The Price of Growth: Evidence of the Pedestrian Nature of Post-Merger Returns

Transcription

1 The Price of Growth: Evidence of the Pedestrian Nature of Post-Merger Returns Sandra Mortal U.S. Securities and Exchange Commission and The University of Memphis Memphis, TN Michael J. Schill Darden Graduate School of Business Administration University of Virginia, Box 6550 Charlottesville, VA February 2011 We thank Ming Dong, Dirk Hackbarth, Vikas Mehrotra, Elvira Sojli, Guoxiang Song, and seminar participants at Brigham Young University, George Mason University, Queens University, the University of Cambridge and the University of Virginia, and conference participants at the meetings for the European Finance Association in Frankfurt and the Northern Finance Association in Winnipeg for helpful comments. Schill acknowledges the Darden Foundation for research support. The Securities and Exchange Commission disclaims responsibility for any private publication or statement of any SEC employee or Commissioner. This paper expresses the author's views and does not necessarily reflect those of the Commission, Commissioners, or other members of the staff.

2 The Price of Growth: Evidence of the Pedestrian Nature of Post-Merger Returns Abstract A large literature examines the returns associated with acquiring firms, presupposing that the returns associated with these firms are in some way unique. We assert that the distinguishing characteristic associated with the returns of these firms is more precisely the magnitude of the expansion of their balance sheet rather than the completion of an acquisition per se. Examining post-deal returns, we observe no difference in returns between firms that grow through merger and those that grow organically at the same rate. In the cross-section, the relatively poor returns documented for stock deals and glamour deals are due rather to the relatively greater asset growth associated with these deals. In addition, the relatively poor returns associated with deals made during high valuation periods are also identical to those made by firms with similar asset growth. The evidence suggests that the long-run returns associated with acquiring a business is not unique to mergers, but is part of a broader asset growth effect. JEL classification: G14; G34 Keywords: Acquisitions; Firm valuation; Market efficiency; Asset growth 1

3 1. Introduction The literature documenting and evaluating the poor post-deal returns of acquiring firms has a long history (see Dewing, 1921; Livermore, 1935; Reid, 1968; Hogarty, 1970; Asquith, 1983; Jensen and Ruback, 1983; Loderer and Martin, 1992; and Agrawal, Jaffe, and Mandelker, 1992). The regularity in post-acquisition returns has been shown to exist in multiple markets (Firth, 1980; Franks and Harris, 1989; Gregory, 1997; Andre, Kooli, and L Her, 2004; and Gregory and McCorriston, 2005) and to vary across several dimensions, including firm size (Malatesta, 1983; Moeller, Schlingemann, and Stulz, 2005), institutional regime (Schipper and Thompson, 1983), form of offer (Dodd and Ruback, 1977; Bradley, Desai, and Kim, 1983), form of payment (Loughran and Vijh, 1997 and Savor and Lu, 2009), short selling activity (David, Drake, and Roulstone, 2010); insider trading activity (Song, 2007; Akbulut, 2009), investor network (Kamath and Yan, 2010); time period (Rosen, 2006; Bouwman, Fuller, and Nain, 2009), econometric design (Mitchell and Stafford, 2000), and performance benchmark (Mandelker, 1974; Langetieg, 1978; Malatesta, 1978; Franks, Harris, and Titman, 1991; and Fama and French, 1993). The leading theory explaining this regularity maintains that market pricing inefficiency plays a role in motivating acquisitions. Following Lakonishok, Shleifer, and Vishny (1994), Rau and Vermaelen (1998) propose that firm stakeholders (management, directors, and investors) in firms with strong performance records extrapolate firms performance in such a way that overestimates management s ability to manage future acquisitions. In support, Rau and Vermaelen find that the underperformance of acquiring firms is concentrated in firms with high past returns and glamour valuation ratios. In a competing model, Shleifer and Vishny (2003) 2

4 argue rather that managers observe firm mispricing and rationally respond by opportunistically acquiring relatively underpriced target shares using relatively overpriced firm stock. Shleifer and Vishny cite existing empirical support for their model in findings of poor post-acquisition performance among stock deals. 1 We contribute to this long literature by proposing that, with respect to post-deal returns, acquisitions in and of themselves are benign corporate events. We maintain that it is the firm s asset expansion that commonly accompanies takeovers that is associated with the pervasive return effects documented in this literature, rather than anything unique to the takeover per se. Traditional theory on corporate investment maintains that investment anticipates expected return changes, to the extent that firms make investments when project values rise in response to discount rate reductions. The time-varying nature of discount rates generates a negative correlation between asset growth and subsequent returns (see, Tobin, 1969; Yoshikawa, 1980; Zhang, 2006; and Xing, 2007). 2 An alternative motivation for an asset growth effect in returns is that managers act opportunistically by exploiting investor underreaction to empire building (Titman, Wei, and Xie, 2004) or by exploiting investor mispricing by expanding assets (organically or through acquisition) when a firm is overpriced and contracting assets when a firm is underpriced (Polk and Sapienza, 2009). Empirical confirmation of a strong negative correlation between asset expansion and subsequent stock returns is found in a growing literature. 3 1 See Lintner (1971) and Rhodes-Kropf and Viswanathan (2004) for additional discussion of related theory. 2 Additional references include Cochrane (1991, 1996). Berk, Green, and Naik (1999), Gomes, Kogan, and Zhang, (2003), and Li, Livdan, and Zhang (2008). 3 See Fairfield, Whisenant, and Yohn (2003), Titman, Wei, and Xie (2004), Broussard, Michayluk, and Neely (2005), Anderson and Garcia-Feijoo (2006), Cooper, Gulen and Schill (2008), Polk and Sapienza (2008), Lyandres, Sun, and Zhang (2008), Xing (2008), and Lipson, Mortal, and Schill (2011). 3

