on Post-‐Buyout Performance in
Private Equity Takeovers
Abstract
This thesis uses a dataset of private equity buyouts completed in the United Kingdom from 2006 to 2010 to research the effect of pre-‐buyout inefficiency on post-‐buyout performance and efficiency gains. Contrary to previous research this thesis focuses on efficiency gains realized by private equity. To measure efficiency the Altman Z-‐score is used, a novel use for this score introduced by this thesis. This thesis finds a significant positive relation between pre-‐buyout inefficiency and the change in ROA and the change in efficiency over a five-‐year period. Firms that have greater pre-‐buyout inefficiency improve their profitability and efficiency score more over the sample period. The same is true for specific inefficiencies such as credit-‐days and collection-‐days. The results are robust to winsorization and to the usage of individual Altman variables rather than the constrained Altman Z-‐score. Tests of the robustness of the contribution of PE are inconclusive.
Name Sebastiaan Tito
Student Number 10002983
Program Economics and Business
Track Finance and Organization
Supervisor Timotej Homar
Date 28-‐06-‐2015
Statement of Originality
This document is written by Student Sebastiaan Tito who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
2. Literature ... 6
2.1 Private Equity and its history ... 6
2.2 Leveraged Buyouts ... 7
2.3 Private Equity Research ... 8
3. Research Methodology ... 10
3.1 Efficiency Scoring ... 10
3.2 Performance Change ... 11
3.3 Efficiency Change ... 12
3.4 Credit and Collection Days ... 13
4. Data ... 14
4.1 Data on Private Equity deals ... 14
4.2 Data on Portfolio Companies ... 15
4.3 Sector Data ... 16
4.4 Market Data ... 17
5. Results ... 18
5.1 Performance Change Analysis ... 18
5.2 Efficiency Change Analysis ... 19
5.3 Credit and Collection Results ... 22
6. Robustness ... 23
6.1 Individual Altman Variables ... 23
6.2 Private Equity Robustness Test ... 25
6.3 Non-‐Winsorized Results ... 26
6.4 ROE Robustness Check ... 29
7. Conclusion ... 30
References ... 32
1. Introduction
In the 1980s the Private Equity (PE) sponsored buyout industry grows rapidly. The total committed capital to PE funds increases from $600 million in the late 1970s to $100 billion in 1994 (Fenn, Liang, & Prowse, 1995). The increased activity in this market attracts the attention of researchers.
Most research into PE focuses on the level of the PE firm. This includes research into the risks of investment in PE, research into the performance of PE adjusted for risks and outperformance of the market, and research into the wealth effects of leveraged buyouts (LBOs). However, the research into the effect of PE ownership on the portfolio company is limited in quantity and in scope. The research that exists is mostly from the 1980s, and focuses on the performance of a portfolio firm post-‐buyout. The results from these papers are ambiguous. Kaplan (1989) finds that PE ownership has a positive effect on performance, as does Jensen (1989). However, Scellato and Ughetto (2013) find that PE ownership has a negative effect, while Acharya, Hahn, and Kehoe (2009) find it has a positive effect using the same techniques.
This thesis intends to slightly broaden the scope of the existing research into pre-‐ and post-‐buyout performance. It will focus on the efficiency of the portfolio company. Central to this research is the PE buyout business model of acquiring a business and making operational changes to trim the fat and create a more efficient company (Mohan, 1990). To a PE investor, pre-‐existing operational and organizational inefficiencies are areas to potentially improve the company. This thesis will research if companies that have more pre-‐buyout inefficiencies manage to make greater improvements during their time as a PE portfolio company. If this is the case then PE does indeed create more firm value in companies that are in worse shape pre-‐buyout. This concerns the function of PE for individual companies as well as the benefit of PE to the economy as a whole.
The research question of this thesis is: what is the effect of pre-‐buyout inefficiencies on post-‐buyout performance? Based on the PE business model the pre-‐buyout inefficiencies can be thought of as potential performance gains. Greater inefficiencies mean more potential for improvement when a PE firm takes over. The circumstances of a buyout with the increased leverage and the
realignment of the incentives can be expected to facilitate this. The hypothesis is that more inefficiency pre-‐buyout should have a positive effect on the change in performance and efficiency over a period of 4 years. This period is at the end of the planned PE holding period which presumably means that the PE firm will have realized most of the efficiency gains they thought possible.
Empirical research will be done to test this hypothesis. The first step is measuring pre-‐buyout efficiency and performance; the second step is measure performance and efficiency 4 years after the buyout. A multiple regression model will then be estimated to test the effect of the pre-‐buyout efficiency on both the change in performance and the change in efficiency over this period. The measure for the pre-‐buyout efficiency is expected to have a negative sign, indicating a positive effect of lower efficiency and a negative effect of greater efficiency pre-‐buyout.
