Liquidity benefits and long-term acquirer performance

Transcription

1 Liquidity benefits and long-term acquirer performance Ning Gao Weimin Liu ABSTRACT Mergers are significant liquidity-enhancing events. Consistent with this hypothesis, we document persistent and significant post-merger enhancements in acquirers liquidity and their exposure to liquidity risk. Acquirers liquidity gain is attributable to improved information environment and shareholder base. The commonly adopted size and book-to-market matching cannot capture acquirers post-merger exposure to liquidity risk. This mismatching is responsible for acquirers low post-merger long-term performance documented in previous studies. After accounting for liquidity risk, acquirers show normal post-merger performance. Keywords: Liquidity risk; acquirer; long-term performance; decomposed buy-and-hold calendar time approach; size and book-to-market matching JEL classification: G12; G14, G34 Ning Gao is from Manchester Business School Weimin Liu is from Nottingham University Business School We gratefully acknowledge helpful comments by Constantinos Antoniou, Michael Brennan, Tim Loughran, Christian Lundblad, Erik Stafford, Huainan Zhao, Norman Strong and Karin Thornburn. We have benefitted from presentations at the EFA 2008 Athens meeting, EFMA 2008 Athens meeting, FMA 2008 Orlando meeting, the University of Nottingham, Shanxi University, and the University of Nottingham Ningbo China. All errors are ours.

2 Liquidity benefits and long-term acquirer performance ABSTRACT Mergers are significant liquidity-enhancing events. Consistent with this hypothesis, we document persistent and significant post-merger enhancements in acquirers liquidity and their exposure to liquidity risk. Acquirers liquidity gain is attributable to improved information environment and shareholder base. The commonly adopted size and book-to-market matching cannot capture acquirers post-merger exposure to liquidity risk. This mismatching is responsible for acquirers low post-merger long-term performance documented in previous studies. After accounting for liquidity risk, acquirers show normal post-merger performance. Keywords: Liquidity risk; acquirer; long-term performance; decomposed buy-and-hold calendar time approach; size and book-to-market matching JEL classification: G12; G14, G34 1

3 1. Introduction Mergers are among the largest and most visible corporate investment activities, where companies are supposed to combine their businesses to generate synergies that benefit shareholders. However, an unsettling observation by extant studies is acquirers long-term low performance post merger. 1 Some studies attribute the phenomenon to market inefficiency in that investors tend to overestimate efficiency gains of mergers (Jensen and Ruback, 1983; and Loughran and Vijh, 1997), overreact to acquirers prior strong performance (Mitchell and Stafford, 2000; Shleifer and Vishny, 2003; and Rau and Vermaelen, 1998), or underreact to poor investment (Roll, 1986). Others ascribe the post-merger low performance to inappropriate benchmark of estimating expected returns (Fama, 1998) or mismeasurement of acquirer long-term performance. 2 In this paper, we study the liquidity benefits acquirers gain from corporate mergers and pursue a rational explanation for the aforementioned unsettling observation. We hypothesize that the documented low post-merger performance is due to acquirers high liquidity characteristics and low liquidity risk ignored by previous studies when measuring acquirer performance. Our interest in liquidity is motivated by a growing literature demonstrating a significant liquidity premium and identifying liquidity as an important source of systematic risk for asset pricing. 3 Our hypothesis also stems from previous studies that relate information environment and shareholder base to firm liquidity. With regard to information environment, mergers are among the largest, most observable and most memorable corporate events. Mergers attract substantial attention from investors, regulators, news press and financial analysts, which intensify public scrutiny of the companies involved. Acquirers managers also have incentive to voluntarily provide information that would not have been disclosed otherwise so that they can convince shareholders and potential investors about the synergy of the deal. Managers also 1 Examples are?,?, and?. 2 There has been a debate over the choice between the long-term buy-and-hold method, advocated by? and?, and the monthly rebalanced traditional calendar-time approach, favoured by? and?. We use the term traditional calendar time approach here to distinguish it from our decomposed buy-and-hold calendar time approach, which we explain later. 3 For examples, see?, Reiganum (1990), Brennan and Subrahmanyam (1995), Amihud, et al. (1997), Eleswarapu (1997),?,?,?,?,?,?,?,?, and?. 2