5 One implication of our proposal is that the post-deal returns associated with acquisitions be completely subsumed by the respective total asset growth rates. This requires that the returns to organic firm growth be no different than those associated with acquired firm growth. Other implications respond to the existing theory for post-merger returns. Our explanation requires that the glamour firm effect used to support the Rau and Vermaelen extrapolation story and the stock-deal effect used to support the Sheifer and Vishny medium of payment story be simply due to cross-sectional differences in the magnitudes of the growth rate for glamour deals and stock deals. Thus our proposal emphasizes that while the occurrence and characteristics of takeovers is pedestrian to returns, the magnitude of firm asset expansion is not. An asset-growth-based view of the merger effect alters other notions of the post-takeover effect, including other tests of the Shleifer and Vishny model by Dong, et al. (2006), the earnings management window dressing of Erickson and Wang (1999) and Louis (2004), the notion that high valuations increase managerial discretion such that managers make poor acquisitions when they run out of good ones (Rau and Vermaelen, 1998; Jensen, 2003; Moeller et al., 2005), and the Roll s management overconfidence hypothesis discussed in Malmendier and Tate (2005, 2008). Massa and Zhang (2009) examine the impact of time-variation in mutual fund flows to growth and value style funds that create windows of opportunity in acquisitions. In all cases, the predictions of these theories are consistent with a more fundamental model where a firm expands assets when their cost of capital declines. In this paper we test this fundamental model. In effect, we examine the basis of the post-deal return literature by testing the supposition that merger-firm returns are unique. Using a broad sample of U.S. merger firms from the Thomson Reuters SDC Platinum Mergers and Acquisitions dataset over the 1981 to 2007 period, we find, as do others, that 4

6 acquiring firms are associated with poor post-deal returns. We then compare the return performance of acquiring firms to a sample of control firms that have not performed a merger in the past three years but have experienced similar levels of asset expansion. We find that the returns for merger firms are no different from the returns observed by firms expanding their balance sheet in similar magnitude through other non-merger related activities. With respect to post-deal returns, there appears to be nothing unique about acquisitions. Rather the defining characteristic that explains the cross-section of returns for these firms is the magnitude of firm asset growth. In short, the asset growth effect appears to completely subsume the long-run merger effect in stock returns. We run a variety of tests to confirm this result. For a sample of 5550 acquirer firm years, the 12 month cumulative abnormal return, starting in January after merger completion, is 4.5%, sizably below the 5.1% average annual rate for Treasury Bills or the 10.1% average annual rate for size and book-to-market-matched firms over this period. Non-acquirer control firms matched on the asset growth rate of the acquirer firms experienced a nearly identical 4.3% average annual rate over the same period. To control for cross-correlation in acquirer firm returns, we form acquirer portfolios where firms enter the portfolio in January after acquiring a firm in the past calendar year and exit after three years of no merger activity. We observe that the acquirer portfolio underperforms traditional benchmark portfolios on an absolute and risk-adjusted basis across various weighting schemes, yet there is no abnormal performance when compared to asset growth rate-based benchmarks. To take a closer look at the interaction between asset growth and acquisition, we sort acquirer firms by asset growth rate for the acquisition year and find that their subsequent returns behave closely to those of non-acquiring firms that experience similar levels of asset growth. For example, acquirers that shrink their assets actually do well in 5

7 subsequent years (average returns of 14.5%) as do other firms associated with non-acquisition asset contraction (average annual returns of 15.8%). We find similar effects with the analysis of operating returns as that of stock returns. We also investigate the leading cross-sectional relations established in the merger literature that poor returns are concentrated in deals that are paid for with stock, completed by firms with glamour book-to-market ratios, or transacted during years with high market-wide valuations. We find that once again the asset growth effect seems to explain these effects in that once the firm s growth rate in total assets is controlled for, these relations disappear. Our evidence refocuses previous research by emphasizing that it is not the deal characteristic that matters, but rather that firms with these deal characteristics tend to grow at similar rates. For example, it is not so much that stock or glamour-firm deals tend to underperform but rather that stock or glamour-firm deals tend to be associated with higher levels of balance sheet expansion and it is the rate of total asset expansion that is more precisely associated with poorer performance. These results fundamentally alter the interpretation of the literature of postacquisition returns in Rau and Vermaelen (1998) and Shleifer and Vishny (2003). Overall, we conclude that the long-run returns associated with takeovers are not unique, but are part of a broader effect related to asset growth. Given our evidence, any explanation that explains mergers must explain the returns associated with any corporate expansion regardless of whether a firm acquires an operating business or any other corporate asset. Our paper emphasizes that the center of the debate is more powerfully focused on the broader phenomenon, the asset growth effect in returns. This work also allows us to sharpen our understanding of the asset growth effect. Cooper, Gulen, and Schill (2008) document some variation in the return effects associated with 6

