Aarhus School of Business and Social Sciences Thesis for Master of Science in Finance and International Business

The Second Era of Private Equity Does Private Equity Fund Ownership (Still) Induce Superior Performance?

Author: Jakob Skau ID: 288036 Academic advisor: Stig Vinther Møller, Associate Professor, Ph.D.

August 2013

Executive Summary This thesis explores value creation within private equity (PE), specifically the operating performance of PEbacked companies (portfolio firms). The majority of previous papers dealing with this matter report post-buyout abnormal performance in the portfolio firms, and some even hail the PE model as a superior organizational form. Opponents of the model, however, argue that PE firms do not generate returns by creating economic value but rather by taking advantage of asymmetric information, market timing and mispricing between debt and equity markets. Common to most previous studies are their outdated data material, which is typically drawn from the first PE wave1 in the late 1980s. Recent research indicates that several structural factors have changed since then, affecting the overall value creation potential of the PE model negatively and the relative significance of the levers through which PE firms create value. Value creation within private equity is typically categorized into three overall levers: financial2, operational3, and governance engineering4. It is claimed that the importance of financial engineering has declined whereas the importance of operational engineering has increased. Hence, this thesis aims to estimate the economic impact of PE fund ownership on portfolio firm performance in order to assess the claimed superiority of private equity as an organizational form. Moreover, it seeks to determine whether value-adding activities related to operational and financial engineering can explain the findings. To the best of my knowledge, no papers have yet investigated this area in depth using fresh European data from the second PE wave 5 . The analysis is based on company data (mainly accounting data) of 130 European portfolio firms acquired during the second PE wave between 2003 and 2008, and a benchmark of 130 companies acquired by industrial corporations (control firms) during the same time period. To analyze the data, the difference-in-difference (DID) method and an event window of three years are applied. OLS regression is used for the DID estimation. The present analysis estimates PE fund ownership to impact performance in portfolio firms positively compared to the control group, specifically on these measures: Return on Asset by 2,8 pp, Operating Return on Assets by 2,3 pp, EBIT-Margin by 4,7 pp and Net Profit Margin by 2,3 pp on average. These findings are robust to econometric specification (i.e. alternative independent variables), choice of performance measures, and matching procedure. This indicates that the PE model is in fact a superior organizational form. However, the magnitude of the estimated impact points towards an overall decline in the model’s value creation potential, compared to studies based on data from the first PE wave. PE funds’ financial engineering abilities do not explain the above findings, but superior operational engineering skills seem to do – particularly related to cost cutting and reduction of capital requirements. These results indicate that a change in significance of the different value creation levers has indeed materialized, namely a decrease in importance of financial engineering and an increase in importance of operational engineering.

1

Historically, two PE waves have been identified. The first in the 1980s and the second in the 2000s. Financial engineering comprises optimization of capital structure and minimization of cost of capital 3 Operational engineering mainly comprise: cost cutting and margin improvements, reduction in capital requirements and improvement of asset efficiency. A common factor among these is that they enhance productivity and effectiveness of operations and optimize resource use 4 Governance engineering mainly revolves around the minimization of agency costs. These are for instance minimized by enhancing incentive alignment and improving monitoring and control of the board and management. 5 This roughly spans from 2003 to the outbreak of the financial crisis in 2008. 2

Table of contents 1.

Introduction ........................................................................................................................................1 1.1.

Background ............................................................................................................................................................... 1

1.2.

Problem Statement................................................................................................................................................ 3

1.3.

Structure ..................................................................................................................................................................... 4

1.4.

Technicalities ........................................................................................................................................................... 5

2.

Private Equity Buyouts and Their Historical Development...................................................5

3.

Value Creation in Private Equity Funds.......................................................................................8 3.1.

4.

5.

The Three Levers of Value Creation ............................................................................................................. 8

3.1.1.

Operational Engineering .......................................................................................................................... 8

3.1.2.

Financial Engineering ................................................................................................................................ 9

3.1.3.

Governance Engineering ....................................................................................................................... 10

3.2.

Change in Relative Significance of the Three Levers of Value Creation ................................. 11

3.3.

Change in Value Creation Potential of the Overall PE Model ....................................................... 11

3.4.

Preliminary Conclusions From the Literature Review .................................................................... 12

Hypotheses ....................................................................................................................................... 13 4.1.

The Superiority Hypothesis (main hypothesis) .................................................................................. 13

4.2.

The Financial Engineering Hypothesis (support hypothesis) ..................................................... 13

4.3.

The Operational Engineering Hypotheses (support hypotheses) ............................................. 13

Empirical Strategy .......................................................................................................................... 15 5.1.

The Difference-In-Difference Method ...................................................................................................... 15

5.1.1. 5.2.

Strategy for Hypotheses Testing ................................................................................................................. 21

5.2.1.

Empirical Testing of The Superiority Hypothesis .................................................................... 21

5.2.2.

Empirical Testing of The Financial Engineering Hypothesis ............................................. 22

5.2.3.

Empirical Testing of The Operational Engineering Hypotheses ...................................... 24

5.3. 6.

Description of Applied Variables ...................................................................................................... 17

Critical Elements and Limitations of the DID Method ..................................................................... 25

Data .................................................................................................................................................... 28 6.1.

Data Sources .......................................................................................................................................................... 28

6.2.

Sample Selection.................................................................................................................................................. 29

6.2.1.

Sampling Strategy for the Portfolio Firms ................................................................................... 29

6.2.2.

Matching Procedure ................................................................................................................................ 30

6.2.3.

Comments on the Final Data Set ....................................................................................................... 32

6.3.

Descriptive Statistics and Sample Characteristics ............................................................................. 33

7.

6.3.1.

Pre-Buyout Descriptive Statistics..................................................................................................... 33

6.3.2.

Post-Buyout Descriptive Statistics................................................................................................... 35

6.3.3.

Other Sample Characteristics ............................................................................................................. 38

Results and Discussion .................................................................................................................. 40 7.1.

Testing of The Superiority Hypothesis .................................................................................................... 40

7.2.

Robustness Checks ............................................................................................................................................. 43

7.2.1.

Econometric Specification.................................................................................................................... 43

7.2.2.

Performance Measures .......................................................................................................................... 44

7.2.3.

Matching Criteria ...................................................................................................................................... 46

7.3.

Testing of the Support Hypotheses............................................................................................................ 48

7.3.1.

The Financial Engineering Hypothesis .......................................................................................... 48

7.3.2.

The Operational Engineering Hypotheses ................................................................................... 52

7.4.

Alternative Explanations................................................................................................................................. 58

7.4.1.

Omitted Variable Bias ............................................................................................................................. 58

7.4.2.

Sampling bias .............................................................................................................................................. 59

7.4.3.

Survivorship bias ...................................................................................................................................... 60

7.4.4.

Differences in Acquisition Motives .................................................................................................. 60

8.

Conclusion ........................................................................................................................................ 61

9.

Future Research .............................................................................................................................. 63

References ................................................................................................................................................ 65

List of Figures Figure 1 – Global Private Equity Investments ............................................................................................................. 6 Figure 2 – Private Equity Investments in Europe...................................................................................................... 7 Figure 3 – Regional Breakdown of Private Equity Investments ........................................................................ 7 Figure 4 – The Sample Transactions Split by Nationality and Type of the Target Company........... 38 Figure 5 – The Sample Transactions Split by Transaction Year ..................................................................... 39

List of Tables Table 1 – Difference-In-Difference Table .................................................................................................................... 16 Table 2 – Dependent Variables ........................................................................................................................................ 17 Table 3 – Control Variables ................................................................................................................................................ 20 Table 4 – Key Explanatory Variables ............................................................................................................................ 21 Table 5 – Summary Statistics of the Sample.............................................................................................................. 34 Table 6 – The Impact of Private Equity Fund Ownership on Firm Performance ................................... 41 Table 7 – Test Summary of OLS Assumptions .......................................................................................................... 43 Table 8 – The Impact of PE Fund Ownership on Firm Performance Adjusted for U.K.-effect ......... 44 Table 9 – The Impact of PE Fund Ownership on Alternative Firm Performance Measures............. 45 Table 10 – The Impact of PE Fund Ownership on Firm Performance (Alt. Matching)........................ 47 Table 11 – The Impact of Financial Engineering on Firm Performance ..................................................... 50 Table 12 – The Impact of Operational Engineering on Firm Performance ............................................... 55

List of Selected Abbreviations CAPEX

Capital Expenditures

Control firm

Benchmark firm, acquired by an industrial corporation

Controls

Control Variables

DID

Difference-In-Difference

FCF

Free Cash Flow

LBO

Leveraged Buyout

OLS

Ordinary Least Squares

PE

Private Equity

Portfolio firm

A PE backed firm

pp

Percentage points

R&D

Research and Development

1. Introduction 1.1. Background During the last decade, there has been remarkable growth in private equity (PE), and the asset class now accounts for about 12% of global M&A activity measured by deal value (McKenzie & Maslakovic). Jensen, (1989) stated that LBO organizations, such as PE firms 6, would become the dominating corporate organizational form, arguing that the PE model combines concentrated ownership, performance-based managerial compensation, highly leveraged capital structures, active governance, and a lean and efficient organization with low overhead costs – i.e., “The Jensen Hypothesis”. According to Jensen and other proponents, this organizational form was superior to the typical public form, which had dispersed ownership, low leverage, and weaker corporate governance (Kaplan & Strömberg, 2008). Generally, the comparative advantages of the PE model are believed to originate from three channels: operational7, financial8, and governance engineering9. The majority of research in the field supports Jensen’s findings by reporting a superior positive impact of PE fund ownership on portfolio firm10 performance (e.g., Baker & Wruck, 1989; Kaplan, 1989a, 1989b; Lichtenberg & Siegel, 1990; Muscarella & Vetsuypens, 1990; Smith, 1990; Holthausen & Larcker, 1996; Wright et al., 1997; Harris et al., 2005; Cressy et al., 2007; Cao & Lerner, 2009; Guo et al., 2011), whereas fewer studies present contradictory findings (e.g., Ravenscraft & Scherer, 1987: Desbriéres & Schatt, 2002). In contrast, studies belonging to the latter group claim that PE firms do not have a beneficial economic impact on their targets (NB. see appendix 1 for a thorough introduction to PE firms and funds and the nature of their transactions). They assert that PE firms are merely good merchants, generating returns by taking advantage of asymmetric information, market timing, and market mispricing between debt and equity markets (Kaplan & Strömberg, 2008). More than two decades have passed since Michael Jensen wrote his seminal piece (Jensen, 1989), and meanwhile, the PE landscape has changed significantly (Jensen et al., 2006; Guo et al., 2011). From the first PE wave in the 1980s to the second and most recent in the 2000s, new deal and target characteristics have emerged. Consequently, the mechanisms through which PE firms create value are likely to have changed as well. Traditionally, financial and governance engineering have been the most frequently discussed levers by researchers, but as an increasing number of industrial corporations have come to understand the value potential of leverage, PE firms have been forced to emphasize other means. One view of PE firms today, which is gradually becoming more common, is that

6

See appendix 1 for a thorough introduction to PE firms and funds. Moreover note that the terms “PE firm” and “PE fund” are used interchangeably in this thesis. 7 Operational engineering mainly comprise: cost cutting and margin improvements, reduction in capital requirements and improvement of asset efficiency. A common factor among these is that they enhance productivity and effectiveness of operations and optimize resource use. 8 Financial engineering comprises optimization of capital structure and minimization of cost of capital 9 Governance engineering mainly revolves around the minimization of agency costs. These are for instance minimized by enhancing incentive alignment and improving monitoring and control of the board and management. 10 The term “portfolio firm” is used through the entire paper to describe a company which has been acquired by a PE fund. It is only characterized as a portfolio firm while under PE fund ownership.

1

portfolio firm performance is enhanced through excellence in operational engineering. This could represent either an actual switch in value creation focus or simply a way of justifying the relatively high fees claimed by PE firms. Nevertheless, it seems curious that PE funds should excel within operational engineering as industrial acquirers (the control group of the present analysis) are expected to have a comparative advantage related to this parameter. Industrial acquirers typically have many years of experience operating within a given industry, whereas PE firms intuitively are expected to be greener. Focus and comparability of the present literature Another motivating factor for researching value creation within private equity is the lack of comparability between existing studies. These typically analyze management buyouts (MBOs) 11 , leveraged buyouts (LBOs)12, and reverse LBOs (RLBOs)13 in combination, although they are rather different by nature (Vinten, 2007). For instance, the impact of ownership in the wake of an LBO is analyzed while the portfolio firm is private, whereas RLBOs are analyzed after they have been relisted. Hence, the latter type of studies also reflects the impact of a new owner which is not a PE fund. Moreover, not all LBOs and MBOs are undertaken with a PE fund as lead acquirer, and consequently, most of the literature focusing on these buyout types does not deal with the impact of PE fund ownership in isolation (e.g., Cressy et al., 2007). Data availability is a major problem in relation to this type of research. In most countries (especially the U.S.) only very limited data are reported on private companies, which results in certain sample selection biases and limitations. Typically, researchers have been forced to focus on the post-exit situation of buyout firms, e.g., RLBOs (Muscarella & Vetsuypens, 1990; Cao & Lerner, 2009) or to use a disproportionate amount of public-to-private transactions, as disclosure of financial data is typically more profound in listed companies compared to those privately owned (e.g., Kaplan, 1989a; Smith, 1990). The former situation is obviously not the ideal setup to assess the impact of PE fund ownership, and the latter is problematic as 80% of deal value and 90% of transactions in Europe were private-to-private during the last decade (McKenzie & Maslakovic). Note, The Jensen Hypothesis predicts that the PE model has the largest beneficial impact in the public-to-private setting, as agency cost savings related to an increase in ownership concentration are expected to be most significant here. Data used in several recent studies (e.g., Archleitner et al., 2010) span time periods which are assessed to be fundamentally different from one another. Thus, it could be beneficial to draw data from the second wave and analyze it in isolation. Or as Kaplan and Strömberg (2008) put it:

11

A Management Buyout (MBO) is a form of acquisition where a company's existing managers acquire a large part or all of the compa ny from either the parent company or the private owners. 12 Described in section 1.1. 13 Reverse Leverage Buyout (RLBO) is used to describe the process of a portfolio firm resuming its previously held publicly trad ed status after having been acquired through a leveraged buyout (LBO).

2

“...the performance of leveraged buyouts completed in the latest private equity wave is clearly a desirable topic for future research.” Furthermore, most of the existing literature is based on U.K. or U.S. data, and not enough is known about the impact of the PE model in continental Europe. The nature of ownership structure in the U.S. and U.K. differs significantly from that most commonly seen in continental Europe. Pre-buyout ownership concentration tends to be higher in continental Europe (e.g., Faccio & Lang, 2002), which according to The Jensen Hypothesis means fewer agency problems ex-ante and thus a lower expected benefit from the PE model. Based on the above, it is found desirable to understand the PE model, and particularly the value creation of PE funds, better. Moreover, it is acknowledged that fresh empirical evidence is needed.

1.2. Problem Statement Thus, this thesis aims to estimate the economic impact of PE fund ownership on portfolio firm performance in order to assess the claimed superiority of private equity as an organizational form. Moreover, it seeks to determine whether value-adding activities related to operational and financial engineering can explain the findings. First, it addresses the following main research question (R.Q.): R.Q. 1: How does PE fund ownership impact the performance of portfolio firms post-buyout? Numerous value-creating factors are identified in the literature, but the acknowledged significance of these differs considerably amongst researchers. The debate on the change in relative significance is a particular focus in this thesis. Financial engineering has historically been considered a highly important lever; hence the following research question is posed: R.Q. 2: Can the supposed superior financial engineering of PE funds (still) explain the findings related to R.Q. 1? However, due to indications of decreasing significance of financial engineering and increasing significance of operational engineering from the first to the second PE wave, the following research question is investigated: R.Q. 3: Can PE funds’ operational engineering abilities explain the findings related to R.Q. 1? Five hypotheses are formulated and tested later in this thesis to answer the above research questions.

3

Main contribution This thesis circumvents many of the limitations related to previous studies and provides fresh evidence on the performance impact of PE fund ownership and its drivers. First, it is solely focused on LBOs following PE ownership. Hence, the data contains no “noise” from other buyout types, such as MBOs and RLBOs. To the best of my knowledge, only a few studies have analyzed the activity in this area after the mid-1990s (e.g., Cressy et al., 2007; Guo et al., 2011). Second, for this thesis a unique set of data has been gathered on private companies (elaborated in section 6), which has made it possible to investigate the portfolio firms while under PE fund ownership (in contrast to previous RLBO studies). In addition, private-to-private PE transactions constitute a reasonable proportion of the data set (85%), which is considerably higher than similar studies (e.g., Cressy et al., 2007; Vinten, 2007). Third, it provides fresh evidence on the economic impact of PE ownership, which is needed as the second PE wave seems incomparable to the first. Hence, it is likely that the findings of this study will differ from previous ones. The data gathered for the present analysis, span from 2003 to 2008, which covers the second PE wave in isolation. Fourth, as described above, most previous research has been based on U.S. or U.K. data, which could be fundamentally different from continental European data. Thus, the forthcoming thesis covers acquisitions in all of Europe – both the U.K. and continental Europe – in order to fill this gap. Delimitation Most of the delimitation is outlined throughout the thesis, but one element in particular, is relevant to highlight up front: This thesis does not deal with returns to PE investors per se. Certain studies (e.g. (Archleitner et al., 2010) decompose the Internal Rate of Return (IRR) on fund investments to assess which factors account for how much (these studies are typically split into: Leverage-effect, EDITDAgrowth, FCF-effect and a multiple-effect). However, the typical research setup of this type of studies is fundamentally different to the one applied in this thesis, and including such a decomposition of the IRR in the present analysis, would broaden its scope excessively.

1.3. Structure The remainder of the thesis is structured as follows. In section 2, the concept of private equity and the historical development of PE transactions are described. Section 3 describes value creation of PE firms and how it might have changed over time. Based on the previous sections’ literature review, a set of hypotheses are formulated in section 4. Section 5 outlines the empirical strategy for testing the hypotheses. Section 6 describes how the data was collected and the final sample’s characteristics. The

4

results of the hypothesis tests are presented and discussed in section 7. Section 8 concludes the analysis. Finally, section 9 highlights relevant areas for future research.

1.4. Technicalities The literature was gathered mainly from databases such as Business Source Complete, JSTOR, Science Direct, Social Sciences Research Network and Wiley Online Library. The data was drawn mainly from Zephyr and Orbis, but Mergermarket.com, EVCA, company websites, and several databases holding full company financial statements (e.g., Greens.dk) were also used. The data was processed using MS Excel 2007, and the statistical analysis was conducted in EViews 7.1. All data can be found on the enclosed CD-ROM.

2. Private Equity Buyouts and Their Historical Development PE transactions PE firms (for a thorough introduction to PE firms and funds, see appendix 1) usually undertake their acquisitions through what are called leveraged buyouts (LBOs). These deals are usually financed with a large majority of debt – hence the term – which prior to the financial crisis would typically amount to 60–90% of the price. Post–financial crisis it has become more complicated to raise debt in the financial market, and consequently, this percentage has declined (Guo et al., 2011). The debt typically consists of a portion which is senior and secured, plus a portion which is junior and unsecured (typically mezzanine - see description in appendix 1). The PE firm then applies capital from its investors as equity to cover the remaining 10–40% of the investment. The management teams which have been picked by the PE firm to run the portfolio company post- acquisition (which could be the current team, a set of executives found externally, or a combination thereof) usually also contribute some equity, though the amount is typically only a small fraction of the total investment. Historical development of private equity and the two waves LBOs first emerged in the 1980s as an important phenomenon but have undergone remarkable growth since. Figure 1 presents the development in global PE investments from 1985 to 2012. During the existence of private equity, two major waves have occurred. The first wave took place in the late 1980s and was primarily a North American and, to some extent, a U.K. phenomenon. From 1985 to 1989, these two regions accounted for 89% of worldwide leveraged buyout transactions and 93% of transaction value. This wave was characterized by large transactions in mature industries (such as manufacturing and retail), and public-to-private deals accounted for almost half of the transaction value. In the late 1980s the junk bond market crashed, followed by declining PE activity in the 1990s, and The Jensen Hypothesis took a hit. A significant number of LBOs eventually ended up in default and bankruptcy, and afterwards LBOs of public companies (public-to-private transactions) nearly disappeared.

5

During the 1990s, smaller and privately owned companies became a more frequent target for PE funds. Although aggregate transaction value declined, twice as many deals were undertaken between 1990 and 1994, compared to 1985–89, on a global basis. The switch in target focus from public to private should be seen as a necessary alteration to the PE model given the economic environment, rather than an admission of its failure (Kaplan & Strömberg, 2008). Figure 1 – Global Private Equity Investments The primary axis depicts the aggregated yearly value of investments made by PE funds globally. Note that these numbers are likely to understate real value, as deal value is often not disclosed in PE transactions.

Aggregated deal value (EURbn)

600

450

300

150

1st PE wave

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

0

2nd PE wave

Source: (Kaplan & Strömberg, 2008; McKenzie & Maslakovic, 2012; EVCA(b), 2013)

As depicted in the figure above, the second wave started in the early 2000s, peaked in 2006 and 2007, which became record-breaking years in terms of global PE investments, and ended with the onset of the financial crisis in 2008. This wave and the characteristics distinguishing it from the first are particularly interesting in the scope of this thesis, as all of the transactions included in its data set fall within this time period. However, in the wake of the financial crisis that began in 2008, activity declined significantly before it was restored in 2010. This was followed by stagnation in 2011 and a small dip in 2012. The latter is attributed to the turmoil in the European debt markets during the first half of 2012, and significant growth is expected in 2013 (McKenzie & Maslakovic). Overall, researchers and practitioners label the present situation “normalization”. The same overall picture as described above is seen in figure 2, which depicts the past decade’s PE investments in Europe. Investments increased significantly from the early 2000s up to the financial crisis in 2008; however, it should be noted that European deals account for a relatively small part of the aggregated activity in 2006 and 2007. During these years, American PE funds in particular increased their investments at a relatively high pace (Kaplan & Strömberg, 2008). 6

Figure 2 – Private Equity Investments in Europe The primary axis shows the aggregated yearly value of investments made by PE funds in Europe. Note that these numbers are likely to understate real value, as deal value is often not disclosed in PE transactions.

Aggregated deal value (EURbn)

80

60

40

20

0 2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

Source: (EVCA(b), 2013)

The figure below provides a regional breakdown of global private equity activity for 1991, 2001, and 2011 (unfortunately, earlier data were not available). From the beginning, private equity was mainly a U.S. and U.K. phenomenon, but by the early 1990s it had established roots in continental Europe as well. In 1991 13% of global PE investments, measured by aggregated deal value, were undertaken in this region. During the late 1990s and early 2000s, the PE model’s popularity increased dramatically, and by 2001 continental European targets accounted for 32% of the total investment value. This percentage declined to 25% in 2011, mainly as a result of the turmoil on the capital markets in Europe and subsequently an increasing popularity of the PE model outside North America and Europe. Figure 3 – Regional Breakdown of Private Equity Investments The figure shows the value of global PE investment split by location of portfolio firms for 1991, 2001, and 2011, respectively. 1991 13%3%

7%

13%

United Kingdom Rest of world

16% 44%

32% 71%

North America Continental Europe

2011

2001

48%

11%

17% North America Continental Europe

25%

United Kingdom Rest of world

North America Continental Europe

United Kingdom Rest of world

Source: (McKenzie & Maslakovic)

7

3. Value Creation in Private Equity Funds 3.1. The Three Levers of Value Creation As touched upon earlier, the superiority of the PE model is believed to originate from three value creation channels: operational, financial, and governance engineering. These are outlined in the following sections. 3.1.1. Operational Engineering Operational engineering can be achieved through several initiatives, which Gottschalg and Berg (2005) categorize as cost cutting and margin improvements, reduction in capital requirements, and improvement of asset efficiency. A common factor among these is that they enhance productivity and effectiveness of operations and optimize resource use. A number of studies have documented a positive effect related to these levers (e.g., Kaplan, 1989b; Muscarella & Vetsuypens, 1990; Wright et al., 2001; Guo et al., 2011). Cost cutting and margin improvements Post-buyout, the PE fund typically starts tightening corporate spending and initiates several costreduction initiatives (Kaplan, 1989b; Magowan, 1989; Anders, 1992; Holthausen & Larcker, 1996). These include substantial changes to how operations are both organized and managed in the portfolio firm, with the aim of reducing costs and improving margins (Muscarella & Vetsuypens, 1990; Wright et al., 2001; Guo et al., 2011). In particular, lower overhead costs play a key role as PE funds tend to make company structures less bureaucratic and leaner (Easterwood, Seth, & Singer, 1989; Butler, 2001). In fact, during the last decade it has become increasingly common to see PE firms hiring exconsultants and ex-industrial executives with vast experience in these fields to assist with operational engineering (Kaplan & Strömberg, 2008). Moreover, margins can be improved by adding value to the goods or services offered by the portfolio firm, enabling it to charge higher prices. This is another area where PE firms allegedly have strengthened their abilities in recent years (Kaplan & Strömberg, 2008). Sometimes these initiatives also include laying off employees, which historically has given PE funds a somewhat bad reputation in the media. Several studies, however, indicate that employment is in fact increased post-buyout (Kaplan, 1989b; Lichtenberg & Siegel, 1990). Other controversial areas of cost reduction are Research and Development (R&D) costs and Capital Expenditures (CAPEX). Opponents of the PE model state that funds typically decrease spending in these two areas in order to boost shortterm cash flows and thus improve the valuation of the portfolio company. This phenomenon has become known as “window dressing” and is investigated frequently in the literature. Findings on this matter are rather contradictory as some studies (e.g., Smith, 1990b; Long & Ravenscraft, 1993; Hoskisson & Hitt, 1994) report a reduction in R&D expenditures post-buyout, whereas other studies

8

(e.g., Kohlberg, 1989; Bull, 1989; Lichtenberg & Siegel, 1990) found no support for the “Window Dressing Hypothesis”. Reduction in capital requirements In addition to cost cutting, it is quite common for PE funds to reduce capital requirements in the portfolio firms, (Lowenstein, 1985; Bull, 1989; Baker & Smith, 1998). Typical ways of achieving this include enhancement of working capital management and carving out underperforming assets (Baker & Wruck, 1989; Smith, 1990; Kester & Luchrman, 1995; Phan & Hill, 1995). In particular, inventory control and management of receivables are tightened (Magowan, 1989; Singh, 1990), which leads to a reduction in inventory requirements and lower receivables post-buyout (Easterwood et al., 1989). Holthausen and Larcker (1996) found that portfolio firms had significantly lower working capital compared to their industry peers while under PE ownership. By reducing working capital, the company’s free cash flows (FCF) are improved, and consequently, its valuation rises 14 . This is regarded the main motivation for PE firms to manage this lever. Improvement of asset efficiency Only a few studies have examined this area. For instance, Muscarella and Vetsuypens (1990) reported that PE firms are particularly strong at enhancing asset efficiency post-buyout in their portfolio firms. They studied 72 RLBOs from the U.S. during the 1980s and found that revenues and Asset Turnover 15 (AT) were higher compared to a random sample of publicly traded firms. Muscarella and Vetsuypens (1990) argue that PE firms enhance the efficiency of the assets in place through superior operational engineering, so they ultimately generate higher sales relative to their peers. 3.1.2. Financial Engineering Financial engineering comprises optimization of capital structure and minimization of cost of capital, and has historically been acknowledged as one of the most important value creation levers applied by PE funds (Gottschalg & Berg, 2005; Kaplan & Strömberg, 2008). Disciplining effect of increased leverage The Free Cash Flow Hypothesis (Jensen, 1986; Jensen, 1989; Palepu, 1990) states that the main wealth benefits from LBOs arise from organizational changes which bring about improvements in operating and investment decisions. The hypothesis in anchored in the thought that the management of companies with large free cash flows will be more likely to engage in “empire building,” i.e., growing the firm through unprofitable investments in order to harvest the prestige of running a large company. This phenomenon, naturally, destroys value for the owners and should be avoided. The FCF hypothesis states that by increasing the company’s leverage, owners can discipline managers to be careful when choosing investments, as they have significant interest payments and debt repayments to

14 15

The Discounted Cash Flow model, in which FCF is a key input, is typically used for the valuation. Asset Turnover is defined as Turnover / Total Assets. It is often used as a proxy for asset efficiency in the literature.

