Private equity-exit through the IPO : is there money on the table? : a sample of European IPOs

(1)

Private Equity-Exit through the IPO;

Is there Money on the table?

A Sample of European IPOs

Universiteit van Amsterdam - Amsterdam Business School

MSc Finance, Asset Management

Master Thesis

Georgallides Theodoros

July 2017

(2)

Abstract: This paper investigates the performance of private equity-owned companies after

the exit of the private equity partner. I use IPOs as a means for exit and manually construct a panel dataset with 237 PE-backed ex-portfolio companies and 847 non-PE control companies for the period 2005-2011 in Europe. I find significantly higher buy-and-hold returns for PE-backed IPOs over one, three and five years. I attribute this to; higher IPO underpricing for PE-backed IPOs, better performance post IPO for PE-backed companies and share price dilution for non-PE companies.

(3)

Statement of Originality

This document is written by Theodoros Georgallides 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.

(4)

Acknowledgments

I am grateful to my family for their endless love, support and financial support, without which I would not have made it.

I am also thankful to my supervisor Florian Peters for his comprehensive feedback into completing this paper.

Lastly, I am grateful to my university for all the best years of my life.

I dedicate this work to my mum, without her love and endless support I would not have made it.

(5)

Given the amount and quality of new research published in recent years in highly reputable journals on the performance and impact of Private Equity (PE) as an industry, the topic is still very young and its impact very uncertain. This is because PE as an industry begun in the early 1980s, and to put this into perspective; the first initial public offering (IPO) to take place in the world happened here in Holland for the Dutch East India Company (VOC) in 1602 (Goetzmann & Rouwenhorst, 2005, p. 14). Given that the PE industry has grown exponentially in the last 30 years (Kaplan & Schoar, 2005) and has increasingly been used as an alternative for great companies such as Uber, Airbnb and Spotify to going public, the industry is only just starting to take off (2016, November 14) The Wall Street Journal. Due to limitations in available data because of its private nature, research until now tends to be biased depending on database and methodology used. I divide the research into three “pillars” which I define below. In this paper I try to shine light to the ever-increasing important industry called private equity.

In my first pillar, research looks into the impact PE has on investors as well as the society in general in terms of impact to employment, GDP growth and company-specific variables such as cost reduction, revenue enhancement and job turnover. All of the variables mentioned above have a direct or indirect impact to society as a whole. In my second pillar, research on private equity focuses on the performance of PE funds and how they compare to the conventional [public equity] funds. Some research finds an outperformance of PE funds as an investment, others find the opposite, depending on database and method of analysis. The third pillar of research into the PE industry is the analysis of, more specifically, PE-owned portfolio companies and the comparison of those to non-PE-PE-owned companies, either publicly- or privately-owned companies. I find this pillar/category of PE research the most important, as how PE impacts their portfolio companies leads to how the other two pillars will look like. I further divide this third pillar into two subcategories. The first is the comparison of the performance of the PE portfolio companies in the post-buyout period or the period in which the company is majority owned by a private equity investment firm. The second subcategory I divide this pillar into, is the period after exit of the private equity investment partner and whether there is persistence or reversal of the effect found in the first subcategory.

The third pillar and consequentially the second subcategory, is the most under-researched topic of the PE academic literature. Whereas data for PE funds is more readily available, there are large limitations to the data of private (portfolio) companies due to the limited requirements for reporting by authorities, especially in the US, and subsequently the

(7)

limitations of readymade databases. In Europe however, data of private companies is more readily available as a consequence of obligatory annual report filings, and any research that does focus on the “third pillar”, does so by using data from European markets (Acharya et al., 2013).

These significant differences in the impact and performance of private equity backed investments compared to non-PE backed investments found in the literature arise from the special way PE investors go about their business (Davis et al., 2013). As explained briefly above, and is explained in detail in section I, I call PE investors corporate restructuring “gurus” who work in hand with company management to decrease costs, increase revenues, maximize efficiencies and at the same time are known to lever up companies they invest in with high levels of debt. All of the above leads to maximization of free cash flows that subsequently lead to higher returns on investment or equity (ROE).

Private equity investors hold their investments for a period of five to seven years. After that period, these investors monetize their investments by selling their stake in the company either back to the owners, the management team, other investors or initiate an initial public offering (IPO) (Robinson & Sensoy, 2013). This is widely known in finance as a private equity exit. Whereas there has been research focusing on how portfolio companies perform post-buyout, during the PE holding period (5-7 years), there has been limited or no research focusing on the post-exit period. The period when PE investors sell their stake in the company and cash out their investment. I find this to be an equally interesting period to the post-buyout period because, firstly, as explained by past relevant research, significant statistical differences exist in all pillars of research on the impact of PE. The consensus agrees on the fact that PE has a strong impact to society, PE funds outperform non-PE funds and portfolio companies, in many ways perform better than their counterparts. These important findings stem out of the performance of the actual portfolio companies PE firms choose to invest in. In other words, if it was not for the portfolio companies themselves, PE investors would not be able to perform the way they do and vice versa. Also, does a PE-exit mean that there are no more abnormal returns to be made? In other words there is either a persistence effect to the portfolio company post-PE-exit or a reversal.

My mission is to see if persistence exists to the direction of the hypotheses formed as a consequence of past research findings to the performance of portfolio companies after the private equity partners’ exit. A priori, my hypothesis is that PE investment partners are better at finding undervalued investments and subsequently, selling at, or close to fully efficient value after having exploited all free cash flows. Meaning, portfolio companies are

(8)

not worth much after the exit of the PE partner. My hypotheses are stated clearly in the next section of this paper.

When it comes to portfolio company specific performance, research finds outperformance of profitability and sales variables as well as returns on equity of PE portfolio companies over their counterparts post-buyout and during the PE holding period. My research question is; Does this outperformance achieved by the portfolio companies carry on after the exit of the PE partners? To answer this I use a variety of variables to firstly see whether ex-portfolio companies are still good investments as compared to “conventional” privately or publicly owned companies and secondly if the answer to my first topic is due to intrinsic productivity variables.

My research focuses on PE-exits through initial public offerings (IPOs). Under an IPO, the PE investment firm sells or partially sells its share to the public on the stock market. I look at internal company data prior to the IPO and post IPO as well as stock performance post IPO and compare the PE-backed companies to the rest. I generate a list of European PE (IPO) exits from Thomson one for the years 2005 to 2011. I then match that list with a list constructed on Standard & Poor’s Capital IQ for all IPOs that took place in Europe during the same period. I manually construct two datasets using the company lists provided by Thomson one and Capital IQ of identical panel data for both groups. My panel consists of data for two years prior to, and five years post-IPO, totaling eight years of data per company. I end up with 233 companies for my treatment group and 847 companies for my control group. My main analysis consists of constructing buy and hold returns (BHRs) using the IPO stock price, and selling after one, three and five years post IPO. I control for book-to-market ratio, revenue growth as well as debt proportion of total capitalization. To cancel out any variation in performance due to differences arising from different years, I use time fixed effects (FE). I also employ a differences-in-differences regression analysis to test intrinsic company variables and whether they experience a reversal or persistence from the PE holding period, which supplements at explaining any out- or under-performance found in my first analysis. I find significant outperformance of private equity-backed IPOs over their “conventional” counterparts with a positive difference of 12% over a holding period of one year, the same outperformance over three years and 16% higher buy-and-hold returns over five years post IPO. These results oppose what I expect a priori that PE investors exploit all cash flows available by buying undervalued companies and selling them at overvalued prices. My results however, are completely in line with recent IPO underpricing literature which finds that Venture Capital (VC) investors underprice their IPOs significantly more

(9)

than their counterparts. I assume that for this reason, Private Equity investment partners also leave more money on the table than conventional IPOs do. For my second hypothesis I test differences in my main explanatory variables and find that PE portfolio companies continue to show higher revenue growth post-IPO. PE portfolio companies tend to have significantly lower book-to-market ratios and I find a small significant reversal in the debt proportion growth after PE-exit/post-IPO.