8 growth from different parts of the balance sheet. In their tests, growth in cash and retained earnings generated little return effects while growth in current operating assets, PP&E, and debt generated quite large return effects. In our tests we observe no variation in return effects between acquired growth and greenfield growth. Some explanations for the asset growth effect rely on time variation in asset risk. The Berk, Green, and Naik model, for example, requires that the exercise of risky growth options reduces firm risk. Although still valid, such explanations seem harder to justify with mergers where the firm is acquiring operating business that may not retain such variation in risk level between exercised and unexercised growth options. The average monthly return for a portfolio of acquiring firms and non-merger control firms is 0.36% and 0.33%, respectively. The average monthly return for Treasury Bills over this same 27 year period is 0.46%. It is hard to justify such performance with any risk-based model. If mispricing is the more plausible explanation, we document that the market appears ambivalent between how it prices organic and acquired growth. 4 It follows that managers are also ambivalent between opportunistic acquisition and greenfield investments. 5 Tests by Li and Zhang (2010) and Lipson, Mortal, and Schill (2011) support the mispricing interpretation of the asset growth effect. We leave for others to sort out these different explanations for the asset growth effect (see Anderson and Garcia-Feijoo, 2006; Chan, Karceski, Lakonishok, and Sougiannis, 2008; Lipson, Mortal, and Schill, 2011; and Li and Zhang, 2011). 4 Others that have examined acquisitions in the context of the asset growth effect include Cooper, Gulen, and Schill (2008) and Chan, Karceski, Lakonishok, and Sougiannis (2008). Both studies maintain that the asset growth is robust to excluding firms that growth through acquisitions. 5 This observation is consistent with what Jensen (2005) calls the agency costs of overvalued equity. He suggests that this problem in the mergers and acquisition market is only a part of the total costs this phenomenon has imposed on firms and society. It extends also to greenfield investments and other major business decisions. In contrast, Harford and Li (2007) examine the management wealth gains to firm expansion. Curiously, they find that CEO wealth improves for corporate acquisitions but not for major capital expenditures. 7

9 The paper is organized as follows. Section 2 describes the sample. Section 3 reports the main empirical tests on returns. Section 4 reports tests on operating returns. Section 5 concludes. 2. Data Our tests require a comprehensive examination of merger activity over our sample period. We use the Thomson Reuters SDC Platinum Mergers and Acquisitions dataset to classify firms. This dataset is widely used for league tables for underwriters who maintain strong incentives to have deals recognized. Due to the importance of the record and the incentives of underwriters to report, it is understood to be nearly comprehensive of all underwritten corporate merger activity. Based on data limitations with the beginning of a comprehensive record we establish a sample period from 1981 to Our sample of firms is all U.S. CRSP-Compustat non-financial firms over the 1981 to 2007 period, domiciled in the U.S. and with a share price greater than $5 at the end of the pre-merger calendar year. Our unit of observation is firm year. Using the Thomson Reuters dataset, a firm year is classified as a merger firm year if during the calendar year at least one M&A transaction of majority or remaining interest by a U.S. firm was reported. For our purposes, we follow the literature in not considering tender offers as mergers, although the number of tender offers is so small this decision has no bearing on our results. To avoid capturing tiny mergers and thus make the tests more discriminating, we classify a firm year as a merger if the annual sum of merger deal value (or the sum of its merger deal values in the case of multiple acquisitions) exceeds five percent of the beginning of year market capitalization of the acquiring firm. We exclude firm 8

10 years as merger or control firms if another confounding transaction occurs in that period (specifically, tender offers, spin offs, and asset sales). Panel A of Table 1 provides the frequency distribution of the sample by year. Panel A contains the number of firm-year observations that fall in a particular category. A firm is classified as control firm if it is not listed as having completed any asset restructuring (including a merger) in a particular year nor in the previous three years. We consider a firm as having had an asset restructuring event if such an event is identified by the Thomson Reuters dataset or the firm acquisitions item from the statement of cash flows is greater than zero. Our classified sample is comprised of 5,550 acquirer firm years and 24,452 control firm-year observations. Of the acquirer firm years, 30% are classified as stock deal years and 34% are classified as cash deal years. We classify firms that pay with a combination of both cash and stock as stock deals. For firms that affected multiple deals in a year with different consideration, we drop firm years that had both a stock deal and a cash deal. The payment type of one third of the acquisition events are not reported so these firms are not used in tests where the payment classification is used. Panel B of Table 1 provides firm characteristics detail by category. Deal Value is as reported by Thomson Reuters. Market value of equity is from CRSP and is shares outstanding multiplied by price as of December of the calendar year during which we count deals. The book to market ratio is computed as defined in Davis, Fama and French (2000) and is the book value divided by market value of equity as of December. Book value of equity is stockholders equity (SEQ) minus preferred stock plus deferred taxes and investment tax credit (TXDITC). Preferred stock is computed in the following order, depending on data availability: redemption value (PSTKRV), liquidation value (PSTKL) or stock capital (PSTK). Total assets corresponds to the variable AT from Compustat. The asset growth rate is defined as the percentage growth in 9

