Abstract
Although buyout investments represent a considerable proportion of private equity volume, so far little research has been done on the exit strategies of buyout investments. This article takes a step in this direction by investigating buyouts in more detail focusing particularly on the divestment process. Since exiting enables the realization of returns and is thus the most important factor for private equity investors, it is important to understand the motivation behind and the determinants influencing this decision. The authors analyze three main exit routes for exiting buyout investments: initial public offerings (IPO), sales and write-offs, using a unique data set for U.S. and European buyout transactions. They examine the determinants influencing the choice of an exit channel by employing a multinomial logit model. The results strongly support the view that private equity investors write-off investments that turn out to be non-performing early, showing their ability to filter out bad investments. They also analyze how the internal rate of return (IRR) influences which exit route is chosen. The results show that only the most profitable ventures are taken public. The results have implications for exiting buyout investments during financial crises.
TOPICS: Private equity, developed, financial crises and financial market history, performance measurement
Although buyout investments represent a considerable proportion of the total private equity1 volume, so far little research has been conducted on the exit routes of buyout investments. This article takes a step in that direction, investigating the strategy of buyouts in more detail, and tries to shed light on the divestment process. Since exiting enables the realization of returns and is thus the most important aspect for private equity investors, it is important to understand the motivation behind and the determinants influencing this decision.
Using a multinomial logit model, we analyze factors affecting the choice of an exit vehicle. The three main exit routes are initial public offerings (IPOs), sales, and write-offs. We find strong support for a signaling effect, implying that private equity investors tend to quickly write off investments that turn out to be non-performing. Through this strategy they show their ability to differentiate between investments that are worth supporting further and those that are not. Although living-dead investments may exist, this shows that holding non-performing investments with the hope of a turnaround is not a common strategy pursued by sophisticated investors. In addition, we find evidence that exit decisions for buyouts are driven by market sentiment and thus seem to be cyclical, as are other investments. On the contrary, we do not find support for market characteristics influencing the occurrence of IPOs. Furthermore, we evaluate the influence of a chosen exit vehicle on the internal rate of return (IRR) of the investment by using a multivariate regression model. There we find support for the common argument that only exceptional firms are taken public. Furthermore, we show that in periods of economic boom returns are extraordinarily higher for IPOs than for sales.
Our prior results have implications for the choice of exit vehicles in the buyout market during the financial crisis. Our analysis suggests that both IPOs and trade sales are adversely affected during recessions, particularly when the credit boom comes to an end. Write-offs, however, are increasingly more likely to occur, for the same reasons why IPOs and trade sales decline, i.e., the lack of capital from investors and financing banks. The decline in IPO and M&A activity in the aftermath of the financial crisis gives supporting evidence for this.
This article is organized as follows. The next section provides a theoretical overview of the investment cycle and the importance of exit strategies within. Theoretical details are supported by interviews we conducted with private equity fund managers from three different buyout funds, each with a different investment focus. We then review the literature and provide hypotheses regarding the influence of specific variables on the choice of an exit vehicle. We next describe the data, and discuss results from our multivariate analysis. The final section concludes.
THE IMPORTANCE OF EXIT STRATEGIES
Characteristics of Buyout Investments
Buyouts are private equity investments in established firms that are in a mature development stage. In general, these companies are characterized by sustainable and strong cash flows and the ability to generate high financial surplus and consequently free available liquidity. Due to their characteristics, special types of exits are more common for buyouts than for other private equity segments, like early-stage ventures, as will be shown later.
Usually, buyouts are classified according to their new owner structure—for instance, management buyouts/management buy-ins (MBOs/MBIs), employee buyouts (EBOs), and institutional buyouts (IBOs). If investments are highly leveraged, they are also classified in accordance with their financial structure as leveraged buyouts (LBOs).
Reasons for buyouts are multifaceted, as Berger [1993] points out. The ones that are mentioned regularly in the literature are outsourcing business units in order to concentrate on core businesses, following up on regulations (especially in family-owned firms), selling loss-generating units, and seeking external funds. Also, privatizations play a very important role in some countries, as shown by Wieser, Wright, and Robbie [1997].
The provision of equity capital is usually by external institutional investors. As presented by Das, Jagannathan, and Sarin [2002], besides intermediating financial support, private equity investors also provide specialized know-how, coordinate the process of leverage financing, place incentives for management, and monitor the firm’s development, especially with regard to cash flow patterns. The latter is essential because of the usual high leverage and risk associated with it. For their venture, they receive the majority of shares and voting rights of the company and aim to control the company through the supervisory board. However, in general private equity investors do not intervene with the operational aspects of the business.
On average the duration of buyout investments spans three to five years. During the investment’s lifetime, investors generally do not expect regular dividend payments. They are rather interested in achieving high capital returns at the end of the investment period. Those returns, realized at the date of exit, have a huge impact on fund reputation and the ability to raise follow on funds, as shown later.
The Life Cycle of Buyout Investments
As shown by Lerner [2000] and Daniels [2004], the cycle of private equity investments can be divided into four phases: fundraising, investment, value-adding, and divestment. All phases and their determinants are critical to the success of an investment. However, the exit phase is mostly regarded as the key driver, as pointed out by Gompers and Lerner [2000a]: “The need to ultimately exit investments influences every aspect of the venture capital cycle, from the availability to raise capital to the types of investments that are made.”
Fundraising
To start an investment, capital has to be raised by choosing appropriate investors. As described by Daniels [2004], these investors should have sufficient capital levels, a long-term investment horizon, and a suitable risk preference. On the other side, investors also have special criteria for selecting a fund, as pointed out by Black and Gilson [1998]. These criteria are based on the fund’s underlying strategy as well as the fund management’s track record. According to Yrkkö et al. [2001], among others, the latter is regarded as the key factor for the acquisition of capital. Thus, successful exits in the past seem to have strong influence on the ability to raise follow on funds and thus are of major importance.
Investment phase
Appropriate investment targets are selected during the second phase. Private equity firms with a good reputation and strong networks in the financial sector are mostly favored for large and high-quality portfolios of target companies. They use the process of screening, due diligence, and valuation to select appropriate investments. Shortly thereafter they structure the deal, whereby they carefully consider exit opportunities as well. The importance of exit considerations during this phase is also described by Schwienbacher [2002] and Daniels [2004]. Additionally, this approach is supported by comments we obtained from interviews with private equity fund managers.
Value-adding phase
The value-adding phase is a key process in a private equity investment. During this phase, private equity fund managers contribute to the success of an investment by providing know-how and offering informal advice and access to their network. They also provide reputational capital that gives their target companies credibility (Black and Gilson [1998]). Furthermore, Cumming and MacIntosh [2002] show that they also participate in strategic decisions that enable them to restructure and monitor their target companies. All in all, this underscores the importance of management know-how and why investors pay special attention to the private equity fund’s track record in the selection processes.
Divestment phase
As shown above, private equity funds consider exit strategies quite early in the investment process since it allows the liquidation of tied-up funds and thus the realization of returns for investors. This argument is supported by Black and Gilson [1998], who point out that exits play a central role in the fund’s viability to capital providers. There are several possibilities for how to exit an investment, but the decision is often quite a complex matter because it depends on a variety of factors that can change over time. Besides the profit that can be realized at the date of exit, factors like the company’s size, management’s goals, and the current market environment have to be considered as well. As shown by Yrkkö et al. [2001], who analyzed exit opportunities in Finland, market characteristics play an important role, as do contractual agreements (Cumming [2002]).
Exit Channels
The divestment phase is crucial for the overall performance of an investment, as noted above. Besides the systematic factors already mentioned, strategic decisions like the timing and the type of exit vehicle have to be considered carefully. Following is a brief overview of some of the different exit opportunities. Due to restrictions in the empirical data and the fact that numerous subtypes of these categories can be considered, only the most important ones—IPOs, sales, and write-offs—are discussed.