9

honour. Thus, the disciplining effect of debt is thought to play an important role as monitoring tool. Jensen (2007) argues that one of the PE model’s benefits is that debt is placed at divisional level rather than at headquarter level, as it is done, in the conglomerate setup for instance. The closer the debt is to the decision makers, the higher a disciplining effect can be expected. Improved capital structure and loan terms The management of PE firms typically consists of ex-investment bankers, who possess an intimate knowledge of capital market mechanisms, and PE firms typically capitalize on the financial expertise of these individuals in the buyout (Anders, 1992). PE firms can assist the management of portfolio firms in several financial activities, such as negotiation of bank loans, bond underwriting, etc. They typically leverage their excellent contacts in the financial community and negotiate terms the portfolio firms would not have been able to obtain stand-alone (e.g., Magowan, 1989; Kaufman & Englander, 1993). The funds’ financial engineering skills help the portfolio firm improve its mix of debt and equity (Anders, 1992). Naturally, this often leads to an increase in debt post-buyout. Tax shield from increased interests In most countries, increased leverage also affects firm value positively through the tax deductibility of interest. However, it is hard to estimate how valuable these tax deductions are. Based on data from the 1980s, Kaplan (1989a) found that tax savings from higher interest payments explain 4–40% of value creation in portfolio firms. These estimates are expected to be lower for buyouts in the 1990s and 2000s, however, since the extent of leverage applied in these deals has declined (Kaplan & Strömberg, 2008). 3.1.3. Governance Engineering Governance engineering mainly revolves around the minimization of agency costs. It is the most exhaustively described and commonly acknowledged lever of value creation (Wright, 2001), and for this reason, no hypothesis is formulated to explicitly test its significance in this thesis. Governance engineering has no directly traceable effect on the bottom line; however, the mechanisms leading to reduction of agency conflict can support the two previously described levers (operational and financial engineering). Benefits related to governance engineering stem from enhanced incentive alignment for the portfolio firm management, higher ownership concentration, and better control of the board and management (Jensen, 1989; Jensen & Murphy, 1990). These elements are outlined in the following sections. Increased ownership concentration and alignment of managerial incentives PE firms pay careful attention to the alignment of management incentives in their portfolio firms. They typically provide large equity upside through stocks and options (Jensen & Murphy, 1990). In fact, Kaplan (1989b) found that management ownership percentages in portfolio firms increase by a factor of four from pre- to post-buyout. In addition, it is normal for PE firms to require management to 10

make a significant investment in the portfolio firms to give them a downside as well. Finally, as the portfolio firms are privately owned, the trading of their ownership stakes is thus illiquid. This serves to reduce management’s incentive to manipulate short-term performance. Even though stock- and optionbased compensation have become more common in public firms since the 1980s, management’s ownership percentages (and upside) remain greater in leveraged buyouts than in public companies (Kaplan & Strömberg, 2008). Jensen (2007) states that the largest gains from this lever are seen in public-to-private transactions, where the increase in ownership concentration is expected to be most significant. Improved controlling PE firms typically control the boards of their portfolio companies, and are more actively involved in governance than are the boards of public companies. Boards of portfolio firms are on average smaller and meet more often, which provides a swifter and more continuous decision making process (e.g., Gertner & Kaplan, 1996; Acharya & Kehoe, 2008). Moreover, boards in portfolio firms are quicker to replace an underperforming management than are public boards. For instance, Acharya and Kehoe (2008) found that one-third of the CEOs in portfolio firms are replaced during the first 100 days, while two-thirds are replaced within a four-year period.

3.2. Change in Relative Significance of the Three Levers of Value Creation As described earlier, there have been two major waves during the existence of PE, and over time the alleged dominant levers of value creation (or returns) have changed. Recall that during the 1980s and 1990s, value was mainly driven by financial and governance engineering. As debt financing got cheaper during the 2000s and industrial companies learned about the benefits of financial engineering, it was argued that “everyone could do financial engineering.” Though the lever was still an important element in the PE model, it had become more of a competitive necessity rather than a differentiating factor (Kaplan & Strömberg, 2008). An increasingly common view regarding value creation today is that PE funds have become experts in operational engineering, through which they create a fundamentally better-performing and healthier company (Heel & Kehoe, 2005; Dobbs, 2006; Kaplan & Strömberg, 2008; Kehoe & Palter, 2009; Matthews et al., 2009; Archleither et al., 2010). This is often backed by the argument that PE firms, to a larger extent than in the past, recruit former industry executives and ex-management consultants to strengthen their operational capabilities. Moreover, PE firms today are generally more specialized within a few industries compared to earlier (Kaplan & Strömberg, 2008).

3.3. Change in Value Creation Potential of the Overall PE Model Besides the change in the relative significance of the value creation levers from the first wave to the second, three factors with potentially great influence on the overall value creation of PE firms changed

11

as well; characteristics of target firms, transaction capital structures, and private-to-private transactions have become increasingly common 16: First, today’s PE firms target not only turnaround companies but also more efficient ones (Jensen, et al., 2006). Two examples include relatively efficient business divisions not reaching their full potential in the larger context they are placed in, and profitable companies positioned to grow but lacking the capital and experience for expansion. Generally, corporations have increased efficiency and tightened cost control since the 1980s, and therefore some of the potential short-term benefits of the PE model seem to have eroded (Wright et al., 2009). Several studies have found the PE model to be more beneficial when target firms are of the turnaround type (e.g., Cuny & Talmor, 2006), which could mean fewer potential benefits from the PE model in the second wave. Hence, PE firms operate in a more competitive environment today, which could mean a lower value creation potential compared to earlier. Second, in recent years PE funds have not increased debt post-buyout in their portfolio firms as much as earlier, which according to The Jensen Hypothesis results in fewer benefits from the disciplining effects of debt (Guo et al., 2011). This is mainly attributed to a relatively high pre-buyout leverage level amongst industrial corporations (potential targets) during the second PE wave. As argued above, industrial managers have become increasingly aware of debt’s value creation potential, and furthermore, debt financing has become relatively cheap (Kaplan & Strömberg, 2008). Third, the proportion of private-to-private transactions has increased relative to public-to-private ones. During the last decade, 80% of deal value and 90% of transactions in Europe were private-to-private (McKenzie & Maslakovic). Ownership in private companies is typically concentrated pre-buyout already, which according to the “Jensen Hypothesis” means fewer potential agency cost savings. It should be noted that despite the structural changes from the first wave to the second, the PE firms’ fundamental motivation was the same: the absence of monitoring within the targeted firms (Renneboog & Simon, 2005; Jensen et al., 2006). Hence, overall the three levers of value creation applied here are the same as earlier (though their relative significance is expected to have changed cf. section 3.2). However, the value potential of the PE model as such is expected to be lower for buyouts undertaken during the second wave - which constitutes the data material of the present analysis – compared to buyouts from the first PE wave.

3.4. Preliminary Conclusions From the Literature Review The majority of the literature reports a positive impact of PE fund ownership on portfolio firm performance. The superiority of PE funds’ abilities to create value stem from three overall levers: operational, financial, and governance engineering. Traditionally, financial and governance 16

PE firms can acquire a target through two type transactions: public-to-private and private-to-private.

12

engineering were given the most importance. Today the significance of financial engineering seems to have declined whereas the importance of operational engineering is argued to have increased. Finally, since the first PE wave in the 1980s, several structural factors – namely, target characteristics, transaction capital structures, and the occurrence of private-to-private transactions – have changed in a way that indicates decreased overall value creation potential for the PE model. So the question is whether the PE model is (still) a superior organizational form, and what the significance of the value creation levers is today. Recall that this is the foundation for the problem statement of this thesis.

4. Hypotheses First, this thesis’s main hypothesis is formulated, which is designed to investigate the overall superiority of the PE model. Next, four supporting hypothesis are set up, intended to explain the findings related to the main hypothesis.

4.1. The Superiority Hypothesis (main hypothesis) PE firms are believed to enhance performance of their portfolio firms relatively more than benchmark (control) firms 17. However, there are indications that the overall value potential of the PE model has declined from the firs to the second PE wave; plus the previous literature has a number of limitations. Thus, it is found highly relevant to test the following hypothesis based on fresh empirical evidence: Hypothesis 1: PE fund ownership induces superior portfolio firm performance relative to other organizational forms (control group)

4.2. The Financial Engineering Hypothesis (support hypothesis) Financial engineering, i.e., optimization of capital structure and minimization of cost of capital, has traditionally been a widely acknowledged value creation lever applied by PE funds (Gottschalg & Berg, 2005). Superior value is expected to be generated by disciplining effects from debt monitoring, unique loan terms facilitated by the management of the PE firm, and corporate tax savings from increased leverage. Hypothesis 2: PE firms are superior at enhancing firm performance through debt monitoring (leverage)

4.3. The Operational Engineering Hypotheses (support hypotheses) The importance of operational engineering has gained increasing acknowledgement amongst researchers during the last decade. As stated earlier, it is typically achieved via cost cutting and margin

17

In this thesis, the applied benchmark is targets with similar characteristics to the portfolio firms, but acquired by industrial corporations. The matching process and applied matching criteria is thoroughly described in section 6.2.2.

13

improvements, reduction of capital requirements, and improvement of asset efficiency. Based on the above literature review, three hypotheses are formulated, related to PE firms’ alleged superior use of operational engineering. Cost cutting and margin improvements First, cost cutting and margin improvement are addressed. As found in the literature review above, PE firms are believed to capitalize on the skills and experience of in-house ex-consultants and industrial executives to tighten cost control, streamline the workforce, and rethink operations within the portfolio firm. This way they manage to reduce bureaucracy and overhead costs by creating a leaner company with improved margins. Therefore, the following hypothesis is formulated. Hypothesis 3: Portfolio firms enjoy higher performance compared to the control firms, as they are subject to tighter cost control Reduction of capital requirements PE firms typically reduce capital requirements though enhancement of working capital management. Inventory control and management of receivables are tightened and streamlined, which leads to a reduction in inventory requirements and lower receivables post-buyout. Often payables are also managed, and longer credit periods are negotiated. Working capital management is generally an overlooked source of value amongst industrial corporations, which historically has enabled PE firms to capitalize on their knowledge within this field. Thus, the following hypothesis is put forth. Hypothesis 4: Post-buyout performance is relatively higher in portfolio firms compared to their industrial peers, due to PE firms’ superior skills within working capital management Improvement of asset efficiency Historically, asset efficiency, conceptualized by “Asset Turnover” has not been given much attention in the literature covering performance in portfolio firms 18 . However, it makes sense to test the significance of it based on fresh data, as the characteristics of the second PE wave differ considerably from those of the first PE wave. AT indicates how much sales volume the management can generate from the company’s asset base, whereas cost cutting and margin improvements (see hypothesis 3) deal with the profitability of the sales. Hence, the two types of key figures complement each other well, and consequently, the following hypothesis is found desirable to test: Hypothesis 5: PE firms enhance the performance of their portfolio firms through superior asset efficiency improvements

18

This could be due to, for instance, limited data availability, as its most common proxy, Asset Turnover, depends on “Turnover,” which is optional for companies to disclose in many countries. However, note that disclosure has improved during the past decade.

14

5. Empirical Strategy In this section, the empirical strategy for testing the hypotheses is outlined. For this purpose the Difference-In-Difference (DID) method is applied. First it is introduced, and arguments for choosing it are provided. Second, the strategy for testing the hypotheses is outlined carefully. Finally, critical elements and limitations of the DID method are described.

5.1. The Difference-In-Difference Method In order to examine the post-buyout impact of PE fund ownership on firm performance, the Difference-In-Difference (DID) method was applied. It is a technique used in econometrics which measures the effect of a treatment or event at a given period in time. The DID estimator represents the difference between the pre/post situation, within-subjects differences of the treatment and control groups19. Thus, it is highly appropriate in contexts where a particular event (in this case, PE buyouts) makes it desirable to investigate an ex-ante and ex-post situation. Moreover, the great appeal of DID estimation stems from its simplicity as well as its potential to circumvent many of the endogeneity problems that typically arise when making comparisons between heterogeneous individuals (Meyer, 1995). In addition, the method is the one most often applied in the previous literature (e.g. by, Smith, 1990; Desbriéres & Schatt, 2002; Vinten, 2007; Guo et al., 2011). Ordinary Least Squares (OLS) is generally used to derive the DID estimate and the related standard errors. This is done in repeated cross sections (or a panel) of data on individuals in treatment and control groups for several years before and after the event of interest (Bertrand et al., 2004). The typical regression is as follows: Yist = As + Bt + cXist + βIst + eist

(1)

Where Yist is the outcome of interest for individual i (e.g., a company) in group s (e.g., PE or industrial ownership) by time t (e.g., a year pre- or post-buyout), Ist is a dummy for whether the event (e.g., PE ownership) has affected group s at time t, As and Bt represent fixed effects for states and year, Xist are controls which are added to enhance the specification of the model, and eist is the error term. β is the DID estimator (also called the treatment effect), which corresponds to the estimated impact of the event (here, PE fund ownership) on the dependent variable of interest (here, firm performance, e.g., Return On Assets (ROA)). Alternatively, the DID table, seen below, can be used to derive the DID estimator. It is less complicated to apply; however, due to its simplicity, it lacks some of the useful features which OLS offers, such as the option to include control variables. Therefore, the table appears in this thesis mainly for illustrative purposes. It illustrates the derivation of the DID estimator, which is found in the bottom

19

The control group and matching process applied in this thesis is thoroughly described in section 6.2.2.

15

right corner. This estimate corresponds to the coefficient β from the regression above, given that no control variables are applied. y11 corresponds to the average pre-buyout observation of the dependent variable of interest (e.g., ROA) for control firms, while y12 shows the same post-buyout. y21 is the average pre-buyout observation for portfolio firms, and y22 post-buyout. Table 1 – Difference-In-Difference Table s denotes the given state and the time t. The lower right corner corresponds to the DID estimator. y is the dependent variable of interest.

yst t=1

s=1 y11

s=2 y21

t=2

y12

y22

Change

y12 - y11

y22 - y21

Difference y21 - y11 y22 - y12 (y22 - y12) - (y21 - y11)

Source: (Bertrand et al., 2004)

Due to the DID table’s simple and inflexible nature, OLS regression is applied to make estimation of more comprehensive models possible. When using OLS in this context, one can use the approach related to equation (1), or make certain alterations to the data at hand in order to circumvent problems associated with autocorrelation. In this thesis, the latter option has been chosen. Thus, the pre- and post-buyout data for each target firm, both control firms and portfolio firms, are collapsed or averaged into one pre-observation and one post-observation, and the difference of these two is used as the observation i. Consider the following example: A given company “i” was acquired in 2007 (by a PE fund or not). Its ROA (and any other variables based on financial numbers applied in this thesis) is computed the following way: 

ROA is defined as: Net Sales / Total Assets



First, ROA was computed for each individual year between 2004 and 2010, given an event window20 of three years.



Subsequently the 2004–2006 measures were averaged into one pre-acquisition observation ROA_prei = (ROAi,t=2006 + ROAi,t=2005 + ROAi,t=2004) / 3



Similarly, the data from 2008–2010 were averaged into one post-acquisition observation ROA_posti = (ROAi,t=2010 + ROAi,t=2009 + ROAi,t=2008) / 3



Finally, the ROA observation for company “i” was computed as: ROAi = ROA_posti – ROA_prei

This computation simplifies the econometric specification since the time series aspect is eliminated, and consequently, the fixed effects can be left out of the model. Moreover, it eliminates common problems related to autocorrelation. A more detailed description of this choice to collapse the data is 20

In order to arrive at the pre/post difference exemplified above, a certain number of years were chosen as foundation for the calculation. This time period is called the “event window”, and for the purpose of this paper, a window of +/ - 3 years was applied. The rationale behind this choice is discussed in section 5.3 along with the DID method’s underlying assumptions and limitations.

16

found in section 5.3. Note that this approach is applied to calculate all variables based on financial numbers used in this analysis. The applied dependent variables are found in table 2 and the independent variables in table 4. Based on the above discussion, the hypotheses of this thesis are analyzed on the basis of the following general econometric model: yi = α + β1PEi + β2LOG_SIZEi + β3LOG_AGEi + β4ROA_PREi + βjINDUSTRYi + βkYEARi + ei (2) Where yi is the set of relevant dependent variables related to firm performance, and PEi is a dummy given the value “1” if company i has been acquired by a PE fund and “0” if has been acquired by an industrial company. LOG_SIZEi, LOG_AGEi, INDUSTRYi, YEARi, and ROA_PREi are relevant control variables commonly used in the literature. All variables applied in the present analysis are described in section 5.1.1 below. Note that when assessing the performance of the portfolio firms, it is crucial to have an appropriate benchmark. Using the DID method, benchmarking is achieved by matching each observation from the treatment group (portfolio firms) with an observation of similar characteristics from a control group (control firms). The applied matching procedure is described carefully in section 6.2.2. 5.1.1. Description of Applied Variables

Dependent Variables The dependent variables investigated in this thesis are defined and abbreviated in the table below. They all serve as measures of firm performance, and it should be emphasized that related studies have used similar measures (e.g., Ravenscraft & Scherer, 1987; Kaplan, 1989a; Smith, 1990; Vinten, 2007; Cressy et al., 2007). Table 2 – Dependent Variables The figure lists the four performance measures applied in the present analysis and how they are abbreviated. The definitions only show which variables have been used to compute the given measure. The actual computation follows the simple aggregation example outlined in section 5.1.

Variable

Abbreviation

Definition

Return on Assets

ROA

Net Earnings / Total Assets

Operating Return on Assets

OROA

EBIT / Total Assets

EBIT-Margin

EM

EBIT / Turnover

Net Profit Margin

NPM

Net Earnings / Turnover

Profit measures Barber and Lyon (1996) and Cressy et al. (2007) argue that operating performance should be measured through operating income (EBIT) rather than Net Earnings, in order not to obscure the measure by 17

special items, tax considerations, and financial elements. However, when assessing the total impact of the PE model, Net Earnings is also an appropriate measure, as some of the benefits from financial engineering (e.g., increased tax shield from high leverage) are not accounted for in EBIT. As EBIT and earnings measure different elements, both will be included in the analysis. EBITDA could also be applied as it represents a measure for operating profit like EBIT. The main difference between the two is that EBIT accounts for the effect of depreciations, and thus CAPEX, whereas EDITDA does not. EDITDA is a good cash flow expression, and thus very applicable when evaluating investments or assets. Moreover, depreciations are often manipulated by management, meaning that EBITDA is likely to be less impacted by “irrelevant” accounting elements. In turn, EBIT is more appropriate when comparing operations across businesses with different characteristics, as it, unlike EBITDA, accounts for CAPEX 21. Hence, EBIT is found more relevant to apply in the present context. Ideally, EBITDA could have been included to boost the robustness of the thesis’ results. However, this was neglected due to data limitations. Scaling of the profit measures In order to compare income across companies, it needs to be scaled. The productivity of the operating assets in place is the parameter of interest here, and therefore it would be ideal to scale operating income by the market value of operating assets (Barber & Lyon, 1996). However, these are not reported separately in financial statements. Alternatively, the market value of total assets (also the theoretical better choice for scaling Net Earnings) could be used, but this measure is not reported in the financial statements either, and only a small fraction of the sample companies are listed. Instead, the book value of total assets is used to scale EBIT and earnings. EBIT / Total Assets is called “Operating Return on Assets” (OROA), and Net Earnings / Total Assets is called “Return on Assets” (ROA). Moreover, it is common to use the average of total assets primo and ultimo, but since a period of six years is already used to construct the observation for each company i, this is redundant. It should be noted that scaling by total assets might be problematic, as firm’s goodwill valuations often change radically post-buyout. This phenomenon boosts total assets and thus induces a downwards pressure on performance ratios, ceteris paribus. Consequently, the impact of PE ownership on portfolio firm performance will be understated. However, most of this asset boosting takes place at holding level (Vinten, 2007), and since this analysis deals with parent companies, it is not considered problematic. Moreover, scaling by total assets should be used with caution if the portfolio firms undertake many acquisitions, as goodwill valuation thereby also changes. However, no literature has been found which provides evidence that portfolio firms are likely to act differently from industrial

21

The classic example is; how do one compare the operating profit of two transportation companies where Company A leases its trucks and Company B owns them. The former company has relatively low depreciations whereas the latter have considerable CAPEX and depreciations. If compared by EBITDA, Company B is likely to show a relatively high operating profit, as a considerable part of its operational expenses (depreciations) are unaccounted for, whereas the leasing expenses defrayed by Company A, have been deducted from the measure. In contrast, EBIT accounts for the depreciations, making the measure more appropriate for comparison.

18

peers. Hence, this effect is assumed to be of equal magnitude across the treatment group (portfolio firms) and control group (industrial firms). Scaling returns by equity was also considered, as it would be an appropriate measure from the viewpoint of the PE funds. However, comparing this measure to performance in control firms is problematic, as these are typically less levered. Note that summary statistics of the present analysis’ sample are presented in section 6.3, where leverage appears to be 8 pp higher in portfolio firms postbuyout. Hence, using return on equity measures would introduce a positive bias in the estimated impact of PE fund ownership. Alternative performance measures Barber and Lyon (1996) also consider alternative measures to operating performance, as ROA and OROA do have some drawbacks. First, total assets are recorded at historic cost, whereas income is recorded in current Euros. A more appropriate measure would be the market value or the replacement costs of total assets (historic cost problem). Second, total assets measures all assets and not only operating ones (non-operating assets problem). And third, operating income is an accrual-based measure easily manipulated by management (earnings problem). The alternatives suggested are Return On Sales (ROS), Return On Market Value of Assets and Cash Flow Return On Assets. Regarding the latter two, they do not suffer from the same drawbacks as those described above, but they are simply not applicable due to lack of data. ROS (which is EBIT and Net Earnings respectively scaled by Turnover) however, is applicable and can overcome the historic cost and non-operating assets problems. However, this measure is not without drawbacks either: ROS does not measure the productivity of assets. For instance, if a firm changes its operations to increase sales (and operating income) while keeping assets constant, the assets become more productive, which OROA or ROA will detect, whereas ROS will not. Nonetheless, the measure detects operating performance changes linked, for instance, to cost reductions in Other Operating Expenses 22 (OOE), or to improvements in production efficiency that reduces Cost of Goods Sold 23 (COGS). Thus, both Return on Sales measures are applied in this analysis. To be able to distinguish them, EBIT / Sales is labelled “EBIT Margin” (EM), and Net Earnings / Sales is labelled “Net Profit Margin” (NPM). An important caveat is that it might be problematic to use “Turnover” or “Employment” for scaling, since it could introduce a bias in the analysis because disclosure is optional in some countries. In order to avoid this bias, elimination of observations due to lack of data availability will not be performed based on variables scaled by Turnover. Instead, the elimination is based on ROA and OROA. If it turns out that an equal number of EM and NPM observations are available, and the data disclosure of the turnover variable is not a problem, the two latter performance measures will be given the same priority as the two former. Since all performance measures have strengths and weaknesses, four have been chosen rather than a single one in order to enhance the robustness of the results. Moreover, three 22 23

OOE comprises costs indirectly related to the production of goods and services COGS are costs directly related to the production of goods and services

19

alternative measures are tested in section 7.2.2, namely Return On Capital Employed (ROCE), Asset Growth (AG), and Sales Growth (SG). Control Variables The control variables (controls) applied in this thesis are listed in table 3 below. These are broadly used in the literature (e.g., Ravenscraft & Scherer, 1987; Smith, 1990; Cressy et al., 2007; Vinten, 2007). Table 3 – Control Variables The figure lists the set of control variables applied in the present analysis plus how they are abbreviated and defined.

Variable

Abbreviation

Definition

Target Size (at buyout time)

LOG_SIZE

Target Age (at buyout time)

LOG_AGE

Log (Total Assets t=0) Log (Buyout Year – Year of Incorporation)

Target's Industry

INDUSTRYj

Ten industry-dummies (NACE classifications)

Year of Buyout

YEARj ROA_PRE

Six year-dummies (2003-2008)

Ex Ante Performance

Pre Buyout ROA (three year average)

Company size (LOG_SIZE) controls for any potential size effects (economies of scale) in the data. Industry dummies (INDUSTRYj) are applied to correct for potential industry variations - e.g., margins differ highly between industries. Ex-ante performance (ROA_PRE) is included as relatively high prebuyout performance is expected to impact performance negatively post-buyout, due to mean reversion in accounting numbers (discussed in section 5.3). These three variables moreover constitute the criteria on which the sample of PE firms is matched with a group of control firms (this procedure is described thoroughly in section 6.2.2). This control group should be seen as a comparable sample of firms having largely the same characteristics as the PE firms, except that they have been acquired by an industrial company rather than a PE fund. Hence, the procedure controls directly for the factors by which the companies have been matched, and indirectly for unobserved factors such as innovative abilities, management’s skills, etc., given that the companies’ similarity on the abovementioned parameters reflects an overall similarity. However, to improve the econometric specification, these matching parameters are also included in the general model as control variables (cf. equation 1). Moreover, it is sensible to control for firm age (LOG_AGE) and year of the deal (YEARj). LOG_AGE is included in order to reduce survivorship bias24 because older firms are likely to perform better as they have survived longer. It is defined as the log of the difference between the year of the acquisition and the year of incorporation. Year dummies (YEARj) are included to control for the different economic environments to which the matched portfolio firm and its control firm have been exposed.

Survivorship bias occurs when an analysis is focused on observations that “survived” some process while systematically excluding those that did not. This typically has the consequence that “successes” are weighted more heavily than “failures” and consequently, a positive bias is introduced. 24

20

Key Explanatory Variables The explanatory variables examined in this thesis are defined and abbreviated in table 4 below. As they are all related to the empirical testing of the hypotheses, they are described and discussed in section 5.2 along with the testing strategies. Table 4 – Key Explanatory Variables The figure lists the explanatory variables applied to analyze this thesis’s five hypotheses. In addition, it shows how they are abbreviated and defined. Note that the computation of all variables other than “PE” and “SD_MON” follows the simple aggregation example outlined in section 5.1. The first category of variables is used to test hypothesis 1 (on the superiority of the PE model), the next hypothesis 2 (on financial engineering), and the last hypotheses 3–5 (on operational engineering).