The rest of the paper is organized as follows. Section I describes the relevant background literature leading up to my hypothesis and research. I give special attention to the relevant literature for this paper. Section II follows with my data and descriptive statistics. Section III describes the methodology I use to tackle my research questions. Section IV presents and analyzes the results of my main regression analyses, and section V presents and analyzes further regression analyses and results as robustness checks. Section VI concludes my paper, provides a final discussion and some ideas for further research on the topic.

I. Literature Review

When assessing historical literature, I find that authors have a certain bias towards the impact and performance of PE, and the outcome of their research meets that bias. Databases containing PE past returns tend to be biased as they suffer of survivorship bias and other biases (Harris et al., 2014). To address these problems, different authors choose different ways to approach their research.

Kaplan and Stromberg (2009) give a very thorough analysis of the private equity industry from the 1980s boom to the bust of the 1990s and the even stronger reemergence of large private equity investors into the mergers and acquisitions scene of the 2000s financial boom market. The authors argue that during the 1980s PE-boom, it was believed that this form of financing would become the dominant form of financing for companies in general. The reason for this assumption stemmed from the fact that private equity combined efficiencies and improvements to common problems such as agency conflicts, present in companies with public capital structures. In the 1980s low leverage, dispersed ownership and weak corporate governance would give its place to high leverage, concentrated ownership and strong, scrutinized corporate governance, where management became liable to expert investors. This was the private equity wave of the 1980s. However, Kaplan and Stromberg (2009) explain that public equities did not in fact give their place to private equity as the bust and crash of the junk bond markets in the 1990s prevented that from

(10)

happening. Private equity is being given a second chance since the early 2000s, where the

leveraged buyout industry has recovered and has experienced a second even larger boom

than the one of the 1980s. The research performed by Kaplan and Stromberg (2009) argues that since the early 2000s and the subsequent financial crisis after 2007, ending in 2009, has had a permanent positive impact onto the PE industry, which backs my personal beliefs. I argue throughout this paper that because of these efficiencies in capital structure and beyond, achieved by PE investors on their portfolio companies, PE is the way to go in company financing.

All private equity funds are organized as limited partnerships. Private equity (investment) firms serve as the general partners (GPs) of the funds and large institutional investors and wealthy individuals or families serve as limited partners (LPs). The GPs have absolute control of the funding and running of the companies whereas the limited partners provide that funding to the PE firm. Hence GPs need to be highly reputable expert investors. Typically general partners invest 1% and limited partners invest the other 99% (Acharya et al. 2013). Typically these partnerships last for ten years, with exceptions on average lasting up to thirteen years. The first five years of a fund’s life are spent performing analysis and due diligence of prospective investments, potential portfolio companies. The typical period funds are invested in companies is five years with a possible extension to eight years (Acharya et al. 2013). If successful, the GPs will raise capital for follow-up funds. GPs compensate themselves generously. Unless otherwise stated, the usual contract has a scheme of 2-20-1. 2% as a constant management fee, 20% as carried interest which is charged on the profits, and all of this for investing just 1% of their own personal money (Robinson & Senso, 2013). GPs may also charge a management fee to the portfolio company itself.

Pillar I- Impact on society

Davis, Haltiwanger, Jarmin, Lerner and Miranda (2013), in a similar manner to my proposition, look at PE portfolio companies and the impact onto the general employment level of the economy as well as the productivity of the companies in question. They look at the employment levels of target portfolio companies after being acquired by PE investors comparing them to a control group of companies in the same industry but not being acquired by PE investors. They find that while portfolio companies experience higher job destruction compared to non-PE-owned counterparts, they also experience much higher job creation than their counterparts. Summing up job creation with job destruction in portfolio companies leads to a negligible difference in employment between the two groups of companies. They also argue that this “refresh” in workforce experienced post-buyout is

(11)

actually beneficial to the society overall. When it comes to productivity, target firms exit less productive establishments and enter new more productive establishments faster than control companies. In essence, PE involvement leads to cleansing and efficiency improvement to the company as a whole and its workforce. In a capitalistic society this is beneficial, however, the actual humane impact is still debatable.

Pillar II- Funds’ performance

Any impact PE GPs bring to portfolio companies they acquire is translated into returns to their LPs (usually pension funds or family offices) who invest into their funds (Kaplan et al. 2014). Any cost reductions or revenue enhancements as well as levering up target companies’ capital structures leads to higher Returns on Equity (ROE) for their LP investors. One of the main differences between private equities and public equities is the disclosure of data such as publishing annual and quarterly reports and subsequently financial statements. Investors and the general public can easily access data and information on companies listed on the stock market therefore valuing such investments is less arbitrary. It is this asymmetric information component of private companies and PE funds in general which allows for dispute about their performance.

Robinson and Sensoy (2013) access a large database provided by a large institutional investor with 837 funds. Having a dataset that is constructed by a limited partner tends to be objective rather than biased towards PE GPs’ exceptional performance. This is because such a database is constructed by a PE GP’s client. They find that more successful GPs are able to demand higher compensation and at the same time perform better (I.e. higher returns on investment) net of fees. An interesting outcome arising from their research is the fact that GPs tend to take more years to exit investments when there is a higher percentage of fixed compensation in the contract, while the opposite is true immediately after GPs start receiving carried interest.

Phalippou and Gottschalg (2008) state that the performance of PE funds is overstated and this is due to inflated accounting valuations of portfolio companies and data biasedness towards better performing funds due to survivorship bias. They find that before fees, PE funds do outperform the S&P 500 by 3% but net of fees, the performance of the PE funds underperforms the S&P 500 by 6% annually. As stated by Kaplan (2012), this research is constrained due to the fact that it is performed on mature PE funds that are liquidated or about to be liquidated. One important issue addressed by this research is the accounting overvaluation of current investments (portfolio companies) by private equity general partners.

(12)

Whereas there seem to be two sides in the research on private equity performance, those who find positive impact and performance of the industry and those who find negative impact and performances that are overstated by the PE firms, authors like Kaplan have at first found PE underperforming and in a later stage, with updated better research have come to the opposite outcome. Kaplan and Schoar (2005) find that leveraged buyout (LBO) funds’ returns slightly underperform the returns of the S&P 500 while they outperform the S&P500 gross of fees. They also find evidence of a persistence effect. Where GPs outperforming with one fund, tend to outperform with the next or prior fund or even second prior fund to the current one. This opposes the effect found on mutual funds, where no such relationship can be proven by the research, and this backs the theory that PE investors perform better not just because of “luck” but because of “skill”. They do not find evidence of any survivorship bias affecting their results. Kaplan and Schoar (2005) find that better GPs may be able to invest in better investments, which brings the persistence. Due to competition, prospective portfolio companies are able to choose the best partner for themselves. In other words, being a good PE (investment) firm goes both ways. Good PE investors are able to choose or be chosen by better target companies. They also find evidence that venture capitalists get better terms when investing in companies, such as lower valuations which in turn improves the performance of the given PE firm.