11 total assets net of cash. Since Cooper, Gulen and Schill (2008) find that cash growth doesn t bear any meaningful effect on returns, we subtract growth in cash from total asset growth, to provide a more powerful benchmark. The removal of cash from total assets eliminates asset growth that is generated without any real investment such as an equity or debt offering where the proceeds are not yet invested in real assets. Specifically, we compute asset growth as indicated in the equation below where is the change operator, total assets is the variable AT from Compustat, and cash the variable CHE. (1) We also report total Deal Value as a percentage of the market value of equity as of December, and as a percentage of total assets. Finally, age is the number of years since the firm reported assets for the first time on Compustat. We observe that acquisition firms are on average larger and faster growing than the control firms. The size differential is greatest among the cash deals and the growth rate differential is greatest among the stock deals. The book-to-market ratio is more similar across merger firms and control firms with ratios of 0.56 and 0.61, respectively. 3. Empirical analysis 3.1. Event study over the calendar year following the merger We begin our tests with a simple event study of returns of acquirers over the calendar year following the completion of the merger. In this test we estimate the average return for firms in the calendar year following a merger event. The estimates are reported in Table 2. In Panel 10

12 A, we report the mean returns by month starting in January following merger completion. We begin in January because we must have the annual financial statements in order to match the merger by asset growth rates over the merger year. Over the year following the merger, the monthly returns by calendar month vary from a mean high of 2.4% in January to a mean low of - 2.0% for July. These monthly seasonal effects are mostly due to broad seasonal effects in the market over this sample period. An estimate of the monthly mean return for the CRSP market indices generates similar monthly patterns in returns. The cumulative monthly return for the acquirer sample is 4.5%. Investors in the acquirers earn on average less than they did on U.S. Treasury Bills that maintained an average annual return over this same period of 5.1%. To document the abnormal returns of merger firms, we follow Fama and French (1993) and Rau and Vermaelen (1998) and match each merger firm with a portfolio of control firms based on a size and book-to-market ratio of the merger firm in a given year. To create these benchmarks we use the full sample of firms (merger firm, other transaction firms, and control firms) to create annual size and book-to-market ratio breakpoints. For size breakpoints, we follow Fama and French (2008) and split the sample at 20 th and 50 th percentiles based on NYSE listed firms. Book to market is split into same-sized quintiles using the whole sample. We use these annual breakpoints to form control firm portfolios based on all firms within the same annual size and book-to-market groupings. The control firms used are firms that have not had any merger or other asset restructuring event in the current year nor in the past three years. For this control group the cumulative return over the same subsequent calendar year is 10.6%. The difference between these two cumulative return estimates is -6.1% and is highly significantly different from zero with a t-statistic of As we look at the returns by month we note that 11

13 the difference in returns is significantly negative for seven of the twelve months (January, February, March, July, August, September, and October). We propose an alternative matching procedure based on the asset growth rate of the acquirers. We establish 10 annual breakpoints for the full sample of firms based on the firm asset growth rate. Using these breakpoints, we form portfolios of control firms that mirror the same asset growth rate as the acquirers. Matching on asset growth rate, the cumulative 12-month return is much lower at 4.3%. The annual return is almost identical to the 4.5% return of the acquisition firms (the t-statistic on the difference is insignificant at 0.27). The monthly means suggest that although returns in January are systematically low the strong returns in March, April, and December make up for it. This first test suggests that the subsequent returns of firms that complete acquisitions are low but not significantly lower than other non-acquisitive firms that grow their assets at the same growth rate. Figure 1 reports simple average cumulative abnormal returns for acquirers matched on portfolios of firms based on the two matching criteria described above: (1) market to book and size, and (2) growth. The figure shows that over the subsequent calendar year, the control sample criteria maintain a dramatic effect on the inference. The acquirers show strong underperformance when compared against a size and book-to-market ratio-based control sample but nothing when compared against an asset growth rate-based control sample. Panel B of Table 2 provides a sense of the year-by-year magnitudes of acquirer and control-firm returns. The panel reports the mean cumulative 12-month difference in returns for the acquirers and the two control groups by calendar year over the sample period. The table emphasizes the systematic tendency for acquirers to underperform size and book-to-market ratio 12

14 portfolios whereas there is less systematic difference in returns between acquirer returns and the asset-growth matched portfolio returns on a year-by-year basis Portfolio returns In the next set of tests we refine the Table 2 work by forming calendar-time portfolio tests of the acquirers to remove any cross-correlation bias in the standard errors (Mitchell and Stafford, 2000). In these tests a portfolio of acquirers is formed where firms enter the acquirer portfolio at the end of December of the year a merger is completed and exit the portfolio three years later. The firms in the portfolio are weighted in three ways: equal-weighted, valueweighted by the firm s market capitalization at the end of the merger year, and merger-volumeweighted where the mergers are equally-weighted within a specific month but across months all mergers are weighted by the aggregate number of firms completing an acquisition during the previous year. The acquirer portfolio excess returns are reported in Table 3 with monthly equalweighted returns of 0.36% (Panel A), value-weighted returns of 0.40% (Panel B), and mergervolume-weighted returns of 0.30% (Panel C). As a first benchmark, the average monthly return on Treasury Bills from 1981 to 2007 based on the monthly estimates by Ibbotson and Associates is 0.46%. In all cases the acquiring firms underperform U.S. Treasury Bills over this 27 year period. In a similar manner, we construct a size and book-to-market ratio matched portfolio returns by forming a benchmark portfolio that matches on size (using the NYSE 20 th and 50 th percentile breakpoints) and book-to-market ratio (based on full sample book-to-market ratio quintiles) at the end of the calendar year in which the merger was completed. This benchmark portfolio is held for three years. The difference in monthly returns between the acquirer portfolio 13

