AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES - AND THEIR RETURN TO ANNOUNCEMENTS OF ACQUISITIONS JULY 2012 BUSINESS AND SOCIAL SCIENCES, AARHUS UNIVERSITY AUTHOR: MATHIAS LETH NIELSEN (287766) LINE OF STUDY: MSC. FINANCE AND INTERNATIONAL BUSINESS ADVISOR: PALLE NIERHOFF DEPARTMENT: DEPARTMENT OF BUSINESS STUDIES
105
Embed
An Explorative Event Study of Listed Private Equity Vehicles
The last decade has seen a remarkable development within private equity. Private equity funds have become a key player in the market for corporate control and now accounts for a major share of the global M&A activity. However, private equity itself has changed. Listed private equity has emerged and approximately 350 private equity vehi-cles are now listed worldwide. Despite the relatively small number of vehicles, the de-velopment is striking since some of the largest and most renowned private equity vehi-cles, such as Blackstone and KKR, have chosen to go public. From an academic point of view, the emergence of listed private equity vehicles significantly expands the oppor-tunities for investigating an industry that is known for being notoriously private.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE
EQUITY VEHICLES
- AND THEIR RETURN TO ANNOUNCEMENTS OF ACQUISITIONS
JULY 2012
BUSINESS AND SOCIAL SCIENCES, AARHUS UNIVERSITY
AUTHOR: MATHIAS LETH NIELSEN (287766)
LINE OF STUDY: MSC. FINANCE AND INTERNATIONAL BUSINESS
ADVISOR: PALLE NIERHOFF
DEPARTMENT: DEPARTMENT OF BUSINESS STUDIES
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN I
EXECUTIVE SUMMARY
The last decade has seen a remarkable development within private equity. Private equity
funds have become a key player in the market for corporate control and now accounts
for a major share of the global M&A activity. However, private equity itself has
changed. Listed private equity has emerged and approximately 350 private equity vehi-
cles are now listed worldwide. Despite the relatively small number of vehicles, the de-
velopment is striking since some of the largest and most renowned private equity vehi-
cles, such as Blackstone and KKR, have chosen to go public. From an academic point
of view, the emergence of listed private equity vehicles significantly expands the oppor-
tunities for investigating an industry that is known for being notoriously private.
The aim of this thesis is to fill a gap in the understanding of private equity; namely how
listed private equity vehicles perform during acquisitions. Numerous researchers have
investigated how listed companies perform when they announce acquisitions, but de-
spite the fact that several authors have suggested that much of the value generation in
private equity is determined during the acquisition phase, no studies have yet investigat-
ed the abnormal return to announcements of acquisitions by listed private equity vehi-
cles. In addition to this, only a handful of studies have investigated the field of listed
private equity. This presents a unique opportunity to influence the research.
Based on an explorative review of the literature within M&A, private equity and listed
private equity, the thesis develops 22 hypotheses. Nine of these hypotheses are selected
for further analysis and tested in an event study, which consists of a battery of tests incl.
parametric, nonparametric and event-induced variance tests. The sample is based on
information from LPX Group and consists of 129 carefully selected deals conducted by
18 listed private equity vehicles in the period 2001-2012. The study finds an insignifi-
cant CAAR of 0.26% to the announcement of acquisitions by listed private equity vehi-
cles. In addition to this, the study finds that the announcement returns depend on the
structure and the experience of the listed private equity vehicle as well as on the period
in which the deal is conducted. These results are in line with former studies within
M&A and private equity.
The fact that listed private equity vehicles earn non-negative abnormal returns from the
announcement of acquisitions suggests that managers of listed private equity vehicles
have a shareholder wealth maximizing motive in acquisitions.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 13
4.2. The Diversity of Listed Private Equity Vehicles
LPEVs are more diverse than traditional PEVs and can be divided into four different
types as shown in exhibit 4.16.
Exhibit 4.1 – Four Types of Listed Private Equity Vehicles
Sources: Adapted from Lahr and Herschke (2009), Talmor and Vasvari (2012) and LPX Group (2012c)
Investment companies are managed by internal investment professionals and invest di-
rectly in portfolio companies. Thus, they offer diversification at portfolio company lev-
el, but not at fund level. Besides the fact that they are committed to the PE business
model, they look like ordinary holding companies.
LPE funds receive investment management from an external management company and
invest directly in portfolio companies. Thus, they offer diversification at portfolio com-
pany level. Investors in LPE funds essentially invest in the limited partnership stake.
Therefore LPE funds are very similar to traditional PE funds. However, LPE funds are
often allowed to invest in other assets.
LPE firms are essentially listed management firms. This implies that they are internally
managed and that they invest in the general partner’s funds. LPE firms therefore offer
6 Section 4.2 is inspired by Talmor and Vasvari (2012).
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 14
investors diversification at both fund and portfolio company level. In addition to this,
investors in LPE firms get a share of the management fees and the carried interests.
LPE funds of funds are externally managed and invest in PE funds. This implies that
LPE funds of funds act as limited partners in several PE funds. Investors are therefore
only indirectly exposed to PE, but in return they get diversification at both fund and
portfolio level. Since LPE funds of funds differ significantly from traditional PE firms,
they will be excluded from the remainder of the study.
The majority of the LPEVs are LPE funds and investment companies, whereas LPE
firms account for a minor share (LPX Group 2012d). However, LPE firms are relevant
to include in the study since they represent a significant proportion when measured by
their market capitalization7 (Talmor & Vasvari, 2012). Geographically, the LPEVs are
concentrated in the U.S. and Europe, which account for 94% of all LPEVs (LPX Group
2012b). Therefore, these will be the geographical focus areas of the study.
4.3. Comparison of Private Equity and Listed Private Equity
Listed private equity differs from unlisted private equity in a number of ways8. First of
all, shares in LPEVs are listed and hence more liquid. This makes it easy to trade shares
in LPEVs. A sale of shares in a PEV is difficult and time consuming since secondary
transactions between limited partners have to be approved by the general partner. Fur-
thermore, the transaction costs are high due to illiquidity. The only transaction cost in
LPE is the bid-ask spread (Bergmann et al. 2009; Talmor & Vasvari 2012).
Secondly, PE funds have high minimum investment requirements whereas there is no
minimum investment requirement in LPE. Hence LPE provides better access for retail
investors. The combination of no minimum investment and liquidity implies that it is
easier for investors to diversify their investments in LPE than in PE. Furthermore, PE
requires investors to have a fixed investment horizon of 7-10 years, whereas LPE allows
for a flexible investment horizon (Bergmann et al., 2009; Talmor & Vasvari, 2012).
Third, PE and LPE differ on a number of structural parameters. PEVs are often large,
have a limited life, return realized proceeds to investors and raise capital for each new
fund. LPEVs, on the other hand, are typically small, have an unlimited life, often retain
7 The Blackstone Group, Partners Group, Onex Corp., American Capital, and Intermediate Capital Corp. 8 See appendix 1 for a tabular overview of the differences
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 15
and reinvest realized proceeds, and have a fixed pool of capital (Brown & Kräussl 2010;
Jegadeesh et al. 2010; Goldman 2011). Therefore, LPEVs do not have to raise new
funds to the same extent as unlisted PEVs. Some scholars have argued that this causes
LPEVs to care less about their returns (Sloan & Benner 2008). However, investors in
LPE avoid the J-curve effect, which is predominant in PE. Thus, they gain immediate
exposure to PE and the underlying portfolio companies (Cumming et al. 2011; Talmor
& Vasvari 2012). However, LPEVs often invest in other assets than PE, while PEVs
typically stick to PE-related assets (Goldman 2011; Talmor & Vasvari 2012). Further
structural differences include that LPEVs do not offer co-investments to their investors
and trade at a discount relative to their net asset value (NAV) (Brown & Kräussl 2010).
Finally, LPE is easier and more transparent than PE, since LPEVs handle cash man-
agement, charge lower fees and allow for easy performance evaluation due to the fact
that prices are quoted (Brown & Kräussl 2010; LPEQ 2012; Talmor & Vasvari 2012)
The differences between LPEVs and PEVs complicate direct comparison. However,
Bergmann et al. (2009) finds that listed and unlisted PEVs behave similar with respect
to their risk and return patterns. This limits the problem of comparability.
To summarize, part 1 has given a number of valuable insights. Chapter 2 showed that
the market for corporate control can be explained by the neoclassical finance theory, the
agency theory and the behavioural finance theory, while Chapter 3 gave an introduction
to PE and explained how a PE fund works. Furthermore, it argued that the current trends
in PE are an increasing focus on operational improvements, specialization within specif-
ic industries or niches, and a movement towards listed private equity. Finally, chapter 4
showed that LPEVs are more diverse than PEVs and can be structured as: investment
companies, LPE funds, LPE firms or LPE funds of funds. The major differences be-
tween PE and LPE are that LPE is more liquid, provide better access for retail investors,
and is easier and more transparent. In addition to this, PE and LPE differ on a number of
structural parameters such as lifetime and size.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 16
5. VALUE GENERATION IN LISTED PRIVATE EQUITY
Having explained what LPE is, the next step is to outline how LPEVs generate value at
the portfolio company level. This will provide a solid basis for developing hypotheses
and explaining the empirical results.
5.1. A Three-Dimensional Framework of Value Generation
Berg and Gottschalg (2005) argue that value generation in PE depends on a number of
levers and can be analysed along three dimensions; phases, causes, and sources. The
phases of value generation can be divided into the acquisition phase, the holding phase,
and the divestment phase9. Value generation in the acquisition phase is determined by
the acquisition price. In addition to this, the initial business plan and the structure of the
buyout are determined. The value generation in the holding phase is determined by the
success of implementing and continuously adjusting the business plan. This includes
implementing the necessary strategic, operational, and organizational changes. The di-
vestment price determines the value generated in the divestment phase. In addition to
this, the mode of divestment is determined in this phase.
The second dimension is the causes of value generation. Essentially, value is generated
by increasing the equity value of the portfolio company. The equity value of a company
can be described by the following equation:
(5.1)
One can therefore distinguish between two causes of value generation; value capturing
and value creation. Value capturing is value generation caused by increasing the valua-
tion multiple while value creation is value generation caused by increasing revenues,
improving margins, or decreasing net debt. Value capturing occurs without changing the
underlying financial performance of the company and is therefore also known as finan-
cial arbitrage. It is achieved by selling a company at a higher valuation multiple than it
was bought at and is thus determined during the acquisition and divestment phase. Val-
ue creation occurs when the underlying financial performance of the company is im-
proved. This can be achieved through both direct and indirect means. Direct value crea-
tion arises through improved financial engineering, operational effectiveness and/or
strategic distinctiveness, i.e. things that directly impact the bottom line. Such levers are
9 The first two phases relate to the investment period in the PE fund lifecycle while the last phase relates to the har-
vesting period.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 17
called primary levers. The primary levers are influenced by secondary levers such as
agency costs, which only have indirect bottom line impact.