Phalippou, Franzoni and Nowak (2012) come up with an academic way to value the returns of PE. They use Pastor and Stambaugh’s (2003) four-factor capital asset pricing model that includes a liquidity factor and find a 3% liquidity risk premium and a total risk premium of 18% for private equity assets. Besides PE funds being illiquid by nature due to long “lock-up” periods, they attribute the connection of PE assets and liquidity to the funding liquidity risk experienced by PE investors, due to the need for credit financing throughout a funds life. Phalippou, Franzoni and Nowak (2012) state that “the apparent high performance of private equity investments over conventional investments is largely explained by the liquidity risk premiums investors need to be compensated for baring these risks” and when controlling for these risks PE funds underperform their public counterparts.

Lastly, Kaplan, Harris and Jenkinson (2014) use the most advanced database up to date and argue with significant proof as to why analysis and results presented previously by Phalippou, Franzoni and Nowak (2008) and are negatively biased. They attribute the negative bias on problems with data provided by databases such as Venture Economics, as well as negatively biased assumptions Phalippou, Franzoni and Nowak (2008) impose onto incomplete data they use. Using the Burgiss database comprised of data from 200

(13)

institutional (LP) investors, Kaplan et al find outperformance of PE funds over the S&P500 of 20% to 27% on average over the lifetime of a fund and more than 3% outperformance over the S&P500 per year.

To sum up, in any way you look at it, PE funds outperform their counterparts, even if these higher returns are due to the higher liquidity risk these funds are exposed to. There is a however, a general dissatisfaction in the industry about the very high fees these investors charge to their clients, even if net of fees’ returns are higher than public funds (2017, April 16) The Wall Street Journal. It is evident that for the funds to outperform, portfolio companies need either outperform their counterparts in terms of better performance, higher IRRs or produce higher returns on equity (ROE) because of PE corporate restructuring and levering up or lastly, the best companies are picked or pick the best PE investors. This leads to my last “pillar”.

Pillar III- Impact on portfolio companies

Putting together the literature puzzle of the most up-to-date private equity research means combining the findings of all three self-defined pillars that make up the bigger picture. From the impact that PE has to society to the impact PE funds have onto investors’ portfolios as well as portfolio companies themselves, it all comes down to the last pillar of the literature; the actual impact PE has to the portfolio companies acquired.

Wilson, Wright, Siegel and Scholes (2011) look at PE in the UK. They use the whole population of PE-backed companies and match them to the whole population of non-PE-backed companies whether publicly- or privately-owned. They use a time-period that overlaps with my sample period. They test internal company variables and draw comparisons with non-PE-owned companies. Their research is very similar to my research, but instead they look at the period post-buyout and compare the performance of PE-owned companies over the holding period. They find a significant positive difference of 5 to 10% in productivity variables and a positive difference of 3% to 5% in profitability of buyout companies over the controls.

In a similar manner, a study linking PE’s corporate governance and its impact onto [PE] portfolio companies performed by Acharya, Gottschalg, Hahn and Kehoe (2013) finds higher abnormal performance benchmarked to a control sample of peers, which is attributed to improvements in sales (revenues) and profitability (namely EBITDA; a perfect indicator used by the industry as the closest variable to assess how much free-cash-flow is left over for investors). The authors go a step ahead, by cross-sectionally analyzing the skills of general partners that lead to these improvements. They find that ex-consultants and partners

(14)

that used to be in the same industry as the company acquired prior to becoming PE investors, tend to outperform due to organic growth, whereas PE investors that were previously bankers or accountants realize outperformance in their portfolio companies through mergers and acquisitions. These findings outline the fact that certain skillsets available to PE partners individually or as a firm, affect the way they go about corporate restructuring in their portfolio companies. These specific skills and the matching of these skills with the appropriate acquired companies lead to an overall outperformance. They find that their sample (large mature PE firms) has a mean IRR of 56.1% of which 34% comes from abnormal inner performance, 50% comes from the higher leverage and the remaining 16% comes from the exposure to the certain industry the company is in. These three findings are exactly the three points PE exploits in order to outperform their benchmarks. Improving the functioning of the companies they acquire, levering up the companies’ capital structures and focusing on sectors of expertise that are found to outperform the rest.

An interesting study performed by Liu (2017) due to its method in establishing validity, looks at a sample of failed leveraged buyouts (LBO) for reasons other than target firm fundamentals. Liu’s (2017) emphasis is the public-to-private PE subsector of companies. Using this sample, Liu (2017) extracts and removes the positive reaction of the stock price due to the announcement of the acquisition by a PE investor and keeps the operating improvements and skills in choosing undervalued companies only. He constructs a sample of successful LBOs, a sample of withdrawn (due to exogenous reasons) LBOs and his control group sample of all other companies. He finds that both treatment groups experience an upward revaluation in a similar manner, with withdrawn, “unsuccessful” LBOs’ targets appreciating by 13.4% over a matched non-PE “affected” sample of companies after the announcement. Liu (2017) finds that the market revalues PE targets upwards, meaning that markets recognize the ability of PE partners at spotting undervalued investments. Above and beyond that effect, he finds significant improvements in operating performance, along with all other studies. Thirdly, he finds that PE general partners adjust the [portfolio] company’s debt structure upwards and lastly he finds a strong impact onto the portfolio companies’ management.

Lastly, a study presented by Aleszczyk, De George, Ertan and Vasvari (2016) looking at buyouts during 1998-2014, focuses on the effects of PE to portfolio companies but, in a similar manner to the rest of the literature in “pillar III”, the study looks at the post-acquisition holding period. The method of research followed by the authors is also similar to the rest of the studies along with mine, testing differences between a treatment (PE) group

(15)

and a control group (here, corporate M&A). Backing Liu (2017) and Kaplan and Schoar (2005), the authors find that PE funds pay significantly lower EBITDA multiples in acquiring targets than conventional (corporate M&A) professionals, making them either expert negotiators or good stock pickers due to their ability to spot undervalued investments, or both. Also, to back literature discussed above that PE investment partners pursue high growth - higher ROE strategies; they find that PE portfolio companies experience an increase in leverage, operating profits, sales and assets. They find that smaller PE funds focus more on increasing revenues whereas more experienced larger funds are able, through corporate restructuring to increase EBITDA more.

Given all the theories discussed above, there are a handful of assumptions that can be extracted with certainty. PE’s impact onto society as a whole is still not certain, but tends to be a strong impact as presented from the high job destruction and creation. PE funds tend to outperform their counterparts either because this outperformance is due to a higher risk premium, or because PE investors are highly skilled at spotting good investments and/ or restructuring these investments. The impact of PE onto portfolio companies results in a stronger performance compared to benchmarks either due to higher, more efficient productivity or due to higher ROE due to capital structure changes, increases in sales or profitability. What is yet to be found and can also be used as a robustness check to the previous literature discussed, is whether PE investors’ positive impact onto portfolio companies is persistent post PE-exit.