4 originates more news coverage of the merger (Ahern and Sosyura, 2014). Years after mergers, companies are still mentioned, discussed and reviewed by market participants (e.g., Hewlett-Packard s merger with Compaq or AT&T s acquisition of NCR). Consequently, the information environment of acquirers improves during and after the event.? show that disclosure, which reduces information asymmetry, improves liquidity of a firm s securities. Balakrishnan et al. (2014) find that voluntarily disclosing more information improve liquidity. Studies by?,?,?,?, Easley and O Hara (2004), and? also suggest that better information environments benefit liquidity and lower the cost of capital. With respect to the link between shareholder base and liquidity, both Merton (1987) and? argue that investors tend to hold stocks that they are familiar with or, at least, aware of. An acquiring firm automatically inherit the shareholder bases of two combining companies. In addition, new information generated during the merger process make merging companies better known and in turn attracts new investors. Beston and Hagerman (1974) report a negative relation between stock bid-ask spread and the number of shareholders.? find that, when Japanese companies reduce their minimum trading units, an exogenous event that increase the base of individual investors, stock liquidity increases. Kadlec and McConnell (1994) also show that when a company commences listing on the New York Stock Exchange (NYSE), both its shareholder base and liquidity improve.? show that liquidity risk is an important channel through which investor recognition affects asset prices, consistent with Merton (1987) and Hou and Moskowitz (2005). To examine the power of liquidity risk in explaining acquirers post-merger performance, we set out by asking whether the size (i.e., market value, MV) and book-to-market (B/M) matching commonly used in previous studies is sufficient to match on acquirers liquidity. Given the multidimensional nature of liquidity, we employ eight liquidity proxies that reflect the four dimensions of liquidity: namely trading costs, trading quantity, trading impact on price, and trading speed. We find that acquirers are more liquid than their MV-B/M matches and mergers exacerbate the poor quality of matching: acquirers liquidity is mismatched in most post-acquisition years. In addition, liquidity beta significantly reduces after transaction for acquirers but not for MV-B/M matching firms. Based on Liu s (2006) liquidity augmented CAPM 3

5 (LCAPM), an average acquirer s post-merger liquidity beta is 0.45 lower than its MV-B/M match s. Ceteris paribus, this implies that acquirer s expected return is 0.25% lower per month than the MV-B/M matched counterpart s. 4 As mentioned above, two most prominent forces that improve acquirers liquidity are better information environment and expanded shareholder base. We verify this argument by showing that acquiring firms have significantly higher analyst coverage and more institutional holders than their MV-B/M matched non-event firms. Our regression results confirm that acquiring firms liquidity gains are positively associated with changes in the number of analysts following and the number of institutional investors. Our evidence demonstrates that one cannot control for the liquidity effect and liquidity risk sufficiently using MV-B/M matching. The evidence also suggests that the previously documented acquirers post-merger low performance can be potentially attributed to the improved liquidity that lowers acquirers exposure to liquidity risk. Indeed, we find that acquiring firms yield normal post-merger performance after adjusting for liquidity risk. 5 For the full sample, the sub-sample of mergers with stock-dominated financing (Loughran and Vijh, 1997), as well as the sub-sample of glamor acquirers (Rau and Vermaelen, 1998), the LCAPM-adjusted acquirer returns are neither economically nor statistically significant. Both glamor acquirers and acquirers of stock-dominated financing load negatively on the liquidity factor of the LCAPM. In particular, glamor acquirers in mergers have a liquidity beta estimate of (t = 5.55) and acquirers in stock-dominated mergers have a liquidity beta estimate of (t = 2.82). Accordingly, ignoring liquidity risk overestimates glamor acquirers expected post-acquisition performance by 0.157% per month (0.084% per month for acquirers with stock-dominated merger deals). Acquirers loadings on the? traded liquidity factor are also significantly negative. Moreover, matched on liquidity measures such as the price impact of?, the dollar trading volume of?, or the trading discontinuity measure of?, acquirers produce no significant long-term buy-and-hold abnormal return following mergers. They do, however, exhibit significant lower performance than their MV-B/M matches. Taken together, 4 The average value of the liquidity risk factor of the LCAPM is 0.55% per month over 7/1985 6/ We utilize two liquidity-risk based models: the Fama French three-factor model extended by the? traded liquidity factor and the LCAPM. To assess liquidity risk, we also use Sadka s (2006) non-traded liquidity factor constructed based on the variable component of price impact and Pastor and Stambaugh s (2003) non-traded liquidity factor. 4