The third dimension is the sources of value generation, which can be either intrinsic or
extrinsic (Berg & Gottschalg 2005). Intrinsic value generation occurs without any form
of knowledge transfer from the LPEV to the portfolio company, while extrinsic value
generation occurs due to knowledge transfer from the LPEV.
5.2. Levers of Value Generation
The levers of value generation are usually split into three groups; financial improve-
ments, operational improvements, and governance improvements (Jensen et al. 2006;
Kaplan & Strömberg 2009).
5.2.1. Financial Improvements
Financial improvements can stem from financial arbitrage and financial engineering.
Financial arbitrage is widely recognized by practitioners, but has received surprisingly
little attention by academics (Berg & Gottschalg 2005; Loos 2005). It is also known as
“multiple riding” and concerns the value generated from selling at a higher valuation
multiple than the portfolio company was acquired at. Financial arbitrage can be based
on the following five factors; changing market valuation multiples, private information
about the portfolio company (MBO), superior market information (proprietary deal
flows and industry expertise), superior deal making capabilities, and conglomerate dis-
2011). Furthermore, PEVs raised in the 1980s have higher returns than PEVs raised in
the 1990s (Wood & Wright 2009). Finally, access to a proprietary deal flow and the
number of deals conducted has a positive impact on PE performance (Loos 2005).
Based on this, experienced LPEVs are expected to earn higher abnormal returns to the
announcement of acquisitions than inexperienced LPEVs.
H17: CAAR to experienced LPE acquirers > CAAR to inexperienced LPE acquirers
6.3.3.3. Investment Strategy
The investment strategy of a LPEV can either be specialization or diversification. Since
a LPEV can specialize within several areas, e.g. industry, size of the target, geography,
and buyout structure, and all of these areas are covered elsewhere, the investment strat-
16 See chapter 3 for more information about the fee structure in PE. 17 See chapter 5 for more information about how LPEVs generate value in acquisitions.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 34
egy will only be examined on an overall level. The reasoning behind the effect of the
investment strategy is that specialization should enable the LPEV to generate more val-
Feito-Ruiz & Menéndez-Requejo 2011), while PE studies are relatively conclusive.
Loos (2005) and Pe’er and Gottschalg (2011) find that geographical specialization
yields higher returns to PEVs, while industry specialization has a negative impact on
returns (Loos 2005). Furthermore, both Cressy et al. (2007) and Kaplan and Strömberg
(2009) find that specialization increases operational performance and profitability of
LPEVs’ targets. Hence, specialized LPEVs are expected to earn higher abnormal returns
in acquisitions than diversified LPEVs.
H18: CAAR to acquisitions by specialized LPEVs > CAAR to acquisitions by diversified
LPEVs
6.3.3.4. Size
The size of the LPEV at the time of the acquisition can be a proxy for skills and hence
have an impact on the abnormal returns (Phalippou & Gottschalg 2009). In addition to
this, Martynova and Renneboog (2011) argue that size can be a proxy for the risk of
management hubris. It is, however, unlikely that LPEVs are exposed to the risk of man-
agerial hubris since their managers essentially own of the management firm. The empir-
ical evidence of the effect of size is inconclusive. Lerner (2007) and Gottschalg et al.
(2010) find that smaller funds outperform larger funds. However, a number of studies
find that PE performance increases with fund size (until a certain point) and that larger
PEVs have better returns than smaller PEVs (Kaplan & Schoar 2005; Loos 2005;
Phalippou & Gottschalg 2009; Acharya et al. 2011). Consequently, the size of the
LPEV at the time of the announcement of an acquisition is expected to have an impact
on the abnormal return, although the direction of the impact is unclear.
H19: The size of the LPEV at the time of the acquisition has an effect on the CAAR
6.3.3.5. Geographical Origin
The geographical origin of the LPEV might have an impact on its announcement return
due to differences in terms of e.g. legal origin19
, industries, taxation, M&A activity or
PE maturity. Especially the PE maturity is interesting, as more experienced LPEVs are
18 See section 5.2.2 for more information about operational improvements. 19 Due to the similarities in the governance structures of LPEVs, the literature regarding the legal origin of acquirers
has not been reviewed.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 35
expected to earn higher abnormal returns. It is reasonable to expect that LPEVs in ma-
ture PE markets have been established earlier than LPEVs in immature PE markets.
Hence, we expect LPE acquirers in countries with mature PE markets, such as the U.S.
and the U.K., to earn higher CAARs than LPE acquirers in countries with less mature
PE markets, such as Continental European countries. The empirical evidence is fairly
inconclusive. Early studies find that U.S. PEVs outperform U.K. PEVs (Cumming &
Walz 2004) or that there are insignificant differences in returns to PEVs in different
regions (Loos 2005; Martynova & Renneboog 2006). More recent studies find that
Scandinavian acquirers earn CAARs which are significantly higher than that of acquir-
ers from other regions (Martynova & Renneboog 2011; Humphery-Jenner et al. 2012).
Based on the mixed empirical evidence, it is expected that the geographical location of
the LPEV has an effect on the abnormal return it earns in acquisitions, and that Scandi-
navian LPE acquirers earn higher CAARs than LPE acquirers from other regions.
H20: CAAR to Scandinavian LPEVs > CAAR to non-Scandinavian LPEVs
6.3.3.6. Management’s Background
As mentioned in chapter 4 one of the trends in PE is that PEVs are adding former ex-
ecutives and management consultants to their teams (Kehoe & Palter 2009; Bain & Co.
2012). Furthermore, both Loos (2005) and Acharya et al. (2011) suggest that the back-
ground of management has an impact on PE returns. The empirical evidence reveals
several interesting findings. First of all, Loos (2005) finds that investment managers
with a background in PE, banking, or corporate management perform well on an indi-
vidual basis, and that a higher share of PE and corporate management backgrounds in a
team also has a positive impact. Secondly, Acharya et al. (2011) find that partners with
operational backgrounds outperform partners with financial backgrounds in organic
deals and vice versa in inorganic deals. Thus, the background of management is ex-
pected to have an effect on the abnormal return to acquisitions by the LPEVs.
H21: The professional background of the management of a LPEV has an effect on the
CAAR it earns in acquisitions
6.3.3.7. Management’s Experience
Since the experience of the LPEV is expected to have a positive impact on abnormal
returns, the experience of the LPEV’s management is likely to have a similar impact.
More experienced managers have larger networks and more deal experience, but none
of these factors has an impact on PE returns (Loos 2005). In addition to this, Loos
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 36
(2005) finds that more experience20
in general leads to lower returns. Likewise, Acharya
et al. (2011) find no difference in PE returns between experienced and inexperienced
partners. Based on this, the experience of a LPEV’s management is not expected to have
an impact on the abnormal return it earns to the announcement of acquisitions.
H22: The experience of the LPEV’s management has no effect on the CAAR it earns in
acquisitions
6.4. Sub Conclusion
Summing up, this chapter has developed 22 hypotheses about the abnormal return to
announcement of acquisitions by LPEVs. Based on this, we expect that LPEVs generate
non-negative abnormal returns to their shareholders upon the announcement of acquisi-
tions, and that these returns depend on certain deal, target, and acquirer characteristics.
7. HYPOTHESES
Since it is outside the scope of the thesis to test all of the 22 hypotheses, this chapter
will prioritize them, in order to end up with a limited number of highly relevant hypoth-
eses. In addition to this, the measurement of the selected hypotheses will be discussed in
order to ensure their validity.
7.1. Presentation and Selection of Hypotheses
The hypotheses have been evaluated against three criteria. The first criterion is that the
hypothesis is relevant in an LPE perspective (C1). Secondly, previous studies need to
provide fairly conclusive results (C2). Third, it is a requirement that the hypotheses can
be tested based on the available dataset (C3). If a hypothesis satisfies all three require-
ments it is selected for further analysis (S). The evaluation can be seen from exhibit 7.1.
From the literature review it was evident that LPEVs avoid hostile bids and primarily
use cash as the means of payment. H7: Type of Bid and H8: Means of payment are there-
fore eliminated based on the first criterion. The second criterion is evaluated based on
the findings in the literature review. Previous empirical evidence provides fairly conclu-
sive results for 11 of the 20 remaining hypotheses. Thus, nine of the hypotheses are
eliminated based on the second criterion. Next, the 11 hypotheses are evaluated against
the third criterion, i.e. that they are testable based on the available dataset. The dataset is
described in detail in chapter 8, so for now we will only focus on the information con-
20 Measured by the average of age, tenure and PE experience.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 37
tained in the dataset. From the dataset information about the necessary inputs is availa-
ble for nine of the 11 hypotheses. Unfortunately, there is no information about the in-
puts for H15: Type of Firm and H22: Management’s Experience. Based on the three-step
procedure nine hypotheses are therefore selected for further research.
H22 Acquirer Management’s Experience The experience of the LPEV’s management
has no effect
Yes Yes No No
Notes: CAAR is the CAAR to LPE acquirers. “C1” means criterion 1 (the hypothesis is relevant in an LPE perspective), “C2” means criterion 2 (previous literature provides clear results), and “C3” means criterion 3 (the hypothesis is testable based on the
dataset). “S.” means that the hypothesis is selected for further analysis. *No studies have been conducted within the impact of
LPEV structure on the CAAR. Hypotheses in italics are the ones chosen for further analysis.
7.2. Specification of Hypotheses
In order to test the selected hypotheses, one needs to specify how they will be measured
and assess whether the measures are valid, i.e. whether they measure what they are sup-
posed to measure. To simplify the notation, the hypotheses have been renumbered as
H1-H9.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 38
H1: Overall CAAR is measured as the cumulative average abnormal return (CAAR),
which is the predominant measure of abnormal returns to announcement of acquisitions
in literature. CAAR is a measure of the announcement return and hence a measure of
the change in investors’ expectations regarding the LPEV’s future profits21
. As ex-
plained in section 6.2.3., the CAAR of LPEVs is not directly comparable to that of other
listed acquirers, mainly due to partial anticipation. Furthermore, for other listed acquir-
ers, the means of payment in an acquisition provides a signal about their view of their
stock price. Since LPEVs almost always pay with cash, and never with shares, their
CAARs are not biased by signalling. Based on the above, CAAR seems to measure
what it is supposed to measure; namely the change in investors’ expectations about the
LPEVs’ profits as a result of announcements of acquisitions.
H2: Deal Period is measured by the year in which the deal is announced. The years are
divided into three periods: 2001-2003, 2004-2008, and 2009-2012. The periods are di-
vided in this way due to the fact that there was a boom in PE buyouts from 2004 to 2008
(Talmor & Vasvari 2012)22
. The aim of this measure is to measure whether CAAR is
declining over time. However, some studies argue that the return depends on the eco-
nomic cycle. Thus, deal period is potentially not just a measure of time, but also of the
economic cycle. Fortunately, one will be able to get an idea about the impact of the eco-
nomic cycle due to the three-period division outlined above; 2004-2008 is a period of
economic boom, while the other two are periods with modest economic growth.
