a fund-manager does not change over a fund life, we can take out this effect by
including fund fixed effects φj , and the resulting estimation corresponds to a within-fund
estimate. The private equity setting lends itself well for within-fund estimations: as
carried interest is based on fund profits, the staggered nature of investing naturally leads
to deals being undertaken under varying ∆j,t .

113
 We include time fixed effects φt as the performance of private equity funds, and
consequently ∆j,t , is related to overall market movements (e.g. Ang et al., 2018). To
the extent that operational improvements are easier to implement in certain states of
the economy, we need to control for market conditions.27
Indicator variables denoting the year in which an investment is made are included to
take out some possible mechanical changes in ∆j,t . Private equity funds often experience
a ”J-curve” effect, where performance is low in the early years due to fees being collected
on committed capital, while gains arise only from invested capital. Including indicator
variables for fund age also serve to mitigate concerns that private equity funds might
invest in different types of companies at different stages of the fund life.28
Finally, we may expect some portfolio companies to be more likely to improve
operations. For example, private equity are more likely to be able to relax credit constraints
of small private companies than of large, publicly traded ones. The model therefore
includes a set of company characteristics Xi,t−1 , measured prior to the buyout, to control
for any impact of observable characteristics. The included controls are (log) revenue,
(log) total assets, and EBITDA/revenue profitability. The book leverage, measured in
the fiscal year of the buyout, is included in Xi,t .
The results for annualized revenue growth are presented in Panel A of Table 4.
The results show that a GP’s pay-for-performance sensitivity is positively associated
with revenue growth for all horizons, though not statistically significant at the t − 1 to
t + 1 horizon, and only at the 10% level for the t − 1 to t + 2 horizon. The coefficient
estimates suggest that the economic magnitude is sizeable. A one standard deviation
change in ∆j,t , 0.0242 in the sample, is associated with an increase between 2.7% and
3.5% in the rate of revenue growth.
Panel B examines company growth using an alternative measure: employee growth.
This measure largely confirms the positive association of ∆j,t with company growth,
though it is only statistically significant for two horizons, t + 2 and t + 3. The point

27
Boucly et al. (2011) and Chung (2011) show that one way in which private equity help portfolio
companies is by relaxing credit constraint to facilitate growth. Presumably, the ability to do so depends
on credit conditions in the market. As an example, Bernstein et al. (2019) document that investment
decreased for all portfolio companies after the financial crisis, though private equity backing mitigated the
decrease relative to matched controls. Davis et al. (2019) find that productivity improvements at target
companies are larger when credit conditions are tight. The type of company invested in may also differ at
different market conditions. Nanda and Rhodes-Kropf (2013) document that venture capital funds invest
in riskier and more innovative companies during hot markets.
28
Arcot et al. (2015) and Degeorge et al. (2016) document that secondary buyouts perform poorly
when made late in the investment period if the GP has significant unspent capital. Among venture capital
funds, Barrot (2017) find that venture capital funds with longer remaining horizons invest in companies
at earlier stages of their development. In terms of GP age, Ljungqvist et al. (2020) document that young
GPs invest in riskier companies to establish a track record. I return to this when examining the impact of
GP reputation on the estimates.

114
estimates suggest that a one standard deviation increase in ∆j,t is associated with
4.1% to 4.6% higher growth rates.
While these results are interesting, they may simply reflect an increased tendency to
engage in capital-intensive growth strategies when ∆j,t is high. Boucly et al. (2011) and
Chung (2011) find that both sales and CAPEX increase following a buyout, which they
interpret as evidence that private equity ownership relaxes credit constraints. Similarly,
add-on acquisitions are common in buyouts, taking place in at least 41.3% of deals
in Cohn et al. (2020).
To examine if this is driving the relationship of ∆j,t with company growth, Panel
C examines growth in total assets as the dependent variable. Somewhat surprisingly,
the relationship between ∆j,t and asset growth is not only statistically insignificant, the
coefficient is negative at all horizons. It appears as if the growth at the company level is
not due to more capital being spent, but instead reflects higher organic growth.29
Panel D presents results examining whether ∆j,t is associated with improved prof-
itability at the portfolio companies. While the coefficient is positive, it is both statistically
insignificant and economically small.
The results provide evidence in support of Hypothesis 1 as higher GP pay-for-
performance sensitivity is positively related to increased company growth that is not
due to increased capital expenditure. To the extent that GPs improve profitability at
their portfolio companies, such improvement appears to not be related to the pay-for-
performance sensitivity of the GP’s payoff.
Given the relatively small sample size of the paper, one may naturally worry that
the documented results are dependent on the specific set of controls included. To
examine the robustness of the results presented, Table A.3 summarizes the main result on
revenue growth for various model specifications. Regardless of specification, the estimated
coefficient is always positive. As in the main table, it is not statistically significant for
the t − 1 to t + 1 horizon. For the t − 1 to t + 2 horizon it is statistically significant in
four out of seven specifications. For the t − 1 to t + 3 and t + 4 horizons the coefficient is
statistically significant in every specification except for the t + 3 horizon with no controls
included. Overall, the result appears robust, especially for the longer time horizons.30
29
Total assets is examined here as it incorporates goodwill that often arises from add-on acquisitions.
However, growth in total assets also captures goodwill from the initial buyout. In untabulated results I
strip out the effect of growth in goodwill and other intangible assets, by replacing total asset growth with
growth in tangible assets. The qualitative patterns in Panel C remains unchanged for all tables presented
in the paper using this alternative measure of capital expenditure.
30
For brevity, similar tables are not presented for the other three measures of operational improvements.
For employee growth, the results are similar to those in Table 4 for all specifications, with a significant
positive coefficient for the horizons t − 1 to t + 2 and t + 3. The main exception is a lack of significance when
only including time fixed effects. The relationship with asset growth remains statistically insignificant,
and is negative in all specifications with the exception of the t + 1 horizon for the first four specifications.
For profitability margin the relationship is positive in almost all specifications, while always economically
small and statistically insignificant.

115
 Table A.3 suggests that the explanatory power of ∆j,t is small, particularly on its
own. In the baseline specification with all controls, the inclusion of ∆j,t increases the
adjusted R2 by between 0.020 to 0.026 in the three longer horizons. For the t − 1 to
t + 1 horizon the contribution is negligible.
Another concern is that the results may reflect the continuation of a pre-trend, where
GPs with high pay-for-performance sensitivity acquire companies with higher growth
rates. As the estimation is essentially a continuous difference-in-difference estimate, we
want to ensure that we have parallel trends prior to the buyout.
To this end, we can pool observations in the years around the buyout, and estimate
separate coefficients for the fund’s ∆j,t for each year around the buyout event. Let t
denote the (fiscal) year in which the buyout takes place, and t + s the years around the
buyout. We can formally test the parallel trends assumption by estimating
5
Yi,j,t+s =αi + λs · ∆j,t + φj + φt + βk 1{fund aget = k} (4)
X X

s k=1

+ γs + Γ1 Xi,t−1 · P ostt+s + Γ2 Xi,t · P ostt+s + i,j,t+s ,

X

where Yi,j,t+s is the company revenue at time t + s, which is s years from the buyout.
Unlike the previous tables, revenue is measured in levels, and consequently we will use log
revenue. αi is a company fixed effect. We estimate a separate coefficient λs for ∆j,t for
each year around the buyout, with t−1 left as the reference year. Similarly, γs are indicator
variables for each of the years around the buyout, with t − 1 as the reference year. As
before, φj , and φt denotes fund fixed effects and fixed effects for the year of the transaction,
respectively. The company control variables Xi,t−1 and Xi,t are here interacted with a
post dummy, taking the value 1 for all years after the buyout event, and 0 otherwise.
Table A.4 presents the results of this estimation. Columns 1–3 have log revenue as
the dependent variable while columns 4–6 presents the results for the EBITDA margin.
Overall, the results are largely consistent with the main table, with a positive relationship
between a GP’s pay-for-performance sensitivity and revenue growth, while there is no
consistent relationship with profitability. As in the main table, the relationship between
∆j,t and revenue is stronger at longer horizons.
It is worth nothing that the coefficients on ∆j,t and revenue are consistently positive,
albeit statistically insignificant, in the pre-buyout years. While it might have been more
comforting if they had been closer to zero, the pattern does not cause much reason to
be alarmed. If the trend observed was one of a monotonically increasing coefficient,
whose increase continued post-buyout, it would appear as if a trend simply continued.
This is not the case however. All estimates are relative to the reference year t − 1.
A positive coefficient for ∆j,t in t − 3 therefore suggests that companies acquired by

116
GPs with a high pay-for-performance sensitivity experienced a slower revenue growth
between t − 3 and t − 1, compared to the average company. In all specifications, the
coefficient is lower in t − 2 than in t − 3. If anything, the trend is going in the opposite
direction and is reversed following the buyout.
While untabulated for brevity, the qualitative patterns are similar to the main table
for the log of total assets and the log of number of employees. The results presented
in the next two subsections are also qualitatively similar to those examined here when
estimating a parallel trends specification.

4.1.1 Impact of reputation

The measure of pay-for-performance sensitivity used in this paper covers the direct part of
pay for performance. However, performance is also relevant for fundraising success and thus
indirectly providing pay for performance. Performance matters less for fundraising success
of highly reputable GPs (Chung et al., 2012; Barber and Yasuda, 2017). Therefore, the
direct component of pay for performance is relatively more important for high reputation
GPs. We now turn to examine Hypothesis 2, which predicts that the observed relationship
between ∆j,t and company growth should be stronger for highly reputable GPs.
The paper follows Barber and Yasuda (2017) in defining a low reputation GP. A
GP is classified as low reputation if it has: cumulatively raised less than $1 billion
prior to the current fund; have raised less than three funds in the past; and have no
past top quartile funds that are five years or older at the time the new fund is raised.
High reputation is the complement to low reputation funds. To construct this measure
I utilize the Preqin universe of funds, and rely on their classification of the quartile
in which a given fund’s performance belongs to. While their classification is based on
IRRs, there are many more funds with this information available than there is with
cash flow data which would allow me to manually construct a quartile ranking based
on, for example, a public market equivalent.
Table 5 presents the results from estimations of an augmented version of (3), where
a separate coefficient for ∆j,t is estimated for high and low reputation GPs, and by
adding a dummy variable indicating whether the acquiring fund in a given deal belongs
to a low reputation GP.
The table makes clear that the results in Table 4 are largely driven by high reputation
GPs, where we expect the direct pay-for-performance sensitivity of payoffs to be relatively
more important. The one exception to this general trend is the t − 1 to t + 4 horizon, where
the estimated coefficient, though statistically insignificant, is higher for low reputation
funds. It is worth noting that at this horizon, there are only 17 observations relates
to transactions made by low reputable funds.

117
 The statistical and economic significance of the estimates are generally higher for
high reputation funds compared to the baseline specification. For revenue growth, the
point estimates suggest that a one standard deviation increase in ∆j,t is associated with
as much as 5.2% higher annual growth rate, though it is reduced to 2.7% at the t + 4
horizon. The corresponding numbers for employee growth are a 5.4% and 4.5% increased
growth rate for the t + 2 and t + 3 horizons, respectively. The pattern is similar to
the baseline specification for growth in total assets and profitability improvements, with
all coefficients being statistically insignificant.
While not tabulated, the estimated coefficients for ∆j,t are only statistically different
for high and low reputation GPs in a few model specifications. In panel A the high-
reputation ∆j,t is statistically higher than the low reputation one at the 10% level for
the t − 1 to t + 1 and t + 2 horizons, and in Panel D at the at the 5% level in the t − 1
to t + 2 horizon. None of the other estimated coefficients for high-reputation ∆j,t are
statistically different from the estimated coefficients for low-reputation ∆j,t .
The evidence is therefore largely consistent with Hypothesis 2, but we cannot rule
out that the observed pattern reflects imprecisely estimated coefficients. In particular,
the evidence for the t − 1 to t + 4 horizon is not consistent with the hypothesis.
At this point it is worth to consider an alternative explanation to our results. It
is possible that the increased growth is not caused by any type of added value to the
company but rather a shift in risk. If high-growth strategies are inherently risky, the
higher growth rate could be accompanied by an increased risk of default. While there is no
apparent relationship between ∆j,t and a deterioration in profitability, this could be due
to a form of survivorship bias in the data: companies that enter insolvency processes tend
to not submit their accounts in the last fiscal year prior to entering insolvency processes.
Similar omissions may as well take place when creditors seize control, if it happens close
to the end of a fiscal year. If high-growth strategies are riskier, we would then observe
more years of the ex post successful ones. With 23.3% of all realized exits in the sample
analyzed here end up with the equity position wiped out, this is a serious concern.
If this drives our results, we expect a fund’s ∆j,t to be positively related to the
probability of default. Table A.5 presents estimates of the probability that a deal ends up
in formal insolvency procedures, gets seized by creditors, or the private equity firm loses
control in a financial restructuring. The first five columns present OLS estimates of a
linear probability model, while the last five columns present estimates from a logit model.
An OLS estimate is presented as an alternative due to the inclusion of fixed effects in
some specification, which may lead to an incidental parameter problem.
Panel A presents estimates for the relationship of a fund’s delta to the risk of
bankruptcy, while Panel B allows for a differential effect depending on whether a fund

118
is managed by a high or low reputation GP. All estimated coefficients in Panel A are
statistically insignificant, are are consistently negative across all specifications. In Panel B,
it is clear that the negative coefficient is driven by high-reputation funds, as the coefficient
is larger in magnitude, while coefficients for low-reputation funds are consistently positive.
Again, none of the coefficients are statistically significant, and neither is the difference
between coefficients of high- and low-reputation GPs. As our previous results refer to
high-reputation funds, these estimates suggest that it is unlikely that they are driven
by risk taking and survivorship bias in the data.31

4.1.2 Deals made close to fundraising events

As the implicit incentive largely stems from the need to perform well in your current
fund to raise another one, studying periods around successful fundraising could be
informative. There are three reasons to expect a stronger relationship for deals made
close to a fundraising event. First, Larocque et al. (2019) document that IRR is of higher
importance to investors than money multiples in decisions to participate in new funds. As
early cash flows are disproportionately influencing IRRs, the implicit incentive to perform
well is much stronger in a newly raised fund. Second, to the extent that investors care
mostly about the performance of the most recent fund, a successful fundraising lowers the
relative importance of implicit incentives in the old fund. Third, when allocating scarce
resources between funds, GPs may shift resources to the one with the most potential
gains. The ∆j,t of the old fund is then increasingly relevant relative to that of the new
fund; if a GP is currently far out of the money there is little incentives to spend scarce
resources on the older fund. Taken together, we expect the importance of ∆j,t to be
larger for deals executed close to a fundrasing event.
To operationalize this, I define the end of a successful fundraise as the first quarter in
which the successor fund has a capital call or, in the absence of cash flow data, the second
quarter of the vintage year of the successor fund. Fundraising campaigns may last up to
more than a year and a half (Barber and Yasuda, 2017), and the first capital call is likely
to occur at some point after successfully raising a new fund. Consequently, a transaction
is therefore classified as being completed “close to fundraising” if it is completed within
eight quarters of the fundraising. This clearly captures some buyouts completed during
the fundraising period, and likely some completed prior to the fundraising campaign
started. It is still considered a useful albeit arbitrary cutoff, as the GP would have
to fundraise with to-date fund performance which is unlikely to include very recent

31
For brevity I do not present similar analysis for the next two subsections, though the patterns are
qualitatively similar.

119
transactions.32 Additionally, most of the evaluated performance of the portfolio company
will cover the period outside of the fundraising period.33
Table 6 presents the results of estimating separate coefficients for ∆j,t for transactions
completed close to fundraising and those far from fundraising. Compared to the baseline
estimations, the estimated coefficient on ∆j,t is higher when the fund is close to fundraising
for both revenue growth and employee growth, for all horizons except the t − 1 to t + 4
one. The economic magnitude for a one standard deviation increase in ∆j,t translates to
between 6.0% and 8.4% higher annual revenue growth for the three horizons for which
the coefficicents are statistically significant. The corresponding numbers for employee
growth are between 5.7% and 7.1% higher growth rates. Interestingly, total asset growth
is negatively related to ∆j,t for deals completed close to fundraising at the t − 1 to t + 2,
t + 3, and t + 4 horizons. This suggests that, if anything, the estimates on revenue growth
are understated. There is also weak evidence of improved profitability at the t − 1 to t + 4
horizon, but as it is not consistent over multiple horizons it is not very convincing.
For revenue growth, the estimated coefficients are larger for deals completed close to
fundraising events than those far from fundraising events. The difference is statistically
significant at the 10% level for the first three horizons. A similar picture is true for
employee growth, though the coefficients are only statistically different at the shortest two
horizons at the 10% level. For total asset growth, the coefficients for statistically smaller
for deals completed close to fundraising events than those estimated far from fundraising
events, significant at the 5% level for the t + 2 horizon, and at the 1% level at the t + 3
and t + 4 horizons. The profitability estimates are statistically different at the 5% level at
the longest horizon. Overall, the results are largely consistent with Hypothesis 3.

4.1.3 Public market movements as exogenous shocks

While previous sections have utilised controls to attempt to control for expected con-
founding factors, it is difficult to argue that the remaining variation is truly exogenous.
Here we will utilize changes that is outside the control of GPs: movements in the stock
market. The aim is to get heterogeneous variation in our measure of pay-for-performance
sensitivity from an exogenous shock, while at the same time control for overall economic
conditions that may otherwise be related to operational improvements.

32
On a practical note, there are only 7 transactions in the sample that took place after the defined
fundraising event.
33
The results presented are robust to alternative windows defining a transaction as being “close to
fundraising” if it is completed between ten to six quarters prior to a fundraising event. The definition
used here would classify a fund as being “far from fundraising” if a successor fund was never raised. In
Brown et al. (2019) they assume that such funds attempt to fundraise towards the end of the fund life. In
this paper such assumption would classify all such transactions as being far from fundraising, leading to
no practical change.

120
 The intuition for using market shocks is relatively straight-forward. The measure
of pay-for-performance sensitivity depends on a fund’s performance, which is partly
determined by the current value of the fund that is impacted by public market movements
(e.g. Ang et al., 2018). For funds operating in a given market, a market shock is common.
Despite the shock being common, the pay-for-performance sensitivity measure will not
be impacted equally for all funds. There are two main reasons for this.
First, the non-linear payoff structure of carried interest implies that the effect on
∆j,t from a shock depends on the fund’s current performance. A fund far out of or far
in the money is barely impacted, whereas a fund at the money sees a big impact from
a change in valuations. Second, the degree to which funds are exposed to any given
market shock depends on the fraction of capital that is currently invested, largely a
function of fund age. As an extreme example, a fund that has yet to make its first
investment has no exposure, as it owns no companies whose valuations are impacted.
The varying sensitivity to performance as a function of fund age was demonstrated in
Figure 1. This heterogeneous response to a common, exogenous, shock provides for the
potential to get utilize exogenous variation in the measure. This strategy is conceptually
similar to that used in Shue and Townsend (2017).
I assume that market movements impact the NAV of a fund by rm,t . Define a
counterfactual NAV
] j,t as the NAV the fund would have had absent the market movement
rm,t . For fund j at time t, it is defined as

] j,t = (NAVj,t − CCj,t + Dj,t ) + CCj,t − Dj,t ,

NAV
1 + rm,t

where CCj,t and Dj,t denotes capital calls and distributions of fund j during quarter t,
respectively. The impact of rm,t on ∆j,t is given by decomposing

∆j,t = ∆
e + ∆m ,
j,t j,t (5)

where the estimation of ∆

e
j,t uses NAVj,t in place of NAVj,t . ∆j,t is defined as the
] m

difference ∆j,t − ∆
e , and is interpreted as the part of ∆ that is attributable to the
j,t j,t

market shock. Fund NAV valuations as well as capital calls and cash flows are available
at a quarterly level, and rm,t is measured over the same interval. As the funds in the
sample may have different geographical exposure, the precise market return used is
either S&P500, FTSE250, or Eurostoxx 600 if the fund is focused on the US, UK, or
Europe, respectively. In practice, US-focused funds are not very common as the sample
of portfolio companies are registered in the UK.
Table 7 presents the results from using only the part of pay-for-performance sensitivity
that is attributable to a market shock. Overall the table confirms the main result of the
paper: higher pay-for-performance sensitivity of a fund is associated with higher company

121
growth. For revenue growth, the estimated coefficient is statistically significant at all
horizons except for the t − 1 to t + 3 horizon, and only at the 10% level for the t − 1
to t + 4 horizon. The economic magnitude for a one standard deviation change in ∆m
j,t
translates to an increased growth rate of between 1.8% and 3.9%.
Compared to the baseline table, the evidence is weaker in support of a relationship
with employee growth. The coefficient is only statistically significant at the 10% level
for the shortest two horizons, while there is no support for higher employee growth
at longer horizons. As before, the coefficient is negative but statistically insignificant
for total asset growth, while for profitability it is mostly positive, always economically
small, and never statistically significant.
By constructing the measure of ∆m
j,t in this way it is implicitly assumed that a PE
fund immediately and fully incorporates any market movement in NAV valuations, and
that a beta of one is appropriate. It is not clear that this is the right approach. Ang
et al. (2018) report an autocorrelation of 0.40 for the Cambridge Associates buyout
index, suggesting that there is some lag in incorporating overall market movements.
As an alternative I therefore utilize the extreme other end, assuming that the lag is a
full quarter. This implies that reported NAVs correspond to the “counterfactual” of
what it would have been absent the market shock, and that the true NAV is given
0
by adding the impact of a market movement. Using NAV
] j,t to denote the NAV after
correcting for the market movement, we have
0
NAV
] j,t = (NAVj,t − CCj,t + Dj,t ) (1 + rm,t ) + CCj,t − Dj,t ,

where CCj,t and Dj,t again denotes capital calls and distributions of fund j in quarter
t, respectively. This estimate is used to estimate a ∆
e 0 by replacing the reported NAV
j,t
0
by NAV
] j,t . The impact of a market movement ∆m j,t is given by

e 0 − ∆ = ∆m
∆ (6)
j,t j,t j,t

Table A.6 presents results using this alternative definition of ∆m

j,t , and they are overall
less convincing than the original specification. For revenue growth, the coefficients are
only statistically different from zero at the t − 1 to t + 1 and t + 2 horizons, and only
at the 10% level. The economic magnitude is also lower, where the highest impact for
one standard deviation change in ∆m
j,t translates to 2.7% higher revenue growth. There
is no evidence of a positive impact on employee growth here, and the coefficient is even
negative, though not statistically different, at the longest two horizons. The results are
qualitatively similar for total assets growth and profitability improvements.
As the results are markedly less convincing using this alternative approach, it begs the
question as to which assumption is the more realistic one. The positive autocorrelation

122
in Ang et al. (2018) suggests that while market movements in a given period are not
fully incorporated, they are at least partially incorporated. The literature documents
a range of market beta estimates for private equity funds. Jegadeesh et al. (2015) use
prices of listed fund-of-funds investing in unlisted private equity funds and find a beta of
1, Driessen et al. (2012) and Franzoni et al. (2012) find a beta of 1.3 for buyout funds,
while Ang et al. (2018) estimate a market beta ranging from 1.18 to 1.77 depending on
model specification. If returns are partially incorporated, and a relevant beta is in excess
of one, the appropriate correction is likely closer to the first approach.
In untabulated analysis I replicate Tables 5 and 6 using ∆m
j,t in place of ∆j,t . The
impact of high reputation is very similar to what is seen in Table 5. Both statistical and
economic significance is larger for the estimated coefficients on ∆m
j,t for high reputation
GPs, compared to estimates in Table 7. Coefficients that were not significant previously
do not become significant. For revenue growt, the coefficients for high reputation GP are
higher than the ones for low reputation GPs, and the differences are statistically different
on at least the 5% level for all horizons except for t − 1 to t + 4. The patterns described
in this paragraph holds true for both versions of ∆m
j,t estimated.
Looking instead at deals completed close a fundraising event the evidence weaker
in support of Hypothesis 3. For revenue growth, the estimates are not consistently
larger for deals completed close to a fundraising event, and the difference in estimated
coefficients is never significant. Only the t − 1 to t + 1 horizon have larger estimated
coefficients for deals close to fundraising compared to the baseline, and it is the only
horizon where any coefficient is statistically significant from zero. Again this pattern
is true for both versions of ∆m
j,t .
For employee growth, the estimated coefficients are higher compared to the baseline
for all horizons but t − 1 to t + 4, but the difference between the estimated coefficients
on ∆m
j,t for deals completed close to a fundraising event are not statistically different
to those completed far from a fundraising event. This pattern is only true for the
first version of ∆m
j,t . For the second version of ∆j,t , neither estimated coefficients are
m

statistically different from zero for employee growth. There are no significant results
for total asset growth and profitability improvements, and no noteworthy patterns that
are different from the baseline estimation.

4.1.4 Summary

The analysis presented tests the hypotheses outlined in section 2.1.1. Hypothesis 1 predicts
a positive relationship between operational improvements in portfolio companies and the
pay-for-performance sensitivity of a GP’s payoff in the acquiring private equity fund. The
evidence is largely supporting this hypothesis in terms of being positively associated with

123
company growth, measured either as revenue growth or growth in number of employees.
There is no evidence of a positive relationship with profitability improvements. The
evidence presented is generally robust to model specification. The result is also robust
to using only exogenous variation in the pay-for-performance sensitivity, utilizing shocks
to the equity market as an exogenous shock.
Hypothesis 2 predicts that the positive relationship in Hypothesis 1 should be more
pronounced for highly reputable GPs, as the implicit incentive is expected to be of relatively
higher importance for low reputation GPs. The evidence is largely supportive of the
hypothesis whether using the main specification or only the exogenous variation in pay-for-
performance sensitivity. Though the documented patterns are very consistent, one needs
to be cautious as the difference in estimated coefficients is not always statistically different.
Hypothesis 3 predicts that the positive relationship in Hypothesis 1 should be more
pronounced when a fund is close to raising a new fund, as the implicit incentive is expected
to be much stronger for the newly raised fund. The evidence is mixed in support of this
hypothesis. While the baseline specification is generally supportive of the hypothesis, it is
not robust to using only exogenous variation in the pay-for-performance sensitivity.
It is worth considering some limitations to the analysis of the paper. First, the analysis
has focused on operational improvements at portfolio companies. However, this is merely
a proxy for the measure that GPs truly care about: deal returns. While improvements
in operational performance of a portfolio company improve deal returns (e.g. Guo et al.,
2011), it is not the only avenue through which returns can be generated. It is comforting
that the documented growth does not appear to be generated via spending more capital,
as it is unclear whether such growth through capital expenditure improves deal returns.
The focus on operational improvements is made out of necessity due to a lack of available
deal-level return data. We may expect that relying on this proxy adds noise to our tests,
which would make it less likely that we find any results.
Second, only the explicit incentive provided by direct pay for performance provided is
formally modelled and tested in this paper. Clearly, implicit incentives play an important
role as demonstrated by Chung et al. (2012). While including a model incorporating
both explicit and implicit incentives is beyond the scope of this paper, the second and
third hypotheses utilizes implications of the role of implicit incentives to sharpen the
predictions of the paper. An interesting avenue for future research is to jointly examine
the role of explicit and implicit incentives.34
34
The implicit incentive from the increased ability to raise a fund when performing better is likely
a matter of performance relative to other funds, as suggested by results in Barber and Yasuda (2017).
By contrast, hurdle rates tend to be fixed. The identification strategy used in this paper of utilizing
exogenous market shocks therefore appears like one interesting approach for disentangling the implicit
and explicit incentives, while changes in the supply of capital from potential LPs could potentially be
used to instrument shocks to implicit incentives.

124
 Third, the model used to estimate pay-for-performance sensitivity of carried interest
treats all funds as being equal. In reality, the terms facing specific funds will differ
between funds, and as such the resulting delta would differ from the one estimated.
Additionally, GPs may have subjective beliefs about their ability that deviates from the
ones imposed by the model. This implies that the pay-for-performance sensitivity faced
by a GP differs from the one suggested by the model. To the extent that GPs have
rational beliefs, the model would underestimate (overestimate) the pay-for-performance
sensitivity of GPs with high (low) ability early in a fund’s life. Such measurement
error should work against the results of the paper, as high (low) ability GPs experience
a smaller increase (decrease) in pay-for-performance sensitivity following good (poor)
performance compared to what is suggested by the model. The results from utilizing
exogenous market shocks may be less exposed to these errors, as it compares a difference
in pay-for-performance sensitivity estimated using the same model. The differencing may
remove part of the impact from a misspecified model, though it is unlikely to completely
resolve the issue for the same reasons that a common shock generates heterogeneous
impacts on the pay-for-performance sensitivity measure.
Finally, it is important to remember that the sample size is relatively small and
limited to UK companies.

5 Conclusion

Carried interest in private equity funds provides a strong direct component of pay-for-
performance as it gives the private equity firm a fraction of fund profits. Carried interest
is typically paid out only if fund performance exceeds a pre-set, fixed, hurdle rate. This
makes carried interest akin to a call option, with the hurdle rate as its strike price. As
with call options, the sensitivity of pay to improved performance is largely a function of
the likelihood of being in the carry. Consequently, the strength of the pay-for-performance
varies over a fund’s life, and this paper examines these dynamics.
The paper builds on the framework in Metrick and Yasuda (2010) to measure expected
pay-for-performance during a fund’s life. Pay-for-performance sensitivity is increasing
in fund performance, and increasingly so as the fund matures. This is analogous to call
options, where the delta of the option increase in a stock’s price, and increasingly
so as time-to-maturity decreases.
Private equity have been shown to improve operations of portfolio companies (e.g.
Bernstein and Sheen, 2016; Davis et al., 2014), which presumably takes effort on behalf
of the GP. Agency theory is clear in its predictions that higher effort is expected
when an agent’s pay is more strongly linked to exerting effort. The paper therefore

125
relates the estimated pay-for-performance sensitivity to the operational performance of
portfolio companies invested in. It does so using a newly assembled dataset on UK
companies acquired in buyouts.
Deals executed by funds at times when pay-for-performance sensitivity is high are
related to higher organic growth. Clearly, pay-for-performance sensitivity is not randomly
assigned to funds and determinants of the measure, in particular managerial ability, are
expected to be related to operational improvements at the portfolio companies. To account
for managerial ability, the paper utilizes within-fund estimations which effectively compare
deals conducted by the same fund at different points during a fund’s life. Private equity
is an ideal setting for conducting within-fund estimations due to the staggered nature of
investments. The further strengthen the interpretation the paper utilizes shocks in the
public market to get exogenous variation in the measure. Public market returns are a good
candidate for such variation as private equity returns are closely linked to public markets
(e.g. Ang et al., 2018), while funds are not equally exposed at any given time. This implies
that a common shock have heterogeneous impact on funds’ pay-for-performance sensitivity.
This paper contributes to our understanding of how the incentives of general partners
in private equity funds impact value creation at the portfolio company level. While
the aggregate value of incentives have been documented (Metrick and Yasuda, 2010;
Chung et al., 2012), this is the first paper to explore the dynamics within a fund. How
managerial incentives from pay-for-performance compensation relate to value creation
is a topic of general interest in finance, although it is notoriously difficult to causally
attribute any observed correlations to an effect of incentives, especially when studying
CEOs (e.g. Edmans et al., 2017). This paper utilizes the private equity setting with
fixed compensation terms to account for many of the complicating features present when
studying CEOs, and to generate plausibly exogenous variation in the measure.
However, the paper only examines the direct component of pay-for-performance
sensitivity. An important avenue for future research is to jointly examine the direct and
indirect components of pay-for-performance compensation.

126
127

Figure 1: Pay-for-Performance Sensitivity Summary

This figure plots the variation in a GP’s pay-for-performance sensitivity ∆, defined in Section 2.2.2, as a function of fund performance, measured as IRR, and fund age,
measured in years. The figure depicts the best fit of the model ∆j,t = α + β1 IRRj,t + β2 Fund Agej,t + β3 IRRj,t × Fund Agej,t + j,t , where j denotes a fund, and t
the fund age measured in quarters from the first cash flow of the fund. The model is fitted for the estimated pay-for-performance sensitivities of 13927 fund-quarters
for buyout funds with available NAV and cash flow data in Preqin. All fund-quarters up until the fund’s 20th quarter is included.
128

Figure 2: Sample Composition - Type of Buyout

This figure presents the sample composition of buyout types by the year in which the buyout transaction completed. The height of each bar represent the number of
buyouts in a given year as displayed on the y-axis. The percentages displayed in each colored bar correspond to the fraction of buyouts within the year of that type.
Percentages are omitted from bars containing a single deal. Buyout type categories are defined in Table D.1.
129

Figure 3: Sample Composition - Exit Route

This figure presents the sample composition of exit routes for the private equity owner of companies acquired in a buyout. The transactions are organized by the year
in which the buyout took place, while the exit may have taken place at a later date, or not yet taken place. The height of each bar represent the number of buyouts in
a given year as displayed on the y-axis. All exit routes are as of 2019-09-30. Percentages displayed in the colored bars correspond to the fraction of that year’s buyouts
that ended up with a given exit route. Percentages are omitted from bars containing a single deal. The “Other” category consists of 5 cases where the portfolio
company was owned by a private equity fund until liquidated and ceased trading. Remaining exit routes are defined in Table D.1.
 Table 1: Sample Construction

This table presents how the final sample in the paper is derived. Transactions without participation of a
private equity firm or where the private equity firm(s) only acquired a majority stake are excluded from
the sample. Company names are taken from the transaction synopsis provided by Capital IQ, and if no
UK company of that name is found the transaction is excluded. The screen for buyout transactions in
Capital IQ was undertaken 2019-01-02.