Hypothesis

All studies discussed above focus on the period post-buyout and during the holding period, where a PE investment firm acquires a company and keeps it in a [PE] fund for the usual time period of five to seven years. An equally important part of a PE investor’s strategy is the way and timing of an exit. This part of a PE investment is overseen by the research. As established by the literature, PE investors are expert professionals, I call them investment “gurus” who can spot a good investment, exploit it on the way and get out on the right time in order to raise new capital or reinvest their profits into the next good investment, in this way they manage to keep their returns higher for longer as opposed to mature companies. I therefore expect companies that used to be PE portfolio companies, to continue to outperform their counterparts when it comes to productivity variables such as revenues and profitability, capital structure differences such as higher debt levels, after the PE-exit. Given the second important assumption that has been shown in recent research that PE investors are corporate restructuring experts, I expect them, in the same manner as when

(16)

spotting or bargaining for undervalued companies, to be able to extract the highest possible price on the imminent exit from these investments. As a means of analysis and data availability I therefore focus on one of the main exit routes for PE investors which is through an IPO. For my first hypothesis I therefore expect that IPOs done by PE investors are overvalued compared to non-PE-backed IPOs. As Phalippou and Gottschalg (2009) explain, this could be either due to the overvaluation of accounting figures, or simply because companies that are exited are not worth as much anymore. On the other hand there is an alternative hypothesis to my first hypothesis that stems out of two causes. First, these companies have acquired special skills not available to their counterparts post-IPO and can continue to perform better. Secondly and more importantly, I draw a link between literature focusing on IPO underpricing in general which finds that PE IPOs are more underpriced than non-PE. I explain some of the implications on IPO underpricing below. These two reasons lead to an alternative hypothesis which states that PE-backed IPOs outperform their controls. For my second hypothesis I extend the analysis of my first hypothesis to analyse the company intrinsic variables that might lead to the under- or outperformance of ex-PE companies post-IPO. I expect a persistence in revenue and profitability outperformance to continue post-IPO, and I expect a reversal in company debt measures post-IPO/PE-exit.

Regarding the implication in my first hypothesis, literature presented by Liu and Ritter (2011) on IPO underpricing, finds that VC IPOs, which are a big part of PE IPOs, experience significantly higher underpricing than non-VC-backed IPOs. Levis (2011) finds that PE-backed IPOs are larger and more profitable companies and in the three year period post-IPO perform better in terms of market and operating performance than controls. Finally Krishnan, Ivanov, Masulis and Singh (2011) find that a venture capitalist’s reputation influences post-IPO performance positively and this is due to a continuation in the involvement of the VC firm in the ex-portfolio company post-IPO. These findings may provide answers to my first alternative hypothesis, that PE-backed IPOs are actually undervalued.

(17)

II. Data

To test my hypotheses I construct a treatment group made up of PE portfolio companies and I compare them to a control group made up of all other companies that underwent an IPO. Due to different rules and regulations between the Europe and the United States, data availability for private companies is more complete for EU companies over American. I therefore use the Thomson One database and get a complete list of all private equity IPO-exits for the period 2005-2011 that took place on a European stock exchange only. I use my treatment group list as my benchmark and match that criteria to construct a list of non-PE-backed IPOs for my control group. I use Standard & Poor’s Capital IQ to construct my control group list with the same criteria. I end up with 330 companies for my PE-backed IPO list from Thomson One and 1735 non-PE IPOs for my control group list. The list produced by Capital IQ initially includes the entire sample of PE-backed IPOs which I remove from my control group for analysis purposes. I also remove all financial institutions such as banks and investment funds from my IPO lists and I also remove all companies which I do not find sufficient data1 for on Capital IQ. Finally, I end up with 237 PE-backed IPOs for my treatment group and 847 non-PE-backed IPOs for my control group. For both groups I manually construct an identical comprehensive panel data list. I handpick all the data I need for each company and each year using Capital IQ, also known as the most reliable database in academic finance and beyond (Bernstein et al. (2010) and Stromberg (2007)). For each company in my panel I include eight years of data, two years prior to the IPO year, the IPO year and five years post-IPO. I use the most commonly used variables found in related literature on PE research. These include productivity variables such as revenue and profit margins per year, and capital structure variables such as debt proportion of total capitalization, long term debt and net debt. I add valuation variables such as book-to-market ratio, market capitalization (post-IPO), enterprise value as well as share price (post-IPO) to my panel data for my valuation analysis after PE-exit.

1_{I remove all companies which do not have data at all either due to going bankrupt or because they are}

(18)

Table 1. Descriptive Statistics

Sample summary statistics for variables of interest for all years and all companies. The table shows summary statistics for key variables for a sample of 847 control and 237 treatment companies. Sample selection criteria are described in Section II. The list of sample companies is provided in the Appendix. The data are from S&P Capital IQ database. ‘‘Revenue’’ is company i’s revenue on year t. ‘‘Debt (%) of total capitalization’’ is the proportion of debt as of total capitalization of company i in year t. ‘‘Book-to-market ratio’’ is company i’s book-‘‘Book-to-market ratio on year t. “EBITDA” is the earnings before interest, tax, depreciation, amortization for company i in year t in thousands ($) dollars. A log transformation is applied to revenue. All variables with “winsorized” have outliers removed. All variables are measures throughout the whole sample period, before and after IPO.

Variable Number Minimum Lower Quartile Median Upper Quartile Maximum Mean Standard Deviation

Treatment group

Revenue 233 -8.54 2.88 4.68 6.35 10.19 4.53 2.51

Revenue (winsorized) 233 -4.27 2.88 4.68 6.35 10.19 4.53 2.49

Book-to-market 233 -10.37 0.23 0.39 0.72 7.69 0.55 0.84

Book-to-market (winsorized) 233 -6.51 0.23 0.39 0.72 7.69 0.54 0.81 Debt (%) of total capitalization 236 -67.98 0.05 0.27 0.53 59.40 0.33 2.32 Debt (%) of total capitalization (winsorized) 236 -2.27 0.05 0.27 0.53 17.62 0.36 0.81

EBITDA 230 -5775.30 249.76 399.67 551.23 6230.00 378.88 359.79 EBITDA (winsorized) 230 -125.30 249.76 399.67 551.23 6230.00 386.12 306.56 Control group Revenue 843 -11.51 2.51 3.84 5.25 16.92 3.87 2.27 Revenue (winsorized) 843 -4.27 2.51 3.84 5.25 10.84 3.86 2.22 Book-to-market 827 -330.67 0.31 0.61 1.14 13164.94 6.73 249.17 Book-to-market (winsorized) 827 -6.51 0.31 0.61 1.14 99.66 1.23 4.74 Debt (%) of total capitalization 843 0.00 0.05 0.27 0.52 141.81 0.44 2.58 Debt (%) of total capitalization (winsorized) 843 0.00 0.05 0.27 0.52 17.62 0.39 0.89

EBITDA 827 -302.59 1.86 7.65 37.10 2254802.00 2045.11 55697.28

(19)

Table 2. Descriptive Statistics - Buy & Hold returns

Sample summary statistics for variables of interest for all years and all companies. The table shows summary statistics for key dependent variables for a sample of 847 control and 237 treatment companies. Sample selection criteria are described in Section II. The list of sample companies is provided in the Appendix. The variables are constructed using data from S&P Capital IQ database. Buy-and-hold returns for 1, 3 and 5 years respectively are calculated using share price returns with IPO quarter closing share price as the buying price. Buy-and-hold returns for 1, 3 and 5 years, Market Cap respectively are calculated using market capitalization returns with IPO market capitalization as the buying price, and outliers have been removed.