H3: Deal Size is measured as the ratio of deal value to the market value of the LPEV at
the announcement date. Usually the maximum of a PE fund’s committed capital that
can be invested in a single portfolio company is 15% (Talmor & Vasvari 2012). The
fund size is approximated by the market value (MV) of the LPEV. Therefore deals that
account for more than 10% of the LPEV’s MV are categorized as large deals, whereas
deals that account for 10% or less of the LPEV’s MV are categorized as small deals.
The MV of LPEVs is not a perfect approximation of the fund size, since a) LPEVs usu-
ally trade at a NAV discount (Phalippou 2010), b) sometimes only part of the underly-
ing PEV is listed (Cheffins & Armour 2007), and c) LPEVs can have investments in
several funds. Thus, the MV of a LPEV provides only a rough estimate of the fund size.
21 The measurement of CAAR is explained in detail in chapter 9. 22 See appendix 2 for an overview of the M&A activity from 2002 to 2012.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 39
Alternatively, the LPEV’s assets under management could have been used, but such
information was unfortunately not available from the dataset.
H4: Target Industry is measured by the target’s two-digit U.S. SIC code. The two-digit
U.S. SIC code provides the most overall industry classification. Based on this, the sam-
ple is divided into the following categories: “Mining & Construction”, “Manufactur-
ing”, “Transportation, Communications, Electric, Gas and Sanitary Services”, “Whole-
sale & Retail Trade”, “Finance, Insurance & Real Estate”, and “Services”23
. Since U.S.
SIC codes are a common measure for industry in the reviewed studies (e.g. Gou et al.
2011), and since they are provided by the U.S. government, the measure is evaluated to
have a satisfactory validity.
H5: Target Legal Origin is measured by the country of the target. Following La Porta et
al. (1998), Continental European countries are classified as civil law countries, while
the U.S. and the U.K. are classified as common law countries. The legal origin is a
measure of the corporate governance system in the country of the target – common law
countries generally have a stronger corporate governance system than civil law coun-
tries (La Porta et al. 1998, 2008). However, countries might differ on other parameters
than the corporate governance system, e.g. the major industries might differ. Hence, the
country of the target might potentially be a measure of more than just the corporate
governance system.
H6: Target Former Ownership is measured by whether the target was listed or unlisted.
Listed targets are classified as public, while unlisted targets are classified private. It is
therefore an unbiased measure of former ownership. However, former ownership is a
proxy for the ownership concentration (Vinten & Thomsen 2008) and, hence, a proxy
for the governance structure. Thus, one cannot say whether a potential effect is due to
the pre-deal ownership concentration or due to the pre-deal governance structure.
H7: LPEV Structure is measured by the PE category. LPEVs categorized as ‘Direct
private equity’ are classified as direct, whereas LPEVs categorized as ‘Private equity
fund managers’ are classified as indirect. Thus, the LPEV structure is a measure of the
degree of diversification offered by the LPEV (cf. section 4.2). It is noteworthy that
scholars are not consistent in their classification of the LPEVs; e.g. Apollo Investment
23 The reader might notice that some of the industry categories differ slightly from the ones provided by the US Gov-
ernment. The reason is that some of the categories have been merged in order to obtain a meaningful sample size.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 40
Corp. is classified as a LPE fund by Lahr and Herschke (2009), but as an investment
company by Talmor and Vasvari (2012). Therefore, the classification in this study is
based on LPX Group (2012c), which classifies the LPEVs based on a thorough and con-
tinuous review of their activities. This ensures a high validity of the measurement.
H8: LPEV Experience is measured by the date of incorporation of the vehicle underly-
ing the LPEV. These dates are provided by Zephyr. Based on Wood and Wright (2009),
vehicles incorporated before 1990 are classified as experienced, while vehicles incorpo-
rated from 1990 and onwards are classified as inexperienced. However, as older LPEVs
are more likely to have access to proprietary deal flows (Kaplan & Schoar 2005), the
date of incorporation is also a measure of the access to a proprietary deal flow. Alterna-
tively, the experience of the LPEVs could have been measured by the number of deals
they have conducted or by their age at the announcement date. However, the former
would be a biased measure of experience since larger LPEVs are likely to conduct more
acquisitions and the latter is complicated by wide dispersion of the age of the vehicles.
H9: LPEV Investment Strategy is measured by the industry focus. LPEVs are classified
as specialized if they focus on specific industries such as ‘IT’ or ‘Cleantech’. Otherwise
they are classified as diversified. The investment strategy is only measured on the in-
dustry dimension and not on other dimensions such as geography, deal size or the type
of firms. Thus, the industry focus is not a complete measure of the LPEV’s investment
strategy. It is, however, one of the most popular measures of investment strategy (see
e.g. Cressy et al. 2007). Hence, it has a satisfactory validity.
7.3. Sub Conclusion
To summarize, part 2 has investigated how LPEVs generate value in acquisitions and
reviewed the relevant literature within M&A, PE and LPE. On the basis of this, 22 hy-
potheses were developed, of which nine were selected for further analysis. The nine
hypotheses concern; 1) the overall abnormal return, 2) the deal period, 3) the deal size,
4) the industry of the target, 5) the legal origin of the target, 6) the former ownership of
the target, 7) the structure of the LPEV, 8) the experience of the LPEV and 9) the in-
vestment strategy of the LPEV. These hypotheses will be tested in part 3, which con-
sists of chapter 8-10. Chapter 8 will outline the sample selection and present descriptive
statistics, while chapter 9 will discuss the methodology used for testing the hypotheses.
Finally, chapter 10 will discuss the empirical findings.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 41
8. DATA AND SAMPLE
The sample selection process is a vital part of the study, since the quality of the data is a
key determinant of the validity of the results. This chapter will therefore outline the se-
lection process and present descriptive statistics of the sample.
8.1. Sample Selection
The LPEVs are identified based on the LPX Composite as of May 20th
2012 (LPX
Group 2012c)24
. Only LPEVs which are based in Europe or the U.S., are categorized as
either ‘Direct private equity’ or ‘Private equity fund manager’, and have an investment
style categorized as either ‘Buyout’ or ‘Growth’ are included in the sample. This gives a
list of 41 LPEVs. Information about their announcement of deals is obtained from
Zephyr, while information about security prices and market indexes is obtained from
Datastream. This procedure yields an initial sample of 495 deals. The use of the LPX
Composite and Zephyr might bias the sample towards a higher share of European deals,
since both data providers (the LPX Group and Bureau van Dijk) are from Europe. In-
stead one could have used U.S.-based sources such as the S&P’s Listed Private Equity
Index, VentureXpert and Dealogic. However, both the LPX Composite and Zephyr are
widely used in LPE studies (see e.g. Bilo et al. (2005), Bergmann et al. (2009), and
Müller and Vasconcelos (2010)). Thus, it is not expected to impose a significant bias.
The next step in the selection process is to impose a number of requirements25
which the
deals have to satisfy in order to guarantee a high quality of the data. The focus of the
study is European and U.S. deals. Hence, targets and LPEVs must be located in Europe
or the U.S. Besides this, one must ensure that the effect of the announcement is not di-
luted. Therefore, only completed deals where the rumour date is the same as the an-
nouncement date, and where the LPEV is stated as the primary acquirer, are included.
Furthermore, deals must have an ISIN number and a deal value of more than USD 1
million. Finally, daily returns need to be available from Datastream for the market index
during the entire period – this excludes deals announced earlier than December 31 2001.
To ensure that the sample provides the necessary inputs for the hypotheses, we only
include deals where the target has a U.S. SIC code and where the former ownership of
the target is known. Secondly, deals are excluded when the LPEV’s date of incorpora-
24 LPX’s requirements for including a LPEV are that a) minimum 50% of the net assets are invested in PE and b) that
the LPEV must be listed. In addition to this, a number of liquidity requirements must be satisfied (LPX Group 2011). 25 See exhibit A.3 in appendix 3 for an overview of the sample selection process.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 42
tion is unknown and when the vehicle was not listed at the time of the announcement.
Third, in order to ensure unbiased event windows, only one announcement is allowed in
the event window and the LPEVs are only allowed to invest in the target once.
To comply with other LPE studies, the liquidity requirements suggested by Bilo et al.
(2005) are imposed on the sample when they are relevant. Bilo et al. (2005) recommend
that LPEVs must have a minimum average trading volume26
of 0.1% per week. Howev-
er, this requirement is too strict as several of the LPEVs trade at small stock exchanges.
Instead, the trading volume requirement is based on Bartholdy et al. (2007), which re-
quire that a stock listed on a small stock exchange must be traded at least 80% of all
trading days. Following Bilo et al. (2005) only deals where the LPEV has an average
market value above USD 2 million and a relative bid-ask spread27
of maximum 20%
during the combined estimation and event period are included. Finally, the sample is
trimmed and the 1.25% most extreme observations in each tail are removed as recom-
mended by e.g. Campbell et al. (2010). Thus, the final sample contains 129 deals28
.
8.2. Descriptive Statistics
The 129 deals are conducted by 18 different LPEVs29
. The majority of the LPEVs are
European and organized as LPE funds. In addition to this, most of the LPEVs are
founded before 1990 and have a diversified investment strategy. Interestingly, 3i Group
PLC accounts for nearly 60% of the deals30
. Thus, the results in chapter 10 are reported
for two samples; the total sample and a sample excl. 3i Group. According to Kasper
Hansen (Associate Director from 3i Group interviewed May 16th
2012), 3i Group is a
LPE fund from 1973 with a diversified investment strategy. It is listed on the London
Stock Exchange and has the highest market value of the all the LPEVs. The average
market value of the LPEVs at the time of acquisition is USD 4.7 billion, but only USD
1.1 billion when 3i Group is excluded.
The targets in the sample are primarily private companies (96.7%) within services
(46.5%) or manufacturing (28.7%). These results are comparable to those found in e.g.
26
(Bilo et al. 2005).
27
(Bilo et al. 2005).
28 See appendix 4 for a list of the 129 deals. 29 See appendix 5 for an overview of the descriptive statistics of the sample. 30 Due to its age and size, 3i Group is dominating samples on LPE deals. Müller and Vasconcelos (2010) report that
3i Group accounts for 54% of the deals in their sample and hence choose to report separate results. The same is done
in this study.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 43
Gottschalg et al. (2010). 38.8% of the targets are from common law countries, while
61.2% are from civil law countries. The most common countries of origin for the targets
are the U.K., the U.S., France, Spain, and Germany.
The deals span from 2001 to 2012, with about two thirds being conducted between 2004
and 2008. It thus seems that the sampled deals are fairly representative of the overall
M&A market31
. Furthermore, most of the deals employ cash as the means of payment.
