The Big Squeeze: Capacity constraints and merger arbitrage hedge fund performance in the last two decades Zaur Rzakhanov∗ Gaurav Jetley University of Massachusetts Boston Analysis Group Department of Accounting and Finance 10 Rockefeller Plaza 100 Morrissey Blvd New York, NY 10020 Boston, MA 02125 Abstract The study proposes a model that explains the evolution of merger arbitrage strategy alpha since early 1990s. The paper demonstrates that the decline in alpha over time is relatedtotheexpansionofmergerarbitragecapitalrelativetoavailablearbitrageopportu- nities,andthatthemergerarbitragespreadactsasanimportantconduitthoughwhichthis expansion has impacted alpha. The results indicate the existence of significant capacity constraints in the merger arbitrage hedge fund strategy and suggest that skill differences amongUS-focusedmergerarbitragefundmanagersareunlikelytobereflectedindifferences in hedge fund returns realized by investors. Keywords: hedge funds; merger arbitrage; capacity constraints. JEL Classification Numbers: G230. ∗Corresponding author. Telephone: 617.287.7331, E-mail:[email protected] 1 1 Introduction Merger arbitrage or risk arbitrage is an investment strategy where the objective is to realize the spread between the acquisition price and the price at which the stock of a target trades subsequent to the announcement of an acquisition. Merger arbitrage is a popular investment strategy. Between 1990 and 2010 assets under management of U.S.-focused, U.S. dollar de- nominated merger arbitrage funds has grown from under $0.5 billion in early 1990s to over $2 billion in 2010 (see e.g., Hedge Fund Research Inc. or HFR, Lipper Tass). In this paper we examine the evolution of alpha of merger arbitrage funds since early 1990s. In particular, we show that a part of the variation in the alpha of merger arbitrage funds can be explained by changes in both the demand and supply of capital devoted to merger arbitrage. In addition, we document several interesting facts such as, the decline of returns of M&A hedge funds and the lack of persistence in performance of M&A funds. Ourpaperisrelatedtostudiesonmergerarbitrageaswellasstudiesthatexamineevolution ofhedgefundalphas. Forexample,LarckerandLys(1987),MitchellandPulvino(2001),Baker and Sava¸soglu (2002), and Jindra and Walkling (2004) report economically and statistically significant excess returns related to merger arbitrage. Anumberofstudieshaveanalyzedtheevolutionofalphasovertimeandtheextenttowhich capital inflows have contributed to the decline in alphas. Fung, Hsieh, Naik, and Ramadorai (2008) analyze fund-of-funds and find structural breaks in hedge fund risk exposures, where structural breaks represent statistically and economically significant changes in systematic risk. Fung, Hsieh, Naik and Ramodarai also find that after taking into account the change in systematic risk, alphas tend to decline following capital inflows. Similarly, Naik, Ramadorai, and Stromqvist (2007) find capital inflows to precede declines in alphas of four out of eight hedge fund strategies. Naik, Ramodarai and Stromquist also take into account the structural break points used by Fung, Hsieh, Naik and Ramodarai. Jetley and Ji (2010) focus on merger arbitrage funds and report a decline in both returns and alphas of merger arbitrage funds since theearly2000s. JetleyandJiindirectlylinkthedeclineinalphasandreturnstocapitalinflows 2 by using volume of trading in a targets’ stock around the announcement date as a proxy for capital flows. A number of other studies, such as Zhong (2008) and Jagannathan, Malakhov, and Novikov (2010) have also analyzed the decline in alphas over time. We extend this literature in several ways. First, we demonstrate a negative relationship between M&A hedge fund alphas and capital devoted to the merger arbitrage strategy. The characteristics of the merger arbitrage strategy are such that it enables one to develop reason- able proxies for both demand and supply of merger arbitrage capital. In addition, the merger arbitrage spread – the amount by which the bid price exceeds the post-announcement price of a target – allows one to precisely measure the unlevered profit potential of the strategy. Our contribution to the literature builds on the inherent measurement-related advantage of merger arbitrage with respect to both the strategy’s capacity (supply and demand of capital) and profitability. We exploit the fact that the size of the merger arbitrage market is limited by the volume of M&A deals at any point in time. In other words, for a given period of time, the dollar value of deals announced