6 by taking into account acquirers liquidity gains and low exposure to liquidity risk, we confirm our conjecture that neglecting liquidity risk causes the failure in explaining acquirer long-term low performance in post-merger years. This paper also deals with the concern about how to measure acquirer portfolio returns by applying a decomposed buy-and-hold calendar time approach (DBHCT), inspired by?. 6 To track investors wealth effect, the obvious metric is the buy-and-hold return over the post-merger period of interest. Nevertheless, researchers have noted that drawing statistical inferences from long-term buy-and-hold abnormal returns (BHAR) is difficult because of the correlation across event firms long-term returns. The crosscorrelation arises because acquisitions tend to cluster in time and using the long-term holding period results in overlapping event windows.? and? suggest using the traditional calendar-time approach, which involves averaging returns across event firms on a monthly basis. While the traditional calendartime approach avoids the cross-correlation problem from using buy-and-hold returns, it is subject to several criticisms. One is that the traditional calendar-time approach is inaccurate in measuring long-term shareholders wealth effect over the assumed holding period, usually three to five years. In addition, the traditional calendar-time approach implies a monthly weight rebalancing of the event-firm portfolio. Rebalancing not only contradicts the assumed long-term holding period but also incurs additional transaction costs and capital gain tax, making the monthly-rebalanced calendar-time approach a poor investment strategy that no investor would seriously consider ex ante. The DBHCT approach address the aforementioned disadvantages while preserving the advantages of both the buy-and-hold method and the rebalanced calendar-time approach. In particular, the DBHCT approach preserves the buy-and-hold property typical of a long-term investor s strategy and, at the same time, addresses the issue related to the cross-correlation of event-firm returns. The rest of this paper is organized as follows. The next section describes the sample, data and the decomposed buy-and-hold calendar-time approach. Section?? assesses acquirers liquidity. In Section 6 A parallel study by? also uses the approach to measure returns of the portfolio consisting of event firms conducting seasoned equity offerings (SEOs). 5

7 ??, we investigate the explanatory power of liquidity risk for acquiring firms long-run post-acquisition performance. Section 5 concludes. 2. Data and methodology 2.1. Sample and data Our sample contains NYSE/AMEX/NASDAQ public acquirers of completed mergers and tender offers drawn from the SDC M&A database. All the deals involve transfer of control rights (i.e., mergers and acquisitions of majority stakes according to the SDC definitions). All the deals are completed during the period July 1984 to June 2008 (to trace the post-event three-year performance, our return data end in 2011). The target companies are either exchange listed or private. Other restrictions imposed on the sample of acquirers are as follows: Similar to?, we require deal value be at least $10 million and no less than 1% of the acquirer s market value of equity at announcement. Both the announcement and completion dates are available from SDC. Acquirers have not made any Initial Public Offerings (IPOs), Seasoned Equity Offerings (SEOs), or similar mergers or acquisitions in the three years prior to deal announcements. This criterion is imposed because these confounding events can inflate the measured low performance of acquirers. The samples of IPOs and SEOs are from the SDC Global New Issue database. Acquirers have monthly returns and MVs from CRSP, and annual book equity data from COM- PUSTAT. We use monthly returns to estimate acquirer post-acquisition performance. We use MV and B/M to find acquirers matching firms. We classify acquirers into glamour, neutral and value acquirers based on B/M breakpoints of all NYSE firms. Information on means of payment, which is used to classify deals into stock- and non-stockdominated offers, is available from SDC. 6