The average deal size is USD 55 million, which is fairly small compared to an average
deal size of USD 126 million for the M&A market in general in the same period. In ad-
dition to this, only 10.9% of the deals have a size corresponding to more than 10% of
the LPEV’s market value at the time of the announcement. Compared to former studies
such as Humphery-Jenner et al. (2012), the sample includes a quite high share of cross-
border deals (58.1%). The reason is that most of the LPEVs are European, and Europe-
an acquirers tend to have a high share of cross-border deals. Finally, all the deals are
outsider-driven. Thus, there is no bias from insider-driven deals in the sample.
9. METHODOLOGY
Having outlined the sample and how it was selected, this chapter will explain the event
study methodology and discuss the different test statistics and their performance.
9.1. Introduction
Event studies are used to measure the effect of an economic event on the value of a
company (Campbell et al. 1997). The event study methodology as we know it today
was developed by Fama, Fisher, Jensen and Roll (1969) and has only been slightly up-
dated since32
. It consists of a seven step procedure where one must; define the event,
select the sample, determine the measurement of abnormal return, outline a procedure
for estimating the abnormal return, outline a procedure for testing the hypotheses, pre-
sent the empirical results, and interpret them (Campbell et al. 1997).
9.2. Definition of the Event
In order to investigate the effect of an event, it is necessary to clearly define what the
event is. In this study the event of interest is the announcement of an acquisition by a
LPEV. In order to measure the effect of the event, one must define the event window.
31 For a comparison, see exhibit A.2 in appendix 2 and exhibit A.5.1 in appendix 5. 32 E.g. by Brown and Warner (1980, 1985).
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 44
The event window must be as narrow as possible in order to avoid distortion from e.g.
releases of other news. On the other hand, it must be wide enough to capture the effect
of the event, even if it is announced after trading closes. Therefore, the study will use an
event window of ± one trading day around the event (i.e. a 3-day event window) as rec-
ommended by e.g. Park (2004), Bartholdy et al. (2007), and Campbell et al. (2010). We
will define the event day as = 0, the first day of the event window as = T1 + 1, and
the last day of the event window as = T2, where is a measure of event time. The
length of the event window can then be defined as L2 = T2-T1 (Campbell et al. 1997).
One could have used more than one event window in order to analyse whether the ab-
normal return is captured (Campbell et al. 2010; Kolari & Pynnonen 2011). That is,
however, outside the scope of this thesis.
9.3. Estimation of Abnormal Return
To enable measurement of the effect of an event one needs to define the dimension
along which the effect is measured. This includes a number of choices. First, the meas-
urement of return needs to be decided upon. In general, we distinguish between measur-
ing returns as simple returns (9.1) and log returns (9.2);
(
) , (9.1)
, (9.2)
where rit and Rit are the simple return and the log return from holding security i from
period t-1 to period t. Pit and Pit-1 are the closing prices for security i at time t and time t-
1 respectively. Log returns are also known as continually compounded returns and have
several advantages (Campbell et al. 1997). One of the main advantages is that the multi-
period log return is simply the sum of the one-period log returns. In addition to this, log-
transformation increases the normality of the returns and eliminates negative values
(Henderson 1990). This is very important in our case for two reasons. First, much of the
event study methodology relies on the assumption of normal distributed returns33
. Se-
cond, Brown and Warner (1985) show that daily abnormal returns tend to be right
skewed. Therefore, log returns will be used going forward.
33 See section 9.4 for more information about the assumptions and appendix 7 for a test of the assumptions.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 45
Secondly, the abnormal return must be defined. The normal return is the return which
would be expected if the event did not occur. The abnormal return is the difference be-
tween the actual return and the expected normal return (MacKinlay 1997). It follows
that in order to estimate the abnormal returns one must estimate the expected returns.
This requires one to choose an estimation model. Here we distinguish between four dif-
ferent models: 1) the constant-mean-return model, 2) the market model, 3) factor mod-
els, and 4) the market-adjusted-return model. The constant-mean-return model assumes
that the expected return is constant through time whereas the market model assumes that
there is a linear relationship between market return and the return of security (MacKin-
lay 1997). The market model is defined as follows:
, (9.3)
where Rit and Rmt are the returns for period t for security i and the market index. εit is the
error term for security i for period t and is expected to be mean zero and have a variance
equal to . Compared to the constant-mean return model, the market model is better
since is removes the part of the return that is related to variation in the market’s return.
This decreases the variance of the abnormal returns (Campbell et al. 1997). The market
model is based on a single factor, namely the market return. Factor models include oth-
er factors than the market return as explanatory variables as well, such as exchange
rates. According to Campbell et al. (1997) the gains from employing multifactor models
are limited, since the marginal explanatory power is small. The market-adjusted-return
model is essentially a restricted form of the market model with αi=0 and βi=1. It is used
when we have no estimation period, or when we do not want to use the estimation peri-
od for estimating the expected returns (Campbell et al. 1997). However, Campbell et al.
(1997) argue that one should only use restricted models as a last resort. Based on this, as
well as the facts that the market model is better than the constant-mean-return model
and that the gains from employing multifactor models is limited, the market model is
chosen as the estimation model. Thus, we can define the expected return as:
| (9.4)
The abnormal return can therefore be defined as:
| (9.5)
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 46
Since the abnormal return is measured over an event window, which comprises several
days, we accumulate the returns and get a measure of the cumulative abnormal return
(CAR) for the individual security across time.
∑ (9.6)
When we investigate the effect of the announcement of acquisitions by LPEVs, we want
to investigate this effect not only across time, but also across securities. Thus, we calcu-
late the average of the CAR for the individual securities. This measure is called the cu-
mulative average abnormal return (CAAR) and is calculated as:
∑
(9.7)
Now, the only thing we lack in order to estimate the abnormal returns is a measure of
Rmt. The market return is measured by the return on a reference (market) index. If one
has securities from more than one country, as is the case for this study, one can choose
between global, regional, or national indices. Besides this, one must choose whether to
use value- or equal-weighted indices and whether to use local or global currency market
indices. Finally, one must choose which index provider to use. Several authors have
discussed which type of market indices to use in multi-country event studies (Park
2004; Campbell et al. 2010). Campbell et al. (2010, p. 3078) argue that “… local-
currency market-model abnormal returns using national market indexes are sufficient.”
In addition to this, value-weighted indices most appropriately reflect the total market
performance (Henderson 1990). For these reasons national MSCI local-currency, value-
weighted indices are used as reference indices for the market return.
In order to estimate the parameters of the market model we need to define an estimation
period. To avoid seasonality, an estimation period of 250 trading days is often recom-
mended since it approximately corresponds to the number of trading days in a calendar
year (see e.g. Campbell et al. (2010) and Corrado (2011)). Some authors argue that the
estimation period should end a number of days prior to the first day of the event win-
dow in order to avoid that information leaks prior to the event affects the estimation
period (Park 2004; Campbell et al. 2010; Corrado 2011). Due to the very private nature
of PE, information leaks prior to the announcement are not expected to occur, and there-
fore the estimation period will end the day before the first day of the event window.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 47
Formally, we can therefore define the estimation period as the period from = T0+1 to
= T1 with a length of L1 = T1-T0.
In order to ensure that the estimation period is unbiased, one normally needs to exclude
events, where the same type of event occurred during the estimation period. In the case
where announcement of events lead to abnormal returns, inclusion of events in the esti-
mation period imply that the estimation period is contaminated. This will bias the ex-
pected returns upwards and hence bias the abnormal returns in the event window
downwards. The idea behind event studies is that the event is unexpected and hence
surprises investors. Thereby it makes them incorporate the new information into their
information set and adjust the value of the stock. However, as explained in section 6.2.3
LPE acquisitions are partially anticipated and thus only lead to minor adjustments of the
stock price of the LPEV. Based on this, the inclusion of announcements of acquisitions
in the estimation period is not expected to impose any significantly bias on the estimat-
ed expected returns.
9.4. General Testing Procedure
Step number five in the event study is to specify the testing procedure. This includes
defining the null hypotheses, specifying the test statistics and evaluating their perfor-
mance. The hypotheses which we want to test (the alternative hypotheses) were speci-
fied in chapter 7, while the null hypotheses are still to be specified. Here it is important
to distinguish between the two types of tests we run. First of all, it is analysed whether
CAAR is different from zero, corresponding to the following null hypothesis:
H0x: CAAR = 0 (9.8)
Secondly, it is analysed whether the CAAR differs depending on certain deal, target,
and acquirer characteristics. This is analysed by dividing the sample into smaller sub-
samples and then testing whether the CAAR differs between these groups. Based on the
literature review we know, for most of the groups, which of the two groups that is ex-
pected to have the highest CAAR. Hence, the null hypothesis is the following;
H0y: CAAR to group 1 ≤ CAAR to group 2, (9.9)
for all the hypotheses except for the hypotheses about the deal period, the industry of
the target and the structure of the LPEV. In these cases the null hypothesis is
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 48
H0z: CAAR to group 1 = CAAR to group 2 (9.10)
Having defined the hypotheses, it is now time to discuss the test statistics. Inspired by
Bartholdy et al. (2007) a battery of nine test statistics is used for the analysis of whether
CAAR is different from zero. Each test statistic will briefly be discussed below34
. The
battery includes both parametric and nonparametric tests, as well as tests that correct for
event-induced variance. This is done in order to increase the robustness of the conclu-
sions (Campbell et al. 1997).
Parametric tests usually take some form of a t-test for differences in means and rely on
three assumptions: the abnormal returns must be normally distributed, have a constant
variance and be uncorrelated across securities35
. When these assumptions are fulfilled,
parametric tests have more power than nonparametric tests. Three parametric tests will
be presented below. They differ by the way the correct for problems inherent in the data
and the degree to which certain assumptions have to be fulfilled in order for them to
obtain a decent performance.
T1 relies on the assumption that the abnormal returns are independent across all securi-
ties in the sample. The test statistic divides the CAAR by its standard deviation, which
is derived from the variance of the abnormal returns of the individual securities during
the estimation period (Bartholdy et al. 2007).
T1 adjusted with adjusted cross-sectional independence applies the so-called Patell ad-
justment (Patell 1976). The reason is that the abnormal returns are forecasts from the
market model. Therefore, one needs to adjust the standard deviation from T1 for the
variance of the forecast error. This is what the Patell adjustment does. The adjustment
factor depends on the number of observed returns during the estimation period; the larg-
er the number of observed returns, the lower the adjustment (Bartholdy et al. 2007).
Since quite strict liquidity requirements were imposed during the sample selection, the
adjustment factor will be relatively small for our sample.
T2 is a t-statistic which standardizes the abnormal returns by scaling them with their
standard deviation (Bartholdy et al. 2007). Thus, it reduces the bias from outliers and
ensures that high abnormal returns will have less weight if the security had a high
standard deviation.