IPOs
Schwienbacher [2002] pointed out that IPOs are an exit channel for highly profitable portfolio companies and thus are supported generally in the literature. Such companies are characterized by a convincing equity story as well as high growth prospects. This indicates that for buyouts in a mature development stage, the probability of exiting via IPO generally should be less than for other private equity investments. This assumption is supported by Das, Jagannathan, and Sarin [2001], who discovered that the probability of an IPO decreases by moving from early to later-stage investments, with buyouts the least probable and thus more likely to be sold. A sub-category of buyout-backed IPOs is reverse LBOs,2 which were analyzed in more detail by Cao and Lerner [2006].
Sales
Sales can be subdivided into trade sales (also referred to as acquisitions), secondary sales, and buybacks. As shown by Cumming and MacIntosh [2002], trade sales, where the portfolio company is sold to a strategic investor, is the most prevalent exit vehicle and also the most profitable of the sales types. Secondary sales (sales to other institutional investors) as well as buybacks (repurchases by the old owner or management) are mostly associated with lower returns and are less preferred, according to Daniels [2004] and Natter [2003]. However, in recent years the preference for secondary buyouts increased significantly, as pointed out by Wright et al. [2006]. The general advantages of sales that are often mentioned in literature, such as by Daniels [2004], are the fast exit opportunities, fewer restrictions compared to IPOs, and the possibility of influencing the investment according to the needs of strategic investors.
Write-offs
Private equity firms generally try to minimize their portfolio failure ratio and its negative effect on performance and reputation. Nevertheless, underperforming investments occur and sooner or later have to be written down or written off. Hence, this exit channel has to be regarded as a last option. However, the interesting question is how fund managers treat these investments and how long it takes until they decide to drop them out of their portfolio. Cumming and MacIntosh [2002] discovered interesting differences between U.S. and Canadian ventures with regard to the handling of those investments that are illustrated in more detail below.
LITERATURE REVIEW AND HYPOTHESES
Overview
Despite their increasing importance, few analyses have focused on the topic of buyouts. And even existing studies have only considered certain aspects of this segment, leaving much space for further research. Many of these studies investigated agency theoretical implications. For example, Thompson and Wright [1991] analyzed monitoring and incentive devices using a large sample of U.K. buyouts. Their results showed that the choice of control strategies is clearly influenced by financing arrangements. Nikoskelainen and Wright [2005] revealed that value increases and return characteristics in LBOs are a result of corporate governance mechanisms, and thus they found evidence for the free cash flow theory developed by Jensen [1986].3 Their study was based on a set of 321 hand-collected U.K. LBOs that were exited between 1995 and 2004. The authors found a positive relationship between the value increase and the management ownership, the number of participants in the equity syndicate, and the leverage and debt coverage.
Other studies, like Groh and Gottschalg [2006], focused on the investigation of buyout performance. By analyzing the risk-adjusted performance of U.S. buyouts, they showed that this strategy clearly outperformed the market benchmark. Similar results were also found by Ick [2006], who investigated the risk and return relationship of private equity relative to public equity. He found private equity returns to be very heterogeneous across the different stages, whereby later-stage investments achieved higher risk adjusted returns.
Furthermore, as stated by Giot and Schwienbacher [2005], less research has focused on the determinants of various exit vehicle choice for private equity investments. Among others, Das, Jagannathan, and Sarin [2002] examined the exits of U.S. venture and buyout funds by estimating the probability of various exit routes. They found high cross-sectional variation in the probability of an exit across different stages, industries, the financing amount, and existing market conditions. For the buyout stage, an exit via sale was the highest probability.
Investment Duration and Exit Strategy
There seems to be a connection between the duration of an investment and different exit strategies. One strategy in this context is called the “grandstanding hypothesis,” introduced by Gompers [1996]. This strategy indicates that younger venture capital funds exit their investment prematurely, and preferably via IPO, in order to build reputation. Among others, Yrkkö et al. [2001], who studied the exit behavior of venture capital funds in Finland and analyzed factors influencing the length of the investment, also supported this hypothesis.
Another exit strategy related to the holding period was presented by Kreuter, Gottschalg, and Zollo [2005], who distinguished between “buy-and-flip” and “buy-and-grow” approaches. The first argues that most of the value creation in a buyout occurs up front. Thus, reselling the portfolio company rapidly leads to a maximization of realized returns. The second approach suggests that higher returns are achieved mainly by adding real value to the investment target. That is only possible over time, but exiting later results in higher proceeds.
A recent study presented by Cao and Lerner [2006] gave ambiguous results for the theory of value creation in investments. By investigating the performance of U.S. reverse LBOs between 1980 and 2002, they found that portfolio companies kept for shorter than the median investment duration performed slightly better than those held longer. Additionally, they investigated investments that were exited within one year, referred to as quick flips, and found that they underperformed the market, whereas other reverse LBOs outperformed the market. However, this difference was not statistically significant, providing weak support for their thesis.
Cumming and MacIntosh [2002] supported the strategy, claiming that adding value is only possible over time. They argued that due to their sophistication, private equity investors can clearly distinguish between good and bad investments, adding value only to good ones. Hence, the duration of an investment can be interpreted as a signal of the firm’s quality. Therefore they suggested that the longer the holding period, the higher the probability of IPOs (because an IPO is regarded as an exit channel for high flyers), followed by secondary sales, acquisitions, buybacks, and write-offs. Most of their results did not show strong evidence of their hypothesis. But they found strong support for the signaling effect with regard to write-offs. The results showed that a shorter holding period increases the probability of an investment being written off. This is in line with their argument that private equity (PE) investors will not stick to portfolio companies whose quality is too low, as it would signal their inability to differentiate between good and bad investments. Thereby Cumming and MacIntosh also implicitly recognized the significance of neglecting living-dead4 investments. They argued that a strategy supporting the maintenance of living-dead investments in the portfolio would in the long run signal an inability by the fund’s management to create value through active participation. Nevertheless, few living-dead investments reveal interesting differences in the regional treatment of such investments. A write-off seems to happen more frequently in the United States than in Canada, where some poorly performing investments are exited through buybacks rather than write-offs.
Findings that write-offs occur quickly were also provided by Cumming [2002]. He analyzed the probability of different exit vehicles of European venture capital funds dependent on their firm characteristics and their contractual agreements with the venture capital fund. With regard to the holding period, he found that the probability of a write-off is highest for the short term. He explained that disadvantageous information is generally exposed rapidly during the investment process.
In line with the value-adding thesis, different studies show that investments not worthy of further capital commitments or efforts are quickly written off instead of being held as living-dead investments. This process signals the PE investor’s ability to quickly identify investments that turn out to be non-performing. Based on these findings and the information we obtained during the interviews, we therefore state:
Hypothesis 1: Write-offs are more likely in shorter holding periods than IPOs and sales.
The Impact of Market Conditions
The general market environment seems to have great influence on the decision to exit private equity investments. Several studies analyzed this assumption by using different approaches and proxies, like stock valuation levels, the market’s overall liquidity, or other variables that proxy boom periods.
Giot and Schwienbacher [2005] argued that the decision to exit an investment has two dimensions, the type and timing of exit. They used competing models for the timing of exits and found that each type of exit has its own dynamics. This is also the case for different industry categories, with Biotechnology and Internet firms the fastest in exiting via IPO and Internet firms being written off most quickly. They also analyzed the effect of the bubble period and found that exits then speeded up.
Cumming and MacIntosh [2002] analyzed determinants influencing the decision on full and partial exits of venture capital investments in the United States and Canada. They investigated the full range of exit vehicles, including IPOs, acquisitions, secondary sales, buybacks, and write-offs. They found market conditions to clearly affect the extent of IPO exits in the U.S. sample.