8 The final sample contains 2,838 deals, made by 2,186 acquirers. We retrieve accounting and market data from the CRSP/COMPUSTAT merged database (CCM) over 1/ /2011 for the following items (data prior to July 1984, the beginning of the post-acquisition testing period, are required to assess acquirers liquidity evolution): (i) daily data including trading volume, number of shares outstanding, share price, and return; (ii) monthly data including MV and return; and (iii) annual accounting data including Total Stockholders Equity (SEQ), Total Common Equity (CEQ), Preferred Stock Carrying Value (PSTK), Total Assets (AT), Total Liabilities (LT), Balance Sheet Deferred Taxes (TXDB), Balance Sheet Investment Tax Credit (ITCB), Preferred Stock Redemption Value (PSTKRV), and Preferred Stock liquidating value (PSTKL). Following?, we calculate the book equity based on annual accounting data, which we assume are available to the public five months after the fiscal year end. To calculate the book equity of the combined firm in the completion month accurately, we adjust the acquiring firm s book value in the completion month by adding the target s book value of equity and deducting the amount of cash payment to the target shareholders. This adjustment is necessary because the book value of equity in the completion month still reflects the stand-alone book value of the acquiring firm s equity. For deals with private targets, we use the acquirer s book value because the target book value is not obtainable. The adjusted book equity data are used to calculate the book-to-market ratio (B/M). When using B/M in the empirical analysis, we exclude negative-b/m stocks. The construction of the following liquidity proxies used in this paper are based on daily data: The return-to-volume ratio of?, RV12, defined as the daily ratio of absolute return on a day to the dollar volume on that day averaged over the prior 12 months. The effective trading costs measure of?, EC12, estimated using daily closing prices over the prior 12 months. We obtain the data (generally available at the end of each year) over 1979 to 2009 from Joel Hasbrouck s website: jhasbrou/ The average of the monthly bid-ask spread estimates of? over the prior 12 months, CS12. Corwin and Schulz estimate the bid-ask spread for a month using daily high and low prices in the 7

9 month. We obtain their monthly spread estimates over from Shane Corwin s website: scorwin/ The quoted bid-ask spread of?, BA12, defined as the average daily relative bid-ask spread over the prior 12 months, where the relative bid-ask spread on a day is the ratio of the difference between ask and bid prices to the average of bid and ask prices on the day. 7 The proportion of zero-return days over the prior 12 months of?, ZR12.? find that the frequency of zero returns is directly related to conventional costs measures such as the quoted bid-ask spread and Roll s (1983) effective spread, and Lesmond et al. rely on daily zero returns to estimate transaction costs. The dollar volume measure of?, DV12, defined as the average daily dollar trading volume over the prior 12 months. The turnover measure of?, TO12, defined as the average daily turnover over the prior 12 months, where daily turnover is the ratio of the trading volume on a day to the number of shares outstanding on the day. The trading discontinuity measure of?, LM12, defined as the standardized turnover-adjusted number of zero daily trading volumes over the prior 12 months, [ LM12 = Number of zero daily volumes in prior 12 months + ] 1/(12-month turnover) Deflator NoTD, (1) where 12-month turnover is the sum of daily turnover (in percentage) over the prior 12 month, NoTD is the number of trading days in the market over the prior 12 months, and Deflator is chosen to be 20,000 so that 0 < 1/(x-month turnover) Deflator < 1 for all sample stocks. The RV12 measure describes the price impact dimension of liquidity. Turnover (TO12) and dollar volume (DV12) capture the trading quantity dimension of liquidity. The four measures, CS12, EC12, 7 Over our sample period, daily bid and ask prices are available after 1992, and before 1992 only in cases when a closing price is missing. Thus, BA12 is calculated based on available daily bid and ask prices in the prior 12 months and the daily spread must be positive. 8