Variable Number Minimum

Lower

Quartile Median

Upper

Quartile Maximum Mean

Standard Deviation

Treatment group

Buy & Hold Return 1-year 236 0.15 0.66 0.98 1.26 4.06 1.01 0.51 Buy & Hold Return 3-year 236 0.00 0.39 0.67 1.18 7.74 0.91 0.88 Buy & Hold Return 5-year 205 0.00 0.26 0.68 1.32 5.91 1.00 1.01 Buy & Hold Return 1-year, Market Cap 236 0.15 0.69 1.01 1.40 58.88 1.57 4.36 Buy & Hold Return 3-year, Market Cap 236 0.02 0.45 0.78 1.35 42.76 1.49 3.60 Buy & Hold Return 5-year, Market Cap 206 0.01 0.39 0.87 1.69 45.56 1.95 4.94

Control group

Buy & Hold Return 1-year 827 0.06 0.54 0.81 1.09 8.13 0.91 0.64 Buy & Hold Return 3-year 827 0.00 0.29 0.59 1.03 8.03 0.81 0.83 Buy & Hold Return 5-year 752 0.00 0.23 0.51 1.01 11.85 0.83 1.12 Buy & Hold Return 1-year, Market Cap 843 0.01 0.57 0.87 1.23 94.96 1.42 5.28 Buy & Hold Return 3-year, Market Cap 843 0.01 0.35 0.69 1.22 75.28 1.56 5.46 Buy & Hold Return 5-year, Market Cap 770 0.01 0.32 0.66 1.33 70.47 1.79 6.29

(20)

The main variables I use for my analysis are presented on table 1 above. I present the data for the treatment and control group separately. For each variable I show the actual variable found on my panel from the data I retrieved from Capital IQ and the same variable having removed outliers. I translate revenue into a logarithm to present it in terms of a base, for statistical comparison purposes. My main regression on buy-and-hold returns (BHR) is run on the PE indicator dummy which is my main dependent variable, and my control variables; revenue, book-to-market ratio (B/M) and debt proportion of total capitalization (debt %). After removing outliers for the logarithm of revenue, the minimum increases and is equal for both treatment and control groups. The maximum, also becomes almost equal for both groups at around 10.5. This removes any biasedness due to outliers. B/M is

winsorized in a similar manner. This variable experiences a large change when removing

outliers for the control group B/M ratios. From the table, it is easily derived that control group companies (non-PE) have larger B/M ratios than treatment group companies. That means that control group companies are undervalued, i.e. their market value is lower than their book value. B/M ratio is calculated using a company’s book value at time t divided by the market capitalization of the company at the same time. A priori, one would expect the variable “debt proportion” to be higher for the treatment companies rather than control companies. It seems however that the control sample includes a few very large outliers which bias the mean at first. After removing outliers for both groups, I am left with more comparable “debt proportions”. Before winsorizing, the “debt proportion” variable is upwardly biased for the control group and downwardly biased for the PE-backed group due to outliers. Due to the peculiar nature of EBITDA values that include a large portion of negative values, comparisons can only be made using the absolute dollar values of this variable. Any other statistical comparison to be made using margins or growth is highly inconsistent. EBITDA variables of the control group are upwardly biased due to outliers and the opposite is true for treatment group companies. After removing outliers, the means reverse, and the treatment group experiences a larger EBITDA mean over the control group.

In an efficient world, a company’s stock price is the best reflection to a company’s performance. Hence, I choose buy-and-hold returns (BHR) to assess the post-IPO performance of the two groups of companies. In table 2., I present these dependent variables of the model I use to test my first hypothesis. Buy-and-hold returns are constructed using the IPO’s price as the buying price, taken from the quarterly report of the company that is provided by the Capital IQ database. The assumption in producing these BHRs is that an investor buys shares at the IPO stock price and sells at the prevailing price in one, three or

(21)

five years after. BHRs are a perfect indicator of the appreciation or depreciation in an asset’s value. I construct a second BHR dependent variable which I use as a robustness test to my first regression. I use market capitalization on the IPO date as the buying price and the prevailing market capitalization in the same period of years as the selling price. Market capitalization is calculated using the number of shares outstanding at time t multiplied by the share price at time t. I expect the two BHRs to be very similar unless there are big differences in share issuance post-IPO. My share price BHRs are not biased from any outliers, hence I keep the raw BHRs calculated from the data I get. Share price BHRs are on average higher for PE portfolio companies. Market capitalization BHRs are highly biased hence I present the value of BHRs after removing outliers. I attribute this biasedness to the large variation in post-IPO share issuance that companies go through.

III. Methodology

To test my first main hypothesis I construct buy-and-hold returns for both groups of companies for one, three and five years. I regress these variables onto my main explanatory variable which is the company being an ex-PE-owned company and onto the control variables I include in my analysis; revenue, debt (%) and B/M ratio. I use a one year lagged B/M ratio in my regressions as I assume that book values adjust slower than market values. In order to remove any time biases I include each IPO year as a dummy (except one) as year fixed effects (FE).

Given the literature I present and its high validity, I expect ex-portfolio companies to perform significantly different in the short to medium term post-IPO, which is the period in which I construct BHRs for. First, given the strong expertise of PE investors when acquiring targets, I expect the same significant effect when exiting those portfolio companies. PE investors are able to identify and acquire undervalued companies, or through their bargaining power and expertise are able to acquire companies at lower multiples (Liu, 2017). I expect these investors to also be able to exit companies at higher multiples than are, for example families or other (non-PE) privately owned companies, selling stakes on the stock market using underwriters. Second, given a PE partner is exiting an investment, I expect the company to have been fully developed and grown to the point that the PE partner has no more abnormal ROE to extract from it. Hence, I expect PE IPOs to not be as appealing as non-PE IPOs to the general public and institutional investors, hence I expect a downward pressure to the IPO price in the short to medium-term. Lastly, I expect companies to “loosen up” after the exit of the PE partner who, as explained in the literature, kept the

(22)

company extremely lean and efficient and held the management liable to any good or bad choices (Acharya et al. 2013). On the other hand, for my alternative hypothesis, given that a priori, companies under PE ownership are developed with superior skills to perform better than their counterparts, I expect this effect to continue into the future and post-PE-exit through a better than average stock price return. Lastly, given recent publications on IPO underpricing, finding significantly more underpricing from VC firms, I expect the PE-backed IPOs to actually be significantly underpriced compared to their peers. Therefore for my alternative hypothesis I expect both these effects to positively impact BHRs for PE IPOs.