Variable

Abbreviation

Definition

PE

Equals "1" if target is acquired by a PE fund, and "0" otherwise

Leverage

LEV

Total Debt / Total Assets

Ex Post Leverage

LEV_POST

Post Buyout Leverage (three year average)

Short Term Debt

SD

Short Term Debt / Total Assets

Short Term Debt Monitoring

SD_MON

Equals "1" if overall leverage and proportion of short term debt increase post buyout, and "0" otherwise

Cost of Goods Sold Margin

COGSM

Cost of Goods Sold / Turnover

Other Op. Expenses Margin

OOEM

Other Operational Expenses / Turnover

Working Capital Ratio

WCR

(Inventory + Debtors - Creditors) / Total Assets

Asset Turnover

AT

Turnover / Total Assets

Superiority Hypothesis Private Equity Ownership Financial Engineering

Operational Engineering

5.2. Strategy for Hypotheses Testing In the forthcoming section, various regressions based on the general econometric model are set up in order to test The Superiority Hypothesis (H1), The Financial Engineering Hypothesis (H2) and The Operational Engineering Hypotheses (H3-H5). 5.2.1. Empirical Testing of The Superiority Hypothesis Hypothesis 1: PE fund ownership induces superior portfolio firm performance relative to other organizational forms (control group) First, it is tested whether PE firms on average are capable of creating higher performance in their portfolio firms relative to industrial equivalents (control group), i.e., whether the PE model is superior in terms of value creation. The test is performed using the regression below, which corresponds to the general specification in section 5.1. yi = α + β1PEi + β2LOG_SIZEi + β3LOG_AGEi + β4ROA_PREi + βjINDUSTRYi + βkYEARi + ei

21

The key variable of interest is PEi, which equals “1” if the company has been acquired by a PE fund and “0” if the buyer was an industrial company. β1 represents the DID estimate, controlled for effects related to company size, age, industry, year of the transaction, and ex-ante performance. In fact, this set of control variables is applied in all of the regressions below. According to the literature review and particularly The Jensen Hypothesis, β1 is expected to be significantly positive. However, recall from section 3.3 that several structural factors in have changed since this hypothesis was formulated 25 all indicating a lower value creation potential for the PE model 26 (i.e. fewer portfolio firms are of the turnaround type, deals are less levered, and fewer public-to-private transactions are undertaken). 5.2.2. Empirical Testing of The Financial Engineering Hypothesis In order to test hypothesis 2, three steps are applied. First, it is tested whether the debt ratio postbuyout has an impact on firm performance, i.e., if a high debt ratio can explain high firm performance. Second, it is tested how a change in capital structure around the buyout impacts firm performance. Finally, it is tested whether short-term financing has an especially strong disciplining effect and thus leads to improved firm performance. Hypothesis 2: PE firms are superior at enhancing firm performance through debt monitoring (leverage) Note that leverage is defined as Total Debt / Total Assets. A theoretically more sound definition is Net Interest Bearing Debt / Total Assets. This measure equals interest-bearing debt minus cash and cash equivalents. Consequently, items such as account payables, provisions, etc. are excluded from the numerator. As these items do not carry interest, their disciplining effect is expected to be relatively low. However, due to data limitations the latter definition could not be applied. The first step is tested by running the regression below. The key explanatory variable is the interaction term between the PE ownership dummy (PEi) and the debt-to-asset ratio post-buyout (LEV_Post)27. This term measures the effect of PE fund ownership combined with the firm leverage level. If hypothesis 2, and thus the Free Cash Flow Hypothesis holds, β1 should be significantly positive. yi = α + β1(PE*LEV_POST)i + β2LEV_POSTi + βjControls + ei The second step is to test whether an increase in leverage post-buyout enhances portfolio firm performance. In order to determine this, the regression below is applied. The interaction term between

25

Most of the PE performance literature is based either entirely or partially on data from the first wave. One, corporations have generally increased efficiency and tightened cost control since the 1980s, and therefore some of the p otential shortterm benefits of the PE model seem to have been eroded (Wright et al., 1997). Two, the capital structures of buyouts today are less levered compared to earlier, which according to The Jensen Hypothesis results in fewer benefits from the disciplining effects of debt and fewer tax savings. Three, in Europe today, 90% of PE transactions are of the private-to-private type, and according to The Jensen Hypothesis, fewer potential agency cost savings are expected to be associated with this type of transaction. These three factors all indicate a lower value creation potential for the PE fund compared to earlier. 27 LEV_POST is calculated as the three-year post-buyout average of leverage, which is defined as Total Debt / Total Assets. 26

22

being owned by a private equity fund (PEi) and change in leverage from pre-buyout to post-buyout (LEV)28 is the key explanatory variable. It measures the effect of PE fund ownership together with being exposed to debt monitoring and more directly tests The Free Cash Flow Hypothesis. For it to hold and for the results to support hypothesis 2, β1 has to be significantly positive. yi = α + β1(PE*LEV)i + β2LEVi + βjControls + ei The third step takes the debt composition into consideration. According to The Jensen Hypothesis (Jensen, 1986 + 1989), it is not the total amount of debt outstanding per se but rather the amount of debt service payments per time period that disciplines or motivates managers to work harder. Thus, the structure or terms of the debt could play an important role in how effectively leverage motivates managers. Debt with a shorter maturity increases the debt service payments per time period (Cotter & Peck, 2001) and is thus expected to increases the incentives for managers to work harder to increase firm value in the early stages of the LBO. Consequently, the third step is aimed to tests whether the disciplining effect of short-term financing is higher than for long-term debt. This is tested though the regression below, where SD_MON is a proxy for short-term debt monitoring. It equals “1” if the overall leverage (LEV) and the share of short term financing have increased post-buyout; otherwise it equals “0”. The interaction term captures the effect on portfolio firm performance of increasing leverage not only around the buyout, but in particular through short-term financing. If hypothesis 2 holds, β1 is expected to be significantly positive. yi = α + β1(PE*SD_MON)i + β2SD_MONi + βjControls + ei As described previously, a substantial fraction of the debt is typically placed at holding company level, which could constitute a problem as this analysis is carried out at parent company level. According to Jensen (2007), PE firms typically place debt at parent level in order to extract its disciplining effects, whereas industrial acquirers more commonly place it at holding level. This has implications for the analysis, as the post-buyout leverage of portfolio firms and control firms are not entirely comparable. However, although we might not see the full impact of increasing leverage on firm performance, it should still be possible to obtain some useful insight on the matter. In addition, recall that companies in general (i.e., potential targets) were more levered pre-buyout during the second PE wave compared to earlier, as the benefits of financial engineering had become known to most executives (Kaplan & Strömberg, 2008). Consequently, the increase in leverage post-buyout was lower here, compared to when the free cash flow hypothesis was formulated in 1989 in the wake of the first PE wave. Hence, it is likely that low or no gains from financial engineering will be found.

28

LEV = LEV_POST – LEV_PRE where “LEV_POST” is calculated as the three-year post-buyout average of leverage and “LEV_PRE” is the three-year pre-buyout average. Leverage is defined as Total Debt / Total Assets.

23

5.2.3. Empirical Testing of The Operational Engineering Hypotheses Three hypotheses were formulated in section 4 to test the superiority of PE firms in relation to operational engineering: Hypothesis 3: Portfolio firms enjoy higher performance compared to the control firms, as they are subject to a tighter cost control Hypothesis 4: Post-buyout performance is relatively higher in portfolio firms compared to their industrial peers, due to PE firms’ superior skills within working capital management Hypothesis 5: PE firms enhance the performance of their portfolio firms through superior asset efficiency improvements Empirical testing of hypothesis 3 – Cost cutting and margin improvements Hypothesis 3 is investigated in order to test the claimed superiority of PE funds’ ability to enhance performance through cost cutting. The hypothesis is tested by running the two regressions below. The key variable of interest in the first regression is PE*COGSM, which captures the impact of PE fund ownership (PE) combined with the effect from the Cost of Goods Sold Margin (COGSM)29. If β1 is significantly positive, hypothesis 3 is supported. yi = α + β1(PE*COGSM)i + β2COGSMi + βjControls + ei The second regression generally tests the same aspect. The only difference is that COGSM has been replaced with OOEM 30, which is short for “Other Operational Expenses Margin.” Hence, the impact of reducing these costs is tested instead of COGS. yi = α + β1(PE*OOEM)i + β2OOEMi + βjControls + ei COGS includes costs directly related to the production of the goods sold, such as raw materials, direct labour, etc. whereas OOEM comprises all costs indirectly related to the production of goods sold, such as commercial costs, administrative expenses, etc. Hence, it is sensible to test both measures. Empirical testing of hypothesis 4 – Reduction of capital requirements The aim of investigating hypothesis 4 is to test the claimed superiority of PE funds’ ability to create value by reducing capital requirements. The hypothesis is tested by running the regression below.

COGSM is defined as COGS / Turnover and computed as follows: COGSM = COGSM_POST – COGSM_PRE where “COGSM_POST” is calculated as the three-year post-buyout average and “COGSM_PRE” is the three-year pre-buyout average. 30 OOEM is defined as OOE / Turnover and computed as follows: OOEM = OOEM_POST – OOEM_PRE where “OOEM_POST” is calculated as the three-year post-buyout average and “OOEM_PRE” is the three-year pre-buyout average. 29

24

Working Capital Ratio (WCR)31 is defined as “inventory + creditors – debtors” scaled by total assets, and serves as a proxy for capital requirements. The scaling is necessary in order to make the measure comparable across companies and has been applied by other studies (e.g., Bergström et al., 2007). The key explanatory variable of interest is PE*WCR, which captures the effect of PE fund ownership (PE) combined with the working capital effect (WCR). If β1 is significantly positive, hypothesis 4 is supported. yi = α + β1(PE*WCR)i + β2WCRi + βjControls + ei Empirical testing of hypothesis 5 – Asset efficiency Hypothesis 5 is investigated in order to assess whether PE firms are more capable than control firms of enhancing portfolio firm profitability by improving asset efficiency. It is tested through the regression below, where PE*AT is the key variable of interest. It measures the effect of PE fund ownership (PE) combined with the asset turnover effect (AT)32. Asset Turnover is defined as Sales / Total Assets, which is commonly used as a proxy for asset efficiency in the literature (Muscarella & Vetsuypens, 1990). If β1 is significantly positive, hypothesis 5 is supported. yi = α + β1(PE*AT)i + β2ATi + βjControls + ei The previous literature (e.g., Muscarella & Vetsuypens, 1990; Bergström et al., 2007), as well as the alleged increased focus on operational engineering within PE firms, bring about expectations of positive effects from the variables described above. However, we also know that corporations generally have increased efficiency and tightened cost control over the last decades, and that consequently, many of the short-term benefits from operational improvements seem to have been erased. Thus, the expected significance of these levers seems ambiguous.

5.3. Critical Elements and Limitations of the DID Method In the following section critical elements and limitations of the DID method are described and discussed. These include the choice of event window, the underlying assumptions (particularly the parallel trend assumption), targeting based on differences, the importance of correcting for autocorrelation, and practical limitations. The choice of event window When deciding on the length of the event window, it is mainly a trade-off between biases caused by the so-called J-curve effect and sample size (due to low data availability when a wide event window is chosen). Here, this trade-off is discussed and a window of +/- 3 years is eventually chosen. WCR = WCR_POST – WCR_PRE where “WCR_POST” is calculated as the three-year post-buyout average of Working Capital Ratio and “WCR_PRE” is the three-year pre-buyout average. Working Capital Ratio is defined as “inventory + creditors – debtors” scaled by total assets. 32 AT = AT_POST – AT_PRE where “AT_POST” is calculated as the three-year post-buyout average of Asset Turnover and “AT_PRE” is the three-year pre-buyout average. Asset Turnover is defined as Turnover / Total Assets. 31

25

The J-curve effect is used to describe the historical tendency of firm performance not to be improved until after the fourth or fifth year of ownership (Wright & Robbie, 1998). Typically, in the post-buyout period, underperforming assets are written off and turnaround activities are undertaken, which can initially pull firm performance in a negative direction. In contrast, positive effects from sources such as synergies and operational improvements take time to materialize. However, it should be noted that both the PE firms and control firms are evaluated based on the realized changes three years postbuyout, and no literature has been found which indicates that PE firms should be either slower or faster at implementing these. Thus, if this J-curve effect is present in my data, both types of firms are expected to have similar exposure. Another consequence of this is that it becomes less likely to detect improvements in performance within the individual groups, but as long as they are affected equally, the DID estimator will not become biased by the J-curve effect. Ceteris paribus, the potential presence of a J-curve effect favours a relatively long event period that captures as much of the PE ownership impact as possible. However, the longer the event period, the higher the data limitations. The Orbis database (from which all accounting data used for the present analysis were extracted) only covers the time period from 2000 to 201133, which means that if an event window of +/- 5 years is chosen, only deals made in 2005 and 2006 are applicable. Another problem resulting from a long event window is that PE funds usually own their portfolio companies for a period of three to seven years, meaning that several deals would have to be excluded from the sample due to relatively early PE fund exit. Thus, based on the discussion above, a window of +/- 3 years is applied. The parallel trend assumption The DID method requires all the same assumptions as OLS estimation with the addition of a “parallel trend assumption,” which can be explained as follows: The DID estimate is an unbiased estimate of the impact of the event if, absent the event, the average change in y (yt=2 − yt=1) would have been the same for the treatment and the control groups. Put differently, given that there is no treatment, over time the two groups are affected by the same things/events and should thus follow the same trend. This assumption is broken if something other than the treatment affects the two groups differently over the event window. For instance, the group of portfolio firms and control firms has not been matched on year of transaction, which is likely to be problematic. Imagine how the financial crisis could have impacted a portfolio firm but not necessarily its matched counterpart (control firm). Fortunately, it seems as if this asymmetry has been averaged out, as the two groups show rather identical distributions related to year of transaction (illustrated in section 6.3.3). Naturally, in economics it is nearly impossible to create two groups which are identical in every aspect except for the treatment, but a well-planned matching procedure and econometric specification of the applied regression models can potentially control for most of the differences. Finally, recall that year dummies are included in the general econometric model, and that the matching procedure applied ensures similarity in terms of 33

2012 data was not available in Orbis when the data collection for this thesis concluded.

26

industry, company size, and pre-acquisition performance. Thus, based on the above discussion, it is not considered likely that the parallel trend assumption is violated in this thesis’s data. Targeting based on differences If unaccounted for, “targeting based on differences” can be one of the DID method’s most restraining features (Meyer, 1995). The term “Ashenfelter’s dip”34 is used to describe the situation where the observed variable (in this case firm performance: ROA, OROA, etc.) takes a dip just prior to “treatment” (here, getting acquired by a PE fund). Cuny and Talmor (2006) found that the PE model works best on turnaround firms, i.e., firms in which performance has taken a dip. PE funds have traditionally had a preference for this type of firm, and consequently, it seems reasonable to expect Ashenfelter’s dips to be present in the data of this analysis. Such a dip complicates measurement of the treatment effect (Meyer, 1995), and since accounting numbers have a natural tendency to mean reversion (Barber & Lyon, 1996), this could potentially lead to an upward bias of the DID estimate of the effect of PE ownership on firm performance. However, recent research points towards a change in target characteristics, and today a broader range of firms are bought by PE funds (Jensen, et al., 2006) – a trend which is expected to limit the problems related to the dips in ex-ante performance. And more importantly, the matching method applied in this thesis controls for them. Even if portfolio firms are characterized by poor ex-ante performance (i.e. are turnaround firms), equally poorly-performing control firms have been picked, as ex-ante performance is used as a high-priority matching criteria. In section 6.3, the descriptive statistics of this thesis’s data are outlined, and it is found that the average ex-ante ROA (ROA_PRE) for portfolio firms is 7,97%, which is very close to 7,84% observed in control firms. Hence, Ashenfelter’s dip is a common problem in data used along with the DID method in general, but limited influence is expected in this analysis due to the nature of the matching procedure applied. Autocorrelation typically a problem Bertrand et al. (2004) investigates 92 papers in which the DID method was applied. They find that the greatest problem related to the DID method is that researchers tend to fail correcting appropriately for autocorrelation in the data. Consequently, the conventional DID standard errors may understate the standard deviation of the estimated treatment effects, which in turn leads to severe overestimation of tstatistics and significance levels. They suggest various solutions to this problem depending on the nature of the analysis at hand. One of these is labelled “simple aggregation” and offers a simple way of circumventing problems related to autocorrelation. The idea is simply to ignore time series information when computing standard errors by collapsing (averaging) time series data into one ex-

Ashenfelter and Card (1985) described a phenomenon which later became known as “Ashenfelter’s dip.” At that time it was commo n to compare wage gains among participants and non-participants in training programs to evaluate the effect of training on earnings. They found that training participants often experience a dip in earnings just before they entered the program (presumably why they enter ed the program in the first place), which then caused the DID estimate to be biased. 34

27

ante observation and one ex-post observation. The solution is popular due to its simplicity and because it produce consistent standard errors. A drawback of this approach is that its statistical power is relatively low and diminishes fast with sample size. However, the size of the sample drawn for the present analysis is rather large (130 portfolio firms and 130 control firms), and thus statistical power is not expected to be a considerable problem.

Thus it is found sensible to apply simple aggregation to avoid problems related to

autocorrelation. Brief summary of the above section The DID method’s limitations are assessed to be of relatively low significance in the setup of this analysis. Moreover, there is little reason to suspect violations of the parallel trend assumption. Therefore, it is sensible to maintain the DID method in combination with simple aggregation as the main hypothesis testing tool.

6. Data This section deals with the data used in the present analysis. First, the data sources are described. Next, the sample selection and matching procedure are outlined. Finally, the descriptive statistics and sample characteristics are described and discussed.

6.1. Data Sources The data was drawn from the Zephyr and Orbis databases (both administered by Bureau Van Dijk), which are rather closely linked together, as deal data from Zephyr can easily be transferred to Orbis in order to obtain company-specific data for the involved parties. Zephyr is a comprehensive database of global deal information. It contains information on mergers and acquisitions (M&A), initial public offerings (IPOs), venture capital (VC) and private equity (PE) deals. Orbis is the most comprehensive database that exists on company information (particularly in terms of data on privately owned companies). It contains firm-level information with emphasis on accounting numbers for more than 100 million companies worldwide. An important caveat is that individual countries have different rules in regards to disclosure of various attributes. In Denmark, for instance, firms are not obligated to disclose information related to turnover and employment (however, most do anyway). Thus, in order to reduce the risk of introducing biases, variables which are constructed using such attributes were given a lower priority in the analysis, and caution was taken when the sample was reduced.

28

6.2. Sample Selection The following section describes the sample selection process by addressing first the sampling strategy for the portfolio firms, next the matching procedure, and lastly comments on the final data set. 6.2.1. Sampling Strategy for the Portfolio Firms First, a sample of acquisitions made by PE firms within Europe in the time period of 01/01/2003– 31/12/2008 was drawn from the Zephyr database, covering the entire second PE wave. Moreover, as company specific data had to be drawn from Orbis later, the database’s coverage sat up a restrain on the time period from which deal data could be drawn. The Orbis database only contains historical company data from 2000 to 2011, and deals made in the first and last three years of this period had to be excluded to allow for the event window (cf. section 5). In addition, the following criteria were applied. First, only fully completed deals of the type “institutional buyout” were included, to filter out deals performed by organizations other than PE funds. This deal type is defined as “an acquisition where a PE firm has taken a 50% stake or more in the Target company.”35. Second, to ensure the deals included were examples of significant change in ownership, only cases where the acquirer had an initial ownership stake of 0% were chosen. This criterion ensures that the “change in ownership” reflects an actual change in managerial influence and thus decision making. In addition, it cleans for cases where PE fund ownership had impacted the portfolio firm prior to the transaction date. Third, cases where the same target had undergone more than one ownership change within the event window were excluded to avoid interference of effects from several ownership changes. For this purpose, a threshold of 10% change in ownership was chosen. Fourth, companies with missing data in regards to basic information such as “industry code” or “date of incorporation” were excluded. See appendix 2 for documentation of the applied search criteria in Zephyr. After completing the steps above, the sample included 2791 deals. For each target company a large range of data (mainly accounting data), was drawn from the Orbis database. Though the database is comprehensive, not all relevant data concerning the target companies were available, and consequently, the sample was reduced to 379 companies. For all the remaining companies, valid observations 36 related to the two main performance variables (ROA and OROA) were available. Regarding the other variables used in the analysis (e.g., Leverage, Working Capital Ratio, etc.) valid observations were not available for all target companies. Thus, the sample sizes vary between the tests performed, depending on which variables are analyzed. This thesis investigates the impact of PE fund ownership on parent company level. Hence, data related to the portfolio companies were gathered at this level and not holding company level. The parent company is the lasting entity, whereas the holding company is established just prior to the acquisition 35

The Zephyr database help function The term “valid observation” is used to describe a situation where a target firm has disclosed data regarding a given variabl e both three years prior to and after the buyout. Thus, disclosed data for the whole event window were available. This is in contrast to an invalid observation where one or several observations in the event window are missing. 36

29

and closed down after the exit of the PE fund. Consequently, no pre-buyout data exists at holding company level. 6.2.2. Matching Procedure Drawing the control group When assessing the performance of the portfolio firms, it is crucial to have an appropriate benchmark. Using the DID method, benchmarking is achieved by matching each observation from the treatment group (portfolio firms) with an observation of similar characteristics from a control group (control firms). Thus, a sample of acquisitions made by industrial companies was drawn using identical steps as described above. The only change was in the parameter “deal type” from “institutional buyouts” to “acquisitions.” The unmatched control group contained 3641 firms (control firms). Subsequently, an elimination process identical to that of the portfolio firms was applied, providing a sample of 2466 control firms with “valid observations” for the two key dependent variables, ROA and OROA. Quality check At this point, two checks were performed to assess the quality of the data from Zephyr and Orbis. First, a test was undertaken to ensure that all targets in the PE group were actually acquired by PE funds. Secondly, the accuracy of the disclosed accounting numbers was investigated. From the data output provided by Zephyr, it is not possible to assess directly whether a particular acquisition has been performed by a PE fund or not. Consequently a sample of 20 acquirers from the sample of 379 PE transactions was drawn, and each acquisition was investigated individually by using a number of resources: E.g. the web, Orbis’ database on detailed deal information, various article databases and Mergermarket.com. It was positive to find that all 20 acquirers turned out to be PE funds. Two of the deals comprised minor MBO elements, but these were assessed insignificant, and consequently, they were not excluded from the sample. A summary of this check is found in appendix 5. Next, the accuracy of the accounting data from Orbis was assessed by matching the downloaded data, with the actual numbers from published annual reports for 20 randomly selected portfolio firms. This analysis indicated a satisfying accuracy of the data provided by Orbis - cf. the summary of the check which is found in appendix 6. Both checks point towards that reliability and quality of the applied data is relatively high. Hence, the present analysis is continued applying this. Matching of the treatment and control group Each portfolio firm was then matched with the best-fitting control firm, based on ex-ante performance, industry, and company size as suggested by Barber and Lyon (1996). They tested various matching procedures related to firm performance studies and found that matching on these three criteria delivers the most well-specified test statistics. They also found that matching on industry and performance is significantly more important than matching on company size.

30

The reasoning behind matching on industry and size is rather straightforward, whereas ex-ante performance is somewhat more complicated. Industry matching controls for structural differences between industries and ensures that the two matched companies are exposed to similar business conditions. The size parameter is included to control for potential economics of scale amongst the sample firms, as research by Fama and French (1995) demonstrates that large firms perform better than smaller equivalents. Matching on performance adjusts for the mean reversion in accounting numbers, which is linked to a transitory component of income – a phenomenon described by Penman (1991) and Fama and French (1995). This component is mainly attributed to accounting methods such as manipulation of numbers or changes in accounting practice or to underlying economic forces such as nonrecurring income or expenses or temporary shifts in product demand. If a company experiences a relatively high ex-ante performance, it is likely that at least some of it is caused by such a transitory component, which will dissipate over time. In addition, the criteria controls for unobserved factors such as skill of management and access to unique resources. As outlined in section 5.1, this ability to circumvent endogeniety problems, is in fact one of the DID methods major advantages. The matching in this thesis is undertaken as follows. Each portfolio firm is matched with a control firm with the same two-digit NACE core code (industry), and the closest possible ROA_PRE (ex-ante performance) within a range of +/- 50% of Total Assets at the time of the buyout (firm size). This provides a total sample of 130 portfolio firms and 130 control firms. The choice of strictness related to the matching criteria is naturally a trade-off between size of the matched sample and similarity of the companies in the portfolio group and control group. The stricter the matching criteria, the smaller but more comparable the sample. Note that applying overly strict criteria could introduce a selection bias. This phenomenon is discussed in section 7.4.2. Calibration of the data set From the 130 matched pairs, only observations related to variables where both the portfolio and control firm could present “valid observations” were included in the sample (i.e., on variable level). For example, a given matched pair of firms may both have valid ROA observations, but in terms of leverage, only the portfolio firm has a valid observation. In this case, the matched pair’s ROA observations will figure in the sample, but none of the leverage observations will (this procedure is labelled “double elimination”). This was done to prevent one group from being overrepresented in the sample, which could introduce a bias. Next, an elimination of outliers was performed in order to reduce bias related to these. Here the same approach of double elimination as described above was applied. This procedure resulted in smaller sample sizes (on variable level) compared to what alternatively could have been obtained by only eliminating the sole extreme observation. However, as the initial sample is relatively large compared to similar studies (e.g., Kaplan, 1989a; Cressy et al., 2007; Guo et al., 2011), a good match is prioritized above the loss of observations. The specific outlier elimination approach can be found in appendix 4.

31

More appropriate benchmark Compared to earlier studies (e.g., Kaplan, 1989a; Smith, 1990; Desbriéres & Schatt, 2002; Cressy et al., 2007; Guo et al., 2011), it is believed that the benchmark applied in the present analysis is more appropriate. First, when the deals were drawn from Zephyr, it was a criterion that the acquirer took a minimum stake of 50% in the target and initially had a 0% stake, whereas the aforementioned studies are considerably less strict. This criterion ensures that the “change in ownership” reflects a change in managerial influence and thus decision making. Second, the matching in this thesis was based on exante performance, which none of the above studies apply. This criterion is described by Barber and Lyon (1996) as the significantly most important one, in combination with “industry.” Third, I have obtained a relatively large sample of closely matched companies compared to other studies with similar focus (e.g., Kaplan, 1989a; Cressy et al., 2007; Guo et al., 2011), allowing me to perform test on smaller sub sample while maintaining a relatively high statistical power. 6.2.3. Comments on the Final Data Set Content of the data set and variable definition The final data set contains all available income statement and balance sheet data, plus acquirer name, target name, date of the deal, target’s year of incorporation, and target’s NACE Rev. 2 industry code. The financial data enabled me to define the set of variables described in section 5.1.1 on a simple aggregation basis using an event window of three years. Potential selection biases Certain elements of the sampling process can potentially have introduced selection biases. The criteria for an observation to be “valid” (i.e., availability of three years’ continuous data pre- and post-buyout) could for instance have introduce survivorship bias37. Moreover, it could increase the bias related to country representation, as the quality of Orbis’s data coverage and disclosure regulations differ significantly across countries. Thus, continuous data is more likely to be available for certain countries. These issues along with other alterative explanations to the present analysis’ main findings are addressed in section 7.4. Distinctive features of the data set This data set is considered somewhat unique mainly for three reasons. One, Orbis is the most comprehensive database on company information (especially private companies) in existence, and until recently, researchers had not been able to utilize this for similar studies. Two, access to historical data from before 2007 is limited and filled with obstacles. Three, recent European legislation ensures relatively good disclosure of financial data for private companies.

Survivorship bias occurs when an analysis is focused on observations that “survived” some process while systematically excluding those that did not. This typically has the consequence that “successes” are weighted more heavily than “failures” and consequently, a positive bias is introduced. 37

32

First, it is not likely that studies with a similar focus as this thesis were able to utilize the Orbis database just a few years ago. Good practice when applying the chosen DID approach requires the researcher to set up an event window of minimum +/- 3 years, and with Orbis’s total coverage period in mind (2000–2011), a sample today can only contain deals made between 2003 and 2008 (six years), which limits the amount of previous papers that can have utilized the database. Second, the part of the data published prior to 2007 is not fully available online. Consequently, researchers need to gain access to offline data, as it does not make sense to carry out an analysis like this using only data from 2007 to 2011. However, the software used for this procedure is outdated and highly incompatible with modern operating systems. Fortunately, I have been able to obtain data covering the full period (after numerous attempts). Third, it is desirable that the data set reflects the nature of the transactions in reality. Recall that 90% of European PE transactions over the last decade were private-to-private, and that disclosure of financial data related to these transactions traditionally has been limited. However, legislation in many European countries has changed during recent years to ensure higher transparency of private companies’ financials. This has enabled me to collect a sample where 92% of the portfolio firms were private before the buyout, which differentiates this thesis from the majority of historical studies, especially U.S.-based studies, as disclosure has not reached the same level there (Cressy et al., 2007).

6.3. Descriptive Statistics and Sample Characteristics In this section the summary statistics of the sample are discussed. Table 5 contains the three-year averaged ex-ante and ex-post measures for the portfolio and control group respectively38, pre- and post-buyout differences between the groups, and preliminary Difference-In-Difference estimates (using the simple DID table approach outlined in the method section). The data relate to firm size, growth, performance, capital structure, and operations. 6.3.1. Pre-Buyout Descriptive Statistics First, table 5 reveals that portfolio firms are significantly larger than control firms, as measured by both total assets and sales, which indicates that PE firms aim to acquire industry leaders to a larger extent than industrial buyers. The notable difference in size is rather surprising, taking the matching procedure into account. One could criticize the matching criteria for not being sufficiently strict; however, tightening them further would have resulted in a significant decrease in sample size. Moreover, Barber and Lyon (1996) found that matching strictly on pre-buyout performance is considerably more important than matching on firm size. This size difference could perhaps be explained by a fundamental difference in acquisition motive between the two groups: Industrial buyers often acquire single business divisions, or relatively small “add-on” targets motivated by a wish to incorporate them into their larger main operation. This type of transactions is much rare amongst PE 38

It should be noted that data related to the year of the ownership change have been left out of the analysis, mainly for two reasons. One, considerable changes are usually implemented in the target firms during the transaction year, and often performance suffers a s managers are occupied with the transaction activities and cannot fully focus on the daily operations of the company. This provides a downwards bias in firm performance (Vinten, 2007). Two, it is problematic to decide whether this particular year should be included as part of the pre- or postbuyout observation.