Ritter and Welch [2002] reviewed the theory and results of economic research on IPO activity. They analyzed in detail the determinants for going public, reasons for the phenomenon of underpricing, and the long-run performance of IPOs. In general, they argued that the asymmetric information theory does not explain the high volatility of IPO activity. Instead non-rational explanations seem to be more appropriate. They found the most interesting unanswered question to be why, after a stock market crash, a downward adjustment in the number of IPO issues takes place instead of an adjustment in prices. Overall, they found stock market conditions to be the predominant factor in decisions to go public.
Gompers and Lerner [2000a] analyzed the private and public financings of venture-backed, privately held biotechnology firms. They showed that venture capitalists take companies public when valuation levels are high (market peaks) and otherwise rely on private equity financing. They also found that experienced venture capitalists are more successful in timing their exits.
As shown by various studies using different approaches, the stock market environment seems to influence PE investments and thus their exit opportunities. In accordance with general findings that a favorable stock market climate drives exit decisions towards IPOs, we suggest the following:
Hypothesis 2: A favorable stock market climate increases the probability of an exit via IPO.
The Influence of Market Characteristics
As mentioned above, the divestment phase has a high impact on the earlier stages of the investment life cycle. Good and stable exit opportunities are therefore crucial to exit successfully and to build reputation, as pointed out by Lerner [2000]. Still, reality does not always meet these requirements. Additionally, there are differences between various regional public markets that influence exit decisions.
Yrkkö et al. [2001] analyzed the attractiveness of the Finnish market for private equity exits. Besides stock market conditions that are of particular interest for younger venture capital funds, they also analyzed the M&A market and the correlation of both markets with each other. They found the M&A market to be quite active and stable, thus enabling constant exit opportunities for PE investors. On the contrary, the IPO market seems to be characterized by high volatility, making an IPO rather unattractive. With regard to the interaction of both markets with each other, they found that they are imperfect substitutes and that the M&A market lags behind the movement of the stock market.
Cumming and MacIntosh [2002], who analyzed partial and full exits in the U.S. and Canadian markets, discovered that regulatory differences have an influence on the frequency of using particular exit channels. In detail they detected that Canadian venture capitalists use IPOs less frequently than do U.S. investors.
Manigart et al. [2002] analyzed determinants of venture capital returns based on a five-country study. They argued that each market has its own characteristics and therefore impacts the venture capital business in unique ways. Special attention was given to the difference between Anglo-American and European countries, as the former are generally regarded as having more developed markets. Their results supported this argument, as they showed that required returns (implying a greater involvement of venture capitalists) in the U.S. and U.K. are higher than those in other European countries.
Cumming [2002] analyzed the impact of contractual terms on the choice of exit routes. By analyzing a hand-collected sample of European venture capital funds, he revealed that the probability of an IPO increases by using common equity as well as a greater number of incentive contingencies in contracts. As European VCs, because of their market structure (shown in the German market by Black and Gilson [1998]), are more likely to use covenants with veto and control rights, an IPO should be less likely in those markets. For a detailed review of the differences between European and U.S. IPO markets, see also Ritter [2003].
Schwienbacher [2002] analyzed differences bet-ween the European and U.S. venture capital markets. Despite many similarities, he also found important exit-stage differences. He revealed that European investors face less-liquid markets with regard to human resources (e.g., the possibility to replace key employees), as well as fewer exit opportunities, than U.S. investors. Hence, these factors make an exit more complicated.
Due to the fact that the U.S stock market is more developed, this article argues:
Hypothesis 3: In the U.S. market, the probability for a buyout-backed IPO is higher than in European stock markets.
IPO as an Exit Channel for High Flyers
As pointed out by Giot and Schwienbacher [2005], in the academic literature IPOs are widely regarded as the most profitable exit route among private equity investments, although trade sales are the most common. They modeled the time to different exits via a competing risk model and showed that the probability of an exit via an IPO is very high at the beginning of the measurement period and decreases rapidly as time goes by, whereas the probability of a trade sale exit is more stable over time. Therefore they conclude that there is a pecking order of exits whereby investors prefer an IPO, followed by a trade sale and, lastly, a write-off.
Schwienbacher [2002] also argued that the increasing profitability of an IPO is due to selection bias: only extraordinary firms are taken public. In a questionnaire sent to venture capitalists in six European countries, they asked for the degree of influence various factors (e.g., quality of the management, expected future profitability) have on the decision to exit via IPO or sale. Their results showed that respondents valued these factors more for IPOs, strengthening their thesis that only outstanding firms are taken public.
Cumming and MacIntosh [2003] analyzed determinants influencing the decision of full and partial exits in Canada and the United States for five different exit vehicles. They showed that exits via initial public offerings or acquisitions are the most profitable, followed by secondary sales, buybacks, and finally write-offs for the lowest-quality firms. However, they found differences between full and partial exits with regard to the average annual return and the variance of those returns.
Furthermore Fenn, Liang, and Prowse [1997] argued that investments are lower in periods with high capital commitment returns.5 They found a possible explanation for poor investing outcomes coexisting with favorable exit opportunities. Due to high valuation levels during boom periods and the possibility of achieving higher returns for earlier ventures, capital commitments may be triggered. They referred to a study performed by Venture Economics in 1970 and 1982 showing that IPOs clearly outperformed sales, with realized gains five times higher.
Nikoskelainen and Wright [2005] examined the IRR of enterprise value and that of invested equity for 321 U.K. buyouts that were exited between 1995 and 2004. In detail, they analyzed the relation between the corporate governance structure and the likelihood of a positive return at the realization date of the investment and revealed that IPOs clearly outperformed trade sales and secondary buyouts.
Because initial public offerings are mainly regarded as an exit option for highly profitable ventures, this article suggests:
Hypothesis 4: An exit via IPO is positively associated with realized returns.
DATA
Overview of Dataset
The data for this article was obtained from CEPRES (Center of Private Equity Research), one of the leading companies offering research-based risk management and information services on the PE industry. The data is derived from the records of CEPRES’s Private Equity Analyzer, a unique online database and analyzing system that is accessible only by CEPRES members. The CEPRES Members Club is the core of a community of private equity funds that provide anonymous data on their private equity transactions and, in exchange, receive exclusive benchmarking services and transparency about their own strengths and challenges. Since the information is obtained anonymously to meet the high confidentiality requirements of the private equity industry, and since the Private Equity Analyzer system aggregates the data in a way that no third parties are ever able to identify individual funds’ or managers’ performances, the data providers do not have an incentive for a positive reporting bias. Moreover, the manager has a strong incentive for a correct reporting, since skewed data would lead to an inability to use the Private Equity Analyzer system appropriately in order to receive real transparency about his transaction characteristics. The database contains precise facts about more than 24,000 worldwide investments ranging from early to later stage. One major advantage of this data is that it incorporates detailed cash flow information on an investment level instead of an aggregated fund level. Cash flows are reported on a gross basis—i.e., unadjusted for management fees, carried interest, and other fund related costs—and are thus unbiased by any externalities. The detailed information about the amount and date of all cash flows to and from the PE investments enables further accuracy in measuring the IRR per investment. Because those IRRs are then reported on a gross basis, they are quite high compared to IRRs reported by other funds. According to Schmidt, Nowak, and Knigge [2004], who performed a separate analysis of 80 private equity funds (those raised in vintage years between 1971 and 1998 and for which net and gross information were available in CEPRES), net IRRs were on average 45% lower than gross IRRs.