To test my main hypothesis I construct the regression equations presented below. Each regression is ordinary least squares (OLS) with [IPO] year fixed effects. For the relationship I am testing, for my first hypothesis, I include a dummy for the company in question having an IPO that is a PE-exit. I expect a negative coefficient to my PE indicator variable for my null hypothesis, and a positive coefficient for my alternative hypothesis. I expect revenue in general to have a positive relationship with BHR, so that higher revenue leads to higher stock price returns. I expect very high levels of debt to have a negative impact on BHRs, due to the bankruptcy distress costs that arise, and lower more sustainable levels of debt to have no significant relation affecting stock price returns. Summing up these two effects of debt levels on BHRs, I expect an overall negative and not highly significant relationship with my dependent variable. Book-to-market ratio is calculated using the book value of year t divided by the market capitalization of year t. Given that [the] B/M ratio is calculated using my dependent variable, there is reverse causality in the relation between these two variables. Therefore, I use a lagged variable for B/M, given that book valuation lags market valuation adjustments. The variable “IPO year” is a time (year) dummy, which is I use to remove any upward or downward effect to BHRs due to differences in years as a whole, time fixed effects (FE). Lastly, I assume heteroskedastic variances, where each company has a different variance, hence I use robust standard errors.

𝐵𝐻𝑅 1𝑦𝑒𝑎𝑟𝑖𝑡= 𝛽0+ 𝛽1𝑙𝑛(𝑟𝑒𝑣𝑒𝑛𝑢𝑒)𝑖𝑡+ 𝛽2𝐷𝑒𝑏𝑡(%)𝑖𝑡+ 𝛽3𝑙. 𝐵/𝑀𝑖𝑡+ 𝛽4𝑃𝐸𝑏𝑎𝑐𝑘𝑒𝑑𝑖𝑡+ 𝑎1𝐼𝑃𝑂𝑦𝑒𝑎𝑟𝑖+ 𝜀𝑖𝑡 (1)

𝐵𝐻𝑅 3𝑦𝑒𝑎𝑟𝑖𝑡= 𝛽0+ 𝛽1𝑙𝑛(𝑟𝑒𝑣𝑒𝑛𝑢𝑒)𝑖𝑡+ 𝛽2𝐷𝑒𝑏𝑡(%)𝑖𝑡+ 𝛽3𝑙. B/M𝑖𝑡+ 𝛽4𝑃𝐸𝑏𝑎𝑐𝑘𝑒𝑑𝑖𝑡+ 𝑎1𝐼𝑃𝑂𝑦𝑒𝑎𝑟𝑖+ 𝜀𝑖𝑡 (2)

𝐵𝐻𝑅 5𝑦𝑒𝑎𝑟𝑖𝑡= 𝛽0+ 𝛽1𝑙𝑛(𝑟𝑒𝑣𝑒𝑛𝑢𝑒)𝑖𝑡+ 𝛽2𝐷𝑒𝑏𝑡(%)𝑖𝑡+ 𝛽3𝑙. B/M𝑖𝑡+ 𝛽4𝑃𝐸𝑏𝑎𝑐𝑘𝑒𝑑𝑖𝑡+ 𝑎1𝐼𝑃𝑂𝑦𝑒𝑎𝑟𝑖+ 𝜀𝑖𝑡 (3)

In terms of sampling, given that the control sample is matched to my treatment sample, the differences found according to the hypotheses set, that distinguish the treatment

(23)

group to the control group, if statistically significant, are causal to the fact that the company is/was a PE portfolio company. For example, given that I use the complete sample of PE-exits and the complete sample of IPOs in Europe, both for the years 2005 to 2011, I expect any relationship found from my model to be causal relationship, that higher or lower BHRs are caused from my main explanatory variable, PE indicator dummy. I control for revenues and debt levels, both of which are important factors in valuing a company on the stock market, its stock price. Including year fixed effects in my regression, I therefore remove variation in stock price due to time differences, leaving any variation to be explained by my main explanatory variable; the company being a PE portfolio company. That means that either PE-backed IPOs are under- or over-priced compared to their peers. As explained in my literature review, these differences are causal due to specific significant differences that arise from the operation of these companies under PE ownership. I explain the main reasons for this causal relationship in the next section where I present the results of my analysis.

I further analyze the variables of interest in which a priori, during the PE holding period, cause PE portfolio companies to have significant differences to their peers. For my second hypothesis, I employ a probit differences-in-differences (DID) regression analysis where I use the PE-exit or IPO as my event. A DID analysis is a multidimensional regression, where differences before and after for both groups, and differences between the two groups are compared. This analysis does not just look at changes to variables before and after PE-exit, but it also benchmarks these changes to the control group which does not go through a PE-exit IPO. DID is a pure probit model where all explanatory variables are dummy variables. I employ this analysis onto revenue and EBITDA variables, as well as debt proportion, long-term debt growth and net debt growth as they evolve with time. In order to produce this analysis I construct a post-IPO period for all companies, which is equal to one if the year in question is post IPO. A PE-indicator dummy for the whole period and a post-IPO/PE-indicator interaction dummy variable which is equal to one if the company is an ex-PE company in the post-IPO period. This compares how each variable looks depending on each of the three given comparison variables/periods.

𝑙𝑛 (𝑟𝑒𝑣𝑒𝑛𝑢𝑒)𝑖= 𝛽0+ 𝛽1𝑃𝐸𝑏𝑎𝑐𝑘𝑒𝑑𝑖+ 𝛽2𝑃𝑜𝑠𝑡𝐼𝑃𝑂𝑖+ 𝑃𝑜𝑠𝑡𝐼𝑃𝑂 ∗ 𝑃𝐸𝑏𝑎𝑐𝑘𝑒𝑑𝑖+ 𝜀𝑖 (1)

𝐸𝐵𝐼𝑇𝐷𝐴𝑖= 𝛽0+ 𝛽1𝑃𝐸𝑏𝑎𝑐𝑘𝑒𝑑𝑖+ 𝛽2𝑃𝑜𝑠𝑡𝐼𝑃𝑂𝑖+ 𝑃𝑜𝑠𝑡𝐼𝑃𝑂 ∗ 𝑃𝐸𝑏𝑎𝑐𝑘𝑒𝑑𝑖+ 𝜀𝑖 (2)

𝐷𝑒𝑏𝑡(%)𝑖= 𝛽0+ 𝛽1𝑃𝐸𝑏𝑎𝑐𝑘𝑒𝑑𝑖+ 𝛽2𝑃𝑜𝑠𝑡𝐼𝑃𝑂𝑖+ 𝑃𝑜𝑠𝑡𝐼𝑃𝑂 ∗ 𝑃𝐸𝑏𝑎𝑐𝑘𝑒𝑑𝑖+ 𝜀𝑖 (3)

(24)

𝑁𝑒𝑡𝑑𝑒𝑏𝑡𝑔𝑟𝑜𝑤𝑡ℎ𝑖= 𝛽0+ 𝛽1𝑃𝐸𝑏𝑎𝑐𝑘𝑒𝑑𝑖+ 𝛽2𝑃𝑜𝑠𝑡𝐼𝑃𝑂𝑖+ 𝑃𝑜𝑠𝑡𝐼𝑃𝑂 ∗ 𝑃𝐸𝑏𝑎𝑐𝑘𝑒𝑑𝑖+ 𝜀𝑖 (5)