33

funds, as they naturally do not have any long term fundamental operations to which they can add-on smaller companies. Instead, they typically acquire full stand alone businesses. Table 5 – Summary Statistics of the Sample The variables in the first column are defined in section 5.1.1. Total assets and sales are measured in Euro (thousands). The column labelled “Pre-buyout” shows the three-year average of the particular measure prior to the buyout. “Port.” refers to portfolio firms and “Con.” refers to control firms. In the same fashion, the “Post-buyout” column refers to the three-year average post-buyout. Below, the standard deviation and sample size are reported. The column labelled named “Pre-buyout difference” contains the difference between portfolio and control firms pre-buyout, and “Post-buyout difference” refers to the difference between the two post-buyout. “DID estimate” refers to the Difference-In-Difference estimate for the given measure. For each of the last three columns, the differences are preliminarily tested through the use of ANOVA, and the results are reported as follows: ***, **, and * indicate that the given measure is significant at a 1% level, 5% level, and 10% level, respectively. The standard errors from the tests are reported below. Pre buyout Port. Con. Size Total Assets (TA) Sales (SALES) Growth Assets (AG) Sales (SG)

30.700

58.627

34.574

17.825

113.752

64.794

123.723

67.688

5.756

130

130

130

130

54.835

35.223

61.071

39.630

19.612

117.933

63.926

114.722

76.497

5.903

6.072

130

130

130

130

0,123

0,123

0,094

0,045

0,000

0,049

0,162 115

0,182 115

0,195 115

0,143 115

0,108

0,150

0,048

0,086

Return on Assets (ROA) EBIT Margin (EM) Net Profit Margin (NPM) Financial levers Leverage (LEV) Long Term Debt (LD) Short Term Debt (SD) Operational levers Cost of Goods Sold Margin (COGSM) Other Operating Exp. Margin (OOEM) Working Capital Ratio (WCR)

Others Age at time of buyout (AGE)

0,133 101

0,297 101

*

21.442

*

0,016

0,107

0,109

0,079

0,009

0,030

0,093 130

0,152 130

0,118 130

0,006

0,008

0,080

0,078

0,075

0,046

0,001

0,029

0,084 130

0,082 130

0,120 130

0,102 130

0,005

0,007

0,068

0,146

0,079

0,110

-0,078

0,086 107

0,089 107

0,115 107

0,053

0,049

0,053

0,068 115

0,086 115

0,100 115

0,627

0,583

0,638

0,209 98

0,189 98

0,193 98

0,191

0,164

0,228

0,189 73

0,176 73

0,204 73

0,498

0,492

0,223 124

0,209 124

0,542 0,214 101

***

-0,031

0,006

0,008

0,030

0,004

0,023

0,079 115

0,005

0,006

0,561

0,044

*

0,078

6.228 1.830 1.967

**

0,049 * 0,015

0,004 0,021

*

0,021 0,008

**

0,028 ** 0,007

**

0,047 *** 0,007

*

0,019 0,006

***

0,034

0,014

0,014

0,013

0,186

0,027

0,043

0,016

0,180 73

0,015

0,016

0,016

0,474

0,468

0,007

0,006

-0,001

0,208 124

0,220 124

0,011

0,014

0,477

0,497

0,475

0,065

0,022

0,225 101

0,190 101

0,219 101

0,026

0,024

0,359

0,304

0,325

0,337

0,054

-0,012

0,246 99

0,222 99

0,225 99

0,198 99

0,025

0,023

0,254

0,244 0,221 122

0,191 98

*

-0,038

0,013

DID estimate

2.183

0,011

0,113 130

0,111 107

*

6.218

0,011

-0,042

24.053

0,117

0,204 122

Asset Turnover (AT)

0,216 101

*

Post buyout difference

48.525

0,141 101

Performance Operating Return on Assets (OROA)

Post buyout Pre buyout Port. Con. difference

0,217

0,271

0,179 122

0,236 122

0,010 0,014

-0,054 0,014

0,011

-0,043 * 0,013

-0,066 ** 0,016

**

-0,064 *** 0,009

1,64

1,57

1,50

1,47

0,07

0,03

-0,04

0,990 119

0,950 119

1,012 119

0,887 119

0,067

0,067

0,028

26

25

27 130

22 130

1

Source: Orbis and own calculations

34

Second, the differences in growth as measured by assets appear to be insignificant, whereas the control firms have higher sales growth at a 10% significance level. This could indicate that industrial firms aim for relatively small growth firms compared to PE firms, which target larger and more consolidated companies. This is consistent with the findings above related to firm size. Third, firm performance as measured by OROA, ROA, and NPM is rather similar, showing only insignificant differences. This makes sense when taking the matching procedure into account. Here, portfolio firms were matched with control firms based on pre-buyout ROA, and as these two measures appear to be nearly identical, it is concluded that the matching was a success in regard to this particular criterion. In contrast, EM appears to be significantly higher for control firms at the 1% level. This finding indicates that PE firms aim to acquire companies with inefficient operations that they can turn around post-buyout, whereas efficient pre-buyout operations seem to be a priority for industrial acquirers. In terms of capital structure, portfolio firms seem to be more levered than the control firms. This difference, however, is only significant at the 10% level, and no significant differences are found for LD or SD. Common to both types of firms is that the pre-buyout leverage level is rather high. This could be explained by the lucrative loan conditions offered prior to the financial crisis, which is when the majority of the deals included in this thesis were performed (cf. figure 5). It might, however, also indicate that industrial corporations have learned about the potential benefits of high leverage, as described in section 3.2, leaving few opportunities for PE funds to pick the “low-hanging fruits” from optimization of the target’s capital structure. Pre-buyout, COGSM, OOEM, and WCR are all higher for portfolio firms relative to control firms. Though PE firms might not target turnaround companies to the same extent as they did earlier, this finding supports the notion that they still screen for relatively inefficient targets. AT is also slightly higher for portfolio firms compared to control firms, indicating the opposite – that portfolio firms are more efficient pre-buyout. However, the differences between all of the measures are found to be insignificant. Finally, the average age of portfolio and control firms is 26 and 25 years respectively. Hence, if survivorship bias related to firm age is present in the data, it is expected to affect the portfolio firm and control groups to a similar extent. 6.3.2. Post-Buyout Descriptive Statistics Table 5 shows two other estimates which are discussed in the following: the post-buyout summary statistics of the portfolio and control firms, and preliminary DID estimates.

35

It should be noted that the statistical tests of the DID estimates in table 5 are only used as a preliminary indicator to describe the sample. Recall that the general econometric model from section 5.1 will be used as the foundation to test all five hypotheses. The results of these tests are presented and discussed in section 7. First, portfolio firms also appear to be larger than control firms post-buyout in terms of both assets and sales; however, none of the DID estimates are significant, indicating that PE ownership does not affect company size per se. The insignificance of the DID estimate related to TA could indicate that portfolio and control firms book goodwill in a similar fashion (i.e. both mainly on either holding or parent company level). This is positive as the alternative would bias the level of total assets and consequently bias the asset-scaled performance measures (e.g. ROA and OROA). Recall section 5.1 where it was described how the literature does not provide evidence of how goodwill is booked post-buyout and whether there tend to be a difference between the two groups. The analysis could potentially be altered to take this “asset boosting” element into account, but it is neglected for reasons discussed in section 7.2.3. Second, both sales growth and asset growth remain positive post-buyout. Furthermore, sales growth still appears to be lower for portfolio firms post-buyout, though not significantly. This is consistent with the growth findings pre-buyout, which indicated that industrial firms target growth firms to a larger extent than PE firms. Contrary to the pre-buyout situation, portfolio firms appear to grow their assets significantly faster than the control firms. This is contradictory to what was concluded in the discussion on the absolute size just above. Note that for both portfolio and control firms, sales and asset growth decrease from pre-buyout to post-buyout. It is likely that this is partly a result of the financial crisis, which affects a large part of the observations post-buyout. The DID estimate for asset growth is significant, though only at a 10% level, whereas sales growth is insignificant and close to zero. This indicates that PE fund ownership induces a relatively higher asset growth compared to industrial ownership. This could either be a result of a relatively large share of goodwill booked at parent level rather than holding level or simply a result of higher organic growth compared to control firms. The latter is assessed most likely, as no evidence is found in the literature pointing towards differences in how the goodwill is booked between the two groups. Third, on a post-buyout basis, portfolio firms’ OROA, ROA, and NPM either remained unchanged or dropped slightly. In contrast, control firms experienced considerable drops for the three measures. Thus, statistical tests of the post-buyout differences show that portfolio firms have significantly higher OROA, ROA, and NPM after being taken over by a PE firm. Regarding EM, the control firms experienced a drop, whereas portfolio firms managed to increase performance. The statistical difference remains significant, however, only at a 5% level. The general drop in absolute profitability for both portfolio and control firms, could to some extent be attributed to the financial crisis, which impacts the majority of this thesis’ sample. On a DID basis, portfolio firms show relative 36

improvements on all four performance measures. The DID estimates related to ROA and EM are both significant at the 5% and 1% levels respectively. This evidence is highly interesting as it points towards a confirmation of this thesis’ main hypothesis: PE ownership induces superior performance in portfolio firms, relative to the control group. These indications are consistent with the majority of the performance literature on the field (e.g., Kaplan, 1989a, 1989b; Lichtenberg & Siegel, 1990; Muscarella & Vetsuypens, 1990; Smith, 1990; Holthausen & Larcker, 1996; Wright et al., 1997; Cressy et al., 2007; Cao & Lerner, 2009; Guo et al., 2011; Alperovych et al., 2013). Moreover, it is interesting that control firms show a significantly higher EM both pre- and post-buyout, while the DID estimate reveals that, in fact, portfolio firms were able to improve this performance measure significantly in a relative scope. This could mean that PE firms use EM as screening key figure, to deliberately target firms performing poorly on this parameter, and then trim the company’s operations post-buyout. The table shows an increase in portfolio firm leverage from pre- to post-buyout and a drop in control firm leverage. Thus, the post difference is significant at a 1% level compared to the 10% level prebuyout. The fact that leverage increases within portfolio firms is consistent with findings in the majority of the literature on the field (e.g., Muscarella & Vetsuypens, 1990; Palepu, 1990; Andrade & Kaplan, 1998). However, a larger increase was expected. An explanation could be that the pre-buyout leverage level for both portfolio and control firms was relatively high, and thus not much room was left for further increases. As described in section 3.2, this initial high leverage level could be caused by the lucrative loan conditions prior to the financial crisis and industrial companies’ increased knowledge about the potential benefits of high leverage. As for the decrease in leverage within control firms, it should be noted that debt is sometimes placed at holding level and not parent level in these transactions, which could explain this finding. The DID estimate indicates that PE ownership stimulates higher leverage compared to control firms; however, this difference is insignificant. In particular, this provides preliminary evidence against hypothesis 2. Portfolio firms manage to decrease COGSM, OOEM, and WCR considerably post-buyout, whereas control firms experience a slight decrease only in COGSM and increases in OOEM and WCR. The post-buyout differences between the two groups shrink for both COGSM and OOEM but increase for WCR. The only significant post-buyout difference is within WCR. AT decreases by roughly the same magnitude for both portfolio firms and control firms. The post-buyout difference remains roughly the same. The DID estimates for COGSM, OOEM, and WCR are significantly negative at a 10%, 5%, and 1% level, respectively, providing preliminary indications of PE firms being superior at managing these operational levers. The results support previous findings in the literature related to cost cutting (e.g., Holthausen & Larcker, 1996; Wright et al., 2001; Guo et al., 2011) and reductions in working capital (e.g., Magowan, 1989; Singh, 1990; Easterwood et al., 1989; Holthausen & Larcker, 1996). For AT, the DID estimate is insignificant, indicating that PE firms are not superior at enhancing asset efficiency, which contradicts previous findings (e.g., Muscarella and Vetsuypens, 1990). These 37

findings could perhaps be explained by the alleged increased focus on operational engineering within PE firms. Recall that PE firms today, for instance, hire ex-consultants and industrial executives to an increasing extent in order to assist the PE firm with improving operations in their portfolio firms. In addition, PE firms tend to specialize within fewer industries today as compared to earlier. Finally, the findings provide evidence against the notion that corporations in general (i.e. potential targets) have increased efficiency and tightened cost control over the last decades to the extent that short-term benefits from operational improvements have been eroded. 6.3.3. Other Sample Characteristics Figure 4 reports how the sample’s portfolio and control firms are distributed by countries of origin. It shows that the majority of the portfolio firms are from either the U.K. or France. Together these two countries account for just below half of the transactions, which is not surprising, as the PE model traditionally has been relatively popular here (Kaplan & Strömberg, 2008). Moreover, the Nordics and Benelux are well represented in the sample, whereas large countries such as Spain, Italy, and in particular Germany account for a relatively insignificant part of the deals. Despite the PE model having been less popular in these countries historically, compared to the U.K. and France, it is still noteworthy that Germany only accounts for five deals. This bias is most likely caused by the relatively small amount of available Orbis data on private German companies 39. In general, biases related to highly varying data availability across countries in Orbis are assessed to be a considerable problem in an analysis like this. This subject is further discusses in section 7.4.4. Figure 4 – The Sample Transactions Split by Nationality and Type of the Target Company The respective terms comprise the following countries: UK = United Kingdom; Nordic = Denmark, Sweden, Norway, Finland, and Iceland; Benelux = The Netherlands, Belgium, and Luxemburg; EM or “Emerging Markets” = Poland, Czech Republic, Romania, Bulgaria, Slovakia, Hungary, and Slovenia; Baltics = Estonia, Latvia, and Lithuania; Other = Austria, Cyprus, Switzerland, Ireland, and Malta.

34

30

28

29 30 21

19

20 13

12 12

Portfolio firms

14 10

10

9 6

5

5 2

4 3

Control firms 1 1

Other

Baltics

Germany

EM

Italy

Spain

Benelux

Nordic

France

0

UK

Number of obs. in the sample

40

Target'slocation Source: Zephyr

39

Orbis website

38

Over the last decade, U.K. targets have constituted about one-third of the total PE investments in Europe, and continental European targets have accounted for the remainder two-thirds (McKenzie & Maslakovic). This pattern is largely reflected in the figure above. Moreover, the distribution of PE transactions corresponds well with the overall geographical distribution of PE fund targets for the period, which is important as the opposite would introduce a geographical bias in the data material (Kaplan & Strömberg, 2008). Recall from this thesis’ introduction, that the PE model is expected to have different value potential across nations. Consequently, it would be sensible to include country dummies in the general econometric model, but this is complicated by small sample sizes for most countries/regions (as seen in the figure above) and thus ignored. Instead, a U.K. dummy is included in the econometric model as a robustness check in section 7.2.1. Figure 4 also shows the “balance” between the origin of portfolio firms and control firms, which appears to be overall acceptable (with the exception of Spain and the Emerging Markets). This is an important finding, as the balance is vital in relation to the parallel trend assumption introduced in the method section. Across countries and regions, various factors – such as legal factors – are expected to develop differently over time, and these could eventually violate the aforementioned assumption. Figure 5 shows how the transactions included in the sample are distributed over time. It reflects the same pattern as seen in figure A2 (from appendix 1), where deal activity increase prior to the financial crisis and subsequently decline. This is positive, as it indicates that no particular time period is highly overrepresented in the sample, relative to the overall PE activity. Figure 5 – The Sample Transactions Split by Transaction Year

Number of obs. in the sample

40

36

33

30

31

26 20

20

16

22

21

20

16

Control firms

11

10

Portfolio firms

8

0

2003

2004

2005

2006

2007

2008

Transaction year Source: Zephyr

Another observation is that the portfolio and control firms appear to be well balanced from year to year (with only a smaller imbalance in 2006 and 2007), which is important as “transaction year” was not used as a matching criterion. By not matching on “transaction year,” the parallel trend assumption is risked violated, as the portfolio and control firms are likely to be affected by fundamentally different 39

time-varying factors. For instance, a portfolio firm acquired by in 2003 and a control firm acquired in 2008 (financial crisis) are likely to be affected fundamentally differently during their respective event windows – in this case, by the macroeconomic environment. As described earlier, in order to control for this phenomenon more thoroughly, year dummies are included in the general econometric model applied to test this thesis’ hypotheses.

7. Results and Discussion In the following section, the stated hypotheses are tested and the results are discussed. First, The Superiority Hypothesis is tested to assess the claimed superior impact of PE ownership on portfolio firm performance. Secondly, a number robustness checks are undertaken. Thirdly, the support hypotheses are tested to assess whether they can explain the findings related to The Superiority Hypothesis. Lastly, alternative explanations (e.g., various biases) are discussed.

7.1. Testing of The Superiority Hypothesis Results In this section, The Superiority Hypothesis is tested. Table 6 shows the estimated impact of PE fund ownership on firm performance measured by ROA, OROA, EM, and NPM. The impact on all four is significantly positive, indicating that the PE model does in fact enhance performance relative to the benchmark firms. Recall that the coefficients related to the PE variables constitute the DID estimate of the given regression. Hence, it measures the change in average performance within the group of portfolio firms (ROA_POST – ROA_PRE), relative to the change in the control group. Thus, on average, the ROA of portfolio firms improves 2.8 pp post-buyout relative to control firms. This estimate is significant at a 5% level. For OROA, it is 2,3 pp and significant at a 10% level; for EM, it is 4,7 pp and significant at a 1% level; and finally for NPM, it is 2,3 pp and significant at a 5% level. The impact is both of a substantial magnitude and robust across different performance measures.

40

Table 6 – The Impact of Private Equity Fund Ownership on Firm Performance The table shows the output from all regressions performed to test hypothesis 1 (The Superiority Hypothesis). For all measures of firm performance (ROA, OROA, EM, and NPM), the impact of PE ownership was estimated using the general econometric specification derived in section 5.1. Next, the insignificant explanatory variables were deleted from the regression one by one, starting with the most insignificant. This procedure was continued until the reduced specifications seen below were reached (see appendix 7). These comprise only explanatory variables that are significant at a minimum 10% level. Note that LOG_SIZE was not significant in any of the regressions, and thus does not appear in the table. The definitions of each of the dependent and independent variables are described in section 5.1.1. For regression (2) and (4) White’s heteroskedastic robust standard errors were applied. ***, **, and * indicate that the given variable is significant at a 1% level, 5% level, and 10% level, respectively. The standard errors are reported below.

ROA

Dependent variables OROA EM

NPM

Independent variables Constant

0,055 *

-0,021 *

0,028

PE fund ownership (PE)

0,028 ** 0,013

Log of firm age (LOG_AGE)

0,011

0,023 * 0,012

0,047 *** 0,013

0,023 ** 0,009

-0,035 * 0,020

Pre buyout performance (ROA_PRE)

-0,579 ***

-0,462 ***

-0,269 ***

-0,395 ***

0,080

0,086

0,081

0,068

Industry dummies (I0-I9)

YES

YES

YES

YES

Year dummies (y2003-y2008)

YES

YES

YES

YES

260 0,189 0,177 NO

260 0,136 0,123 YES

260 0,165 0,144 NO

260 0,166 0,155 YES

Other charateristics Number of observations R-squared Adjusted R-squared White standard errors Source: Orbis; Zephyr; Own analysis

The estimated positive performance effect corresponds to findings in the majority of previous studies (e.g., Baker & Wruck, 1989; Kaplan, 1989a, 1989b; Lichtenberg & Siegel, 1990; Muscarella & Vetsuypens, 1990; Smith, 1990; Holthausen & Larcker, 1996; Wright et al., 1997; Harris et al., 2005; Cressy et al., 2007; Cao & Lerner, 2009; Guo et al., 2011) and contradicts fewer studies which find a negative performance impact of PE ownership (e.g., Ravenscraft & Scherer, 1987; Desbriéres & Schatt, 2002). Note that the former group of studies reports a highly varying magnitude of the effect, which emphasizes the incomparability of the existing literature as outlined in the introduction. Kaplan (1989a) documented an increase in operating income of 42% during the first three years post-buyout for 48 MBOs undertaken in the U.S. between 1980 and 1986. Cressy et al. (2007) studied a sample of 122 U.K. buyouts over the period 1995–2000 and found that operating profitability in portfolio firms increases about 4 to 5 pp post-buyout. However, Guo et al. (2011) reported that operating performance of portfolio firms slightly exceeds the applied benchmarks, based on a U.S. sample between 1990 and 2006.

41

Note that the SIZE and AGE controls largely have no significant impact on firm performance (except for AGE on ROA). However, a high pre-buyout performance appears to have a highly significant negative impact across all four performance measures post-buyout. Several industry and year dummies also appear to be significant (cf. appendix 7). Note that findings related to the applied controls support Barber and Lyon’s (1996) notion that matching on ex-ante performance and industry is much more important than matching on firm size. It also supports the notion presented in this thesis, that matching on year of transaction would have been ideal40. Discussion The results found in this thesis indicate that PE fund ownership (still) induces superior performance in the portfolio firms, which is in line with findings in the majority of previous studies. The magnitude of the estimated positive impact is rather similar to those found in studies based on more recent data (e.g., Cressy et al., 2007; Guo et al., 2011) but significantly less positive compared to studies based on older data (e.g., Kaplan, 1989a). This could reflect an overall decrease in the value creation potential of the PE model. Recall that a number of structural changes in the PE landscape were presented in this thesis’ introduction, all of which indicated lower benefits from the PE model in today’s business environment – i.e., fewer portfolio firms are of the turnaround type, deals are less levered, and fewer public-to-private transactions are undertaken. Consequently, it is likely that these changes can partly explain the decrease of the model’s value creation potential. The variations in magnitude of the estimated impact relative to Cressy et al. (2007) and Guo et al. (2011) should also be addressed. Cressy et al. (2007) uses data from a time period when the overall value creation potential of the PE model could be slightly higher. In addition, he only investigates buyouts from the U.K. Recall that prebuyout ownership concentration tends to be lower in the U.K. and U.S. (Faccio & Lang, 2002), which according to The Jensen Hypothesis means more agency problems ex-ante and thus higher expected benefits from the PE model (NB. findings supporting this notion are presented in this thesis’s section 7.2.1). The U.S. data and the time period from which Guo et al. (2011) drew data bring about expectations of relatively large benefits from the PE model. Surprisingly, they find the opposite, which perhaps could be attributed to their choice of research design (i.e. their choice of method, choice of a different benchmark, etc.). Finally, the usual checks related to OLS were performed for all of the regressions in table 6 above (see Appendix 7 for a detailed description). In the table below, a summary of these tests is presented. Here no multicollinearity is found; however, heteroskedasticity appears to be present in the OROA and NPM models. Thus, White’s heteroskedastic robust standard errors are applied to remedy this. No tests for autocorrelation are performed due to reasons described in section 5.3. Recall that the pre- and post-buyout observations were collapsed, or averaged, in order to avoid problems related to

40

Recall that applying year of transaction as matching criteria would have reduced the sample size significantly and compromised the preperformance criteria.

42

autocorrelation. Finally, this procedure of testing the OLS assumptions is applied for all regressions in this thesis. Table 7 – Test Summary of OLS Assumptions This table gives an overview of the usual tests of the OLS assumptions. The numbers reported are p-values of White’s test.

Subject of test

Applied test

Multicollinearity

Correlogram1

Heteroskedasticity

White's test

Dependent variable of regression ROA OROA EM NPM No

No

No

No

0,66

0,01

0,99

0,00

Note 1) Multicollinearity is assessed to be a problem when pair-wise correlation between explanatory variables is high. In this thesis, high correlation implies one correlation value above 0,8 and/or several correlation values exceeding 0,5 (Bergström et al., 2007). “Yes” indicates that multicollinearity is present; “no” indicates the opposite.

7.2. Robustness Checks In this section several checks are performed to enhance the robustness of the results from section 7.1 above. First, the econometric specification is altered to assess whether the choice of control variables affects the result significantly. Second, alternative performance measures are applied. Finally, different matching procedures are tested in order to account for potential selection biases. It appears that none of the alterations significantly change the overall conclusion from section 7.1; that PE fund ownership improves firm performance relative to the benchmark. 7.2.1. Econometric Specification First, a U.K. dummy is included in the econometric models applied to test The Superiority Hypothesis (see table 8 below). As described in the previous section, benefits from the PE model are expected to be higher in the U.K. compared to continental Europe, as pre-buyout ownership concentrations tends to be lower here. This, combined with the strong representation of U.K. targets in the sample, could cause an upwards bias in the evaluation of the performance impact of PE ownership in general. The U.K. dummy equals “1” if the target company was located in the U.K. at the time of the buyout and “0” otherwise. Table 8 shows that the impact of PE fund ownership remains significantly positive on three out of four performance measures. ROA and EM are still significant at a 5% and 1% level respectively, whereas NPM is now significant only at a 10% level. The impact on OROA appears to be insignificant. The magnitude of the impact on all four performance measures has dropped, notably for OROA (2,3 to 1,2 pp) and NPM (2,3 to 1,7 pp), but less for ROA (2,8 to 2,5 pp) and EM (4,7 to 4,5 pp). The impact of the U.K. dummy is significantly positive on three out of four performance measures, namely ROA, EM, and NPM, but insignificant on OROA. The magnitude of the impact is rather large, varying between 2,8 pp and 4,3 pp. These findings indicate that there is indeed a higher positive ex43

post performance impact from buyouts in the U.K. compared to continental Europe, which supports the predictions from The Jensen Hypothesis. However, the findings from section 7.1 remain robust to the inclusion of a U.K. dummy. Table 8 – The Impact of PE Fund Ownership on Firm Performance Adjusted for U.K.-effect The table shows the output from all regressions performed to test The Superiority Hypothesis adjusted for the “U.K. effect.” For all measures of firm performance (ROA, OROA, EM, and NPM), the impact of PE ownership was estimated using the general econometric specification derived in section 5.1. Next, the insignificant explanatory variables were deleted from the regression one by one, starting with the most insignificant. This procedure was continued until reduced specifications were reached (see attached CD-ROM). These only comprise explanatory variables that are significant at a minimum 10% level. The firm performance variables are defined in section 5.1.1, and “U.K. dummy” equals “1” if the target firm is from the U.K. and “0” otherwise. ***, **, and * indicate that the given variable is significant at a 1% level, 5% level, and 10% level, respectively. The standard errors are reported below.

ROA

Dependent variables OROA EM

NPM

Independent variables PE fund ownership (PE)

0,025 ** 0,011

UK dummy Usual controls

0,043 ***

0,012 0,012

0,028

0,045 *** 0,012

0,033 **

0,017 * 0,010

0,036 ***

0,015

0,017

0,015

0,014

YES

YES

YES

YES

260 0,204 0,195

260 0,137 0,123

260 0,219 0,188

260 0,191 0,176

Other charateristics Number of observations R-squared Adjusted R-squared Source: Zephyr; Orbis; Own analysis

Besides the U.K. dummy above, several other controls were tested, but none of these changed the significance of PE fund ownership on any of the performance measures. In particular, level versions of the SIZE and AGE variables, geographical region dummies (based on the categorization of the European countries in figure 4), a cross-border dummy41, and several other variables were applied. 7.2.2. Performance Measures Table 9 shows the impact of PE fund ownership on alternative performance measures, specifically Return on Capital Employed (ROCE), Asset Growth (AG) and Sales Growth (SG). Of the three, ROCE is by construction most similar to the four operational measures initially applied. It is defined as EBIT / ( TA – Current Liabilities), or EBIT divided by fixed assets plus working capital. Thus, it is identical to OROA, except the denominator is adjusted for current liabilities. Asset Growth and Sales Growth are, as the names indicate, considered growth measures rather than performance measures per se. However, they have been applied as such in previous literature (e.g., Vinten, 2007). The expected causality between PE fund ownership and all three variables is positive.

41

The cross boarder dummy equals “1” if the acquirer and target originated from different countries and “0” otherwise.

44

Table 9 shows a positive impact from PE fund ownership on all three performance measures. The effect on ROCE and AG is significant at a 10% and 5% level respectively, whereas it is insignificant on SG. PE ownership impacts ROCE by 5,2 pp and AG by 6,6 pp. The magnitude of the estimated impact on ROCE is higher compared to the original measures (ROA, OROA, EM, and NPM), whereas the impact on AG is not directly comparable. These results do not change the initial conclusion from section 7.1 that PE fund ownership enhances portfolio firm performance. Note that the coefficients of determination (adjusted R-squared) of the regressions from table 9 are considerably lower than those reported in table 6 - especially for the ROCE regression - which supports my choice of initial specification. Table 9 – The Impact of PE Fund Ownership on Alternative Firm Performance Measures The table shows the output from all regressions performed to test the robustness of the results from section 7.1 related to the choice of performance measure. For all measures of firm performance (ROCE, AG, and SG), the impact of PE ownership was estimated using the general econometric specification derived in section 5.1. Next, the insignificant explanatory variables were deleted from the regression one by one, starting with the most insignificant. This procedure was continued until reduced specifications were reached (see attached CD-ROM). These only comprise explanatory variables that are significant at a minimum 10% level. Return on Capital Employed (ROCE) is defined as EBIT / TA – Current Liabilities. The computation of Asset Growth (AG) and Sales Growth (SG) is straightforward. ***, **, and * indicate that the given variable is significant at a 1% level, 5% level, and 10% level, respectively. The standard errors are reported below.