Chapter 2 will go more in-‐depth into the history of PE and the existing literature. The goal of this chapter is to outline the different views on PE and to substantiate the research decisions. Chapter 3 will outline the methodology. It states how the data will be used to generate different variables and what the actual research models are. Chapter 4 will describe and test the dataset used for this research. The goal of this chapter is to promote transparency in this research and enable the reader to judge the validity of the data. Chapter 5 presents the results from this research. The outcomes of the different models and their meanings will be explained. Chapter 6 is the robustness section. This section tests several aspects of the thesis for robustness. The goal of this chapter is to give the reader insight into the effect of the research setup on the results. Chapter 7 is the conclusion, it answers the research question and discusses the limitations of this research.
2. Literature
2.1 Private Equity and its history
A PE fund is a non-‐transparent investment vehicle that invests the funds of its investors in companies (Fraser-‐Sampson, 2007). The PE fund belongs to a PE firm that usually manages several of these funds and the companies these funds hold in their investment portfolios (Cumming, 2010). The PE firm intends an exit after 5 years and aims for an annualized return in excess of 20% (Pearl & Rosenbaum, 2013). In practice The PE firm manages these companies for about 4-‐7 years (Fenn et al., 1995). In this period the PE firm can make operational changes as well as investments in the portfolio company to boost its resale value. In practice PE firms are usually characterized as Venture Capital (VC) or as buyout funds (Fraser-‐Sampson, 2007). Venture Capital funds focus on early investments in startup companies that have no access to traditional capital markets. In return for a share of the control they provide the startup with access to capital and general business knowledge. Their goal is usually to sell most of their investment in an initial public offering (IPO) (Damodaran, 2010). Fraser-‐ Sampson (2007) states that VC is a powerful tool for economic growth. He further states that by the end of 2000, VC is directly responsible for the creation of 8 million jobs, roughly one job for every $36 thousand invested (2007).
Buyout funds target established companies rather than startups (Fraser-‐ Sampson, 2007). The deals often contain a large portion of debt financing, and buyout firms usually take a more active approach in managing the portfolio company (2007). Especially in Europe they are seen as less favorably by governments than VC (2007). This is because their operational restructures reduce jobs and their financial restructures reduce tax yield (2007). IPO exits are less likely for buyout funds than for VC, especially in recent years so called secondary buyouts (selling to another investment firm) are becoming more prevalent (Jong, Roosenboom, Verbeek, & Verwijmeren, 2007).
Private Equity firms can be found as far back as the 1940s, although these firms focus mainly on Venture Capital. There are not many firms involved in buyouts at this time (Liles, 1977). It is not until the development of the limited
partnership in the 1970s that both buyout and VC firms become more popular (Bygrave & Timmons, 1992). However, the regulatory changes and tax benefits of the late 1970s and the early 1980s are what really start the boom in PE (Pratt, 1982). It is during this boom that leveraged buyouts become very popular. In 1987 a record size buyout fund by Kohlberg, Kravis, and Roberts (KKR) raises $5.6 billion, almost twice the total commitments to VC in that year (Fenn et al., 1995). In 1988 KKR executes one of the largest LBOs in history, and the largest ever at that time, when it buys RJR-‐Nabisco for $31.4 billion (Copeland, Koller, & Murrin, 2000). The increasing size of the industry and the total committed funds draw more attention to PE. Especially the LBO as an instrument comes under scrutiny (Axelson, Strömberg, & Weisbach, 2008).
2.2 Leveraged Buyouts
The LBO is the deal type most commonly associated with a PE buyout, and refers to a special form of deal financing. Although Fraser-‐Sampson (2007) emphasizes that really all buyouts are ‘leveraged’, because they all involve some form of debt financing, there is a marked difference between a regular deal and an LBO. Central to an LBO is the use of a target’s own assets to secure acquisition debt (Fraser-‐Sampson, 2007). On average about 75% of the acquisition is financed using debt (Jong et al., 2007).
There are different views on the benefits and/or drawbacks of LBO deals. After the M&A wave of the 1960s and 1970s there are a lot of large inefficient conglomerates (Jong et al., 2007). In this landscape LBOs provide a key economic benefit, because they enable buyout companies to purchase and restructure these conglomerates (Jong et al., 2007). However, in the period 1986-‐1990, 41% of the buyout value creation is the result of the increased debt, rather than operational improvement (2007). These leveraged recapitalizations are of significant importance in the operations of PE firms and have come under political scrutiny (Fraser-‐Sampson, 2007).
However, there are positive effects of PE ownership and the associated high amount of debt. Jensen (1989) argues that active investors are a return to form for companies since active investors are able to maximize firm value. This is consistent with the empirical results of research by Donaldson (1984).
Donaldson finds that top managers are more concerned with maximizing the corporate purchasing power of management, rather than maximization of firm value (1984). LBOs facilitate the active PE investors in realigning these companies to make them perform more effectively with the same resources (Jensen, 1989). In a different paper Jensen (1986) discusses the agency costs of free cash flow in an organization. His theory states that management has incentives to grow the business beyond its optimal size, because this increases their remuneration (1986). This is consistent with the empirical findings of Murphy (1985), who found a strong positive relation between firm growth and increase in managerial remuneration. The solution Jensen (1986) proposes is more debt, because the increased debt service payments force management to use funds more efficiently and stop them from undertaking value reducing projects. This makes management more aligned with the broader stakeholders.