Number of UK buyouts identified in Capital IQ from 1999 - 2018 1964

Less: Buyouts not included in the 850 randomly selected (1114)

Number of PE transactions processed 850

Less: Transactions where no UK company of a stated name was found (19)

Less: Transactions without PE firm participation (15)

Less: Transactions where majority ownership did not pass to a PE firm (10)
Less: Duplicate transactions identified (7)
Less: Transactions initially identified but no longer in Capital IQ (5)

Less: Transactions of companies exempted from submitting filings (3)

Number of buyouts with Entry and Exit information 791

Less: Transactions without accounting data in year t-1 and t+1 (230)
Less: Transactions of financial companies or public utilities (69)

Initial sample with financials available 494

Less: Transactions where acquiring fund is involved in only (248)
one transaction or lack cash flow and NAV data in Preqin

Final sample 246

130
 Table 2: Sample Composition

This table presents the composition of buyout types, exit route for the private equity owner, and holding
periods. Panel A presents the composition of the type of buyout in the sample. Panel B presents exit
routes. Panel C presents descriptives on holding periods, expressed in number of years from entry to exit,
of all deals with an exit as of 2019-09-30. The time-periods represents the year in which the buyout took
place. The “Other” category consists of 5 cases where the portfolio company was owned by a private
equity fund until liquidated and ceased trading. All other variable definitions are found in Table D.1.

Panel A: Buyout Type

Full Sample 1999 - 2003 2004 - 2008 2009 - 2013 2014 - 2018
Public-to-Private 0.134 0.172 0.146 0.105 0.061
Secondary Buyouts 0.341 0.159 0.385 0.374 0.586
Divisional 0.231 0.343 0.212 0.193 0.091
Private-to-Private 0.263 0.318 0.243 0.240 0.232
Distressed 0.030 0.009 0.014 0.088 0.030
N 791 233 288 171 99

Panel B: Exit Route

Year of Entry
Full Sample 1999 - 2003 2004 - 2008 2009 - 2013 2014 - 2018
Lost Control 0.162 0.163 0.240 0.094 0.051
IPO 0.062 0.082 0.056 0.070 0.020
Trade Sale 0.290 0.382 0.288 0.263 0.121
Sale to Fin. Inst. 0.316 0.348 0.375 0.304 0.091
Still Owned 0.159 0.004 0.035 0.257 0.717
Other 0.006 0.021 - - -
Not Known 0.005 - 0.007 0.012 -
N 791 233 288 171 99

Panel C: Holding Periods (exited deals)

Year of Entry
Full Sample 1999 - 2003 2004 - 2008 2009 - 2013 2014 - 2018
Mean 4.99 4.92 5.56 4.33 2.85
Median 4.43 4.58 5.46 4.04 2.85
Standard Deviation 2.74 2.87 2.91 1.83 0.91
25th Percentile 2.99 3.04 3.01 3.22 2.52
75th Percentile 6.65 6.27 7.84 5.26 3.34
N 661 232 277 124 28

131
 Table 3: Summary Statistics

This table presents descriptive statistics of the financials of companies in the final sample. Panel A
presents characteristics as of the most recent fiscal year prior to the buyout, referred to as year -1. Panel
B presents changes in these measures from year -2 to year -1, while Panel C presents changes from year -3
to year -1. Panel D presents the industry composition.

Panel A: Portfolio company characteristics in year prior to buyout

N Mean Median SD 25th pct 75th pct
Revenue (M£) 246 157.2 69.3 248.7 23.8 166.1
EBITDA (M£) 246 19.6 8.8 28.9 3.5 20.2
Total Assets (M£) 246 156.4 66.1 240.5 18.5 168.1
Number of Employees 238 1797 614 2741 212 1919
EBITDA Margin 246 0.159 0.141 0.106 0.084 0.212

Panel B: 1-year pre-buyout growth

N Mean Median SD 25th pct 75th pct
Revenue 227 0.177 0.119 0.276 0.035 0.249
EBITDA 227 0.165 0.182 1.268 -0.019 0.502
Total Assets 229 0.176 0.071 0.391 -0.009 0.247
Number of Employees 217 0.112 0.062 0.253 -0.012 0.169
EBITDA Margin 227 0.009 0.010 0.061 -0.008 0.029

Panel C: 2-years pre-buyout growth

N Mean Median SD 25th pct 75th pct
Revenue 214 0.499 0.285 0.750 0.092 0.624
EBITDA 213 0.792 0.414 2.762 0.063 1.060
Total Assets 218 0.452 0.194 0.869 0.018 0.535
Number of Employees 203 0.300 0.112 0.653 -0.024 0.416
EBITDA Margin 213 0.020 0.012 0.071 -0.009 0.042

Panel D: Industry Composition (%)

Mining 1.3
Construction 2.1
Manufacturing 23.0
Transportation 7.6
Wholesale 4.7
Retail 16.0
Services 45.4

132
 Table 4: Operational Improvements and Pay-for-Performance Sensitivity

This table presents estimates of operational improvements in relation to the fund’s pay-for-performance
sensitivity ∆j,t . The four columns represent improvements from the year prior to the buyout, t − 1, to
one, two, three, and four years after the buyout. The dependent variable is annualized revenue growth in
Panel A, annualized employee growth in Panel B, annualized growth in total assets in Panel C, and the
cumulative improvement in the EBITDA/Revenue margin in Panel D. Each model specification includes
time, fund, and industry fixed effects; dummies for the fund year in which the transaction took place; and
log revenue, log total assets, and EBITDA margin in t − 1, as well as book leverage post transaction. All
variables are winsorized at a 1% level. Standard errors, reported in parenthesis under each estimate, are
clustered at the acquiring fund level. Estimates followed by the symbols ***, **, or * are statistically
significant at the 1%, 5%, and 10% levels, respectively.

Panel A - Revenue Growth (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆j,t 0.536 1.466∗ 1.346∗∗ 1.113∗∗∗
(0.747) (0.786) (0.629) (0.346)

Adjusted R2 0.299 0.310 0.472 0.599

N 246 218 178 127

Panel B - Employee Growth (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆j,t 0.312 1.885∗∗ 1.700∗∗ 0.489
(0.914) (0.725) (0.675) (1.024)

Adjusted R2 0.219 0.301 0.411 0.282

N 238 207 169 120

Panel C - Growth Total Assets (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆j,t −0.035 −1.508 −1.151 −1.017
(1.839) (1.356) (1.006) (1.301)
Adjusted R2 0.143 0.138 0.156 0.126
N 246 218 178 127

Panel D - Cumulative change in Return on Sales (EBITDA/Revenue)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆j,t 0.016 0.028 0.024 0.137
(0.187) (0.179) (0.180) (0.314)
Adjusted R2 0.031 0.126 0.166 0.329
N 246 218 178 127
Transaction-Year FE Yes Yes Yes Yes
Fund FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
Fund-Year FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes

133
Table 5: Operational Improvements and Pay-for-Performance Sensitivity - Impact of Fund
Reputation

This table presents estimates of operational improvements in relation to the fund’s pay-for-performance
sensitivity ∆j,t interacted with a dummy indicating if the acquiring fund is run by a low or high reputation
GP. A GP is classified as low reputation if it has: cumulatively raised less than $1 billion dollar prior to
the current fund; raised less than three funds in the past; and have no past top quartile fund that is at
least five years old at the time of the current fund. High reputation GP is the complement. The four
columns represent improvements from the year prior to the buyout, t − 1, to one, two, three, and four years
after the buyout. The dependent variable is annualized revenue growth in Panel A, annualized employee
growth in Panel B, annualized growth in total assets in Panel C, and the cumulative improvement in the
EBITDA/Revenue margin in Panel D. Each model specification includes time, fund, and industry fixed
effects; dummies for the fund year in which the transaction took place; and log revenue, log total assets,
and EBITDA margin in t − 1, as well as book leverage post transaction. All variables are winsorized at a
1% level. Standard errors, reported in parenthesis under each estimate, are clustered at the acquiring
fund level. Estimates followed by the symbols ***, **, or * are statistically significant at the 1%, 5%, and
10% levels, respectively.

Panel A - Revenue Growth (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆j,t × 1{High Rep} 0.819 2.164∗∗∗ 1.457∗∗ 1.096∗∗∗
(0.819) (0.777) (0.636) (0.262)
Fund ∆j,t × 1{Low Rep} −1.537 −0.366 0.650 2.729
(0.947) (0.733) (1.210) (1.903)
1{Low Rep} 0.280 0.415∗∗∗ 0.141 −0.460∗∗∗
(0.175) (0.138) (0.143) (0.169)

Adjusted R2 0.299 0.321 0.467 0.603

N 246 218 178 127

Panel B - Employee Growth (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆j,t × 1{High Rep} 0.669 2.240 ∗∗∗
1.877 ∗∗
0.694
(1.048) (0.743) (0.893) (0.866)
Fund ∆j,t × 1{High Rep} −1.230 0.044 0.968 5.094
(1.276) (0.663) (0.959) (4.060)
1{Low Rep} 0.275 0.319∗∗ 0.160 −0.768
(0.220) (0.144) (0.209) (0.568)

Adjusted R2 0.213 0.300 0.403 0.315

N 238 207 169 120
Time FE Yes Yes Yes Yes
Fund FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
Fund-Year FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes

134
 Table 5 — Continued

Panel C - Growth Total Assets (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆j,t × 1{High Rep} −0.136 −1.256 −1.097 −1.063
(2.165) (1.803) (1.386) (1.479)
Fund ∆j,t × 1{Low Rep} −0.251 −0.265 −0.555 −4.764
(3.842) (3.825) (3.062) (4.528)
1{Low Rep} −0.013 0.085 0.060 0.943∗
(0.403) (0.469) (0.352) (0.535)

Adjusted R2 0.135 0.128 0.155 0.127

N 246 218 178 127

Panel D - Cumulative change in Return on Sales (EBITDA/Revenue)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆j,t × 1{High Rep} 0.039 0.188 0.047 0.116
(0.251) (0.194) (0.217) (0.298)
Fund ∆j,t × 1{Low Rep} −0.478 −0.602 0.124 −1.049
(0.468) (0.410) (0.198) (1.296)
1{Low Rep} 0.061 0.102∗ −0.003 0.163
(0.088) (0.056) (0.045) (0.167)

Adjusted R2 0.040 0.159 0.154 0.365

N 246 218 178 127
Time FE Yes Yes Yes Yes
Fund FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
Fund-Year FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes

135
Table 6: Operational Improvements and Pay-for-Performance Sensitivity - Impact of Completed
Fundraising

This table presents estimates of operational improvements in relation to the fund’s pay-for-performance
sensitivity ∆j,t , interacted with a dummy indicating if a transaction was completed close to a fundraising
event, where close means at most eight quarters before the fundraising event. A fundraising event is
the first quarter a cash flow is reported in a successor fund, or the second quarter of a fund’s vintage
year in the absence of cash flow data. The four columns represent improvements from the year prior
to the buyout, t − 1, to one, two, three, and four years after the buyout. The dependent variable is
annualized revenue growth in Panel A, annualized employee growth in Panel B, annualized growth in
total assets in Panel C, and the cumulative improvement in the EBITDA/Revenue margin in Panel D.
Each model specification includes time, fund, and industry fixed effects; dummies for the fund year in
which the transaction took place; and log revenue, log total assets, and EBITDA margin in t − 1, as well
as book leverage post transaction. All variables are winsorized at a 1% level. Standard errors, reported
in parenthesis under each estimate, are clustered at the acquiring fund level. Estimates followed by the
symbols ***, **, or * are statistically significant at the 1%, 5%, and 10% levels, respectively.

Panel A - Revenue Growth (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
∆j,t × 1{Close to Fundraising} 3.466 ∗
2.904 ∗∗
2.504 ∗∗
0.768
(1.885) (1.116) (1.119) (0.886)
∆j,t × 1{Far from Fundraising} 0.372 0.789 0.461 0.135
(0.860) (0.609) (0.404) (0.513)
1{Close to Fundraising} −0.465∗ −0.301∗ −0.269 −0.052
(0.268) (0.174) (0.175) (0.115)

Adjusted R2 0.307 0.337 0.478 0.598

N 246 218 178 127

Panel B - Employee Growth (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
∆j,t × 1{Close to Fundraising} 2.957∗ 2.358∗ 2.681∗ 0.706
(1.730) (1.195) (1.542) (1.044)
∆j,t × 1{Far from Fundraising} −0.234 0.483 0.216 −0.127
(0.660) (0.530) (0.446) (0.763)
1{Close to Fundraising} −0.385 −0.201 −0.304 −0.072
(0.265) (0.196) (0.248) (0.170)

Adjusted R2 0.256 0.357 0.405 0.291

N 238 207 169 120
Time FE Yes Yes Yes Yes
Fund FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
Fund-Year FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes

136
 Table 6 — Continued

Panel C - Growth Total Assets (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
∆j,t × 1{Close to Fundraising} −3.232 −4.834 ∗
−4.471 ∗∗∗
−4.278∗∗∗
(4.052) (2.448) (1.202) (1.106)
∆j,t × 1{Far fom Fundraising} 1.758 0.613 0.278 0.400
(1.989) (1.244) (1.119) (1.199)
1{Close to Fundraising} 0.607 0.764∗ 0.706∗∗∗ 0.763∗∗∗
(0.746) (0.407) (0.258) (0.257)

Adjusted R2 0.135 0.128 0.155 0.127

N 246 218 178 127

Panel D - Cumulative change in Return on Sales (EBITDA/Revenue)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
∆j,t × 1{Close to Fundraising} 0.226 0.072 0.177 0.506∗
(0.598) (0.336) (0.340) (0.262)
∆j,t × 1{Far from Fundraising} −0.071 −0.140 −0.089 −0.203
(0.218) (0.202) (0.168) (0.272)
1{Close to Fundraising} −0.048 −0.016 −0.029 −0.093
(0.110) (0.071) (0.064) (0.057)

Adjusted R2 0.040 0.159 0.154 0.365

N 246 218 178 127
Time FE Yes Yes Yes Yes
Fund FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
Fund-Year FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes

137
Table 7: Operational Improvements and Pay-for-Performance Sensitivity due to Market Shocks

This table presents estimates of operational improvements in relation to the part of a fund’s pay-for-
performance sensitivity that is due to a recent market shock, ∆m j,t , as defined in section 4.1.3. In this
table, the NAV of the PE fund is assumed to have fully incorporated the market shock. The four columns
represent improvements from the year prior to the buyout, t − 1, to one, two, three, and four years
after the buyout. The dependent variable is annualized revenue growth in Panel A, annualized employee
growth in Panel B, annualized growth in total assets in Panel C, and the cumulative improvement in the
EBITDA/Revenue margin in Panel D. Each model specification includes time, fund, and industry fixed
effects; dummies for the fund year in which the transaction took place; and log revenue, log total assets,
and EBITDA margin in t − 1, as well as book leverage post transaction. All variables are winsorized at a
1% level. Standard errors, reported in parenthesis under each estimate, are clustered at the acquiring
fund level. Estimates followed by the symbols ***, **, or * are statistically significant at the 1%, 5%, and
10% levels, respectively.

Panel A - Revenue Growth (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆m
j,t 4.671 ∗∗
3.605 ∗∗
0.765 2.188∗
(1.840) (1.504) (0.997) (1.155)
Adjusted R2 0.305 0.314 0.469 0.596
N 246 218 178 127

Panel B - Employee Growth (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆m
j,t 3.718 ∗
3.111 ∗
1.559 0.185
(1.911) (1.622) (2.178) (3.056)
Adjusted R2 0.225 0.299 0.407 0.272
N 238 207 169 120

Panel C - Growth Total Assets (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆m
j,t −0.040 −0.395 −0.724 −1.876
(7.725) (4.560) (3.220) (4.809)
Adjusted R2 0.140 0.135 0.151 0.114
N 246 218 178 127

Panel D - Cumulative change in Return on Sales (EBITDA/Revenue)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆m
j,t −0.030 0.019 0.011 0.099
(0.175) (0.168) (0.177) (0.299)
Adjusted R2 0.030 0.114 0.154 0.316
N 246 218 178 127
Time FE Yes Yes Yes Yes
Fund FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
Fund-Year FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes

138
Appendices
A Appendix Figures and Tables

139
 $100u2 , Cuu = 0.2 max 100u2 − 100, 0

$100u
p

$100 $100ud, Cud = 0.2 max (100ud − 100, 0)

1−
p
$100d

$100d2 , Cdd = 0.2 max 100d2 − 100, 0

Figure A.1: Illustration of a Simplified Portfolio Company Path

This figure depicts a simplified illustration of possible paths from a simulation of the value of a private
equity fund. It corresponds to the value of a fund that makes a single investment of $100 at t = 0 and is
liquidated at t = 2. The investment increase in value by u > 1.01 with probability p, and decrease in
value with d = 1
u
with probability 1 − p. At each potential end node, Cuu , Cud and Cdd denotes the value
of the payoff to the GP from carried interest at the node. Carried interest is calculated in the absence of
a hurdle rate to simplify the illustration, and the GP receives 20% of any profits.

140
 ∆ = 0.2u2
Cuu
Cu∆ = 0.2u

∆1u = e−r 0.2 (up + d(1 − p))

p

∆0 = e−2r 0.2 u2 p2 + 2udp(1 − p) Cu∆ = 0.2u

∆ = 0.2ud
Cud
Cd∆ = 0.2d
1−
p

∆1d = e−r 0.2up

∆ = 0
Cdd
Cd∆ = 0

Figure A.2: Illustration of Calculation of Pay-for-Performance Sensitivity Measure

This figure complements figure A.1 by providing the value of a GP’s ∆ at every node prior to t = 2
for the same model setup outlined for that figure. ∆t is defined as the present value of the expected
increase in carried interest paid to the GP from adding $1 of additional value at node t. When relevant, a
subscript indicates the path taken from the original node at t = 0 to the specific ∆t . C ∆ denotes the
increase in carried interest paid at the t = 2 nodes from increasing the value of the investment by $1 at an
earlier node. The subscripts to C ∆ indicates the path taken from the node at which value was added. For
example, Cuu
∆
is defined as the increase in carry payment Cuu from adding $1 of additional value at the
start node at t = 0. Without additional value added, the carry payment Cuu = 20(u2 − 1), while with the
additional value added it is Cuu = 20(u2 − 1) + 0.2u2 , making the difference Cuu
∆
= 0.2u2 . C ∆ is positive
at the middle node as the fund profit in absence of any additional value added is zero, and no hurdle rate
is used in this illustration. C ∆ is zero at the bottom node at t = 2 since the added value is not sufficient
to offset the loss in value from the path taken there.

141
142

Figure A.3: Number of Deals - Comparison Random Sample and Full Set

This figure depicts the distribution of buyouts per year for the deals that are part of the sample analyzed in the paper, as well as for the full set of UK buyouts
identified in the transaction scan done in Capital IQ. The height of each bar represents the number of buyouts in a year. The left-hand side y-axis refers to the sample
in the paper, while the right-hand side y-axis represent all UK buyouts identified in Capital IQ.
143

Figure A.4: Exit Route by Year of Exit

This figure depicts the sample composition of the exit route of the private equity owner from companies acquired in a buyout. The transactions are organized by the
year in which the exit took place. Companies still under private equity ownership as of 2019-09-30 are not included in the figure. The height of each bar represents the
number of buyouts exited in a given year and is displayed on the y-axis. Percentages displayed in the colored bars correspond to the fraction of the exits in a year that
were of that specific exit route. Percentages are omitted from bars containing a single deal. The “Other” category consists of 5 cases where the portfolio company was
owned by a private equity fund until liquidated and ceased trading. Remaining exit routes are defined in Table D.1.
144

Figure A.5: Outcome Per Year of Active Buyouts

This figure depicts the composition of the outcome of all active buyouts per year. The height of each bar represents the number of companies owned by private equity
during at least part of the year. Percentages displayed in the colored bars correspond to the fraction of deals in a year where private equity either: lost control; sold
the company; acquired the company; or retained the company. The “Sale” category contains the exit types “IPO”, “Trade Sale”, and “Sale to Financial Institution”.
A “Retained Company” is one that is owned at the start of the year and still held at the end of the year. Percentages are omitted from bars representing less than 1%
of active deals of the year. The “Other” category consists of 5 cases where the portfolio company was liquidated and ceased trading. All exit routes are defined in
Table D.1.
145

Figure A.6: Holding Periods

This figure depicts the distribution of holding periods of all buyouts completed between 1999 and 2015 and that are fully exited (659 in total). The figures are
organized by the year in which the buyout took place. The holding period is measured as the number of years from the date the transaction was completed until the
date of the exit, displayed on the y-axis.
 Nine Individuals ECI Partners

28.8% 71.2%

MPM Topco Consolidating accounts

Six Individuals MPM Bidco

MPM Products MPM Products

Pre-buyout Post-buyout

Figure A.7: Corporate Structure Example

This figure depicts the change in ownership structure following a buyout of MPM Products by ECI
Partners in 2016. Prior to the buyout, the company consisted of a single entity directly owned by six
individuals. To facilitate the acquisition, ECI created two vehicles, “MPM Bidco” and “MPM Topco”.
MPM Bidco is wholly owned by MPM Topco, and is the entity acquiring all shares of MPM Products.
Following the buyout, MPM Topco prepares consolidated group accounts and the ultimate ownership of
MPM topco is split between ECI Partners and nine individuals. ECI Partners’ ownership stake in MPM
Topco goes through an additional three layers of wholly owned entities which are omitted from the figure.
The ownership percentages are compiled from the first Confirmation Statement submitted by MPM Topco
in March 2017.

146
 Table A.1: Pay-for-Performance Sensitivity and Fund Performance

This table presents estimates of the estimated pay-for-performance sensitivity ∆j,t of a PE fund j estimated
in quarter t of the fund’s life. ∆j,t is estimated as described in Section 2.2.2 for all quarters and all buyout
funds with NAV valuations and cash flow data available in Preqin. All models include an intercept that is
omitted from the table. Fund IRR is measured at the time the delta is estimated, using the reported NAV
as a final distribution. Fund age is measured in number of quarters since the first fund cash flow. All
variables are winsorized at a 1% level. Standard errors, reported in parenthesis under each estimate, are
double clustered at the PE fund level and the fund age level. Estimates followed by the symbols ***, **,
or * are statistically significant at the 1%, 5%, and 10% levels, respectively.

Dependent Variable: Pay-for-performance sensitivity ∆j,t

(1) (2) (3)
Fund IRR 0.0338 ∗∗∗
−0.0151∗∗
(0.0060) (0.0067)
Fund Age −0.0007∗ −0.0019∗∗∗
(0.0004) (0.0005)
Fund IRR× Fund Age 0.0095∗∗∗
(0.0016)

Adjusted R2 0.136 0.012 0.468

N 13927 13927 13927

147
 Table A.2: Summary Statistics - Comparison with Full Buyout Sample

This table presents a comparison of descriptive statistics between companies in the sample and the full set of companies with financials available. The two sample
constructions are described in Table 1. Panel A compares statistics for the most recent fiscal year prior to the buyout, referred to as year -1. Panel B presents the
change in these accounting measures from year -2 until year -1. Panel C presents changes from year -3 until year -1. Panels D, E, and C compares the buyout types,
exit routes, and industry composition, respectively. The final column in panels A, B and C tests differences in means, and numbers followed by ***, **, or * indicate
statistically significant means at the 1%, 5%, and 10% levels, respectively.
Panel A: Portfolio company characteristics in year prior to buyout
Sample in Paper Full Sample
N Mean Median SD N Mean Median SD Mean diff.
Revenue (M£) 246 157.2 69.3 248.7 494 128.4 56.5 206.7 28.80
EBITDA (M£) 246 19.6 8.8 28.9 494 16.5 7.1 25.2 3.15
Total Assets (M£) 246 156.4 66.1 240.5 494 139.1 53.7 220.1 17.39
Number of Employees 238 1797 614 2741 482 1428 456 2397 368.60∗
EBITDA Margin 246 0.159 0.141 0.106 494 0.150 0.136 0.108 0.01
148

Panel B: 1-year pre-buyout growth

Sample in Paper Full Sample
N Mean Median SD N Mean Median SD Mean diff.
Revenue 227 0.177 0.119 0.276 460 0.145 0.099 0.251 0.03
EBITDA 227 0.165 0.182 1.268 460 0.184 0.155 1.436 −0.02
Total Assets 229 0.176 0.071 0.391 464 0.140 0.041 0.400 0.04
Number of Employees 217 0.112 0.062 0.253 448 0.095 0.043 0.252 0.02
EBITDA Margin 227 0.009 0.010 0.061 460 0.006 0.007 0.068 0.00

Panel C: 2-years pre-buyout growth

Sample in Paper Full Sample
N Mean Median SD N Mean Median SD Mean diff.
Revenue 214 0.499 0.285 0.750 433 0.378 0.203 0.608 0.12∗∗
EBITDA 213 0.792 0.414 2.762 432 0.583 0.327 2.347 0.21
Total Assets 218 0.452 0.194 0.869 442 0.360 0.140 0.792 0.09
Number of Employees 203 0.300 0.112 0.653 422 0.253 0.110 0.586 0.05
EBITDA Margin 213 0.020 0.012 0.071 432 0.012 0.010 0.071 0.01
 Table A.2 — Continued

Panel D: Buyout type

Sample in Paper Full Sample
Public-to-Private 0.211 0.134
Private-to-Private 0.273 0.263
Divisional Buyout 0.182 0.231
Secondary Buyout 0.326 0.342
Distressed Buyout 0.008 0.030
N 246 791

Panel E: Exit Route

Sample in Paper Full Sample
Lost Control 0.202 0.162
IPO 0.079 0.062
Trade Sale 0.269 0.290
Sale to Financial Institution 0.306 0.316
Still Owned 0.136 0.159
Other - 0.006
Not Known 0.008 0.005
N 246 791

Panel F: Industry Composition

Sample in Paper Full Sample
Mining 1.3% 1.2%
Construction 2.1% 3.5%
Manufacturing 23.0% 27.7%
Transportation 7.6% 7.0%
Wholesale 4.7% 5.6%
Retail 16.0% 12.7%
Services 45.4% 42.2%

149
 Table A.3: Revenue and Pay-for-Performance Sensitivity - Alternative Specifications

This table presents estimates of annualized revenue growth in relation to the fund’s pay-for-performance
sensitivity ∆j,t . The seven columns indicate specifications with various levels of controls included, as
indicated at the bottom of the table. “Controls” include log revenue, log total assets, and EBITDA
margin in t − 1, as well as book leverage post transaction. All variables are winsorized at a 1% level.
Standard errors, reported in parenthesis under each estimate, are clustered at the acquiring fund level.
Estimates followed by the symbols ***, **, or * are statistically significant at the 1%, 5%, and 10% levels,
respectively.

Panel A - Revenue Growth t − 1 to t + 1 (N = 246)

Fund ∆j,t 0.549 0.764 0.693 0.330 0.277 0.588 0.536
(0.425) (0.360) (0.452) (0.628) (0.713) (0.762) (0.747)

Adjusted R2 0.001 0.114 0.045 0.143 0.141 0.306 0.299

Panel B - Revenue Growth t − 1 to t + 2 (N = 218)

Fund ∆j,t 0.608∗∗ 0.842∗∗ 0.673 0.957 0.953 1.445∗∗ 1.466∗
(0.292) (0.360) (0.600) (0.604) (0.700) (0.724) (0.786)

Adjusted R2 0.003 0.121 0.056 0.159 0.158 0.328 0.310

Panel C - Revenue Growth t − 1 to t + 3 (N = 178)

Fund ∆j,t 0.374 0.742∗ 0.851∗ 1.284∗∗ 1.237∗∗ 1.434∗∗ 1.346∗∗
(0.340) (0.395) (0.485) (0.492) (0.589) (0.576) (0.629)

Adjusted R2 0.001 0.164 0.056 0.207 0.205 0.476 0.472

Panel D - Revenue Growth t − 1 to t + 4 (N = 127)

Fund ∆j,t 0.884∗∗∗ 1.004∗∗ 1.078∗∗ 1.161∗∗∗ 1.101∗∗ 1.156∗∗∗ 1.113∗∗∗
(0.310) (0.446) (0.420) (0.401) (0.430) (0.417) (0.346)

Adjusted R2 0.023 0.188 0.058 0.227 0.219 0.564 0.599

Transaction-Year FE No Yes No Yes Yes Yes Yes
Fund FE No No Yes Yes Yes Yes Yes
Industry FE No No No No Yes Yes Yes
Fund-Year FE No No No No No No Yes
Controls No No No No No Yes Yes

150
 Table A.4: Operational Improvements over Time

This table presents estimates of operational performance measure in relation to the fund’s pay-for-
performance sensitivity ∆j,t . The dependent variable is (log) revenue in columns 1–3, and EBITDA
margin in columns 4–6. A separate coefficient is estimated for the fund’s pay-for-performance sensitivity
∆j,t , for each year around the buyout, using t − 1 as the reference year. Note that ∆j,t itself does not
vary across those years. A separate coefficient is similarly estimated for dummies for each year t + s
with t − 1 being the reference point. The dummy coefficients are omitted from the presentation of the
table. All specifications include fiscal year fixed effects, company fixed effects, dummies for the fund-year
in which the transaction takes place, and company controls interacted with a Post dummy. The Post
dummy takes the value of 1 in the post-buyout period, and 0 otherwise. Company controls included are
log revenue, log total assets, and EBITDA margin measured in t − 1, as well as book leverage measured
post transaction. Columns 2 and 5 add fixed effects for the acquiring fund, and columns 3 and 6 add
industry fixed effects. The pre-buyout period includes up to three years prior to the buyout, while the
post-period includes the year in which the buyout took place, and up to four years after or until an exit
occurs, whichever happens first. All continuous variables are winsorized at a 1% level. Standard errors,
reported in parenthesis under each estimate, are clustered at the acquiring fund level. Estimates followed
by the symbols ***, **, and * are statistically significant at the 1%, 5%, and 10% levels, respectively.

(log) Revenue EBITDA/Revenue

t − 3 × Fund ∆j,t 1.126 0.766 1.212 0.032 0.000 −0.067

(1.228) (1.033) (1.208) (0.268) (0.263) (0.266)

t − 2 × Fund ∆j,t 0.604 0.254 0.659 0.029 0.021 −0.060

(0.689) (0.798) (0.688) (0.224) (0.221) (0.221)

t × Fund ∆j,t 0.722 0.171 0.852 0.176 0.217∗ 0.110

(0.676) (0.944) (0.682) (0.140) (0.126) (0.126)

t + 1 × Fund ∆j,t 1.158∗∗ 0.737 1.334∗∗ 0.164 0.210 0.078

(0.538) (0.869) (0.542) (0.197) (0.187) (0.181)

t + 2 × Fund ∆j,t 1.953∗∗∗ 1.639 2.192∗∗∗ 0.154 0.199 0.069

(0.733) (1.035) (0.715) (0.239) (0.238) (0.233)

t + 3 × Fund ∆j,t 1.990∗ 2.148∗ 2.245∗∗ 0.059 0.032 −0.048

(1.068) (1.187) (1.060) (0.249) (0.243) (0.221)

t + 4 × Fund ∆j,t 4.563∗∗∗ 5.225∗∗∗ 4.921∗∗∗ 0.137 0.102 0.078

(1.218) (1.384) (1.239) (0.324) (0.268) (0.284)

Adjusted R2 0.957 0.943 0.958 0.766 0.714 0.763

N 1702 1702 1702 1701 1701 1701

Company FE Yes Yes Yes Yes Yes Yes

Fiscal Year FE Yes Yes Yes Yes Yes Yes
Fund FE No Yes Yes No Yes Yes
Industry FE No No Yes No No Yes
Fund-Year FE Yes Yes Yes Yes Yes Yes
Controls Yes Yes Yes Yes Yes Yes

151
 Table A.5: Likelihood of Bankruptcy

This table presents estimates of the likelihood that a buyout ends up in insolvency procedures, creditors seize control, or a financial restructuring resulting in a new
owner taking over without compensating previous equity holders. The dependent variable takes the value of 1 if this happens, and zero for all other types of exits.
Deals in which the company is still controlled by the private equity firm as of 2019-09-30 are not included in the estimation. The variable of interest is the fund’s
pay-for-performance sensitivity ∆j,t . In Panel A it is presented alone, while it is interacted with a high and low reputation indicator variable in Panel B. Columns 1-5
presents estimations using OLS, while columns 6-10 presents the estimations of a Logit model. The five columns differ in the set of control variables included. All
models includes time fixed effects. From columns 2 and 7, fund fixed effects are included. Columns 3 and 8 add industry fixed effects. In columns 4 and 9, indicator
variables are included for the year in a fund’s life the deal took place. Columns 5 and 10 add company controls. Company controls include (log) revenue, (log) total
assets, the EBITDA margin, and book leverage. Company controls are measured at the end of the fiscal year prior to the transaction, with the exception of leverage
which is measured after the transaction. For the Logit model, the reported coefficient estimates correspond to marginal effects evaluated at sample averages. All
continuous variables are winsorized at a 1% level. Standard errors, double clustered at the fund level and the year of the deal in the OLS model, are reported in
parenthesis under each estimate. ***, **, and * denotes statistical significance at the 1%, 5%, and 10% levels, respectively.