ROCE

Dependent variables AG

SG

0,066 **

0,011

Independent variables PE fund ownership (PE) Usual controls

0,052 * 0,029

0,027

0,041

YES

YES

YES

244

260 0,148 0,114

260 0,074 0,051

Other charateristics Number of observations R-squared Adjusted R-squared

0,042 0,034

Source: Zephyr; Orbis; Own analysis

It is surprising that portfolio firms seem to grow their assets more than control firms post-buyout. From the descriptive statistics in section 6.3, we know that the sample’s average portfolio firm is significantly larger than the average control firm pre-buyout. Thus, one would expect the control firms to be in a better position to grow further. This could either emphasize the overall superiority of the PE model or indicate differences in how goodwill is booked between the two groups – recall that goodwill valuations often change radically in target firms post-buyout. However, no evidence was found in the literature indicating that PE funds book goodwill at parent company level rather than holding level to a larger extent than control firms 42 . Thus, the former explanation seems more viable. Lastly, other alternative performance measures (both profit measures and means of scaling) were considered, but disregarded for reasons discussed in section 5.1.1. 42

This is not paramount, however. Assuming portfolio firms tend to book goodwill at parent level rather than holding level, and the control firms tend to do the opposite, it would boost the asset base of the portfolio firms and thus decrease the performance measures scaled by total assets (ROA and OROA). Consequently, the impact of PE fund ownership on firm performance would be negatively biased. Such finding would only serve to strengthen the robustness of the conclusion from section 7.1, and thus probably not provide much valuable knowledge.

45

Based on the discussion of alternative performance measures, it is sensible to continue the analysis with the measures initially applied, namely ROA, OROA, EM, and NPM. 7.2.3. Matching Criteria Selection bias might be present in the sample, particularly due to the matching procedure applied. Therefore, alternative matching criteria are tested in this section in order to enhance the robustness of the results from section 7.1. Recall that for the final sample, only 130 out of 379 randomly collected portfolio firms were matched with a control firm. This is problematic and could introduce a significant selection bias, given the characteristics of the two groups (portfolio and control firms) were fundamentally different prematching43. Also recall the matching criteria initially applied: the best possible match based on prebuyout performance (ROA_PRE) within a 50% +/- range of firm size strictly matching on a two-digit industry code. In order to reach the alternative sample analyzed in table 10 below, the same matching procedure was applied, except the criterion “company size” 44 was removed as this parameter turned out to be insignificant in nearly all of the regressions performed in this analysis (cf. appendix 7 and attached CD ROM). Moreover, Barber and Lyon (1996) describe this as the least important of the three criteria. Applying only industry and ex-ante performance as criteria produced a matched sample of 344 portfolio and 344 control firms 45. The two groups remain similar in regards to pre-buyout performance (average of 6,8% and 6,6% respectively). However, the magnitude of both measures is significantly lower compared to the original sample (8,0% and 7,8% respectively), which indicates the presence of some selection bias in the original data. Table 10 shows the estimated impact of PE fund ownership on firm performance for the sample based on the alternative matching procedure. The impacts on ROA, EM, and NPM all remain significantly positive in this test at a 1% level, whereas the impacts on ROA and NPM were significant only at a 5% level using the original sample. The impact on OROA also remains positive but is now insignificant. The magnitude of the impact decreased for all performance measures – slightly for ROA (2,8 pp to 2,6 pp) and NPM (2,3 pp to 2,0 pp), but more drastically for OROA (2,3 pp to 1,3 pp) and EM (4,7 pp to 2,6 pp). 43

To illustrate how this matching procedure could have biased the sample, consider the following rather extreme example: The buyouts in the portfolio firm group, pre-matching, are highly concentrated in one industry – say 50% in Industry1 – where only 5% of the control firms represent this industry. When a matching procedure similar to the one applied in this paper is used, only a small fraction of the 50% will appear in the matched sample, which is problematic as various firm and market characteristics can differ greatly across industries. For instance, the performance improvements in portfolio firms from Industry1 could happen to be below average compared to the portfolio group as a whole. Only including a small fraction of these firms would cause a positive bias in the estimated impact of PE fund ownership on firm performance in general. Note that if certain industries are in fact over- or underrepresented in the final sample of the 2 x 130 matched firms, controlling with industry dummies in the econometric specification does not help reduce the bias. 44 Measured by Total Assets 45 The randomly collected sample of portfolio firms totalled 379, but 35 of these could not be matched with a unique control firm due to the strict industry criterion. The potential bias is, however, assessed to be minimal.

46

The results of this robustness test support the present analysis’ initial conclusion from section 7.1 that PE fund ownership has a relatively strong positive impact on portfolio firm performance. The decrease in the estimated impact on all measures, and particularly the large decrease for EM, could indicate that some selection bias is present in the original sample. As expected, the larger sample size increases the statistical power of the tests, and overall the findings become more significant due to decreased standard errors. Table 10 – The Impact of PE Fund Ownership on Firm Performance (Alt. Matching) The table shows the output from all regressions performed to test the robustness of the results from section 7.1 related to the choice of matching criteria. For all measures of firm performance (ROA, OROA, EM, and NPM), the impact of PE ownership was estimated using the general econometric specification derived in section 5.1. Next, the insignificant explanatory variables were deleted from the regression one by one, starting with the most insignificant. This procedure was continued until reduced specifications were reached (see attached CD-ROM), which only comprise explanatory variables that are significant at a minimum 10% level. The firm performance variables are defined in section 5.1.1. ***, **, and * indicate that the given variable is significant at a 1% level, 5% level, and 10% level, respectively. The standard errors are reported below.

ROA

Dependent variables OROA EM

NPM

Independent variables PE fund ownership (PE) Usual controls

0,026 ***

0,013

0,026 ***

0,020 ***

0,007

0,009

0,008

0,007

YES

YES

YES

YES

660 0,263 0,257

649 0,157 0,148

661 0,131 0,121

629 0,150 0,141

Other charateristics Number of observations R-squared Adjusted R-squared Source: Zephyr; Orbis; Own analysis

Lowering the strictness of the matching criteria has three main consequences; the potential selection bias is reduced, the sample size is increased, but the comparability of the two groups is decreased. Since it has already been argued that sample size of the original sample is not a considerable problem, the choice is a trade-off between selection bias and comparability. Despite the likelihood of selection bias having some impact on the results from section 7.1, comparability is prioritized, and consequently the analysis is continued using the original sample. Note that several other alterations to the matching procedure were tested – for instance, year of transaction, country of origin, and stricter requirements to match on firm size. However, these produced neither statistically different estimates nor remarkably higher coefficients of explanation. The results reported in the table above are the only ones described in detail since they represent the largest deviation from the original findings. Note that certain seemingly relevant robustness checks

47

have been neglected altogether: specifically goodwill adjusted performance and reduction of the Jcurve effect46. To sum up section 7.2, the general econometric specification is robust to alterations and the results from section 7.1 are robust to alternative matching criteria. Hence, none of the robustness checks provided substantial reason to change the conclusion that PE fund ownership has a significantly positive impact on firm performance. In addition, it was found that a positive selection bias could impact the magnitude of the estimates to some extent.

7.3. Testing of the Support Hypotheses In this section, The Financial Engineering Hypothesis and The Operational Engineering Hypotheses (the support hypotheses) are tested in order to assess whether PE firms’ claimed superior abilities within these fields can explain this thesis’s main findings from section 7.1. 7.3.1. The Financial Engineering Hypothesis The formulation of the hypothesis is found in section 4, and the testing strategies are outlined in section 5.2, but a brief brush-up follows here. Recall that The Free Cash Flow Hypothesis (Jensen, 1986; Jensen, 1989; Palepu, 1990) states that the main wealth benefits from LBOs arise from organizational changes which bring about improvements in operating and investment decisions. In particular, the benefits arise from the disciplining effect of high leverage on management and tax savings from interest payments. Moreover, the Free Cash Flow Hypothesis predicts that the closer debt is to the decision makers, the higher the disciplining effect it has. PE funds have traditionally placed debt at parent company level (close to the decision makers), whereas industrial acquirers more commonly place parts of it at holding level (Jensen, 2007). Hence, the expected benefits from leverage are predicted to be higher in buyouts performed by PE funds. Hypothesis 2: PE firms are superior at enhancing firm performance through debt monitoring (leverage) Note that OROA and EM capture the operational performance effects of increasing leverage, as both measures have EBIT in the numerator, whereas ROA and NPM (Net Earnings in the numerator) also includes financial and tax effects, since Net Earnings comprise the numerator here. Thus, the two types of performance measures embrace different aspects of the expected benefits from increased leverage.

46

Assessing goodwill-adjusted performance could potentially have remedied problems related to asset boosting, but unfortunately the necessary data was not available. Moreover, performing such a test is found redundant since portfolio firms are expected to be exposed to asset boosting to a larger extent than control firms (cf. the discussion of the descriptive statistics in section 6.3). Consequently, the asset base by which Net Earnings and EBIT are scaled is biased upwards, and ROA and OROA biased downwards. Hence, there is no point in undertaking such an analysis as a robustness check. The J-curve effect could have been reduced by expanding the event window to +/-5 years. However, this would have decreased the sample size significantly and narrowed the sample time period, so the entire second PE wave could not have been captured. Finally, no evidence is found that the J-curve effect should be more articulated amongst portfolio firms compared to the control firms.

48

Results As described in section 5.2, hypothesis 2 is examined through three tests, which are performed in the following sections. The related results are reported in table 11 below. First, the impact of PE fund ownership combined with the post-buyout debt ratio level is tested. In this test the key explanatory variable is the interaction term47 between the PE ownership dummy (PEi) and the debt-to-asset ratio post-buyout (LEV_POST) 48 . If hypothesis 2 holds, the interaction term coefficient should be significantly positive. In table 11 we see that the results of the first test show some support of hypothesis 2. The level of leverage post-buyout (LEV_POST) has a highly significant negative impact on firm performance sample-wide, but the impact of leverage on portfolio firm performance in particular (PE*LEV_POST), is significantly positive on three of four measures (ROA, EM, and NPM). However, taking into account the magnitude of the coefficients, this finding indicates that the impact on portfolio firm performance is only less negative compared to the control group. In other words, on a relative basis PE firms do seem slightly superior in terms of enhancing performance from increasing leverage, as they destroy less value compared to the industrial acquirers. However, this method does not take the prebuyout capital structure or debt decomposition into account, and therefore tests two and three are performed in the following sections.

47

Before addressing the results, it is important to clarify how interaction terms in regressions should be interpreted. If the effect of a variable (e.g., LEV_POST) is expected to have the same impact on firm performance for both portfolio firms and control firms, there is no need to include the interaction term. Hence, the impact of LEV_POST on firm performance is simply the variable’s beta coefficient. Ho wever, if different impacts are expected across the two groups, an interaction term can be included to account for this. Regard the following regression, used for this section’s first test. yi = α + β1(PE*LEV_POST)i + β2LEV_POSTi + βjControls + ei By including the interaction term, the impact of leverage post-buyout (LEV_POST) on firm performance also depends on whether the target company is owned by a PE fund (PE = 1) or an industrial acquirer (PE = 0). The ceteris paribus impact of post -buyout leverage becomes β1 * PE + β2. Hence, for control firms the impact is β 2 and for portfolio firms β 1 + β2. From the literature review it was found that PE firms are expected to be particularly good at exploiting financial and operational engineering. Hence, a different impact is expected across PE fund and industrial ownership, and consequently, interaction terms are widely applied in this paper’s econometric specifications. 48 LEV_POST is calculated as the three-year post-buyout average of leverage, which is defined as Total Debt / Total Assets.

49

Table 11 – The Impact of Financial Engineering on Firm Performance The table shows the output from all regressions performed to test The Financial Engineering Hypothesis. For all measures of firm performance (ROA, OROA, EM, and NPM), the impact of PE ownership was estimated using the general econometric specification derived in section 5.1. Next, the insignificant explanatory variables were deleted from the regression one by one, starting with the most insignificant. This procedure was continued until reduced specifications were reached (see attached CD-ROM). These only comprise explanatory variables that are significant at a minimum 10% level. All variables are defined in section 5.1.1. ***, **, and * indicate that the given variable is significant at a 1% level, 5% level, and 10% level, respectively. The standard errors are reported below.

ROA

Dependent variables OROA EM

NPM

Independent variables (Test 1) PE*LEV_POST

0,045 ** 0,020

LEV_POST Usual controls

-0,081 ***

0,002

0,040 *

0,023

0,021

-0,095 **

-0,120 ***

0,031 * 0,017

-0,121 ***

0,022

0,039

0,037

0,031

YES

YES

YES

YES

196 0,242 0,206

196 0,144 0,122

162 0,181 0,149

178 0,206 0,178

0,019

-0,104

-0,027

-0,023

0,101

0,076

0,085

0,060

Other charateristics Number of observations R-squared Adjusted R-squared Independent variables (Test 2) PE*LEV LEV Usual controls

-0,145 ***

-0,092 *

-0,089

-0,145 ***

0,041

0,055

0,061

0,041

YES

YES

YES

YES

196 0,210 0,193

196 0,155 0,142

162 0,125 0,103

178 0,270 0,244

0,029

0,009

0,028

0,034

0,030

0,033

0,033

0,026

Other charateristics Number of observations R-squared Adjusted R-squared Independent variables (Test 3) PE*SD_MON SD_MON Usual controls

-0,048 **

-0,008

-0,021

0,024

0,026

0,026

-0,058 *** 0,021

YES

YES

YES

YES

196 0,209 0,176

196 0,065 0,055

162 0,093 0,070

178 0,258 0,194

Other charateristics Number of observations R-squared Adjusted R-squared Source: Zephyr; Orbis; Own analysis

The second test is performed in order to account for the pre-buyout capital structure. It is designed to test whether an increase in leverage post-buyout enhances portfolio firm performance. The interaction term between being owned by a private equity fund (PEi) and change in leverage from pre-buyout to post-buyout (LEV) 49 is the key explanatory variable. It measures the effect of PE fund ownership

49

LEV = LEV_POST – LEV_PRE where “LEV_POST” is calculated as the three-year post-buyout average of leverage and “LEV_PRE” is the three-year pre-buyout average. Leverage is defined as Total Debt / Total Assets.

50

together with being exposed to debt monitoring and in a more direct way tests The Free Cash Flow Hypothesis. The second test provides no support for hypothesis 2. Table 11 shows a significantly negative impact on three out of four measures (ROA, OROA, and NPM) for both control and portfolio firms, caused by an increase in leverage post-buyout (LEV). Contrary to the first test, PE fund ownership does not affect the impact as all interaction terms are statistically insignificant. The results indicate that a 1 pp increase in leverage post-buyout decreased performance between 0.092 pp and 0.145 pp sample-wide. The third test is applied in order to take debt composition into account. It tests whether the disciplining effect of short-term financing is higher than for long-term debt, as it is typically increases the debt service payments per time period relatively more (Cotter & Peck, 2001). The key variable is PE*SD_MON, which captures the effect of not only increasing leverage post-buyout but in particular through short-term financing. SD_MON is a proxy for short-term debt monitoring. It equals “1” if the overall leverage and the share of short-term financing have increased post-buyout, and “0” otherwise. Table 11 shows that debt monitoring through short-term financing has a negative impact on two out of four measures (ROA and NPM) for both type of firms. The results indicate that firms increasing shortterm debt monitoring (represented by SD_MON) on average have a 4,8 pp lower ROA than the alternative group, and a 5,8 pp lower NPM, which is quite a surprising magnitude. The remaining coefficients are insignificant. These findings are similar to the ones in test two as they provide no support for hypothesis 2. Overall, the results of the three tests show no, or weak, support the Free Cash Flow Hypothesis and thus contradict findings from the majority of previous studies (e.g., Kaplan, 1989a; Cotter and Peck, 2001; Nikoskelainen & Wright, 2007; Cressy et al., 2007). Consequently, they do not explain the findings related to The Superiority Hypothesis, but do indicate that financial engineering has lost significance from the first to the second PE wave. Discussion The findings could be explained by a number of factors. First, the disciplining effect is less likely to occur in private-to-private transactions as private companies tend to have strong and influential owners already. Thus, debt as a monitoring tool might have less impact on firm performance in this thesis’s data material as the included deals took place during the second PE wave (where 90% were private-to-private). Secondly, it has been described how corporations in general have realized the potential of increasing leverage. Hence, most potential PE targets were relatively highly levered pre-buyout during the second PE wave compared to the first, which also indicates a lower potential for applying debt post-buyout as 51

a monitoring tool. In fact, this notion is backed by findings reported in this thesis’s descriptive statistics. Here it was found that the average leverage in portfolio firms was 62,7% pre-buyout and 58,3% in control firms. These debt levels are considerably higher than previous findings in the literature (e.g., Cotter & Peck, 2001; Nikoskelainen & Wright, 2007). Consequently, portfolio firms on average only increased leverage by 1.2 pp post-buyout, which is considerably less compared to findings based on data from the first PE wave (e.g., Kaplan, 1989a; Palepu, 1990). Thirdly, in the descriptive statistics it was also found that control firms lowered their average leverage by 2,2 pp post-buyout. This could indicate that a certain amount of debt is in fact placed at holding level rather than parent level in industrial buyouts. If this portion varies greatly between portfolio firms and control firms, the estimated debt levels for control firms could be negatively biased. Finally, some subsamples are rather small, which is likely to decrease the statistical power of the tests and distort the conclusions. For instance, no more than 30 portfolio firms were exposed to short-term debt monitoring. 7.3.2. The Operational Engineering Hypotheses In this section, The Operational Engineering Hypotheses are tested to assess whether they can explain a part of the findings related to The Superiority Hypothesis. Three hypotheses are investigated to test the overall superiority of PE firms’ ability to perform operational engineering within their target companies. – more specifically, hypothesis 3 on cost cutting, hypothesis 4 on reduction of capital requirements, and hypothesis 5 on efficiency improvements. Testing of Hypothesis 3 – Cost Cutting This hypothesis is investigated in order to assess whether PE firms are more capable of enhancing portfolio firm performance through cost cutting, compared to control firms. The hypothesis is examined by applying two tests. Hypothesis 3: Portfolio firms enjoy higher performance compared to the control firms, as they are subject to a tighter cost control The first test’s key variable is PE*COGSM, which measures the effect of PE fund ownership (PE) combined with the Cost of Goods Sold Margin effect (COGSM50). The key variable of the second test is PE*OOEM, which measures the effect of PE fund ownership (PE) combined with the Other Operating Expenses Margin effect (OOEM 51 ). In order to support hypothesis 3, β1 has to be significantly negative, indicating that a decrease in the particular cost margin enhances firm

COGSM is defined as COGS / turnover and computed as follows: COGSM = COGSM_POST – COGSM_PRE where “COGSM_POST” is calculated as the three-year post-buyout average and “COGSM_PRE” is the three-year pre-buyout average. 51 OOEM is defined as OOE / turnover and computed as follows: OOEM = OOEM_POST – OOEM_PRE where “OOEM_POST” is calculated as the three-year post-buyout average and “OOEM_PRE” is the three-year pre-buyout average. 50

52

performance. This association is rather obvious, but the interesting aspect is whether PE firms manage to enhance performance through these margins relatively more than their industrial peers. Two different cost margins are examined as they reflect fundamentally different costs. COGS are costs directly related to the production of goods, whereas OOE comprises costs indirectly related to the production, such as Sales, General & Administrative (SG&A) costs. Note that EM and NPM are scaled by turnover, which is identical to the two cost margins. This relationship could be problematic, and consequently, ROA and OROA are given a higher priority when the results are interpreted. Results Table 12 presents the results related to the two tests of hypothesis 3. For both COGSM and OOEM all the interaction terms have a significant negative impact on firm performance – i.e., an increase in cost margin causes a decrease in performance, and a decrease in cost margin causes an increase in performance. As described earlier, this relationship is not unexpected. However, it is surprising that no significant impact is found for control firms (i.e., COGSM and OOEM as standalone variables are insignificant in all four regressions), whereas the beta coefficient of all interaction terms is significant. The impact of the COGSM combined with PE ownership is significant at a 10% level and 5% level on ROA and OROA respectively. For OOEM it is significant at a 5% level for both performance measures. For COGSM the estimated impact of a 1 pp decrease is a 0,38 pp and 0,50 pp increase in performance for ROA and OROA respectively. For OOEM, it is 0,49 pp and 0,43 pp respectively. Based on the magnitudes of the findings, improvements in COGSM and OOEM respectively explain between 54%-115% and 30%-64% of the superior performance amongst portfolio firms (see appendix 8) from a ceteris paribus perspective. With the usual reservations to the interpretation of ceteris paribus estimates, it is concluded, not what the exact magnitude of the impact on portfolio firm performance is, but rather that the impact is overall considerable and plays a large role. These results support the findings in the majority of the previous literature (e.g., Holthausen & Larcker, 1996; Wright et al., 2001; Guo et al., 2011). The outcome of the above tests indicates that PE firms in particular focus on cost cutting in order to enhance portfolio firm performance, and that they succeed in implementing these cost reductions without compromising the revenue side. These results could potentially explain the findings related to hypothesis 1, and are discussed further at the end of this section. Finally, as a side note it is interesting to discover that OOEM plays such a significant role. Recall that Other Operating Expenses (OOE) comprises indirect production costs, such as R&D and advertising spending, that PE funds are often accused of lowering. Doing so tends to improve performance in the short run (captured by the three-year event window applied in this analysis) but become harmful in the longer run. Thus, it would be interesting to decompose OOEM even further and investigate whether the enhancement of firm performance stems from sustainable sources such as lowered bureaucracy (lowered administrative costs) or more unsustainable sources such as advertising and R&D. Findings 53

in previous literature regarding the latter are rather contradictory, as some studies (e.g., Smith, 1990b; Long & Ravenscraft, 1993; Hoskisson & Hitt, 1994) report a reduction in R&D expenditures postbuyout whereas other studies (e.g., Kohlberg, 1989; Bull, 1989; Lichtenberg & Siegel, 1990) fail to support the “Window Dressing Hypothesis” (as it is called). A common argument against this hypothesis, which I regard highly relevant, is that potential buyers nearly always hire financial advisors to undertake a thorough financial due diligence of the potential target, pre-buyout. Thus, if R&D- and capital expenditures have been reduced significantly to boost FCF in the short run, it most likely will be detected and affect the valuation accordingly. Due to the intuitive appeal of this argument, and to limit the scope of the present analysis, window dressing is not investigated further in this thesis.

54

Table 12 – The Impact of Operational Engineering on Firm Performance The table shows the output from all regressions performed to test The Operational Engineering Hypotheses. For all measures of firm performance (ROA, OROA, EM, and NPM), the impact of PE ownership was estimated using the general econometric specification derived in section 5.1. Next, the insignificant explanatory variables were deleted from the regression one by one, starting with the most insignificant. This procedure was continued until reduced specifications were reached (see attached CD-ROM). These only comprise explanatory variables that are significant at a minimum 10% level. The variables are defined in the data chapter (section 5.1.1). ***, **, and * indicate that the given variable is significant at a 1% level, 5% level, and 10% level, respectively. The standard errors are reported below. “H3” denotes that the given test is related to “hypothesis 3,” and so forth.

ROA

Dependent variables OROA EM

NPM

Independent variables (H3) PE*COGSM COGSM Usual controls

-0,380 *

-0,501 **

0,206

0,242

-0,567 *** 0,202

-0,586 *** 0,180

-0,123

-0,140

0,031

0,131

0,138

0,163

0,155

0,126

YES

YES

YES

YES

202 0,321 0,292

202 0,299 0,269

184 0,280 0,255

192 0,352 0,310

Other charateristics Number of observations R-squared Adjusted R-squared Independent variables (H3) PE*OOEM OOEM Usual controls

-0,494 **

-0,431 **

-0,422 **

0,151

0,182

0,171

-0,387 *** 0,138

0,018

0,018

-0,171

-0,062

0,108

0,130

0,124

0,095

YES

YES

YES

YES

198 0,358 0,334

198 0,282 0,255

166 0,346 0,316

172 0,330 0,291

Other charateristics Number of observations R-squared Adjusted R-squared Independent variables (H4) PE*WCR WCR Usual controls

-0,221 **

-0,212 **

0,031

0,102

-0,023 0,117

0,100

0,089

0,042

0,026

-0,029

0,061

0,072

0,064

0,053

YES

YES

YES

YES

244 0,214 0,191

244 0,089 0,074

208 0,131 0,109

220 0,255 0,219

-0,018

0,039

0,030

0,034

-0,009

0,007

0,020

0,024

0,019

0,018

YES

YES

YES

YES

238 0,208 0,195

238 0,102 0,086

200 0,168 0,133

214 0,228 0,201

-0,099 *

Other charateristics Number of observations R-squared Adjusted R-squared Independent variables (H5) PE*AT AT Usual controls

-0,074 *** 0,028

0,058 ***

-0,037 0,027

-0,002

Other charateristics Number of observations R-squared Adjusted R-squared Source: Zephyr database; Orbis database; Own analysis

55

Testing of Hypothesis 4 – Capital Requirements The hypothesis below is investigated to test whether PE firms in particular manage to enhance performance of their portfolio firms by efficient management of working capital. Working Capital Ratio (WCR) 52 is defined as “inventory + creditors – debtors” scaled by total assets. The key explanatory variable of interest is the interaction term PE*WCR, which captures the effect of PE fund ownership (PE) combined with the working capital effect (WCR). Hypothesis 4: Post-buyout performance is relatively higher in portfolio firms compared to their industrial peers, due to PE firms’ superior abilities within working capital management Results The results reported in table 12 provide support for hypothesis 4, as three out of four performance measures (ROA, EM, and NPM) are impacted significantly positively by a decrease in Working Capital Ratio. For ROA and EM no significant effects are found amongst control firms (WCR standalone is insignificant). However, the interaction term is significant at a 5% level for both measures, which indicates that PE firms in particular enhance performance of their portfolio firms by lowering working capital. The impact on NPM is only found significant for the variable WCR at a 10% level, indicating that both portfolio and control firms manage to create value from working capital management. For each pp WCR is decreased for portfolio firms, ROA and EM is increased by 0,21pp. NPM is estimated to increase 0,099 pp per 1 pp decrease in WCR for both portfolio and control firms. From the calculations in appendix 8 it is found that WCR*PE from a ceteris paribus perspective explains 29% and 16% of the improvements in ROA and EM amongst portfolio firms, respectively. Note, that the same reservations, as mentioned in the COGSM and OOEM discussion, apply here. These results support the overall findings from the majority of previous studies (e.g., Magowan, 1989; Singh, 1990; Easterwood et al., 1989; Holthausen & Larcker, 1996). However, the magnitude of the impact is somewhat incomparable as the aforementioned studies suffer from several of the factors described in section 1.1 (incomparability of previous literature). Holthausen & Larcker (1996) is, for instance, based on a sample of RLBOs, which induces certain biases in terms of portfolio firm characteristics. Moreover, the four studies are based on data from the first PE wave. These results are also supported by the findings of this thesis’s preliminary analysis. In the descriptive statistics from section 6.3, it was found that Working Capital Ratio is decreased on average by 3,7 pp post-buyout in portfolio firms and increased 2,7 pp in control firms. This corresponds to a (significant) DID estimate of 6,4 pp, which indicates that PE firms do focus on improvements in this area. The implications of the results from testing hypothesis 4 are discussed at the end of this section.

WCR = WCR_POST – WCR_PRE where “WCR_POST” is calculated as the three-year post-buyout average of working capital ratio and “WCR_PRE” is the three-year pre-buyout average. Working capital ratio is defined as “inventory + creditors – debtors” scaled by total assets. 52

56

Testing of Hypothesis 5 – Asset Efficiency This hypothesis is investigated in order to assess whether PE firms are more capable of enhancing portfolio firm performance through improvements in asset efficiency, compared to control firms. Asset Turnover is defined as Turnover / Total Assets and is often used as a proxy for asset efficiency in the literature (Muscarella & Vetsuypens, 1990). The key variable in the econometric specification is PE*AT, which measures the effect of PE fund ownership (PE) combined with the asset turnover effect (AT)53. Hypothesis 5: PE firms enhance the performance of their portfolio firms through superior efficiency improvements Results Table 12 presents the results of the regressions performed to test hypothesis 5. It is found that EM is the only performance measure impacted significantly by AT. The impact on the control group (the beta coefficient of AT) is positive at a 1% significance level, whereas the interaction term is significantly negative at a 1% level. If the magnitude of the coefficients is regarded, we see that the results indicate a slight negative impact of asset turnover on portfolio firm performance, as it is given by: -0,074 + 0,058 = -0,16. Intuitively, it makes sense that the impact of AT on EM is negative. Regard the construction of the DuPont Pyramid54 where EM and AT are the two main drivers of OROA, and NPM and AT the main drivers of ROA. Value is often created through one driver at the other’s expense. This notion has the consequence that EM and NPM are not entirely applicable as dependent variables in this particular regression. Instead, ROA and ORAO are given a higher interpretive priority. Thus, as the impact on both of these measures is found to be insignificant, it is concluded that this analysis provides no support of hypothesis 5, which is that PE firms in particular enhance profitability through improvements in asset efficiency (Asset Turnover). These results contradict the findings of the few studies which investigate this particular association. For instance, Muscarella and Vetsuypens (1990) studied 72 RLBOs from the U.S. during the 1980s and found that revenues and AT in portfolio firms were improved post-buyout, compared to a random sample of publicly traded firms. As argued earlier, the motivation for testing this particular hypothesis was based not on previous studies but rather on an intuitive idea that asset turnover could have become an important factor in PE firms’ value creation during the second PE wave. This does not seem to be the case, however. Instead, it is concluded that PE firms on average prioritize margin improvements and cost cutting over asset efficiency improvements.