during that period is a good proxy for the demand of merger arbitrage capital. Further, assets under management of merger arbitrage funds may be used as a proxy for the supply of merger arbitrage capital. Thus by being able to separately track the demand and supply of merger arbitrage capital we develop a measure a merger arbitrage cap- ital abundance (MACA), which we use to explain the evolution of alphas of merger arbitrage funds. Consistent with Jetley and Ji (2010) and Fung, Hsieh, Naik, and Ramadorai (2008), we find that large hedge fund returns and alpha related to merger arbitrage prevailing in the 1990s have declined substantially. We directly link the decline of alphas to an increase in MACA – our measure of merger arbitrage capital abundance – and to the decline in merger arbitragespread. Weestablishthislinkbyfirstusingtheseven-factormodeldevelopedbyFung and Hsieh (2004) to obtain an out-of-sample forecast of merger arbitrage strategy alpha over time. We then explore the relationship between forecasted alpha and two variables: one that proxies for the amount of capital available for merger arbitrage relative to existing arbitrage opportunities and another that proxies for merger arbitrage profitability. We find negative 3 relationship between merger arbitrage capital abundance variable and alpha, as well as a positiverelationshipbetweenarbitragespreadandalpha. Therelationshipsalsosurvivevarious robustness checks designed to address possible biases related to hedge fund data collection, return properties and model specification. Lastly, we find that these two variables accurately predict the evolution of average merger arbitrage strategy alpha over time. We also document lack of long term persistence in the performance of M&A hedge funds, both in absolute and relative terms. Absolute persistence refers to the ability to generate similar returns over time. So if a hedge fund was able to generate a 6 percent return in say 1998, absolute persistence would imply returns of about 6 percent in years subsequent to 1998. The decline in average absolute performance is consistent with the observed increase in capital devoted to risk arbitrage as well as the decline in arbitrage spread. Thelackofpersistenceinrelativereturns–theinabilityofmergerarbitragefundstoconsis- tently report returns that would put them in the top quartile – has interesting implications. A lack of relative persistence suggests that either skills do not vary much across merger arbitrage funds or difference in skill are less relevant than strategy-wide factors. The lack of persistence in relative returns coincides with narrowing of the distribution of arbitrage spreads over the lasttwodecades(seee.g.,JetleyandJi(2010)). Becausearbitragespreadistheprimarysource ofprofitformergerarbitragehedgefunds, narrowingarbitragespreaddistributionimpliesthat particularly skilled fund managers, if they exist, would find it more difficult to differentiate themselves from mediocre managers. These results build on the previous studies of persistence in hedge funds’ performance. Brown, Goetzmann, and Ibbotson (1999) using 1989-1995 data found no persistence in hedge fund returns for 1 year horizons, Agarwal and Naik (2000) found evidenceofpersistenceforquarterlyhorizons, suggestingthatpersistenceinreturnsexistsover short time horizons. Fung, Hsieh, Naik, and Ramadorai (2008) found that between 1994 and 2005 only 22% of all hedge funds delivered positive and statistically significant alpha. A more resent paper by Jagannathan, Malakhov, and Novikov (2010) indicates persistence over 3 year horizon for top hedge fund managers and no persistence for the rest. As far as we know, our paper is the first to take into account both the demand and supply 4 of merger arbitrage capital to explain the evolution of alphas of merger arbitrage funds. Our findings also show that cross-sectional variation in returns has declined sharply in recent years, indicating that merger arbitrage hedge fund managers have found it difficult to distinguish themselves from each other. This conclusion is further corroborated by the lack of long term persistence in performance of merger arbitrage hedge funds. Taken together these results suggest that managerial ability (or skill) to deliver returns did not vary much across merger arbitrage hedge funds in the recent years due to increased capital availability, competition among funds and an overall decline in profitability. The rest of the paper is structured as follows. The next section describes stylized facts regarding the performance of merger arbitrage hedge funds over the last two decades, and developsourtestablehypotheses. Section3describestheempiricalstrategy. Section4discusses the results. Section 5 concludes. 