Given the research I present, I expect all dependent variables of my second analysis to be higher for PE portfolio companies. This translates into positive significant coefficients for the independent PE-dummy variable. This is, PE companies have higher revenues, EBITDAs, and debt levels than control companies before and after IPO. I expect a negative relationship with “PostIPO” and debt variables, as I expect all companies to decrease debt levels post-IPO, however I do not expect a significant relationship. I expect the variable “PostIPO” to have a positive relationship to both revenues and EBITDA, as I expect companies to grow in both those variables after they go public. The last variable, the interaction of the period variable “PostIPO” with the PE-dummy variable shows the DID outcome. Where the difference (after minus before) for the control group is subtracted from the same difference of the treatment (PE) group. This is the variable of interest of a DID analysis. Given my assumptions deduced for my first null hypothesis, I expect PE companies to perform worse than non-PE companies after the PE-exit (IPO) due to the departure of the PE partner, so in this case I expect a reversal. Alternatively, I expect the interaction variable to have a positive relationship with revenues and EBITDA, i.e. the PE company is superior due to the skills gained while under PE ownership. On the other hand I expect a reversal for all three debt variables after PE-exit (IPO) for both null and alternative first hypotheses. This is because, PE investors lever up portfolio companies while under their ownership, and I expect a gradual decrease of debt after PE-exit. Hence, I use net-debt growth and long-term debt growth for this test. I expect that growth in debt is decreasing or even negative post-IPO. The strength of a DID regression analysis is that any results that are proven to be significant, and especially for my variable of interest the interaction variable, mean the relation found is causal. This is due to the fact that DID regression analysis controls for many dimensions as well as benchmarks any differences to similar moves in the control group. In section IV, I explain in more detail with means of tables 4 and 5 how DID works and interpret my results.

(25)

IV. Results

Table 3. Main Regression - Share price BHRs regressed with PE indicator & controls

Buy-and-hold returns and private equity indicator dummy, revenue, proportion of debt as of total capitalization and book-to-market ratio. The table shows results from cross-sectional and longitudinal (panel data) regressions of buy-and-hold returns for companies with IPO years 2005 to 2011 on private equity indicator dummy, revenue per year, debt percentage and book-to-market per year. ‘‘Private Equity indicator dummy’’ is a dummy variable equal to “1” if the company (IPO) is a private equity-exit and equal to zero otherwise. ‘‘Revenue’’ is company i’s revenue on year t. ‘‘Debt (%) of total capitalization’’ is the proportion of debt as of total capitalization of company i in year t. ‘‘Book-to-market ratio’’ is company i’s book-to-market ratio on year t-1. A log transformation is applied to revenue. A 1 year lagged B/M (t-1) to BHR year (t) is used. The firm characteristics are measured at each year post-IPO. Buy-and-hold returns for 1, 3 and 5 years respectively are calculated using share price returns with IPO quarter closing share price as the buying price. Yearly fixed effects are included by controlling for each IPO year and is indicated by “Year FE”. All independent variables are winsorized to remove outliers. All values presented are interpreted as percentage (%) changes onto BHRs. Standard errors are reported in parentheses. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.

(1) (2) (3)

Dept. Variables: BHR 1-Year BHR 3-Year BHR 5-Year Private Equity indicator dummy 12.00*** 12.70** 15.00*

(3.74) (6.14) (8.07)

Revenue 1.10 3.50*** 8.00***

(0.90) (1.10) (1.20) Debt (%) of total capitalization -11.50*** -15.80*** -21.00***

(2.40) (0.60) (0.60)

Book-to-market ratio -2.40 15.25 -18.38

(0.50) (16.75) (16.20)

Constant 114.50*** 59.30*** 48.30***

(6.10) (7.30) (10.10)

Year FE Yes Yes Yes

Observations 969 964 860

R-squared 12.0 8.0 10.3

To test my first main hypothesis I run three regressions on BHRs for one, three and five years. I find very strong results rejecting my null hypothesis and not rejecting my alternative hypothesis. My main hypothesis, a priori, is that PE investors are shroud bargainers and are able to get the lowest possible price for their portfolio companies when “entering”, as well as experts in corporate restructuring, so I expect them to be able to extract the highest possible price when “exiting”. With high statistical significance (at the 1% level) a company that is in my treatment group has a 12% higher one year buy-and-hold returns post-IPO than companies which have not been PE-backed. At the 5% significance

(26)

level, I find that companies which undergo an IPO and are PE portfolio companies experience 12.7% higher BHRs over three years, and 15% higher BHRs over a five year holding period post IPO. This result contrasts the assumptions deduced about PE investors from the literature I present, being superior at extracting the best possible price out of their investments. However this result is perfectly in line with my alternative hypothesis, that companies that used to be PE portfolio companies, are either highly underpriced on their IPO or gain skills and expertise that are unique to them and are able to keep performing better than their counterparts even after the PE-exit/post-IPO. I find significant positive relationship between revenues and BHRs for three and five years holding periods, however, I do not find a significant positive relationship for one year buy-and-hold return. I attribute this to the high variability in revenue values on the IPO year. In the three and five years BHRs, revenue has a positive and at the 1% level, highly statistically significant impact onto the BHRs. For every 1 log unit increase in a company’s revenue, there is on average of 3.5% increase in BHRs for three years and an 8% increase in five year BHRs. This is as expected given higher revenues are a good indication of higher free cash flows, and a stock price is the present value of a company’s future cash flows. My second control variable, “debt proportion of total capitalization” I find a highly significant negative relationship for all buy-and-hold years. For every 1% increase in a company’s debt proportion I find a significant decrease of BHRs ranging from 11.5% for one year to 15.8% decrease in BHRs for three years and about 20% decrease in five year BHRs. This result is a surprise as a priori, I do not expect such a strong impact from debt onto the short-term BHRs I construct as dependent variables. This could be attributed to the fact that high outliers in my sample, such as companies with extreme levels of debt, companies close to bankruptcy, also experience extreme falls in stock price valuation for the same reason at the same period. This is why I find such a strong relationship between debt proportion and BHRs. Book-to-market ratio is not significant at explaining variation in BHRs.

The main alternative hypothesis that explains the result of highly significant positive coefficients for my PE dummy variable says that instead of focusing on the company itself and whether it being a PE portfolio company makes it overvalued because of intrinsic performance related variables or whether it performs better post-IPO compared to its counterparts for the same reasons, it focuses instead on the IPO valuation itself. A study published by Liu and Ritter (2011) that focuses on IPO underpricing, finds a significant positive impact on IPO underpricing if the IPO is done by VC investors and even more if these IPOs have “all-star” analyst coverage. That is, IPOs of VC-backed companies are

(27)

significantly undervalued compared to their counterparts. They attribute this to the impact of analyst coverage and oligopolistic market power. In my research I do not distinguish between VC or buyout companies (BO) however, both these subcategories are part of the category called Private Equity. Hence, these effects found by Liu and Ritter (2011) may apply to whole PE industry.

Table 4. Differences-in-Differences Probit regressions

Revenue, EBITDA, debt proportion, long-term debt growth, net debt growth and PE indicator dummy, post-IPO period dummy and post-post-IPO/PE interaction. The table shows results for differences-in-differences regressions of revenue, EBITDA, debt proportion, long-term debt growth, net debt growth for companies for IPO years 2005 to 2011 on PE indicator dummy, post-IPO period dummy and post-IPO/PE interaction. ‘‘Revenue’’ is company i’s revenue on year t. “EBITDA” is earning before interest tax depreciation amortization for company i at year t. ‘‘Debt proportion’’ is the proportion of debt of total capitalization of company i in year t. “Long-term debt growth” is the growth rate of long-term debt of company i at year t, calculated using net debt in year t-1. “Net debt growth” is the growth rate of company i at yeat t, calculated using net debt in year t-1.‘‘Private Equity indicator dummy’’ is a dummy variable equal to “1” if the company (IPO) is a private equity-exit and equal to zero otherwise. “Post-IPO” is a dummy variable equal to “1” if the dependent variable analyzed is in the periods after IPO and equal to zero otherwise. “Post-IPO/PE interaction” is dummy variable equal to “1” if the company is PE-backed and the period is post-IPO and equal to zero otherwise. All dependent variables are winsorized to remove outliers. Standard errors are reported in parentheses. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.