AT = AT_POST – AT_PRE where “AT_POST” is calculated as the three-year post-buyout average of Asset Turnover and “AT_PRE” is the three-year pre-buyout average. Asset Turnover is defined as Turnover / Total Assets. 54 The DuPont Pyramid is a framework which breaks down profitability measures, such as ROA and ORAA, into a margin component and an efficiency measure. Its further breakdown is not relevant in this context. 53

57

Further discussion of results for H3-H5 In summary, the results of the cost cutting and the capital requirements analysis roughly support the previous findings in the literature, whereas the results from the AT analysis do not. The alleged increased prioritization of operational engineering within PE firms is emphasized as the main explanation to the lever’s high significance during the second PE wave. Historically, a positive impact of operational engineering has not been widely acknowledged in the literature, but during the last decade, it has become an increasingly common explanation as to how PE firms create value (e.g., Heel & Kehoe, 2005; Dobbs, 2006; Kaplan & Strömberg, 2008; Kehoe & Palter, 2009; Matthews et al., 2009; Archleither et al., 2010; Alperovych et al., 2013). This is often backed by the argument that they to an increasing extent recruit former industry executives and ex-management consultants to strengthen their operational capabilities. Moreover, PE firms today are generally more specialized within fewer industries compared to earlier (Kaplan & Strömberg, 2008). In addition, the findings provide evidence against the notion that corporations (potential targets) have generally increased efficiency and tightened cost control over the last decades to the extent that shortterm benefits from operational improvements have been eroded. It seems that inefficiencies still exist on which the PE firms can capitalize. Finally, an important caveat is that a considerable part of the impact on firm performance attributed to operational or financial engineering is likely to be caused by governance engineering. The impact of the three levers are highly interconnected and therefore complicated to isolate. Governance engineering has no directly traceable effect 55 on the bottom line, however, the mechanisms leading to the reduction of agency problems can support the other two (i.e., operational and financial engineering).

7.4. Alternative Explanations Despite the findings from section 7.1 passing the robustness checks and being partially explained by the results of the hypotheses tested above, there are additional uncontrolled factors to consider. Therefore, a number of alternative explanations are discussed in the following section. 7.4.1. Omitted Variable Bias Omitted variable bias (an endogeneity problem) occurs when one or several important variables are left out of the econometric model. As a consequence, variables closely correlated with the error term become biased.

55

Recall that benefits related to governance engineering stem from enhanced incentive alignment, higher ownership concentration, and better control of the board and management (Jensen, 1989; Jensen & Murphy, 1990).

58

Proxy or instrument variable estimation can potentially remedy omitted variable bias, but these options are not used in the present analysis since feasible proxies typically are hard to obtain, and application of the two methods has not been found in relevant previous literature. Moreover, it is unlikely that omitted variable bias is a major issue, as one of the DID method’s relative strengths is its ability to circumvent endogeneity problems. Nevertheless, it is still likely that some bias is present in the data. A relevant unobserved variable in the present analysis could be, for instance, managerial ability. PE firms are generally considered prestigious, and it is likely that on average they are capable of attracting “brighter” employees compared to industrial companies. Say, for instance, that the impact of managements’ ability is significant but unaccounted for in the econometric model. In that case, the PE dummy becomes positively biased. One could probably use “IQ of top executives” as a reasonable proxy for managerial ability; however, such data is not available (the typical proxy problem). 7.4.2. Sampling bias Sampling bias is a persistent error that arises due to flaws in the sample selection process. Sampling bias can occur if a sample is not collected randomly. If this is the case, some entities are more (or less) likely to be chosen than others, which cause these to be overrepresented (or underrepresented) in the sample. This subject was touched upon in section 7.2, where a robustness check was performed to account for possible sampling bias caused by the matching procedure. However, some sampling bias could be present in the data even before matching, caused by the lack of a “balanced” disclosure in the Orbis database. The database naturally does not cover all company data published. In fact, it is likely that data collection for large private companies have been prioritized, and that a size bias is present in the data. Thus, the 379 portfolio firms gathered randomly (the foundation for the matching process in this thesis) could suffer from this bias. The impact of this phenomenon is reduced in the present analysis by controlling for company size in the econometric models, however, it is likely that some bias still remain. Another issue associated with the balance of the Orbis data relates to “target’s country of origin.” From testing the U.K. dummy in section 7.2.1, it was found that the impact of PE fund ownership on portfolio firm performance can differ significantly across countries. Hence, if certain countries are overrepresented in the sample pre-matching, biases are introduced – positive if the effect of PE fund ownership is stronger in these, and negative if the opposite is the case. Ideally this should have been accounted for in the econometric specification of the models applied; yet, this was not possible due to small sub samples for several countries. The consequences hereof are however assessed to be limited56.

No differences in the PE model’s value potential between countries within continental Europe have been identified in the literature. It only tends to distinguish between U.K. and continental Europe as a whole. In addition, it was found in section Fejl! Henvisningskilde ikke fundet. that the impact of PE fund ownership on portfolio firm performance is still significant for buyouts made in continental Europe, when 56

59

7.4.3. Survivorship bias Survivorship bias occurs when an analysis is focused on observations that “survived” some process while systematically excluding those that did not. This typically has the consequence that “successes” are weighted more heavily than “failures” and consequently, a positive bias is introduced. Recall that only “valid observations” 57 are included in the present analysis. This could potentially introduce survivorship bias in the data material, as bad investments are likely to be exited earlier than three years post-buyout, and consequently not appear in the sample. It is expected that PE funds exit their investments early more often than industrial buyers, as they are subject to a limited time frame in which desired changes (e.g. performance improvements) have to materialize. Industrial acquirers are assessed to be less likely to exit early as they are not motivated by prospects of short term performance improvements to the same extent, but tend to apply a more long-term investment perspective. If the early exit is indeed more common amongst portfolio firms, then a positive bias exists, resulting in an upwards pressure on the estimated impact of PE fund ownership on firm performance. 7.4.4. Differences in Acquisition Motives Differences in acquisition motives between PE funds and industrial corporations could perhaps explain some of the positive estimated impact of PE fund ownership on portfolio firm performance. Even though the second PE wave was different from the first in many aspects, the main motivation of the PE fund buyouts was still the absence of monitoring within the target companies (e.g., Prowse, 1998; Brealey & Myers, 2003; Renneboog & Simons, 2005; Jensen et al., 2006). PE funds typically acquire firms with certain characteristics assessed capable of reaching a certain performance within a relatively short time horizon (max. 10 years) (e.g., Cuny & Talmor, 2006). In contrast, industrial acquirers tend to have a broader range of motives. Examples of motives are: Access to specific resources (e.g., patents, technology, key personnel, etc.), internalization of previously sourced services or goods (e.g. single business divisions or relatively small targets), entry into a new market (on which it can take a while before profitability materializes), large long-run growth potential of target (the descriptives revealed that control firms have more articulated pre-buyout growth characteristics compared to portfolio firms), and empire building (in which profitability is not key). Common to these is that improvements in short-term performance in the target company are not the main objective. The matching procedure applied in this thesis circumvents some of the problems related to differences in acquisition motives as the portfolio and control firms have relatively similar characteristics pre-buyout (cf. the descriptive statistics). However, it does not eliminate the issue completely. Thus, it is likely that this phenomenon explains some of the positive impact of PE fund ownership on portfolio firm performance estimated in section 7.1. Note that only a limited amount of literature deals with this matter in detail. Therefore, focusing future research on it would be beneficial. “the U.K. effect” has been accounted for. Thus, it is likely that adding individual country dummies would not alter the significance of this analysis’s main results. 57 The term “valid observation” is used to describe a situation where a target firm has disclosed data regarding a given variable three years both prior to and after the buyout.

60

8. Conclusion This thesis provides fresh evidence on the impact of PE fund ownership on portfolio firm performance during the second PE wave. Moreover, it assesses the funds’ skills within financial and operational engineering. By analyzing a sample of 130 European PE buyouts and 130 industrial buyouts between 2003 and 2008, I find empirical evidence of increased performance in portfolio firms relative to control firms, specifically on these measures: Return on Asset by 2,8 pp, Operating Return on Assets by 2,3 pp, EBIT-Margin by 4,7 pp and Net Profit Margin by 2,3 pp. These findings are robust to econometric specification, choice of performance measures and matching procedure. However, a positive selection bias could have affected the magnitude of the estimates. This main finding supports the claim that the PE model induces superior firm performance. Consequently, the thesis supports the majority of findings in the previous literature (e.g., Baker & Wruck, 1989; Kaplan, 1989a, 1989b; Lichtenberg & Siegel, 1990; Muscarella & Vetsuypens, 1990; Smith, 1990; Holthausen & Larcker, 1996; Wright et al., 1997; Harris et al., 2005; Cressy et al., 2007; Cao & Lerner, 2009; Guo et al., 2011) and contradicts fewer studies which find a negative performance impact of PE ownership (e.g., Ravenscraft & Scherer, 1987; Desbriéres & Schatt, 2002). Note that the magnitude of the estimated impact is lower compared to the majority of the aforementioned studies, which supports the notion that the PE model has lost value potential from the first PE wave in the 1980s to the second in the 2000s. These findings could be explained by the structural changes in the PE landscape which were presented in this thesis’ introduction – namely, fewer portfolio firms are of the turnaround type, deals are less levered, and fewer public-to-private transactions are undertaken. In order to explain how PE fund ownership induces superior performance, four support hypotheses were examined: one constructed to measure the significance of financial engineering, and three assessing the impact of operational engineering. Traditionally, financial engineering has been given a significant role in explaining how PE firms create value in their portfolio firms; however, recent indications point towards a declining significance in today’s business environment. Consequently, hypothesis 2 (constructed to test The Free Cash Flow Hypothesis) was tested to provide up-to-date evidence. Neither testing the impact of “raw” leverage on firm performance in isolation, nor taking pre-buyout capital structure and debt decomposition into account, produces supporting evidence for the Free Cash Flow Hypothesis. These results contradict findings from the majority of previous studies (e.g., Kaplan, 1989a; Cotter & Peck, 2001; Nikoskelainen & Wright, 2007; Cressy et al., 2007) and do not explain the findings related to The Superiority Hypothesis. Instead, they indicate that financial engineering has lost significance from the first to the second PE wave. The findings related to hypothesis 2 are also explained by certain structural changes in the private equity landscape. During the second PE wave, most potential PE targets were relatively highly levered pre-buyout compared to the first wave, which results in a lower potential for debt as a monitoring tool. Moreover, 90% of the PE transactions were of the private-to-

61

private type, where the disciplining effect of increased leverage is expected to be lower 58 as private companies tend to have strong owners already. Three hypotheses were tested to assess the claimed superiority of PE firms in relation to operational engineering and its increasing significance in today’s business landscape. Hypothesis 3 was designed to test PE firms’ skills within cost control, hypothesis 4 dealt with the ability to reduce capital requirements, and finally hypothesis 5 investigated asset efficiency. When testing hypothesis 3 it was found that PE firms in particular focus on cost cutting in order to enhance portfolio firm performance, and that they succeed in implementing these cost reductions in a way that improves overall firm performance. This could explain a substantial part of the findings related to The Superiority Hypothesis (hypothesis 1). The test results related to hypothesis 4 on capital requirements indicated that PE firms are superior at performing working capital management in a value-adding manner. The magnitude of the impact on portfolio firm performance is estimated to be significantly positive, but considerably lower compared to the impact from cost cutting. Hence, this could also explain a part of the findings related to this thesis’ Superiority Hypothesis. When asset efficiency (hypothesis 5) was investigated, no significant performance impact was found. It indicates that this lever is managed similarly in portfolio and control firms, and that neither uses it to create value by improving it postbuyout. Combined with the findings related to hypothesis 3 (cost cutting), it also indicates that PE firms prioritize cost cutting and margins improvements over asset efficiency (asset turnover). In contrast with the two former tests, these results do not serve to explain the findings related to The Superiority Hypothesis. The results of the cost cutting and the capital requirements analyses roughly support the previous findings in the literature, whereas the results from the AT analysis do not. The alleged increased priority of operational engineering within PE firms is emphasized as the main explanation to the above findings regarding hypotheses 3–5. Several authors argue that this lever has become increasingly important in today’s business environment (e.g., Heel & Kehoe, 2005; Dobbs, 2006; Kaplan & Strömberg, 2008; Kehoe & Palter, 2009; Matthews et al., 2009; Archleither et al., 2010; Alperovych et al., 2013). It is argued that PE firms to an increasing extent recruit former industry executives and ex-management consultants to strengthen their operational capabilities, and that PE firms today are generally more specialized within fewer industries compared to earlier (Kaplan & Strömberg, 2008). Finally, the findings provide evidence against the notion that corporations (potential targets) have generally increased efficiency and tightened cost control over the last decades to the extent that short-term benefits from operational improvements have been eroded. In fact, it seems that inefficiencies still exist on which PE firms in particular capitalize. Subsequently, a number of alternative explanations were investigated: endogeneity (from omitted variable bias), sampling bias, survivorship problems and differences in acquisition motives. The role

58

Relative to public-to-private transactions.

62

of the two former elements is assessed to be limited, whereas survivorship bias and the differences in acquisition motives could impact the results to some degree. To wrap it all up and briefly answer this thesis’s research question: PE fund ownership is estimated to impact performance in portfolio firms positively, specifically on these measures: Return on Asset by 2,8 pp, Operating Return on Assets by 2,3 pp, EBIT-Margin by 4,7 pp and Net Profit Margin by 2,3 pp. This indicates that the PE model is in fact (still) a superior organizational form. However, the magnitude of the estimated impact points towards an overall decline in the model’s value creation potential, compared to studies based on data from the first PE wave. PE funds’ financial engineering abilities do not explain the above findings, but superior operational engineering skills seem to do – particularly abilities related to cost cutting and reduction of capital requirements. These results indicate that a change in significance of the different value creation levers have indeed materialized, namely the decreasing importance of financial engineering and increasing importance of operational engineering.

9. Future Research The empirical evidence from the literature strongly suggests that private equity activity creates economic value on average. Kaplan & Strömberg (2008) suspect that increased investments by PE firms in operational engineering will ensure that this phenomenon continues into the future. Because private equity creates economic value, they believe that it has a substantial permanent component. In this context, the following topics are particularly interesting for future research. Part of the impact on firm performance attributed to operational or financial engineering is likely to be caused by governance engineering. Governance engineering has no directly traceable effect on the bottom line; however, the mechanisms leading to the reduction of agency problems can support the other two (i.e., operational and financial engineering) 59. It would be beneficial if future research could establish a method for isolating the effect of the individual levers. Given the explanations provided in this thesis, it is still odd that PE firms should be able to outperform industrial acquirers in terms of operational engineering. Hence, our understanding of value creation within PE funds could benefit from additional research aimed at explaining this phenomenon. This thesis highlights differences in acquisition motives between PE funds and industrial acquirers as an obvious area to investigate further. A number of structural differences between the first and the second PE wave have been identified. These explain some of the changes in the PE model’s overall value creation potential; however,

59

Recall that benefits related to governance engineering stem from enhanced incentive alignment, higher ownership concentration, and better control of the board and management (Jensen, 1989; Jensen & Murphy, 1990).

63

Wright et al. (2009) emphasize that additional future research is needed to gain a broader understanding of the PE model in today’s environment. The present analysis indicates a positive impact of PE fund ownership on portfolio firm performance in the short run, but what is the persistence of this impact? More research is generally needed on this matter, according to Kaplan & Strömberg (2008).

64

References Acharya, V. V., Gottschalg, O. F., Hahn, M., & Kehoe, C. (2012). Corporate Governance and Value Creation: Evidence from Private Equity. Oxford University Press , pp. 369-400. Acharya, V., & Kehoe, C. (2008). “Corporate Governance and Value Creation Evidence from Private Equity.”. Journal of Applied Corporate Finance 19 (4) , pp. 44-53. Alperovych, Y., Amess, K., & Wright, M. (2013). Private equity firm experience and buyout vendor source: What is their impact on Efficiency. European Journal of Operational Research . Anders, G. (1992). The Barbarians in the Boardroom. Harvard Business Review , pp. 79-87. Andrade, G., & Kaplan, S. (1998). How Costly is Financial (Not Economic) Distress? Evidence from Highly Leveraged Transactions That Became Distressed. Journal of Finance 53(5) , pp. 1443-1493. Archleitner, A.-K., Braun, R., Engel, N., Figge, C., & Tappeiner, F. (2010). Value Creation Drivers in Private Equity Buyouts: Empirical Evidence from Europe. Journal of Private Equity , pp. 17-27. Ashenfelter, O., & Card, D. (1985). Using the Longitudinal Structure of Earnings to Estimate the Effect of Training Programs. Review of Economics and Statistics, vol. 67, no. 4 , pp. 648-660. Baker, G. P., & Smith, G. D. (1998). The New Financial Capitalists: Kohlberg Kravis Roberts and the Creation of Corporate Value. Cambridge University Press . Baker, G. P., & Wruck, K. H. (1989). Organizational Changes and Value Creation in Leveraged Buy-Outs: The Case of O.M. Scott & Sons Company. Journal of Financial Economics, vol. 25, no. 2 , pp. 163-190. Barber, B. M., & Lyon, J. D. (1996). Detecting abnormal operating performance: The empirical power and specification of test statistics. Journal of Financial Economics 41 , pp. 359-399. Bergström, C., Grubb, M., & Jonsson, S. (2007). The Operating Impact of Buyouts in Sweden: A study of Value Creation. The Journal of Private Equity , pp. 22-38. Bertrand, M., Duflo, E., & Sendhil, M. (2004). How Much Should We Trust Difference-inDifferences Estimates? The Quarterly Journal of Economics . Brealey, R., & Myers, S. (2003). The Principles of Corporate Finance. Irwin McGraw-Hill Professional Publishing. Bull, I. (1989). Financial Performance of Leveraged Buyouts: An Empirical Analysis. Journal of Business Venturing, vol. 4 , pp. 263-289. Butler, P. A. (2001). The Alchemy of LBOs. The McKinsey Quarterly, no. 2 , pp. 140-151. Cao, J., & Lerner, J. (2009). The Performance of Reverse Leveraged Buyouts. Journal of Financial Economics, vol. 91, no. 2 , pp. 139-157. Cochrane, J. (2005). The Risk and Return of Venture Capital. Journal of Financial Economics (75) , pp. 3-52.

65

Cotter, J., & Peck, S. (2001). The Structure of Debt and Active Equity Investors: The Case of the Buyout Specialist. Journal of Financial Economics(59) , pp. 101-147. Cressy, R., Munari, F., & Malipiero, A. (2007). Playing to their strengths? Evidence that specialization in private equity industry confers competetive advantage. Journal of Corporate Finance , pp. 647-669. Cuny, C., & Talmor, E. (2006). A Theory of Private Equity Turnarounds. Workin paper . Desbriéres, P., & Schatt, A. (2002). The Impacts of LBOs on the Performance of Acquired Firms: the French Case. Journal of Business Finance and Accounting, vol 29 , pp. 695-729. Dobbs, R. (2006). Creating value, The debate over public vs private ownership. McKinsey On Finance - Number 21 , pp. 1-8. Easterwood, J. C., Seth, A., & Singer, R. F. (1989). The Impact of Leveraged Buyouts on Strategic Direction. California Management Review, vol. 31, no 1 , pp. 30-41. EVCA. (2013). Retrieved from EVCA: http://evca.eu/facts/ EVCA(b). (2013). 2012 Pan-European Private Equity and Venture Capital Activity. European private equity and venture capital association , pp. 1-76. Faccio, M., & Lang, L. (2002). The Ultimate Ownership of Western European Corporations. Journal of Financial Economics, vol. 65 , pp. 365-395. Fama, E. F., & French, K. (1995). Size and Book-To-Market Factors in Earnings and Returns. Journal of Finance, vol. 31 , pp. 131-155. Freelink, S., & Volosovych, V. (2012). Do Secondary Buyouts Create Value? Evidence from the United Kingdom. Working Paper . Gertner, R., & Kaplan, S. N. (1996). “The Value Maximizing Board.”. Unpublished paper, University of Chicago . Gompers, P., & Lerner, J. (1996). The Use of Covenants: An Empirical Analysis of Venture Partnership Agreements. Journal of Law and Economics 39 (2) , pp. 463-498. Gottschalg, O., & Berg, A. (2005). Understanding Value Generation in Buyouts. INSEAD - Faculty & Research - Doctoral dissertation . Guo, S., Hotchkiss, E., & Song, W. (2011). Do Buyouts (Still) Create Value? The Journal of Finance Vol. LXVI, No. 2 , pp. 479-517. Harris, R., Siegel, D. S., & Wright, M. (2005). Assessing the Impact of Management Buyouts on Economic Efficiency: Plant-level Evidence from the United Kingdom. Review of Economics and Statistics (87) , pp. 148-153. Heel, J., & Kehoe, C. (2005). Why some private equity firms. The McKinsey Quarterly . Holthausen, R. W., & Larcker, D. F. (1996). The Financial Performance of Reverse Leveraged Buyouts. Journal of Financial Economics, vol 42, no. 3 , pp. 293-332. 66

Hoskisson, R. E., & Hitt, M. A. (1994). Downscoping: How to Tame the Diversified Firm. Oxford University Press . Hwang, M., Quigley, J., & Woodward, S. E. (2005). An Index for Venture Capital, 1987–2003. Contributions to Economic Analysis and Policy . Jensen, M. (1986). Agency Costs of Free Cash Flow, Corporate Finance and Takeovers. American Economic Review 76(2) , pp. 323-329. Jensen, M. (1989). Eclipse of the Public Corporation. Harvard Business Review, Edition: 67 , pp. 61-74. Jensen, M. (2007). The Economic Case for Private Equity. Harvard NOM Research Paper No. 07-02 . Jensen, M., & Murphy, K. J. (1990). Performance Pay and Top Management Incentives. Journal of Political Economy 98(2) , pp. 225-264. Jensen, M., Kaplan, S. S., Ferenbach, C., Feldberg, M., Moon, J., Hoesterey, B., et al. (2006). Morgan Stanley Roundtable on Private Equity and its import for Public Companies. Journal of Applied Corporate Finance 18(3) , s. 8-37. Kaplan, S. N. (1989a). Management Buy-outs: Evidence on taxes as a Source of Value. Journal of Finance 44(3) , pp. 611-632. Kaplan, S. N. (1989b). The Effects of Management Buyouts on Operating Performance and Value. Journal of financial Economics 24(2) , pp. 217-254. Kaplan, S. N., & Strömberg, P. (2008). Leveraged Buyouts and Private Equity. Journal of Economic Perspectives. Volume 22, Number 4 , pp. 1-28. Kaplan, S., & Schoar, A. (2005). Private Equity Performance: Returns, Persistence, and Capital Flows. Journa lof Finance (60) , pp. 1791-1823. Kaufman, A., & Englander, E. (1993). Kohlberg, Kravis Roberts & Co. and the Restructuring of American Capitalism. Business History Review vol 67, no. 1 , pp. 52-97. Kehoe, C., & Palter, R. (2009). The Future of Private Equity. Perspectives on Corporate Finance and Strategy, no. 31 . Kester, C. W., & Luchrman, T. A. (1995). Rehabilitating the Leveraged Buyout. Harvard Business Review , pp. 119-130. Kohlberg, K. R. (1989). Leveraged Buyouts. Journal of Applied Corporate Finance , pp. 64-70. Lichtenberg, F., & Siegel, D. (1990). The Effects of Leveraged Buyouts on Productivity and Related Aspects of Firm Behaviour. Journal of Financial Economics (27) , pp. 165-194. Ljungqvist, A., & Richardson, M. (2003). The Cash Flow, Return, and Risk Characteristics of Private Equity. Working paper No. 9495 NBER .

67

Long, W. F., & Ravenscraft, D. J. (1993). LBOs, Debate and R&D Intensity. Strategic Management Journal, vol. 14 , pp. 119-135. Lowenstein, L. (1985). Management Buyouts. Columbia Law Review, vol. 85 , pp. 730-785. Magowan, P. A. (1989). The Case for LBOs: The Safeway Experience. California Management Review, vol. 23, no. 1 , pp. 9-19. Matthews, G., Bye, M., & Howland, J. (2009). Operation Improvement: The Key to Value Creation in Private Equity. Journal of Applied Corporate Finance, vol. 21, no. 3 . McKenzie, D., & Maslakovic, M. (n.d.). Private Equity. Financial Markets Series , 1-8. Metrick, A., & Yasuda, A. (2010). The Economics of Private Equity Funds. The review of Financial Studies , pp. 2303-2341. Meyer, B. D. (1995). Natural and Quasi-Experiments in Economics. Journal of Business & Economic Statistics, vol. 13, no. 2, , pp. 151-161. Muscarella, C. J., & Vetsuypens, M. R. (1990). Efficiency and Organizational Structure: A study of Reverse LBOs. Journal of Finance vol 45. no. 5 , pp. 1389-1413. Nikoskelainen, E., & Wright, M. (2007). The impact of corporate governance mechanisms on value increase in leveraged buyouts. Journal of Corporate Finance vol. 13(4) , pp. 511-537. Osborne, J. W., & Overbay, A. (2004). The power of outliers (and why researchers should always check for them). Practical Assessment, Research & Evaluation, 9(6) . Palepu, K. (1990). Consequences of Leveraged Buyouts. Journal of Financial Economics, vol. 27 , pp. 247-262. Penman, S. (1991). An Evaluation of Accounting Rate of Return. Journal of Accounting, Auditing and Finance , pp. 233-255. Phalippou, L., & Gottschalg, O. (2009). The Performance of Private Equity Funds. The Review of Financial Studies (22) , pp. 1747-1776. Phan, P. H., & Hill, C. W. (1995). Organizational Restructuring and Economic Performance in Leveraged Buyouts: An Ex Post Study. Academy of Management Journal, vol. 38, no. 3 , pp. 704739. Prowse, S. D. (1998). The Economics of Private Equity Market. Economic Review (3) . Ravenscraft, D., & Scherer, F. (1987). Life after Takeover. Journal of Industrial Economics, vol. 36, no. 2 , pp. 147-156. Renneboog, L., & Simon, T. (2005). Public-to-Private Transactions: LBOs, MBOs, MBIs and IBOs. ECGI Finance working paper no 94 . Singh, H. (1990). Management Buyouts: Distinguishing Characteristics and Operating Changes Prior to Public Offerings. Strategic Management Journal, vol. 11 , p. 111.

68

Smith, A. (1990). Corporate Ownership Structure and Performance. Journal of Financial Economics (27) , pp. 143-164. Smith, A. (1990b). The Effects of Leveraged Buyouts. Business Economics, vol. 25, no. 2 , pp. 19-25. Vinten, F. (2007). The Performance of Private Equity Buyout Fund Owned Firms. Working paper . Wright, M. (1996). The Investor-led Buy-out: A New Strategic Option. Long Range Planning, vol. 29, no. 5 , pp. 691-702. Wright, M., & Robbie, K. (1998). Venture Capital and Private Equity: A Review and Synthesis. Journal of Business Finance and Accounting, vol. 25 , pp. 521-570. Wright, M., Amess, K., Wier, C., & Girma, S. (2009). Private Equity and Corporate Governance: Retrospect and Prospect. Corporate Governance: An International Review , pp. 353–375. Wright, M., Hoskinsson, R. E., & Busenitz, L. W. (2001). Firm Rebirth: Buyouts as Facilitators of Strategic Growth and Entrepreneurship. Academy of Management Executive, vol 15, no. 1 , pp. 111-125. Wright, M., Wilson, N., & Robbie, K. (1997). The Longer term Performance of Management Buyouts. Frontier of Entrepreneurship Research , pp. 555-569.