34 See appendix 6 for a specification of the test statistics used in the event study. 35 See appendix 7 for a test of the assumptions.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 49
T3 differs from T2 by including Patell’s adjustment. Thus, the standard deviation is
adjusted for the variance of the forecast errors. Besides this, it is equal to T2.
Nonparametric tests are used when the data is ordinal or when the assumption of nor-
mality is unsatisfied (Keller, 2005). Instead of analysing the difference in means, non-
parametric tests analyse whether the locations of the populations differ. The advantage
of nonparametric tests is that they are free of specific assumptions about the distribution
of the returns (Campbell et al. 1997)36
. This implies that they can be used to analyse
samples with quite few observations. Three nonparametric tests are used in this study.
T4 is a rank test. It converts the abnormal returns into a uniform distribution and assigns
a rank to each return. The rank is standardized, since we know that the security will not
be trading all trading days (Corrado & Zivney 1992). The expected rank (0.5) is then
subtracted from the rank of each security and the sum of these differences is divided by
the standard deviation of the ranks to get the test statistic (Bartholdy et al. 2007).
T5 is a sign test and works by converting the abnormal returns into nominal data. It re-
lies on the assumption that abnormal returns are independent across securities and that
the probability of observing either a positive or a negative abnormal return is 0.5 respec-
tively (Campbell et al. 1997). This implies that the sign test might be poorly specified
when the distribution of the abnormal returns is skewed, since the expected proportion
of positive abnormal returns in this case will be different from 0.5 (Campbell et al.
1997). If the probability of observing either a positive or a negative abnormal return is
0.5, then the expected sign of the abnormal return will be zero. The sign test therefore
analyses whether the average observed sign is different from zero.
T6 is a generalized sign test. Instead of assuming that the probability of observing either
a positive or negative abnormal return is 0.5 respectively, T6 estimates the probability
from the estimation period (Bartholdy et al. 2007). The test then compares the propor-
tion of positive abnormal returns during the event window with the proportion of posi-
tive abnormal returns during the estimation period (Renneboog et al. 2007).
Event-induced variance tests are employed in the case where the variance changes
around the event day. An increase in the variance will cause the expected standard devi-
ations based on the estimation period to underestimate the standard deviation in the
36 Therefore, nonparametric tests are also known as distribution-free tests (Keller 2005).
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 50
event window. Hence, the test statistics will be upward biased. Due to the fact that an
event gives investors new information, and investors react to new information cf. the
EMH, it is likely that the variance will increase around the event day. Therefore, two
event-induced variance tests are conducted.
T7 is a parametric test with adjusted variances and is also known as the BMP test37
. It
corrects for the problem of event-induced variance by using standardized abnormal re-
turns and a variance that is estimated from the event window rather than from the esti-
mation period (Boehmer et al. 1991).
T8 is a nonparametric rank test with adjusted variances. As opposed to T4, this test
standardizes the abnormal returns before they are ranked. For the abnormal returns in
the estimation period the test uses the same standardization procedure as T2, while it
uses standard deviations based on Patell’s adjustment for the standardization of the ab-
normal returns in the event window (Bartholdy et al. 2007). The abnormal returns on
the event date are then standardized by the standard deviation across all securities on the
event date. Afterwards, the ranking procedure from T4 is followed.
9.5. Testing Procedure for Tests of Differences
For the analysis of whether CAAR differs depending on certain deal, target and acquirer
characteristics, two types of tests are carried out. First, a t-test and a Wilcoxon rank sum
test are used to analyse whether there are differences in means and locations between
two groups (Keller 2005). The t-test is a parametric test and is adjusted whenever the
two groups have unequal variances. To shed light on this an F-test is employed. The
Wilcoxon rank sum test is a nonparametric test which analyses the differences in the
ranks between two groups (Keller 2005). Secondly, an ANOVA test and a Kruskal-
Wallis test are used to test for differences in means and locations between more than
two groups. The ANOVA test is a parametric test which simultaneously compares the
means of a number of groups (Keller 2005), while the Kruskal-Wallis test is a nonpara-
metric test which simultaneously compares the ranks of a number of groups.
In addition to the abovementioned tests, one could have made a multiple regression
analysis. However, the value added from conducting such an analysis would be limited
in our case. The interpretation would be complicated by the fact that most of the explan-
37 After Boehmer, Musumeci & Poulsen (1991)
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 51
atory power would be captured in the intercept since most of the variables are dummy
variables. Furthermore, some of the variables have a sample size as low as five. This
would imply that several of the underlying assumptions would be unsatisfied. For these
reasons, a multiple regression analysis has not been included in the test procedure38
.
9.6. Performance of Test Statistics
The performance of a test is measured by means of its size and power. The size of a test
is the probability of committing a type I error, i.e. rejecting the null hypothesis when it
is true. When the probability of committing a type I error is equal to the size of the test,
the test is well-specified (Kothari & Warner 2006). The power of a test is the probabil-
ity of finding abnormal returns when they are present (Kothari & Warner 2006), i.e. one
minus the probability of committing a type II error. Thus, the goal is well-specified tests
with high power.
In order to analyse the performance of the test statistics one could have conducted a
Monte Carlo simulation. This is however outside the scope of this thesis. Instead the
performance is evaluated based on former studies. In general, parametric tests have
higher power than nonparametric tests when their assumptions are fulfilled (Bartholdy
et al. 2007). However, nonparametric tests dominate parametric tests in terms of power
and size for multi-country studies with three-day event windows (Campbell et al. 2010).
Among the nonparametric tests, the rank test dominates the sign test in detecting small
abnormal returns (Corrado & Zivney 1992). Furthermore, it works well for small sam-
ples and has better performance than T7 (the BMP test) when the estimation period is
contaminated, i.e. includes other events (Aktas et al. 2007). In the case of event-induced
variance, T7 and T8 have higher power than the other tests, with T8 being the most
powerful in multi-country studies (Aktas et al. 2007; Campbell et al. 2010; Corrado
2011). The abnormal returns in our sample are approximately normally distributed39
,
but some of the subsamples suffer from small sample sizes40
. In addition to this, the
business model of LPEVs implies that the estimation periods are most likely contami-
nated41
. Finally, there are indications of event-induced variance in the data. In case of
doubt, emphasis will therefore be placed on the results of T8.
38 That being said, a multiple regression analysis would not have changed the results significantly. 39 See appendix 7 for a test of the assumptions. 40 See exhibit 10.1 in chapter 10. 41 See section 9.3 for a discussion of this.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 52
10. EMPIRICAL FINDINGS
Having explained the event study methodology, this chapter will present the empirical
findings and discuss the validity and the reliability of the results.
10.1. Discussion of the Results
The results are presented in exhibit 10.1. A separate exhibit with results for the sample
excl. 3i Group can be found in appendix 8. These results are stated in brackets in the
following sections.
10.1.1. Overall CAAR
The study finds an overall 3-day CAAR of 0.26% (-0.10%)42
. The CAAR is insignifi-
cant across all tests for both samples. This result implies that announcement of acquisi-
tions by LPEVs does not generate significant short run abnormal returns to their share-
holders. Consequently, H1: CAAR to the announcement of acquisitions by LPEVs ≥ 0 is
confirmed. This result is in line with the literature presented in chapter 6. The CAAR to
acquirers in general is insignificant or positive up to 1.2% (Andrade et al. 2001; Moeller
et al. 2005). Thus, the CAAR to LPEV acquirers does not seem to differ from that of
other acquirers. In addition to this, the result is supported by Jegadeesh et al. (2010)
who find that the market expects long-run abnormal returns of -2% to 2% for LPEVs.
As explained in chapter 6, the CAAR is likely to be affected positively by the facts that
PEVs generate more value in acquisitions than other acquirers (Achleitner et al. 2009,
2010) and are able to capture a larger share of the value generated (Bargeron et al.
2008; Officer et al. 2010; Stotz 2011). On the other hand, the magnitude of the CAAR
is likely to be negatively affected by partial anticipation from investors (Schipper &
Notes: CAAR is defined as the sum of CAR [-1; 1] divided by N. The significance level is indicated by a = 10%, b = 5%, and c = 1%. * implies that the tests for differences are the ANOVA test and the Kruskal-Wallis test. W-test = the Wilcoxon Rank Sum test. The test
for differences which is relied upon is marked with bold; the t-test is relied upon when the observations of both groups are normally
distributed according to a Jarque-Bera test; when they are not, the Wilcoxon Rank Sum test is relied upon. The industries are as follows: ‘Min. & Con.” = Mining & Construction, “T., C., E., G. & S. S” = Transportation, Communication, Electric, Gas &
Sanitary Services, “W. & R. Trade” = Wholesale & Retail Trade, and “F., I. & R. E.” = Finance, Insurance & Real Estate”.
10.1.2. Deal Period
The study shows that deals conducted in the period 2001-2003 had a CAAR of 1.70%
(2.56%). For the total sample, this is statistically significant at a 10% significance level
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 54
according to all tests, but none of the tests show significance when 3i Group is exclud-
ed. However, the latter sample only consists of two deals. Deals conducted in the peri-
ods 2004-2008 and 2009-2012 had statistically insignificant CAARs of 0.00% (-0.29%)
and 0.02% (-0.04%) respectively. Comparing the three periods, 2001-2003 yields a sig-
nificantly higher CAAR (approx. 1.70%) than the other periods for the total sample, but
an insignificantly higher CAAR when 3i Group is excluded. This is most likely due to
the small sample size in the period 2001-2003 when 3i Group is excluded. The CAAR
decreases for both 2004-2008 and 2009-2012 when 3i Group is excluded, while it in-
creases for 2001-2003. This suggests that the difference between 2001-2003 and the
other periods increases after removing 3i Group. Based on the discussion above, the
CAAR can therefore be concluded to differ between 2001-2003 and the other periods.
Hence, H2: CAAR to LPE acquirers in 2001-2003 > CAAR to LPE acquirers in 2004-
2008 > CAAR to LPE acquirers in 2009-2012 is partly confirmed. This result is sup-
ported by previous empirical studies which find that abnormal returns have declined
over time in both M&A and PE (Martynova & Renneboog 2008; Acharya et al. 2011).
The decline in the CAAR from the first period to the latter periods can be explained by
three things. First, financial markets have most likely become more efficient in their
valuation of the LPEVs, since an increasing number of LPEVs have emerged along with
LPE indices - both of which have increased the transparency of performance. Secondly,
the increasing number of PEVs and LPEVs during the period has increased the competi-
tion in the market for corporate control (Wright et al. 2006; Bain & Co. 2012). Third,
premiums and hence prices have been driven up (Gou et al. 2011). The fact that CAAR
has not declined from the period 2004-2008 to the period 2009-2012 indicates that the
business cycle effect counteracts the overall decline in returns. If CAARs decline over
time, but LPEVs do better during busts (2009-2012) than during booms (2004-2008) as
suggested by (Kaplan & Schoar 2005; Achleitner et al. 2009, 2010), then one can ex-
pect approximately the same CAARs during a boom as during the following bust.