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

Dept. Variables: Revenue EBITDA

Debt proportion Long-term debt growth Net debt growth Private Equity indicator 60.1*** 255.50*** 2.00 -18.60 -39.30***

(10.00) (25.79) (2.10) (118.50) (6.16) Post-IPO 52.80*** 39.35** -1.10 -81.10 -35.82*** (6.11) (15.54) (1.27) (71.00) (3.48) Post-IPO PE interaction 11.50 -21.04 -4.45* 154.10 27.68*** (12.60) (32.45) (2.60) (138.60) (7.21) Constant 352.60*** 119.10*** 38.80*** 296.40*** 51.00*** (48.90) (12.49) (0.019) (61.4) (3.00) Observations 7,844 8,242 7,889 4,744 7,100 R-squared 2.90 3.00 0.00 0.00 2.00

To test my second hypothesis, I employ a differences-in-differences (DID) probit regression analysis. The aim of this test is to look if there is a trend continuation or reversal from the pre-IPO period to the post-IPO period, i.e. PE-exit, focusing on my treatment group. In regression (1), results show that if a company is a PE portfolio company, it has 60% higher

(28)

revenue than control companies over the whole period (pre- and post-IPO). Post-IPO revenue is 52.8% higher than pre-IPO revenue for all companies. Hence, companies experience a jump in revenue over the five years post-IPO. In order to get a better understanding of the interaction variable, I also provide table 5. below, to breakdown how my DID analysis works. The Post-IPO/PE interaction variable shows (insignificant at 10%) that PE companies experience 11.5% increase to the difference they have with the control group in the pre-IPO/exit period. This finding adds to my first analysis which I perform for my first hypothesis, where I find that given BHRs, PE-backed IPOs are more undervalued. As explained in the previous part of this section, my results are either attributed to the fact the PE IPOs are underpriced more than their controls or because of intrinsic company performance variables performing better than controls post-IPO. Hence, I find persistence, where revenues increase more post-IPO for ex-PE companies, meaning that there is a continuation in better performance as a priori.

Due to the peculiarity of a large portion of EBITDAs being negative, I use absolute value of EBITDA in thousands of US dollars. PE portfolio companies have $255 thousand higher EBITDA than controls over the whole sample period. All companies have on average $39 thousand higher EBITDA after IPO compared to pre-IPO. Taking into account the change of the benchmark (control group), EBITDA actually decreases by $21 thousand post IPO for ex-portfolio companies. Whereas taking the benchmark into account, revenue increases post IPO for ex-PE companies, EBITDA decreases. Ex-PE portfolio companies therefore experience a deterioration to their profitability especially when summing the two variables, an increase in revenues post IPO and a (benchmarked) decrease in EBITDA leads to an even bigger deterioration of profit margins. Hence, PE-exit does have a negative effect when it comes to keeping the company efficient and lean.

In regression (3) I find that PE companies have only 2% higher debt proportion as of total capitalization. Companies have 1.1% lower debt post IPO as compared to the whole period. The interesting result is the significant (at the 10% level) decrease in debt proportion as of total capitalization for PE-backed companies post IPO. Benchmarking to the control group, I find that ex-PE portfolio companies experience a decrease of 4.45% post-IPO which backs my second hypothesis assumption, a reversal in debt variables. In regression (4) however, PE companies experience 18.6% lower long-term debt growth over the whole sample period. All companies experience 81% lower long-term debt growth. PE companies however, controlling for non-PE companies experience a 154% increase in their long-term debt growth. This result contrasts any assumption made a priori, however this result is

(29)

statistically insignificant. Lastly, for regression (5) I test net-debt growth. Net-debt is the result after company’s cash holding are subtracted from total debt. I find that ex-PE portfolio companies experience 39% lower net-debt growth over the whole sample period, with significant results. Given that the results in regression (4) are insignificant at explaining any difference between the two groups, I attribute the significance to regression (5) to the cash holdings variable which is a component of net-debt. PE portfolio companies might be increasing cash holdings at the same time as increasing long-term debt, and the effect of increasing cash holdings makes the results for PE indicator highly significant and negative. In the period post-IPO, adding to the decrease in long term debt as assumed, the increase in cash holdings, I get a very significant negative coefficient of 35.8% decrease in net debt. And lastly, controlling for changes in non-PE portfolio companies, ex-PE companies experience a significant increase of 27% to the growth of net debt. This is in line with the less significant increase in long-term debt in regression (4). All in all, given my second hypothesis, I expect a decrease or reversal in all debt variables, which is the case with the significant decrease in debt proportion of ex-PE-backed companies post-IPO. Hence, the result of regression (3) rejects my null (second) hypothesis, that there is no significant change in debt proportion post PE-exit. There is no significant impact on long-term debt post-IPO/PE-exit, which is explained by the fact that long-term debt is by nature long-term, and the period I am testing is considered short-term. Hence, there is no decrease in long-term debt. Lastly, in regression (5), ex-PE companies’ net debt grows 27%, controlling for non-PE companies and I attribute this result to a probable decrease in cash holding post-IPO, but cannot prove it.

(30)

Table 5. Breakdown of Differences-in-Differences analysis

Revenue, EBITDA, debt proportion, long-term debt growth, net debt growth and difference (treatment-control) before IPO, difference (treatment-control) after IPO, differences-in-differences result. The table shows results for differences-in-differences regressions of revenue, EBITDA, debt proportion, long-term debt growth, net debt growth for companies for IPO years 2005 to 2011 broken down into differences before and after IPO, as well as the difference in those differences values. ‘‘Revenue’’ is company i’s revenue on year t. “EBITDA” is earning before interest tax depreciation amortization for company i at year t. ‘‘Debt proportion’’ is the proportion of debt of total capitalization of company i in year t. “Long-term debt growth” is the growth rate of long-term debt of company i at year t, calculated using net debt in year t-1. “Net debt growth” is the growth rate of company i at yeat t, calculated using net debt in year t-1. The values shown under “difference-in-difference” are the same as the value of post-IPO/PE interaction variable’s values in table 4. All dependent variables are winsorized to remove outliers. Standard errors are reported in parentheses. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.

Before After

Difference-in- Difference Independent Variables:

PE-backed Control Difference

(1) Revenue 4.13 3.53 0.60*** 4.77 4.05 0.72*** 0.12 (0.10) (0.08) (0.12) (2) EBITDA 375.14 119.20 255.95*** 393.28 158.58 234.70*** -21.24 (25.78) (19.70) (32.45) (3) Debt proportion 0.39 0.38 0.02 0.33 0.36 -0.03 -0.05* (0.02) (0.02) (0.03)

(4) Long-term debt growth 4.74 4.63 0.11 5.18 3.43 1.76 1.65

(2.57) (1.56) (3.01)

(5) Net debt growth 1.17 5.10 -3.93*** 0.36 1.52 -1.16*** 2.77***

Private equity-exit through the IPO : is there money on the table? : a sample of European IPOs