2 Stylized facts and hypotheses development The last two decades saw a significant increase in popularity of hedge funds. The amount of assets under management has grown by fourteen fold from $118 billion in 1997 to $1.7 trillion in 2010. Assets under management of U.S. focused merger arbitrage hedge funds grew fourfold to over $2.0 billion in 2010. Figure 1 plots the natural logarithm of the ratio1 of assets under management of U.S. focused, U.S. dollar denominated merger arbitrage funds to value of announced M&A deals over time.2 Additionally, we estimate time trend of this measure using locally weighted scatter plot smoothing (lowess). The figure indicates a strong upward trend in merger arbitrage capital relative to the value of deals announced from 1994 through 1ForeachcalendarmonththenumeratoristhesumofassetsundermanagementofU.S.-focusedhedgefunds whoseassetsaredenominatedinU.S.dollarsasreportedbyHFRandLipperTassdatabases. Thedenominator isthesumofvaluesofM&Abidsannouncedinagivencalendarmonthnetofassumedliabilitiesasreportedby Thomson ONE Banker database. We used only bids that, if completed, will result in change in control defined as increase in ownership by the bidder from less than 50% to more than 50%. 2Thisratioislikelytounderstatecapitalabundanceaswedonottakeintoaccountassetsundermanagement for funds that are not specifically U.S.-focused. Such funds may also allocate a part of their portfolios to U.S. mergers and acquisitions. We identified U.S. focused funds by using regional investment focus variable in HFR and Lipper Tass databases. We used fund asset currency denomination variable to identify U.S. dollar denominated funds. 5 early 2000s. Subsequently, the ratio declines somewhat, and then seems to level off starting in 2006. [Insert Figure 1 here] The increase in merger arbitrage capital through early 2000s has been accompanied by a downward trend in merger arbitrage spread over the same period. Figure 2 depicts the evolution of median quarterly arbitrage spread for friendly cash only bids.3 Consistent with Jetley and Ji (2010) we observe a decline in the spread between 1994 and early 2000s, reaching the bottom around 2002 and 2003. After 2003 the median arbitrage spread increases slightly, spiking in the last quarter of 2008 and the first quarter of 2009 and then sharply declining in the latter part of 2009 and in 2010. Despite these fluctuations the merger arbitrage spreads exhibits a broad decline between 1990s and 2000s. The same conclusions hold when we use locally weighted scatterplot smoothing to trace the median arbitrage spread trend over time. [Insert Figure 2 here] Correspondingtotheincreaseinmergerarbitragecapitalandtothebroaddeclineinmerger arbitrage spread over time is the decline in merger arbitrage hedge fund returns. Given trends reported in Figures 1 and 2 we compare hedge fund performance between 1994 to 2002 period and 2003 to 2010 period. The average monthly return for HFR merger arbitrage hedge fund index fell from 88 basis points in the 1994 – 2002 period to 50 basis points in the 2003 – 2010 period. The decline in average returns can also be seen for individual merger arbitrage hedge funds.4 In our sample of hedge funds formed prior to 2003, the median return declines from 82 basispointspermonthin1994-2002periodto41basispointspermonthin2003–2010period. Figure 3 shows the box plot of the distribution of the M&A hedge funds’ returns indicating 3As in Figure 1 we used only bids that, if completed, will result in change in control defined as increase in ownership by the bidder from less than 50% to more than 50%. To calculate the arbitrage spread we subtract target’s stock price one trading day after offer bid announcement from the offer’s bid price and divide the difference by target’s stock price one trading day after offer bid announcement. For multiple bid auctions we use the first offer’s bid price and stock price one trading day after the day of the first bid announcement. Multiple bidders are defined as in Bates, Becher, and Lemmon (2008) and Bates and Lemmon (2003). We remove outliers by excluding the top and the bottom 1% of merger arbitrage spread values and use remaining sample of offer bids to calculate quarterly median arbitrage spread. 4Some of those funds are not in HFR M&A hedge fund index. 