Panel A
OLS Logit

Fund ∆j,t −0.418 −1.202 −1.017 −0.804 −1.103 −0.430 −1.771 −1.864 −1.735 −2.444
152

(1.751) (2.023) (2.016) (1.982) (2.089) (1.215) (1.558) (1.530) (1.551) (1.567)
Adjusted (Pseudo) R2 0.000 0.031 0.030 0.016 0.000 0.075 0.162 0.219 0.232 0.234
N 213 213 213 213 213 213 213 213 213 213

Panel B
OLS Logit

Fund ∆j,t × 1{High Rep} −1.329 −2.432 −2.656 −2.895 −1.103 −1.272 −2.681 −3.234 −3.273 −2.444
(1.726) (2.015) (2.023) (2.013) (2.089) (1.460) (1.577) (1.587) (1.588) (1.567)
Fund ∆j,t × 1{Low Rep} 2.118 0.693 2.543 2.145 2.081 3.458 1.900 3.035 2.863 1.121
(2.373) (1.546) (2.067) (2.017) (1.979) (2.784) (3.251) (2.894) (2.985) (3.055)
Adjusted (Pseudo) R2 0.001 0.039 0.044 0.029 0.014 0.106 0.228 0.258 0.277 0.284
N 213 213 213 213 213 213 213 213 213 213
Transaction Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Fund FE No Yes Yes Yes Yes No Yes Yes Yes Yes
Industry FE No No Yes Yes Yes No No Yes Yes Yes
Fund-Year FE No No No Yes Yes No No No Yes Yes
Controls No No No No Yes No No No No Yes
Table A.6: Operational Improvements and Pay-for-Performance Sensitivity due to Market Shocks
- Alternative Specification

This table presents estimates of operational improvements in relation to the part of a fund’s pay-for-
performance sensitivity that is due to a recent market shock, ∆m j,t , defined in section 4.1.3. In this table,
the NAV of the PE fund is assumed to not have incorporated the market shock yet. The four columns
represent improvements from the year prior to the buyout, t − 1, to one, two, three, and four years
after the buyout. The dependent variable is annualized revenue growth in Panel A, annualized employee
growth in Panel B, annualized growth in total assets in Panel C, and the cumulative improvement in the
EBITDA/Revenue margin in Panel D. Each model specification includes time, fund, and industry fixed
effects; dummies for the fund year in which the transaction took place; and log revenue, log total assets,
and EBITDA margin in t − 1, as well as book leverage post transaction. All variables are winsorized at a
1% level. Standard errors, reported in parenthesis under each estimate, are clustered at the acquiring
fund level. Estimates followed by the symbols ***, **, or * are statistically significant at the 1%, 5%, and
10% levels, respectively.

Panel A - Revenue Growth (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆m
j,t 2.666∗ 2.389∗ 0.352 1.374
(1.430) (1.218) (0.900) (1.136)
Adjusted R2 0.304 0.307 0.462 0.577
N 246 218 178 127

Panel B - Employee Growth (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆m
j,t 2.873 1.155 −0.970 −1.427
(1.838) (0.841) (0.849) (1.564)
Adjusted R2 0.221 0.284 0.396 0.281
N 238 207 169 120

Panel C - Growth Total Assets (annualized)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆m
j,t −4.662 −1.234 −1.385 −2.098
(3.451) (2.560) (1.708) (2.633)
Adjusted R2 0.140 0.135 0.153 0.119
N 246 218 178 127

Panel D - Cumulative change in Return on Sales (EBITDA/Revenue)

t − 1 to t + 1 t − 1 to t + 2 t − 1 to t + 3 t − 1 to t + 4
Fund ∆m
j,t 1.094 0.349 0.916 −0.966
(1.104) (1.891) (2.027) (2.217)
Adjusted R2 0.032 0.122 0.161 0.321
N 246 218 178 127
Time FE Yes Yes Yes Yes
Fund FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
Fund-Year FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes

153
B Additional Data Description

This appendix complements the main data description in Section 3, by providing

details the sample construction process and some additional details on the accounting
data used in the paper.

B.1 Sample Construction

I follow Strömberg (2008) and use Capital IQ (CIQ) to identify buyout transactions.
The transaction screen was undertaken 2019-01-02. The following criteria were used in a
“Capital IQ Transaction Screen” to generate the initial set of transactions:

• “Transaction Primary Features” set to: Merger/Acquisition – Acquisition of Equity

Stake or Private Placement – PIPE or Private Placement – Growth capital/Private
Equity;

• “Merger/Acquisition Features” set to: “Going Private Transaction” or “Leveraged

Buy Out (LBO)” or “Management Buyout” or “Platform” or “Privatization of Gov-
ernment Entity” or “Management Participated” or “Secondary LBO” or “Acquisition
of Majority Stake”;

• “Secondary features” set to: “Leveraged Buy Out (LBO)”, or “Management Buyout”,
or “Secondary LBO”, or “Going Private Transaction”;

• “Investment Firm Type (Buyers/Investors)” is listed as “PE/VC”;

• Transaction is listed as “closed” or “effective”.

The initial screen picked up transactions that were initial toe-hold investments, a
purchase of remaining minor stakes, or the purchase of a minority stake. The focus here
are transactions in which a private equity buyout fund gains control of the company.
Therefore, transactions that are classified as either “Minority Stake Investment” or
“Majority Shareholder Purchasing Remaining Shares”, or where the investor was classified
as a “VC”, are dropped from the sample. This leaves an initial set of 11,549 transactions.
For all unique companies in these transactions, the following items are extracted:
company name; country of incorporation; filing currency; and the country of primary
office location. All UK companies are selected based on being incorporated in the UK
or, if that item is not available, either have GBP as its filing currency or has its primary
office in the UK. Due to the limited number of years with financials available in the
Bureau van Dijk (BvD) databases, we restrict the sample to transactions completed
between 1999 and 2018. This leaves the initial set of 1964 buyouts of UK companies
that forms the first row of Table 1.

154
 Of the identified buyouts of UK companies 850 are randomly selected to be processed.
Each transaction requires a substantial amount of manual work, necessitating working
with a random part of the sample. Figure A.3 outlines the composition of buyouts over
time of those randomly selected versus those not processed. Overall, the distribution of
buyouts over time is quite similar, though Capital IQ coverage is low for the later
years in the sample.
57 transactions are dropped from the sample for the following reasons:

• 19 companies are dropped as no UK company could be found either in Companies

House or Fame under the name of the acquired company in Capital IQ today or,
when it differed, the name of the acquired company in the transaction synopsis
provided by Capital IQ. Both Companies House and Fame contains previously
registered names of companies, so when no match can be found it likely reflects
either a mistake in the naming in Capital IQ or that some non-UK companies are
identified during the process of selecting UK-registered companies.

• 15 transactions are dropped as no PE firm participated. PE firm participation is

verified using annual reports and ownership filings.

• 10 transactions are dropped because the PE firm did not acquire a controlling stake.
This information is gathered from post-buyout ownership filings.

• 7 transactions are duplicates, where two separate company IDs in Capital IQ have
been assigned to the same transaction, with two separate transaction IDs.

• 5 transactions are dropped as the transaction ID identified in the initial screen were
no longer present in Capital IQ at the time a match was attempted to be established
with Fame.

• 3 transactions are dropped because the acquired companies were too small and were
exempted from submitting filings throughout the period of interest.

The remaining 791 buyouts are fully processed with the type of buyout classified and the
type of exit documented with the date and the type of exit. For the 16.2% of transactions
that are classified as still being owned by private equity this is as of 2019-09-30, when
a potential sale was last searched for. A lack of news about a sale is not sufficient to
classify a company as still in private equity hands. Ownership filings are used to verify
the ownership status and searches for insolvency filings are undertaken in Companies
House. There are 4 cases where the outcome could not be determined. In one of these
it is suspected that creditors seized control as the ownership passed to a syndicate of

155
financial institutions after a period of large reported losses. However, to be consistent,
only events which are explicitly mentioned in the annual reports are included as it is
possible that they paid the previous owner to acquire the company. The remaining three
unknown cases concerns individuals taking over control of the company with no mention
as to what happened to the previous private equity owners.

B.2 Additional Details on the Accounting Data

The main source of the accounting data used in the paper is Fame, a BvD database.
The filings available in Companies House are scanned documents, and BvD provides a
digitalized version of them in a panel format for individual company IDs. The ID assigned
to a company in BvD is the one the company is registered with in Companies House.
Fame provides financials for up to 20 years worth of annual reports. For the
earlier transactions in the sample this implies that the pre-buyout period is not always
covered. For these deals I manually complement the financials from annual reports
downloaded from Companies House. There are two additional scenarios when I manually
complement the Fame data.
First, occasionally there are fiscal years inexplicably missing from Fame, despite
the annual reports being available in Companies House. I add these years manually
to avoid gaps in the data where there are none.
Second, the acquisition event typically gives rise to a large amount of goodwill in
the acquiring parent company, and subsequent add-on acquisitions may further increase
the goodwill post in the balance sheet. The goodwill is typically amortised over a
period of years and may be subject to impairment charges. Fame provides an EBITDA
number that accounts for depreciation and amortisation. However, it is not uncommon
that one or both of the two numbers are missing from individual years. Additionally,
impairment charges are only rarely included by Fame, though they are both common
and usually of considerable magnitude. I have manually gone through Fame to detect
missing depreciation/amortisation entries, and scanned annual reports for mentions of
impairment charges to verify whether they are picked up by Fame.35 When Fame miss
impairment charges I complement the EBITDA calculations.36 This process is applied to
the pre-buyout years as well as the period under private equity ownership to avoid any
bias arising from different treatment of the data during the period of PE ownership.

35
Impairment and write-downs can be made for fixed assets and intangible assets such as goodwill.
Fame tends to not account for either. The charge always shows up somewhere in Fame as a cost to make
net income calculations correct. While there is some variation of where they place it, the most common
appears to be to the cost in general “Administration Expenses”, and sometimes as an exceptional item.
36
The increase in interest payments and the typical increase in amortisation is the reason why EBITDA
measures are commonly used to evaluate the performance of PE portfolio companies.

156
 Not all companies are required to provide the same level of detail in their annual
reports. In particular, companies classified as “small” are only required to submit abridged
accounts, which contains limited information and excludes a profit and loss statement.
These filings therefore cannot be used in the data analysis. The classification requirements
for small companies has changed over time, and any company-year in which a company
qualified as a small company is dropped from the sample.37

37
To be classified as a small company, it needs to be smaller in at least two out of three qualifying
measures: turnover, total assets, and number of employees. The limit on number of employees has been
stable at 50 over time. The following requirements and time periods are relevant for the sample in this
paper, with numbers for turnover (total assets): prior to 2004-01-10, £2.8m (£1.4m); prior to 2008-04-06,
£5.6m (£2.8m); prior to 2015-04-06, £6.5m (£3.26m); since 2015-04-06, £10.2m (5.1m). Classifications
are from Section 247 of the Companies Act 1985 and Section 382 of the Companies Act 2006, both links
accessed 2019-08-22. I am grateful to the policy team at Companies House for providing these links.

157
C Details on Monte Carlo Procedure

This appendix provides additional details on the simulation process used for estimating
expected carried interest described in section 2.2.
The estimation of expected carried interest relies on Monte Carlo simulations of the
value paths of individual investments in a private equity portfolio. The simulation consists
of several step, that largely match those outlined in Figure 1 in Metrick and Yasuda
(2010), with one added step on the calibration part. The following steps describe the
baseline process when the model is calibrated:

1. Model inputs are defined. These include contract terms defining fee payments, as
well as number of investments, timing of investments, and other model parameters.
All model parameters are defined in Table C.1, which also provides sources justifying
the values used.

2. Initial values are set for the value added by GPs, and the initial parameter σ. Both
of these are set as the midpoint of a range containing wide enough to span the value
which the calibration converges on. The range is decreased in an iterative process
according to point 5.

3. 100,000 simulations of a private equity fund portfolio are generated. For each
investment in the portfolio, the following is drawn:

(a) A holding period is drawn from the empirical distribution of holding periods.
If the fund age at the exit date implied by the holding period exceeds the
fund’s 12th year, the investment is instead exited at the end of that year.

(b) Portfolio company paths are generated using the process defined in 2.2.1.

(c) The entire generated path is checked to ensure that the value of equity > 0
at all times. If not, company passes to creditors and the private equity fund
writes off the investment.

(d) Cash flows of individual investments are aggregated to the fund level, and
distributed to the GP and LPs according to the contract terms defined in step
1.

(e) The resulting cash flows are discounted to the fund inception date using a
risk-neutral rate r.

4. The expected values of fees and cash flows are computed as average values over the
100,000 trials.

158
 5. The input values of either σ or value added by the GP is adjusted. As the two
items interact, they are adjusted alternatively in an iterative process according to
the following rules:

(a) σ is calibrated to ensure that the likelihood of default equals the empirically
observed one, as documented in Table 3. The low value of σ is set to the
current midpoint in the range if model output of default < empirically observed
default level. If the model output of default > empirically observed default
level, the high value of the σ range is set to the current midpoint.
(b) Value added by the GP is calibrated to ensure that the present value of LPs
capital call equals the present value of cash distributions to LPs. If the net
present value of cash flows to LPs < 0, the low value of the value added range
is set to the current midpoint, while if net present value > 0, the high value
added range is set to the current midpoint.

Steps 2 to 5 are repeated until σ is accurate to within three decimal points (17.4%),
and value added is determined with two decimal points per $100 of committed capital
($9.23). The process of updating guesses requires increasingly precise estimates to ensure
the update converges. As the range halves every iteration, the number of simulations
need to increase by a factor four to decrease the standard deviation of the estimated
value by half. I ensure that the range stays within six standard deviations at all times,
which the 100,000 runs guarantees for most of the iterations. However, the number of
simulations needs to be updated towards the more precise estimates, and 25.6 million
simulations are run in the final iteration.
Changes to σ and value added feed into the calibrated value of each other. Therefore,
to reduce the computational time required, the first few rounds of the iterations only
estimate σ to an accuracy of two decimal points (whole percent) and value added to
one decimal point. Note that each estimation of σ (and value added) itself requires a
number of iterations to reach the desired accuracy. When this process converges, the
number of simulations increase to reach the desired accuracy is achieved. The final
output of the model is summarized in Table C.2.
The main focus of the paper is estimating the change in expected GP pay from
adding one dollar of additional value to one investment, as outlined in section 2.2.2.
When empirically estimating the process for each reported NAV in Preqin, a seed is
utilized to ensure that the same portfolio company paths are generated with and without
an additional dollar of value added, making that the only source of variation. This
significantly reduces the standard error of the estimation, ensuring that 10,000 simulated
portfolio paths yields an accurate estimate of the pay-for-performance sensitivity measure.

159
 The paper does not present any analysis of variation in contract terms. Metrick and
Yasuda (2010) document how GPs’ expected pay from various fee sources varies when
the contract terms vary. While this is important for overall fee levels, the impact is
much smaller on the pay-for-performance sensitivity measure that is the focus of this
study. The main reason is that the equilibrium approach alters the expected value added
by a GP to ensure that LPs break even in expectation. Thus, altered contract terms
change the calibrated values, but have little impact on the estimated pay-for-performance
sensitivity measure. Importantly, the qualitative patterns summarized in Figure 1 are
virtually unchanged for alternative contract terms.

160
 Table C.1: Model Parameters

Parameter Value Source

ρ 0.2 Used in Metrick and Yasuda (2010), num-

bers from estimates in Campbell et al.
(2001) for small-cap stocks.

σ 17.4% Calibrated so fraction of simulated portfolio

paths ending in default equals the empir-
ically observed rate of 19.4% of all exited
deals documented in Table 2.

r 5% By assumption.

Carried interest level 20% Used by 83.3% of buyout funds in Metrick

and Yasuda (2010), and 96.0% out of the
860 funds in Preqin with information about
level of carried interest.

Carry basis Committed Used by 83.3% of buyout funds in Metrick

capital and Yasuda (2010). LPs receive their
committed capital, and hurdle rate, back
before carried interest is paid out to the
GP.

Catch-up rate 100% Reported by Metrick and Yasuda (2010)

to be used by a majority of funds with a
hurdle rate.

Exit date Drawn from the empirical distribution of

holding periods of exited deals in the sam-
ple, described in Table 2 and Figure A.6.
An investment is terminated early if the
value of the portfolio company at any point
equals debt. Any remaining investments at
the end of a fund’s 12th year are liquidated.

Hurdle rate 8% Used by a majority of buyout funds in

Metrick and Yasuda (2010), and by 94.5%
out of the 474 buyout funds in Preqin with
information about hurdle rates.

Investment pace 26%, 23%, Investment pace per year as reported in

25%, 18%, Metrick and Yasuda (2010). For estima-
and 8% tions where funds are partly invested, the
investment pace is scaled as required to
ensure a fully invested fund at the end of
the investment period. Arcot et al. (2015)
and Degeorge et al. (2016) provide evidence
that GPs accelerate investments towards
the end of the investment period if there is
unspent committed capital.

Leverage (D/E) 2:1 Leverage of buyouts in western Europe

documented by Axelson et al. (2013). Here
expressed as D/E.

Management fee 2% Most common management fee level among

buyout funds in Preqin, used by 75.3% (out
of 799).
(Continues)

161
 Table C.1 — Continued

Parameter Value Source

Management fee basis Committed capital during investment pe-

riod, net invested capital thereafter. Used
by 84% of buyout funds in Metrick and
Yasuda (2010).

Monitoring fees (% of Total enter- 0.72% Sum of “regular monitoring” (0.44%) and
prise value) “accelerated monitoring” fees (0.28%) as a
fraction of TEV in Phalippou et al. (2018).

Investments per fund 16 Average number of deals per fund in Braun

et al. (2017).

Transaction fees (% of Total enter- 0.88% Sum of “transaction fees” (0.81%) and “add-
prise value) on transaction fees” (0.07%) as a fraction
of TEV in Phalippou et al. (2018).

LP rebate rate of monitoring fees 70% Average rebate on monitoring fees in Phalip-
pou et al. (2018).

LP rebate rate of transaction fees 72% Average rebate on transaction fees in

Phalippou et al. (2018).

Table C.2: Model Output

This table summarizes the model outputs for simulations run at fund inception, after calibrating σ
and value added by GPs as described in Section 2.2. The values presented are for a fund with $100 of
committed capital. It compares the model output to those for buyout funds in Metrick and Yasuda (2010)
using comparable fund terms. The calibrated value added that make LPs break even is not available
for buyout funds in Metrick and Yasuda (2010). They do provide values for VC funds which is $18.71.
Average total revenues to buyout funds in their paper is $5.36 lower, suggesting that value added is likely
around $13.35. In their paper, σ is not calibrated but assumed to be 0.60. As this paper adds leverage at
the portfolio company level, the proper comparison to their assumed value is the volatility of the equity
position σE = 0.521.

Output in paper Metrick and Yasuda (2010)

Carry per $100 $4.06 $5.08
Management fees per $100 $10.48 $11.09
Transaction fees per $100 $0.56 $1.68
Monitoring fees per $100 $0.19 $0.87
Total GP revenue per $100 $15.26 $18.72
Calibrated σ 17.36% n/a
Calibrated value added per $100 $9.23 n/a

162
D Variable Definitions

Table D.1: Variable Definitions

Variable name Description

Buyout Types

Distressed Buyout A transaction where the acquired company was in financial

distress prior to the buyout. It is defined as being acquired
from one of the following situations: acquired from bankruptcy
procedures (called “administration” in the UK); acquired in a
refinancing or restructuring deal where creditors are not fully
compensated; and one case where the acquiring private equity
firm was paid by the previous owner to take on the company and
its debt.

Divisional Buyout A transaction where the acquired company was a subsidiary or

division of a larger corporation prior to the buyout.

Private-to-Private Buyout A transaction where the acquired company was privately trading
prior to the buyout, with individuals controlling the company,
and is not classified as a distressed buyout.

Public-to-Private Buyout A transaction where the acquired company was publicly listed
at the time of the buyout.

Secondary Buyout A transaction where the acquired company was a portfolio

company of another private equity firm prior to the transaction.
This category includes cases where the portfolio company is
passed between private equity ownership more than once.

Exit Routes

IPO An exit is classified as an IPO if the private equity owner exits

via an initial public offering. Usually, a private equity owner
do not sell all shares at the IPO, instead exiting over a period
of time following the IPO. The exit date is defined as the date
at which majority ownership is lost, whether that is the IPO
date or a later date. The exit route is classified as an IPO if the
stake is reduced below a majority stake while the company is
still publicly trading.

Lost Control A private equity owner is defined as having lost control of the
company if any of the following happens: creditors seize control
of the company; the company is placed in bankruptcy procedures
(called “administration” in the UK); or control passes to a new
owner in a financial restructuring of the company, where the
equity holders receive no compensation. These are events that
are generally not disclosed publicly, and the eventual outcome is
found through disclosures in Annual Reports and, in the case of
bankruptcy procedures, through reports submitted submitted to
Companies House by the court-appointed administrators.

(Continues)

163
 Table D.1 — Continued

Variable name Description

Other (Exit Route) There are five cases in the sample where the exit route cannot be
classified into either of the other five exit route categories. These
five cases involve the private equity firm maintaining ownership
of the portfolio company until the company has been sold in
pieces or liquidated by other means, and afterwards seize trading.

Trade Sale A sale is classified as a trade sale when the company is sold to a
non-investment firm or to individuals.

Sale to Financial Institution A sale is classified as a sale to a financial institution when the
company is sold to a investment firm. The vast majority of these
cases are sales to private equity funds but also includes sales
to pension funds, sovereign wealth funds and other investment
firms.

Still Owned A company is classified as still being owned if there are no

recorded transactions as of 2019-09-30, and the most recent
annual report or ownership list still indicates the private equity
firm as the ultimate majority owner.

Dependent Variables

Employee Growth The percentage increase in number of employees from t − 1 to

t + x. Represented as an annualized growth rate over the x + 1
years.

Growth Tangible Assets The percentage increase in tangible assets from t − 1 to t + x.

Represented as an annualized growth rate over the x + 1 years.
This measure is used as an alternative to growth in total assets
to exclude the impact of goodwill on the growth rate.

Growth Total Assets The percentage increase in total assets from t − 1 to t + x.

Represented as an annualized growth rate over the x + 1 years.
Total assets includes goodwill arising from acquisitions, including
that generated by the buyout itself.

Revenue Growth The percentage increase in revenue from t−1 to t+x. Represented
as an annualized growth rate over the x + 1 years.

Return on Sales (ROS) Defined as Earnings before Interest, Taxes, Depreciation and
Amortisation (EBITDA) divided by Revenue. EBITDA is
adjusted for impairment charges, write-offs and revaluations
to tangible and intangible assets. In the analysis, the cumulative
change in ROS from t − 1 until t + x is used.

(Continues)

164
 Table D.1 — Continued

Variable name Description

Explanatory Variables

Close to Fundraising An indicator variable taking the value of 1 if a transaction took

place close to a fundraising event, and zero otherwise. This is
defined as a fundraising event taking place within eight quarters
of the closing date of the transaction. A fundraising event is
defined as the first quarter in which a cash flow is observed in
the follow-on fund or, when cash flow data is not available, the
second quarter of the follow-on fund’s vintage year.

Far from Fundraising An indicator variable taking the value of 1 if a transaction did
not take place close to a fundraising event, and 0 otherwise.

Fund ∆j,t The expected increase in a GP’s payoff from carried interest
from adding $1 to the next investment made, estimated using
the framework described in section 2.2.

Fund ∆m
j,t The part of ∆j,t attributable to market movements in the most
recent quarter. See section 4.1.3 for details on construction.

Fund IRR The internal rate of return of the fund. In this paper the IRR is
measured prior to a fund being fully liquidated. To calculate it,
the current NAV valuation is accounted for as a final distribution.

Fund Age Age in quarters, with the fund’s first quarter is the one in which
the first capital call takes place.

High Reputation An indicator variable taking on the value of 1 in buyouts where

the acquiring GP is classified as a high reputation GP. A GP is
classified as high reputation if it is not classified as low reputation.

Low Reputation An indicator variable taking on the value of 1 in buyouts where

the acquiring GP is classified as a low reputation GP. A GP is
classified as low reputation if it has: cumulatively raised than
$1 billion prior to the current fund; have raised less than three
funds in the past; and have no past top quartile funds that are
five years or older at the time the new fund is raised.

165
References

Aboody, David and Ron Kasznik, “CEO stock option awards and the timing of
corporate voluntary disclosures,” Journal of Accounting and Economics, 2000, 29 (1),
73 – 100.
Abowd, John M., “Does Performance-Based Managerial Compensation Affect Corporate
Performance?,” ILR Review, 1990, 43 (3), 52–S–73–S.
Acharya, Viral V., Oliver F. Gottschalg, Moritz Hahn, and Conor Kehoe,
“Corporate Governance and Value Creation: Evidence from Private Equity,” The Review
of Financial Studies, 2013, 26 (2), 368–402.
Achleitner, Ann-Kristin, Reiner Braun, Bastian Hinterramskogler, and
Florian Tappeiner, “Structure and Determinants of Financial Covenants in Leveraged
Buyouts*,” Review of Finance, 2011, 16 (3), 647–684.
Agarwal, Vikas, Naveen D. Daniel, and Narayan Y. Naik, “Role of Managerial
Incentives and Discretion in Hedge Fund Performance,” The Journal of Finance, 2009,
64 (5), 2221–2256.
Agrawal, Anup and Charles R. Knoeber, “Firm Performance and Mechanisms
to Control Agency Problems between Managers and Shareholders,” The Journal of
Financial and Quantitative Analysis, 1996, 31 (3), 377–397.
Ang, Andrew, Bingxu Chen, William N. Goetzmann, and Ludovic Phalippou,
“Estimating Private Equity Returns from Limited Partner Cash Flows,” The Journal
of Finance, 2018, 73 (4), 1751–1783.
Arcot, Sridhar, Zsuzsanna Fluck, José-Miguel Gaspar, and Ulrich Hege, “Fund
managers under pressure: Rationale and determinants of secondary buyouts,” Journal
of Financial Economics, 2015, 115 (1), 102 – 135.
Axelson, Ulf, Per Strömberg, and Michael S. Weisbach, “Why Are Buyouts
Levered?,” The Journal of Finance, 2009, 64 (4), 1549–1582.
, Tim Jenkinson, Per Strömberg, and Michael S Weisbach, “Borrow Cheap,
Buy High? The Determinants of Leverage and Pricing in Buyouts,” Journal of Finance,
2013, 68 (6), 2223–2267.
Barber, Brad M and Ayako Yasuda, “Interim Fund Performance and Fundraising in
Private Equity,” Journal of Financial Economics, 2017, 124 (1), 172–194.
Barrot, Jean-Noël, “Investor Horizon and the Life Cycle of Innovative Firms: Evidence
from Venture Capital,” Management Science, 2017, 63 (9), 3021–3043.
Bebchuk, Lucian Arye, Jesse M. Fried, and David I. Walker, “Managerial Power
and Rent Extraction in the Design of Executive Compensation,” The University of
Chicago Law Review, 2002, 69 (3), 751–846.
Berk, Jonathan B. and Richard C. Green, “Mutual Fund Flows and Performance
in Rational Markets,” Journal of Political Economy, 2004, 112 (6), 1269–1295.

166
Bernstein, Shai and Albert Sheen, “The Operational Consequences of Private Equity
Buyouts: Evidence from the Restaurant Industry,” The Review of Financial Studies,
2016, 29 (9), 2387–2418.
, Josh Lerner, and Filippo Mezzanotti, “Private Equity and Financial Fragility
during the Crisis,” The Review of Financial Studies, 2019, 32 (4), 1309–1373.
Boucly, Quentin, David Sraer, and David Thesmar, “Growth LBOs,” Journal of
Financial Economics, 2011, 102 (2), 432 – 453.
Braun, Reiner, Tim Jenkinson, and Ingo Stoff, “How persistent is private equity
performance? Evidence from deal-level data,” Journal of Financial Economics, 2017,
123 (2), 273 – 291.
Brav, Omer, “Access to Capital, Capital Structure, and the Funding of the Firm,” The
Journal of Finance, 2009, 64 (1), 263–308.
Brenner, Menachem, Rangarajan K Sundaram, and David Yermack, “Altering
the terms of executive stock options,” Journal of Financial Economics, 2000, 57 (1),
103 – 128.
Brown, Gregory W., Oleg R. Gredil, and Steven N. Kaplan, “Do private equity
funds manipulate reported returns?,” Journal of Financial Economics, 2019, 132 (2),
267 – 297.
, Robert S. Harris, Tim Jenkinson, Steven N. Kaplan, and David Robinson,
“What Do Different Commercial Data Sets Tell Us About Private Equity Performance?,”
2015. Working paper.
Campbell, John Y., Martin Lettau, Burton G. Malkiel, and Yexiao Xu, “Have
Individual Stocks Become More Volatile? An Empirical Exploration of Idiosyncratic
Risk,” The Journal of Finance, 2001, 56 (1), 1–43.
Chung, Ji-Woong, “Leveraged Buyouts of Private Companies,” 2011. Working paper.
, Berk A. Sensoy, Léa Stern, and Michael S. Weisbach, “Pay for Performance
from Future Fund Flows: The Case of Private Equity,” The Review of Financial Studies,
2012, 25 (11), 3259–3304.
Cohn, Jonathan B., Edith S. Hotchkiss, and Erin M. Towery, “The Motives for
Private Equity Buyouts of Private Firms: Evidence from U.S. Corporate Tax Returns,”
2020. Working Paper.
, Lillian F. Mills, and Erin M. Towery, “The evolution of capital structure and
operating performance after leveraged buyouts: Evidence from U.S. corporate tax
returns,” Journal of Financial Economics, 2014, 111 (2), 469 – 494.
Datta, Sudip, Mai Iskandar-Datta, and Kartik Raman, “Executive Compensation
and Corporate Acquisition Decisions,” The Journal of Finance, 2001, 56 (6), 2299–2336.
Davis, Steven J., John Haltiwanger, Kyle Handley, Ben Lipsius, Josh Lerner,
and Javier Miranda, “The Economic Effects of Private Equity Buyouts,” 2019.
NBER Working Paper 26371.
, , , Ron Jarmin, Josh Lerner, and Javier Miranda, “Private Equity, Jobs,
and Productivity,” American Economic Review, 2014, 104 (12), 3956–90.
Degeorge, Francois, Jens Martin, and Ludovic Phalippou, “On secondary buyouts,”
Journal of Financial Economics, 2016, 120 (1), 124 – 145.
Demsetz, Harold and Belén Villalonga, “Ownership structure and corporate
performance,” Journal of Corporate Finance, 2001, 7 (3), 209 – 233.
Driessen, Joost, Tse-Chun Lin, and Ludovic Phalippou, “A New Method to
Estimate Risk and Return of Nontraded Assets from Cash Flows: The Case of Private
Equity Funds,” Journal of Financial and Quantitative Analysis, 2012, 47 (3), 511–535.