15 and the size and book-to-market ratio-matched portfolio is -0.27% (t-stat = -3.53), -0.21% (t-stat = 1.84), and -0.36% (t-stat=-4.29), respectively, for the equal-weighted, value-weighted, and merger-volume-weighted portfolios. These findings are again consistent with past evidence of acquirer-firm underperformance. As an alternative benchmark, we form asset growth control firm portfolios that match the acquirer sample in the same manner as the size and book-to-market ratio matching procedure conducted in the last test. The asset growth control firm portfolios are delineated based on asset growth deciles based on annual full sample breakpoints. The control firm portfolio is thus balanced to maintain the same average asset growth rate as the merger portfolio. We find that these benchmark portfolio returns are systematically lower, and in fact lower than the acquirer portfolio returns. The difference in returns of the acquirer portfolios and the asset growth portfolio returns, however, are no longer statistically significantly different, 0.04% (t-stat=0.35), 0.07% (t-stat=0.38) and 0.07% (t-stat=-0.55), respectively for the equal-weighted, valueweighted, and merger volume-weighted portfolios. We repeat our analysis using risk-adjusted portfolio returns using the Fama-French three factor model. The conclusions are similar: acquirer portfolio returns are systematically poor using standard benchmarks, yet comparable to asset growth rate-matched portfolio returns. There is some concern regarding the effect of accounting treatment for mergers (pooling versus purchase merger accounting). The accounting method used to record the merger affects the asset growth rate of the bidding firm. To ensure that our results are not driven by accounting method, we observe that over 90% of the bidder firms use purchase accounting and that for this subsample of firms our results remain unchanged. 14

16 In looking at the cross-section of asset growth rates associated with acquirers, we observe that the distribution of the asset growth rate of acquiring firms is dispersed across a wide range of values. Although most acquiring firms tend to grow their balance sheet substantially in the acquisition year, some firms actually shrink their asset base in the year of acquisition. To investigate the independent return effects to acquiring firms across various rates of asset growth, we form portfolios based on quintiles of asset growth. With the independent sorts we generate a clean test of the dual effects of merger and asset growth. We begin this analysis by sorting the acquirer firms and control firms into five groups based on annual breakpoints of the full sample population of asset growth rate. Based on this categorization, the average number of merger firms in the five annual asset growth quintile groups is 7.3, 9.3, 19.1, 45.6, and 124.3, respectively, by order of increasing growth rate. Table 4 reports various firm characteristic statistics for the acquirer and non-merger firms by asset growth rate quintile. The mean of the yearly median asset growth rate is -9%, 2%, 9%, 19%, and 57%, respectively for the acquirers, by order of increasing growth rate. The respective values for the non-merger firms are reasonably comparable at -7%, 2%, 8%, 17%, and 40%. To better appreciate the source of the asset growth rate, we decompose the asset growth rate into four asset-side components and separately four financing-side components. The decomposition follows that of Cooper, Gulen, and Schill (2008). On the assets side we compute growth in cash (Compustat data item CHE), current assets, property plant and equipment (Data item PPENT) and other assets. We present the cash growth rate in this panel even though as described earlier we do not use it in the total asset growth rate calculation. Each of the growth measures is computed as the yearly change in the specific item scaled by total assets. Current assets is computed net of cash (Data item ACT minus CHE). The 15

17 other assets line item is computed as total assets minus property plant and equipment minus current assets minus cash. On the liabilities and equity side we compute the change in stock, retained earnings (Data item RE), debt, and current liabilities, each scaled by total assets. Stock is common stock (Data item CEQ) plus minority interest (Data item MIB) plus preferred stock (Data item PSTK) minus retained earnings (Data item RE). Debt is long term debt (Data item DLTT) plus short term debt (Data item DLC). The current liabilities item is computed as total assets (Data item AT) minus retained earnings (Data item RE) minus stock minus debt. The majority of the balance sheet growth comes through increases in non-cash current assets such as accounts receivable and inventory for both the merger and non-merger groups. For example, among the high growth mergers, the decomposition of the asset growth rate was 0% cash growth, 18% non-cash current assets growth, 13% PPE growth, and 19% other asset growth. For the non mergers, the numbers were respectively, 2% cash growth, 18% non-cash current assets growth, 12% PPE growth, and 2% other asset growth. The comparison shows that other asset growth, for example growth in goodwill, is, as expected, not as large a part of nonmerger firm growth, and cash growth is a much larger component in the non-merger growth. For the low asset growth group, the decomposition of asset growth for the merger firms was 0% cash growth, -5% non-cash current assets growth, -1% PPE growth, and -2% other asset growth; whereas the figures for the benchmark firms were 1% cash growth, -4% non-cash current assets, -1% PPE growth, and 0% other asset growth. Clearly non-cash current assets is a major driver of asset growth across both groups of firms. On the financing side, the high growth merger firms finance the growth with 12% contribution from operating liabilities, 21% debt financing, 11% equity financing, and 6% 16