10 ZR12 and BA12, reflect the trading costs dimension of liquidity. The LM12 proxy primarily measures the trading speed dimension of liquidity (i.e., the potential delay in executing an order, or the probability of no trade), but it also captures other dimensions of liquidity such as trading quantity, price impact, and trading costs, as illustrated in?. In addition, LM12 reflects liquidity risk, the danger of being unable to liquidate a position opportunely at a fair price.? shows that the trading discontinuity measure generates the most pronounced liquidity premium among the proxies of liquidity. When constructing TO12, DV12 and LM12, we require daily trading volumes to be available over the prior 12 months. Similar to?, to construct RV12, we require at least 80% daily trading volumes available in the prior 12 months and the construction excludes zero trading volumes. To compute the proportion of zero-return days, we require CRSP daily returns available in the prior 12 months. As previously discussed, mergers are expected to improve acquirers liquidity via an enhanced information environment and a broader shareholder base. We use analyst coverage and institutional ownership to verify their association with acquirers liquidity gains. Analyst coverage in a month is the number of analysts who issue earnings forecasts one-fiscal-period ahead in the prior 12 months. We obtain the number of analysts from the summary file of the I/B/E/S database. If an I/B/E/S stock has CRSP price information available but no analyst following over the prior 12 months, its analyst coverage is set to zero. If a CRSP stock is not in the I/B/E/S database, its analyst coverage is set to missing. Based on the data of CDA Spectrum 13F filings, we calculate institutional ownership, which is the percentage of shares owned by 13-F institutional investors. For an eligible firm, we also compute the number of institutional shareholders holding the firm s stocks. Table?? reports the sample distribution by year and by acquirer/deal characteristics. Among the 2,838 transactions, there are 2,462 mergers and 376 tender offers. We examine mergers and tender offers separately because previous studies find that acquirers exhibit low performance in the long term after mergers but not after tender offers (Agrawal and Jaffe, 2000).? find that low performance comes mainly from glamour acquirers. We therefore divide the sample into glamor, neutral and value acquirers according to the 30th and 70th percentiles of B/M of all NYSE firms in each year. There are 1,389 9

11 glamor, 1,178 neutral, and 271 value acquirers.? and? find that acquirers of stock-dominated offers underperform in the long run. Accordingly, we separate the sample into stock dominated and non-stock dominated offers, where stock dominated offers are those transactions financed at least half by stock. There are altogether 1,300 stock-dominated and 1,538 non-stock dominated offers. In the 1980s, there are more non-stock-dominated offers than stock dominated offers, but this pattern reverses in the 1990s through to 2001 when the internet bubble burst. There are 1,372 acquirers listed on NYSE/AMEX and 1,466 listed on NASDAQ with more listed on NASDAQ in the later sample period. There are 1,027 diversifying transactions (i.e., the acquirer and the target are from two different industries) and 1,811 non-diversifying transactions. The average deal value is $ million in nominal term. Deal value relative to acquirer s market cap is, on average, 44.35%, which means that these transactions are large investments made by acquirers. An average transaction takes days to complete. There is a notable trend that deal duration decreases over time, suggesting improved deal making skills of deal makers. In some years, the number of deals is less than 10 for some categories, e.g., there are only four tender offers in The decomposed buy-and-hold calendar-time approach To deal with the conflict on the choice between the monthly-rebalanced calendar-time approach and the buy-and-hold method, we adopt a decomposed buy-and-hold calendar-time (DBHCT) approach in the sprit of?. Specifically, we form an event-firm portfolio at the end of June each year starting from The portfolio contains all acquirers that complete at least one acquisition in the past 12 months. An acquirer is included only once if it completes several transactions in the past 12 months. We hold this portfolio for the next three years. We calculate the monthly portfolio returns in the holding period using the algorithm recommended by?: R p,τ = N i=1 w i τ 1 t=1 (1 + R i,t) Nj=1 w j τ 1 t=1 (1 + R j,t) R i,τ, τ = 2,..., m; R p,1 = N w i R i,1, (2) i=1 10