This article focuses on exits of buyout investments. We therefore restrict our datasets to buyouts (i.e., MBOs/MBIs and LBOs) where information about the type of exit was given. We concentrate on liquidated investments only and, hence, only include fully realized investments and investments that had to be written off.6
In addition, certain macroeconomic variables from external sources are incorporated and merged with the CEPRES data to perform the analysis. To proxy the public and private market environment, we use information of IPO issues and M&A deals we obtained from Thomson Research. To proxy the economic activity, we further use the CFNAI (Chicago Fed National Activity Index) and the EuroCOIN Index. Both indices are based on an extension of the methodology used to construct the original Stock-Watson XCI.7 The CFNAI is a weighted average of 85 existing monthly indicators of national economic activity in the United States. It is constructed to have an average value of zero, and thus a positive index corresponds to growth above trend and a negative index corresponds to growth below trend. The EuroCOIN Index, provided by the Centre of Economic Policy Research (CEPR), is the leading real-time business cycle indicator for the Euro area. The indicator provides an estimate of the monthly growth of Euro area GDP after the removal of measurement errors and seasonal and other short-run fluctuations. As reported by CEPR, the quarterly growth rate of the GDP averaged 0.59 over the sample period 1988–2003 (Exhibit 1). Thus, a value of the EuroCOIN Index exceeding 0.59 indicates a growth above trend, and an index value below 0.59 indicates, if positive, a growth below trend. Similar to earlier research—e.g., Steffen [2007]—we define a boom period as both indices being above their trend for at least four consecutive quarters. As a result, we found the following periods to be boom phases: second-quarter 1994 through fourth-quarter 1994, second-quarter 1997 through first-quarter 1998, and second-quarter 1999 through first-quarter 2000.
Exhibit 1
EuroCOIN Index
The EuroCOIN Index, provided by the Centre of Economic Policy Research (CEPR), is the leading real-time business cycle indicator for the Euro area. The indicator provides an estimate of the monthly growth of Euro-area GDP, after the removal of measurement errors and seasonal and other short-run fluctuations. As reported by CEPR, the quarterly growth rate of the GDP averaged 0.59 over the sample period (1988–2003). Thus, a value of the EuroCOIN Index exceeding the long-term average 0.59 indicates growth above trend, and a positive index value below 0.59 indicates growth below trend.
While the data sample consisted of 888 observations from different parts of the world, the vast majority were European and U.S. investments. Since this article focuses on European and U.S. investments, others were removed from the study, reducing the dataset to 871 observations. Because of missing values in the dataset, observations were further reduced to the final number of 672 in the case of the multinomial regression model. Additionally, to perform the OLS regression, the IRR was included. Because of missing values the number of observations decreased to 666. Since we further excluded investments that were written off to perform the OLS regression, the final number of observations was 567.
Descriptive Statistics
An overview of all variables used in this article and their definitions are provided in Exhibit 2.
Exhibit 2
Definition of Variables
The 666 observations after the exclusion of missing values span an investment period of 16 years, from 1990 through 2005. Exhibit 3 presents the descriptive statistics for the data sample.
Exhibit 3
Descriptive Statistics
Panel A in Exhibit 3 reports the frequency of exit vehicles within the data sample. Overall, the most frequently used exit vehicle is a sale, followed by an IPO and finally a write-off. However, by splitting the data into European and U.S. investments, differences are evident. Although sales are the most common exit channel in both sub-samples, write-offs are more common than IPOs in the U.S. accounting. Additionally, the data shows a bias towards European investments that accounts for 538 observations, compared to 134 U.S. observations. The rows underneath the European allocation shed more light on the European distribution. It can be seen that U.K. investments clearly dominate the sub-sample, followed by French, Finnish, German, and Spanish investments. Overall and in contrast to the Europe/U.S. ratio, the proportions reflect the appropriate sizes of the European private equity market. Panel A furthermore reports the occurrence of exit channels by industry segment.8 Consumer Discretionary and Industrial Production are by far the largest segments, together making up more than 54% of the investments. Examining further the proportion of each industry segment per exit channel, it is shown that sales are the most common exits. Even though write-offs are the second-most-common exit vehicle for the Communications and Services segments, overall IPOs are the most-used exit vehicle after sales. Exhibit 4 presents the frequency of each exit vehicle over time.
Exhibit 4
Chicago Fed National Activity Index (CFNAI)
The CFNAI is a weighted average of 85 existing monthly indicators of national economic activity in the United States. It is constructed to have an average value of zero, and thus a positive index corresponds to growth above trend and a negative index corresponds to growth below trend.
Over the whole period, sales are the predominant type of exit vehicle. Further, the data shows that the proportion of sales compared to IPOs has grown over time, especially since the second half of the 1990s. Due to the large proportion of European buyouts, this can be explained by the European private equity market reaching its largest growth rates during that time. Further, the figure reveals that changes in IPOs and sales do not seem to be strongly correlated. Especially during the New Economy Boom, there was greater influence on the number of sales than the number of IPOs. However, by investigating the data sample by region, it turns out that in the U.S. sub-sample no entry is made for an IPO in 1999, the heyday of the New Economy Boom, indicating the limitations of the current data sample as a proxy for reality because of missing values.
Exhibit 3, Panel B sheds more light on the statistics of the holding period of the investments. There it is shown that on average investments are held for around four years. This is the case for investments exited via IPO and sale as well as MBOs/MBIs and LBOs. Differences are evident only for write-offs whose average holding periods are at least two years. The shortest-term investments are exited right at the beginning, whereas the longest duration is almost 16 years.
Summary statistics on IRR are provided in Exhibit 3, Panel C. The IRR measures the average annual return based on gross cash flow payments and is thus not affected by individually charged fees and costs. As investments analyzed herein are fully realized, the IRR can be calculated exactly and is not an interim reported measure. Due to missing values, the data sample is restricted to 666 observations. Overall, the sample generates a median of 29% and a mean of 38%. This is in line with calculations from Knigge, Nowak, and Schmidt [2004], who calculated a mean gross IRR of 39% for buyouts in their data sample. Returns for IPOs range from -13% to 1,287%, with a mean of 111%. Looking at the median of 70%, it seems that huge outliers bias the overall IRR upwards. Sales range from -100% to 900%, generating a mean of 49% and a median of 31%. There also seems to be an upward bias here, although not as strong. By comparing the IRR statistics across exit channels, IPOs reach the highest values, indicating that those exits are more profitable. By industry, the highest returns are achieved in the Information Technology sector followed by Services and Communication. The lowest returns are achieved for Consumer Discretionary, Materials, and Industrial Production. Exhibit 5 also sheds light on the average distribution of the returns over time. There it can be seen that the average returns in the sample correspond to general market movements, reaching their highest values during the boom phases in the early 1990s and in the New Economy Boom.
Exhibit 5
Frequency of Exit Channels Over Time (1990–2005)
This graph shows the annual frequency of exit channels.
METHODOLOGY AND RESULTS
Methodology
Multinomial Logit Model (MNL Model)
When the dependent variable is discrete, linear regression models are not appropriate due to the fact that basic assumptions of the model are violated, as shown by Wooldridge [2006] and Verbeek [2000]. Therefore we use the multinomial logit (MNL) approach to analyze determinants influencing the choice of an exit vehicle.
In this article we use the following three dependent variables:
I—initial public offering (IPO)
S—sale
W—write-off
whereby sale is used as the reference category, as it includes the most observations. Thus, formally the model consists of two equations, with lnΩ being the logarithm of the appropriate odds (also called logits), x representing the vector of independent variables that are presented in more detail below, Pr being the estimated probability for the success of the respective outcome category, and ß being the vector of the estimated coefficients.