167
Edmans, Alex, Xavier Gabaix, and Dirk Jenter, “Executive Compensation: A
Survey of Theory and Evidence,” in Benjamin Hermalin and Michael Weisbach, eds.,
The Handbook of the Economics of Corporate Governance, Amsterdam: Elsevier, 2017,
pp. 383–539.
Fama, Eugene F., “Agency Problems and the Theory of the Firm,” Journal of Political
Economy, 1980, 88 (2), 288–307.
and Michael C. Jensen, “Agency Problems and Residual Claims,” The Journal of
Law and Economics, 1983, 26 (2), 327–349.
and , “Separation of Ownership and Control,” The Journal of Law and Economics,
1983, 26 (2), 301–325.
Franzoni, Francesco, Eric Nowak, and Ludovic Phalippou, “Private Equity
Performance and Liquidity Risk,” Journal of Finance, 2012, 67 (6), 2341–2373.
Frydman, Carola and Dirk Jenter, “CEO Compensation,” Annual Review of Financial
Economics, 2010, 2 (1), 75–102.
Gibbons, Robert and Kevin J. Murphy, “Optimal Incentive Contracts in the
Presence of Career Concerns: Theory and Evidence,” Journal of Political Economy,
1992, 100 (3), 468–505.
Grossman, Sanford J. and Oliver D. Hart, “An Analysis of the Principal-Agent
Problem,” Econometrica, 1983, 51 (1), 7–45.
Guo, Shourun, Edith S. Hotchkiss, and Weihong Song, “Do Buyouts (Still) Create
Value?,” The Journal of Finance, 2011, 66 (2), 479–517.
Hall, Brian J. and Jeffrey B. Liebman, “Are CEOs Really Paid Like Bureaucrats?*,”
The Quarterly Journal of Economics, 1998, 113 (3), 653–691.
Harris, Robert S., Tim Jenkinson, and Steven N. Kaplan, “Private equity
performance: What do we know?,” Journal of Finance, 2014, 69 (5), 1851–1882.
Holmström, Bengt, “Moral Hazard and Observability,” The Bell Journal of Economics,
1979, 10 (1), 74–91.
Humphery-Jenner, Mark, Ling Lei Lisic, Vikram Nanda, and Sabatino Dino
Silveri, “Executive overconfidence and compensation structure,” Journal of Financial
Economics, 2016, 119 (3), 533 – 558.
Hüther, Niklas, David T. Robinson, Sönke Sievers, and Thomas Hartmann-
Wendels, “Paying for Performance in Private Equity: Evidence from Venture Capital
Partnerships,” Management Science, 2020, 66 (4), 1756–1782.
Jegadeesh, Narasimhan, Roman Kräussl, and Joshua M Pollet, “Risk and
Expected Returns of Private Equity Investments: Evidence Based on Market Prices,”
Review of Financial Studies, 2015, 28, 3269–3302.
Jensen, Michael, “Eclipse of the Public Corporation,” Harvard Business Review, 1989,
67 (5), 61–74.
Jensen, Michael C. and Kevin J. Murphy, “Performance Pay and Top-Management
Incentives,” Journal of Political Economy, 1990, 98 (2), 225–264.
and William H. Meckling, “Theory of the firm: Managerial behavior, agency costs
and ownership structure,” Journal of Financial Economics, 1976, 3 (4), 305 – 360.
Kaplan, Steven, “The effects of management buyouts on operating performance and
value,” Journal of Financial Economics, 1989, 24 (2), 217 – 254.
Kaplan, Steven N. and Per Strömberg, “Leveraged Buyouts and Private Equity,”
Journal of Economic Perspectives, 2009, 23 (1), 121–146.
Larocque, Stephannie A., Sophie Shive, and Jennifer Sustersic Stevens, “Private
Equity Returns, Cash Flow Timing, and Investor Choices,” 2019. Working paper.

168
Lichtenberg, Frank R. and Donald Siegel, “The effects of leveraged buyouts on
productivity and related aspects of firm behavior,” Journal of Financial Economics,
1990, 27 (1), 165 – 194.
Ljungqvist, Alexander, Matthew Richardson, and Daniel Wolfenzon, “The
investment behavior of buyout funds: Theory and evidence,” Financial Management,
2020, 49 (1), 3–32.
McConnell, John J. and Henri Servaes, “Additional evidence on equity ownership
and corporate value,” Journal of Financial Economics, 1990, 27 (2), 595 – 612.
Mehran, Hamid, “Executive compensation structure, ownership, and firm performance,”
Journal of Financial Economics, 1995, 38 (2), 163 – 184.
Metrick, Andrew and Ayako Yasuda, “The Economics of Private Equity Funds,”
Review of Financial Studies, 2010, 23 (6), 2303–2341.
Michaely, Roni and Michael R. Roberts, “Corporate Dividend Policies: Lessons
from Private Firms,” The Review of Financial Studies, 2012, 25 (3), 711–746.
Nanda, Ramana and Matthew Rhodes-Kropf, “Investment cycles and startup
innovation,” Journal of Financial Economics, 2013, 110 (2), 403 – 418.
Otto, Clemens A., “CEO optimism and incentive compensation,” Journal of Financial
Economics, 2014, 114 (2), 366 – 404.
Phalippou, Ludovic, Christian Rauch, and Marc Umber, “Private equity portfolio
company fees,” Journal of Financial Economics, 2018, 129 (3), 559 – 585.
Renneboog, Luc, Tomas Simons, and Mike Wright, “Why do public firms go
private in the UK? The impact of private equity investors, incentive realignment and
undervaluation,” Journal of Corporate Finance, 2007, 13 (4), 591 – 628. Private Equity,
Leveraged Buyouts and Corporate Governance.
Robinson, David T. and Berk A. Sensoy, “Do Private Equity Fund Managers Earn
Their Fees? Compensation, Ownership, and Cash Flow Performance,” The Review of
Financial Studies, 2013, 26 (11), 2760–2797.
Shue, Kelly and Richard R. Townsend, “How Do Quasi-Random Option Grants
Affect CEO Risk-Taking?,” The Journal of Finance, 2017, 72 (6), 2551–2588.
Smith, Abbie J., “Corporate ownership structure and performance: The case of
management buyouts,” Journal of Financial Economics, 1990, 27 (1), 143 – 164.
Strömberg, Per, “The new demography of private equity,” in Lerner J Gurung A, ed.,
Globalization of Alternative Investments Working Papers Volume 1: Global Economic
Impact of Private Equity, (World Economic Forum USA, New York), 2008, pp. 3–26.
Yermack, David, “Do corporations award CEO stock options effectively?,” Journal of
Financial Economics, 1995, 39 (2), 237 – 269.
, “Good Timing: CEO Stock Option Awards and Company News Announcements,”
The Journal of Finance, 1997, 52 (2), 449–476.

169
5 | Managerial Ownership and Op-
erational Improvements in Buy-
outs∗

This paper examines the impact of managerial ownership on operational performance for
companies backed by private equity. High post-buyout CEO equity ownership stakes are
associated with improved profitability. This result could arise for several reasons, such
as the impact of incentives or matching of talented CEOs to companies. To distinguish
between these explanations, the paper utilizes that the CEO is often replaced in buyouts.
The ownership-performance result is only present when a CEO is retained, suggesting that
the matching explanation is unlikely to be a main driver. In addition, the result is stronger
when absolute changes in ownership are used rather than ownership levels. Taken together,
these results suggest that changes in incentives from managerial ownership stakes induced
by buyouts likely contribute to improved operational performance.

1 Introduction

Private equity has become a prominent ownership form, with $4.1 trillion in assets
under management globally (Preqin 2020) and sponsoring more than 4000 deals during
2018 in the US alone (Pitchbook 2019). Given its large role in the economy, it is
important to understand if and how private equity impacts the performance of portfolio
companies. Although the existing literature suggests that private equity generates
improvements to portfolio companies’ operations, we do not know which actions taken
generate these improvements.1
Starting with Jensen and Murphy (1990), there has been a long debate over whether the
equity ownership of CEOs provides sufficient incentives to generate value for shareholders.
A salient feature of private equity acquisitions is that managerial equity ownership
increases (e.g. Kaplan and Stein, 1993; Cronqvist and Fahlenbrach, 2013). Increasing the
∗
I gratefully acknowledge the Private Equity Institute at the Saïd Business School for providing financial
support. I thank Dan Zaitsev and Kresimir Krasovic for dedicated research assistance, and wish to thank
Vimal Balasubramaniam, Pedro Bordalo, Thomas Hellmann, Tim Jenkinson, Alan Morrison, Ludovic
Phalippou, Enrique Schroth, Martin Schmalz, Joel Shapiro, Mungo Wilson and seminar participants at
the Saïd Business School for many helpful comments and constructive suggestions.
1
Examples of studies documenting operational improvements are Acharya et al. (2013); Bernstein and
Sheen (2016); Boucly et al. (2011); Davis et al. (2014, 2019); Guo et al. (2011); and Kaplan (1989).

170
equity ownership stake increases the pay-for-performance sensitivity of the management
team, providing stronger incentives to generate value. Gompers et al. (2016) document
that private equity firms believe that improving incentives is the second most important
source of value creation.2 This paper examines how the change in managerial ownership
induced by a buyout is related portfolio company performance.
To examine this question I assemble a unique dataset containing operational perfor-
mance and managerial ownership of companies acquired in a buyout. Since companies
under private equity ownership are privately trading, this data is generally not available
unless the company subsequently goes public or has publicly traded debt, limiting the
potential sample and raising concerns that the sample is not representative (Cohn et al.,
2014; Davis et al., 2019). I overcome this challenge by focusing on portfolio companies
registered in the United Kingdom. All UK companies, public and private, are required to
submit filings to the national registrar office, Companies House, which makes the filings
publicly available. Required filings include annual reports as well as annual breakdowns
of the precise ownership of every shareholder in the company. I compile ownership filings
by hand to measure managerial ownership before and after the buyout.3 Using a random
sample of 322 UK portfolio companies, this paper provides new insights into the impact
of managerial ownership on the performance of companies acquired in a buyout.
The main finding of the paper is that the post-buyout CEO ownership share is
positively correlated with improvements in profitability (EBITDA/Revenue).4 The results
suggest that companies with a 10 percentage points higher CEO equity ownership stake
(one standard deviation) improve their profitability margin by 3.1%—3.6% from the
unconditional average profitability. The relationship is more pronounced for changes in
CEO ownership stakes, where a 10 percentage points increase in the ownership translates
to a 4.3%—5.0% increase in profitability.
Clearly, the interpretation of any observed relationship between managerial ownership
and operational performance depends on whether the CEO is retained or replaced. Cornelli
and Karakas (2015) document that the CEO of a target company is often replaced by
the acquiring private equity firm, which is the case in 39% of the buyouts in my sample.
Interestingly, the positive relationship between managerial ownership and profitability

2
The most important source is “improving revenue or improve demand factors” (Table 23). As
improved incentives may improve revenue, these are not mutually exclusive.
3
This is not the first paper to utilize the UK setting to study private companies. Brav (2009) and
Michaely and Roberts (2012) used the UK setting go gain insights from privately trading companies, while
Bernstein et al. (2019) study how private equity ownership impacted portfolio companies through the
financial crisis.
4
Results are similar for gross profitability and can be provided upon request to the author. Buyout
transactions usually create large goodwill posts, increasing total assets as well as significant goodwill
amortisation posts. Because of this, the focus is on EBITDA as a profit measure, and return on sales as a
profitability measure rather than return on assets.

171
improvements found in this study is driven entirely by the subsample of buyouts where
the CEO is retained. By contrast, the coefficient is statistically insignificant and close
to zero when the CEO is replaced. The economic magnitude is up to 60% higher when
the CEO is retained compared to the full sample.
The positive relationship between operational performance and increased managerial
equity ownership may reflect matching of talented CEOs to companies as in Gabaix and
Landier (2008). Higher ability CEOs should be able to bargain for a bigger share, as
well as accept compensation packages that are more dependent on performance. The
results could also correspond to the outcome of efficient bargaining from managers
with inside information (Blanchflower et al., 1996). Additionally, they may capture
an effect of incentives provided through the ownership stake. The arrival of private
equity almost invariably induces a shock to the managerial ownership share. To the
extent managers have limited say over the post-buyout share, the change in ownership
level represents a shock to CEO incentives.
The different sample splits help differentiating between these potential explanations. If
the positive relationship between managerial ownership and performance reflects attracting
talented CEOs, we would expect the results to be stronger when the CEO is replaced as
the pool of potential CEOs is larger. Instead, the positive relationship is only observed
when the CEO is retained, which casts doubt on this explanation. When a CEO is
retained, a change in the ownership stake implies a change in incentives, while CEO ability
is held constant. Furthermore, the incentive story suggests that changes in ownership are
more relevant than the level of ownership, which is what I find. In contrast, the bargaining
story suggests that the relevant metric is the ownership share of the management team.
To further sharpen the interpretation of the results I examine the determinants of
CEO equity ownership post-buyout. These results are largely consistent with a desire to
provide incentives. There is a negative relationship between ownership stake and initial
profitability. Taken together with the main results of the paper, this is consistent with
private equity firms setting contracts to maximize value. Providing management with
higher equity stakes comes at a cost as it limits the upside of the acquirer. Therefore,
we expect higher management stakes in companies where profitability improvements are
particularly valuable: those with low initial profitability. Furthermore, the pre-buyout
ownership stake is positively related to the post-buyout stake, but only when the CEO is
retained. This is consistent with an attempt to increase incentives from pre-buyout levels.
Taken together, the results of the paper support the view that improving managerial
incentives plays an important part in value creation by private equity firms.
A natural question to ask is how profitability improvements come about. Improved
profitability may reflect reduced costs or increased revenues with the existing cost base.

172
The results of the paper indicate that the profitability improvement related to increases
in managerial ownership stems from reducing costs rather than increasing productivity.
This could indicate that private equity firms provide CEOs with larger stakes when cost
cutting is desired. If cost cutting is associated with a high private cost to the CEO,
this is consistent with an optimal contracting view.
Any study on private equity portfolio companies has to deal with the fact that
companies are not randomly assigned to be buyout targets. Thus, simply comparing
buyout targets with non-targeted companies is tricky. Therefore the analysis relies on
cross-sectional variation in managerial ownership among buyout targets, rather than
comparing buyout targets and other companies. Following Boucly et al. (2011), the
analysis includes matched control companies to account for industry-time fixed effects,
but the variation in managerial ownership is only among the targets of a buyout.
However, even among buyout targets there is a concern that the optimal compensation
package varies with unobserved company characteristics. But this explanation cannot
explain why it is only in the retained CEO subsample that we see positive relationships
between managerial ownership and operational performance, and between pre- and
post-buyout CEO ownership stakes. Presumably, a high pre-buyout ownership stake
would indicate a higher CEO ownership stake in the optimal contract whether the
CEO is retained or not.
Finally, distinguishing between the efficient bargaining and the incentive story is
problematic as they often give similar predictions. For example, imagine a manager of
a division in a conglomerate with knowledge of how to improve that division. She may
approach a private equity firm with the intention to increase her ownership stake prior
to implementing the necessary improvements. The larger stake provides incentives to
take action, but presumably the manager’s inside information allows her to bargain for a
larger stake. Absent exogenous shocks to the ownership stake provided to CEOs, this
paper cannot claim causality, merely offering robust supportive evidence of the idea that
higher CEO incentives lead to improved operational performance.
This paper speaks directly to the literature on operational improvements provided
by private equity. Early studies by Kaplan (1989), Lichtenberg and Siegel (1990), and
Smith (1990), showed that management buyouts are associated with improved operations,
and that this generates value for the private equity sponsors. More recently, Guo et al.
(2011), and Acharya et al. (2013) document the importance of operational improvements
for generating value in buyouts. Boucly et al. (2011), using a sample of French portfolio
companies, suggest one important channel through which private equity can help their
portfolio companies: relaxing credit constraints to facilitate growth. Davis et al. (2014) and
Davis et al. (2019) provide evidence that private equity facilitates reallocation of capital to

173
more productive establishments, generating an overall increase in total factor productivity.
This paper adds to this literature by documenting how a specific change induced by a
buyout — changes in managerial ownership — relates to operational improvements.
As discussed in Jeng and Lerner (2016), data limitations is always an issue in private
equity research, and certainly when studying privately held portfolio companies with
limited disclosure requirements. Cohn et al. (2014) directly show how this may impact the
interpretation of results, as companies with publicly available financials perform better
than those without. Davis et al. (2019) highlight that not all buyouts are the same, with
public-to-private buyouts differing markedly from other types of buyouts. This speaks
to one of the strengths of this study, as the UK setting allows the inclusion of all types
of buyouts, and all companies are required to submit annual reports.
The UK setting also allows for collecting a large-scale sample of managerial ownership
both pre- and post-buyout, and linking it to operational performance. This is a significant
improvement over the existing large-scale study focusing on managerial ownership of
private equity targets by Leslie and Oyer (2009).5 They utilize a set of public-to-private
buyouts to obtain pre-buyout ownership data, while using a different set of companies
going public through an IPO to measure managerial ownership levels after a period of
private equity ownership. Since these sets are different, the authors are unable to measure
changes induced at the time of the buyout. Additionally, the authors lack company
financials during the period of private equity ownership.
Beginning with Kaplan and Strömberg (2003), private equity and venture capital
have been used to explore optimal governance, examining the choices taken by value-
maximizing principals. Cornelli and Karakas (2015) study the role of boards and find
that CEO turnover decreases and becomes less contingent on performance when private
equity sponsors have higher involvement in the company. Edgerton (2012) shows that
private equity sponsors curb excessive usage of corporate jets. Cronqvist and Fahlenbrach
(2013) provide a detailed overview how CEO contracts are changed for a set of 20 private
equity deals. This paper contributes to this stream of papers by linking how changes
in managerial ownership is related to operational performance.
The question of how managerial ownership impacts the value of a company is an
important question in the agency literature. Agency theory suggests that if managers
can exert costly effort to increase performance, then we expect better performance

5
Cronqvist and Fahlenbrach (2013) also examines CEOs of companies acquired in public-to-private
buyouts. They obtain detailed information about the contract offered to CEOs in 20 large deals, offering
insights into optimal contracting. While this paper lacks that level of detail, it covers more transactions
and combines the ownership data with operational performance data.

174
when the pay of the manager is more tightly linked to performance.6 Existing studies
on managerial ownership in publicly held companies have to overcome multiple issues:
compensation packages are the outcome of a matching and bargaining process; managers
may influence the process in which compensation is set (Bebchuk et al., 2002) or time
the delivery of news to the awarding of stock options (Yermack, 1997; Aboody and
Kasznik, 2000); compensation packages can be renegotiated ex post (Brenner et al.,
2000); and companies design compensation packages to exploit managerial characteristics
such as optimism and overconfidence (Otto, 2014; Humphery-Jenner et al., 2016). To
the extent that a buyout shocks the ownership level of managers, this study provides
evidence of the importance of managerial ownership.
The next section presents the institutional setting. Section 3 describes the dataset
and provides summary statistics. Section 4 presents the empirical approach and the
results of the paper, while section 5 concludes.

2 Institutional Setting
A private equity fund is a partnership between a group of investors (referred to as Limited
Partners, or LPs), and a private equity firm (referred to as the General Partner, or GP).
“Private equity” encompasses many different strategies that varies in the the stage of a
company’s life at which investments are made, and the level of control the GP gets. This
paper focuses on buyouts in which the GP acquires a controlling equity stake in the target
company. Venture capital and growth investments are therefore not part of the study.
When a fund is raised, LPs commit capital to the fund but the capital is not provided
up front. The GP is provided significant freedom over what to invest in and the timing of
those investments, though the fund has a maximum lifetime and an investment period
during which to invest the committed capital. Once an investment is identified, the GP
“calls” capital from LPs. When investments are exited, the resulting cash is distributed
to LPs and usually cannot be recycled.
The terms of the partnership are defined in the limited partnership agreement which
stipulates the length of the investment period, the maximum allowed lifetime of the
fund, and the compensation structure to the GP. As documented in Metrick and Yasuda
(2010), the most common compensation arrangement for buyout funds includes a 2%
management fee and 20% carried interest, allowing the GP to share in the profits if
6
Jensen and Meckling (1976), Fama (1980), Fama and Jensen (1983b,a) are examples of this theoretical
agency literature. Holmström (1979) and Grossman and Hart (1983) provide models of optimal contracts
when effort is unobservable and impacts contractable output with noise. There is some empirical evidence
that company value is positively impacted from increased managerial ownership and when compensation
is more sensitive to company performance (e.g. Abowd, 1990; McConnell and Servaes, 1990; Mehran,
1995). However, the evidence is mixed. Agrawal and Knoeber (1996) and Demsetz and Villalonga (2001)
do not find evidence of a relationship between managerial ownership and company performance.

175
they perform well. In addition to the direct pay for performance, Chung et al. (2012)
demonstrates that the indirect pay from improving performance — the increased ability
of a GP to successfully raise new funds that provide future fees — is of at least the
same magnitude to the direct compensation. The structure of a GP’s compensation
package therefore provides strong incentives to focus on taking actions that are value
maximizing for their portfolio companies.
As GPs obtain a controlling stake in companies, they can implement changes deemed
value increasing. The clear focus on value maximization, coupled with control rights,
implies that private equity firms approximate strong principals in agency models. For
this reason, the actions taken by GPs may be informative of governance actions taken to
maximize firm value. An early example is Kaplan and Strömberg (2003). They provide
empirical evidence that contracts set up by venture capitalists largely conform to what
optimal contracting theory suggests they should be. In a similar vein, Cronqvist and
Fahlenbrach (2013) provide a detailed examination of changes in CEO contracts for a
set of 20 large companies taken private in buyouts in order to learn about what strong
principals change in contracts with agents. The most significant change they find is that
a significant portion of equity grants performance-vests based on prespecified measures,
and that CEOs forfeit unvested equity in the event they are fired.7
Proponents of private equity, like Jensen (1989), argue that private equity firms add
value to companies by focusing managers on generating cash flow, and by providing high-
powered incentives that align the interests of managers with those of investors. Kaplan
and Stein (1993) and Cronqvist and Fahlenbrach (2013) document that the percentage
ownership of CEOs increases following a buyout, supporting this view. The increase
takes place despite a reduced dollar amount invested by retained CEOs. A simultaneous
“cash-out” and increased ownership is possible due to the increase in leverage following
the buyout. The reduced equity share of the enterprise value implies that any dollar
amount invested represents a higher equity share post-buyout.
To further increase the pay-for-performance sensitivity of the management team,
the acquiring private equity firm can split the equity share using multiple share classes.
For example, the fraction held by the management team of the residual claim can be

7
Private equity is also used to study the impact of boards by Cornelli and Karakas (2015). Their main
results are that CEOs are less likely to be fired, and termination is less sensitive to performance, when the
private equity owner has seats on the board. They interpret their findings as private equity owners having
a longer horizon and being more patient, and utilizing information learned while monitoring. Edgerton
(2012) examines actions taken by private equity owners to evaluate if (some) CEOs in listed companies
enjoy excessive perks. He focuses on the case of corporate jets, which are costly but may be justified if
increasing the productivity of CEOs. While hard to disentangle these views in any given company, he
shows that jets are kept after being acquired by private equity in companies with modest jet ownership,
suggesting a valuable role. However, companies with the biggest fleets experience a clear reduction in jets,
suggesting that at least a subset of companies have excessive perks.

176
increased through a use of preference shares and ordinary shares. We now turn to the
data used in the paper that allows us to precisely measure the fraction of the residual
ownership held by the management team.

3 Data

This section describes the dataset utilized in the paper. The dataset contains information
on the operational performance of UK-registered portfolio companies acquired in a buyout,
complemented with hand-collected information of the managerial ownership pre- and
post-buyout. The accounting dataset partially overlaps with that used in the second
paper of the thesis where it is discussed in more detail. To avoid unnecessary repetition
for the reader, some details on the data collection collection process, considered not
essential for the current paper, are relegated to Appendix B.
As private companies lack market valuations, one needs to resort to accounting data
to assess the performance of portfolio companies. Accessing accounting information on
privately held companies is a challenge in private equity research. Consequently, many
studies on private equity portfolio companies rely on samples where the researchers have
been given access to deals from a particular investor (e.g. Acharya et al., 2013), or on
companies with publicly available financials (e.g. Guo et al., 2011). Accounting data
may be publicly available if companies have publicly traded debt or eventually goes
public through an IPO, and in that process submits past accounts. In both of these
scenarios, one does not obtain a comprehensive sample and there is a concern that the
resulting sample is not fully representative.8
Precise ownership information tends to be even less accessible. Existing studies rely
on subsets of companies that have either been or subsequently become publicly listed (e.g.
Kaplan and Stein, 1993; Leslie and Oyer, 2009). The study of changes in contractual terms
for CEOs by Cronqvist and Fahlenbrach (2013) relies on a sample of 20 public-to-private
buyouts, for which they obtain detailed information on CEO contracts.
In this paper I focus on UK companies to overcome challenges of data availability. All
companies in the UK, public and private, are required to submit filings to “Companies
House”, the United Kingdom official national registrar office. Companies House makes
all accounts publicly available in the form of scanned PDFs. The information available
from Companies House is gathered in electronic form by Bureau van Dijk (BvD) in their

8
Cohn et al. (2014) use tax data on all public-to-private deals to show that there is a marked difference
in the performance between companies with publicly available financials and those without, with the latter
showing worse performance. Davis et al. (2019) find a stark difference between public-to-private buyouts
and buyouts of already private targets, with the former (latter) experiencing sharp declines (increases) in
number of employees.

177
FAME database.9 Companies House also requires annual filings containing a list of all
shareholders in the company, with their names and precise ownership documented.10

3.1 Sample Selection

This paper follows Strömberg (2008) and uses Capital IQ to identify UK companies that
are targets of a buyout in the years 1999-2018. Appendix B.1 describes the screening
process in detail, but essentially it captures buyouts, as defined by Capital IQ, in which a
majority ownership stake is sold. This means that venture capital investments, growth
investments and other minority stake positions are not included in this study.
Table 1 outlines the process for constructing the sample. The initial screen in Capital
IQ identifies 1964 companies targeted in a buyout. Of these, 850 are randomly selected
to be processed. There is a substantial amount of manual work required in processing
the data, which explains why only part of the entire dataset is utilized in the paper.
57 transactions are dropped for reasons outlined in the table. To enable an analysis of
changes in operational performance of target companies, a company is required to have
financial data available in at least the year prior to the buyout, t−1, and the year following
the buyout t + 1. This restriction reduces the sample size with a further 230 transactions.
69 companies operating in the financial (SICs 600-699) or public utility (SIC 489-493)
sectors are dropped to be consistent with earlier studies (e.g. Bernstein et al., 2019).
This leaves a sample of 494 buyouts for which financials are available in the pre- and
post-buyout period. A further 134 transactions are dropped because ownership data is
inaccessible in either the pre- or post-buyout period, or ownership filings are missing.
The main reason for inaccessible ownership information is that the ultimate holding
company, of which the management team potentially holds a stake, is registered in the
Cayman Islands, Jersey, or some other jurisdiction outside of the UK. This practice has
become increasingly common over time, and consequently the sample is tilted towards the
earlier part of the sample period, relative to the full set of buyouts. For the pre-buyout
period, ownership data is sometimes inaccessible for divisional or secondary buyouts,
if the ultimate parent company is not a UK company.
9
Information in FAME is available for a given registered company for up to the 20 most recent annual
reports. Note that this restriction is better than the 10 year requirement present in Orbis and Amadeus,
two other commonly used BvD databases which provides coverage over a wider region than just the UK.
The UK setting, and BvD databases, have been utilized to gain insights from privately held companies by
Brav (2009) and Michaely and Roberts (2012).
10
I am not the first to utilize the UK setting to gain insights on PE-sponsored companies. The setting
is used by Bernstein et al. (2019) to study how companies backed by private equity fared during the
financial crisis, and by Chung (2011) to examine the performance of private-to-private buyouts. To the
best of my knowledge, this is the first study which utilizes UK ownership filings to measure the managerial
ownership share. Chung (2011) uses these filings to document the change in number of shareholders and
the percentage held by the largest shareholder following a buyout, but he does not specifically measure
managerial ownership stakes, nor does he link these changes to subsequent improvements in operational
performance.

178
 Finally, the paper utilizes a set of matched control companies to account for industry
trends by including industry-year fixed effects, as suggested in Gormley and Matsa (2013).
A match cannot be found for all companies, and consequently 38 additional transactions
are dropped from the sample. This leaves 322 transactions in the final sample.

3.1.1 Accounting Data

The main source for financials is Fame, a Bureau van Dijk (BvD) database. BvD provides
a digitalized database of information contained in the annual reports that all UK-registered
companies are required to submit to Companies House. While Fame provides a coherent
time-series for each registered entity, it is not sufficient to track the target company to
get a comparable series of pre- and post-buyout financials.
A typical buyout is completed through the creation of one or several acquisition
vehicles which have no economic activity on their own. Upon acquiring the target
company, these newly created vehicles become parent companies to the target company
(and any subsidiaries). In the UK, wholly owned subsidiaries in a corporate group are
not required to submit accounts, as long as at least one parent provides consolidated
accounts. In virtually all buyouts, one of the newly created entities take on the role of
providing group consolidated accounts. Sometimes the consolidating entity shifts multiple
times during the period under private equity ownership, even when the ultimate owner
remains the same. While the target company may continue to submit accounts, they
are usually unconsolidated. Unconsolidated accounts often provide an unrepresentative
view of operating performance as they do not net out intra-group transactions, tend
to understate costs, occasionally misrepresent turnover, and frequently lack relevant
financial information altogether.
For every transaction in the sample I document the relevant reporting entity for every
company-year. This information is gathered directly from annual reports of the target
portfolio company, downloaded from Companies House. The annual report contains
information of the ultimate consolidating entity if the company is a wholly owned
subsidiary in a group. In the case of divisional buyouts, the unconsolidated accounts
of the division are used for pre-buyout financials, as the larger consolidated group
accounts include divisions that are not acquired in the buyout. This process provides
a time-series of the reporting entity for every transaction in the sample. The financials
downloaded from Fame are merged using the correct reporting entity, providing a coherent
time-series linked to each transaction.
Table 2 provides a summary of the target companies in the sample, measured in the
last fiscal year prior to the buyout, denoted t − 1. Panel A provides these descriptives
in levels. The average (median) company has revenues of £119.6 (£54.9) million, an

179
EBITDA margin of 16.0% (14.2%), assets in place worth £117.3 (£48.5) million, and
have 1432 (448) employees. Panel B provides growth rates in the period t − 2 to t − 1.
The average (median) company increases revenue by 17.8% (12.5%), assets expand by
17.0% (7.0%), the number of employees increases by 11.6% (6.4%), and the EBITDA
margin increases by 1.0 (0.9) percentage points.
Panel C summarizes the buyout types in the sample. The most common is secondary
buyouts representing 38.8% of buyouts in the sample, followed by private-to-private
buyouts (25.8%), divisional buyouts (20.8%), public-to-private buyouts (13.7%), and
distressed buyouts (0.9%). Secondary buyouts have grown in importance over time,
with almost all other categories declining. Secondary buyouts represent a majority of
the deals in the sample after the financial crisis.
Panel D summarizes the exit routes of the buyouts in the transaction. The time-
period splits correspond to the year the buyout was completed in, not the year of the
exit. Consistent with the increased prevalence of secondary buyouts, sales to financial
institutions — of which the by far most common is another private equity firm — represents
the most common exit route with 37.0% of all exits. The second most common exit route,
representing 26.4% of all exits, is a trade sale, where another company or individuals acquire
the company. 17.1% of all buyouts with an exit end with the private equity owners losing
control without being compensated. This category includes formal insolvency procedures
(“administration” in the UK), creditors seizing control, and financial restructuring where a
new owner comes in, only partially compensating creditors while previous equity owners are
wiped out. 12.7% of all deals have yet to see an exit as of 2019-09-30, with most of the recent
deals still being owned. 6.5% of the deals are exited through a listing on the public market.
Table A.1 provides a comparison of the company characteristics of companies in
the sample with the larger set for which financials are available. Such a comparison is
instructive to evaluate whether the subsample with ownership data available systematically
differs from a typical buyout.11 , Panels A, B, and C compare company characteristics
and growth rates leading up to the buyout. While not statistically different due to
large standard deviation, the companies in the sample is somewhat smaller and more
profitable that the typical buyout. Pre-buyout revenues are £8.8 million, or 6.9% lower,
for the subset of companies in the sample compared to the larger sample. Revenue
has however increased at a more rapid pace, perhaps because of the smaller size. The

11
While we would ideally like to compare the sample with the whole population of buyout deals,
financials are only available for the randomly selected subset that has been processed. In the second paper
of the thesis, a comparison is provided of the distribution of deals over time for the randomly selected
sample compared to all 1964 identified buyout transactions in Capital IQ. The distribution over time is
found to be very similar, suggesting a random selection from the population.

180
companies in the sample have higher EBITDA/revenue profitability margins at 16.0%,
compared to 15.0% for the overall sample.12
Panels D, E and F of Table A.1 provide comparisons of the types of buyout, exit
route, and industry composition. In these dimensions, the subset used in the paper
is quite representative of the overall sample. Secondary buyouts are more common as
both entry and exit route. The service industry is overrepresented by 4.5 percentage
points, mainly at the expense of the manufacturing industry which is underrepresented
by 3.2 percentage points.
The composition of type of buyouts and exits are largely consistent with Strömberg
(2008), though with a higher proportion of secondary buyouts at entry, and selling to
financial institutions at exit. Both of these have grown in importance over time and partly
reflect the more recent sample examined here. The 17% figure of deals which end up in
bankruptcy procedure or where the company is seized by creditors is high compared to the
8% of UK buyouts ending in bankruptcy reported by Strömberg (2008). However, 11%
of the exits in his study are not known, with 22% of deals in 2006-2007 being unknown.
Presumably, deals without a succesful sale are more likely to have an unknown outcome,
such as creditors seizing control and continuing running the business. In addition, the
timing of his study implies that the impact of the financial crisis is not part of his exit
figures. For most of the companies in this sample that ended up in administration or
lost control to creditors following the financial crisis, this took placed during the period
2009-2012, following attempts to restructure the debt.13 The figures in this sample are
comparable to the 18% of deals ending up in bankruptcy in Degeorge et al. (2016).