12 where N is the number of stocks in the portfolio, w i is the portfolio weight of stock i (for equally weighting, w i = 1/N), m is the number of months in the holding period (m = 36 for the three-year holding period), R i,t is the month-t return of stock i, and R p,τ is the month-τ return of the portfolio in the m-month holding period. Equation (??) indicates that the computation of monthly portfolio returns does not involve rebalancing portfolio weights or revising portfolio constituents over the multi-month holding period. Because we form portfolios every year and hold them for multiple years, this procedure involves some overlapping portfolios (e.g., three overlapping portfolios if the holding period is three years). We employ the technique of? to aggregate the monthly returns of these overlapping portfolios to get the non-overlapping monthly returns over the full testing period 7/1985 6/2011. With the three-year holding period, for example, the return in March 2000 is the average of the returns in March 2000 of the three portfolios formed at the end of June 1997, 1998, and To form the portfolio, we require the number of constituent stocks to be at least ten in order to minimize heterosckedasticity. This procedure is essentially a calendar time approach. It is, therefore, not subject to the cross-correlation problem of the buy-and-hold approach. Moreover, this approach preserves the buy-and-hold property, and appropriately reflects a representative long-term investor s wealth experience. Unlike the monthly-rebalanced traditional calendar-time approach, the DBHCT approach minimizes transaction costs and represents a feasible investment strategy. With the time series of the monthly portfolio returns of event firms computed using the DBHCT approach, we estimate acquirers long-term post-event performance using five asset pricing models: the FF3FM, the FF3FM extended by the? momentum factor, the Fama French five-factor model (FF5FM), the FF3FM extended by the Pastor and Stambaugh (2003) traded liquidity factor (FFPS4FM), and the LCAPM of?. 11

13 3. The liquidity evolution of acquiring firms To evaluate acquirers performance, previous studies commonly use MV-B/M matching or an asset pricing model incorporating MV and B/M factors. We hypothesize that the documented post-merger low performance of acquiring firms is due to their liquidity improvement and low exposure to liquidity risk; and the MV and B/M characteristics or risk factors fail to account for the liquidity effect. In this section, we assess acquirers liquidity evolution relative to their MV-B/M matching firms Acquirer and matching firm characteristics before and after the transaction For each acquirer, we find its match at the end of the deal completion month. The matching firm is in the same MV decile of NYSE firms and, of firms belonging to the same MV decile, has B/M closest to the acquirer s B/M. Moreover, we require the matching firm is not involved in any IPOs, SEOs or the same types of M&A transactions in the three years prior to the time of matching. Table??, Panel A shows that the MV-B/M matching is successful for MV and Stdev: there is no significant difference in MV (Stdev) between an average acquirer and its matching firm except in the first (second) post-acquisition year. The MV-B/M matching is partially successful on B/M: acquirers, on average, have a B/M similar to the matched firms in years 1 and 3 after deal completion, but significantly different B/M in years 2, 4 and 5. We also try an alternative matching procedure used by?, which requires the MV of a matching firm to be no less than 70% and no more than 130% of the acquirer s MV one month after the completion month, and a qualified firm with B/M closest to the acquirer s B/M is selected. This alternative procedure improves B/M matching with the acquirers and matching firms having similar B/M ratio in years 1, 3 and 4 after deal completion. With the alternative, however, the matching on MV is less successful, with the acquirers and matching firms having similar MV only in year 5. We note that previous studies do not report the quality of matching. Our results demonstrate that the commonly used dual-dimensional matching can be inaccurate in matching the dimensions that it is designed to match. Hence, it may be no surprise if the MV-B/M matching cannot match on liquidity. 12