2.3 Private Equity Research
Research into the performance of PE firms shows that unlike mutual funds, PE can be a better investment than a diversified portfolio. Kaplan and Schoar (2005) do empirical research into the performance of PE firms. They find that on average the risk-‐adjusted fund returns net of fees equal the S&P500 index, this is consistent with findings by Phalippou and Gottschalg (2007) and Phalippou (2009). But Kaplan and Schoar (2005) do find significant evidence of heterogeneity across funds. They find significant evidence of persistent outperformance for better performing funds, indicating that well-‐managed funds do provide an alpha for investors (Kaplan & Schoar, 2005). However, their results show that PE funds suffer from diseconomies of scale, and that as an outperforming firm becomes more popular and its committed capital to funds grows, it delivers poorer results (2005). The diseconomies of scale are also found in empirical research by Lopez-‐de-‐Silanes, Phalippou, and Gottschalg (2013). However, contrary to these findings, a recent empirical research by Harris, Jenkinson, and Kaplan (2014) finds that buyout funds outperform the market by an average of 20%-‐27% when adjusted for risk and net of fees.
As mentioned in the introduction the research into the effect of PE ownership on portfolio companies is ambiguous. Kaplan (1989) does empirical
research into the performance of portfolio companies of buyouts completed between 1980 and 1986. He finds that these companies have increased operating income, decreased capital expenditures, and increases in net cash flow 3 years after the buyout when he controls for economy-‐wide and industry effects. He also controls for divestitures and layoffs. His results show that buyouts experience an improvement in performance that is due to operational improvements rather than layoffs or managerial exploitation of shareholders.
Scellato and Ughetto (2013) use propensity score matched peers to better control for the effect of PE buyout on a portfolio company. They find that a buyout by a generalist fund negatively impacts operating profitability, while a buyout by a turnaround specialist positively impacts operating profitability. This is contrary to empirical findings by Acharya et al. (2009). They use the same propensity score matched peers but find that PE ownership positively affects profitability. Bacon, Wright, Meuleman, and Scholes (2012) do empirical research into the effect of a PE buyout on human resources. They find that a PE buyout increases average skill level and job satisfaction.
3. Research Methodology
3.1 Efficiency ScoringThis thesis intends to research the effect of portfolio company pre-‐buyout inefficiencies on post-‐buyout performance. The first step is therefore to determine the level of pre-‐buyout efficiency. An extensive due diligence is required for an accurate evaluation of existing inefficiencies within a company. However, in the context of this research that is not possible. The estimation of this measure is therefore limited to the financial and operational data reported by the portfolio companies and the ratios computed from this data.
It is important that this score is a broad measure of both the operational efficiency as well as the efficiency of the structure of a firm. This requires a measure that takes multiple firm characteristics into account. The measure needs to score liquidity, profitability, leverage, solvency, and activity. To get a reliable measure with adequate weighing of these different categories this research will make use of the Altman Z-‐score. This score is designed specifically as a measure for these characteristics (Altman, 1968).
Because this research focuses on companies owned by a PE firm, the measure of inefficiency will be the Altman Z-‐score for private companies. This so called Z’-‐score focuses on book value where the original Z-‐score focuses on market value (Altman, 2000). For simplicity there will be no focus on the original Altman score. All further mentions of the ‘Z-‐score’ will refer to the Z’-‐score for private companies.
The Altman Z-‐score has the following form:
𝑍 = 0.717 𝑋! + 0.847 𝑋! + 3.107 𝑋! + 0.420 𝑋! + 0.998 𝑋! . Where: 𝑋! = 𝑊𝑜𝑟𝑘𝑖𝑛𝑔 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠, 𝑋! = 𝑅𝑒𝑡𝑎𝑖𝑛𝑒𝑑 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠, 𝑋! = 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝐵𝑒𝑓𝑜𝑟𝑒 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑎𝑛𝑑 𝑇𝑎𝑥𝑒𝑠 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠, 𝑋! = 𝐵𝑜𝑜𝑘 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝐸𝑞𝑢𝑖𝑡𝑦 𝑇𝑜𝑡𝑎𝑙 𝐿𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠, 𝑋! = 𝑂𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑅𝑒𝑣𝑒𝑛𝑢𝑒 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠.
Originally designed by Edward I. Altman in 1968, his revisited paper from 2000 contains this version designed for the analysis of private companies. His goal was to design a score that mapped the default probability of a company using a weighed scoring of broad financial measures from several key firm characteristics. This score allows the comparing of different firms based on their overall health. For the purposes of this research the Altman Z-‐score is a great way to benchmark efficiency of portfolio companies in a way that allows for comparison across structures and business models. The inclusion of control variables for industry allows for an even broader comparison. A higher Altman Z-‐score thus reflects greater efficiency and a lower score reflects the existence of more inefficiencies, and thus more potential unrealized efficiency and performance gains. The use of Altman Z-‐scores as a measure of efficiency is a novel use of the score this research introduces and it goes a long way to overcome the obstacle of not being able to do extensive due diligence.
In this thesis the Z-‐score is central to the identification of inefficiency. However, the main regressions are also estimated using the individual Altman variables. This is to test the robustness of using the Z-‐score without the weight-‐ constraints suggested by Altman. For these regressions please refer to Section 6.1.