18 retained earnings. For the control firms the respective figures are 9%, 5%, 9%, and 9% for operating liabilities, debt financing, equity financing, and retained earnings. 6 In Panel B, we report return statistics for the portfolios over the sample period. The first set of tests report mean equal-weighted portfolio return values by asset quintile. In increasing order of asset growth rate, the mean monthly return values are 1.21%, 0.91%, 0.91%, 1.16%, and 0.55%, respectively, for the acquirer sample, and 1.32%, 1.18%, 1.13%, 1.07%, and 0.56%, respectively, for the non-merger group. Both acquirer firms and non-merger firms experience a strong asset growth effect, with the difference in portfolio returns between the low growth and high growth portfolio returns being statistically different from zero: 0.66% (t-stat=2.67) for the merger firms and 0.75% (t-stat=5.64) for the non-merger firms. The difference in returns across these portfolios between acquirer and non-merger firms, however, is not significantly different from zero. The t-statistics for the difference in returns between acquirer and non-merger firms is -0.44, -1.39, -1.62, 0.73, -0.10, respectively, for the low growth to high growth quintiles. The returns of acquirers, particularly those with larger asset growth do not appear to underperform their non-merger firm counterparts. The panel also reports results for value-weighted and merger-weighted portfolios. The results generate the same inference. As a test-case we perform the same exercise based on a book-to-market ratio-based sorting (Panel C). In this case the difference in returns is much larger with equal-weighted 6 To provide a sense of the type of control firms used in the sample consider the following control firms from 2006: Ampco Pittsburgh, a specialty steel manufacturer, grew assets at a rate of 51% primarily due to proceeds from asbestos insurance claim settlement. Goodrich Petroleum, an oil and gas concern grew assets at a rate of 66% due to the purchase of oil & gas properties funded by long-term debt. Hansen Natural, a beverage company, grew assets at a rate of 50% due to strong business profitability with return on assets of 60%. Jones Soda, another beverage manufacturer, grew assets at a rate of 81% due to increases in accounts receivable and other accrual accounts funded through an equity offering. Las Vegas Sands, a hotel and gaming enterprise, grew assets at a rate of 75% due to the construction of new properties in Macao and Las Vegas and funded with new debt. Spartan Motors, a specialty vehicles manufacturer, grew assets at a rate of 53% through expansion of receivables, inventory, and property, plant and equipment funded by debt, stock option exercises, and high profitability. 17

19 acquirer less non-merger portfolio returns of -0.09% (t-stat=-0.61), -0.39% (t-stat=-3.01), -0.35% (t-stat=-2.85), -0.24% (t-stat=1.91), and -0.46% (t-stat=-3.02), respectively, across the book-tomarket quintiles in increasing order. On a value-weighted basis this effect is largely muted, whereas it is similar on a merger volume-weighted basis. Clearly, the asset-growth rate controls once again do better at explaining the acquirer-firm effect. To provide evidence of the return performance of the portfolios over a larger event window, we plot the mean monthly returns of the two extreme asset growth portfolios (high growth and low growth) for both acquirer and non-merger firms for the seven years surrounding the year of the sort. These plots are reported in Figure 2. We observe that the high growth acquirer portfolios behave in a similar manner as non-merger portfolios with very high returns prior to the sorting year followed by a similar pattern of low returns following the sorting year. The opposite is true for the low growth firms with relatively low returns prior to the sorting year and high returns after. The return reversal pattern of the plot appears very similar to that of Cooper, Gulen, and Schill (2008) and suggests that controlling for the firm asset growth rate the returns to acquirers do not appear to be substantially different than the returns of firms that grow at similar rates without acquisitions Cross-sectional regression estimates We now conduct a series of tests of the merger effect using an alternative full panel Fama-MacBeth cross-sectional approach. In this test we estimate monthly cross-sectional regressions with firm return on the left-hand side of the regression and log-book-to-market ratio and log market capitalization on the right-hand side as control variables, and log of one plus Deal Value relative to market value of equity (Deal/Cap) and Merger Dummy as test variables in turn. 18

20 The monthly returns are from January to December of a given year, while independent variables are as of the year before. The variable Merger Dummy is an indicator variable for the firm having a deal value greater than 5% of the market value of equity. If the firm effected multiple mergers in a given year, then deal value is the total value spent in all deals performed in that year. In Table 5 we report the Fama-MacBeth time-series average of the monthly coefficients. In this first regression, we observe that the Deal/Cap variable is highly negatively correlated with subsequent returns. Firm returns are lower for those firms that experienced larger mergers in the preceding year. In regression 2, we observe that the merger dummy variable is also negative and statistically significant with merger firms experiencing a significant -0.4% less in returns each month. Next, we add the log of one plus the firm s asset growth rate to the right-hand side of the regression. In regression 3 we find that the coefficient on the asset growth rate is negative and highly significant following a specification similar to Lipson, Mortal, and Schill (2011). When the firm asset growth rate is added to the regression, however, the explanatory power of the two merger variables disappears. Both the coefficient on the merger dummy and the deal percentage (Deal/Cap) are now no longer significantly different from zero. As an additional test, we add an interaction term to the regressions to examine whether the merger effect maintains some joint properties with asset growth. In both specifications (regressions 5 and 6) we find no significant correlation with either the independent or interaction merger variables, while the asset growth rate maintains large negative explanatory power. 19