14 Table??, Panel B shows that the MV-B/M matching indeed fails to match on liquidity. It is clear that acquirers are significantly more liquid than their MV-B/M matches in the post-acquisition years. Inspecting the trading discontinuity measure (LM12) reveals that, in the first year after deal completion, an average acquirer has zero volume days, which is significantly lower (t = 6.23) than the of its matched firm. The difference declines slightly over the next four years but remains significant through all five post-deal years. For the RV12 measure, acquirers display significantly lower price impact than their MV-B/M matches in each of the five post-acquisition years. Acquirers exhibit a significantly lower bid-ask spread estimate of? than their matches in the five post-acquisition years except for the fifth year. Acquirers are also more heavily traded than their matches with the differences in dollar volume (DV12) between acquirers and their matches being positive in each of the five post-acquisition years and statistically significant at 1% for the first two years. With four other liquidity proxies (TO12, EC12, BA12, and ZR12), we also find that acquirers are generally more liquid than their MV-B/M matches in the post-deal years, especially in the first year after deal completion. Based on all eight liquidity proxies, Figure?? depicts acquirers liquidity evolution relative to their MV-B/M matching firms. The inability of MV-B/M to match on liquidity, together with the substantial liquidity premium documented in the asset pricing literature, suggests that the previously documented acquirer low post-merger performance is due to the failure to account for liquidity risk. There is also a clear tendency that liquidity of acquiring firms themselves improves after transactions. Taking the trading discontinuity measure of liquidity (LM12) as an example, while an average acquirer has zero-volume days in year Y 2 (i.e., the second year before the deal announcement), it has zero-volume days in the first year after deal completion, a highly significant (t = 13.8) improvement in liquidity. Comparing acquirer liquidity in post-acquisition years with that in the first or third year before the deal announcement does not alter this observation. Examining other liquidity proxies reveals a similar tendency of improved acquirer liquidity. To assess acquirers information environment and breadth of shareholder base, we examine the evolution of acquirers analysts following and institutional ownership relative to their MV-B/M matched firms. 13

15 Table??, Panel C shows that the number of analysts (NoAnalysts) following acquirers is persistently greater than that of the matched firms in the post-transaction years. For instance, acquirers have, on average, analysts following in the second year after deal completion, which is significantly greater than the average of analysts for the matched firms (t = 3.54). Also, the NoAnalysts differences between acquirers and their matches are all positive and significant for the other four post-transaction years. Acquirers also have a significantly larger number of institutional shareholders (NoInstitutions) than their matches in each of the five post-transaction years. In the second year after deal completion, for example, an average acquirer has institutional shareholders, which is greater than , the number of institutional shareholders for an average matching firm (t = 3.85). In terms of acquirers institutional ownership (InstOwnership), it is generally greater (but not significantly so) than that of MV-B/M matched firms in the post-acquisition years. It is also apparent that following deal completion, acquirers have more analysts following, more institutional shareholders and a greater percentage of institutional holdings compared with the years prior to the deal announcement. Looking at institutional ownership (InstOwnership), for example, institutional shareholders hold 50.7% of an acquirer s shares outstanding two years after deal completion (Y +2 ), which is significantly greater than 43.6%, the level two years before the deal announcement (Y 2 ). We observe similar increases in InstOwnership for all other post-acquisition years. Compared with their pre-announcement levels, the number of analysts following (NoAnalysts) and the number of institutional holders (NoInstitutions) also exhibit significant increases in all post-transaction years Acquirer and matching firm liquidity beta before and after transaction To ascertain the evolvement of acquirers liquidity risk, for each acquirer and its MV-B/M match we estimate the following time-series regression using monthly returns from 36 months before the announcement to 36 months after the completion: 8 R t R f t = α + β 1 (R mt R f t ) + β 2 LIQ t + β 3 D t (R mt R f t ) + β 4 D t LIQ t + ε t, (3) 8 Results are similar if we use monthly stock returns from 60 months before announcement to 60 months after completion. 14