6 significantdeclineinthemedian,25thand75thpercentilesofthereturndistributionovertime. Thus, Figures 1 through 3 suggest that increase in merger arbitrage capital prior to 2003 not only reduced the returns but also muted the differences between competing M&A arbi- trage hedge funds. A closer examination of the individual hedge funds’ returns points to the compression of hedge fund returns over time. Figure 3 indicates that while during the 1994 – 2002 period the interquartile range for average monthly merger arbitrage hedge fund returns was 47 basis points, in the 2003 – 2010 period, the interquartile range has shrunk by more than 50 percent to 19 basis points per month. This result is consistent with the Jetley and Ji (2010) who report a compression of merger arbitrage spreads during roughly the same period.5 [Insert Figure 3 here] Another interesting aspect of hedge fund performance is its persistence. Is good (bad) performance in one period followed by good (bad) performance in the subsequent period? Ability to identify funds with persistent performance helps investors to identify funds that are consistently good and avoid funds that are consistently bad. Early research found no or little persistence over short time horizons (quarterly or annual), while more recent literature found differences across funds and some intermediate term (3 year horizon) persistence among alpha producing funds (see e.g. Brown, Goetzmann, and Ibbotson (1999); Agarwal and Naik (2000); Fung, Hsieh, Naik, andRamadorai(2008); Zhong(2008); Jagannathan, Malakhov, and Novikov (2010)). Our results also suggest lack of persistence in the merger arbitrage hedge funds’ performance. Figure 4 plots average merger arbitrage hedge fund returns in the 1994 – 2002 period against average merger arbitrage hedge fund returns in 2003 – 2010 period. The average returns for individual funds exhibit substantial variation in the earlier period which declines noticeably in the later period. This can be gauged by comparing the distribution of returns along the horizontal axis to those along the vertical axis. As shown in Figure 4, the reduction is not limited only to best performing funds. Generally, funds in the upper part of returndistributioninthe1990shaveseentheirperformancedeclinesubstantially.6 Conversely, 5See Table 2, page 57 of Jetley and Ji (2010). 6They are located below the 45 degree line on Figure 4. 7 funds that underperformed in the 1994 – 2002 period improved their performance in the later period.7 [Insert Figure 4 here] The decline in returns as well as the reduction in variation of returns shown in Figures 3 and 4 is consistent with Zhong (2008) and Fung, Hsieh, Naik, and Ramadorai (2008). Zhong found that the decrease in hedge funds’ alpha is driven primarily by decline in alphas among top performing funds. Fung, Hsieh, Naik and Ramadorai found that capital inflows in alpha- producing funds have decreased the ability of such funds to continue delivering alpha. Becausehedgefundsreportreturnsandassetsundermanagementonvoluntarybasishedge fund databases are subject to a number of well-known biases. Our primary concern is with survivorship, backfill and serial correlation biases (e.g., Jagannathan, Malakhov, and Novikov (2010)). The backfill bias arises when a fund joins a hedge fund database such as HFR or Tass databases bringing with it its return history. If a hedge fund chooses to join a database during theperiodofrelativelyhighperformanceusingbackfillobservationmaycreateanappearanceof decliningperformanceovertime. Thesurvivorshipbiasexistswhenperformancemeasurements do not take into account performance of funds that have been liquidated or stopped reporting. Ifthedecisiontoliquidateorstopreportingissystematicallyrelatedtoperformance,theresults will be biased upward when performance of such funds is not taken into account. Investments by hedge funds in relatively illiquid securities are known to be one of the sources of the serial correlation bias. It is unlikely that the results shown in Figures 3 and 4 are due to the survivorship or backfill biases. To begin with, Liang (2000) finds that the performance of merger arbitrage funds in HFR and Tass databases exhibit little or no survivorship bias. However to investi- gate survivorship bias we analyze the performance of both living funds and funds that either liquidated or stopped reporting as of the end of the sample period (2010). Out of 30 hedge funds in our HFR/Tass sample with date of inception prior to 2003, 14 funds continued to be 7We find similar result when we compare individual hedge funds’ Sharpe’s ratios. 