3.1.2 Managerial Ownership Data

The most novel piece of data utilized in this paper is ownership data of CEOs in
companies acquired in buyouts. Ownership information is available in the UK as all
12
The companies in this paper are smaller than those studied in the second paper in the thesis.
Companies there are larger than the typical buyout, a consequence of the acquiring fund being larger
than the typical fund due to the requirement that the acquiring fund has cash flow data in Preqin. The
typical portfolio company in this paper is still relatively large compared to other studies, except for those
focusing exclusively on public-to-private buyouts. In studies of UK buyouts, the average target company
has sales of £158.6 million in Renneboog et al. (2007) (public-to-private only), £10.4 million in Chung
et al. (2012) (private-to-private only), and £98 million in Bernstein et al. (2019). Comparing to studies
in other countries: the average French company has €32.6 million in sales in Boucly et al. (2011); the
average company acquired in a private-to-private buyout in the US has sales of $117.4 million in Cohn et
al. (2020), while the average company acquired in public-to-private buyouts has sales of $604.2 million in
Cohn et al. (2014).
13
In the event a financial restructuring took place that concluded with the private equity firm retaining
a majority stake, it is not part of the 17% figure. While the restructuring process is not formally studied,
there are multiple cases where the private equity owner injects equity as part of a restructuring deal with
creditors to maintain its majority stake. These examples are consistent with the evidence in Bernstein et
al. (2019) that private equity owners can provide support to portfolio companies to get through financially
challenging periods.

181
companies annually submit a filing called “Annual Return” (more recently, “Confirmation
Statements”). In the annual return, a company provides details about the directors of
the company, the amount of all classes of outstanding shares, and the precise ownership
of every individual shareholder in the company. Interestingly, this information is more
detailed for private companies as any ownership, no matter how small, is included with
the name of the shareholder. By contrast, a similar list is not publicly available for
publicly listed companies, though the annual report discloses large shareholders as well
as information on the ownership of directors in the company.
The filings are available in a scanned format, and are sometimes handwritten, especially
for the earlier part of the sample. I manually compile the ownership status by name,
to measure the fraction of ownership belonging to the CEO, non-CEO directors, other
individuals, and the private equity firm. This information is collected at two points:
the last annual return filed prior to the buyout; and the first filing after the buyout.
This allows measuring not only the level of ownership but also changes in ownership
shares. In buyouts, different classes of shares are often constructed and distributed with
various voting rights, though the different versions of ordinary shares tend to rank pari
passu for cash flow rights. It is also common to utilize various classes with different cash
flow rights, for example preference shares. In the UK in particular, there is a common
practice to use shareholder loans. Shareholder loans are not part of the annual return,
instead showing up as liabilities on the balance sheet.
Two complications arise when collecting ownership data pre-buyout. First, when a
company is publicly listed prior to the buyout, only the ownership of the CEO and other
directors is available in, and collected from, annual reports. Outstanding options are
included in the ownership share of directors by estimating the delta of these options,
assuming that the Black-Scholes option pricing formula holds. Other long-term incentive
share plans in place are counted in full, as if the requirements would be met in the future.
In practice, including or excluding these items makes almost no difference to the results
of the paper as they correspond to a tiny fraction of the ownership shares of public
companies. Second, in divisional buyouts, any managerial ownership stakes tend to be in
shares of the ultimate parent company. The pre-ownership measures for these directors
therefore represents ownership in the ultimate parent company.14
Table 3 summarizes the common stake held by CEOs in the sample. Panel A
displays the post-buyout ownership stake held by the CEO, and Panel B summarizes
14
It is not uncommon that at least one of the directors in the targeted division is also a director at the
ultimate parent. Commonly, the acquired division represents a small fraction of the parent, and parent
directors do not join the acquired division in the buyout. It is likely that these directors’ main focus is not
on the (relatively) small division but rather on the larger company. Therefore, any ownership by directors
having a directorship in the ultimate parent is excluded from the management team ownership share for
these divisions, unless a director followed the acquired company which happens in two single cases.

182
CEO ownership before the buyout takes place. The first row in each panel displays
the overall sample numbers, while the remaining five rows provides a break down by
type of buyout. The average (median) CEO ownership stake is 11.5% (8.5%) after the
buyout, compared to 17.4% (6.9%) before the buyout.
The average reduction in CEO ownership following the buyout contrasts earlier studies
by Kaplan and Stein (1993), Leslie and Oyer (2009), and Cronqvist and Fahlenbrach
(2013). This difference can be reconciled with the type of buyouts included in these
studies, as they study only public-to-private buyouts while this paper includes all types
of buyout. Indeed, average CEO ownership increases for public-to-private buyouts
here as well, from 7.1% to 10.1%. The main category driving the high pre-buyout
ownership is private-to-private buyouts, where the average (median) CEO ownership is
38.3% (38.8%). The high ownership share reflects that a substantial fraction of private-
to-private buyouts are companies owned by a founder, or family-owned companies in
which the CEO holds a large stake.
The average post-buyout ownership is highest in private-to-private buyouts (15.7%),
followed by secondary buyouts (10.5%), distressed buyouts (10.3%), public-to-private
buyouts (10.1%), and divisional buyouts (8.5%). The pattern is similar in the pre-buyout
ownership numbers, with secondary buyouts switching place with distressed buyouts.
Table A.2 provides additional descriptives on the ownership of all directors. Post-
buyout, the CEO almost always holds the largest share of any other personnel, and in the
average company the CEO share represent 40% of the entire management team’s share.

4 Results

This section develops the empirical approach taken in and presents the findings of the analy-
sis.
The goal of the paper is to examine how CEO ownership stakes relate to improvements
in operational performance of companies acquired in a buyout. Ideally, this would entail
comparing otherwise identical companies, randomly selected to be acquired in a buyout
and where the managerial ownership stakes are randomly decided. That is not the reality
of the data however. Private equity firms do not select their targets randomly, and
the ownership stake provided to a CEO is likely related to observed and unobserved
characteristics of the company, the CEO, and possibly the company’s future prospects.
With the ideal comparison not available, this paper utilizes a difference-in-difference
approach to get closer to the ideal setting.
Following Boucly et al. (2011) and Bernstein et al. (2019), a control group is constructed
of companies that are comparable to target companies on observable characteristics. This

183
enables us to control for industry-wide shocks by including industry-year fixed effects,
as suggested in Gormley and Matsa (2013).
The focus of the paper is not on the overall difference in operational improvements
of companies acquired in buyouts compared to their matched peers. Instead, we ask
whether any differential improvements are a function of the ownership stake awarded to
a CEO post-buyout. Again, this stake is clearly not randomly selected. Edmans et al.
(2017) note that although models are clear on predictions that higher incentives should
increase company value gross of CEO compensation, that does not necessarily translate
to a relationship between company value and CEO ownership in the data. If incentives
are set optimally, cross-sectional variation in incentives relate to different fundamental
inputs in the optimization process. To the extent that there is deviation from optimality,
we would however expect improvements when incentives are set closer to the optimum.
Changes made by private equity firms have been interpreted as those taken by strong
principals to improve governance in for example Cronqvist and Fahlenbrach (2013) and
Edgerton (2012). Therefore, the paper will focus on changes in the ownership stakes
rather than the ownership level. Changes are particularly informative when a CEO is
retained, as that corresponds to changed incentives for a given individual.15
Identification ultimately relies on the extent to which the approach taken plausibly
deals with the endogeneity issues present. The paper evaluates multiple potential
interpretation of the results, and utilizes the private equity setting to sharpen these
predictions whenever possible.

4.1 Cross-Sectional Variation in Ownership

To set the scene, we begin by documenting how cross-sectional variation in post-buyout

CEO ownership stakes relates to operational improvements in companies acquired in a
buyout. Several company outcome variables will be examined. To facilitate the discussion
I will refer to revenue. As the focus is on operational improvements, it is natural to
compare a pre- and a post-buyout period, allowing for a differential growth in revenue as
a function of the ownership stake given to the CEO following the buyout. We include
a pre-period of up to three years pre-buyout, and a post-buyout period of up to four
full fiscal years post-buyout. The following model is estimated:

Yi,t,j =αi + αt + αj + βCEO Ownershipi ∗ Posti,t (1)

+ λPosti,t + θXi ∗ Posti,t + εi,j,t ,

15
It is important to point out here that awarding the CEO with a higher stake dilutes the PE firm’s
stake. Hence, increasing incentives through higher CEO ownership stakes may improve operational
performance without improving the value of the PE firm’s stake, and the evidence should be interpreted
as gross-of-pay effects.

184
where Yi,j,t is (the log of) revenue for company i, operating in industry j, measured
in year t; αi , αj , and αt are company, industry, and year fixed effects, respectively;
CEO Ownershipi measures the fraction of ordinary shares held by the CEO post-buyout;
Posti,t is an indicator variable which takes the value of 1 if the end of a fiscal-year falls in
the period after the buyout is completed, and zero if it is before; Xi contains controls
at the company level, measured at the year prior to the buyout.
The inclusion of company fixed effects together with a post period implies that the
estimates capture changes within a company. The interaction of the post-period dummy
with the CEO Ownership variable allows for a differential impact in companies as a
function of the ownership stake held by the CEO. Time fixed effects and industry fixed
effects are included in some specifications to take out the impact of economy-wide shocks
and differences in levels between industries.
Company characteristics are included as controls for two reasons. First, it is conceivable
that companies of certain characteristics are more likely to improve in specific dimensions.
For example, Boucly et al. (2011) and Cohn et al. (2020) show that private equity
ownership support company growth for small companies by relaxing financial constraints.
It is also conceivable that companies with an initial low profitability have more room
for profitability improvements. Second, characteristics of the company are likely to be
related to the CEO ownership stake. For example, CEO risk aversion suggests lower
stakes in larger companies to limit a CEO’s wealth at risk. Similarly, companies with low
profitability may require restructuring. If that is a particularly a risky undertaking it could
suggests lower stakes held due to risk aversion, or larger stakes because more compensation
is required to entice the CEO to take the job in the first place. Indeed, section 4.4 presents
evidence consistent with the idea that the post-buyout CEO ownership stake is related to
company characteristics. The general idea is that we need to control for these variables
to avoid an omitted variable bias in the coefficient on the CEO’s ownership stake.
The set of company controls included are: log revenue, EBITDA margin, and deal
leverage. These are measured in the year prior to the buyout, with the exception of
deal leverage which is the book value of leverage, measured at the first annual report
after the buyout is completed.
Private equity firms typically focus on EBITDA as the measure of company profitability,
and indeed even write compensation-based contracts on the measure (Cronqvist and
Fahlenbrach, 2013). If a private equity firm wishes to focus on improving EBITDA, there
are essentially two ways to achieve this: growing the company with a constant profit
margin, or improving profitability. Revenue growth and EBITDA margin, defined as
EBITDA/Revenue, are therefore the main dependent variables in the study.

185
 Table 4 presents the estimates of (1), with standard errors clustered at the company
level. Columns 1-3 present results with log revenue as the dependent variable, while
columns 4-6 display results using the EBITDA margin as the dependent variable. Both
dependent variables are estimated with three models, varying in the set of fixed effects
included. The first model (columns 1 and 4) includes company fixed effects and year
fixed effects. The second (columns 2 and 5) adds industry fixed effects. The third
removes company controls.
The results show that companies with higher post-buyout CEO ownership are
associated with improved profitability, while the coefficient on revenue suggests an
economically small and statistically insignificant positive association to CEO ownership.
In terms of economic magnitude, the results suggest that companies with a 10 percentage
points higher CEO ownership experience improvements in the profitability margin of
40–50 basis points. This corresponds to a 2.5%–3.1% increase from the unconditional
average pre-buyout profitability margin. In comparison, the statistically insignificant
coefficients for revenue suggest an increase in revenue of 1.24%–1.77%.
The post dummy indicates that, controlling for the ownership stake given to the CEO,
companies acquired in buyouts experience large growth in revenue following the acquisition,
while profitability, if anything, is reduced. However, note that this growth is not compared
to any matched peers and may have happened even in the absence of a buyout.

4.1.1 Controlling for industry effects

The results presented so far compare cross-sectional variation among companies acquired
in a buyout. There is a natural concern that companies with varying levels of managerial
ownership vary significantly in other characteristics. If managerial ownership differs
between industries, we wish to filter out any variation in performance between industries
from the impact of managerial ownership. This study follows the methodology of Boucly
et al. (2011) and compare changes in operational performance with that of matched peers.
This facilitates the inclusion of industry-year fixed effects to account for industry-wide
shocks as suggested by Gormley and Matsa (2013). This may be particularly relevant to
control for in a private equity setting, as Jenkinson et al. (2018) provide evidence that
private equity may exhibit some skill in selecting outperforming industries.
The matching procedure used in this paper is similar to that used by Boucly et al.
(2011) and Bernstein et al. (2019). A portfolio company is matched with up to five
companies that i) belong to the same 2-digit SIC industry; ii) are not owned by another
private equity firm; iii) are within a ±15% revenue band of the acquired company in the
year prior to the buyout; and iv) have an EBITDA profitability deviating by at most ±5

186
percentage points of the target acquired company in the year prior to the buyout.16 In
the event that more than five potential matches are found, the closest five in a squared
distance sense are used, with revenue and profitability given equal weight. The results
presented are qualitatively similar for alternative bands, as well as the inclusion of number
of employees as a matching variable, with results available from the author upon request.17
With matched control companies, we augment the main specification (1) to account
for variation between companies acquired by private equity and controls companies.
Thus, the model to be estimated is,

Yi,t,j =αi + αt × αj + βCEO Ownershipi ∗ Posti,t ∗ PEi (2)

+ γPosti,t ∗ PEi + λPosti,t + θXi ∗ Posti,t + εi,j,t ,

where Yi,j,t is (the log of) revenue for company i, operating in industry j, measured
in year t; αi , αj , and αt are company, industry, and year fixed effects, respectively;
CEO Ownershipi measures the fraction of ordinary shares held by the CEO post-buyout;
Posti,t is an indicator variable which takes the value of 1 if the end of a fiscal-year
falls in the period after the buyout is completed, and zero if it is before; Xi contains
controls at the company level measured at the year prior to the buyout. The newly
added term PEi takes the value of 1 if company i was acquired in a buyout, and 0
if it is a matched control company.
By including a matched control set, the estimation now examines the increase in the
post-buyout value of the dependent variables for companies acquired in a buyout relative
to matched control companies. The focus is again on β1 , which captures the extent to
which the difference in operational improvements between PE-backed companies and
matched peers is different in companies with higher post-buyout CEO ownership shares.
CEO ownership is measured only for PE-backed companies and not for the matched peers,
which is why the additional interactions are not included in the estimation.18
16
As the entity reporting consolidated accounts is normally a holding company while under PE
ownership, the reporting entity’s SIC code is usually starting with 67, corresponding to a holding company
of an investment office. This classification is not informative of the activity of the portfolio company,
and instead the SIC classification of the portfolio company is used in the matching process as well as in
construction of industry fixed effects.
17
As pointed out by Boucly et al. (2011), while tighter bands improve the similarity in matched
companies, it leads to more companies ending up with fewer, or no, matched controls. The alternative
bands that have been attempted are revenue within ±5% and ±30%, and profitability margins of either
±3 percentage points or ±10 percentage points. When number of employees are included, a similar band
as for revenue has been utilized.
18
The reason we do not measure CEO ownership stakes is primarily because we are interested in the
change in the ownership stake resulting from the buyout acquisition, as will be the focus from the next
section. While one can measure this for peers as well, in practice the CEO ownership stake is rarely
changed, especially in private companies. If it would be measured, the vast majority of changes would
therefore take the value of 0 and it is not expected to have much impact on the results of the paper. In
addition, changes in ownership stakes that are not shocked bring it own set of endogeneity issues as it
reflect active choices by companies.

187
 Table 5 reports the estimates of (2). Columns 1-3 use log revenue as the dependent
variable, while columns 4-6 instead use the EBITDA margin of the company as the
dependent variable. The first model (columns 1 and 4) includes company fixed effects,
company controls and year fixed effects. The second model (columns 2 and 5) replaces
the year fixed effects with industry-year fixed effects. The third model (columns 3 and 6)
drops the company control variables. Standard errors are clustered at the company level.
The results are similar to the previous table. Companies with higher post-buyout CEO
ownership are associated with increased profitability, but not higher revenue, following
the buyout. In terms of economic magnitude, the point estimate suggests that the
improvement in profitability margin relative to matched peers is 49–58 basis points
higher in a company in which the CEO ownership is 10 percentage points higher. This
corresponds to a 3.1%–3.6% increase from the unconditional mean profit margin prior
to the buyout. The point estimates for revenue suggests a decrease of 2.2%–3.0% for a
company with a 10 percentage points higher CEO ownership, although the estimated
coefficients are statistically insignificant.
The direct comparison of companies acquired in a buyout compared to the matched
peers indicates a sharp difference in revenue: buyout targets grow more rapidly. Buyout
targets increase revenue by about 32% more than their peers. To the extent there is a
difference in profitability margin improvements, it is driven by companies with higher
CEO ownership stakes. It is worth noting that profitability margins are lower in the
post-period for both PE targets and the matched peers, except for buyout companies
with high levels of CEO ownership.

4.2 Interpretation of the Results

Tables 4 and 5 establish the main correlation of the paper: profitability improves more
for companies acquired in a buyout where the CEO receives a higher ownership stake.
There are however several explanations as to what may cause this correlation. Three
possible explanations are examined here, and the private equity setting is utilized in an
attempt to tease out which of these appear to be plausible.
First, the correlation observed may reflect talented CEOs being matched to companies
as in Gabaix and Landier (2008). The idea is that there is a distribution of ability among
CEOs, and the companies that are willing to offer the highest pay will attract the best
CEOs. As a higher ownership stake implies a higher payoff in a successful exit, companies in
which a CEO is offered a larger stake attract higher ability CEOs that improve operations.
Second, the observed relationship might reflect private information held by the
management team. If the existing CEO stays on and have company-specific knowledge
of costly but value-increasing actions, they gain bargaining power in the negotiations

188
with the private equity acquirer. This inside information explanation would suggest
that the management team obtain a higher ownership share when there is greater
potential to improve operations.
Third, the higher ownership stakes may provide the management team with incentives
to take costly actions that increases value. The role of incentives is theoretically clear,
but convincing empirical evidence of a causal effect of incentives is notoriously hard
to come by (Edmans et al., 2017).

4.2.1 Changes in ownership stakes

In a buyout, the ownership share held by the CEO is virtually always altered. This
allows the econometrician to work with changes in ownership and not only cross-sectional
variation. This is helpful as we can control for pre-buyout ownership levels. To the
extent that unobservable, company-specific, characteristics determine the optimal level
of ownership in a given company, controlling for pre-buyout ownership stakes allows
us to partially control for it.
In terms of the three proposed explanations for the observed results, the attracting
talented CEO explanation suggests that the post-buyout stake is what matters in attracting
the best CEOs, regardless of what the pre-buyout stake was. Similarly, the bargaining
story suggests that management with better information will extract a larger post-buyout
share. Thus, in both of these explanations the post-buyout ownership stake should be
the most relevant factor for improvements in operational performance. By contrast,
the incentive mechanism suggests that only increased ownership stakes should lead to
improved performance as that is when the incentive is altered. Merely retaining the
post-buyout stake at a high, pre-buyout, level does not increase incentives, and hence
we would not expect an improvement of operational performance. It should be noted
that the bargaining story is based on private information of the CEO. A high pre-buyout
ownership share might enable CEOs with positive private information to reinvest a
larger share to increase the post-buyout stake, so this is not a conclusive test between
those this explanation and that of incentives.
Table 6 presents the results from running the previous model with the post-buyout
ownership level decomposed into a pre-buyout ownership level, and the change from the
pre- to the post-buyout period. The results make clear that the relationship between
the post-buyout CEO ownership stake and profitability improvements is largely driven
by changes in CEO ownership. As before, revenue is not consistently related to the
ownership stake. In terms of economic magnitude, a 10 percentage unit increase in
CEO ownership translates to a 68–80 basis point increase in the EBITDA margin, or
a 4.3%–5.0% increase from the average pre-buyout levels.

189
 Although the estimated coefficient is higher for changes in ownership than for the
pre-buyout level of ownership, the estimated coefficients are not statistically different
from each other. Nonetheless, these results are consistent with the notion that a change
in ownership level is more relevant for improving operations.

4.2.2 The parallel trends assumption

The analysis relies on a triple difference-in-difference estimation, comparing operational

improvements among buyout targets to that of matched peers, allowing any difference to
vary with the ownership stake awarded to the CEO. As in any difference-in-difference
estimation, the identifying assumption is that the measured difference would not had taken
place had the “treatment” not happened. As previously argued, we expect cross-sectional
variation in ownership stakes due to company fundamentals. What the buyout acquisition
brings in form of “treatment” is shocking the ownership stake of the CEO. Therefore, the
change in CEO ownership stakes, rather than the CEO ownership level, is what could
plausibly be inferred as a treatment brought about by a buyout.
It is not easy to convincingly argue that the acquired companies would have behaved
similarly had the treatment not taken place. A minimum requirement, though not
sufficient, is that the continuously measured treatment do not display any pre-trends
that merely continues, so that at least in this dimension companies that experienced a
large increase in the CEO’s stake are similar to those that did not, controlling for other
covariates. We can formally test this by estimating the model

Yi,t,j =αi + αt × αj + βs (∆CEO Ownershipi ∗ PEi )

X

s
 
+ υs CEO Ownershipi,t−1 ∗ PEi (3)
X

s
+ γs (PEi ) + λs + θXi ∗ Posti,t + εi,j,t ,
X X

s s

where βs , υs , γs , and λs are estimated separately for each year in the pre- and post-buyout
period, with t − 1 used as the reference year.
Table 7 presents the results of this estimation. Overall, there is a substantial increase
in the relationship between changes in CEO ownership stakes and company profitability
following the buyout. Economically, the increase appears to be largest for the first two
full years under PE ownership (t + 1 and t + 2), while becoming weaker both economically
and statistically the further out we go.
While the jump in relationship between profitability and ∆CEO ownership around
the buyout event is comforting, the table is not quite as strongly suggesting a lack of
relationship in the pre-period as one might wish for. The coefficients are consistently
positive, though never statistically significant.

190
 However, it is worth thinking about what we are worried about. If the trend observed
was that of a monotonically increasing coefficient, whose increase continued post-buyout,
it is not plausible to argue that there is an impact of the buyout. This is not the case
however. All estimates are relative to the reference year t − 1. A positive coefficient
for ∆CEO ownership in t − 3 therefore suggests that companies that experienced high
increases in CEO ownership were more profitable in t − 3 than in t − 1. In all specifications,
the coefficient is lower in t − 2 than in t − 3. If anything, the trend is going in the opposite
direction, and is reversed following the buyout. One possible interpretation is that the
acquiring private equity firm increases the CEO stake particularly much in companies
that have experienced a recent drop in profitability.
Turning to the results on revenue there is a consistent negative, though statistically
insignificant, correlation between ∆CEO ownership in all years. Unlike the profitability
regressions, there is no consistent increase or decrease in the coefficients that suggest
an impact of the treatment.
To alleviate potential concerns that the observed trends are driven by the inclusion of
the specific set of matched peers I repeat the estimation excluding all matched peers. The
results are presented in Table A.3. There is a similar reversal in trend on profitability
in this table which indicates that the inclusion of matched peers are not driving the
results. The estimates on revenue provide some weak evidence that revenue increase more
when the CEO ownership is increased more, though it is not consistently estimated for
all periods or over all models so should be interpreted carefully.
Overall, the results are largely supportive that the main result of a positive relationship
between changes in CEO ownership stakes and profitability improvements is not driven by
the continuation of a pre-trend. However, this is not sufficient to conclude a causal effect
of increasing a CEO’s ownership stake. To accept that, we have to believe that no other
simultaneous changes takes places which drives profitability improvements. Since buyouts
may bring about multiple unobserved changes that potentially impact profitability, this
is difficult to argue. If such changes are correlated with increased CEO stakes we suffer
from omitted variable bias and would incorrectly attribute the effect to the change in
CEO stakes. For example, if GPs attempt to incentivize CEOs by providing larger stakes
when they believe a restructuring is necessary to improve profitability, they may take
a more active role in other aspects of running the company.
We can take some steps to mitigate this concern. First, note that companies acquired in
buyouts do not experience improvements in profitability on average. This true both on an
absolute level, and when compared to matched peers. Details on the results comparing the
matched peers are provided in Appendix C, which provides additional descriptive statistics
on the acquired companies in the sample compared to the set of matched companies. To

191
briefly summarize the results, buyout targets increase revenue substantially more than
matched peers, though there is an existing trend of higher growth among buyout targets.
However, buyout targets experience no improvement in profitability, either on an absolute
scale or relative to matched peers. Hence, only targets in which the CEO ownership stake
increase experience profitability improvements. It is therefore unlikely that the results of
the paper captures a common effect arising from transitioning into PE ownership.

4.2.3 Retained or replaced CEO

An important feature of buyouts is that the CEO is often replaced by the acquiring PE
house (e.g. Cornelli and Karakas, 2015). A natural split of the sample is therefore deals
in which the CEO is retained and those in which the CEO is replaced. In this paper, the
CEO is retained in 197 transactions (61%), and replaced in the remaining 125 (39%).
The act of replacing a CEO is clearly endogenously chosen by the acquirer, potentially
due to a perceived inability of the current management team to perform. While this
suggests that the two subsets may perform differently on average, a sample split can
nonetheless be informative of the relevant interpretation of the paper’s main result.
Suppose that the main underlying mechanism for the result is that the most talented
CEOs are attracted to companies offering the highest ownership stakes. Arguably, this
explanation is more relevant when the CEO is replaced, since the acquirer needs to attract
a new person to the role and the pool of potential CEOs is large. By contrast, the story
of inside information and bargaining power assumes that the CEO is an insider, not
brought in from the outside. Similarly, the incentive story is more plausible when a CEO
is retained, as an ownership change then represents an change in incentives for the same
manager. In this scenario it cannot be that a higher ownership stake attracts a higher
ability CEO, as the CEO remains the same person.19 This may also be instructive to the
discussion on whether increases in CEO ownership correlates with other actions taken by
the acquiring PE firms. If PE firms that take a more active role in running the company
are more likely to replace the CEO, we may learn whether this concern is small or large
depending on which subsample drives the results of the paper.
Table 8 presents the result of estimating the model in the paper separately for deals
in which the CEO is retained and deals in which the CEO is replaced. Panel A contains
the results for the subsample in which the CEO is retained, while Panel B presents the
results for the subsample in which the CEO is replaced.
As is evident from the two tables, the documented relationship between improved
profitability and changes in ownership is driven by the subsample where the CEO is
replaced, and it is economically stronger for the change in ownership stake compared
19
It is not quite clear how to interpret an “increase” in ownership when a CEO is replaced.

192
to the level. In terms of economic magnitude, a 10 percentage point increase in CEO
ownership when the CEO is retained is related to a 77–102 basis points increase in
profitability, corresponding to a 4.8%–6.4% increase from the unconditional average pre-
buyout level. By contrast, in the subsample where the CEO is replaced, the coefficient is
statistically indifferent from zero, and the magnitude of the coefficient is much smaller. In
neither sample is the coefficient on pre-buyout ownership level statistically significant for
profitability improvements. For revenue, there is no statistically significant relationship
between changes in revenue and CEO ownership when the CEO is retained. By contrast,
higher ownership stakes in companies with a replaced CEO are associated with a reduction
in revenue. However, the statistical significance is somewhat weak, so these results
should be interpreted with caution.
As buyouts commonly employ a large amount of leverage, the junior ownership stake
held by the CEO is akin to an option. It is possible that CEOs with larger stakes pursue
riskier strategies as higher volatility increase the value of an option claim. This could
be problematic, as companies that enter financial difficulties tend to not submit their
reports in the last fiscal year prior to entering insolvency processes. Hence, if strategies
with higher variance are employed and we observe fewer of the bad outcomes, it could
bias our results in the direction of what we find.
To evaluate whether this is a concern we can estimate whether CEO ownership
stakes are related to the probability that the company is seized by creditors or enter
insolvency procedures. Table A.4 presents such estimates for all exited deals. The model
is estimated using a linear probability model in the first five columns, while a logit model
specification is estimated in the last five columns. It is clear from the table that increases
in CEO ownership stakes do not appear to be related to an increase in the probability of
default, particularly when the CEO is retained. All estimated coefficients are statistically
insignificant, and the sign is always negative. Consequently, the results of the paper
do not appear to be driven by increased risk-taking.
Naturally, one may wonder whether companies in which the CEO is retained differs
systematically from those in which the CEO is replaced. Private equity firms may for
example be inclined to replace CEOs in companies that are doing poorly. Clearly, whether
a PE firm attempts to retain a CEO or not is an endogenous choice, as is the decision of
the CEO to stay on if offered the opportunity. We cannot observe the decision-making
process, but we can compare these companies on observable dimensions.
Table A.5 compares companies for which the CEO is retained versus those for which the
CEO is replaced. There is only small differences between these companies in levels in the
year prior to the buyout. This is quite remarkable given that these two sets are not matched
in any dimensions. The average company in which the CEO is replaced is slightly larger,

193
though the median is smaller. Companies in which the CEO is replaced have more assets,
and this is the only difference which is statistically significant. Importantly for the main
result of the paper, there is no difference in profitability margins for the average company,
and the median company in which the CEO is replaced have higher profitability margins.
There are slightly more differences in trends, as companies that retain the CEO grow
faster in the period leading up to the buyout, though this is only statistically different
for revenue at the one-year horizon. Interestingly, there is no clear trend on profitability
margin improvements: companies with a retained CEO has improved it a bit more in
the last year, but not over the two-year horizon.
We can also compare differences in CEO ownership levels between companies with
retained versus replaced CEOs. Table A.6 present this for both the post- and pre-buyout
ownership levels in Panels A and B, respectively. The ownership levels post-buyout are
remarkably similar, suggesting that there is a common level of ownership that private equity
firms prefer their managers to hold. There is some evidence that retained CEOs have higher
post-buyout ownership shares for private-to-private buyouts and secondary buyouts, where
the mean difference is statistically significant at the 10% level. Presumably, this is because
these are the categories with the highest pre-buyout ownership status. It is also clear that
the distribution of buyouts in which the CEO is replaced or retained is different: private-
to-private buyouts are overrepresented among buyouts in which the CEO is replaced.
There is far clearer differences in the pre-buyout ownership levels: companies in which
the CEO is replaced have much higher pre-buyout ownership levels, driven largely by a
higher fraction of private-to-private buyouts and a significantly higher ownership level
among private-to-private buyouts. The average (median) private-to-private buyout in
which the CEO is replaced has pre-buyout CEO ownership of 45.2% (50.0%). These
are often founder-owned or family-run businesses with one or a few, large, shareholders.
If improving incentives is an important contribution of buyouts leading to improved
operational performance, these are not the type of buyouts where one expects buyouts to
play a large role, as there is no pre-buyout issues of separation of ownership and control.
Taken together, the results presented in Tables 6, 8, and 8 are mostly consistent with
the incentive story, while matching of talented CEOs to companies does not appear to be
the main explanation. While the results of the paper are consistent with the incentive story,
they are generally consistent with the explanation of managerial bargaining power with
private information on value-increasing actions. To give a strong claim of the incentive
story we would need a plausibly exogenous instrument for changes in the ownership
stake, and such clean identification is not provided here.

194
4.3 Sources of Profitability Improvements

If we believe the previously presented results, a natural follow-up question is how these
profitability gains are achieved. How private equity generate returns is a hotly debated
topic, with proponents arguing that private equity improves operations while critics argue
that they are detrimental to other stakeholders of the company. For example, Boucly
et al. (2011) show that PE ownership can release financial constraints of companies,
while Bernstein and Sheen (2016) provide convincing evidence that restaurants over
which private equity firms have control becomes cleaner and safer compared to franchised
restaurants in the same chain. By contrast, Antoni et al. (2019) document that private
equity targets in Germany experience a reduction in employment and employee earnings.
Eaton et al. (2019) show that buyouts lead to higher tuition and per-student debt, while
there is lower graduation rates, loan repayment rates, and earnings among graduates.
In light of this debate it is interesting to examine how changes in CEO ownership
are associated with the components of the profitability improvements. In principle,
profitability improvements can be the result of increasing revenue with costs growing at
a slower rate, or cutting costs while revenue decreases at a slower rate. We therefore
turn to examine how changes in CEO ownership is related to three items: operational
costs; revenue; and EBITDA.
Table 9 presents results for the subsamples in which the CEO is retained in Panel
A, while Panel B presents results when the CEO replaced. This split is relevant as the
main result of the paper originates from the retained CEO subsample, so we expect
differences along this dimension. To make observations comparable between companies of
different sizes, each variable is standardized to be presented on a per employee basis. The
dependent variable in columns 1-3 is the log of operating costs/employee, in columns 4-6
it is the log of revenue/employee, and in columns 7-9 it is the log of EBITDA/employee.20
For the subsample where the CEO is retained, costs per employee is negatively related
to both increases and the pre-level of CEO ownership stakes, though only statistically
significant at the 10% level. Revenue per employee is also negatively related to ownership
stakes, though not statistically significant. As both costs and revenue are reduced, the
impact on EBITDA/employee is relatively small. While the coefficient is (statistically
insignificant and) negative, the magnitude is smaller than the effect on revenue, resulting
in a net improvement of profitability. This suggests that the profitability improvements

20
Other standardisations are possible, such as dividing by total assets. However, as goodwill from the
acquisition impact total assets it is distorted by the acquisition price, which is why number of employees
are used to standardize. Taking logs exclude negative EBITDA observations, and we consequently lose
some observations for EBITDA. Operational costs are not available for every company, which is why the
number of observations is lower than that of revenue.