10.1.3. Deal Size
The CAAR to large deals is 0.58% (0.58%) while the CAAR to small deals is 0.22%
(-0.34%). The results are not significantly different from zero for any of the samples.
The difference in the CAAR between large and small deals is insignificant for the over-
all sample. Since 3i Group has a positive effect on CAAR and only conducts small
deals, the difference increases after removing 3i Group. Thus we find a difference that is
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 55
significant at the 10% level for the sample excl. 3i Group. It is, however, not enough to
conclude a difference across both samples and hence H3: CAAR to large acquisitions by
LPEVs > CAAR to small acquisitions by LPEVs is rejected. This result is supported by
Acharya et al. (2009) who find that value generation is independent of the deal size. It
does, however, contradict most of the other previous studies which suggest that larger
deals should lead to higher returns (Loos 2005; Wright et al. 2006; Martynova et al.
2006; Bargeron et al. 2008; Officer et al. 2010). A potential explanation is that the deal
size is measured on a relative basis in this study, while several of the other studies have
measured deal size on an absolute basis. As explained in chapter 6, one would expect
better performance of large deals, since larger targets e.g. can bear more debt (Achleit-
ner et al. 2009) and have more dispersed ownership (Faccio & Lang 2002). However,
the LPEVs are likely to have larger bargaining power when the targets are smaller. This
counteracts the positive effects of larger targets and can therefore explain why the deal
size does not have an effect on the CAAR.
10.1.4. The Industry of the Target
The study finds CAARs ranging from -0.08% to 1.59% (-0.60% to 1.20%) depending
on the industry of the target. The results are insignificantly different from zero across all
industries in the samples. Furthermore, the CAARs are not significantly different be-
tween the industries. Therefore, one can reject H4: The CAAR to LPE acquirers depends
on the industry of the target. This contradicts previous studies which find that PE re-
turns are industry dependent (Loos 2005; Cumming et al. 2007; Gottschalg et al. 2010).
There might be several reasons for these results. First of all, four of the six industry cat-
egories have sample sizes of 13 (6) or less. Thus, the results should be interpreted with
caution. Secondly, M&As and PE activity tend to cluster in different industries in dif-
ferent periods (Andrade et al. 2001; Vinten & Thomsen 2008). This is likely to drive up
prices and hence drive down returns. Third, due to the low sample sizes, high level in-
dustry categories have been used. Thus, one might find different results if a more de-
tailed industry categorization is applied.
10.1.5. The Legal Origin of the Target
The CAAR to deals involving civil law targets is 0.29% (0.03%) while the CAAR to
deals involving common law targets is 0.22% (-0.73%). None of these results are signif-
icantly different from zero, although T6 and T7 report significance at the 10% signifi-
cance level for common law targets in the sample excl. 3i Group. The CAAR does not
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 56
differ significantly between civil and common law targets for any of the samples. Thus,
H5: CAAR to acquisitions of targets from civil law countries by LPEVs > CAAR to ac-
quisitions of targets from civil law countries by LPEVs is rejected. This contradicts the
existing (but scarce) empirical evidence which finds that CAAR depends on the legal
origin of the target, although they disagree about which legal origin that is preferable
Easy to diversify investments due to liquidity and
no minimum required investment
Hard to diversify due to limited liquidity and high
minimum required investment
Unlimited life (evergreen) Limited life
Capital is permanent Capital is raised for each new fund
Realization proceeds are often retained and rein-
vested
Realization proceeds are returned to investors
Mainly structured as LPE funds, but quite diverse Mainly structured as LPE funds
Investors buy shares at a NAV discount Investors buy shares at the asset value
Invest in other assets than private equity Do not invest in non-private equity assets
Allow for a flexible investment horizon Fixed investment horizon of 7-10 years
No J-curve effect J-curve effect
Transparent and decent disclosure Limited disclosure and transparency
Handle cash management Cash management is handled by the limited part-
nership
Lower fees 2% management fee and 20% carried interest
Do not offer co-investment opportunities Offer co-investment opportunities
Small Large
Note: *NAV = Net Asset Value
Sources: Bauer et al. 2001; Bilo et al. 2005; Bergmann et al. 2009; Lahr & Herschke 2009; Brown & Kräussl 2010; Jegadeesh,
Kräussl & Pollet 2010; Cumming et al. 2011; Goldman 2011; LPEQ 2012; LPX Group 2012a; Talmor & Vasvari 2012.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 74
Appendix 2 – Overview of M&A Activity, 2002-2012
Exhibit A.2 – M&A Activity, 2002-2012
Notes: The data includes all deals completed from 2002 until 20th of May 2012, where both targets and acquirers are located in Europe or the US.
Source: Zephyr
-
4.000
8.000
12.000
16.000
20.000
24.000
28.000
-
500.000
1.000.000
1.500.000
2.000.000
2.500.000
3.000.000
3.500.000
No
. o
f d
eals
Ag
gre
gate
d d
eal
valu
e (
US
D m
n.)
Year
No. of deals Aggregated deal value (USD mn.)
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 75
Appendix 3 – Sample Selection Process
Exhibit A.3 – Sample Selection Process
No. Criterion Deleted
deals
Remaining
deals
1 Initial number of deals from Zephyr - 495
2 The LPEV is on the list from LPX 64 431
3 Acquirer and target are from Europe or the U.S. 4 427
4 The deal is completed - 427
5 Rumour date equals announcement date 37 390
6 The deal has an ISIN number - 390
7 The deal has a value of more than $1 million 9 381
8 The former ownership of the target is known 88 293
9 The target has a U.S. SIC code - 293
10 The acquirer’s date of incorporation is known - 293
11 The acquirer’s stock price is available 8 285
12 There is only one announcement in the event window 50 235
13 The trade volume of the acquirer’s stock is available 5 230
14 The acquirer’s stock trades minimum 80% of all trading days 13 217
15 The average market value of the acquirer is above $2 million 0 217
16 The bid-ask spread of the acquirer’s stock is maximum 20% 0 217
17 Daily returns are available for the market index 57 160
18 The acquirer only invests once in the target 27 133
19 Trimming (removal of the 1.25% most positive and the 1.25%
most negative observations)
4 129
20 Final number of deals in the sample - 129
Sources: LPX Group (2012c), Zephyr, Datastream and own analysis
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 76
Appendix 4 – List of Deals
Exhibit A.4 – List of Deals
Acquirer name Acquirer
country
code
Target name Target
country
code
3I GROUP PLC GB CONSILIUM TECHNOLOGIES LTD GB
3I GROUP PLC GB LAMY SA FR
3I GROUP PLC GB PETROFAC LTD GB
3I GROUP PLC GB PROVIMAR SA ES
3I GROUP PLC GB PIXOLOGY GB
3I GROUP PLC GB ASPECTS SOFTWARE LTD GB
3I GROUP PLC GB LA CHEMIAL SPA IT
3I GROUP PLC GB DISPLAY PRODUCTS TECHNOLOGY
LTD
GB
3I GROUP PLC GB RICH XIBERTA SA ES
3I GROUP PLC GB PROSOL GESTION SA FR
3I GROUP PLC GB EPIGENOMICS AG DE
3I GROUP PLC GB TOP LAYER NETWORKS INC. US
3I GROUP PLC GB JDH HOLDINGS LTD GB
3I GROUP PLC GB JUTEL OY FI
3I GROUP PLC GB TRANSMOL LOGÍSTICA SL ES
3I GROUP PLC GB HUNTSWOOD PLC GB
3I GROUP PLC GB NEUROTECH SA FR
3I GROUP PLC GB LDV LTD GB
3I GROUP PLC GB VANYERA 3 SL ES
3I GROUP PLC GB REPUBLIC LTD GB
3I GROUP PLC GB WILLIAMS LEA GROUP LTD GB
3I GROUP PLC GB MACTIVE AB SE
3I GROUP PLC GB ROLLER STAR SA ES
3I GROUP PLC GB MICROSULIS LTD GB
3I GROUP PLC GB MICROEMISSIVE DISPLAYS LTD GB
3I GROUP PLC GB VIBRATION TECHNOLOGY LTD GB
3I GROUP PLC GB NOVEM CAR INTERIOR DESIGN
GMBH
DE
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 77
3I GROUP PLC GB ARCCURE TECHNOLOGIES GMBH DE
3I GROUP PLC GB STI SPA IT
3I GROUP PLC GB INFINITE DATA STORAGE LTD GB
3I GROUP PLC GB GRIES-DECO-COMPANY GMBH DE
3I GROUP PLC GB PHARMETRICS INC. US
3I GROUP PLC GB NOVEXEL SA FR
3I GROUP PLC GB INCLINE GLOBAL TECHNOLOGY
SERVICES LTD
GB
3I GROUP PLC GB VONAGE HOLDINGS CORPORATION US
3I GROUP PLC GB DAALDEROP BV NL
3I GROUP PLC GB BRAINSHARK INC. US