8 active through the end of the sample period, another 14 funds liquidated and 2 funds stopped reporting after 2002. Notably 11 out of 14 funds that liquidated did so between March and November 2008, while 2 funds that stopped reporting did so in June 2008.8 Thus, over 50 percent of funds in our sample that existed prior to 2003 liquidated or stopped reporting in 2008. Because the two funds that have stopped reporting did so in 2008, it is unlikely that their decision is driven by superior performance. Consequently, it is highly improbable that our results are driven by the survivorship bias. Additionally,tocheckiftheresultsreportedinFigures3and4aresensitivetotheinclusion of observations close to financial crisis of 2008, we exclude funds’ performance subsequent to December 2006. The results remained the same – we still find the decline in and compression of merger arbitrage hedge fund returns as well as lack of persistence for active and liquidated funds. The presence of data from the recent financial crisis does not appear to influence our conclusions. To check if the results are influenced by backfill bias we exclude all observations prior to date of inclusion in HFR or Tass databases. Again, our results continue to hold.9 Getmansky, Lo, and Makarov (2004) argue that the trading in illiquid securities by hedge funds may lead to serial correlation in hedge fund returns that would make hedge fund returns appear to be less volatile and hence bias results in favor of finding performance persistence. In general, onewouldnotexpectmergerarbitragehedgefundreturnstobesignificantlyimpacted by serial correlation bias. This is because merger arbitrage funds invest in relatively liquid securities – common equity of target firms. Our sample of hedge funds confirms this intuition. Wefindthattheaveragefirstorderautocorrelationinhedgefunds’returnsis0.16,andthatthe autocorrelation estimates are statistically insignificant for 75% of hedge funds in our sample. Because the existence of serial correlation in hedge fund returns would work against finding results reported in Figures 3 and 4, we conclude that serial correlation is not strong enough to affect conclusions reported in those figures. Nevertheless, to account for the impact of 8We use the last date of reported assets as an indicator of when the fund liquidated or stopped reporting. 9Results available upon request. 9 autocorrelation in hedge fund returns we use serial correlation robust Newey and West (1987) standard errors in our hedge fund performance models. We find that our results are not sensitive to assumptions regarding autocorrelation of hedge funds returns. We have also looked at persistence of relative performance. Again there appears to be no relationship between funds’ relative performance rankings in 1994 – 2002 and in 2003 – 2010 periods - the Spearman’s rho10 for the average monthly merger arbitrage hedge fund returns is only –0.03 and the p-value is 0.89. Thus, a performance leader or laggard in an earlier period is no more or less likely to remain in the same position relative to other funds in the latter period. This inability to deliver to investors consistently superior (or inferior) returns relative to peers is indicative of limited or no role of skill in merger arbitrage strategy during the period. This result is also illustrated in Table 1. The table compares hedge fund rankings using average return quartiles in 1994 – 2002 and 2003 – 2010 periods. In each period we rank each fund of by average raw monthly return within that period, and in each period we sort each fund into a return quartile. The table examines whether a fund was able to remain in the same quartile during the two time periods – 1994 – 2002 and 2003 – 2010. The table shows whether the relative performance improved or declined across the two periods. If there was persistence in relative performance one should observe all hedge funds clustering on the main diagonal. Instead, we see that performance rank in 1994 – 2002 period is typically followed by either increase or decline in rank in the 2003 – 2010 period. This result survives if we exclude backfilled observations from our sample. [Insert Table 1 here] The stylized facts point to the possibility that the decline in merger arbitrage hedge fund returns over the last two decades has been driven by the increasing merger arbitrage capital and the decline in the level and variability of merger arbitrage spreads. The findings presented above are consistent with Jetley and Ji (2010) who document a 400 basis points decline in merger arbitrage spread between early 1990s and mid-2000s. 10Using30mergerarbitragehedgefundswithdateofinceptionpriorto2003. Definedascorrelationbetween the rankings of values of two variables. 10