with

Independent variables incorporate the private equity firm’s experience, investment characteristics, and macroeconomic factors. Including them in the equations above, the following linear models are estimated:


AGE is the age of the private equity firm at the date the investment started and is included as a proxy for the fund management’s experience. USA is a dummy variable equal to one if the investment was made in the USA and equal to zero if the investment was made in Europe. HOLDING PERIOD is included to analyze the impact of the length of the investment, as there are different assumptions on strategies that are associated with the duration of the investment and its success. The variable is scaled in years. To proxy the market conditions, we use the NO. OF IPO ISSUES, which is the number of worldwide initial public offerings per year, and NO. OF M&A DEALS, being the number of worldwide M&A transactions per year, as well as BOOM, a dummy variable that is used as a proxy for the economic activity. Based on the U.S. index CFNAI and the European index EuroCOIN, which measure GDP growth, BOOM equals one if both indices are above their long-term average for at least four quarters and zero otherwise. HIGHTECH is a dummy variable that equals one if the investment was made in the Healthcare & Other, Information Technology, or Communications segments and zero otherwise. This variable is considered to control for the fact that IPOs are more common in the high-technology segment than in other segments.
The effects of the individual coefficients considering all outcomes are tested using the Likelihood-ratio test as well as the Wald test. The appropriateness of the right-hand-side variables in the model are determined by comparing the Akaike and Bayesian information criteria, among other factors. The overall fit of the model is determined by analyzing results of the Likelihood-ratio test, McFaddens’s R2, and adjusted R2. The log-likelihood function was estimated by Newton’s method.
Furthermore, the independence from irrelevant alternatives (IIA) is tested. This property is an important and restrictive assumption of the multinomial logit model, as pointed out by McFadden [1973]: “Application of the model should be limited to situations where the alternatives can plausibly be assumed to be distinct and weighted independently in the eyes of each decision maker.”10 Two tests are used to check this property: the Hausman-type test proposed by Hausman and McFadden and an approximate likelihood ratio test proposed by McFadden, Tye, and Train and later improved by Small and Hsiao. Results of both tests show that the IIA property is not violated.
OLS Model
For analyzing the impact of the chosen exit vehicle on realized returns, an OLS model is used, with IRR being the dependent variable. The IRR is a standard performance measure in the private equity industry. It presents the discount rate that makes the net present value of an investment’s cash inflows equal with its cash outflows.11 For the average IRR per exit for each year in the period 1990 through 2005, see Exhibit 6.
Exhibit 6
Average IRR Over Time (1990–2005)
Average IRRs per annum are calculated on the basis of gross cash flows for fully realized investments only. The gross IRRs presented are on average 45% higher12 than their corresponding net values.
Investigating the differences in positive performance of the outcome categories, we exclude from the analysis write-offs, as they are always associated with an IRR of –100%. Thus, the data sample reduces to 567 observations. As a result, only IPOs and sales are analyzed as competing exit vehicles. They are included in the analysis via the dummy variable IPO that equals one if the investment is exited via IPO and zero if it is exited via sale.
To control for endogenous as well as external factors influencing the IRR, different variables are included. The ones that were already mentioned above (see also Exhibit 2) are the HOLDING PERIOD, NO. OF IPO ISSUES, and NO. OF M&A DEALS. In addition, we incorporate variables that control for various industry segments13 as well as INRATE, a variable that measures the risk-free rate at the date the investment was made.
Results
Since the interpretation of estimates of nonlinear models is more complex than that of linear models, we use different approaches to describe the relationship between the relevant variables and each outcome as comprehensively as possible. The first approach is based on predicted probabilities that are estimated for each outcome category. Predicted probabilities enable one to understand the influence of each independent variable on the respective exit channel. But although they provide a useful way to understand the magnitudes of effects in the model, they are limited in that they do not indicate the dynamics among the outcomes of the dependent variables. Therefore, for interpretation we use an additional approach based on the odds ratios. To obtain the odds ratios, the linear logit model has to be transformed by taking the exponential on both sides of the equation. The result is a multiplicative model in which its outcomes, the odds, are more intuitive measures than the logits.
Before investigating the individual influences of the variables, we first provide overview on the range of predicted probabilities for each of the three outcomes in Exhibit 7.
Exhibit 7
Predicted Probabilities per Exit vehicle
This exhibit shows the summary statistics of the predicted probabilities obtained from the multinomial logit model. The statistics are presented for each of the dependent variables (IPO, sale, and write-off) and display the mean, standard deviation, minimum value, and maximum value. The range is the difference between the maximum and minimum predicted probability.
The exhibit shows that the range of the predicted probability of a write-off is the widest starting at a value close to zero and extending to 90%. Next is a sale, whose range spans 81%, starting at the level of 9%. The range for an IPO is 66%, starting at 1%. The distributions indicate that there are nonlinearities in the model that have to be taken into account, as predicted probabilities occur below 20% and above 80%. Further conclusions can be drawn by looking at Exhibit 8, which summarizes the effect of each variable on the predicted probabilities per exit channel.
Exhibit 8
Effect of Each variable on predicted probabilities
This exhibit shows the statistics on predicted probabilities for a variation of the independent variable, shown in the first row, holding all other variables constant at their mean. The changes are reported as average changes above all exit channels in the second column and for each exit channel in the third, fourth, and fifth columns. The respective first row under each variable presents the range of the predicted probability by changing the independent variable from its minimum to its maximum value. The following two rows show the change in the predicted probability as the independent variable increases by one unit from 1/2 unit below to 1/2 unit above the mean value and from 1/2 standard deviation below to 1/2 standard deviation above the mean value. Mean values are listed in the equation below the exhibit. For dummy variables the exhibit shows only one row indicating the change in predicted probability as the variable changes from 0 to 1.
Overall, the figures in Exhibit 8 indicate that the range of probabilities for most of the variables is large enough to examine the impact of the variable on the respective probability of the exit channel. There are exceptions, such as the influence of the variable AGE on an exit via IPO. A change from the minimum to the maximum value of AGE changes the probability of an IPO by only 1%, thus reducing the validity of an interpretation. In general, interpretations of changes in probabilities by varying one of the independent variables are based on the assumption that the investment is average in all other characteristics.
Hypothesis 1: Write-offs are more likely in shorter holding periods than are IPOs and sales.
It was assumed that a write-off of an investment occurs near the beginning of the investment process, implying that underperforming investments are identified quickly and non-profitable investments are rather written off soon, instead of being held as living-dead investments.
As shown in Exhibit 9, an increase of the holding period from its minimum to its maximum value (0 up to almost 16 years) reduces the predicted probability of a write-off, whereas the probability for an exit via IPO or sale increases.
Exhibit 9
Change in Predicted Probabilities for Varying Holding Periods
Predicted probabilities are obtained from the multinomial logit model with the dependent variables IPO, sale, and write-off. The exhibit shows the changes in probabilities as the holding period increases from its minimum to its maximum value, holding all other variables constant at their mean.
It can be further seen that changes in predicted probabilities for write-offs are greatest in the first years of the investment. The same result can be drawn for sales, although the probability rises. On the contrary, the probability of realizing an exit via IPO seems to be quite stable over the duration of the investment. Looking at Exhibit 8, the observations above are also presented numerically. Varying the holding period from its minimum to its maximum value decreases the predicted probability of being written off by 53%. In contrast, the probability for exiting via IPO increases by 15% (38% for sales). By further investigating the standard deviation changes, it can be seen that the holding period has a huge influence on write-offs compared to other variables and outcomes. A standard deviation change in holding period centered around the mean decreases the probability of a write-off by 11%. To further investigate the dynamics of exit channels over each other, we analyze the logit estimates that are obtained from the multinomial logit regression and presented in Exhibit 10.
Exhibit 10
Logit Estimates on the determinants of the choice of Exit channel
The following results are obtained by using a multinomial logit model with the dependent variables IPO, sale, and write-off. All independent variables included in the model are listed in the first column. The second column shows the logits and the third column includes the coefficients whose sign indicates the direction of the effect. The significance levels for the p-value of the estimated coefficients are included in the fourth column and are denoted by *** for 1%, ** for 5%, and * for 10%. The effects as well as the standardized effects are presented in last columns.
There it can be seen that the effect of increasing the holding period by one year reduces the odds of realizing a write-off relative to an IPO by 50%. The odds of realizing a write-off relative to a sale is reduced by 48%. Both results are significant at the 1% level.