195
associated with increased CEO ownership come about primarily through cutting costs,
rather than increasing revenue.
By contrast, the subsample where the CEO is replaced display a very different pattern.
Here costs per employee generally increases in companies with a higher post-buyout
CEO ownership stake, though the coefficient is not statistically significant in all models.
Revenue, while positive, is much smaller in magnitude, resulting in a modest negative
relationship between CEO ownership and EBITDA per employee. Note that the coefficients
for EBITDA and revenue are not statistically significant, so one needs to be careful in
reading too much into these point estimates. Still, as this essentially corresponds to a
decomposition of the profitability effects documented in earlier tables, it is suggestive
that the effect of CEO ownership stakes differs markedly between the subsamples where
the CEO is retained and where the CEO is replaced.

4.4 Determinants of Post-Buyout CEO Ownership

Having established a positive association between increases in CEO ownership and

profitability improvements, we now turn to examine which factors are relevant for
determining the ownership stake given to CEOs.
Looking at company characteristics, we expect company size, leverage, and profitability
to play a role. A given percentage of ownership represents more wealth invested in a
larger company. If CEOs are risk averse, we therefore expect them to have lower
percentage ownership stakes in larger companies. Similarly, conditional on size, a dollar
invested provides a higher percentage equity ownership stake in companies with higher
leverage. Working in the opposite direction, leverage increases the risk of the equity in a
company. Finally, a company with low profitability may require substantial restructuring.
If such actions are risky endeavours, a risk averse CEO may demand pay to be less
contingent on performance. However, if such restructuring events are dependent on CEO
effort, the acquiring private equity firm could be inclined to provide higher-powered
incentives. Additionally, the results presented so far suggest that higher ownership stakes
are associated with improved profitability. All else equal, an improvement of profitability
of one percentage point is worth more if the original profitability is 1% than if it is 10%.
This suggests that PE firms may wish to award higher CEO ownership stakes in companies
with low initial profitability. A higher CEO ownership stake implies less upside for the
private equity firm, and a higher stake should therefore be awarded when it worth more.
In addition to company characteristics, there are several reasons to expect that the
pre-buyout ownership stake held by the CEO is positively related to the post-buyout
CEO ownership stake. First, if unobserved company characteristics contribute to the
optimal CEO ownership stake, we may suspect that companies with higher ownership

196
stakes pre-buyout will have a high optimal ownership stake post-buyout. This should
be relevant whether the CEO is retained or replaced. Second, to the extent that the
private equity acquirer is attempting to improve incentives for retained CEOs we would
expect the pre-buyout ownership stake to positively correlate with the post-buyout stake;
a relatively small post-buyout ownership stake can improve incentives if pre-buyout
ownership level is even lower, while a large pre-buyout stake requires an even higher
post-buyout stake to increase incentives. Finally, for retained CEOs we should expect
a wealth effect: CEOs with a large pre-buyout stakes receive large payouts from the
acquisition and become wealthier. All else equal, this suggest that CEOs with higher
pre-buyout stakes should be more willing to invest a large dollar amount, as it represents
a smaller fraction of their wealth. Indeed, Cronqvist and Fahlenbrach (2013) and Kaplan
and Stein (1993) provide evidence that while CEOs cash-out in dollar terms, the average
ratio of post-LBO % ownership / pre-LBO % ownership is 2.86.
To examine the determinants of post-buyout CEO ownership, a Tobit model as well
as an OLS model are estimated. As ownership stakes are bounded between 0 and 1, the
Tobit model appears appropriate. However, Tobit models may struggle with the inclusion
of fixed effects and as such an OLS model is estimated for comparison. Fortunately, the
two models provide very similar estimates for all models suggesting that this is not a
major issue. Transaction year effects are included to control for time-varying trends in
ownership. Deal type dummies (e.g. public-to-private, secondary LBO) are included as
they have quite different pre-buyout ownership stakes as reported in Table 3.
The estimates of the two models are reported in Table 10. OLS estimates are presented
in columns 1-3, while the Tobit estimates are presented in columns 4-6. In columns 1
and 3, only company characteristics are included. Columns 2 and 5 adds the pre-buyout
CEO ownership stakes together with an interaction term indicating whether the CEO
is retained or not. In columns 3 and 6, acquisition type dummies are included.
Company size is negatively related to the post-buyout ownership stake of a CEO,
consistent with the ideas of risk averse CEOs and wealth constraints. The coefficient on
leverage is positive, but not statistically significant. Profitability is negatively related to
the ownership stake, suggesting that a desire to provide incentives to improve profitability
may dominate. Interestingly, a CEO’s pre-buyout ownership stake is positively related
to the post-buyout ownership stake, but only when the CEOs is retained. This suggests
that the desire to improve incentives is important for private equity acquirers.
It is somewhat surprising that there is no relation to be found in companies where the
CEO is replaced. One might expect the pre-buyout CEO stake contains some information
about the optimal stake of ownership in the company. However, this result could reflect
the higher proportion of very high pre-buyout ownership stakes in founder-controlled

197
companies. However, if that is the explanation we would expect that the inclusion of
buyout type dummies would mitigate this, as this is only the case in private-to-private
buyouts. This is not the case. In unreported results I have looked at whether the results
are different if separately estimate a coefficient for pre-ownership levels below and above
50%, but find no significant relationship between the pre- and post-buyout ownership
stakes in companies where the CEO is replaced.
To summarize, these results are consistent with CEO risk-aversion and an attempt by
private equity firms to provide incentives to the CEO of the acquired company.

5 Conclusion

This paper examines the impact of managerial ownership on operational performance

of companies acquired in a buyout. Knowledge of how specific changes implemented by
private equity firms relate to operational performance in portfolio companies is scarce,
and the main contribution of the paper is to examine the impact of one of the salient
features of private equity ownership: changes in managerial equity ownership stakes. The
paper assembles a new, partially hand-collected, dataset of financial information and
managerial ownership details for private equity portfolio companies registered in the UK.
The paper finds that companies with higher CEO equity ownership stakes experience
improved profitability. The results are economically stronger for changes in ownership
stakes compared to the level of ownership, and are driven by deals in which the CEO
is retained. For retained CEOs, a change in the ownership stake corresponds to altered
incentives, while managerial ability is held constant. High ownership stakes are awarded
in companies where improving profitability is particularly valuable: those with low
initial profitability. Stakes are higher when the CEO ownership stake is high pre-
buyout, but only when the CEO is retained, consistent with an attempt to improve
incentives from their pre-buyout levels.
The results are consistent with an interpretation that incentives play an important
role, though it cannot be ruled out that they reflect bargaining power of management
team’s with knowledge of valuable actions. It is important to note that the results
of the paper should not be interpreted as causal. A potential avenue for future re-
search is utilizing the private equity setting to examine plausibly exogenous variation
in ownership stakes awarded.

198
 Table 1: Sample Construction

This table presents how the final sample in the paper is derived. Transactions without participation of a
private equity firm or where the private equity firm(s) only acquired a majority stake are excluded from
the sample. Company names are taken from the transaction synopsis provided by Capital IQ, and if no
UK company of that name is found the transaction is excluded. The screen for buyout transactions in
Capital IQ was undertaken 2019-01-02. When ownership data is inaccessible it is typically because the
ultimate holding company is registered in the Cayman Islands, Jersey, Luxembourg, or another country
outside of the UK where ownership details are not available. The 8 transactions with missing ownership
filings are due to missing pages from the scanned documents in Companies House, and one case where the
submitted annual return is empty.

Number of UK buyouts identified in Capital IQ from 1999 - 2018 1964

Less: Buyouts not included in the 850 randomly selected (1114)

Number of PE transactions processed 850

Less: Transactions where no UK company of a stated name was found (19)
Less: Transactions without PE firm participation (15)

Less: Transactions where majority ownership did not pass to a PE firm (10)
Less: Duplicate transactions identified (7)
Less: Transactions initially identified but no longer in Capital IQ (5)
Less: Transactions of companies exempted from submitting filings (3)

Number of buyouts with entry and exit information 791

Less: Transactions without accounting data in year t-1 and t+1 (230)
Less: Transactions of financial companies or public utilities (69)

Number of buyouts with financials available 494

Less: Transactions with inaccessible ownership data post-buyout (99)
Less: Transactions with inaccessible ownership data pre-buyout (27)
Less: Transactions with missing ownership filings (8)

Less: Transactions for which no matched control companies is found (38)

Final sample 322

199
 Table 2: Summary Statistics

This table presents descriptive statistics of the portfolio companies in the sample. Panel A describes
characteristics of companies acquired in buyouts, measured at the most recent fiscal year prior to the
buyout, referred to as year -1. Panel B presents the change in these accounting measures from year -2 to
year -1. Panel C presents the distribution of buyout types. Panel D presents the distribution of exit
routes as of 2019-09-30. If no exit had occurred at that point it is indicated that the company is still
owned. The time-periods represents the year in which the buyout took place. The “Other” category
consists of 1 case where the portfolio company was owned by a private equity fund until it ceased trading.

Panel A: Portfolio company characteristics in year prior to buyout

N Mean Median SD 25th pct 75th pct
Revenue (£ million) 322 119.6 54.9 198.3 21.5 130.6
EBITDA (£ million) 322 15.2 7.8 21.3 3.4 16.5
Total Assets (£ million) 322 117.3 48.5 186.3 17.9 124.7
Number of Employees 317 1432 448 2445 186 1235
EBITDA Margin 322 0.160 0.142 0.103 0.092 0.218

Panel B: 1-year pre-buyout growth

N Mean Median SD 25th pct 75th pct
Revenue Growth 302 0.178 0.125 0.261 0.042 0.255
EBITDA 302 0.205 0.191 1.450 -0.000 0.446
Total Assets Growth 305 0.166 0.070 0.413 -0.024 0.238
Employee Growth 295 0.116 0.064 0.251 0.000 0.175
∆EBITDA Margin 302 0.010 0.009 0.055 -0.008 0.027

Panel C: Type of Acquisition

Full Sample 1999-2003 2004-2008 2009-2013 2014-2018
Public-to-Private 0.137 0.200 0.125 0.057 0.107
Private-to-Private 0.258 0.295 0.213 0.283 0.286
Divisional Buyout 0.208 0.333 0.184 0.094 0.071
Secondary Buyout 0.388 0.162 0.463 0.566 0.536
Distressed Buyout 0.009 0.010 0.015 - -
N 322 105 136 53 28

Panel D: Type of Exit

Full Sample 1999-2003 2004-2008 2009-2013 2014-2018
Lost Control 0.171 0.162 0.257 0.019 0.071
IPO 0.065 0.105 0.044 0.057 0.036
Trade Sale 0.264 0.381 0.228 0.208 0.107
Sale to Fin. Inst. 0.370 0.343 0.426 0.434 0.071
Still Owned 0.127 - 0.044 0.283 0.714
Other 0.003 0.010 - - -
N 322 105 136 53 28

200
 Table 3: Summary Statistics - CEO Ownership

This table presents descriptive statistics of the pre- and post-buyout CEO ownership stakes in portfolio
companies acquired in buyouts. Panel A presents the percentage ownership of ordinary shares that is
held by the CEO following a buyout of the company. The statistics are presented for the full sample on
the first row, and then split up by the different types of buyouts in the remainder of the rows. Panel B
presents statistics for the pre-buyout CEO ownership stakes.

Panel A: CEO Ownership after the Buyout

N Mean Median SD 25th pct 75th pct
Full sample 322 0.115 0.085 0.099 0.049 0.143

Public-to-Private Buyout 43 0.101 0.073 0.103 0.041 0.112

Private-to-Private Buyout 85 0.157 0.110 0.126 0.063 0.224

Divisional Buyout 61 0.085 0.068 0.067 0.037 0.113

Secondary Buyout 127 0.105 0.088 0.081 0.047 0.137

Distressed Buyout 6 0.103 0.101 0.037 0.082 0.142

Panel B: CEO Ownership prior to the Buyout

N Mean Median SD 25th pct 75th pct

Full sample 322 0.174 0.069 0.246 0.010 0.201

Public-to-Private Buyout 43 0.071 0.032 0.139 0.011 0.072

Private-to-Private Buyout 85 0.383 0.388 0.321 0.043 0.569

Divisional Buyout 61 0.054 0.000 0.184 0.000 0.007

Secondary Buyout 127 0.127 0.090 0.138 0.042 0.163

Distressed Buyout 6 0.162 0.105 0.207 0.000 0.278

201
 Table 4: Operational Improvements and CEO Ownership

This table reports the estimates of a (continuous) difference-in-difference fixed effect model of CEO equity
ownership on measures of company performance. The dependent variable is (log) revenue in columns 1-3,
and EBITDA margin in columns 4-6. The main independent variable is “CEO Ownership”, defined as the
fraction of outstanding common equity that is owned by the CEO of the company targeted in the buyout,
measured just after the buyout takes place. All specifications include company fixed effects and calendar
year fixed effects. Columns 1, 2, 4, and 5 include company level controls, interacted with a Post dummy.
The Post dummy takes the value of 1 in the post-buyout period, and is 0 otherwise. The company level
controls are measured in the fiscal year prior the buyout and are: (log) revenue; EBITDA margin; and
book leverage (measured after the buyout). Specifications in columns 2 and 5 add industry fixed effects.
The pre-buyout period includes up to three years prior to the buyout, while the post-period includes the
year in which the buyout took place, and up to four years after or until an exit occurs, whichever happens
first. All continuous variables are winsorized at a 1% level. Standard errors are clustered at the company
level and presented in parentheses below the estimate. ***, **, and * denotes statistical significance at
the 1%, 5%, and 10% levels, respectively.

(log) Revenue EBITDA/Revenue

CEO Ownership × Post 0.136 0.177 0.124 0.044∗∗ 0.050∗∗ 0.040∗

(0.153) (0.154) (0.135) (0.018) (0.021) (0.021)

Post 0.407∗∗∗ 0.394∗∗∗ 0.530∗∗∗ −0.007∗∗ −0.008∗∗ −0.002

(0.096) (0.093) (0.093) (0.003) (0.004) (0.005)

Adjusted R2 0.938 0.943 0.927 0.697 0.699 0.687

N 1573 1573 1573 1572 1572 1572
Company FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Industry FE No Yes Yes No Yes Yes
Company Controls Yes Yes No Yes Yes No

202
 Table 5: Operational Improvements and CEO Ownership, with Matched Peers

This table reports the estimates of a (continuous) difference-in-difference fixed effect model of CEO equity
ownership on measures of company performance. The dependent variable is (log) revenue in columns
1-3, and EBITDA margin in columns 4-6. The main independent variable is “CEO Ownership”, defined
as the fraction of outstanding common equity that is owned by the CEO of the company targeted in
the buyout, measured just after the buyout takes place. Each buyout target is matched with up to
five potential peers that are included in the estimation. Section 4.1.1 contains details of the matching
procedure. The dummy “PE” takes the value of 1 if the target was acquired in a buyout, and 0 otherwise.
All specifications include company fixed effects. Columns 1, 2, 4, and 5 include company level controls,
interacted with a Post dummy. The Post dummy takes the value of 1 in the post-buyout period, and is 0
otherwise. The company level controls are measured in the fiscal year prior the buyout and are: (log)
revenue; EBITDA margin; and book leverage (measured after the buyout). Columns 1 and 4 includes year
fixed effects, which are replaced by industry-year fixed effects in the remaining columns. The pre-buyout
period includes up to three years prior to the buyout, while the post-period includes the year in which
the buyout took place, and up to four years after or until an exit occurs, whichever happens first. All
continuous variables are winsorized at a 1% level. Standard errors are clustered at the company level and
presented in parentheses below the estimate. ***, **, and * denotes statistical significance at the 1%, 5%,
and 10% levels, respectively.

(log) Revenue EBITDA/Revenue

CEO Ownership × Post −0.261 −0.220 −0.297 0.054∗∗ 0.058∗∗ 0.049∗

(0.248) (0.233) (0.229) (0.027) (0.028) (0.027)

Post × PE 0.338∗∗∗ 0.320∗∗∗ 0.320∗∗∗ 0.001 0.001 −0.002

(0.057) (0.049) (0.047) (0.005) (0.009) (0.007)

Post 0.175∗∗∗ 0.056 0.095∗∗∗ −0.009∗∗∗ −0.012∗∗∗ −0.010∗∗

(0.042) (0.048) (0.028) (0.002) (0.002) (0.004)

Adjusted R2 0.898 0.902 0.889 0.523 0.525 0.512

N 7071 7071 7071 7052 7052 7052
Company FE Yes Yes Yes Yes Yes Yes
Year FE Yes No No Yes No No
Industry × Year FE No Yes Yes No Yes Yes
Company Controls Yes Yes No Yes Yes No

203
 Table 6: Operational Improvements and Changes in CEO Ownership

This table reports the estimates of a (continuous) difference-in-difference fixed effect model of CEO equity
ownership on measures of company performance. The dependent variable is (log) revenue in columns 1-3,
and EBITDA margin in columns 4-6. The main independent variables from Table 5, the post-buyout
ownership share held by the CEO, is here split up into a pre-buyout ownership share and the change.
“∆CEO Ownership” is defined as the change from the pre-buyout to the post-buyout level in fraction of
outstanding common equity that is owned by the CEO of the company targeted in the buyout, while
“CEO Ownershipt−1 ” corresponds to the pre-buyout ownership share. Each buyout target is matched
with up to five potential peers that are included in the estimation. Section 4.1.1 contains details of the
matching procedure. The dummy “PE” takes the value of 1 if the target was acquired in a buyout, and 0
otherwise. All specifications include company fixed effects. Columns 1, 2, 4, and 5 include company level
controls, interacted with a Post dummy. The Post dummy takes the value of 1 in the post-buyout period,
and is 0 otherwise. The company level controls are measured in the fiscal year prior the buyout and
are: (log) revenue; EBITDA margin; and book leverage (measured after the buyout). Columns 1 and 4
includes year fixed effects, which are replaced by industry-year fixed effects in the remaining columns.
The pre-buyout period includes up to three years prior to the buyout, while the post-period includes the
year in which the buyout took place, and up to four years after or until an exit occurs, whichever happens
first. All continuous variables are winsorized at a 1% level. Standard errors are clustered at the company
level and presented in parentheses below the estimate. ***, **, and * denotes statistical significance at
the 1%, 5%, and 10% levels, respectively.
(log) Revenue EBITDA/Revenue

∆CEO Ownership × Post × PE −0.247 0.009 0.100 0.068∗∗ 0.080∗∗ 0.080∗∗

(0.172) (0.218) (0.190) (0.031) (0.039) (0.040)

CEO Ownershipt−1 × Post × PE −0.051 −0.041 0.020 0.049 0.071 0.051

(0.187) (0.222) (0.217) (1.416) (1.459) (0.040)

Post × PE 0.211∗∗ 0.131 0.243∗∗∗ −0.007 −0.029 0.008

(0.102) (0.106) (0.079) (0.021) (0.026) (0.006)

Post 0.055 0.114∗∗ 0.034 0.028∗∗∗ 0.030∗∗ −0.015∗∗∗

(0.042) (0.055) (0.025) (0.011) (0.010) (0.003)

Adjusted R2 0.900 0.904 0.891 0.523 0.525 0.512

N 7071 7071 7071 7052 7052 7052

Company FE Yes Yes Yes Yes Yes Yes

Year FE Yes No No Yes No No
Industry × Year FE No Yes Yes No Yes Yes
Company Controls Yes Yes No Yes Yes No

204
 Table 7: Operational Improvements over Time

This table reports the estimates of the parallel trends test outlined in section 4.2.2. The dependent
variable is (log) revenue in columns 1-3, and EBITDA margin in columns 4-6. “∆CEO Ownership” is
defined as the change from the pre-buyout to the post-buyout level in fraction of outstanding common
equity that is owned by the CEO of the company acquired in the buyout. A separate coefficient is
estimated for ∆CEO Ownership for each year around the buyout, using t − 1 as the reference year. Note
that ∆CEO Ownership itself does not vary. A separate coefficient is similarly estimated for each year
around the buyout event for CEO Ownershipt−1 , as well as dummies for t + s × P E and t + s, though
omitted from the presentation of the table. Each buyout target is matched with up to five potential peers
that are included in the estimation. Section 4.1.1 contains details of the matching procedure. The dummy
“PE” takes the value of 1 if the target was acquired in a buyout, and 0 otherwise. All specifications
include company fixed effects as well as company controls interacted with a Post dummy. The Post
dummy takes the value of 1 in the post-buyout period, and 0 otherwise. The company controls are
measured in the fiscal year prior the buyout and are: (log) revenue; EBITDA margin; and deal leverage.
Specifications in columns 1 and 4 include year fixed effects, which are replaced by industry-year fixed
effects in remaining columns. The pre-buyout period includes up to three years prior to the buyout, while
the post-period includes the year in which the buyout took place, and up to four years after or until an
exit occurs, whichever happens first. All continuous variables are winsorized at a 1% level. Standard
errors, clustered at the company level, are presented in parentheses below each estimate. ***, **, and *
denotes statistical significance at the 1%, 5%, and 10% levels, respectively.

(log) Revenue EBITDA/Revenue

t − 3 × PE × ∆CEO Ownership −0.172 −0.309 −0.219 0.066 0.075 0.064

(0.340) (0.262) (0.320) (0.044) (0.053) (0.061)

t − 2 × PE × ∆CEO Ownership −0.245 −0.336 −0.298 0.047 0.029 0.014

(0.216) (0.241) (0.240) (0.042) (0.048) (0.049)

t × PE × ∆CEO Ownership −0.285 −0.339 −0.399 0.119∗ 0.082 0.080

(0.284) (0.246) (0.325) (0.072) (0.074) (0.070)

t + 1 × PE × ∆CEO Ownership −0.243 −0.290 −0.200 0.129∗∗ 0.099∗ 0.118∗∗

(0.154) (0.232) (0.176) (0.053) (0.055) (0.059)

t + 2 × PE × ∆CEO Ownership −0.105 −0.190 −0.117 0.122∗∗∗ 0.127∗∗∗ 0.153∗∗∗

(0.198) (0.274) (0.224) (0.043) (0.043) (0.055)

t + 3 × PE × ∆CEO Ownership −0.207 −0.312 −0.292 0.109∗∗ 0.098∗ 0.114∗∗

(0.187) (0.214) (0.204) (0.055) (0.056) (0.057)

t + 4 × PE × ∆CEO Ownership −0.064 −0.167 −0.101 0.093 0.086 0.086

(0.217) (0.247) (0.255) (0.061) (0.085) (0.079)

Adjusted R2 0.916 0.919 0.929 0.544 0.548 0.555

N 7071 7071 7071 7052 7052 7052

Company FE Yes Yes Yes Yes Yes Yes

Year FE Yes No No Yes No No
Industry × Year FE No Yes Yes No Yes Yes
Company Controls Yes Yes No Yes Yes No

205
Table 8: Operational Improvements and Change in CEO Ownership - Impact of CEO staying or
leaving

This table reports the estimates of a (continuous) difference-in-difference fixed effect model of changes
in CEO equity ownership on measures of company performance. Panel A reports estimates for the
subsample in which the CEO is retained, while Panel B does it for the subsample in which the CEO is
replaced. The dependent variable is (log) revenue in columns 1-3, and EBITDA margin in columns 4-6.
The main independent variables are “∆CEO Ownership”, defined as the change from the pre-buyout to
the post-buyout level in fraction of outstanding common equity that is owned by the CEO of the company
targeted in the buyout, while “CEO Ownershipt−1 ” corresponds to the pre-buyout ownership share. Each
buyout target is matched with up to five potential peers that are included in the estimation. Section 4.1.1
contains details of the matching procedure. The dummy “PE” takes the value of 1 if the target was
acquired in a buyout, and 0 otherwise. All specifications include company fixed effects. Columns 1, 2, 4,
and 5 include company level controls, interacted with a Post dummy. The Post dummy takes the value of
1 in the post-buyout period, and is 0 otherwise. The company level controls are measured in the fiscal
year prior the buyout and are: (log) revenue; EBITDA margin; and book leverage (measured after the
buyout). Columns 1 and 4 includes year fixed effects, which are replaced by industry-year fixed effects in
the remaining columns. The pre-buyout period includes up to three years prior to the buyout, while
the post-period includes the year in which the buyout took place, and up to four years after or until an
exit occurs, whichever happens first. All continuous variables are winsorized at a 1% level. Standard
errors are clustered at the company level and presented in parentheses below the estimate. ***, **, and *
denotes statistical significance at the 1%, 5%, and 10% levels, respectively.

Panel A: Subsample with retained CEO

(log) Revenue EBITDA/Revenue

∆CEO Ownership × Post × PE −0.128 −0.093 −0.110 0.077∗∗ 0.099∗∗ 0.102∗∗

(0.228) (0.207) (0.273) (0.035) (0.049) (0.048)

CEO Ownershipt−1 × Post × PE −0.092 −0.065 −0.060 0.019 0.040 0.050

(0.205) (0.233) (0.220) (0.032) (0.034) (0.037)

Adjusted R2 0.893 0.900 0.908 0.605 0.606 0.616

N 4451 4451 4451 4432 4432 4432

Panel B: Subsample with replaced CEO

(log) Revenue EBITDA/Revenue

∆CEO Ownership × Post × PE −0.843∗∗ −0.726∗ −0.606 0.015 0.005 0.026

(0.331) (0.380) (0.406) (0.061) (0.083) (0.089)

CEO Ownershipt−1 × Post × PE −0.744∗∗ −0.551 −0.419 −0.017 −0.002 0.017

(0.361) (0.365) (0.428) (0.058) (0.086) (0.090)

Adjusted R2 0.896 0.900 0.914 0.524 0.525 0.541

N 2620 2620 2620 2520 2620 2620

Company FE Yes Yes Yes Yes Yes Yes

Year FE Yes No No Yes No No
Industry × Year FE No Yes Yes No Yes Yes
Company Controls Yes Yes No Yes Yes No

206
 Table 9: CEO Ownership and Employee Effects

This table reports the estimates of a (continuous) difference-in-difference fixed effect model of changes in CEO equity ownership on measures of company performance.
Panel A reports estimates for the subsample in which the CEO is retained, while Panel B does it for the subsample in which the CEO is replaced. The dependent
variable is (log) operating cost per employee in columns 1-3, (log) revenue per employee in columns 4-6. and (log) EBITDA per employee in columns 7-9. The main
independent variables are “∆CEO Ownership”, defined as the change from the pre-buyout to the post-buyout level in fraction of outstanding common equity that is
owned by the CEO of the company targeted in the buyout, and “CEO Ownershipt−1 ”, the pre-buyout ownership share. Estimates of variables not shown but included
in the model are available upon request to the author. Each buyout target is matched with up to five potential peers that are included in the estimation. Section 4.1.1
contains details of the matching procedure. The specifications in the table are as in Table 6. All continuous variables are winsorized at a 1% level. T-statistics are
reported in parenthesis under each estimates, with standard errors clustered at the company level. ***, **, and * denotes statistical significance at the 1%, 5%, and
10% levels, respectively.

Panel A: Subsample with retained CEO

(log) Cost per Employee (log) Revenue per Employee (log) EBITDA per Employee
∆CEO Ownership × Post × PE −0.251 ∗
−0.303 ∗
−0.300 ∗
−0.107 −0.113 −0.139 −0.065 −0.032 -0.056
(0.130) (0.178) (0.171) (0.113) (0.174) (0.167) (0.379) (0.453) (0.385)
CEO Ownershipt−1 × Post × PE −0.242∗ −0.288∗ −0.286∗ 0.069 0.054
207

−0.199 −0.185 −0.206 −0.119

(0.131) (0.154) (0.146) (0.132) (0.171) (0.161) (0.344) (0.403) (0.385)

Adjusted R2 0.901 0.907 0.903 0.908 0.915 0.916 0.762 0.762 0.764
N 4075 4075 4075 4159 4159 4159 3873 3873 3873

Panel B: Subsample with replaced CEO

(log) Cost per Employee (log) Revenue per Employee (log) EBITDA per Employee
∆CEO Ownership × Post × PE 0.470 0.622 0.839∗∗ 0.004 0.178 0.413 −0.163 0.145 -0.158
(0.374) (0.445) (2.239) (0.180) (0.308) (1.020) (0.616) (0.580) (-0.384)
CEO Ownershipt−1 × Post × PE 0.546 0.479 0.755∗ −0.043 0.086 0.237 −0.021 0.007 -0.272
(0.422) (0.397) (1.883) (0.192) (0.297) (0.612) (0.683) (0.601) (-0.595)

Adjusted R2 0.875 0.881 0.882 0.937 0.943 0.943 0.801 0.800 0.809
N 2357 2357 2357 2460 2460 2460 2231 2231 2231
Company FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes No No Yes No No Yes No No
Industry × Year FE No Yes Yes No Yes Yes No Yes Yes
Company Controls Yes Yes No Yes Yes No Yes Yes No
 Table 10: Determinants of CEO Ownership

This table presents estimations of the determinants of the ownership share held by the CEO of a portfolio
company directly after a buyout. The CEO ownership share is defined as the fraction of ordinary
shares owned by the CEO. Columns 1-3 presents estimations using OLS, while columns 4-6 presents
the estimations of a Tobit model. The accounting variables are measured at the end of the fiscal year
prior to the transaction. The one exception is leverage, which is measured at the time of the transaction.
“CEO Retained” is an indicator variable taking the value of 1 if the CEO is retained, and 0 otherwise.
All models includes time fixed effects. Columns 3 and 6 includes a separate intercept term for the five
types of acquisitions “public-to-private”, “private-to-private”, “divisional buyouts”, “secondary buyouts”,
and “distressed buyouts”. For the Tobit model, the reported coefficient estimates correspond to marginal
effects evaluated at sample averages. All continuous variables are winsorized at a 1% level. Standard
errors are reported in parenthesis under each estimates. ***, **, and * denotes statistical significance at
the 1%, 5%, and 10% levels, respectively.

OLS Tobit

(Log) Revenue −0.024∗∗∗ −0.016∗∗∗ −0.017∗∗∗ −0.022∗∗∗ −0.015∗∗∗ −0.016∗∗∗

(0.005) (0.005) (0.005) (0.004) (0.004) (0.005)
Leverage 0.031 0.019 0.017 0.036 0.025 0.024
(0.033) (0.032) (0.030) (0.029) (0.028) (0.028)
EBITDA/Revenue −0.134∗∗∗ −0.112∗∗∗ −0.151∗∗∗ −0.101∗∗∗ −0.117∗∗∗ −0.151∗∗∗
(0.051) (0.040) (0.051) (0.034) (0.036) (0.045)
CEO pre-Ownership 0.027 0.020 0.027 0.023
(0.027) (0.029) (0.026) (0.026)
CEO pre-Ownership 0.225∗∗∗ 0.225∗∗∗ 0.204∗∗∗ 0.204∗∗∗
×1{CEO Retained} (0.044) (0.044) (0.039) (0.039)
1{CEO Retained} −0.005 −0.003 −0.002 0.000
(0.014) (0.014) (0.014) (0.013)

Adjusted (Pseudo) R2 0.109 0.255 0.263 0.089 0.207 0.213

N 322 322 322 322 322 322
Transaction Year FE Yes Yes Yes Yes Yes Yes
Acquisition type dummies No No Yes No No Yes

208
Appendices
A Appendix Tables

209
 Table A.1: Summary Statistics - Comparison with Full Buyout Sample

This table presents a comparison of descriptive statistics between companies in the sample and the full set of companies with financials available. The two sample
constructions are described in Table 1. Panel A compares statistics for the most recent fiscal year prior to the buyout, referred to as year -1. Panel B presents the
change in these accounting measures from year -2 until year -1. Panel C presents changes from year -3 until year -1. Panels D, E, and C compares the buyout types,
exit routes, and industry composition, respectively. The final column in panels A, B and C tests differences in means, and numbers followed by ***, **, or * indicate
statistically significant means at the 1%, 5%, or 10% levels, respectively.
Panel A: Company characteristics in year prior to buyout
Sample in Paper Full Sample
N Mean Median SD N Mean Median SD Mean diff.
Revenue (M£) 322 119.6 54.9 198.3 494 128.4 56.5 206.7 −8.78
EBITDA (M£) 322 15.2 7.8 21.3 494 16.5 7.1 25.2 −1.34
Total Assets (M£) 322 117.3 48.5 186.3 494 139.1 53.7 220.1 −21.74
Number of Employees 317 1432 448 2445 482 1428 456 2397 3.88
EBITDA Margin 322 0.160 0.142 0.103 494 0.150 0.136 0.108 0.01
210

Panel B: 1-year pre-buyout growth

Sample in Paper Full Sample
N Mean Median SD N Mean Median SD Mean diff.
Revenue 302 0.178 0.125 0.261 460 0.145 0.099 0.251 0.033∗
EBITDA 302 0.205 0.191 1.450 460 0.184 0.155 1.436 0.022
Total Assets 305 0.166 0.070 0.413 464 0.140 0.041 0.400 0.026
Number of Employees 295 0.116 0.064 0.251 448 0.095 0.043 0.252 0.021
EBITDA Margin 302 0.010 0.009 0.055 460 0.006 0.007 0.068 0.004

Panel C: 2-years pre-buyout growth

Sample in Paper Full Sample
N Mean Median SD N Mean Median SD Mean diff.
Revenue 289 0.469 0.289 0.681 433 0.378 0.203 0.608 0.091∗
EBITDA 288 0.839 0.400 2.592 432 0.583 0.327 2.347 0.256
Total Assets 294 0.414 0.196 0.785 442 0.360 0.140 0.792 0.055
Number of Employees 283 0.294 0.158 0.610 422 0.253 0.110 0.586 0.041
EBITDA Margin 288 0.017 0.011 0.065 432 0.012 0.010 0.071 0.005
 Table A.1 - Continued

Panel D: Buyout type

Sample in Paper Full Sample
Public-to-Private 0.137 0.134
Private-to-Private 0.258 0.263
Divisional Buyout 0.208 0.231
Secondary Buyout 0.388 0.342
Distressed Buyout 0.009 0.030
N 322 791

Panel E: Exit Route

Sample in Paper Full Sample
Lost Control 0.171 0.162
IPO 0.065 0.062
Trade Sale 0.264 0.290
Sale to Financial Institution 0.370 0.316
Still Owned 0.127 0.159
Other 0.003 0.006
Not Known - 0.005
N 322 791

Panel F: Industry Composition

Sample in Paper Full Sample
Mining 0.8% 1.2%
Construction 3.0% 3.5%
Manufacturing 24.5% 27.7%
Transportation 7.5% 7.0%
Wholesale 3.8% 5.6%
Retail 13.6% 12.7%
Services 46.8% 42.2%

211
 Table A.2: Summary Statistics - Managerial Ownership

This table presents descriptive statistics of the pre- and post-buyout managerial ownership in buyout
targets. Panel A presents the percentage ownership of ordinary shares that is held by the management
team following a buyout of the company. The statistics are presented for the full sample on the first
row, and then split up by the different types of buyouts in the remainder of the rows. Panel B presents
statistics for the pre-buyout managerial ownership stakes.