3.2 Performance Change
A performance measure needs to be specified to measure the effect of inefficiencies. Based on existing literature on the subject of performance measures, this research will use Pre-‐Tax Return on Assets. The advantage of using PTROA is that is it not influenced by possible tax benefits from accounting changes or leverage. For simplicity all further mentions of ROA will concern Pre-‐ Tax Return on Assets. The advantage of using ROA as a measure is that it measures the return to the broader stakeholders, rather than just equity holders. It is also less influenced by changes in leverage than ROE, which is an important advantage since leverage varies widely across PE portfolio companies and across the holding period (Acharya et al., 2009). Dess and Robinson (1984) also describe ROA as both a measure of economic performance of a firm as well as efficiency of a firm in regard to the profitable use of its total asset base.
Because of the existence of negative ROAs, the change in performance is measured as the simple difference in ROA stated as:
𝑅𝑂𝐴 𝐶ℎ𝑎𝑛𝑔𝑒 = 𝑅𝑂𝐴!!− 𝑅𝑂𝐴!"#
To research the effect of the pre-‐buyout efficiency on post-‐buyout performance this research uses multiple regression models that add control variables to control for outside influences. The full regression model is:
𝑟𝑜𝑎_𝑐ℎ𝑎𝑛𝑔𝑒! = 𝛼 + 𝛽! 𝑧_𝑝𝑟𝑒! + 𝛽! 𝑟𝑜𝑎_𝑝𝑔𝑚𝑒𝑑_𝑐ℎ𝑎𝑛𝑔𝑒! + 𝛽! 𝑓𝑡𝑠𝑒_𝑟𝑒𝑡 +
𝛽! 𝑠𝑖𝑧𝑒! + 𝜀!
Where: 𝑟𝑜𝑎_𝑐ℎ𝑎𝑛𝑔𝑒! Change in ROA
𝑧_𝑝𝑟𝑒! Pre-‐Buyout Altman Z-‐score
𝑟𝑜𝑎_𝑝𝑔𝑚𝑒𝑑_𝑐ℎ𝑎𝑛𝑔𝑒! Change in Peer Group Median ROA 𝑓𝑡𝑠𝑒_𝑟𝑒𝑡 FTSE100 return between pre and t4
𝑠𝑖𝑧𝑒! Log(Total Assets)
3.3 Efficiency Change
The efficiency is scored pre-‐buyout and at t4 using the Altman Z-‐score
formula stated above. This research will also focus on the effect of this pre-‐ buyout efficiency on change in efficiency at t4. Because of the existence of
negative Z-‐scores this efficiency change is the simple difference in Altman Z-‐ score stated as:
𝑍 𝐶ℎ𝑎𝑛𝑔𝑒 = 𝑍!!− 𝑍!"#
To research the effect of the pre-‐buyout efficiency on post-‐buyout efficiency changes this research uses multiple regression models that add control variables to control for outside influences. The full regression model is:
𝑧_𝑐ℎ𝑎𝑛𝑔𝑒! = 𝛼 + 𝛽!(𝑧_𝑝𝑟𝑒!) + 𝛽! 𝑟𝑜𝑎_𝑝𝑔𝑚𝑒𝑑_𝑐ℎ𝑎𝑛𝑔𝑒!
+ 𝛽! 𝑧_𝑝𝑔𝑚𝑒𝑑_𝑐ℎ𝑎𝑛𝑔𝑒! + 𝛽! 𝑓𝑡𝑠𝑒_𝑟𝑒𝑡 + 𝛽! 𝑠𝑖𝑧𝑒! + 𝜀! Where: 𝑟𝑜𝑎_𝑐ℎ𝑎𝑛𝑔𝑒! Change in ROA
𝑧_𝑝𝑟𝑒! Pre-‐Buyout Altman Z-‐score
𝑟𝑜𝑎_𝑝𝑔𝑚𝑒𝑑_𝑐ℎ𝑎𝑛𝑔𝑒! Change in Peer Group Median ROA 𝑧_𝑝𝑔𝑚𝑒𝑑_𝑐ℎ𝑎𝑛𝑔𝑒! Change in Peer Group Median Z-‐score 𝑓𝑡𝑠𝑒_𝑟𝑒𝑡 FTSE100 return between pre and t4
3.4 Credit and Collection Days
An important part of the efficient functioning of a company is the time it has to pay back creditors and the time it takes to collect on its bills. This section will research the changes in these credit and collection periods specifically. To enable comparison across industries the periods are benchmarked against the median periods in their peer group. The benchmark is calculated for the year pre-‐buyout and for t4. The dependent variable in the regressions will be
the change of this benchmark. These measures can be defined as:
𝐶𝑟𝑒𝑑𝑖𝑡𝑑𝑎𝑦𝑠 𝐵𝑒𝑛𝑐ℎ𝑚𝑎𝑟𝑘 = 𝐶𝑜𝑚𝑝𝑎𝑛𝑦 𝐶𝑟𝑒𝑑𝑖𝑡𝑑𝑎𝑦𝑠 𝑝𝑟𝑒 𝑃𝑒𝑒𝑟 𝐺𝑟𝑜𝑢𝑝 𝑀𝑒𝑑𝑖𝑎𝑛 𝐶𝑟𝑒𝑑𝑖𝑡𝑑𝑎𝑦𝑠 𝑝𝑟𝑒 𝐶𝑜𝑙𝑙𝑒𝑐𝑡𝑑𝑎𝑦𝑠 𝐵𝑒𝑛𝑐ℎ𝑚𝑎𝑟𝑘 = 𝐶𝑜𝑚𝑝𝑎𝑛𝑦 𝐶𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑑𝑎𝑦𝑠 𝑝𝑟𝑒 𝑃𝑒𝑒𝑟 𝐺𝑟𝑜𝑢𝑝 𝑀𝑒𝑑𝑖𝑎𝑛 𝐶𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑑𝑎𝑦𝑠 𝑝𝑟𝑒 𝐶𝑟𝑒𝑑𝑖𝑡𝑑𝑎𝑦𝑠 𝐵𝑒𝑛𝑐ℎ𝑚𝑎𝑟𝑘 𝐶ℎ𝑎𝑛𝑔𝑒 = 𝐶𝑜𝑚𝑝𝑎𝑛𝑦 𝐶𝑟𝑒𝑑𝑖𝑡𝑑𝑎𝑦𝑠 𝑡4 𝑃𝑒𝑒𝑟 𝐺𝑟𝑜𝑢𝑝 𝑀𝑒𝑑𝑖𝑎𝑛 𝐶𝑟𝑒𝑑𝑖𝑡𝑑𝑎𝑦𝑠 𝑡4− 𝐶𝑜𝑚𝑝𝑎𝑛𝑦 𝐶𝑟𝑒𝑑𝑖𝑡𝑑𝑎𝑦𝑠 𝑝𝑟𝑒 𝑃𝑒𝑒𝑟 𝐺𝑟𝑜𝑢𝑝 𝑀𝑒𝑑𝑖𝑎𝑛 𝐶𝑟𝑒𝑑𝑖𝑡𝑑𝑎𝑦𝑠 𝑝𝑟𝑒 𝐶𝑜𝑙𝑙𝑒𝑐𝑡𝑑𝑎𝑦𝑠 𝐵𝑒𝑛𝑐ℎ𝑚𝑎𝑟𝑘 𝐶ℎ𝑎𝑛𝑔𝑒 = 𝐶𝑜𝑚𝑝𝑎𝑛𝑦 𝐶𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑑𝑎𝑦𝑠 𝑡4 𝑃𝑒𝑒𝑟 𝐺𝑟𝑜𝑢𝑝 𝑀𝑒𝑑𝑖𝑎𝑛 𝐶𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑑𝑎𝑦𝑠 𝑡4− 𝐶𝑜𝑚𝑝𝑎𝑛𝑦 𝐶𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑑𝑎𝑦𝑠 𝑝𝑟𝑒 𝑃𝑒𝑒𝑟 𝐺𝑟𝑜𝑢𝑝 𝑀𝑒𝑑𝑖𝑎𝑛 𝐶𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑑𝑎𝑦𝑠 𝑝𝑟𝑒
The benchmarks serve as measures of how efficient a company manages its credit and its collections compared to its peer group. A credit-‐days benchmark >1 is indicative of favorable credit terms for the company. A collection-‐days benchmark >1 indicates a company takes longer than the peer group median to collect from debtors.
The regressions test both the effect of the pre-‐buyout benchmark as well as the effect of the pre-‐buyout Z-‐score on the change in the benchmark. Here, the pre-‐buyout Z-‐score is again used to measure the general level of efficiency. However, the pre-‐buyout benchmark is also included to measure the pre-‐buyout efficiency specific to the credit and collection periods. The full regression model for both benchmark changes is:
4. Data
4.1 Data on Private Equity deals
This analysis focuses on Private Equity buyouts of UK companies, executed between 2006 and 2010. The focus is on UK companies because the availability of financial and operational data on private companies is greater for UK companies than for US companies. The UK is also a large market with well-‐ developed PE, making it unnecessary to include other countries to obtain an adequate sample size (Engel & Stiebale, 2014). Deal data is collected from the Zephyr M&A database, with Bureau van Dijk identification numbers (BvDIDs) for the target companies. To accurately examine the effect of PE ownership on portfolio company management and operations, the research is limited to full buyouts. This dataset has 677 buyouts. The sample has been reduced to 133 buyouts, because this research requires financial and operational data not available for all companies in the full dataset. Financial and operational data for the portfolio companies is collected from the Orbis database using the companies’ BvDIDs. For an overview of the number of buyouts per year please refer to Table 1.
Table 1 Deal Distribution by Year
Deals Population Sample 2006 150 29 2007 195 51 2008 141 31 2009 60 16 2010 131 6 Total 677 133
Several firm characteristics from the sample are subjected to T-‐testing against the known values in the population to test the representativeness of the sample. The means of the sample and the population: total assets (t=0.7472), number of employees (t=0.3734), and operating revenue (t=0.8522) are tested. The null hypotheses are not rejected at an alpha of 10%, 5%, or 1%. There is
therefore no evidence of a difference in the means introduced by using this sample. For detailed results of these tests please refer to Appendix 1.