21 3.4. Exploring cross-sectional effects in merger returns We use the Fama-MacBeth regressions to examine the explanatory power of form of payment in the cross section. We form dummy variables that indicate whether the buyer paid for the deal at least partly with stock (Stock Dummy) or all with cash (Cash Dummy). If the firm affected multiple deals in the same year, where some of the deals were paid for at least in part with stock while others were completely paid for with cash, we set the form of payment to missing. We interact these variables with the merger variables to test for cross-sectional effects. Consistent with the literature (Loughran and Vijh, 1997 and Savor and Liu, 2009), we find that the abnormal returns associated with acquirers tend to be concentrated among deals paid for with stock. In our tests, the coefficient on the stock dummy interaction variable is negative and significantly different from zero in both specifications (regressions 7 and 9), but not the cash dummy. This result follows the predictions of Shleifer and Vishny (2003). We now add the asset growth rate to the right-hand side of the regression and re-estimate the equation. Again, the coefficient on the asset growth rate is negative and highly statistically significant, but the addition of the regressor eliminates the cross-sectional stock effect (regressions 8 and 10). The coefficient on the stock dummy interaction is no longer significant when the asset growth rate is included. In Panel B of Table 4 we repeat the same exercise with regressions weighted by firm market capitalization. We again observe the same effect that the explanatory power of the merger variables (including the stock payment effect) is unique only when the broader asset growth rate variable is not included in the regression. We extend our look at the cross-sectional effects in merger returns by forming calendartime portfolio returns based on various characteristics of the acquirers. Our first test follows on 20

22 from the Fama-MacBeth tests to look at portfolios of acquirers that use stock and those that use cash to pay for the deal. We report the characteristics of these portfolios in Panel A of Table 6. We observe that the acquirers using stock to pay for deals tend to be somewhat smaller, younger, with a lower book-to-market ratio, yet the deals tend to be larger relative to the acquiring firm. In addition, it is particularly important to note that the asset growth rate of firms who pay for deals with stock is 42% whereas the asset growth rate of firms who use cash is 28%. We find that the mean monthly portfolio return of the stock deals is 0.37%, while that for cash deals is 0.58%. When we compute abnormal returns based on size and book-to-market ratio matched control firm portfolio, we observe a -0.28% (t-stat = -2.14) for the stock deals and -0.06% (t-stat = -0.32) for the cash deals. This finding is similar to our above finding in that the abnormal performance appears to be concentrated among stock deals. When we control for the higher asset growth rate among the stock deals, however, we find that the underperformance disappears. The abnormal return based on asset growth rate portfolio control is 0.05% (t-stat = 0.49) for stock deals and 0.10% (t-stat = 0.45) for cash deals. The cross-sectional effect disappears once we have control for the higher asset growth rate of the stock deals. These results fundamentally alter the interpretation of the literature of post-acquisition returns in Shelifer and Vishny (2003). Rau and Vermaelen (1998) observe that merger firm underperformance is concentrated in low book-to-market ratio glamour firms. They claim this finding is consistent with managers and the market over-extrapolating the acquiring firm s valuation ratio with respect to the target firm. In Panel B we examine the interaction between the value effect and the merger effect. We begin by sorting the acquirers into five book-to-market ratio quintiles. We report the mean monthly return for firms within each book-to-market ratio quintile. The mean returns are 0.22%, 0.26%, 0.40%, 0.56%, and 0.39% for the five quintiles in order of increasing book-to-market 21

23 ratio. The evidence is consistent with Rau and Vermaelen: the worst returns are concentrated among glamour acquirers. If we match the acquirer returns by size and book-to-market ratio, we find that the acquirers continue to underperform the benchmarks. In looking at the firm characteristics by quintile we observe that the asset growth rate is largest for the glamour firms and smallest for the value firms. Specifically, the asset growth rate decreases from 45% for the glamour acquirers to 41%, 29%, 31%, and 25% for the value acquirers. We now match each acquirer with an asset growth rate matched portfolio. When we subtract the mean asset growth rate matched portfolio return from the mean acquirer portfolio return we find that none of the differences in returns are significant with the t-statistic for the low book-to-market ratio glamour firms returning the same stock performance regardless of whether or not the firm is an acquirer. Bouwman, Fuller, and Nain (2009) observe that acquirers that execute their acquisitions during non-bust years (years associated with relatively high price-to-earnings ratios) tend to do particularly poorly. We use our portfolio set up to create merger portfolios based on the state of market valuation. In effect each acquisition is categorized by year into three market states: boom, neutral, and bust years based on the market price-to-earnings ratio defined by the S&P 500 index. 7 We use the annual average of monthly data on market price-to-earnings ratio. Because the market price-to-earnings ratio has a strong upward trend over the sample period, we detrend the data, in the same spirit as Bouwman, et. al. We first remove the best straight-line fit from the price-to-earnings ratio series, and then we subtract from the residuals their five year moving average. The top half of the above-average years are classified as high valuation markets and the bottom half of the below average years are classified as low-valuation markets. All other years are classified as neutral-valuation markets. 7 We obtain the market price-to-earnings ratio data from Robert Shiller s web site ( 22