3I GROUP PLC GB FIOS INC. US
3I GROUP PLC GB SALAMANDER ENERGY (THAI-
LAND) LTD
GB
3I GROUP PLC GB TRANSPORTS ALLOIN SAS FR
3I GROUP PLC GB SCREENTONIC SA FR
3I GROUP PLC GB COMBINATURE BIOPHARM AG DE
3I GROUP PLC GB MERIDEA FINANCIAL SOFTWARE
OY
FI
3I GROUP PLC GB CHRONICLE SOLUTIONS (UK) LIM-
ITED
GB
3I GROUP PLC GB AMBERWAVE SYSTEMS CORPORA-
TION
US
3I GROUP PLC GB HTC SWEDEN AB SE
3I GROUP PLC GB SONIM TECHNOLOGIES INC. US
3I GROUP PLC GB FOTOLOG INC. US
3I GROUP PLC GB GIRAFFE CONCEPTS LTD GB
3I GROUP PLC GB NETRONOME SYSTEMS INC. US
3I GROUP PLC GB AZELIS SA LU
3I GROUP PLC GB CERENICIMO SAS FR
3I GROUP PLC GB KINETO WIRELESS INC. US
3I GROUP PLC GB VETTE CORPORATION US
3I GROUP PLC GB KNEIP COMMUNICATION SA LU
3I GROUP PLC GB ADVANCED POWER AG CH
3I GROUP PLC GB AOPTIX TECHNOLOGIES INC. US
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 78
3I GROUP PLC GB FINNET OY FI
3I GROUP PLC GB INTALIO INC. US
3I GROUP PLC GB METASTORM INC. US
3I GROUP PLC GB CAIR LGL SA FR
3I GROUP PLC GB CAMBRIDGE SEMICONDUCTOR LTD GB
3I GROUP PLC GB UNIÓN RADIO SA ES
3I GROUP PLC GB TSMARINE (CONTRACTING) LTD GB
3I GROUP PLC GB VELOCIX LTD GB
3I GROUP PLC GB DATANOMIC LTD GB
3I GROUP PLC GB COREMETRICS INC. US
3I GROUP PLC GB LABCO SAS FR
3I GROUP PLC GB WELLPARTNER INC. US
3I GROUP PLC GB ENOCEAN GMBH DE
3I GROUP PLC GB NUJIRA LTD GB
3I GROUP PLC GB GARLIK LTD GB
3I GROUP PLC GB VNU MEDIA BV NL
3I GROUP PLC GB REFRESCO HOLDING BV NL
3I GROUP PLC GB STORK MATERIALS TECHNOLOGY
BV
NL
3I GROUP PLC GB GO OUTDOORS LTD GB
ALTAMIR ET COMPAGNIE SCA FR FRANCE TÉLÉCOM MOBILE SATEL-
LITE COMMUNICATIONS SA
FR
ARQUES INDUSTRIES AG DE ACTEBIS COMPUTERS BV NL
ARQUES INDUSTRIES AG DE OXXYNOVA GMBH & CO. KG DE
BURE EQUITY AB SE RUSHRAIL AB SE
BURE EQUITY AB SE VITTRA AB SE
BURE EQUITY AB SE CYGATE AB SE
BURE EQUITY AB SE CARL BRO A/S DK
CANDOVER INVESTMENTS PLC GB THULE AB SE
CANDOVER INVESTMENTS PLC GB EXTRAPRISE GROUP INC. US
CAPMAN OYJ FI SCAN·JOUR A/S DK
CAPMAN OYJ FI NEOVENTA MEDICAL SE
CAPMAN OYJ FI VARESVUO PARTNERS OY FI
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 79
CAPMAN OYJ FI SILEX MICROSYSTEMS AB SE
DEA CAPITAL SPA IT FIRST ATLANTIC REAL ESTATE
HOLDING SPA
IT
DEA CAPITAL SPA IT LANIFICIO LUIGI BOTTO SPA IT
DEUTSCHE BETEILIGUNGS AG DE CLYDE BERGEMANN GMBH MAS-
CHINEN- UND APPARATEBAU
DE
DINAMIA CAPITAL PRIVADO SCR
SA
ES BESTIN SUPPLY CHAIN SL ES
DINAMIA CAPITAL PRIVADO SCR
SA
ES SOCIEDAD GESTORA DE TELE-
VISIÓN NET TV SA
ES
DINAMIA CAPITAL PRIVADO SCR
SA
ES SAFE 2000 SL ES
DINAMIA CAPITAL PRIVADO SCR
SA
ES ÉMFASIS BILLING & MARKETING
SERVICES SL
ES
DINAMIA CAPITAL PRIVADO SCR
SA
ES SAINT GERMAIN GRUPO DE INVER-
SIONES SL
ES
DINAMIA CAPITAL PRIVADO SCR
SA
ES ESTACIONAMIENTOS Y SERVICIOS
SA
ES
DINAMIA CAPITAL PRIVADO SCR
SA
ES SERVICIO DE VENTA AUTOMÁTICA
SA
ES
ELECTRA PRIVATE EQUITY PLC GB FORTHPANEL LTD GB
EURAZEO SA FR GALAXIE SA FR
EURAZEO SA FR AIR LIQUIDE SA FR
GIMV NV BE INTERWISE INC. US
GIMV NV BE OPENBRAVO SL ES
GIMV NV BE VANDEMOORTELE NV BE
GIMV NV BE UBIDYNE GMBH DE
GIMV NV BE EASYVOYAGE SA FR
GIMV NV BE RES HOLDING BV NL
GIMV NV BE MCPHY ENERGY SA FR
GIMV NV BE PRIVATE OUTLET SAS FR
GIMV NV BE EDEN CHOCOLATES BE
GIMV NV BE PE INTERNATIONAL GMBH DE
GIMV NV BE AMBIT BIOSCIENCES CORPORA-
TION
US
GIMV NV BE EBUZZING FR
GIMV NV BE TINUBU SQUARE FR
GIMV NV BE PRONOTA NV BE
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 80
GIMV NV BE PROSENSA HOLDING BV NL
HELIAD EQUITY PARTNERS GMBH
& CO. KGAA
DE VANGUARD AG DE
INTERNET CAPITAL GROUP INC. US VCOMMERCE CORPORATION US
INTERNET CAPITAL GROUP INC. US CHANNEL INTELLIGENCE INC. US
INTERNET CAPITAL GROUP INC. US STARCITE INC. US
INTERNET CAPITAL GROUP INC. US ICG COMMERCE INC. US
KOHLBERG KRAVIS ROBERTS &
COMPANY LP
US FOTOLIA LLC US
MARFIN INVESTMENT GROUP
HOLDINGS SA
GR SUNCE KONCERN DOO HR
MARFIN INVESTMENT GROUP
HOLDINGS SA
GR FAI RENT-A-JET AG DE
RATOS AB SE ARCORUS AB SE
RATOS AB SE ARCUS-GRUPPEN AS NO
RATOS AB SE INWIDO FINLAND OY FI
WENDEL SA FR STAHL GROUP BV NL
Source: Zephyr
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 81
Appendix 5 – Descriptive Statistics
Deal Characteristics
Exhibit A.5.1 – Deal Activity per Year
Notes: Pre-2003 includes 1 deal from 2001 and 9 deals from 2002. 2012 includes deals until 20.05.2012.
Source: Zephyr
Exhibit A.5.2 – Deal Period
Deal Period No. of deals Percentage
2001 – 2003* 20 15.5%
2004 - 2008 86 66.7%
2009 – 2012** 23 17.8%
Total 129 100.0%
Notes:*The period 2001-2003 includes only one deal from 2001, **the period 2009-2012 includes deals until 20.05.2012. This
includes 4 deals from 2012.
Source: Zephyr
-
4
8
12
16
20
24
28
-
250
500
750
1.000
1.250
1.500
1.750
No
. o
f d
eals
Ag
gre
gate
d d
eal
valu
e (
US
D m
n.)
Year
No. of deals Aggregated deal value (USD mn.)
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 82
Exhibit A.5.3 – Deal Size
Deal Size No. of deals Percentage
Less than 10% of the LPEV’s MV 115 89.1%
More than 10% of the LPEV’s MV 14 10.9%
Total 129 100.0%
Notes: Deal Size = Deal Value in USD reported by Zephyr / Market Value of the Acquirer at the announcement date in USD report-
ed by Datastream.
Sources: Zephyr and Datastream
Exhibit A.5.4 – Deal Size Statistics
Measure USD mn.
Maximum deal size 623.2
Minimum deal size 1.6
Average deal size 55.0
Source: Zephyr
Table A.5.5 – Deal Geographical Scope
Geographical Scope of the deal No. of deals Percentage
Domestic deals 54 41.9%
Cross-border deals 75 58.1%
Total 129 100.0%
Source: Zephyr
Exhibit A.5.6 – Deal Type of Buyout
Type of Buyout No. of deals Percentage
Outsider-driven* 129 100.0%
Insider-driven** 0 0.0%
Total 129 100.0%
Notes: *Outsider-driven buyouts are defined as MBIs and IBOs. **Insider-driven buyouts are defined as MBOs.
Source: Zephyr
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 83
Exhibit A.5.7 – Deal Means of Payment
Means of Payment No. of deals Percentage
Cash 105 81.4%
Mixed 1 0.8%
Undisclosed 23 17.8%
Total 129 100.0%
Source: Zephyr
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 84
Target Characteristics
Exhibit A.5.8 – Target Industry
Note: Targets are classified according to their primary US SIC codes.
Sources: Zephyr, NAICS Association (2008), and U.S. Securities and Exchange Commission (2011).
Table A.5.9 – Target Legal Origin
Legal Origin of the Target No. of deals Percentage
Common law country 50 38.8%
Civil law country 79 61.2%
Total 129 100.0%
Source: Zephyr
Table A.5.10 – Target Former Ownership
Former Ownership of the Target No. of deals Percentage
Public ownership 4 3.1%
Private ownership 125 96.9%
Total 129 100.0%
Source: Zephyr
- 10 20 30 40 50 60 70
Construction & Mining
Finance, Insurance & Real Estate
Transportation, Communications,
Electric, Gas and Sanitary Services
Wholesale & Retail Trade
Manufacturing
Services
No. of targets
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 85
Exhibit A.5.11 – Target Geographical Origin
Country code (country) No. of deals Percentage
BE (Belgium) 3 2.3%
CH (Switzerland) 1 0.8%
DE (Germany) 12 9.3%
DK (Denmark) 2 1.6%
ES (Spain) 14 10.9%
FI (Finland) 5 3.9%
FR (France) 17 13.2%
GB (the U.K.) 26 20.2%
HR (Croatia) 1 0.8%
IT (Italy) 4 3.1%
LU (Luxembourg) 2 1.6%
NL (Netherlands) 8 6.2%
NO (Norway) 1 0.8%
SE (Sweden) 9 7.0%
US (the U.S.) 24 18.6%
Total 129 100.0%
Source: Zephyr
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 86
Acquirer Characteristics
Exhibit A.5.12 – LPEV Structure
LPEV Structure No. of deals Percentage
Direct LPEVs 124 96.1%
Indirect LPEVs 5 3.9%
Total 129 100.0%
Notes: Direct LPEVs = LPE funds and Investment Companies, Indirect LPEVs = LPE firms.
Sources: Zephyr and LPX Group (2012c)
Exhibit A.5.13 – LPEV Experience
Experience of the LPEV No. of deals Percentage
Experienced 109 84.5%
Inexperienced 20 15.5%
Total 129 100.0%
Notes: Experienced = Established before 1990, Inexperienced = Established from 1990 and onwards
Source: Zephyr
Exhibit A.5.14 – LPEV Investment Strategy
LPEV Investment Strategy No. of deals Percentage
Diversified 119 92.2%
Specialized 10 7.8%
Total 129 100.0%
Sources: Zephyr and LPX Group (2012c)
Exhibit A.5.15 – LPEV Size
Measure USD mn.
Maximum LPEV size 11,133.2
Minimum LPEV size 67.4
Average LPEV size 4,697.4
Notes: LPEV size = the market value of the LPEV at the announcement date reported by Datastream in USD million. 3i Group PLC has an average size of USD 7,175.0 million. When 3i Group PLC is excluded, the average LPEV size is USD 1,144.7 million.