4.2 Data on Portfolio Companies
Comprehensive annual data is gathered on each of the 133 portfolio companies through the Orbis database. This data includes complete balance sheets, key ratios, and operational data. This research will focus on the company pre-‐buyout (that is, one year prior to the buyout year) and 4 years after the buyout. This annual data is reworked for each company to fit the uniform labels of ‘pre’ and ‘t4’. This is done to enable comparisons between companies with
different buyout years.
The reliability of this research is highly dependent on the quality of the gathered data. To control for the influence of severe outliers or possible faulty reporting, the data is winsorized at the 1st and 99th percentile. Section 6.3
contains the regression results from non-‐winsorized variables. This will provide this research with transparency and allows the reader to make inferences on the effect of the winsorization on the outcomes and conclusions. For descriptive statistics on the most important portfolio company variables please refer to Table 2. For full descriptive statistics please refer to Appendix 2.
Table 2 Descriptive Statistics Portfolio Companies
N Mean Median Std. Dev. Min Max
Altman Pre 132 2.59 2.35 1.91 -‐0.57 8.54 Altman t4 131 2.49 2.53 1.75 -‐1.59 7.12 Altman Change1 131 -‐0.09 -‐0.12 1.53 -‐5.41 5.65 ROA Pre 133 10.13 8.15 19.02 -‐72.65 58.55 ROA t4 133 7.50 6.77 17.72 -‐59.34 69.08 ROA Change1 133 -‐2.63 -‐1.81 22.07 -‐117.89 59.35 OR Pre2 133 180.75 48.31 427.89 0.46 3077.41 OR t42 133 188.21 71.78 378.17 1.96 2791.56 OR Change (%) 133 99.24% 22.00% 328.81% -‐74.36% 2734.94%
Total Assets Pre2 133 279.15 32.34 765.36 0.55 4416.10
Credit-‐Days Pre 130 32.75 28 39.25 0 298
Credit-‐Days t4 131 31.13 27 27.45 0 176
Collection-‐Days Pre 130 49.12 52 31.00 0 133
Collection-‐Days t4 132 45.71 43 31.45 0 139
1: Simple difference t4 -‐ pre
2: In $ Millions
4.3 Sector Data
The portfolio companies are widely spread across sectors. The sample of 133 companies has 76 unique four-‐digit NACE sector identifications and one company missing a NACE identifier. The highest concentration of NACE identifications is 17 companies with NACE 7010, the indication for head office of a holding company. This constitutes 12.78% of the sample. Other significant concentrations are 7 companies (5.26%) in ‘other business support’ (NACE 8299), 7 companies (5.26%) in ‘other financial brokerage’ (NACE 6499), and 5 companies (3.76%) in ‘other information technology’ (NACE 6209). For information on sectors with a lower frequency as well as a complete breakdown of the sample by sector please refer to Appendix 3.
This wide spread of sectors across the sample makes an internal comparison with this sample size impossible. Therefore, to control for sector-‐ wide cyclical performance changes as well as booms and busts there is a need for specialized peer groups. These peer groups are based on four-‐digit NACE sector and company size. Using the Orbis standard peer groups that consist of comparable international companies operating within the same NACE sector, the peer group is narrowed down to the 10 most comparable companies by total assets in the buyout year. For some companies it was impossible to construct a representative peer group and some peer group variables were unavailable. This means that including peer group data will limit the sample by approximately 10-‐ 13 observations, depending on the combination of variables used.
Comprehensive annual financial and operating data is gathered on all peer group companies for several periods. To further control for outliers, the median value of the financial and operating variables of the peer group is used. These median values are then matched with their respective companies and together form the ‘peer group median’ variables. These variables are again reworked into the format of ‘pre’ and ‘t4’. Peer group variables are winsorized to
the same fraction as the portfolio company data. For descriptive statistics on the most important peer group variables please refer to Appendix 2.
4.4 Market Data
Based on the availability of data, the coverage period of this research is between 2005 and 2014. Analyzing one year pre-‐buyout and four years after the buyout means the sample period for this thesis is deals between 2005 and 2010. To control for the obvious effects of the global financial crisis as well as several mini-‐booms and mini-‐busts surrounding this period there is need for market data. To proxy for these events the FTSE100 index will be used. This index consists of the 100 largest publicly traded companies in the UK, thus it should provide an accurate approximation of the broad UK market. Yearly returns are calculated on a simple basis as the percent difference in the index compared to one-‐year prior. These returns are again reworked to fit the buyout timeline for each company depending on the buyout year. The variable FTSE return is therefore the simple return of the FTSE between ‘pre’ and the end of ‘t4’. The
FTSE index scores are winsorized to the same fraction as the portfolio company data.
5. Results
5.1 Performance Change Analysis
The regression analysis starts with the simple regression of ROA change and the pre-‐buyout Altman Z-‐score. Subsequent regressions are estimated by adding one variable at a time to control for outside effects. Table 3 contains the results of these regressions. Column 1 is the simple regression with just the pre-‐ buyout Altman Z-‐score and a constant. The Z-‐score has a negative sign and is significant at 1%. This is congruent with the notion that more inefficiency (a lower Z-‐score) has a positive effect on the change in performance. The R-‐squared of this regression is 0.1394.