24 We report the mean returns for the portfolios organized around these three classifications. The boom year mergers are associated with monthly returns of 0.39%. When matched against size and book-to-market ratio matched control portfolios, the monthly abnormal return is -0.34% (t-stat = -3.21). Mergers during neutral years also tend to underperform the benchmark with monthly abnormal returns of -0.33% (t-stat = -3.17). Mergers during bust years are not associated with any abnormal return. These findings are comparable to those of Bouwman et al. Following our other tests, we now match these with asset growth rate matched portfolio returns. In this case we find again that the asset-growth rate matched portfolio returns are also low for the boom and neutral years such that the abnormal return for these years is no different between merger firms and non-mergers firms. For the bust years, we find that the asset growth rate matched portfolios do even worse than the merger firms with an abnormal return of 0.45% (t-stat = 3.00). If anything, we observe that merger firms tend to do well relative to firms with similar asset growth rates. 4. Operating returns As a last set of tests, we examine the operating performance of acquiring firms around acquisitions. Our tests are motivated by other event studies that have looked at operating performance of acquiring firms around mergers (Ravenscraft and Scherer, 1989; Healy, Palepu, and Ruback, 1992; and Bouwman, Fuller and Nain, 2009). We define operating returns as operating income before depreciation (Compustat data item OIBDP) divided by total assets (Data item AT). The sample of operating returns is winsorized at the 1 percent and 99 percent levels. We create an industry and performance matched sample as a performance benchmark following the approach recommended by Barber and Lyon (1996). Specifically, we restrict the control 23

25 firms to be those non-merger firms with the same two-digit SIC codes and total assets between 0.25 to 2.0 times that of the sample firm. From this group we make two alternative matching restrictions. For our baseline control, we follow practice and match on operating returns within 5 percentage points of the sample merger firm in the year of the merger and use the median firm from this group as our control firm. For our test control, we match on asset growth rate within 5 percentage points of the sample merger firm in the year of the merger and use the median firm from this group as our alternative control firm. 8 We present results in Table 7 for the full merger sample (Panel A), for the sample where the consideration for all mergers by the firm in the year are at least partially financed with stock (Panel B) and for the sample where the consideration for all mergers by the firm in the year are entirely with cash (Panel C). We document operating performance for seven years centered at the merger year (Year 0) for the merger firm, the benchmark, and the difference between the two. We compute sample means for the merger firms, the two sets of control firms, the differences, and t-tests that the differences in mean operating returns are different from zero. For our baseline control firms, we find that, merger firms tend to outperform their benchmark during pre-merger years, but modestly underperform during the post-merger years. This pattern appears correlated with the consideration paid. For stock deals, the pre-merger performance is normal (note we do match on merger year performance), whereas the post-merger years are associated with statistically significant poor performance by merger firms. The opposite is the case for the cash deals. 9 These results are highly sensitive to the benchmark. 8 If for a particular merger firm, there are no potential control firms within the specified criteria, we relax the industry constraint, but require firms to be within 10% in regards to size, if we are still unable to find eligible matches we also relax the size constraint. Our operating performance matching criteria is similar to that of Bouwman, Fuller and Nain (2009). 9 These results are somewhat different from Bouwman, Fuller and Nain (2009) due to discrepancies in the samples, most notably that their sample excludes subsidiary acquisitions. When we exclude subsidiary acquisitions from our sample, our results mirror those of Bouwman, et al. 24

26 When we substitute the asset growth rate-matched control firms, the merger firms actually outperform their benchmark following the merger for the full sample. 10 We conclude that the post-deal underperformance in operating returns is also subsumed by the asset growth effect. 5. Conclusions We examine the returns associated with corporate acquisition deals and the well documented finding of poor long-run stock returns. We propose that, with respect to post-deal returns, acquisitions in and of themselves are benign corporate events. As an alternative explanation, we following existing investment theory and maintain that it is the firm s asset expansion that commonly accompanies takeovers that is associated with the pervasive return effects documented in this literature, rather than anything unique to the takeover per se. In support of this explanation, we observe that the return performance of acquiring firms is, in fact, no different from that of firms that grow their balance sheet organically at the same rate. The asset growth effect completely subsumes the long-run merger effect in stock returns. Moreover, we find that cross-sectional effects such as the relatively poorer returns for stock financed deals and glamour firm deals appear to be more precisely due to systematic variation in firm asset growth rate, rather than other unique characteristics of the deal or acquirer. The return effects associated with stock deals and glamour firm deals are simply due to a tendency for these deals to maintain systematically higher assert growth rates. In addition, we find that the cyclical effect of merger firms to maintain particularly low returns during high market valuation years also disappears after controlling for asset growth. 10 It is worth noting that some of the underperformance of the asset growth control firms is due to the higher persistence of asset growth rate for these firms. In effect, the control firms continue to expand the denominator at a faster rate which possibly depresses the operating return. 25