69

Appendices Appendix 1 - Introduction to Private Equity ............................................................................................................. 71 Appendix 2 - Zephyr Search Criteria ............................................................................................................................. 76 Appendix 3 - Deals Included in the Sample ............................................................................................................... 77 Appendix 4 - Outlier Elimination .................................................................................................................................... 86 Appendix 5 - Quality Check of Data Disclosure (Orbis) – Identity of the Acquirer ............................... 87 Appendix 6 - Quality Check of Data Disclosure (Orbis) – Key Financials................................................... 88 Appendix 7 - Regression Procedure and Testing of OLS Assumptions ....................................................... 90 Appendix 8 - Magnitude of the Impact From Operational Engineering ...................................................... 95 Appendix 9 - Content of Attached CD-ROM ............................................................................................................... 96

70

Appendix 1 - Introduction to Private Equity PE firms establish PE funds, by raising pools of capital (private equity) from investors such as pension funds, insurance companies and wealthy individuals and place it in a fund. The capital is then invested in operating companies typically with an investment horizon of three to seven years60. The PE firm collects fees on an ongoing basis, and when the investments are concluded, the fund is closed, and the remaining equity is handed back to the investors (with a positive or negative return). This construction is elaborated in the following. PE firms PE firms are typically organized as a partnership or a limited liability corporation – see figure . In the US TPG, Blackstone and KKR are some of the most prominent names, whereas CVC, Apax, and EQT are some of the largest players in Europe. Jensen (1989) described PE frims as “lean decentralized organizations with relatively few investment professionals and employees”. He found that the firms on average employ 13 investment professionals, mainly with an investment banking background. Today they are larger, but still very small compared to the portfolio firms they acquire. The largest PE firms today employ around 100 investment professionals. In general PE firms tend to employ professionals with a more diverse background and a wider verity of skill compared to 20 years ago. It has for instance become more common to hire ex industrial executives and ex consultants in order to get a broader understanding of the potential targets pre acquisition and to strengthen operations post acquisition (Acharya et al., 2012; Kaplan & Strömberg, 2008). PE funds As illustrated in figure , a PE firm raises capital through a PE fund, in which investors commit to provide a certain amount of capital to cover investments in companies and management fees to the PE firm itself. These funds are typically established as closed-end vehicles, meaning that investors cannot withdraw their funds until the fund is terminated. This is in strong contrast to mutual funds, where investors can withdraw their capital whenever they like. The funds are organized as limited partnerships where general partners (representing the PE firm) manage the fund and limited partners provide the majority of the capital. The general partners typically provide considerably less capital. However it is customary for them to provide at least 1% of the total. Limited partners are typically institutional investors, such as corporate and public pension funds, endowments, and insurance companies, as well as wealthy individuals.

60

Private equity funds’ investments in European companies are held for about 5 years on average (EVCA, 2013)

71

Figure A1 – Simplified Management and Ownership Structure in Private Equity Private equity firm (General partners / manager of the funds)

Investors (Limited partner)

Private equity fund A

Private equity fund B

(Limited partnership)

(Limited partnership)

Ownership

Target 1

Target 2

Ownership

Target N

(Portfolio company) (Portfolio company) (Portfolio company)

Target 1

Target 2

Target N

(Portfolio company) (Portfolio company) (Portfolio company)

Source: Own creation

This fund typically has a fixed life of ten years which can be extent up to three years. When the capital is committed, the PE firm usually has five years to deploy the capital in a number of target companies (portfolio firms), and subsequently five to eight years to return it to the investors (Kaplan & Strömberg, 2008). When the fund agreement is signed, limited partner have little influence on how the general partners manage the capital, as long as its basic covenants are followed. The agreements usually contain restrictions related to in which securities the fund can invest, how much capital can be deployed in one company and the amount of debt at the fund or holding level. The latter is in contrast to debt at the portfolio company level, which is not restricted (Gompers & Lerner, 1996). Note, that the terms “PE firm” and “PE fund” are used interchangeably in this thesis. Compensation The general partners (or the PE firm) typically receive compensation in three forms. First, they receive an annual management fee during the full lifetime of the fund, which is computed as a percentage of total capital committed and, as investments are realized, a percentage of the capital employed. Second, the general partners earn a share of the fund’s profits, which is referred to as “carried interest” (or simply “carry”) and typically amounts to 20%. Third, sometimes deal and monitoring fees are charged to the companies in which the general partners invest (Metrick & Yasuda, 2010). Phalippou & Gottschalg (2009) find that the average fees claimed by PE funds amount to 6% of the value invested, yearly. A simplified version of the overall cash flow structure is presented in figure below.

72

Figure A2 – Cash Flow Structure of a Typical Private Equity Setup Private equity firm

GP commitment

Management f ees + carried interest

(General partners / managers of the fund)

Equity

Investment

Target

Private equity fund

Investor

(Portfolio company)

(Limited partnership)

(Limited partner)

Gross distribution

Net distributions

Source: Own creation

Transaction and financing A PE firm usually acquire a company, or the majority stake of it, through what is called a leverage buyout or LBOs. These deals are usually financed with a large majority of debt - hence the term which prior to the financial crisis would typically amount to 60-90% of the price. Post financial crisis it has become more complicated to raise debt in the financial market, and consequently this percentage has declined (Guo et al., 2011). The debt typically consists of a portion which is senior and secured plus a portion which is junior and unsecured (see description of mezzanine capital further below). The PE firm then applies funds from its investors as equity to cover the remaining 10 to 40% of the investment. The management teams, which has been picked by the PE firm to run the portfolio company post acquisition (could be the current team, a set of executives found externally or a combination hereof) usually also contribute with some equity, though the amount is typically only a small fraction of the total investment. Mezzanine capital, which is a hybrid of debt and equity financing, is often used applied in these deals. It is basically debt capital that gives the lender the rights to convert to an ownership or equity interest in the company if the loan is not paid back fully, in time. It is typically subordinated to debt provided by senior lenders (e.g. banks and venture capital companies). As mezzanine financing is usually provided fast, with little due diligence on the part of the lender and little or no collateral on the part of the borrower, this type of financing is relatively costly. Private equity firms typically use this form of financing to reduce the amount of equity capital required to finance a leveraged buyout. Other less common investment strategies within private equity are venture capital and growth capital. These are not dealt with thoroughly in this thesis however, as the data material only comprise LBO transactions. In brief, venture capital is at large, very similar to a typical private equity investment. Its most distinguishing feature is that mainly small start-up companies are targeted. This type of investment is most often seen in the application of new technology, new marketing concepts and new products that do not have a proven track record or stable revenue streams. Growth Capital refers to 73

investments in mature companies that are looking for capital to grow further. In this setup, private equity companies usually supply a minority investment and thus, do not obtain control of the target. Popularity, returns and fees It is not within the scope of this thesis to assess the role and magnitude of fees claimed by PE funds, but a brief description hereof is in order to aid the readers understanding of private equity as a concept. As seen above, the popularity of private equity funds have increased dramatically over the past 30 years. In 1980 about 5 EURbn was committed to these funds on a global basis, and at their peak in 2007, the number reached 530 EURbn (McKenzie & Maslakovic). According to the media, this spectacular growth in popularity can mainly be attributed to the high past performance of the funds an assertion that is not unambiguously supported by the literature, where results are largely conflicting. Several papers point towards that private equity firms historically have been able to outperform relevant market indices such as the S&P-500 (e.g. Cochrane, 2005; Hwang et al., 2005), whereas other researchers find that PE firms historically have underperformed these (e.g. Ljungqvist & Richardson, 2003; Kaplan & Schoar, 2005). The conclusions of the papers in the latter group are mainly based on returns net-of-fees, and they criticize the authors of the first group for not recognizing the major role which fees play. The majority of recent research can be put in the latter group. Phalippou & Gottschalg (2009) for instance find that the average fees claimed by PE funds amount to 6%, and returns of the funds are 3% lower compared to the S&P 500 Index net-of-fees, and 3% higher gross-of-fees. The marketing material of PE firms, however, typically reports gross-of-fee returns which naturally make their funds seem very attractive. Regardless the size of the fees and the performance net hereof, the popularity of PE funds has increased sensationally since the 1980s, and pioneer researchers within finance, such as Michael Jensen and Steven Kaplan, have repeadetly concluded, that private equity as organizational form, is here to stay. Increasing occurrence of secondary buyouts A secondary buyout (SBO) has the distinguishing characteristic that both seller and buyer are private equity funds. During the last decade SBOs have emerged as an important phenomenon. In the 1980s, the SBO sub-market accounted for only 2% of the total PE investments, but in the four years in the wake of the financial crisis, its share grew to 26% (Kaplan & Strömberg, 2008). Thus, today one in four PE deals in Europe is an SBO. Particularly one notion about SBOs has induced scepticism of the concept amongst researchers, and constitutes an interesting aspect in the context of the present analysis: The potential of operating performance improvements following a SBO could be limited, as the first PE fund expectedly has implemented the easiest improvements with the largest impact already. Because a certain homogeneity in PE fund skills is expected, it is questionable whether the second PE fund can capitalize much on an unexploited improvement potential. With offset in this logic, the PE model was not indented to become a close circuit structure, where PE funds traded with each other. Several studies support this

74

hypothesis by reporting a negative operational performance impact in portfolio firms after an SBO (e.g. Freelink & Volosovych, 2012). Naturally, the acquiring PE fund would not take part in the transaction, if the general partners did not believe they could make a positive return of the investment. One explanation of how they achieve this, could be that the second PE fund leverage superior market timing and negotiating skills, and consequently does not add economic value to their SBO portfolio firms (e.g. Freelink & Volosovych, 2012). Despite this scepticism, some argue that SBOs do create economic value in the portfolio firms in certain situations (e.g. Freelink & Volosovych, 2012). However, this discussion is found beyond the scope of this thesis. If anything, the anticipated effect of including SBOs in the present analysis, is a downwards pressure on portfolio firm performance, which only serves to validate the main findings of this thesis (i.e. PE fund ownership induce superior performance in their portfolio firms).

75

Appendix 2 - Zephyr Search Criteria Search criteria – portfolio firms Product name

Zephyr

Update number

30

Software version

30.0

Data update

11/03/2013 (n° 3074106)

Username

Aarhus Business School-8246

Export date

12/03/2013 Step result Search result

1. Deal types: Institutional buy-out 2. Time period: on and after 01/01/2003 and up to and including 31/12/2008 (completed, include updated deals, include new deals) 3. World regions: European Union enlarged (27) 4. Percentage of stake: Percentage of initial stake (max: 0 %); Percentage of acquired stake (min: 51 %)

18,616 567,967

18,616 18,443

382,414 346,151

8,618 6,124

5. Deal status: Completed

729,986

4,157

Search criteria – control firms Product name

Zephyr

Update number

30

Software version

30.0

Data update

11/03/2013 (n° 3074106)

Username

Aarhus Business School-8246

Export date

12/03/2013 Step result Search result

1. Deal types: Acquisition

456,986

456,986

2. Time period: on and after 01/01/2003 and up to and including 31/12/2008 (completed, include updated deals, include new deals)

567,967

171,86

3. World regions: European Union enlarged (27)

382,414

46,191

4. Percentage of stake: Percentage of initial stake (max: 0 %); Percentage of acquired stake (min: 51 %)

346,151

16,991

5. Deal status: Completed

729,986

9,497

76

Appendix 3 - Deals Included in the Sample Note that deals where the acquirer is labelled “MBO TEAM – COUNTRY” have all been checked to ensure, that the lead acquirer in fact was a PE fund. List of Acquisitions Performed by Private Equity funds Included in the Sample Acquirer

Location Target

Location Year

ANDLINGER & COMPANY

BE

PRIMUS

BE

2003

2026140 ONTARIO INC.

GB

WET AUTOMOTIVE SYSTEMS AG

DE

2003

LØNMODTAGERNES DYRTIDSFOND

DK

NORISOL A/S

DK

2003

ABN AMRO VENTURES BV

NL

LABIANA LIFE SCIENCES SA

ES

2003

CASTLE HARLAN INC.

US

TECHNIFOR SA

FR

2003

MBO TEAM - FRANCE

FR

E-CITY

FR

2003

CHEQUERS CAPITAL PARTNERS SA

FR

EUROFARAD SAS

FR

2003

DOLPHINS SCHOOLS LTD, THE

GB

HONORMEAD SCHOOLS LTD

GB

2003

MBO TEAM - UNITED KINGDOM

GB

HOBBS LTD

GB

2003

SOVEREIGN CAPITAL LTD

GB

LAUREL MANAGEMENT SERVICES LTD

GB

2003

MBO TEAM - UNITED KINGDOM

GB

NCC GROUP PLC

GB

2003

MBO TEAM - BELGIUM

BE

VELLEMAN COMPONENTS NV

BE

2004

ARMAND CAPITAL GROUP, THE

US

SALUC SA

BE

2004

GENESIS CAPITAL SRO

CZ

PIETRO FILIPI SRO

CZ

2004

HANNOVER FINANZ GMBH

DE

HEGO PARTNER HOLDING GMBH

DE

2004

BACPE FINLAND HOLDINGS OY

FI

JANTON OYJ

FI

2004

MANAGEMENT

FR

COMPAGNIE FRANCAISE DA SA

FR

2004

NORDSTJERNAN AB

SE

ADMV

FR

2004

LIME ROCK PARTNERS LLC

US

SERIMER DASA SAS

FR

2004

NORDSTJERNAN AB

SE

KALIX SA

FR

2004

CICLAD PARTICIPATIONS SAS

FR

FAURÉ HERMAN SAS

FR

2004

ALCHEMY PARTNERS LLP

GB

SWIFT ADVANCES PLC

GB

2004

ADAPTGROUND LTD

GB

TRACS LTD

GB

2004

RUTLAND PARTNERS LLP

GB

HARVEY & THOMPSON LTD

GB

2004

MBO TEAM - UNITED KINGDOM

GB

OYEZSTRAKER GROUP LTD

GB

2004

77

HICKS MUSE TATE & FURST INC.

US

J CHOO LTD

GB

2004

CVC CAPITAL PARTNERS LTD

GB

ANI PRINTING INKS

SE

2004

BV CAPITAL PARTNERS

BE

MATCO NV

BE

2005

MBO TEAM - BELGIUM

BE

PACKING CREATIVE SYSTEMS NV

BE

2005

MBO TEAM - BELGIUM

BE

EMERSON & CUMING MICROWAVE PRODUCTS

BE

2005

LITORINA KAPITAL MANAGEMENT AB

SE

GM-ITM A/S

DK

2005

ACTIVA CAPITAL SAS

FR

PRO NATURA SAS

FR

2005

MBO TEAM - FRANCE

FR

NORPROTEX SA

FR

2005

ROCAFIN

FR

ROCAMAT SA

FR

2005

ATRIA CAPITAL PARTENAIRES SA

FR

EUROPEAN HOMES FRANCE

FR

2005

TSL EDUCATION HOLDINGS LTD

GB

TSL EDUCATION LTD

GB

2005

MBO TEAM - UNITED KINGDOM

GB

SUPERGLASS INSULATION LTD

GB

2005

SANCTUARY SPA HOLDINGS LTD, THE

GB

SANCTUARY CONNECTIONS LTD, THE

GB

2005

MBO TEAM – UNITED KINGDOM

GB

FRANK SMYTHSON LTD

GB

2005

MBO TEAM - ITALY

IT

LUXY SRL

IT

2005

EUROPAINT INTERNATIONAL BV

NL

DUFA ROMANIA SRL

RO

2005

LITORINA KAPITAL MANAGEMENT AB

SE

LBC SWEDEN AB

SE

2005

NORDSTJERNAN AB

SE

NILS HANSSON ÅKERI AB

SE

2005

SEGULAH AB

SE

HEXAGON AUTOMATION AB

SE

2005

SCHOELLER WAVIN SYSTEMS NV

NL

ARCA SYSTEMS INTERNATIONAL AB

SE

2005

NC ADVISORY AB

SE

ATOS MEDICAL AB

SE

2005

NOVAX AB

SE

DESIGNTORGET AB

SE

2005

CHEQUERS CAPITAL PARTNERS SA

FR

TCR INTERNATIONAL NV

BE

2006

MBO TEAM - GERMANY

DE

VESTOLIT GMBH & CO. KG

DE

2006

AXCEL INDUSTRIINVESTOR A/S

DK

NETCOMPANY A/S

DK

2006

LD INVEST EQUITY

DK

SFK SYSTEMS A/S

DK

2006

MANAGEMENT

ES

INDIBA SA

ES

2006

QUALITAS EQUITY PARTNERS SA SGECR

ES

FOMENTO Y PREFABRICADOS DE ARAGÓN

ES

2006

EXPLORER I

PT

AYSAV SL

ES

2006

LABORATORIOS BONIQUET

ES

SPARCHIM SL

ES

2006

78

PALUEL-MARMONT CAPITAL SA

FR

ANTEL TÉLÉBAT SA

FR

2006

AXA INVESTMENT MANAGEMENT PE EU

FR

AIXAM-MEGA SA

FR

2006

LBO FRANCE GESTION SAS

FR

SAM + SA

FR

2006

MANAGEMENT

FR

SIGNAUX LAPORTE SAS

FR

2006

SOCADIF SA

FR

GEOLINK SAS

FR

2006

EURAZEO SA

FR

EUROPCAR INTERNATIONAL SAS

FR

2006

BLACKSTONE CAPITAL PARTNERS LP

US

TRIANON PALACE HÔTEL SA

FR

2006

PROMOTION CHIRURGICALE LOIRE ÎLE

FR

CLINIQUE LAMBERT SA

FR

2006

BUTLER CAPITAL PARTNERS SA

FR

SOCIÉTÉ NATIONALE MCM SA

FR

2006

MBO TEAM - UNITED KINGDOM

GB

WHITWORTHS HOLDINGS LTD

GB

2006

3I GROUP PLC

GB

MARKEN LTD

GB

2006

CZURA THORNTON LTD

GB

CHILTERN INTERNATIONAL LTD

GB

2006

SOVEREIGN CAPITAL PARTNERS LLP

GB

NATIONAL FOSTERING AGENCY LTD, THE

GB

2006

UK SOFTWARE HOLDINGS LTD

GB

XKO SOFTWARE LTD

GB

2006

MBO TEAM - UNITED KINGDOM

GB

PARASOL LTD

GB

2006

SANPAOLO IMI SPA

IT

BIOLCHIM SPA

IT

2006

SHAREHOLDERS

IT

CONTROLS SRL

IT

2006

MANAGEMENT

SE

SCANDBOOK AB

SE

2006

BANK FÜR TIROL UND VORARLBERG AG

AT

TIROLER RÖHREN- UND METALLWERKE AG

AT

2007

FORTIS PRIVATE EQUITY NV

BE

ANTILOPE NV

BE

2007

MBO TEAM - BELGIUM

BE

PENNE NV/SA

BE

2007

MANAGEMENT

BE

WOLLUX

BE

2007

ARQUES INDUSTRIES AG

DE

VAN NETTEN GMBH

DE

2007

GCI BRIDGECAPITAL AG

DE

BOHNACKER AG

DE

2007

POLARIS MANAGEMENT A/S

DK

HAMLET PROTEIN AS

DK

2007

POLARIS MANAGEMENT A/S

DK

SKAMOL A/S

DK

2007

ESPIGA CAPITAL GESTIÓN SGECR SA

ES

2003 SA

ES

2007

VISTA CAPITAL DE EXPANSION SA SGECR

ES

LABORATORIOS INDAS SA

ES

2007

BERLING CAPITAL OY

FI

AIR FINLAND OY

FI

2007

FINANCIÈRE AIGLE 2

FR

IMV TECHNOLOGIES SA

FR

2007

79

SASA HOLDING ET MANAGEMENT

FR

SASA INDUSTRIE SA

FR

2007

BARCLAYS PRIVATE EQUITY LTD

GB

PARKEON SAS

FR

2007

SPIRIT CAPITAL SA

CH

ETS DEBRISE DULAC ET CIE SA

FR

2007

MANAGEMENT

FR

PROFIMO SAS

FR

2007

FINANCIÈRE GAILLON 8 SAS

FR

KAUFMAN & BROAD SA

FR

2007

MBO TEAM - FRANCE

FR

MENUISERIE GRÉGOIRE SA

FR

2007

JLL PARTNERS INC.