Panel A: Management Team Ownership after the Buyout

N Mean Median SD 25th pct 75th pct
Full sample 322 0.283 0.249 0.183 0.150 0.384

Public-to-Private Buyout 43 0.238 0.182 0.200 0.118 0.314

Private-to-Private Buyout 85 0.346 0.310 0.190 0.185 0.423

Divisional Buyout 61 0.222 0.192 0.147 0.133 0.250

Secondary Buyout 127 0.291 0.266 0.178 0.169 0.386

Distressed Buyout 6 0.217 0.248 0.079 0.142 0.280

Panel B: Management Team Ownership prior to the Buyout

N Mean Median SD 25th pct 75th pct

Full sample 322 0.369 0.232 0.364 0.051 0.645

Public-to-Private Buyout 43 0.143 0.093 0.163 0.033 0.184

Private-to-Private Buyout 85 0.783 0.985 0.368 0.489 1.000

Divisional Buyout 61 0.094 0.000 0.244 0.000 0.030

Secondary Buyout 127 0.333 0.264 0.236 0.160 0.500

Distressed Buyout 6 0.328 0.172 0.421 0.002 0.600

212
 Table A.3: Operational Improvements over Time - Buyout Targets Only

This table reports the estimates of the parallel trends test outlined in section 4.2.2, excluding the matched
control sample. The dependent variable is (log) revenue in columns 1-3, and EBITDA margin in columns
4-6. “∆CEO Ownership” is defined as the change from the pre-buyout to the post-buyout level in fraction
of outstanding common equity that is owned by the CEO of the company acquired in the buyout. A
separate coefficient is estimated for ∆CEO Ownership for each year around the buyout, using t − 1 as
the reference year. Note that ∆CEO Ownership itself does not vary. A separate coefficient is similarly
estimated for each year around the buyout event for CEO Ownershipt−1 , as well dummies for t + s × P E
and t + s, though omitted from the presentation of the table. Each buyout target is matched with up to
five potential peers that are included in the specification. Section 4.1.1 contains details of the matching
procedure. The dummy “PE” takes the value of 1 if the target was acquired in a buyout, and 0 otherwise.
All specifications include company fixed effects as well as company controls interacted with a Post dummy.
The Post dummy takes the value of 1 in the post-buyout period, and 0 otherwise. The company controls
are measured in the fiscal year prior the buyout and are: (log) revenue; revenue growth; EBITDA margin;
and deal leverage. Specifications in columns 1 and 4 include year fixed effects, which are replaced by
industry-year fixed effects in remaining columns. The pre-buyout period includes up to three years prior
to the buyout, while the post-period includes the year in which the buyout took place, and up to four
years after or until an exit occurs, whichever happens first. All continuous variables are winsorized at
a 1% level. Standard errors, clustered at the company level, are presented in parentheses below each
estimate. ***, **, and * denotes statistical significance at the 1%, 5%, and 10% levels, respectively.

(log) Revenue EBITDA/Revenue

t − 3 × ∆CEO Ownership 0.101 0.109 0.094 0.039∗∗ 0.038∗∗ 0.042∗∗

(0.080) (0.072) (0.091) (0.017) (0.018) (0.018)
t − 2 × ∆CEO Ownership 0.084 0.085 0.065 0.028 0.028 0.031
(0.075) (0.074) (0.090) (0.026) (0.027) (0.029)

t × ∆CEO Ownership 0.020 0.038 0.070 0.020 0.021 0.022

(0.082) (0.086) (0.097) (0.018) (0.018) (0.019)
t + 1 × ∆CEO Ownership 0.089 0.134∗ 0.125 0.085∗∗∗ 0.090∗∗∗ 0.088∗∗∗
(0.075) (0.076) (0.097) (0.026) (0.027) (0.028)
t + 2 × ∆CEO Ownership 0.098∗ 0.149∗∗∗ 0.116∗∗ 0.106∗∗∗ 0.116∗∗∗ 0.105∗∗∗
(0.052) (0.055) (0.059) (0.019) (0.020) (0.022)
t + 3 × ∆CEO Ownership 0.021 0.076 0.033 0.064∗∗∗ 0.067∗∗∗ 0.067∗∗∗
(0.141) (0.135) (0.142) (0.024) (0.023) (0.025)
t + 4 × ∆CEO Ownership 0.059 0.136 0.064 0.068∗∗ 0.070∗∗ 0.079∗∗∗
(0.128) (0.141) (0.146) (0.027) (0.029) (0.027)

Adjusted R2 0.955 0.956 0.957 0.711 0.712 0.712

N 1573 1573 1573 1572 1572 1572

Company FE Yes Yes Yes Yes Yes Yes

Year FE Yes Yes Yes Yes Yes Yes
Industry FE No Yes Yes No Yes Yes
Company Controls Yes Yes No Yes Yes No

213
 Table A.4: Likelihood of Bankruptcy

This table presents estimations of the likelihood that a buyout ends up in insolvency procedures, creditors seize control, or a financial restructuring resulting
in a new owner taking over with the old equity holders receiving no compensation. The dependent variable takes the value 1 if this happens, and zero for
all other types of exits. Deals that are still controlled by the private equity firm as of 2019-09-30 are not included in the estimation. A CEO’s ownership
stake is defined as the fraction of ordinary shares owned by the CEO. “CEO Ownershipt−1 ” measures this stake in the pre-buyout period, while “∆ CEO
Ownership” measures the change in this measure from the pre-buyout to the post-buyout period. “1{CEO Retained}” is an indicator variable taking the value
of 1 if the CEO is retained, and 0 otherwise. Columns 1-5 presents estimations using OLS, while columns 6-10 presents the estimations of a Logit model. The
five columns differ in the set of control variables included. All models includes time fixed effects. From columns 3 and 8, industry fixed effects are included.
Columns 4 and 9 add company controls, and columns 5 and 10 includes a separate intercept term for each of the five types of acquisitions “public-to-private”,
“private-to-private”, “divisional buyouts”, “secondary buyouts”, and “distressed buyouts”. Company controls include (log) revenue, the EBITDA margin, and book
leverage. They are measured at the end of the fiscal year prior to the transaction, with the exception of leverage which is measured at the time of the transaction.
For the Logit model, the reported coefficient estimates correspond to marginal effects evaluated at sample averages. All continuous variables are winsorized
at a 1% level. Standard errors are reported in parenthesis under each estimates. ***, **, and * denotes statistical significance at the 1%, 5%, and 10% levels, respectively.

OLS Logit

∆ CEO Ownership −0.385 −0.118 −0.260 −0.195 −0.219 −0.381 −0.094 −0.334 −0.221 −0.284
214

(0.243) (0.383) (0.387) (0.408) (0.409) (0.253) (0.365) (0.399) (0.402) (0.409)

CEO Ownershipt−1 −0.158 0.175 0.030 0.156 0.178 −0.179 0.166 −0.067 0.117 0.105
(0.231) (0.378) (0.382) (0.404) (0.414) (0.239) (0.361) (0.394) (0.401) (0.411)

∆ CEO Ownership −0.379 −0.344 −0.364 −0.273 −0.467 −0.406 −0.478 −0.296
×1{CEO Retained} (0.515) (0.514) (0.530) (0.527) (0.538) (0.546) (0.550) (0.545)

CEO Ownershipt−1 −0.544 −0.429 −0.494 −0.520 −0.582 −0.460 −0.566 −0.544
×1{CEO Retained} (0.484) (0.483) (0.497) (0.494) (0.490) (0.507) (0.509) (0.508)

1{CEO Retained} 0.091 0.086 0.101 0.109 0.099 0.095 0.116 0.131
(0.077) (0.077) (0.079) (0.079) (0.077) (0.078) (0.079) (0.080)
Adjusted (Pseudo) R2 0.036 0.032 0.044 0.040 0.056 0.110 0.112 0.139 0.148 0.178
N 281 281 281 281 281 281 281 281 281 281
Transaction Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE No No Yes Yes Yes No No Yes Yes Yes
Company Controls No No No Yes Yes No No No Yes Yes
Acquisition Type Dummies No No No No Yes No No No No Yes
 Table A.5: Summary Statistics of Companies with Retained and Replaced CEO

This table presents a comparison of descriptive statistics between companies in the sample in which the CEO was retained versus companies in which the CEO was
replaced. Panel A present company descriptives for the most recent fiscal year prior to the buyout, referred to as year -1. Panel B presents the change in these
accounting measures from year -2 until year -1. Panel C presents changes from year -3 until year -1. The final column tests differences in means, and numbers followed
by ***, **, or * indicate statistically significant means at the 1%, 5%, and 10% levels, respectively.

Panel A: Company characteristics in year prior to buyout

CEO Retained CEO Replaced
N Mean Median SD N Mean Median SD Mean diff.
Revenue (M£) 197 117.1 60.0 167.5 125 123.7 47.6 241.0 −6.55
EBITDA (M£) 197 15.3 8.8 20.0 125 14.9 6.0 23.5 0.44
Total Assets (M£) 197 108.7 50.2 151.1 125 131.1 39.8 232.5 −22.52∗
Number of Employees 194 1330 582 2198 123 1599 340 2804 -269
EBITDA Margin 197 0.160 0.140 0.101 125 0.160 0.154 0.102 0.00
215

Panel B: 1-year pre-buyout growth

CEO Retained CEO Replaced
N Mean Median SD N Mean Median SD Mean diff.
Revenue 185 0.203 0.136 0.290 117 0.136 0.120 0.199 0.068∗∗
EBITDA 185 0.273 0.199 1.466 117 0.094 0.178 1.423 0.179
Total Assets 187 0.160 0.068 0.411 118 0.176 0.087 0.420 −0.016
Number of Employees 179 0.127 0.069 0.237 116 0.098 0.061 0.272 0.028
EBITDA Margin 185 0.015 0.011 0.057 117 0.002 0.007 0.051 0.013

Panel C: 2-years pre-buyout growth

CEO Retained CEO Replaced
N Mean Median SD N Mean Median SD Mean diff.
Revenue 177 0.490 0.289 0.685 112 0.434 0.289 0.676 0.056
EBITDA 176 0.901 0.488 2.381 112 0.738 0.369 2.916 0.163
Total Assets 180 0.443 0.209 0.813 114 0.366 0.170 0.739 0.077
Number of Employees 172 0.333 0.165 0.646 111 0.226 0.118 0.540 0.107
EBITDA Margin 176 0.015 0.012 0.063 112 0.019 0.010 0.069 −0.003
 Table A.6: Summary Statistics - CEO Ownership

This table presents descriptive statistics of the pre- and post-buyout CEO ownership in buyout targets, comparing the subsample of companies in which the CEO was
retained with those in which the CEO was replaced. Panel A presents the percentage ownership of ordinary shares that is held by the CEO following a buyout of the
company. The statistics are presented for the full subsample on the first row, and then split up by the different types of buyouts in the remainder of the rows. Panel B
presents statistics for the pre-buyout CEO ownership stakes. The final column tests differences in means, and numbers followed by ***, **, or * indicate statistically
significant means at the 1%, 5%, and 10% levels, respectively.

Panel A: CEO Ownership after the Buyout

CEO Retained CEO Replaced
N Mean Median SD N Mean Median SD Mean diff.
Full sample 197 0.117 0.093 0.097 125 0.107 0.083 0.089 −0.010

Public-to-Private Buyout 27 0.103 0.076 0.103 16 0.116 0.083 0.089 −0.013

Private-to-Private Buyout 35 0.178 0.144 0.138 50 0.126 0.100 0.095 0.052∗

Divisional Buyout 41 0.084 0.071 0.059 20 0.094 0.072 0.114 −0.010
216

Secondary Buyout 91 0.114 0.097 0.082 36 0.086 0.068 0.076 0.028∗

Distressed Buyout 1 0.143 0.143 n/a 5 0.112 0.085 0.032 0.031

Panel B: CEO Ownership prior to the Buyout

CEO Retained CEO Replaced
N Mean Median SD N Mean Median SD Mean diff.
Full sample 197 0.111 0.050 0.172 125 0.258 0.112 0.293 −0.147∗∗∗

Public-to-Private Buyout 27 0.046 0.028 0.045 16 0.144 0.040 0.247 −0.098

Private-to-Private Buyout 35 0.243 0.132 0.273 50 0.452 0.500 0.302 −0.209∗∗∗

Divisional Buyout 41 0.006 0.000 0.019 20 0.144 0.006 0.310 −0.138∗

Secondary Buyout 91 0.127 0.087 0.149 36 0.110 0.090 0.095 0.017

Distressed Buyout 1 0.500 0.500 n/a 5 0.154 0.118 0.070 0.346

B Data Description

This appendix provides additional details on the data collection process. Note that most
of this is covered in the second paper of the thesis.

B.1 Sample Construction

I follow Strömberg (2008) and use Capital IQ (CIQ) to identify buyout transactions.
The transaction screen was undertaken 2019-01-02. The following criteria were used in a
“Capital IQ Transaction Screen” to generate the initial set of transactions:

• “Transaction Primary Features” set to: Merger/Acquisition – Acquisition of Equity

Stake or Private Placement – PIPE or Private Placement – Growth capital/Private
Equity;

• “Merger/Acquisition Features” set to: “Going Private Transaction” or “Leveraged

Buy Out (LBO)” or “Management Buyout” or “Platform” or “Privatization of Gov-
ernment Entity” or “Management Participated” or “Secondary LBO” or “Acquisition
of Majority Stake”;

• “Secondary features” set to: “Leveraged Buy Out (LBO)”, or “Management Buyout”,
or “Secondary LBO”, or “Going Private Transaction”;

• “Investment Firm Type (Buyers/Investors)” is listed as “PE/VC”;

• Transaction is listed as “closed” or “effective”.

The initial screen picked up transactions that were initial toe-hold investments, a
purchase of remaining minor stakes, or the purchase of a minority stake. The focus here
are transactions in which a private equity buyout fund gains control of the company.
Therefore, transactions that are classified as either “Minority Stake Investment” or
“Majority Shareholder Purchasing Remaining Shares”, or where the investor was classified
as a “VC”, are dropped from the sample. This leaves an initial set of 11,549 transactions.
For all unique companies in these transactions, the following items are extracted:
company name; country of incorporation; filing currency; and the country of primary
office location. All UK companies are selected based on being incorporated in the UK
or, if that item is not available, either have GBP as its filing currency or has its primary
office in the UK. Due to the limited number of years with financials available in the
Bureau van Dijk (BvD) databases, we restrict the sample to transactions completed
between 1999 and 2018. This leaves the initial set of 1964 buyouts of UK companies
that forms the first row of Table 1.

217
 Of the identified buyouts of UK companies 850 are randomly selected to be processed.
Each transaction requires a substantial amount of manual work, necessitating working
with a random part of the sample. Figure A.3 in Chapter 4 outlines the composition
of buyouts over time of those randomly selected versus those not processed. Overall,
the distribution of buyouts over time is quite similar, though Capital IQ coverage is
low for the later years in the sample.
57 transactions are dropped from the sample for the following reasons:

• 19 companies are dropped as no UK company could be found either in Companies

House or Fame under the name of the acquired company in Capital IQ today or,
when it differed, the name of the acquired company in the transaction synopsis
provided by Capital IQ. Both Companies House and Fame contains previously
registered names of companies, so when no match can be found it likely reflects
either a mistake in the naming in Capital IQ or that some non-UK companies are
identified during the process of selecting UK-registered companies.

• 15 transactions are dropped as no PE firm participated. PE firm participation is

verified using annual reports and ownership filings.

• 10 transactions are dropped because the PE firm did not acquire a controlling stake.
This information is gathered from post-buyout ownership filings.

• 7 transactions are duplicates, where two separate company IDs in Capital IQ have
been assigned to the same transaction, with two separate transaction IDs.

• 5 transactions are dropped as the transaction ID identified in the initial screen were
no longer present in Capital IQ at the time a match was attempted to be established
with Fame.

• 3 transactions are dropped because the acquired companies were too small and were
exempted from submitting filings throughout the period of interest.

The remaining 791 buyouts are fully processed with the type of buyout classified and
the type of exit documented with the date and the type of exit. The type of buyout is
determined based on the type of ownership prior to the buyout. To establish the exit
route I utilize a combination of sources. Capital IQ, Zephyr, and news articles are scanned
for information regarding sales. When a sale is identified I corroborate the information
through reading annual reports of the acquired company and, when possible, the acquirer.
If no mention of a sale is identified I look at whether the company ended up in insolvency
procedures, called being placed in “administration” in the UK. When a company is placed

218
in administration, the administrator regularly submit filings to Companies House on the
progress.21 A company, or part of it, can be acquired out of administration. There are
cases in the sample where Capital IQ (and, less commonly, Zephyr) incorrectly attribute
the seller in these cases as the PE firm, even when none of the proceeds go to the PE
firm. This is why sales identified in the first step are corroborated by reading annual
report and searching for potential insolvency.22
For the remaining unclassified cases I utilize ownership filings to detect when a
PE firm cease to be the ultimate owner. This process either confirms that the PE
firm remains the ultimate owner, or document when a change in ownership takes place.
In the latter case I consult annual reports to find out the reason behind the change
in ownership. This exercise picks up some sales not picked up by the other sources.
Typically, these concern small companies when sold to the management team or other
individuals. Importantly, this exercise also picks up cases where creditors seize control of
a company by, for example, a debt-to-equity swap or other events in which the equity
holders receive no compensation for their stakes.
For the 12.7% of transactions in the paper that are classified as still being owned
by private equity that is as of 2019-09-30, when a potential sale was last searched for.
A lack of news about a sale is not sufficient to classify a company as still in private
equity hands. Ownership filings are used to verify the ownership status and searches
for insolvency filings are undertaken in Companies House.

B.2 Additional Details on the Accounting Data

In this paper I utilize the first year in which annual reports are submitted following a
buyout, referred to as year t = 0. Except by accident, the first annual report does not
cover exactly one calendar year worth of trading. The annual report itself will refer
to the period covered from the day of incorporation until the fiscal year end, but in
practice no trading takes place prior to the completion of the buyout acquisition. For the
first annual report of newly created companies, I collect the date that trading activity
commences from the annual report, and annualize income statement numbers through
linear extrapolation. Note that balance sheet items are snapshots and therefore requires

21
Administration is the UK equivalent of Chapter 11 in the US. The filings submitted by an administrator
include: a “Statement of administrator’s proposal”, which provides background information on the events
leading to administration, the strategy and progress to date of the administrator(s), and an assessment of
the likely outcome for the various classes of creditors; and semi-annual “Administrator’s progress report”
outlining the progress and outcome for creditors.
22
Since the Enterprise Act 2002, administrators can be appointed out of court. This enabled pre-
packaged administration, where a sale is negotiated prior to the company being placed in administration,
with the sale taking place immediately after. Almost all misclassified sales by Capital IQ relate to
pre-administration packages. Even in pre-packaged administration cases the administrator submits filings
to Companies House.

219
no adjustment. If the trading period covers less than 4 months of a year, the income
statement items are treated as missing since a shorter period may not be representative
of the full year. The results of the paper are robust to varying this accepted window to
between 3 and 6 months, or dropping t = 0 completely from the analysis.23
Fame provides financials for up to 20 years worth of annual reports. For the earlier
transactions in the sample this implies that the pre-buyout period is not always covered.
For these deals I manually complement the financials from annual reports. There are two
additional cases which prompts me to manually complement the Fame data.
First, occasionally there are fiscal years inexplicably missing from Fame even when
the annual report is available in Companies House. I add these years manually to
avoid gaps in the data.
Second, the acquisition event typically gives rise to a large amount of goodwill in the
acquiring parent company, and subsequent add-on acquisitions may further increase the
goodwill post in the balance sheet. The goodwill is amortised over a period of years and
may be subject to impairment charges. Fame does provide an EBITDA number that
accounts for depreciation and amortisation. However, it is not uncommon that one or both
of the two numbers are missing from individual years. Additionally, impairment charges
are only rarely included by Fame, though they are both common and of considerable
size. I have manually gone through Fame to detect missing depreciation/amortisastion
entries, and read annual reports to detect mentions of impairment charges to check
whether Fame picked them up.24 When impairments are not picked up by Fame I
complement the EBITDA calculations.25 This process is applied to the pre-buyout years
as well as the period under private equity ownership to avoid any bias arising from
different treatment of the data when under PE ownership. In total, financials for 843
company-years are either added or amended.26
Not all companies have to provide the same level of detail in their annual reports. In
particular, companies classified as “small” are only required to submit abridged accounts
containing limited information which excludes profit and loss information. These filings
23
Earlier studies on operational improvements generally do not include the year of acquisition in their
analysis and hence do not suffer from this problem. As noted earlier, the consolidating parent company
sometimes shift multiple times during holding period in which case it is relevant to account for this
problem outside of t = 0.
24
Impairment and write-downs can concern both fixed assets and intangible assets such as goodwill,
and both are commonly not accounted for by Fame. The charge always shows up somewhere in Fame,
and while there is some variation, the most common approaach appears to be to include it in general
“Administration Expenses”, and sometimes as an exceptional item.
25
The increase in interest payments and the typical increase in amortisation is the reason why EBITDA
measures are commonly used to evaluate the performance of PE portfolio companies.
26
A similar approach is taken to ensure that the accounts of matched peers are correct and available.
The vast majority of amendments are made for companies acquired in a buyout and relate to missed
impairment charges on goodwill, an item that is virtually always present in buyout targets and much less
frequently a major item in matched peers.

220
therefore cannot be used in the data analysis. The requirements to be classified as
a small company has changed over time, and any company-year in which a company
qualified as small is dropped from the sample.27

27
To be classified as a small company, a company needs to be smaller in at least two out of three
qualifying measures: turnover, total assets, and number of employees. The limit on number of employees
has been stable at 50 over time. The following requirements and time periods are relevant for the sample
in this paper, with numbers for turnover (total assets): prior to 2004-01-10, £2.8m (£1.4m); prior to
2008-04-06, £5.6m (£2.8m); prior to 2015-04-06, £6.5m (£3.26m); since 2015-04-06, £10.2m (5.1m).
Classifications are from Section 247 of the Companies Act 1985 and Section 382 of the Companies Act
2006, both links accessed 2019-08-22. I am grateful to the policy team at Companies House for providing
these links.

221
C Comparison PE targets and Matched Peers

This appendix provides a brief comparison of the quality of the matching procedure
outlined in section 4.1.1.
Table C.1 provides a comparison of the companies in the sample versus that of
matched control companies, measured at the end of the last fiscal year end prior to the
buyout. Panel A presents the results in levels, Panel B the changes over the last year
and Panel C the changes over the last two years.
As the control set was matched on companies being similar along the lines of revenue
and EBITDA margin, these two items and and EBITDA are very similar in levels. The
matched controls have more total assets, but fewer employees, although neither difference
is statistically significant due to large variation within each group. When it comes
to the changes leading up to the buyout, the typical buyout target have experienced
higher growth rates, especially measured over the 2-year horizon. Though both groups
have experienced improvements in the EBITDA margin, that improvement is larger
among the matched peers.
Table C.2 formally tests differences in (log) revenue and EBITDA profitability margins
using the same parallel trends specification as in section 4.2.2 in the main text. The
table makes clear that targets acquired in a buyout grow much faster than the matched
control group. However, the table also emphasize the point in the descriptives: these
companies were already on a path of higher growth, though the buyout appears to
accelerate the pace of the growth. There is no significant difference in profitability
improvements between buyout targets and the matched control group. If anything,
companies acquired in buyouts experience a slight worsening of profitability after the
buyout relative to the control group, though this is not consistent over all models and
horizons and is never statistically significant.
These tables make clear that the matching procedure have not produced a perfect
match given the existence of pre-trends. It is important to note that the identification
strategy of the paper does not rely on the parallel trend assumption being met in this
difference-in-difference specification, as long as the trends are not different between buyout
targets experience large increases in CEO ownership and those that do not. It is also
worth pointing out that the pre-trend growth in revenue that appears to accelerate after
the buyout is similar to that observed in Boucly et al. (2011).

222
 Table C.1: Summary Statistics - Comparisons Buyout Targets versus Matched Controls

This table presents a comparison of descriptive statistics between companies in the sample and the set of matched control companies. Panel A compares statistics for
the most recent fiscal year prior to the buyout, referred to as year -1. Panel B presents the change in these accounting measures from year -2 until year -1. Panel C
presents changes from year -3 until year -1. The final column tests differences in means, and numbers followed by ***, **, or * indicate statistically significant means
at the 1%, 5%, and 10% levels, respectively.

Panel A: Company characteristics in year prior to buyout

Sample in Paper Full Sample
N Mean Median SD N Mean Median SD Mean diff.
Revenue (M£) 322 119.6 54.9 198.3 1190 116.4 54.7 135.6 3.24
EBITDA (M£) 322 15.2 7.8 21.3 1190 14.5 7.5 18.1 0.69
Total Assets (M£) 322 117.3 48.5 186.3 1189 139.1 59.5 183.6 −19.80
Number of Employees 317 1432 448 2445 1107 1233 402 1708 199.00
EBITDA Margin 322 0.160 0.142 0.103 1190 0.156 0.140 0.091 0.004
223

Panel B: 1-year pre-buyout growth

Sample in Paper Full Sample
N Mean Median SD N Mean Median SD Mean diff.
Revenue 302 0.178 0.125 0.261 1058 0.165 0.108 0.239 0.013
EBITDA 302 0.205 0.191 1.450 1055 0.226 0.146 2.738 −0.021
Total Assets 305 0.166 0.070 0.413 1076 0.154 0.106 0.222 0.012
Number of Employees 295 0.116 0.064 0.251 985 0.078 0.042 0.154 0.038∗
EBITDA Margin 302 0.010 0.009 0.055 1055 0.024 0.009 0.058 −0.014∗∗

Panel C: 2-years pre-buyout growth

Sample in Paper Full Sample
N Mean Median SD N Mean Median SD Mean diff.
Revenue 289 0.469 0.289 0.681 873 0.381 0.269 0.690 0.088
EBITDA 288 0.839 0.400 2.592 865 0.362 0.379 3.027 0.477∗
Total Assets 294 0.414 0.196 0.785 917 0.363 0.246 0.555 0.051
Number of Employees 283 0.294 0.158 0.610 819 0.212 0.118 0.367 0.082∗
EBITDA Margin 288 0.017 0.011 0.065 865 0.031 0.013 0.073 −0.014∗∗
 Table C.2: Operational Improvements of Buyout Targets versus Matched Controls

This table reports the estimates of a parallel trends test comparing companies acquired in a buyout
to that of matched peers. The procedure is similar to that outlined in section 4.2.2, excluding CEO
ownership variables. The dependent variable is (log) revenue in columns 1-3, and EBITDA margin in
columns 4-6. A separate coefficient is estimated for a PE dummy for each year t − s, as well as an
indicator variable for each year t − s that is omitted for brevity. The year t − 1 is used as a reference
point. The PE dummy takes the value 1 if a company was acquired in a buyout, and 0 if it is a matched
control company. All specifications include company fixed effects as well as year fixed effects. Industry
fixed effects are added in columns 2 and 5, and company controls interacted with a Post dummy are
dropped in columns 3 and 6. The Post dummy takes the value of 1 in the post-buyout period, and 0
otherwise. The company controls are measured in the fiscal year prior the buyout and are: (log) revenue;
revenue growth; EBITDA margin; and deal leverage. The pre-buyout period includes up to three years
prior to the buyout, while the post-period includes the year in which the buyout took place, and up to
four years after or until an exit occurs, whichever happens first. All continuous variables are winsorized at
a 1% level. Standard errors, clustered at the company level, are presented in parentheses below each
estimate. ***, **, and * denotes statistical significance at the 1%, 5%, and 10% levels, respectively.

(log) Revenue EBITDA/Revenue

t − 3 × PE −0.086∗∗∗ −0.092∗∗∗ −0.091∗∗ 0.004 0.002 0.002

(0.030) (0.030) (0.038) (0.007) (0.007) (0.007)
t − 2 × PE −0.014 −0.017 −0.023 0.009 0.008 0.008
(0.040) (0.040) (0.042) (0.008) (0.008) (0.008)

t × PE 0.223∗∗∗ 0.228∗∗∗ 0.205∗∗∗ −0.007 −0.001 −0.015

(0.066) (0.068) (0.076) (0.019) (0.018) (0.023)
t + 1 × PE 0.293∗∗∗ 0.298∗∗∗ 0.271∗∗∗ −0.001 0.005 −0.010
(0.058) (0.061) (0.064) (0.020) (0.019) (0.022)
t + 2 × PE 0.336∗∗∗ 0.340∗∗∗ 0.323∗∗∗ 0.000 0.006 −0.010
(0.064) (0.067) (0.070) (0.021) (0.021) (0.026)
t + 3 × PE 0.380∗∗∗ 0.384∗∗∗ 0.370∗∗∗ −0.008 −0.002 −0.015
(0.074) (0.076) (0.079) (0.019) (0.019) (0.025)
t + 4 × PE 0.374∗∗∗ 0.379∗∗∗ 0.368∗∗∗ −0.008 −0.001 −0.012
(0.087) (0.091) (0.082) (0.022) (0.021) (0.024)

Adjusted R2 0.927 0.927 0.931 0.546 0.550 0.556

N 7087 7087 7087 7068 7068 7068

Company FE Yes Yes Yes Yes Yes Yes

Year FE Yes Yes Yes Yes Yes Yes
Industry FE No Yes Yes No Yes Yes
Company Controls Yes Yes No Yes Yes No

224
D Variables in the data set

Table D.1: Variable Definitions

Variable name Description

Buyout Types
Distressed Buyout A transaction where the acquired company was in financial
distress prior to the buyout. It is defined as being acquired
from one of the following situations: acquired from bankruptcy
procedures (called “administration” in the UK); acquired in a
refinancing or restructuring deal where creditors are not fully
compensated; and one case where the acquiring private equity
firm was paid by the previous owner to take on the company and
its debt.

Divisional Buyout A transaction where the acquired company was a subsidiary or

division of a larger corporation prior to the buyout.

Private-to-Private Buyout A transaction where the acquired company was privately trading
prior to the buyout, with individuals controlling the company,
and is not classified as a distressed buyout.

Public-to-Private Buyout A transaction where the acquired company was publicly listed
prior to the buyout.

Secondary Buyout A transaction where the acquired company was a portfolio

company of another private equity firm prior to the transaction.
This category includes cases where the portfolio company is
passed between private equity ownership more than once.

Exit Routes
IPO An exit is classified as an IPO if the private equity owner exits
via an initial public offering. Usually, a private equity owner
do not sell all shares at the IPO, instead exiting over a period
of time following the IPO. The exit date is defined as the date
at which majority ownership is lost, whether that is the IPO
date or a later date. The exit route is classified as an IPO if the
stake is reduced below a majority stake while the company is
still publicly trading.

Lost Control A private equity owner is defined as having lost control of the
company if any of the following happens: creditors seize control
of the company; the company is placed in bankruptcy procedures
(called “administration” in the UK); or control passes to a new
owner in a financial restructuring of the company, where the
equity holders receive no compensation. These are events that
are generally not disclosed publicly, and the eventual outcome is
found through disclosures in Annual Reports and, in the case of
bankruptcy procedures, through reports submitted submitted to
Companies House by the court-appointed administrators.

(Continues)

225
 Table D.1 − Continued

Variable name Description

Other (Exit Route) There are five cases in the sample where the exit route cannot be
classified into either of the other five exit route categories. These
five cases involve the private equity firm maintaining ownership
of the portfolio company until the company has been sold in
pieces or liquidated by other means, and afterwards seize trading.