As stated earlier in this article, according to findings from Cumming and MacIntosh [2002], attention should also be paid to regional differences for the results presented above. First of all, by examining Exhibit 10 it can be seen that the probability that an investment is written off instead of exited via IPO or sale increases if the investment was made in the United States, indicating that write-offs are more common there. In the following, we split the data sample into U.S. and European investments and perform the same MNL regression as for the whole sample. The results are shown in Exhibit 11.
Exhibit 11
Logit Estimates for Rregional Sub-Samples
The following results are obtained by using the multinomial logit model with the dependent variables IPO (I), sale (S) and write-off (W). Independent variables are listed in the first column. The following columns include the logits with the appropriate estimated coefficients. The t-statistic is included below the coefficients whereby significance levels for the p-value are denoted by *** for 1%, ** for 5%, and * for 10%.
Here it can be seen that there are no apparent differences between the regional sub-samples when analyzing the treatment of write-offs. Both sub-samples indicate that the occurrence of write-offs compared to other exit channels decreases with an increase of investment duration. Furthermore, we analyze a possible shift from non-performing investments toward the exit vehicle sale to examine the existence of the living-dead phenomenon found by Cumming and MacIntosh [2002] in the Canadian data sample. Therefore, we investigate the size of underperforming European and U.S. investments whose internal rate of return is smaller than -90% and that are exited via sale. The U.S. sub-sample includes only seven investments exited via sale whose IRR was negative, although not smaller than -90%. Consequently, none of those investments are living-dead ventures. There are 49 European investments with a negative IRR exited via sale, although only five of those have a smaller IRR than -90%. Thus, their proportion of European write-offs (covering 64 observations) is only around 8%, which seems to be too small to support the living-dead phenomenon in the European data.
Overall, the results shown above provide strong support for Hypothesis 1. Given an average duration of four years, it can be stated that non-profitable investments are being written off early, as predicted probabilities decrease more rapidly during the first years. Consequently, it seems that rather than holding underperforming investments too long, funds use the more disciplined write-off strategy. This supports the argument that signaling plays an important role for private equity investors in the United States as well as in Europe. Further, Exhibit 10 reveals that the older and thus more experienced the PE fund, the probability of being written off increases compared to both other exit channels (IPO and sale). This is also in line with Hypothesis 1, as it indicates that signaling is especially important for sophisticated PE managers.
Hypothesis 2: A favorable stock market environment increases the probability of an exit via IPO.
It was argued that buyouts, like most other asset classes, are driven by the market sentiment, especially with regard to IPO exits. To proxy the stock market environment, we use the number of worldwide initial public offerings. Exhibit 11 shows the probabilities for each exit channel as the number of worldwide IPOs increases.
The probability of an exit via IPO in the data sample clearly increases with an increase in the number of worldwide IPOs. In contrast, the probability of an exit via sale decreases with almost the same dynamic. The probability of being written off does not seem to be influenced very strongly by the stock market climate. Exhibit 8 provides a more detailed view of the probability changes. It shows that the probability of a buyout-backed IPO increases by 5% with an increase in the number of worldwide IPO issues, whereas the probability of a sale decreases by 4%.
With regard to Exhibit 10, presenting the odds, it is shown that the IPO climate also influences the odds of an IPO versus a sale in the buyout segment. An increase of IPO issues by 1,000 per annum is expected to increase the chance of realizing an IPO compared to a sale by 38%. The result is significant at the 1% level. Furthermore, we investigate the effect of other macroeconomic variables on the odds of IPO versus sale. To compare the effects of variables that are scaled differently, we use standardized estimates that are shown in the right column in Exhibit 10. In the percent column of the standardized effect it can be seen that a period with high economic growth (measured through the variable BOOM) has the largest effect on the odds of realizing an IPO relative to a sale, with an increase of 46%, followed by the number of IPO issues, with an increase of 25%. On the other side, an increase in the worldwide M&A activity reduces the chance of realizing an IPO relative to a sale by 32%. This is in line with general arguments that private and public markets are at least partly substitutes. By looking at the odds of write-off versus IPO (and respectively sale) for the variable BOOM, it can be further summarized that the occurrence of write-offs decreases during times with strong economic growth, which is an intuitive result.
Exhibit 12
Effect of the IPO Climate on Predicted Probabilities for Each Exit
Predicted probabilities are obtained from the multinomial logit model with the dependent variables IPO, sale, and write-off. The exhibit shows the changes in probabilities as the number of worldwide IPOs increases, holding all other variables constant at their mean.
Additionally, in Exhibit 13 we split the data sample according to investments that were exited before the New Economy Boom (2000 and earlier) and those afterwards (since 2001) to analyze the differences that could have been arisen through the stock market crash.
Exhibit 13
Logit Estimates Before and after the new Economy Boom
The following results are obtained by using the multinomial logit model with the dependent variables IPO (I), sale (S), and write-off (W). Independent variables are listed in the first column. The following columns include the logits with the appropriate estimated coefficients. The t-statistic is included below the coefficients whereby significance levels for the p-value are denoted by *** for 1%, ** for 5%, and * for 10%.
The results show that for buyout investments that went public until the end of 2000, the IPO climate had a significant positive effect, whereas for investments that went public since the beginning of 2001, the IPO climate has not provided statistically significant estimates. With regard to the M&A activity, we found that results before the crash were similar to those for the whole sample, and after the crash there have been insignificant results as well.
Generally it can be confirmed that even for the buyout segment, the decision to go public is driven by the market sentiment, and thus Hypothesis 2 is supported. However, after a dramatic market crash like the fall of the New Economy Boom, these findings do not hold any longer. A possible explanation therefore could be found in behavioral finance and the phenomenon called herding, where the decisions of market participants are clearly influenced by observations they made in the past. Well-performing IPOs thus have a positive effect on the occurrence of herding, whereas during a market collapse herding seems to disappear.
Hypothesis 3: In the U.S. market the probability of a buyout-backed IPO is higher than in European stock markets.
We argue that the market characteristics of the more developed stock markets in the United States facilitate more buyout-backed IPOs than occur in European countries. As shown in Exhibit 8, the probability of realizing an IPO is 3% smaller for U.S. investments than for European investments. This is also the case for an exit via sale, where the probability reduces by 5% if an investment is made in the United States. Instead, the predicted probability for realizing a write-off increases by 8% for U.S. investments. By investigating the effects in Exhibit 10, it can be seen that the likelihood of a write-off is greater when an investment is made in the United States. However, the chance of realizing an IPO relative to a sale shows insignificant results for U.S. investments. Consequently, the results do not support the hypothesis.
It is also important to consider the fact that the proportion of U.S. investments is quite small, including only 134 observations, or only 20% of the data sample. This share does not represent an overall picture of the PE market. Therefore the European data sample was reduced to 134 observations, equal to the number of U.S. observations. Although the revised sample still does not represent the real-world proportion, it approaches toward it and meets the requirement for how large a data sample must be to allow estimations to be performed. To avoid arbitrary selection, the collection of the reduced European sub-sample was made randomly. In addition, we controlled for the proportion of each industry segment that is now in the reduced sample so that it is equal to the whole data sample. Results of the reduced sample analysis are provided in Exhibit 14.
Exhibit 14
Multinomial Logit Estimates of the reduced data Sample
The following results are obtained by using the multinomial logit model with the dependent variables IPO (I), sale (S), and write-off (W). Independent variables are listed in the first column. The following columns include the logits with the appropriate estimated coefficients. The t-statistic is included below the coefficients whereby significance levels for the p-value are denoted by *** for 1%, ** for 5%, and * for 10%.