US

LINREAD LTD

GB

2007

LATTIMER HOLDINGS LTD

GB

LATTIMER LTD

GB

2007

HERO ACQUISITIONS LTD

GB

HSS HIRE SERVICE GROUP LTD

GB

2007

MBO TEAM - UNITED KINGDOM

GB

FERNAU AVIONICS LTD

GB

2007

MBO TEAM - UNITED KINGDOM

GB

BECK & POLLITZER ENGINEERING LTD

GB

2007

ADELIE FOOD HOLDINGS LTD

GB

BRAMBLES FOOD LTD

GB

2007

MBO TEAM - UNITED KINGDOM

GB

MONEYBOOKERS LTD

GB

2007

MB AEROSPACE HOLDINGS LTD

GB

MB AEROSPACE LTD

GB

2007

QUADRIVIO SGR SPA

IT

ARBO SRL

IT

2007

F.LLI ELIA SRL

IT

GIOVANNI AMBROSETTI AUTO LOGISTICA SPA

IT

2007

ASKEMBLA ASSET MANAGEMENT AB

SE

UAB KAUSTA GUDER

LT

2007

MG BALTIC INVESTMENT UAB

LT

MEDIAFON UAB

LT

2007

LODE HOLDINGS LTD

CY

LODE AS

LV

2007

ALTA CAPITAL PARTNERS

LV

RIGAS PIENSAIMNIEKS AS

LV

2007

LYNX PROPERTY BV

NL

DOMO RETAIL SA

RO

2007

GED CAPITAL DEVELOPMENT

ES

HAPPY TOUR SRL

RO

2007

NEWCO

SE

NACKA NÄRSJUKHUS PROXIMA

SE

2007

ENTERPRISE INVESTORS SP ZOO

PL

STD SLOVAKIA SRO

SK

2007

INDUFLEX HOLDING

BE

ROGERS INDUFLEX NV

BE

2008

ATITLAN ALPHA SGCR

ES

VERDIFRESH SL

ES

2008

MAGNUM INDUSTRIAL PARTNERS SL

ES

PRETERSA PRENAVISA EDH SL

ES

2008

SANTANDER INFRAESTRUCTURAS FCR

ES

TERMINAL MARITIMA DE GRANELES SL

ES

2008

SENTICA PARTNERS OY

FI

DAREKON OY

FI

2008

MANAGEMENT

FI

KOJA TEKNIIKKA OY

FI

2008

80

MBO TEAM-FRANCE

FR

OCÉDIS

FR

2008

LLOYDS TSB DEVELOPMENT CAPITAL LTD

GB

BULLOCK CONSTRUCTION LTD

GB

2008

SEQUOIA CAPITAL

US

DATA CONNECTION LTD

GB

2008

CDR TABASCO LTD

GB

BODYCOTE TESTING GROUP LTD

GB

2008

EVANS HOLDINGS LTD

GB

FW EVANS CYCLES (UK) LTD

GB

2008

MBO TEAM - UNITED KINGDOM

GB

COFFEE NATION LTD

GB

2008

STRATEGIC TEAM GROUP LTD

GB

STRATEGIC TEAM MAINTENANCE.CO LTD

GB

2008

MBO TEAM - UNITED KINGDOM

GB

XLN TELECOM LTD

GB

2008

CREDEM PRIVATE EQUITY SGR

IT

POPLAST SRL

IT

2008

RIELLO INVESTIMENTI SPA

IT

C BLADE SPA

IT

2008

MBO TEAM - ITALY

IT

MICROTECNICA SRL

IT

2008

MBO TEAM - ITALY

IT

GLASS IDROMASSAGGIO SRL

IT

2008

EGERIA BV

NL

H TEN HERKEL BV

NL

2008

CARGO PARTNER GROUP AS

NO

GLOBEX TRANSPORT AB

SE

2008

LITORINA KAPITAL MANAGEMENT AB

SE

TEXTILIA TVÄTT & TEXTILSERVICE AB

SE

2008

81

List of Acquisitions Performed by Industrial Companies Included in the Sample

Acquirer

Location

Target

Location

OMEGA PHARMA NV

BE

ROIG FARMA SA

ES

29-012003

EMAP COMMUNICATIONS LTD

GB

AGOR SAS

FR

05-022003

DLF TRIFOLIUM A/S

DK

CEBECO SEEDS GROUP BV

NL

28-022003

HOMESERVE PLC

GB

SERVOTOMIC LTD

GB

03-032003

CORPORACIÓN NOROESTE SA

ES

HORMIGONES Y MINAS SA

ES

11-072003

GIRO GH SA

ES

RODA PACKING SA

ES

17-072003

ARCOTRONICS ITALIA SPA

IT

ROEN EST SRL

IT

20-092003

CARTA MUNDI NV

BE

GAMES AND PRINT SERVICES LTD

GB

06-112003

AVIZA FRANCE

FR

ASML FRANCE SARL

FR

25-012004

TAYLOR NELSON SOFRES PLC

GB

AREA INVESTIGACIÓN SA

ES

11-022004

GAMMA TELECOMMUNICATIONS LTD

GB

UNI WORLD COMMUNICATIONS LTD

GB

27-022004

REXAM PLC

GB

PLASTIC OMNIUM MEDICAL SA

FR

08-042004

WISDOM GROUP, THE

ES

PROEIN SL

ES

27-042004

KONINKLIJKE VOLKER WESSELS STEVIN NV

NL

FITZPATRICK PLC

GB

17-052004

MM PACKAGING ROMANIA SA

RO

RODATA SA

RO

27-052004

LUNCHEON VOUCHERS LTD

GB

CAPITAL INCENTIVES & MOTIVATION LTD

GB

04-062004

MAPLES FINANCE JERSEY LTD

GB

GARTMORE FUND MANAGERS LTD

GB

10-062004

NOVASEP SAS

FR

APPLEXION SA

FR

21-062004

MR NICOLAS WARD

GB

EAST YORKSHIRE ALUMINIUM & GLASS LTD

GB

27-072004

FININFO SA

FR

DUN & BRADSTREET FRANCE SAS

FR

01-102004

GALAXY SA

FR

TECHNIQUE BÉTON

FR

31-102004

AUTOCOMMERCE DD

SI

ADRIA MOBIL DOO

SI

12-112004

OMERIN SAS

FR

FLEXELEC SA

FR

03-122004

ALLTRACEL PHARMACEUTICALS PLC

IE

WESTONE PRODUCTS LTD

GB

22-122004

STARHOTELS SPA

IT

CASTILLE SAS

FR

29-012005

CP KELCO APS

DK

NOVIANT OY

FI

07-022005

Date

82

GROUPE DOUCET

FR

PRIMALAB SA

FR

28-022005

SHAREHOLDERS

ES

STEEL BETON ESPAÑOLA SA

ES

23-042005

NATRA SA

ES

CHOCOLATERIE JACALI

BE

27-052005

BONFIGLIOLI RIDUTTORI SPA

IT

TECNOINGRANAGGI RIDUTTORI SRL

IT

31-052005

SOCIEDADE COMERCIAL OREY ANTUNES SA

PT

AGENCIA MARITIMA DE CONSIGNACIONES SA

ES

01-062005

CORROSION MATERIALS LTD

GB

SPECIAL METALS SERVICES SA

FR

01-062005

BEKAERT SA/NV

BE

SOUTHWEST SCREENS & FILTERS SA

BE

02-062005

BERGMAN & BEVING AB

SE

G SOHLBERG AB

SE

15-062005

ALK-ABELLÓ A/S

DK

ALLERBIO SA

FR

29-062005

DHL IBERIA

ES

ÁLVAREZ SILVA

ES

09-072005

MATERIS HOLDINGS LUXEMBOURG SA

LU

ZOLPAN SA

FR

29-072005

CRH PLC

IE

MARMORITH NV

BE

31-072005

JAMES DONALDSON & SONS LTD

GB

MGM TIMBER (SCOTLAND) LTD

GB

12-082005

VOLATI AB

SE

SVEICO AB

SE

18-082005

RAGLETH LTD

GB

ANGLIA MALTINGS HOLDINGS LTD

GB

20-092005

ID LOGISTICS SAS'S MANAGEMENT

FR

LA FLÊCHE CAVAILLONAISE

FR

06-102005

LOOKERS PLC

GB

APEC LTD

GB

12-102005

EXPO-AN SA

ES

GRUPO INMOCARAL SA

ES

21-112005

ACOTEX - VELOUTA NV

BE

ARTILAT NV

BE

14-122005

TORRECID SA

ES

COLORES CERÁMICOS DE TORTOSA SA

ES

14-122005

PERSIMMON PLC

GB

SENATOR HOMES LTD

GB

14-122005

BANQUE MARTIN MAUREL SA

FR

INTERNATIONAL CAPITAL GESTION SA

FR

06-012006

LOHJA RUDUS OY AB

FI

ELPOTEK OY

FI

13-012006

ROTORK PLC

GB

OMAG SNC

IT

13-012006

LEKKERLAND

DE

MACROMEX SRL

RO

13-012006

PASQUIER SA

FR

SYMPHONIE

FR

01-022006

KONINKLIJKE REESINK NV

NL

PACKO AGRI

BE

10-022006

ARK-H

GB

DISS PROMOTIONAL SERVICES

GB

28-022006

FINANCIÈRE ACCÈS INDUSTRIE

FR

ACCÈS INDUSTRIE SA

FR

06-032006

83

ALTIA OYJ

FI

MOBIL PLUS ADV SIA

LV

24-032006

BODYCOTE INTERNATIONAL PLC

GB

SAAB METECH AB

SE

27-032006

DISTRIPAR SA

BE

CLUB NV

BE

01-042006

PIRELLI RE AGENCY

IT

PEKAO DEVELOPMENT SP ZOO

PL

03-042006

SOMFY SA

FR

COTHERM SAS

FR

17-052006

CAISSE DES DÉPÔTS ET CONSIGNATIONS

FR

MÉDIPRÉMA SA

FR

19-052006

VINCI ENERGIES BELGIUM NV

BE

PROMATIC-B NV

BE

31-052006

MR PEKKA LAITINEN

FI

VAROVA OY

FI

16-062006

FELDING FINANCE BV

NL

DONOVAN MEDICAL EQUIPMENT LTD

IE

30-062006

SITA UK LTD

GB

HEMMINGS WASTE MANAGEMENT LTD

GB

05-072006

AVENIR TELECOM SA

FR

AKS INTERNATIONAL EOOD

BG

07-072006

TALLINK GRUPP AS

EE

SILJA OYJ ABP

FI

19-072006

SAS COMPONENT GROUP A/S

DK

AIRLINE ROTABLES LTD

GB

21-072006

E2V TECHNOLOGIES SAS

FR

ATMEL GRENOBLE SAS

FR

31-072006

EURIZON FINANCIAL GROUP

IT

ISYDE SRL

IT

15-082006

INSPACE PLC

GB

WILLMOTT DIXON HOUSING LTD

GB

01-092006

PIKOLIN SA

ES

SMATTEX Y COLCHONES MEDITERRÁNEO SA

ES

06-092006

PAMESA CERÁMICA SL

ES

ARCILLAS ATOMIZADAS SA

ES

18-102006

ILIRIJA RAZVOJ, PROIZVODNJA IN TRZENJE KOZMETICNIH IZDELKOV DD, LJUBLJANA

SI

LEK KOZMETIKA

SI

13-112006

COLLIERS CRE PLC

GB

PAUL & COMPANY

GB

13-122006

TICKET TRAVEL GROUP AB

SE

MZ TRAVEL AB

SE

13-122006

SOMACO INDUSTRIES SAS

FR

PATRICOLA ENTREPRISE SAS

FR

31-122006

FUJITSU SERVICES OVERSEAS HOLDING LTD

GB

TDS INFORMATIONSTECHNOLOGIE AG

DE

18-012007

SPIE SUD-EST SAS

FR

MOUILLOT & CIE SA

FR

25-012007

MR JEAN-CLAUDE LE BLEIS

FR

NKE SA

FR

07-022007

NEW-STEEL OY

FI

JORMET OY

FI

12-022007

MNF CAPITAL

PT

MERCATUS SA

PT

14-022007

POLKACREST LTD

GB

ATTERO SERVICES LTD

GB

19-022007

84

WELLNESS FOODS LTD

GB

STREAMFOODS LTD

GB

28-022007

NEXTRADIOTV SA

FR

GROUPE TESTS SA

FR

04-042007

PASTIFICIO RANA SPA

IT

MAMA LUCIA

BE

24-042007

CESKE VINARSKE ZAVODY AS

CZ

VINIUM AS

CZ

27-042007

MAISONS FRANCE CONFORT SA

FR

PCA MAISONS SAS

FR

02-052007

AQUARIA DE INV CORP SA

ES

IRRIGARONNE SAS

FR

29-052007

SOMARO SA

FR

NORDGALVA SAS

FR

31-052007

PINGUIN SA/NV

BE

PADLEY VEGETABLES LTD

GB

04-062007

LATVIJAS MOBILAIS TELEFONS SIA

LV

ZETCOM SIA

LV

06-062007

AUTODISTRIBUTION SA

FR

AD POLSKA SP ZOO

PL

26-062007

PREGIS GMBH

DE

PETROFLAX SA

RO

05-072007

CREATUFT

BE

TASIBEL NV

BE

06-072007

OA ACQUISITIONS LTD

GB

OXFORD AVIATION SERVICES LTD

GB

20-072007

LAUNET FINANCE SAS

FR

LAUNET

FR

31-072007

BODYCOTE INTERNATIONAL PLC

GB

NITRUVID SAS

FR

01-082007

JARO SA

PL

FABRYKA PORCELANY WALBRZYCH SA

PL

28-082007

IMMUNODIAGNOSTIC SYSTEMS HOLDINGS PLC

GB

BIOCODE HYCEL

BE

31-082007

RED HOUSE SA

LU

UNIBRA SA

BE

03-092007

SCHENKER AG

DE

SPAIN TIR TRANSPORTES INTERNACIONALES SA

ES

28-092007

EUROMEDIC INTERNATIONAL BV

NL

CLÍNICA CENTRAL DO BONFIM SA

PT

03-102007

SCEPTRE LEISURE SOLUTIONS LTD

GB

CROWN LEISURE PLC

GB

18-102007

KRKA DD

SI

TAD PHARMA GMBH

DE

09-112007

POLYPEPTIDE LABORATORIES GROUP BV

NL

NEOMPS SA

FR

19-112007

COOPERNIC

BE

PALINK UAB

LT

28-112007

ARCELOR MITTAL NV

NL

VALLOUREC PRÉCISION SOUDAGE SAS

FR

11-122007

LABCO SA

FR

GENERAL LAB SA

ES

21-122007

KOOPERATIVA FÖRBUNDET

SE

DAGLIVS

SE

31-122007

ELEKTRA SA

ES

ELECTRICIDAD GUERRA SA

ES

14-012008

MITISKA SA

BE

MG FINANCES SA

BE

31-012008

85

WPP DIGITAL

GB

HEATHWALLACE LTD

GB

14-022008

INYECCIONES PLASTICAS MECACONTROL SL

ES

PLASTICOS SOPLADOS TECNICOS SA

ES

03-032008

INVITEL ZRT

HU

MEMOREX TELEX COMMUNICATIONS AG

AT

05-032008

NGC MEDICAL SPA

IT

AVIONORD SRL

IT

14-032008

MITIE GROUP PLC

GB

DW TILLEY LTD

GB

27-032008

MERCOR SA

PL

TECRESA PROTECCIÓN PASIVA SL

ES

09-042008

KONECRANES OYJ

FI

AUSIÓ SISTEMAS DE ELEVACIÓN SL

ES

21-042008

EXPRIVIA SPA

IT

SPEGEA - SCUOLA DI MANAGEMENT SCARL

IT

19-052008

PASPORTE LTD

GB

27-052008

MANAGEMENT BHJ A/S

DK

CORSA PETFOOD SL

ES

02-062008

OFFICE2OFFICE PLC

GB

ACCORD OFFICE SUPPLIES LTD

GB

03-062008

POLSKIE PRZEDSIEBIORSTWO WYDAWNICTW KARTOGRAFICZNYCH IM EUGENIUSZA ROMERA SA

PL

DLUGIE ROZMOWY SA

PL

04-072008

FRÖSUNDA LSS AB

SE

ARGOLIS AB

SE

11-072008

SPIE OUEST-CENTRE SAS

FR

STE GEFCA SARL

FR

14-102008

KIRK KAPITAL A/S

DK

DIFKO A/S

DK

05-122008

EUROMEDIC ROMANIA SRL

RO

NEFROMED DIALYSIS CENTERS SRL

RO

11-122008

KESWICK ENTERPRISES LTD

GB

LINK LOGISTICS LTD

GB

31-122008

CARBONE LORRAINE SA

FR

CALCARB LTD

GB

31-122008

Appendix 4 - Outlier Elimination Osborne & Overbay, (2004) suggest deleting data points which are three or more standard deviations from the mean. However, this approach can produce problems when applied to highly skewed distributions. Therefore, the distribution of each individual variable was assessed before the rule was applied. In cases of highly skewed distributions, outliers were eliminated based on graphical inspection. In addition several checks were made for logically inconsistent figures, for instance: 

Growth rates below “-1”



Debt ratios with a value below “0” or above “1”



Working capital ratios with a value below “-1” or above “1”



Cost margins below “0”

86

Appendix 5 - Quality Check of Data Disclosure (Orbis) – Identity of the Acquirer Acquiror Name

PE Fund? Responsible PE Fund

Target Name

Comment

BV CAPITAL PARTNERS

BE

MATCO NV

BE

YES

BV CAPITAL PARTNERS

POLARIS MANAGEMENT A/S

DK

HAMLET PROTEIN AS

DK

YES

POLARIS MANAGEMENT A/S

CVC CAPITAL PARTNERS LTD

GB ANI PRINTING INKS

SE

YES

CVC CAPITAL PARTNERS LTD

CARGO PARTNER GROUP AS

NO GLOBEX TRANSPORT AB

SE

YES

CAPMAN PARTNERS OY

Acquired via Cargo Partners Group AS

MBO TEAM - BELGIUM

BE

PENNE NV/SA

BE

YES

FORTIS PRIVATE EQUITY

Fortis Private Equity w as the leading acquirer

ARMAND CAPITAL GROUP, THE

US

SALUC SA

BE

YES

ARMAND CAPITAL GROUP, THE

MBO TEAM - GERMANY

DE

VESTOLIT GMBH & CO. KG

DE

YES

STRATEGIC VALUE PARTNERS

ABN AMRO VENTURES BV

NL

LABIANA LIFE SCIENCES SA

ES

YES

ABN AMRO VENTURES BV

BACPE FINLAND HOLDINGS OY

FI

JANTON OYJ

FI

YES

BANK OF AMERICA PE - FINLAND

NORDSTJERNAN AB

SE

ADMV

FR

YES

NORDSTJERNAN AB

LIME ROCK PARTNERS LLC

US

SERIMER DASA SAS

FR

YES

LIME ROCK PARTNERS LLC

ROCAFIN

FR

ROCAMAT SA

FR

YES

LBO FRANCE GESTION SAS

FINANCIÈRE GAILLON 8 SAS

FR

KAUFMAN & BROAD SA

FR

YES

PAI PARTNERS SAS

CZURA THORNTON LTD

GB CHILTERN INTERNATIONAL LTD

GB

YES

CZURA THORNTON LTD

CDR TABASCO LTD

GB BODYCOTE TESTING GROUP LTD

GB

YES

CLAYTON DUBILIER & RICE INC.

HICKS MUSE TATE & FURST INC.

US

J CHOO LTD

GB

YES

HICKS MUSE TATE & FURST INC.

QUADRIVIO SGR SPA

IT

ARBO SRL

IT

YES

QUADRIVIO SGR SPA

ALTA CAPITAL PARTNERS

LV

RIGAS PIENSAIMNIEKS AS

LV

YES

ALTA CAPITAL PARTNERS

SCHOELLER WAVIN SYSTEMS NV

NL

ARCA SYSTEMS INTERNATIONAL AB

SE

YES

STIRLING SQUARE CAPITAL PARTNERS LLP

ENTERPRISE INVESTORS SP ZOO

PL

STD SLOVAKIA SRO

SK

YES

ENTERPRISE INVESTORS SP ZOO

Strategic Value Partners w as the leading acquirer

Acquired via Financíere Gaillon 8 SAS

Acquired via CDR Tabasco LTD

Acquired via. Schoeller Wavin Systems

87

Appendix 6 - Quality Check of Data Disclosure (Orbis) – Key Financials All data drawn form Orbis are measured in EUR, whereas the data obtained from annual reports are denominated in the target country’s home currency. Consequently the majority of the deviations are attributed to differences in applied exchange rates by Orbis and the present analysis. Thus, a relatively high tolerance is applied to judge the magnitude of the deviations. In the two figures below, all deviations exceeding 5% numerically have been marked with red. Through a qualitatively assessment, the deviations are not regarded problematic. Figure A3 – Disclosure Deviations in Selected Key Financials for Portfolio firms This figure shows the deviations in selected key financials between Orbis and the original annual reports. For this check, twenty portfolio firms from the sample of the present analysis have been randomly drawn, and annual reports have been gathered from national databases such as Greens.dk and individual company web sites. The percentage in each column indicates the key figure’s highest observed deviation, and the ones exceeding 5%, positively or negatively, are highlighted. All data from Orbis is reported in EURth.

Deviation in Selected Key Financials Total Assets Turnover EBIT Net Result

Target name MATCO NV

BE

1,45 %

-1,83 %

0,13 %

-1,17 %

PACKING CREATIVE SYSTEMS NV

BE

-1,25 %

2,33 %

0,91 %

-0,60 %

HAMLET PROTEIN AS

DK

-0,48 %

-2,37 %

-1,15 %

0,52 %

VERDIFRESH SL

ES

0,68 %

0,67 %

2,71 %

-1,17 %

JANTON OYJ

FI

-1,05 %

0,18 %

-0,30 %

2,30 %

ANTEL TÉLÉBAT SA

FR

-0,49 %

-1,48 %

1,31 %

-0,30 %

SASA INDUSTRIE SA

FR

-0,90 %

-0,95 %

0,86 %

-1,78 %

PARKEON SAS

FR

0,26 %

2,42 %

-2,86 %

-1,40 %

KALIX SA

FR

-0,88 %

2,67 %

-0,10 %

-0,76 %

LINREAD LTD

GB

1,49 %

1,80 %

-4,55 %

-0,86 %

HSS HIRE SERVICE GROUP LTD

GB

7,07 %

3,81 %

0,29 %

0,55 %

TRACS LTD

GB

-1,81 %

2,12 %

6,92 %

2,51 %

BECK & POLLITZER ENGINEERING LTD

GB

-0,61 %

2,66 %

5,23 %

3,82 %

BRAMBLES FOOD LTD

GB

3,08 %

-0,76 %

1,51 %

-1,44 %

PARASOL LTD

GB

-1,29 %

0,82 %

2,25 %

0,68 %

BIOLCHIM SPA

IT

0,48 %

0,31 %

-1,17 %

2,26 %

MICROTECNICA SRL

IT

-2,06 %

-0,93 %

-4,44 %

2,73 %

LUXY SRL

IT

1,63 %

-0,53 %

-1,16 %

2,07 %

LBC SWEDEN AB

SE

-4,31 %

2,17 %

-5,80 %

3,52 %

SCANDBOOK AB

SE

-2,09 %

-5,32 %

-1,59 %

1,94 %

Source: Zephyr; Orbis; National company information databases such as greens.dk; Company websites

88

Figure A4 – Disclosure Deviations in Selected Key Financials for Control firms This figure shows the deviations in selected key financials between Orbis and the original annual reports. For this check, twenty control firms from the sample of the present analysis have been randomly drawn, and annual reports have been gathered from national databases such as Greens.dk and individual company web sites. The percentage in each column indicates the key figure’s highest observed deviation, and the ones exceeding 5%, positively or negatively, are highlighted. All data from Orbis is reported in EURth,

Deviation in Selected Key Financials Total Assets Turnover EBIT Net Result

Target name HORMIGONES Y MINAS SA

ES

2,20 %

1,33 %

0,54 %

0,50 %

AREA INVESTIGACIÓN SA

ES

-0,83 %

-1,36 %

0,91 %

-0,40 %

APPLEXION SA

FR

-1,31 %

-1,99 %

0,32 %

5,04 %

WESTONE PRODUCTS LTD

GB

2,02 %

2,40 %

2,44 %

1,11 %

CHOCOLATERIE JACALI

BE

1,93 %

-8,30 %

-2,88 %

-2,13 %

ALLERBIO SA

FR

-0,50 %

-0,35 %

2,04 %

-2,00 %

APEC LTD

GB

-5,84 %

2,02 %

4,91 %

-1,14 %

ELPOTEK OY

FI

0,10 %

-2,13 %

-0,95 %

0,86 %

ACCÈS INDUSTRIE SA

FR

0,54 %

0,34 %

0,05 %

0,18 %

DONOVAN MEDICAL EQUIPMENT LTD

IE

-0,66 %

-2,48 %

1,44 %

-2,82 %

ISYDE SRL

IT

0,70 %

-1,63 %

-1,69 %

3,31 %

PAUL & COMPANY

GB

0,54 %

2,28 %

6,79 %

3,13 %

MERCATUS SA

PT

0,27 %

-1,52 %

-1,22 %

0,14 %

PCA MAISONS SAS

FR

0,90 %

-1,97 %

-1,18 %

1,52 %

FABRYKA PORCELANY WALBRZYCH SA

PL

6,74 %

3,56 %

-2,45 %

1,86 %

NEOMPS SA

FR

-0,50 %

-1,83 %

1,12 %

-4,64 %

HEATHWALLACE LTD

GB

-1,49 %

-2,64 %

-6,33 %

0,65 %

TECRESA PROTECCIÓN PASIVA SL

ES

-4,86 %

-3,80 %

-2,17 %

1,30 %

STE GEFCA SARL

FR

1,68 %

0,98 %

2,39 %

-1,12 %

LINK LOGISTICS LTD

GB

7,20 %

4,96 %

1,14 %

2,33 %

Source: Zephyr; Orbis; National company information databases such as greens.dk; Company websites

89

Appendix 7 - Regression Procedure and Testing of OLS Assumptions First, the impact of PE ownership on firm performance (e.g. ROA) is estimated using the general econometric specification derived in section X. Next, the insignificant explanatory variables are deleted from the regression one by one, starting with the most insignificant. This procedure is continued until a reduced specification is reached; only comprising significant variables (at a 10 percent level). Subsequently, tests related to multicollinearity and heteroskedasticity are performed. Finally, it should be noted that this testing procedure is applied for all regressions in this thesis.

Arriving at the Reduced Econometric Specification (The Superiority Hypothesis) Initial econometric specification Dependent Variable: ROA Method: Least Squares Date: 07/29/13 Time: 11:54 Sample (adjusted): 1 260 Included observations: 260 after adjustments Variable

Coefficient

Std. Error

t-Statistic

Prob.

C PE LOG_AGE LOG_SIZE ROA_PRE I0 I1 I2 I3 I4 I5 I6 I7 I8 Y2003 Y2004 Y2005 Y2006 Y2007

0.034259 0.027014 -0.031214 0.002404 -0.612182 -0.086829 -0.030200 -0.033285 -0.031873 -0.043133 -0.050099 0.011716 -0.013442 0.039021 0.085368 0.059305 0.059175 0.030454 0.030327

0.071878 0.013158 0.020487 0.014218 0.084356 0.085570 0.049073 0.044505 0.051714 0.045359 0.049170 0.048996 0.048845 0.051384 0.029292 0.025304 0.023273 0.021958 0.021157

0.476623 2.053136 -1.523602 0.169117 -7.257103 -1.014717 -0.615423 -0.747906 -0.616329 -0.950925 -1.018885 0.239119 -0.275193 0.759390 2.914337 2.343754 2.542606 1.386915 1.433408

0.6341 0.0411 0.1289 0.8658 0.0000 0.3113 0.5389 0.4552 0.5383 0.3426 0.3093 0.8112 0.7834 0.4484 0.0039 0.0199 0.0116 0.1667 0.1530

R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic)

0.248689 0.192575 0.103122 2.562825 231.6203 4.431823 0.000000

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter. Durbin-Watson stat

-0.018259 0.114762 -1.635541 -1.375337 -1.530935 1.989763

90

Reduced econometric specification Dependent Variable: ROA Method: Least Squares Date: 08/04/13 Time: 12:08 Sample (adjusted): 1 260 Included observations: 260 after adjustments Variable

Coefficient

Std. Error

t-Statistic

Prob.

C PE LOG_AGE ROA_PRE I8

0.054810 0.027975 -0.034745 -0.579343 0.057735

0.028439 0.012919 0.019567 0.079704 0.028728

1.927287 2.165402 -1.775693 -7.268709 2.009742

0.0551 0.0313 0.0770 0.0000 0.0455

R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic)

0.189464 0.176750 0.104128 2.764851 221.7563 14.90166 0.000000

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter. Durbin-Watson stat

-0.018259 0.114762 -1.667356 -1.598882 -1.639829 1.976123

Testing of OLS Assumptions In the following section, the relevant OLS assumptions are tested for the reduced econometric specification above (with ROA as the dependent variable). The approach however, applied is used for all relevant models in this paper. First, an assessment of multicollinearity is conducted (see below). This is done by assessing the correlation between the explanatory variables in a correlation matrix as seed below. A general rule states that one pair-wise correlation above 0,8 (in numerical terms) is problematic and so is several above 0,5 (Bergström et al., 2007). The correlations in this case are generally low however, which indicates that there is no multicollinearity amongst them. The most notable correlation amounts to 0,17 and is found between the variables ROA_PRE and LOG_AGE. Hence, there is no need to exclude any variables from the model due to multicollinearity. Next, a test for heteroskedastisity is performed, which occurs when the error term has non-constant variance. If heteroskedastisity is present, the OLS estimator is still unbiased, but becomes inefficient; i.e. it is no longer considered BLUE. The estimated variances of the OLS estimates are then biased, making the t-statistics unreliable. White’s test (see below) investigates whether the residual variance of the variables in the regression model are constant. Hence, the null hypothesis maintains that the errors are constant (or homoskedastic). From the relatively high P-value (0,66) reported below, the null hypothesis cannot be rejected, and therefore homoskedastisity is assumed.

91

Multicollinearity assessment

LOG_AGE ROA_PRE I8 PE

LOG_AGE 1.000000 -0.171521 -0.006237 -0.023557

ROA_PRE -0.171521 1.000000 0.089802 0.007852

I8 -0.006237 0.089802 1.000000 3.31E-18

PE -0.023557 0.007852 3.31E-18 1.000000

Heteroskedasticity test Residual graph .4 .2 .0 .4

-.2

.2

-.4

.0 -.2 -.4 25

50

75

100 Residual

125

150 Actual

175

200

225

250

Fitted

White’s test for heteroskedasticity F-statistic Obs*R-squared Scaled explained SS

0.789047 9.598940 17.68342

Prob. F(12,247) Prob. Chi-Square(12) Prob. Chi-Square(12)

0.6616 0.6511 0.1256

Test Equation: Dependent Variable: RESID^2 Method: Least Squares Date: 07/29/13 Time: 11:27 Sample: 1 260 Included observations: 260 Collinear test regressors dropped from specification Variable

Coefficient

Std. Error

t-Statistic

Prob.

C PE PE*LOG_AGE PE*ROA_PRE

0.010396 -0.001719 0.002732 0.030017

0.016993 0.011241 0.007978 0.032261

0.611780 -0.152940 0.342365 0.930465

0.5412 0.8786 0.7324 0.3530 92

PE*I8 LOG_AGE LOG_AGE^2 LOG_AGE*ROA_PR E LOG_AGE*I8 ROA_PRE ROA_PRE^2 ROA_PRE*I8 I8

0.001735 0.001584 -0.001982

0.012153 0.023564 0.008438

0.142756 0.067211 -0.234883

0.8866 0.9465 0.8145

-0.078391 -0.004525 0.105829 -0.079702 -0.028437 0.004684

0.054703 0.029563 0.084216 0.130887 0.050095 0.039457

-1.433033 -0.153055 1.256630 -0.608938 -0.567654 0.118701

0.1531 0.8785 0.2101 0.5431 0.5708 0.9056

R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic)

0.036919 -0.009870 0.020955 0.108461 642.7424 0.789047 0.661632

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter. Durbin-Watson stat

0.010634 0.020852 -4.844172 -4.666138 -4.772600 2.060047

Outputs From the Three Remaining Tests of The Superiority Hypothesis The same approach as outlined above is applied to reach viable econometric specifications and test OLS assumptions for the remaining three dependent variables (OROA, EM and NPM). The reduced models and estimates follow here: Dependent Variable: OROA Method: Least Squares Date: 07/29/13 Time: 14:12 Sample (adjusted): 1 260 Included observations: 260 after adjustments White heteroskedasticity-consistent standard errors & covariance Variable

Coefficient

Std. Error

t-Statistic

Prob.

PE ROA_PRE I6 I8 Y2007

0.022626 -0.462406 0.085360 0.101822 -0.025373

0.011967 0.086456 0.038372 0.030657 0.014184

1.890667 -5.348479 2.224513 3.321341 -1.788875

0.0598 0.0000 0.0270 0.0010 0.0748

R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat

0.136457 0.122911 0.118901 3.605053 187.2609 2.041034

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter.

-0.018349 0.126959 -1.402007 -1.333532 -1.374479

93

Dependent Variable: EM Sample (adjusted): 1 260 Included observations: 214 after adjustments Variable

Coefficient

Std. Error

t-Statistic

Prob.

C PE ROA_PRE I4 I8 Y2004

-0.020611 0.046849 -0.268915 -0.032733 0.073972 0.042534

0.011290 0.012667 0.081088 0.016523 0.026119 0.019029

-1.825688 3.698548 -3.316345 -1.981067 2.832128 2.235277

0.0693 0.0003 0.0011 0.0489 0.0051 0.0265

R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic)

0.164515 0.144431 0.092647 1.785359 208.4873 8.191442 0.000000

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter. Durbin-Watson stat

-0.012745 0.100162 -1.892405 -1.798031 -1.854269 2.104766

Dependent Variable: NPM Sample (adjusted): 1 260 Included observations: 230 after adjustments White heteroskedasticity-consistent standard errors & covariance Variable

Coefficient

Std. Error

t-Statistic

Prob.

PE ROA_PRE I8 Y2004

0.022744 -0.395324 0.082823 0.036110

0.009430 0.067785 0.032738 0.010036

2.411725 -5.832025 2.529838 3.597980

0.0167 0.0000 0.0121 0.0004

R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat

0.165681 0.154606 0.088704 1.778265 232.8249 1.943054

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter.

-0.009073 0.096475 -1.989781 -1.929989 -1.965662

Summary of OLS Assumption Tests Subject of test

Applied test

Multicollinearity

Correlogram1

Heteroskedasticity Normality

Dependent variable of regression ROA OROA EM NPM No

No

No

No

White's test

0,66

0,01

0,99

0,00

Jarque-Bera

0,08

0,11

0,14

0,18

Note 1) Multicollinearity is assessed to be a problem when pair-wise correlation between explanatory variables is high. In this paper, high correlation implies one correlation value above 0,8 or several correlation values exceeding 0,5 Source: Bergström 2007). “Yes” indicates that multicollinearity is present, otherwise “no”.

94

Appendix 8 - Magnitude of the Impact From Operational Engineering Figure A5 reports a rough estimate of the magnitude of the estimated impact of the variables used to test The Operational Engineering Hypotheses (H3-H5). It is important to stress that these numbers should be interpreted with the usual reservations to ceteris paribus estimates (for instance, the estimated beta coefficients only apply for marginal changes in the given variable). Hence, they have only been calculated to provide a rough indication of whether the individual variable plays a significant role in explaining firm performance. They are not very applicable when the exact magnitude of the impact has to be estimated. Thus, we cannot conclude that COGS for instance explains 61% of the estimated impact of PE fund ownership on ROA, OOEM 59% and WCR 29%. Rather we can conclude, that they all explain a significant part, and that the two former measures are likely to explain more than WCR. Figure A5 - Estimated magnitude of the impact from operational engineering The figure shows the beta coefficients of all independent variables investigated in H3-H5 which are estimated to have a significant impact on the firm performance measures (ROA, OROA, EM and NPM). It also reports the sample’s average post/pre difference for each variable. This number is multiplied with the variable’s beta coefficient to estimate the average impact on each performance measure. Below, it is calculated how large a percentage of the PE variable’s beta coefficient (which represents the impact of PE fund ownership on portfolio firm performance) is explained by the given independent variable on average – ceteris paribus.

Independent variables COGSM Beta coefficients of COGSM*PE Estimated impact on perfomance % of Beta coefficient of PE

Post-pre differences -0,045

OOEM Beta coefficients of OOEM*PE Estimated impact on perfomance % of Beta coefficient of PE

-0,033

WCR Beta coefficients of WCR*PE Estimated impact on perfomance % of Beta coefficient of PE

-0,037

Beta coefficients of PE

ROA

Dependent variables OROA EM

NPM

-0,380 0,017 61%

-0,501 0,022 99%

-0,567 0,025 54%

-0,586 0,026 115%

-0,494 0,016 59%

-0,431 0,014 64%

-0,422 0,014 30%

-0,387 0,013 57%

-0,221 0,008 29%

insig.

-0,212 0,008 16%

insig.

0,028

0,023

0,047

0,023

95

Appendix 9 - Content of Attached CD-ROM 1. Eviews Outputs for All Tests Performed 2. Data – Control Firms 3. Data – Portfolio Firms 4. Final Sample (Eviews data) 5. Final Sample (Excel data) 6. Matching as Robustness Check 7. Matching of the Original Sample

96