Trade Sale A sale is classified as a trade sale when the company is sold to a
non-investment firm or to individuals.

Sale to Financial Institution A sale is classified as a sale to a financial institution when the
company is sold to a investment firm. The vast majority of these
cases are sales to private equity funds but also includes sales
to pension funds, sovereign wealth funds and other investment
firms.

Still Owned A company is classified as still being owned if there are no

recorded transactions as of 2019-09-30, and the most recent
annual report or ownership list still indicates the private equity
firm as the ultimate majority owner.

Company Variables
CEO Ownership The fraction of common shares held by the CEO of the company.
This variable is compiled by hand from ownership filings (“Annual
Returns” or, more recently, “Confirmation Statements”) contain-
ing precise ownerships of individual shareholders. The variable is
measured at two points: in the last annual return submitted prior
to the buyout and the first annual return submitted following the
completion of the transaction. The measure ∆CEO Ownership
is defined as the difference in the two measures.

EBITDA (ROS) Earnings before Interest, Taxes, Depreciation and Amortisation

(EBITDA). EBITDA is adjusted for impairment charges, write-
offs and revaluations to tangible and intangible assets.

Replaced CEO An indicator variable taking the value 1 if the CEO pre-buyout
does not remain as CEO post-buyout, and 0 otherwise.

Retained CEO An indicator variable taking the value 1 if the CEO pre-buyout
remains as CEO post-buyout, and 0 otherwise.

226
References

Aboody, David and Ron Kasznik, “CEO stock option awards and the timing of
corporate voluntary disclosures,” Journal of Accounting and Economics, 2000, 29 (1),
73 – 100.
Abowd, John M., “Does Performance-Based Managerial Compensation Affect Corporate
Performance?,” ILR Review, 1990, 43 (3), 52–S–73–S.
Acharya, Viral V., Oliver F. Gottschalg, Moritz Hahn, and Conor Kehoe,
“Corporate Governance and Value Creation: Evidence from Private Equity,” The Review
of Financial Studies, 2013, 26 (2), 368–402.
Agrawal, Anup and Charles R. Knoeber, “Firm Performance and Mechanisms
to Control Agency Problems between Managers and Shareholders,” The Journal of
Financial and Quantitative Analysis, 1996, 31 (3), 377–397.
Antoni, Manfred, Ernst Maug, and Stefan Obernberger, “Private equity and
human capital risk,” Journal of Financial Economics, 2019, 133 (3), 634 – 657.
Bebchuk, Lucian Arye, Jesse M. Fried, and David I. Walker, “Managerial Power
and Rent Extraction in the Design of Executive Compensation,” The University of
Chicago Law Review, 2002, 69 (3), 751–846.
Bernstein, Shai and Albert Sheen, “The Operational Consequences of Private Equity
Buyouts: Evidence from the Restaurant Industry,” The Review of Financial Studies,
2016, 29 (9), 2387–2418.
, Josh Lerner, and Filippo Mezzanotti, “Private Equity and Financial Fragility
during the Crisis,” The Review of Financial Studies, 2019, 32 (4), 1309–1373.
Blanchflower, David G., Andrew J. Oswald, and Peter Sanfey, “Wages, Profits,
and Rent-Sharing,” The Quarterly Journal of Economics, 1996, 111 (1), 227–251.
Boucly, Quentin, David Sraer, and David Thesmar, “Growth LBOs,” Journal of
Financial Economics, 2011, 102 (2), 432 – 453.
Brav, Omer, “Access to Capital, Capital Structure, and the Funding of the Firm,” The
Journal of Finance, 2009, 64 (1), 263–308.
Brenner, Menachem, Rangarajan K Sundaram, and David Yermack, “Altering
the terms of executive stock options,” Journal of Financial Economics, 2000, 57 (1),
103 – 128.
Chung, Ji-Woong, “Leveraged Buyouts of Private Companies,” 2011. Working paper.
, Berk A. Sensoy, Léa Stern, and Michael S. Weisbach, “Pay for Performance
from Future Fund Flows: The Case of Private Equity,” The Review of Financial Studies,
2012, 25 (11), 3259–3304.
Cohn, Jonathan B., Edith S. Hotchkiss, and Erin M. Towery, “The Motives for
Private Equity Buyouts of Private Firms: Evidence from U.S. Corporate Tax Returns,”
2020. Working Paper.

227
 , Lillian F. Mills, and Erin M. Towery, “The evolution of capital structure and
operating performance after leveraged buyouts: Evidence from U.S. corporate tax
returns,” Journal of Financial Economics, 2014, 111 (2), 469 – 494.
Cornelli, Francesca and Oğuzhan Karakas, “CEO Turnover in LBOs: The Role of
Boards,” 2015. Working paper.
Cronqvist, Henrik and Rüdiger Fahlenbrach, “CEO contract design: How do strong
principals do it?,” Journal of Financial Economics, 2013, 108 (3), 659 – 674.
Davis, Steven J., John Haltiwanger, Kyle Handley, Ben Lipsius, Josh Lerner,
and Javier Miranda, “The Economic Effects of Private Equity Buyouts,” 2019.
NBER Working Paper 26371.
, , , Ron Jarmin, Josh Lerner, and Javier Miranda, “Private Equity, Jobs,
and Productivity,” American Economic Review, 2014, 104 (12), 3956–90.
Degeorge, Francois, Jens Martin, and Ludovic Phalippou, “On secondary buyouts,”
Journal of Financial Economics, 2016, 120 (1), 124 – 145.
Demsetz, Harold and Belén Villalonga, “Ownership structure and corporate
performance,” Journal of Corporate Finance, 2001, 7 (3), 209 – 233.
Eaton, Charlie, Sabrina T Howell, and Constantine Yannelis, “When Investor
Incentives and Consumer Interests Diverge: Private Equity in Higher Education,” The
Review of Financial Studies, 2019. Forthcoming.
Edgerton, Jesse, “Agency Problems in Public Firms: Evidence from Corporate Jets in
Leveraged Buyouts,” The Journal of Finance, 2012, 67 (6), 2187–2213.
Edmans, Alex, Xavier Gabaix, and Dirk Jenter, “Executive Compensation: A
Survey of Theory and Evidence,” in Benjamin Hermalin and Michael Weisbach, eds.,
The Handbook of the Economics of Corporate Governance, Amsterdam: Elsevier, 2017,
pp. 383–539.
Fama, Eugene F., “Agency Problems and the Theory of the Firm,” Journal of Political
Economy, 1980, 88 (2), 288–307.
and Michael C. Jensen, “Agency Problems and Residual Claims,” The Journal of
Law and Economics, 1983, 26 (2), 327–349.
and , “Separation of Ownership and Control,” The Journal of Law and Economics,
1983, 26 (2), 301–325.
Gabaix, Xavier and Augustin Landier, “Why has CEO Pay Increased So Much?,”
The Quarterly Journal of Economics, 2008, 123 (1), 49–100.
Gompers, Paul, Steven N. Kaplan, and Vladimir Mukharlyamov, “What do
private equity firms say they do?,” Journal of Financial Economics, 2016, 121 (3), 449
– 476.
Gormley, Todd A. and David A. Matsa, “Common Errors: How to (and Not to)
Control for Unobserved Heterogeneity,” The Review of Financial Studies, 08 2013, 27
(2), 617–661.
Grossman, Sanford J. and Oliver D. Hart, “An Analysis of the Principal-Agent
Problem,” Econometrica, 1983, 51 (1), 7–45.
Guo, Shourun, Edith S. Hotchkiss, and Weihong Song, “Do Buyouts (Still) Create
Value?,” The Journal of Finance, 2011, 66 (2), 479–517.
Holmström, Bengt, “Moral Hazard and Observability,” The Bell Journal of Economics,
1979, 10 (1), 74–91.
Humphery-Jenner, Mark, Ling Lei Lisic, Vikram Nanda, and Sabatino Dino
Silveri, “Executive overconfidence and compensation structure,” Journal of Financial
Economics, 2016, 119 (3), 533 – 558.

228
Jeng, Leslie and Josh Lerner, “Making Private Data Accessible in an Opaque Industry:
The Experience of the Private Capital Research Institute,” American Economic Review,
2016, 106 (5), 157–160.
Jenkinson, Tim, Stefan Morkoetter, and Thomas Wetzer, “Buy Low, Sell High?
Do Private Equity Fund Managers Have Market Timing Abilities?,” 2018. Working
Paper.
Jensen, Michael, “Eclipse of the Public Corporation,” Harvard Business Review, 1989,
67 (5), 61–74.
Jensen, Michael C. and Kevin J. Murphy, “Performance Pay and Top-Management
Incentives,” Journal of Political Economy, 1990, 98 (2), 225–264.
and William H. Meckling, “Theory of the firm: Managerial behavior, agency costs
and ownership structure,” Journal of Financial Economics, 1976, 3 (4), 305 – 360.
Kaplan, Steven, “The effects of management buyouts on operating performance and
value,” Journal of Financial Economics, 1989, 24 (2), 217 – 254.
Kaplan, Steven N. and Jeremy C. Stein, “The Evolution of Buyout Pricing and
Financial Structure in the 1980s*,” The Quarterly Journal of Economics, 1993, 108 (2),
313–357.
and Per Strömberg, “Financial Contracting Theory Meets the Real World: An
Empirical Analysis of Venture Capital Contracts,” The Review of Economic Studies,
2003, 70 (2), 281–315.
Leslie, Phillip and Paul Oyer, “Managerial incentives and value creation: Evidence
from private equity,” 2009. Working paper.
Lichtenberg, Frank R. and Donald Siegel, “The effects of leveraged buyouts on
productivity and related aspects of firm behavior,” Journal of Financial Economics,
1990, 27 (1), 165 – 194.
McConnell, John J. and Henri Servaes, “Additional evidence on equity ownership
and corporate value,” Journal of Financial Economics, 1990, 27 (2), 595 – 612.
Mehran, Hamid, “Executive compensation structure, ownership, and firm performance,”
Journal of Financial Economics, 1995, 38 (2), 163 – 184.
Metrick, Andrew and Ayako Yasuda, “The Economics of Private Equity Funds,”
Review of Financial Studies, 2010, 23 (6), 2303–2341.
Michaely, Roni and Michael R. Roberts, “Corporate Dividend Policies: Lessons
from Private Firms,” The Review of Financial Studies, 2012, 25 (3), 711–746.
Otto, Clemens A., “CEO optimism and incentive compensation,” Journal of Financial
Economics, 2014, 114 (2), 366 – 404.
Renneboog, Luc, Tomas Simons, and Mike Wright, “Why do public firms go
private in the UK? The impact of private equity investors, incentive realignment and
undervaluation,” Journal of Corporate Finance, 2007, 13 (4), 591 – 628. Private Equity,
Leveraged Buyouts and Corporate Governance.
Smith, Abbie J., “Corporate ownership structure and performance: The case of
management buyouts,” Journal of Financial Economics, 1990, 27 (1), 143 – 164.
Strömberg, Per, “The new demography of private equity,” in Lerner J Gurung A, ed.,
Globalization of Alternative Investments Working Papers Volume 1: Global Economic
Impact of Private Equity, (World Economic Forum USA, New York), 2008, pp. 3–26.
Yermack, David, “Good Timing: CEO Stock Option Awards and Company News
Announcements,” The Journal of Finance, 1997, 52 (2), 449–476.

229
6 | Conclusion

In this dissertation, I propose three empirical analyses on private equity, an industry

that has grown rapidly over the past forty years and is now a prominent ownership form.
The common idea of these studies is to examine private equity investments. Investments
in, and by, private equity funds differ substantially from investments in, and by, other
common fund structures such as mutual funds.
Chapter 1 of this thesis introduces the institutional setting and the broad field of
private equity within which these papers contribute. In Chapter 2, I review the extant state
of the literature. The chapter offers a detailed review on three specific areas: i) investing in
private equity funds and issues facing investors; ii) how private equity funds are structured,
and how that structure is related to potential agency conflicts between investors (limited
partners, or LPs) and fund managers (general partners, GPs); and iii) the economic impact
of private equity ownership on the portfolio companies that private equity funds invest in.
The three papers of the thesis contribute to each of these three areas. The first
paper, The accuracy of net asset values in private equity: Evidence from the secondary
market, examines pricing in the secondary market for private equity fund stakes, using a
confidential dataset provided by a London-based intermediary in the market. This market
offers investors early liquidation of their otherwise illiquid private equity fund investments.
The second paper explores the dynamics of pay-for-performance sensitivity in a GP’s
compensation that arises due to the carried interest provision. The paper links variation in
the pay-for-performance sensitivity to operational improvements at the portfolio companies
acquired by the private equity fund. The third paper focuses on the how operational
improvements are related to a change at the portfolio company induced by buyouts: a
change in a CEO’s ownership stake. In the following, I discuss the general implications
and limitations of each paper separately, and suggest directions for future work.
As investments of private equity funds are privately held, one of the characterising
features of private equity is a lack of market valuations and publicly available data.
Consequently, GPs provide regular appraisals of the fund value to LPs in the form of
net asset value (NAV) estimations. However, NAVs are not always accurate (e.g. Barber
and Yasuda, 2017; Brown et al., 2019). In the first paper of the thesis I document that

230
cross-sectional variation in discounts is informative of the underlying value of NAVs.
Funds that receive high (low) bids report high (low) returns for the quarter in which
the bid is placed, and bids predict fund performance in the subsequent years. These
findings have several important implications.
First, the secondary market has the potential to alleviate the illiquidity traditionally
associated with private equity investments. The secondary market remained marginal for
a long time due to restrictions on the transferability of private equity fund stakes (Lerner
and Schoar, 2004). The market has however grown rapidly since the beginning of the
2000s. If liquidating stakes become a viable option for LPs facing liquidity shocks, that
will lower the returns required to invest in the primary market (Bollen and Sensoy, 2016;
Maurin et al., 2020). This could widen the potential investor base and encourage further
growth in the market. The paper adds to our understanding of how this market functions.
Second, how much of a liquidity relief the secondary market offers depends on how
much of an haircut investors have to accept for transacting in it. A common interpretation
of discounts in the market is that they represent compensation to providers of liquidity
(e.g. Nadauld et al., 2019). The results of the paper suggest that a more nuanced view
is warranted, as discounts at least partly reflects rational expectations about future
fund performance. The model of Bollen and Sensoy (2016) directly takes the discount
as a measure of how well-functioning the secondary market is. To the extent that
reported NAVs are not always fair, reported discounts are not necessarily capturing
the true cost of transacting in the market.
Third, an interesting approach for evaluating private equity investments is taken by
Boyer et al. (2019) that develop a private equity index based on transactions in the
secondary market. Implicit in their approach is that secondary market transactions
reflect the market value of a fund stake. The results of my paper complement theirs
by providing supportive evidence of this assumption in the cross-section of discounts.
Additionally, results in my paper suggest a potential remedy to a problem of relying solely
on transactions: the lack of frequent transactions. Conditional on transacting at least
once, the average fund in their sample only transacts 2.7 times in the period 2006 to 2018.
Results in my paper suggest that bids are informative of underlying NAV values and
behave similarly to transaction values. If bids, gathered as a limit order book, can be used
to complement transaction prices, the frequency of data observations is greatly increased.
The lack of a large amount of transactions in my database is a limitation of my study.
Though I take steps to verify that these bids are serious, there will always be a concern that
they do not properly reflect transaction prices. It therefore appears as if is the best way
forward is to combine information inherent in bids with realized market valuations from
transactions. Constructing an index in this way seems like a fruitful area for future research.

231
 For the second and third papers I assemble a representative dataset of UK portfolio
companies acquired in a buyout. The dataset contains information on the private equity
exit, company-level financials in the pre- and post-buyout periods, and precise ownership
data. The UK offers a unique setting for studying portfolio companies controlled by private
equity firms. All companies in the UK are required to submit financials and ownership
information annually to Companies House, and these filings are publicly available.
The second paper, The dynamics of pay-for-performance sensitivity in private equity
funds, explores how the sensitivity of GPs’ pay to performance varies over a fund’s life.
GP compensation is linked to fund performance through the carried interest provision,
providing the GP with a fraction of fund profits conditional on beating a hurdle rate.
Consequently, a GP’s pay-for-performance sensitivity is increasing in fund performance,
and increasingly so as the fund matures. I estimate this sensitivity using an extended
version of the model framework developed by Metrick and Yasuda (2010).
I examine whether variation in pay-for-performance sensitivity is related to operational
improvements in portfolio companies. GPs often claim that they add value to their
portfolio companies, and several papers document operational improvements following
buyouts (e.g. Bernstein and Sheen, 2016; Boucly et al., 2011). Presumably, adding value
to portfolio companies takes effort on behalf of the GP, and agency theory predicts
that we should expect higher effort exerted when an agent’s pay is more sensitive to
performance that can be improved from exerting effort.
Exploiting the staggered nature of investments by private equity funds, I find that
within-fund variation in pay-for-performance sensitivity is positively related to growth in
portfolio companies acquired in buyouts. The effect is stronger when the importance of
the direct component of pay for performance is high relative to the indirect component of
pay for performance. To rule out possible confounding factors, public market movements
are utilized to generate exogenous variation in the measure. Private equity valuations are
positively impacted by public market movements (e.g. Ang et al., 2018), but a common
market shock has a heterogeneous impact on the pay-for-performance sensitivity of private
equity funds. This is primarily driven by the amount of capital invested at the time of
the shock, as well as the non-linear payoff structure of carried interest. The main finding
of the paper holds using only exogenous variation in pay-for-performance sensitivity. The
findings of the paper are important for several reasons.
First, a large part of compensation in private equity funds is varying with performance
(Chung et al., 2012; Metrick and Yasuda, 2010), which is seen by some as drivers of strong
returns in private equity funds (e.g. Jensen, 1989; Kaplan and Strömberg, 2009). However,
we have only limited evidence on how variation in pay-for-performance sensitivity relates
to performance in private equity funds.

232
 Robinson and Sensoy (2013) examine variation in fund terms and how it relates to
fund performance, finding little evidence of differential net-of-fees returns. For venture
capital funds they find no relationship, and for buyout funds they document a positive
relationship between the carried interest level and fund performance. However, only 3%
of buyout funds in their sample has a carried interest level different from the typical
20%, implying that their results are driven by a small subset of funds. Hüther et al.
(2020) show that, among venture capital funds, performance is higher among funds with
“GP-friendly” contract terms, where carry is paid on a deal-by-deal basis. In both of
these studies, it is difficult to attribute any results to an effect of contract terms. More
able GPs will in expectation perform better, and that performance might allow them to
extract better contract terms from investors, as in Berk and Green (2004). By contrast,
my paper utilizes the dynamics of pay-for-performance sensitivity to perform within-fund
estimations. I also utilize exogenous variation in pay-for-performance sensitivity generated
by public market movements to confirm the robustness of my findings.
Second, my findings have implications for the more general question of how variation in
pay-for-performance sensitivity in compensation to managers relates to performance. This
is a question that has received considerable attention in the CEO literature, though it is
difficult to establish any causal relationship (e.g. Edmans et al., 2017). Many problematic
features when studying CEO pay relate to the fact that setting compensation is an ongoing
process. The manager may influence the process (Bebchuk et al., 2002), engage in ex
post renegotiating which makes ex ante incentives unclear (Brenner et al., 2000), time the
delivery of news (Aboody and Kasznik, 2000; Yermack, 1997), and compensation may be
set in a way which exploits managerial optimism and overconfidence (Humphery-Jenner
et al., 2016; Otto, 2014). By contrast, in private equity funds, contract terms are fixed at
inception, which implies that many of these issues are not relevant. A similar argument is
made by Agarwal et al. (2009) in their study of hedge fund compensation. They provide
evidence that cross-sectional variation in pay-for-performance sensitivity among hedge
funds is positively related to fund performance. Similarly to Hüther et al. (2020) and
Robinson and Sensoy (2013), their results may reflect variation in managerial ability that
influences contract terms. Thus, to the extent that my empirical approach is convincing,
these results are relevant beyond the private equity field.
Third, by linking the pay-for-performance sensitivity to operational improvements
at the portfolio company I add to our understanding of when private equity firms
contribute to value creation in portfolio companies. However, this focus is one of the
major limitations of the study. GPs’ pay for performance is linked to fund returns, and
improving operations of portfolio companies is not the only way in which profits can
be generated. While deal-level cash flow data was not widely available when this study

233
was undertaken, Brown et al. (2020) provide a first look at a dataset that Burgiss is
assembling, containing deal-level data. This suggest an avenue for improving the study
of this paper by using deal-level performance data.
Another limitation of the study is its focus on the direct component of pay for
performance. As shown by Chung et al. (2012), the indirect component is at least as large
in magnitude. While I exploit variation in the relative importance of the direct versus
the indirect component to strengthen the interpretation of my results, I do not measure
it directly. As Chung et al. (2012) note, private equity appears ideally suited for such an
endeavour, as the repeated fundraising allows a quantification of the indirect component of
pay, something that is difficult in other settings. An interesting avenue for future research
is to jointly consider the role of direct and indirect pay for performance on performance.
It is likely that one can disentangle these effects in private equity, as hurdle rates are
fixed numbers while fundraising success is likely about performance relative to peers.
The third paper, Managerial Ownership and Operational Improvements in Buyouts,
investigates the relevance of CEO ownership stakes for operational improvements of
companies acquired in buyouts. I manually collect data on managerial ownership stakes
pre- and post-buyout from ownership filings, required to be submitted annually by all
companies in the UK. This data is combined with the portfolio company financials
assembled for the second paper of the dissertation.
I find that post-buyout CEO equity ownership stakes are positively associated with
profitability improvements. The results are stronger when the ownership stake increases,
and are driven entirely by the subsample in which the CEO is retained. This paper
contributes to our understanding of the impact of private equity ownership on portfolio
companies in multiple ways.
First, while many studies document the impact of private equity ownership on various
aspects of the operations of portfolio companies (e.g. Antoni et al., 2019; Bernstein and
Sheen, 2016; Boucly et al., 2011; Davis et al., 2014), we know relatively little about how
these changes come about. There is evidence that private equity can support growth by
relaxing credit constraints and injecting capital (e.g. Bernstein et al., 2019; Boucly et al.,
2011; Cohn et al., 2020), and two recent papers offer insights into what GPs say they
do. Biesinger et al. (2020) examine value creation plans prepared for individual deals.
They document a significant heterogeneity in strategies, but find that ex ante selection
of strategy is not related to deal returns. Instead they document that, somewhat less
surprisingly, returns are high in deals in which the GP claims they successfully executed
their strategies. Gompers et al. (2016) survey GPs and ask what they believe generate
value in their deals. The top three expected sources of value creation are “increase revenue

234
or improve demand”, “improve incentives”, and “follow-on acquisitions”. Clearly, these
sources are not mutually exclusive as, for example, follow-on acquisitions increase revenue.
My paper contributes to this literature by focusing on a salient change induced by
buyouts: the change in CEO ownership stakes. By relating changes in CEO ownership
stakes to subsequent operational improvements, I partially open up the black box of value
creation. If we interpret changes in a CEO’s ownership stake as changing incentives to
the CEO, the paper validates the belief of GPs documented in Gompers et al. (2016) that
improved incentives matter for value creation. The results are consistent with an effect
of incentives, especially as improved profitability is primarily found in companies with
a retained CEO that experiences an increase in the ownership stake. However, most of
the results are also consistent with an explanation of a CEO with private information
that bargain for higher shares. As the paper does not offer any evidence from exogenous
variation in ownership stakes, the findings should be cautiously interpreted.
Second, private equity firms have been viewed as strong principals, whose actions can
inform us of optimal governance policies (e.g. Cornelli and Karakas, 2015; Cronqvist and
Fahlenbrach, 2013; Kaplan and Strömberg, 2003). However, these papers typically do not
link variation in the changes made by private equity firms to subsequent improvements in
operational performance. Presumably, not all companies acquired in buyouts require the
same amount of change to approach optimality, as is the case for perks in Edgerton (2012).
My paper contributes to this literature by focusing on a specific action — changes in CEO
ownership stakes — and examining how variation in this action relates to subsequent
improvements in operational performance. A drawback of this focus is that I do not
account for the full range of changes implemented by private equity firms. It cannot
be ruled out that changes in CEOs’ ownership stakes are related to other, unobserved,
changes implemented, which in turn drive the performance. However, an alternative
explanation would have to be able to explain why this is only the case in the subsample
where the CEO is retained, and not when the CEO is replaced.
Finally, to the extent that buyouts are seen as a shock to the ownership stake held
by a CEO, the paper contributes to the wider question of how pay for performance of
CEOs relates to company performance. As noted above, it is unlikely that CEOs have
no bargaining power, so one should be careful in interpreting these results. Still, the
results provide evidence of the impact of a shock to ownership stake, and future research
may be able to exploit the private equity setting to generate exogenous variation in
the ownership stake awarded.
Three limitations of the study are worth discussing, as they provide an avenue for
future research. First, the CEO ownership stake is measured as the fraction of ordinary
shares held. However, due to the often significant amount of leverage imposed by private

235
equity, the position is akin to a barrier option. Higher performance not only leads to higher
payoffs, but also lowers the risk that the equity position is wiped out. Consequently, the
capital structure matters for the pay-for-performance sensitivity. Second, an additional
component of pay for performance is the likelihood of being fired. While Cornelli and
Karakas (2015) document that CEO turnover becomes less contingent on performance, it
is an important aspect that is omitted from the analysis. The omission is to allow an
ex ante measure, whereas a firing decision is an endogenous choice. Still, the analysis
could be improved by including an ex ante estimation of the risk of being fired due to
low performance. Third, CEOs typically cash out in a buyout, even as their ownership
stakes increase (Kaplan and Stein, 1993; Cronqvist and Fahlenbrach, 2013). It clearly
matters whether ownership stakes increase because CEOs invest a large fraction of their
wealth, or due to an active capital structure choice made by the private equity firm
to generate high-powered incentives. Improving on these limitations will further our
understanding of the role that managerial incentives play in improving operations of
portfolio companies in private equity.
Taken together, these three studies contribute to the growing body of work on private
equity investments and the impact of private equity ownership on companies acquired in
buyouts. Hopefully, they will open the door to further research on these topics.

236
References

Aboody, David and Ron Kasznik, “CEO stock option awards and the timing of
corporate voluntary disclosures,” Journal of Accounting and Economics, 2000, 29 (1),
73 – 100.
Agarwal, Vikas, Naveen D. Daniel, and Narayan Y. Naik, “Role of Managerial
Incentives and Discretion in Hedge Fund Performance,” The Journal of Finance, 2009,
64 (5), 2221–2256.
Ang, Andrew, Bingxu Chen, William N. Goetzmann, and Ludovic Phalippou,
“Estimating Private Equity Returns from Limited Partner Cash Flows,” The Journal
of Finance, 2018, 73 (4), 1751–1783.
Antoni, Manfred, Ernst Maug, and Stefan Obernberger, “Private equity and
human capital risk,” Journal of Financial Economics, 2019, 133 (3), 634 – 657.
Barber, Brad M and Ayako Yasuda, “Interim Fund Performance and Fundraising in
Private Equity,” Journal of Financial Economics, 2017, 124 (1), 172–194.
Bebchuk, Lucian Arye, Jesse M. Fried, and David I. Walker, “Managerial Power
and Rent Extraction in the Design of Executive Compensation,” The University of
Chicago Law Review, 2002, 69 (3), 751–846.
Berk, Jonathan B. and Richard C. Green, “Mutual Fund Flows and Performance
in Rational Markets,” Journal of Political Economy, 2004, 112 (6), 1269–1295.
Bernstein, Shai and Albert Sheen, “The Operational Consequences of Private Equity
Buyouts: Evidence from the Restaurant Industry,” The Review of Financial Studies,
2016, 29 (9), 2387–2418.
, Josh Lerner, and Filippo Mezzanotti, “Private Equity and Financial Fragility
during the Crisis,” The Review of Financial Studies, 2019, 32 (4), 1309–1373.
Biesinger, Markus, Çağatay Bircan, and Alexander Ljungqvist, “Value Creation
in Private Equity,” 2020. Working Paper.
Bollen, Nicolas P B and Berk A Sensoy, “How much for a haircut? Illiquidity,
secondary markets, and the value of private equity,” 2016. Fisher College of Business
Working Paper.
Boucly, Quentin, David Sraer, and David Thesmar, “Growth LBOs,” Journal of
Financial Economics, 2011, 102 (2), 432 – 453.
Boyer, Brian, Taylor D. Nadauld, Keith Vorkink, and Michael S. Weisbach,
“Private Equity Indices Based on Secondary Market Transactions,” 2019. Dice Center
WP 2018-21.
Brenner, Menachem, Rangarajan K Sundaram, and David Yermack, “Altering
the terms of executive stock options,” Journal of Financial Economics, 2000, 57 (1),
103 – 128.

237
Brown, Gregory, Robert S. Harris, Wendy Hu, Tim Jenkinson, Steven N.
Kaplan, and David Robinson, “Private equity portfolio companies: A first look at
Burgiss holdings data,” 2020. Working Paper.
Brown, Gregory W., Oleg R. Gredil, and Steven N. Kaplan, “Do private equity
funds manipulate reported returns?,” Journal of Financial Economics, 2019, 132 (2),
267 – 297.
Chung, Ji-Woong, Berk A. Sensoy, Léa Stern, and Michael S. Weisbach, “Pay
for Performance from Future Fund Flows: The Case of Private Equity,” The Review of
Financial Studies, 2012, 25 (11), 3259–3304.
Cohn, Jonathan B., Edith S. Hotchkiss, and Erin M. Towery, “The Motives for
Private Equity Buyouts of Private Firms: Evidence from U.S. Corporate Tax Returns,”
2020. Working Paper.
Cornelli, Francesca and Oğuzhan Karakas, “CEO Turnover in LBOs: The Role of
Boards,” 2015. Working paper.
Cronqvist, Henrik and Rüdiger Fahlenbrach, “CEO contract design: How do strong
principals do it?,” Journal of Financial Economics, 2013, 108 (3), 659 – 674.
Davis, Steven J., John Haltiwanger, Kyle Handley, Ron Jarmin, Josh Lerner,
and Javier Miranda, “Private Equity, Jobs, and Productivity,” American Economic
Review, 2014, 104 (12), 3956–90.
Edgerton, Jesse, “Agency Problems in Public Firms: Evidence from Corporate Jets in
Leveraged Buyouts,” The Journal of Finance, 2012, 67 (6), 2187–2213.
Edmans, Alex, Xavier Gabaix, and Dirk Jenter, “Executive Compensation: A
Survey of Theory and Evidence,” in Benjamin Hermalin and Michael Weisbach, eds.,
The Handbook of the Economics of Corporate Governance, Amsterdam: Elsevier, 2017,
pp. 383–539.
Gompers, Paul, Steven N. Kaplan, and Vladimir Mukharlyamov, “What do
private equity firms say they do?,” Journal of Financial Economics, 2016, 121 (3), 449
– 476.
Humphery-Jenner, Mark, Ling Lei Lisic, Vikram Nanda, and Sabatino Dino
Silveri, “Executive overconfidence and compensation structure,” Journal of Financial
Economics, 2016, 119 (3), 533 – 558.
Hüther, Niklas, David T. Robinson, Sönke Sievers, and Thomas Hartmann-
Wendels, “Paying for Performance in Private Equity: Evidence from Venture Capital
Partnerships,” Management Science, 2020, 66 (4), 1756–1782.
Jensen, Michael, “Eclipse of the Public Corporation,” Harvard Business Review, 1989,
67 (5), 61–74.
Kaplan, Steven N. and Jeremy C. Stein, “The Evolution of Buyout Pricing and
Financial Structure in the 1980s*,” The Quarterly Journal of Economics, 1993, 108 (2),
313–357.
and Per Strömberg, “Financial Contracting Theory Meets the Real World: An
Empirical Analysis of Venture Capital Contracts,” The Review of Economic Studies,
2003, 70 (2), 281–315.
and Per Strömberg, “Leveraged Buyouts and Private Equity,” Journal of Economic
Perspectives, 2009, 23 (1), 121–146.
Lerner, Josh and Antoinette Schoar, “The illiquidity puzzle: theory and evidence
from private equity,” Journal of Financial Economics, 2004, 72 (1), 3–40.
Maurin, Vincent, David T. Robinson, and Per Strömberg, “A Theory of Liquidity
in Private Equity,” 2020. Working paper.

238
Metrick, Andrew and Ayako Yasuda, “The Economics of Private Equity Funds,”
Review of Financial Studies, 2010, 23 (6), 2303–2341.
Nadauld, Taylor D., Berk A. Sensoy, Keith Vorkink, and Michael S. Weisbach,
“The liquidity cost of private equity investments: Evidence from secondary market
transactions,” Journal of Financial Economics, 2019, 132 (3), 158 – 181.
Otto, Clemens A., “CEO optimism and incentive compensation,” Journal of Financial
Economics, 2014, 114 (2), 366 – 404.
Robinson, David T. and Berk A. Sensoy, “Do Private Equity Fund Managers Earn
Their Fees? Compensation, Ownership, and Cash Flow Performance,” The Review of
Financial Studies, 2013, 26 (11), 2760–2797.
Yermack, David, “Good Timing: CEO Stock Option Awards and Company News
Announcements,” The Journal of Finance, 1997, 52 (2), 449–476.

239