Compared to the results obtained by using the whole data sample including 672 observations, the analysis with the reduced number of observations and equal proportion of U.S. and European investments shows few differences. Although the coefficient for the odds of IPO versus sale has the expected sign, thus indicating that the chance of realizing an IPO over sale is higher for U.S. investments, the result is statistically insignificant. Since stock markets in the U.S. and U.K. are often regarded as quite similar, we also analyze the effects when splitting the data sample into U.K. and U.S. investments (425 observations) and other European investments (247 observations). The results there show neither the expected positive sign nor statistical significance.
Consequently, with the reduced data sample as well as when merging investments made in the U.S. and U.K., Hypothesis 3 is not supported. Thus, market characteristics herein defined through the higher degree of development and experience do not have an influence on the occurrence of IPOs relative to sales.
Hypothesis 4: An exit via IPO has a positive effect on realized returns.
Herein it was argued that IPOs are a more successful exit channel than sales. To test the hypothesis, we use an OLS regression whose results are presented in Exhibit 15.
Exhibit 15
Determinants of the Internal Rate of Return (IRR)
Estimates are based on an OLS regression with IRR as the dependent variable. IPO is a dummy variable equal to 1 if the portfolio company is exited via IPO, and 0 if via sale. HOLDING PERIOD measures the investment duration in years. Further variables comprise industry segments14 and variables that proxy for the market environment. The two right columns split the data sample into phases with different economic activity. BOOM = 1 indicates that GDP growth was for at least four quarters above its long-term average, otherwise BOOM equals 0. Significance levels for the p-value of the estimated coefficients are denoted by *** for 1%, ** for 5%, and * for 10%.
This exhibit includes regression results for the full sample as well as for two sub-samples. The sub-samples are divided according to periods that are defined as boom phases. We define a period being a boom phase when both indices (for Europe the EuroCOIN index and for the U.S. the CFNAI) are above their long-term average for at least four consecutive quarters. IPO presents a dummy variable that equals one if the investment was exited via an IPO and zero if it was sold. The results of the whole data sample show a difference of 0.65 for the dummy variable IPO, meaning that the IRR that can be achieved through an exit via IPO is on average 0.61 times higher than for sales. By investigating the sub-sample results, it is shown that the magnitude of the effect is stronger in boom times, yielding 1.1 times greater returns than are produced in times with low economic activity, where returns of IPOs are only 0.35 times larger than those for sales. The results are significant at the 1% level.
Consequently, it can be concluded that taking a portfolio company public yields higher returns than an exit via sale. Even if at a first glance the magnitude is not what we would expect for high flyers, differences arise by investigating phases with extraordinarily high economic growth. There it is shown that the IRR of buyouts is clearly driven by the market sentiment, and thus the valuation of buyouts seems to be cyclical too. This result is in line with the one from Hypothesis 2 stating that the market climate has a strong influence on the decision to take buyout companies public.
The OLS regression further reveals statistically significant results for the investment duration, indicating that with an increase of the holding period by one year the IRR decreases on average by 12%. The decrease is also observable in both sub-samples. During boom periods an increase of one year leads to an average IRR reduction of 26%, compared with 10% in other periods. There it can be seen that fast exits during boom periods are an important factor, so timing seems to play a central role. Nevertheless, the negative sign raises questions, as it does not seem to support the value-adding thesis introduced above. Since theories about value creation in buyouts differ with regard to the holding period, we cluster the duration of the investments to analyze possible differences. Based on the clustering performed by Cao and Lerner [2006], we differentiate between investments that are exited within one year (also referred to as quick flips), investments that have an average holding period, and investments with an above-average holding period. The latter considers the quality signal sent out by investments exceeding a certain holding period. As pointed out by Cumming and MacIntosh [2002], those investments signal an inability to exit because their quality is regarded as too low.
In all sub-samples, we found average IRRs to decline with an increase in investment duration. Thus, the value-adding thesis does not seem to be supported. A possible explanation could be the reinvestment problem presented by Manigart et al. [2002]. This states that PE managers face difficulties in finding new appropriate investments when a portfolio company is liquidated too quickly. Thus, PE managers require a higher return for an investment with a short holding period. Additionally, recent studies have shown that various factors influence the performance and holding period, as pointed out by Ljungqvist and Richardson [2003], who reviewed different studies and stated that exits are based on individual strategies of PE fund managers. Those can be based, for instance, on current market conditions or considerations affecting their reputation. Also, Wright et al. [1994] analyzed the longevity of buyouts and pointed out that there are numerous influences on the investment’s duration. Overall, this makes the isolated analysis of the relationship between the duration of the investment and the process of adding value difficult and comprehensive, although it opens opportunities for further research. Besides, the analysis reveals that companies in the Information Technology and Services segments achieve high returns. However, results for the sub-samples are ambiguous, being statistically significant only for periods outside boom phases. Thus, it can be stated that special industry segments are able to achieve higher returns, although booming periods with probably high valuation levels seem to blur this effect. Additionally, results of the OLS regression show the negative impact of the number of worldwide IPOs on the IRR. According to Fenn, Liang, and Prowse [1997], lower returns during periods with greater capital availability could be due to the fact that within those times there are more unsophisticated private equity investors operating in the market. Additionally, during booming periods deals close more quickly, leaving less time for thorough due diligence, deal structuring, and pricing, resulting in lower returns.
DISCUSSION
Exit Strategies During the Financial Crisis
The global financial markets have experienced an unprecedented financial crisis since the summer of 2007 that is likely to carry over into 2010 and 2011 and constitutes an extraordinary challenge for all market participants. Since the onset of this crisis, companies with access to the available liquidity have had a major advantage, as banks and non-bank lenders have hoarded cash and grown sharply less able and willing to lend. The highly leveraged buyout market has been particularly affected. Our prior results have implications for the choice of exit vehicles in the buyout market during the financial crisis. The analysis suggests that both IPOs and trade sales are adversely affected during recessions, particularly when a credit boom comes to an end. Write-offs, however, are increasingly more likely to occur, for the same reasons that IPOs and trade sales decline—i.e., the lack of capital from investors and financing banks.
The validity of these predictions has become obvious over the last months. At the writing of this article, stock markets are at historical low valuations (in some cases down 70% from 12 months earlier) and are experiencing high volatility and increasing risk aversion among investors. In an uncertain economic environment, companies also lack a convincing equity story, due to their current earnings profile. As a result, the IPO market declined substantially: in 2008, the worldwide number of IPOs was down by 61% and issuing volume was down by 67% compared to the previous year (Ernst & Young [2009]).15
The volume and number of deals in the market for M&A were also substantially lower. The number of buyouts in Europe fell by 20% to 1,043 in 2008, the volume was down one third (to EUR 84 billion), and valuations were 50% lower compared to 2007 partly because of investors’ great uncertainty as to how firms would meet their own earnings targets.16 Private equity investors are selling only when necessary to avoid selling at fire-sale prices. Investors are hoarding cash (with Treasury-bill yields at extremely low levels) and buying only “to make a bargain,” yet another reason why private equity investors are trying to hold investments in their portfolios and rather pursue a buy-and-build strategy, if possible.
The years before the onset of the financial crisis can be characterized as a credit boom, with easy and cheap credit facilities. This led to deals being largely financed with debt that needed to be partly repaid or refinanced in the next couple of months. Furthermore, the ongoing recession causes firms to breach their covenants and to negotiate new loan terms with their lenders. Recent research by Ivashina and Scharfstein [2009] has shown that banks substantially reduced credit activity in the United States. To avoid covenant breaches, private equity investors are restructuring their portfolio companies.17
Taken together, a lack of alternative exit options (IPOs and trade sales) and also tight credit markets suggest that we see default rates of private equity-backed deals (and subsequently write-offs) substantially increasing.18
Alternative Tests
We perform a variety of robustness checks to support our results. We use alternative measures for the funds experience as well as macroeconomic factors. The variable EXPERIENCE is a dummy variable that equals one if at the date the investment was made the PE investor had already raised a fund, and zero if it was a first-time fund. VCBOOM is a dummy variable that proxies for IPO activity in the private equity market. It equals one if private equity-backed IPOs exceed their long-term average (1990–2004), and otherwise it equals zero. HOTIPO is also a dummy variable that accounts for the New Economy Boom and equals one if the investment was exited in the year 1999 or 2000, and zero otherwise. Furthermore, we include the volume of worldwide M&A deals through the variable VOL.OF M&A DEALS.