Source: Zephyr
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 87
Exhibit A.5.16 – LPEV Country of Origin
Country code (country) No. of deals Percentage
BE (Belgium) 15 11.6%
DE (Germany) 4 3.1%
ES (Spain) 7 5.4%
FI (Finland) 4 3.1%
FR (France) 4 3.1%
GB (the U.K.) 79 61.2%
GR (Greece) 2 1.6%
IT (Italy) 2 1.6%
SE (Sweden) 7 5.4%
US (the U.S.) 5 3.9%
Total 129 100.0%
Source: Zephyr
Exhibit A.5.17 – LPEV Region of Origin
Region No. of deals Percentage
Northern Europe ex. UK and Scandinavia 19 14.7%
Scandinavia 11 8,5%
Southern Europe 15 11.6%
UK 79 61.2%
US 5 3.9%
Total 129 100.0%
Source: Zephyr
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 88
Exhibit A.5.18 – Deals per LPEV
Note: Arques Industries AG was formerly known as Gigaset and can be found under this name in the Excel sheets on the CDROM
Source: Zephyr
0 20 40 60 80
ALTAMIR ET COMPAGNIE SCA
DEUTSCHE BETEILIGUNGS AG
ELECTRA PRIVATE EQUITY PLC
HELIAD EQUITY PARTNERS GMBH…
KOHLBERG KRAVIS ROBERTS &…
WENDEL SA
ARQUES INDUSTRIES AG
CANDOVER INVESTMENTS PLC
DEA CAPITAL SPA
EURAZEO SA
MARFIN INVESTMENT GROUP…
RATOS AB
BURE EQUITY AB
CAPMAN OYJ
INTERNET CAPITAL GROUP INC.
DINAMIA CAPITAL PRIVADO SCR SA
GIMV NV
3I GROUP PLC
No. of deals
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 89
Appendix 6 – Presentation of Test Statistics
Test Statistics for the Analysis of H0: CAAR = 0
T1 – T-test with unadjusted variances (Bartholdy et al. 2007)
√ , (A.1)
where
√∑
∑ (A.2)
T1 adj. – T-test with Patell’s adjustment (Patell 1976; Bartholdy et al. 2007)
√ , (A.3)
where
√∑
∑
∑
(A.4)
T2 – T-test with standardized abnormal returns (Bartholdy et al. 2007)
∑
∑
∑
√ , (A.5)
where
(A.6)
and √
∑
(A.7)
T3 – T-test with standardized abnormal returns and Patell’s adjustment (Patell 1976;
Corrado 2011)
∑√
√
, (A.8)
46
where ∑ , (A.9)
with
(A.10)
and √
∑
∑
(A.11)
46 One could have used the adjustment for forecast errors suggested by Salinger (1992). The adjustment is, however,
only relevant if one uses longer event windows.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 90
T4 – The Rank test (Corrado & Zivney 1992; Bartholdy et al. 2007)
∑
√ ∑
√ , (A.12)
where √
∑
√ ∑
, (A.13)
and
, (A.14)
where (A.15)
T5 – The Sign test (Campbell et al. 1997; Bartholdy et al. 2007)
∑
√ ∑
√ , (A.16)
where √
∑
√ ∑
(A.17)
and (A.18)
T6 – The Generalized Sign test (Bartholdy et al. 2007; Renneboog et al. 2007)
∑
√ , (A.19)
where ω is the number of positive CAARs in the event window for the sample
and SD √ , (A.20)
with
∑
∑
, (A.21)
and if and if (A.22)
T7 – The BMP test: A t-test with variance-adjusted standardized abnormal returns
(Boehmer et al. 1991; Bartholdy et al. 2007)
∑
√ , (A.23)
where √(
) ∑
∑
(A.24)
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 91
and ∑ (A.25)
with
, (A.26)
and √
∑
∑
(A.27)
T8 – The Rank test with variance-adjusted standardized abnormal returns (Patell
1976; Bartholdy et al. 2007)
∑
√ ∑
√ , (A.28)
where √
∑
√ ∑
(A.29)
and
, (A.30)
where (A.31)
When then is used in (A.31),
where
, (A.32)
with √
∑
for the estimation period (A.33)
and √
∑
∑
for the event window (A.34)
When then is used instead of in (A.31),
where
, (A.35)
with √
∑
∑
(A.36)
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 92
Test Statistics for the Analysis of Differences between Groups
T-test for differences between two means with equal variances (Keller 2005)
√
, (A.37)
where
(A.38)
The test statistic has degrees of freedom.
T-test for differences in between two means with unequal variances (Keller 2005)
√
, (A.39)
with (
)
(
)
(
)
degrees of freemdom.
F-test for differences in variances (Keller 2005)
To test whether there are unequal variances, an F-test is used:
, (A.40)
with and degrees of freedom.
Wilcoxon Rank Sum test for differences between two locations (Keller 2005)
, (A.41)
where ∑ , (A.42)
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 93
, (A.43)
and √
(A.44)
ANOVA test for differences between two or more means (Keller 2005)
, (A.45)
where
, with ∑
, (A.46)
and
, with ∑ ∑
(A.47)
Kruskal-Wallis test for differences between two or more locations (Keller 2005)
[
∑
] , (A.48)
where ∑ for sample j (A.49)
.
.
.
.
.
.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 94
Appendix 7 – Test of Assumptions
According to Campbell et al. (1997) parametric tests in event studies rely on the as-
sumption of independent and identically distributed (i.i.d.) abnormal returns:
(A.50)
This assumption is evaluated, both for the total sample and for a sample excl. 3i Group
PLC, by making a histogram of the abnormal returns along with a P-P plot of the stand-
ardized abnormal returns. In addition to this, measures of kurtosis and skewness are
calculated and a Jarque-Bera test is performed. A sample is evaluated to follow the
normal distribution if it has a kurtosis close to three and a skewness of zero. Alterna-
tively, a sample is said to follow the normal distribution if the Jarque-Bera statistic is
insignificant.
From exhibit A.7.1 one can see that the skewness is approximately zero and that the
kurtosis is close to three. In addition to this, the Jarque-Bera statistic is insignificant,
since it has a p-value of 0.426. Thus, the assumption is evaluated to be fulfilled for the
total sample. This is confirmed by looking at the histogram and the P-P plot of the
standardized abnormal returns in exhibit A.7.2.
Exhibit A.7.1 – Histogram, Descriptive Statistics and Test of Normality
Sources: Own calculations in MS Excel 2007 and EViews 5.0
0
4
8
12
16
20
-0.0250 -0.0125 -0.0000 0.0125 0.0250
Series: CAR
Sample 1 129
Observations 129
Mean 0.001142
Median 0.000680
Maximum 0.024072
Minimum -0.024581
Std. Dev. 0.008890
Skewness -0.028432
Kurtosis 3.560549
Jarque-Bera 1.706285
Probability 0.426074
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 95
Exhibit A.7.2 – P-P plot of Standardized Abnormal Returns
Sources: Own calculations in MS Excel 2007 and EViews 5.0
From exhibit A.7.3 one can see that the skewness is -0.493 and that the kurtosis is
3.852. This is reasonably close to the required values of zero and three respectively. In
addition to this, the Jarque-Bera statistic is insignificant, since it has a p-value of 0.153.
Thus, the assumption of i.i.d. is evaluated to be fulfilled for the sample which excludes
3i Group PLC. This is confirmed by looking at the histogram and the P-P plot of the
standardized abnormal returns in exhibit A.7.4, although the confirmation is not as
strong as for the total sample.
Exhibit A.7.3 – Histogram, Descriptive statistics and Test of Normality (excl. 3i Group PLC)
Sources: Own calculations in MS Excel 2007 and EViews 5.0
0
2
4
6
8
10
12
14
-0.02 -0.01 -0.00 0.01 0.02
Series: CAR2
Sample 1 129
Observations 53
Mean -0.000428
Median -0.000651
Maximum 0.017984
Minimum -0.024581
Std. Dev. 0.008124
Skewness -0.493947
Kurtosis 3.852071
Jarque-Bera 3.758496
Probability 0.152705
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 96
Exhibit A.7.4 – P-P plot of Standardized Abnormal Returns (excl. 3i Group PLC)
Sources: Own calculations in MS Excel 2007 and EViews 5.0
The assumption of uncorrelated abnormal returns across securities is satisfied when
there is no significant clustering of the events in calendar time (Campbell et al., 1997).
From exhibit A.7.5 one can see that 15.5% of the events to some degree suffer from
calendar clustering. This suggests that the abnormal returns are not completely uncorre-
lated. However, since the securities are listed in ten different countries, and since the
targets are located in 15 different countries, the abnormal returns from the deals that
suffer from calendar clustering are expected to be fairly uncorrelated. Thus, the assump-
tion of uncorrelated abnormal returns across securities is evaluated to be fulfilled for the
total sample.
Exhibit A.7.5 – Event Window Clustering
Days Between Events Frequency Percentage
Zero days 4 3.1%
One day 8 6.2%
Two days 6 4.7%
Three days 2 1.6%
Total 20 15.5%
Source: Zephyr
From exhibit A.7.6 one can see that 11.3% of the events in the sample (which does not
include 3i Group PLC) to some degree suffer from calendar clustering. One the other
hand, the securities are listed in 10 different countries and the targets are located in 13
different countries. Thus, the assumption of uncorrelated abnormal returns across secu-
rities is evaluated to be fulfilled for the sample that does not include 3i Group PLC.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 97
Exhibit A.7.6 – Event Window Clustering (excl. 3i Group PLC)
Days Between Events Frequency Percentage
Zero days 2 3.8%
One day 2 3.8%
Two days 2 3.8%
Three days 0 0.0%
Total 6 11.3%
Source: Zephyr
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 98
Appendix 8 – Results for the Sample excl. 3i Group PLC
Exhibit A.8 – CAAR to Announcements of Acquisitions by LPEVs (excl. 3i
Group PLC)
Parametric tests Nonparametric tests Var. adj. tests Differences
Sample N CAAR T1 T1 adj. T2 T3 T4 T5 T6 T7 T8 T-test W-test
Notes: CAAR is defined as the sum of CAR [-1; 1] divided by N. The significance level is indicated by a = 10%, b = 5%, and c = 1%. * implies that the tests for differences are the ANOVA test and the Kruskal-Wallis test. W-test = the Wilcoxon Rank Sum test. The test
for differences which is relied upon is marked with bold; the t-test is relied upon when the observations of both groups are normally
distributed according to a Jarque-Bera test; when they are not, the Wilcoxon Rank Sum test is relied upon. The industries are as
follows: ‘Min. & Con.” = Mining & Construction, “T., C., E., G. & S. S” = Transportation, Communication, Electric, Gas & Sanitary Services, “W. & R. Trade” = Wholesale & Retail Trade, and “F., I. & R. E.” = Finance, Insurance & Real Estate”.
AN EXPLORATIVE EVENT STUDY OF LISTED PRIVATE EQUITY VEHICLES
JULY 2012 MATHIAS LETH NIELSEN 99
Appendix 9 – Test for Differences in CAAR
Exhibit A.9.1 – Test for Differences in CAAR between the Overall Sample and the Sample excl. 3i Group
Notes: CAR = CAAR for the total sample, CAR2 = CAAR for the sample excl. 3i Group PLC.