Table 3 Regression ROA Change and Pre-‐Buyout Altman Z-‐score
Dependent variable: Change in ROA
(1) (2) (3) (4) Constant 8.72*** (2.88) 10.44*** (3.42) 10.49*** (3.44) 28.9** (2.26) Prebuyout Z -‐4.34*** (-‐3.60) -‐3.9*** (-‐3.96) -‐3.91*** (-‐3.98) -‐4.4*** (-‐4.03) Ch in PGROA1 1.62*** 1.61*** 1.62*** (3.25) (3.19) (3.17) FTSE ret -‐1.58 -‐2.28 (-‐0.28) (-‐0.41) Size2 -‐1.59 (-‐1.62) R-‐Squared 0.1394 0.2830 0.2832 0.2979 # obs 132 122 122 122
Note: t-‐stat with robust std. errors in parentheses, significance level * p<0.1, ** p<0.05, *** p<0.01 1: Value at t4 -‐ value pre
2: Log(Total Assets)
Adding control variables has a positive effect on the R-‐squared. Column 2 is the regression with the added sector control. This control is the change in median peer group ROA. The Z-‐score has a negative sign and is still significant at
1%. The sector control has a positive sign and is also significant at 1%. This is in line with logical expectations, if the median peer group ROA goes up during a period this could point to a boom in the sector, which positively affects the ROA of a company operating within that sector. The R-‐squared of this regression is 0.2830. This jump is in line with the logical importance of controlling for sector conditions in analyzing the performance of a company. As mentioned before, the observations drop from 132 to 122 because of some missing peer group values. The regression in column 3 controls for broad market conditions with the inclusion of the FTSE100 index returns. This variable is not significant at 10% and the inclusion leads to a marginally higher R-‐squared (0.2832). It is not directly obvious why including a control for broad market conditions would be insignificant, especially considering the market volatility in the sample period. However, a logical explanation could be that the change in median peer group ROA is already sufficiently dependent on market conditions to explain this influence. For the purpose of completion the FTSE return is left in because it marginally improves on the explanatory power of the model and leaves the significance of the other variables unaffected.
Column 4 adds a pre-‐buyout size control to the regression. The peer group selection already intrinsically controls for size in its calculation. However, for completion this inclusion controls for any remaining size and scale effects. As expected the size control is not significant. The Altman Z-‐score and the peer group control are still significant at 1% and the R-‐squared increases to 0.2979. Based on the coefficients of regression 3 and 4, and the standard deviations reported in the descriptive statistics, the effect of pre-‐buyout efficiency on the change in ROA is significant. A firm scoring one standard deviation lower in pre-‐buyout Altman Z-‐score has an expected change in ROA 7.43% higher than mean-‐efficiency firms.
5.2 Efficiency Change Analysis
The regression analysis starts with the simple regression of Altman Z-‐ score change and the pre-‐buyout Altman Z-‐score. Subsequent regressions are estimated by adding one variable at a time to control for outside effects. Table 4 contains the results of these regressions. Column 1 is the simple regression with
just the pre-‐buyout Altman Z-‐score and a constant. The Z-‐score has a negative sign and is significant at 1%. This is congruent with the notion that more pre-‐ buyout inefficiency (a lower Z-‐score pre-‐buyout) has a positive effect on the change in efficiency at t4. The R-‐squared of this regression is 0.2528.
The second regression controls for the sector performance by adding the change in peer group median ROA. The control is not significant and the R-‐ squared increases slightly to 0.2619. The sign and significance of the pre-‐buyout Z-‐score is unaffected. The control is added not because of a specific direct link between sector performance and firm efficiency but to control for any distortions. A potential sector boom could lead to higher operating revenue and/or higher EBIT, causing a higher Altman Z-‐score without any changes in operational efficiency.
Columns 3, 4, and 5 add the change in median peer group Altman Z-‐score in the place of the change in peer group median ROA. This controls for sector-‐ wide efficiency changes such as technological developments or cyclical sector effects. These regressions omit the change in peer group median ROA because of possible multicollinearity, since the ROA is a part of the Altman Z-‐score. Just as the sector ROA control, the peer group change in Altman Z-‐score is not significant and slightly increases the R-‐squared to 0.2622. The inclusion of the peer group Z-‐score gives a slightly higher R-‐squared and its t-‐stat (0.83) is also marginally higher than that of the peer group ROA (0.52). Therefore, regressions 4 and 5 include the peer group Z-‐score, but do not include the peer group ROA control.