Results for the multinomial logit model are shown in Exhibit 16.
Exhibit 16
Robustness Checks for the Multinomial Logit Model
The following coefficients are estimated by using a multinomial logit model with the dependent variables IPO (I), sale (S), and write-off (W). Independent variables are listed in the first column. The following columns include the logits for each model with the outcome sale as base category. The t-statistic is included below the estimated coefficients whereby significance levels for the p-value are denoted by *** for 1%, ** for 5% and * for 10%.
Overall, the figures presented in this article support the findings for Hypotheses 1-3. They are also in line with the signaling effect, showing that signaling plays an important role, especially for sophisticated investors.
Results of the robustness checks for the OLS model and thus for Hypothesis 4 are presented in Exhibit 17.
Exhibit 17
Robustness Checks for the OLS Model
Estimates are based on an OLS regression with IRR as the dependent variable. IPO is a dummy variable equal to 1 if the portfolio company is exited via IPO, and 0 if via sale. Significance levels for the p-value of the estimated coefficients are denoted by *** for 1%, ** for 5%, and * for 10%.
It can be seen that the result for Hypothesis 4 is also supported by including alternative variables. Furthermore, to analyze the “money-chasing deal” argument that emerged in line with the OLS regression, we incorporate the variable COMMITED CAPITAL, being the committed capital to the private equity market at the exit date of the investment. However, we could not find additional statistical support for the negative impact on liquidity or competing fund inflows on returns within the present data.
CONCLUSION
This article focused on buyout exit strategies in Europe and the United States. We analyzed the three main exit routes (IPO, sale, and write-off) by using a unique data set derived from the CEPRES Private Equity Analyzer. This database contained anonymous cash flow information on the individual investment level. We analyzed the motivation behind and determinants influencing the choice of an exit vehicle by using a multinomial logit model. We found strong support for the signaling effect, implying that PE investors tend to quickly write off investments that turn out to be non-performing, instead of holding them in their portfolios as so called living-dead investments. Through this strategy, PE investors show their ability to differentiate between investments that are worth supporting further and those that are not. Additionally, we found evidence that exit decisions for buyouts are also driven by the market sentiment and thus seem to be cyclical like other investments. On the contrary, we could not find support for the idea that market characteristics influence the choice of exit vehicles. Despite this, we evaluated the influence of a chosen exit vehicle on the IRR of the investment by using a multivariate regression model. There, we found support for the common argument that only exceptional firms are taken public as they achieve higher returns. We found returns superior for IPOs relative to sales in periods with high GDP growth. In future research it would be insightful to analyze the category of sales in more detail, since that exit vehicle comprises several sub-types that could not be recognized herein. It would also be interesting to analyze the influence of the holding period on buyout returns, as this relationship is still unclear.
ENDNOTES
We thank Mark Wahrenburg and Sina Borgsen (discussant) and the participants at the German Finance Association Annual Meeting (DGF, 2007) for valuable comments. All remaining errors are our own.
↵ 1The term “private equity” as used in this article comprises all stages of investments where private equity capital is provided.
↵ 2Prior public companies that have been subject to LBOs and now are again returned to the public equity market are referred to as reverse LBOs.
↵ 3This theory supports the positive role of leverage financing as motivation and a controlling instrument. It is shown that through an LBO agency costs can be reduced and the operating efficiency can be improved, leading to an increase in firm value.
↵ 4Underperforming investments held in portfolios instead of being written off as necessary are often referred to as living-dead.
↵ 5They argued that low returns in boom periods are due to tough competition and the fact that deals are closed quickly, leaving less time for careful due diligence and deal structuring. They also stated that the presence of less-experienced PE investors is higher in boom periods. In general their argument supported the “money-chasing deal argument” presented by Gompers and Lerner [2000b].
↵ 6Although a possible bias could arise by excluding write-downs, it allows an accurate classification of non-performing investments. Types of exit vehicles in the data sample consist of four categories: “realized-IPO,” “realized sale,” “realized public merger/sale,” and “write-off.” However, due to the similarity and the small number of observations related to the exit category “realized-public merger/sale,” those data are incorporated in the category “realized sales.”
↵ 7The Stock-Watson XCI was one of the leading coincident indicators for U.S. economic activity until 2003. It was developed by James Stock of Harvard University and Mark Watson of Princeton University in an article titled “Forecasting Inflation” published in the Journal of Monetary Economics, 44 (1999), pp. 293-335.
↵ 8The 27 industry classifications provided by CEPRES were aggregated to the following industry segments: Financial Business Services: Financial Services, Business Services, Fund of Fund Investments; Consumer Discretionary: Consumer Industry/Food, Hotel, Leisure, Retail, Textiles; Healthcare & Other: Healthcare/Life Sciences; Industrial Production: Construction, Industrial/Manufacturing, Traditional Products; Information Technology: High Tech, IT, Semiconductor, Software; Communication: Internet, Media, Telecom; Materials: Materials, Natural Resources/Energy; Services: Environment, Logistics, Transportation, Waste/Recycling, Other Services; Others.
↵ 9The 27 industry classifications provided by CEPRES were aggregated to the following industry segments: Financial Business Services: Financial Services, Business Services, Fund of Fund Investments; Consumer Discretionary: Consumer Industry/Food, Hotel, Leisure, Retail, Textiles; Healthcare & Other: Healthcare/Life Sciences; Industrial Production: Construction, Industrial/Manufacturing, Traditional Products; Information Technology: High Tech, IT, Semiconductor, Software; Communication: Internet, Media, Telecom; Materials: Materials, Natural Resources/Energy; Services: Environment, Logistics, Transportation, Waste/ Recycling, Other Services; Others.
↵ 10See Agresti [2002], p. 299.
↵ 11The definition is obtained from the glossary of the NVCA Yearbook 2005.
↵ 12See Schmidt, Nowak, and Knigge [2004] for more details.
↵ 13The following industry segments were incorporated in the model: Financial Business Services, Healthcare & Other, Information Technology, Communication, Materials, Services, and Others. Consumer Discretionary and Industrial Production were excluded due to high collinearity.
14The 27 industry classifications provided by CEPRES were aggregated to the following industry segments: Financial Business Services: Financial Services, Business Services, Fund of Fund Investments; Consumer Discretionary: Consumer Industry/Food, Hotel, Leisure, Retail, Textiles; Healthcare & Other: Healthcare/Life Sciences; Industrial Production: Construction, Industrial/Manufacturing, Traditional Products; Information Technology: High Tech, IT, Semiconductor, Software; Communication: Internet, Media, Telecom; Materials: Materials, Natural Resources/Energy; Services: Environment, Logistics, Transportation, Waste/Recycling, Other Services; Others.
↵ 15The main IPO activity took place in United States (with respect to IPO volumes) and China (with respect to the number of IPOs). In the first quarter of 2009, there were 51 IPOs worldwide.
↵ 16Source: “European Private Equity Review,” Mergermarket [2009].
↵ 17According to a PWC survey, 71% of all international private equity funds started to restructure portfolio companies in 2008 and 79% expected to restructure or continue restructuring in 2009 (PWC [2009]).
↵ 18A recent study by the Boston Consulting Group even suggests that 50% of private equity portfolio companies might default within the next three years (Meerkatt and Liechtenstein [2008]).
- © 2010 Pageant Media Ltd