Table 4 Regressions Z-‐Change and Pre-‐Buyout Altman Z-‐score
Dependent variable: Change in Altman Z-‐Score
(1) (2) (3) (4) (5) Constant 0.95*** (4.38) 1*** (4.28) 0.97*** (4.31) 0.98*** (4.26) (0.61) 0.53 Prebuyout Z -‐0.4*** (-‐5.27) -‐0.4*** (-‐5.08) -‐0.41*** (-‐5.09) -‐0.41*** (-‐5.06) -‐0.39*** (-‐4.63) Ch in PGROA1 0.01 (0.52) Ch in PG Z1 0.09 0.09 0.1 (0.83) (0.80) (0.91) FTSE ret -‐0.32 -‐0.3 (-‐0.50) (-‐0.48) Size2 0.04 (0.57) R-‐Squared 0.2528 0.2619 0.2622 0.2649 0.2667 # obs 131 121 121 121 121
Note: t-‐stat with robust std. errors in parentheses, significance level * p<0.1, ** p<0.05, *** p<0.01 1: Value at t4 -‐ value pre
2: Log(Total Assets)
Column 4 adds a control variable for the FTSE100 index return. This because the FTSE control can filter out broader market-‐effects that could affect the change in Z-‐score. The control is not significant and the R-‐squared increases to 0.2649. The sign and significance of the pre-‐buyout Z-‐score is unaffected. Lastly, regression 5 adds a size control. This variable controls for possible scale effects not controlled for by the Z-‐score. The size control is not significant and the R-‐squared increases to 0.2667. Although the t-‐stat drops a little (from -‐5.06 in regression 4 to -‐4.63 in regression 5), the sign and significance of the pre-‐buyout Z-‐score is unaffected.
Based on the coefficient of regression 3, and the standard deviations reported in the descriptive statistics, the effect of pre-‐buyout efficiency on the change in efficiency is significant. A firm scoring one standard deviation lower in pre-‐buyout Altman Z-‐score has an expected change in Altman Z-‐score 0.77 points higher than mean-‐efficiency firms.
5.3 Credit and Collection Results
Table 5 contains the results from the regressions on the credit and collection period benchmarks. For both dependent variables the pre-‐buyout benchmark is significant at 1% with a negative sign. The interpretation of this is that if a company scored worse on its pre-‐buyout benchmark, it is expected to improve more. The pre-‐buyout benchmarks appear to be capable predictors of post-‐buyout efficiency gains achieved by the company. The broad Z-‐score is not significant for either regression. An explanation for this could be that a general measure such as the Z-‐score cannot account for very specific inefficiencies such as those associated with the credit and collection period.
Table 5 Credit and Collection Period Regressions
Dependent variable: Change in Creditdays Benchmark
Change in Collectiondays Benchmark (1) (2) (1) (2) Constant 1.25*** (7.38) 1.44*** (4.83) 0.99*** 1.17*** (8.06) (4.19) Benchmark Pre -‐0.86*** (-‐9.28) -‐0.87*** (-‐9.19) -‐0.79*** -‐0.79*** (-‐11.38) (-‐11.29) Pre-‐Buyout Z -‐0.06 -‐0.06 (-‐0.85) (-‐0.84) R-‐Squared 0.7063 0.7078 0.7512 0.7512 # obs 114 114 114 114
Note: t-‐stat with robust std. errors in parentheses, significance level * p<0.1, ** p<0.05, *** p<0.01
6. Robustness
6.1 Individual Altman Variables
To measure efficiency this thesis uses the Altman Z-‐score. As is explained in section 3.1, the Altman Z-‐score is a weighed score of measures of liquidity, profitability, leverage, solvency, and activity. It has the following form:
𝑍 = 0.717 𝑋! + 0.847 𝑋! + 3.107 𝑋! + 0.420 𝑋! + 0.998 𝑋! . Where: 𝑋! = 𝑊𝑜𝑟𝑘𝑖𝑛𝑔 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠, 𝑋! = 𝑅𝑒𝑡𝑎𝑖𝑛𝑒𝑑 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠, 𝑋! = 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝐵𝑒𝑓𝑜𝑟𝑒 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑎𝑛𝑑 𝑇𝑎𝑥𝑒𝑠 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠, 𝑋! = 𝐵𝑜𝑜𝑘 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝐸𝑞𝑢𝑖𝑡𝑦 𝑇𝑜𝑡𝑎𝑙 𝐿𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠, 𝑋! = 𝑂𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑅𝑒𝑣𝑒𝑛𝑢𝑒 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠.
For this thesis it sufficed to calculate the Z-‐score with the coefficients suggested by Altman and use this as a measure of overall efficiency. However, understanding the effects of the individual variables is an important step in determining the direction of future research. Therefore, this section will perform some of the core regressions of the thesis with the individual Altman variables, rather than the broad Z-‐score. The results of these regressions can be found in Table 6 and Table 7. For the ROA regression, the change in peer group control is the change in median peer group ROA. For the Altman Z-‐score regression, the change in peer group control is the change in median peer group Altman Z-‐score.
Table 6, column 1 contains the regression results for the change in ROA regression with the individual pre-‐buyout Altman variables. When all of the variables are included only the EBIT and the Equity Book Value variables are significant at 1% and 5% respectively. The peer group control is also significant at 1%.
Interestingly, when the individual Altman variables are included, the R-‐ squared of the regression (0.4438) is much higher than in the corresponding regression using the Z-‐score (0.2830). An explanation for this could be that the Z-‐score is designed to score a company’s default risk, and that the constraints imposed on the joint effect of the variables limit the explanatory power when looking at profitability.