16 where R t is the month-t return of the acquirer or its MV-B/M match, R f t if the risk-free rate for month t, R mt is the month-t return of the market portfolio, LIQ t is the month-t value of Liu s (2006) liquidity risk factor, and D t is a dummy variable that takes the value of 1 in a post-completion month and 0 in a month before announcement. The hypothesis is that β 4 is negative for acquirers. Table??, Panel D reports the average of coefficient estimates across acquirers (and their MV-B/M matching firms). The results show that β 4 is significant at 0.24, implying that acquirer liquidity beta, on average, reduces by 0.24 in post-merger years compared to years prior to merger. In contrast, MV- B/M matching firms β 4 is insignificant at 0.21, indicating that there is no significant change in MV-B/M matching firms liquidity beta. Since matching firms do not experience any IPOs, SEOs or M&As, which are liquidity-changing events, it is not surprising that their liquidity betas do not change. Importantly, the post-merger liquidity beta (i.e., β 2 + β 4 ) is negative at 0.29 for acquirers, which is 0.45 lower than the matching firms post-merger liquidity beta of The evidence reveals that MV-B/M matching also fails to match on acquirers liquidity risk, and the relatively low liquidity risk to which acquirers expose could be the cause of the observed low acquirer post-merger performance Sources of acquirer liquidity improvement As we developed in the introduction, mergers are salient information generating process for both the short and long term. Acquirers also expand their shareholder base during and post merger. Previous literature establishes assoiation between better information environment (e.g., Diamond and Verrecchia, 1991; Balakrishnan, Billing, Kelly, and LjungQvist, 2014) or greater shareholder base (e.g., Merton, 1987; Grullon, Kanatas,and Weston, 2004) with higher liquidity. Further, Brennan and Subrahmanyam (1995), Irvine (2003), Roulstone (2003) argue that greater analyst coverage mitigates information asymmetry and improves stock liquidity. Irvine (2003) recommends firms to improve analyst coverage to enhance liquidity. Rubin (2007) demonstrates that liquidity is positively associated with institutional holdings. Therefore, we make two hypotheses 1) a greater increase in analyst coverage relates to a greater increase in acquirer post-merger liquidity and 2) a greater increase in the number of institutional shareholders 15

17 leads to a greater imporvement in acquirer post-merger liquidity. For completeness, we also examine the relation between change in institutional ownership measured as percentage of shares outstanding and change in liquidity. However, theories suggest that it is the number of shareholders that are more relevant. These hypotheses above are also consistent with our previous findings that acquirers experience liquidity improvements post transaction, and have more analysts following and a larger institutional shareholder base. As a result, we test our above hypotheses by relating acquirer liquidity to the number of analysts (NoAnalysts) and the number of institutional shareholders (NoInstitutions). To evaluate the extent to which acquirers pre-acquisition liquidity persists in the post-acquisition years, we regress the post-merger liquidity (PostLiq) rather than the change in liquidity on the above variables, with premerger liquidity (PreLiq) included as one of the control variables. Specifically, we run the following two regressions: PostLiq i = α 0 + α 1 NoAnalysts i + α 2 NoInstitutions i + β Controls i + γ Dummies i + ε i, (4) PostLiq i = α 0 + α 1 NoAnalysts i + α 3 InsOwnership i + β Controls i + γ Dummies i + ε i, (5) where PostLiq i is acquirer i s post-acquisition liquidity measured by the average of the liquidity proxy of interest, NoAnalysts i is the change in the number of analysts, NoInstitutions i is the change in the number of institutional holders, InstOwnership i is the change in institutional ownership, Controls i is a column vector of control variables (such as ln(mv), ln(b/m), etc.), and Dummies i is a column vector of dummy variables (including year dummies, industry dummies, etc.). Any change in a variable involved is calculated as the difference between the average of this variable over the two years after deal completion, skipping the completion year, and the average of the variable over the two years prior to the deal announcement, skipping the announcement year. We skip the completion year and announcement year to mitigate the influence of trading activities related to the transaction. However, the results are qualitatively the same if we do not skip the years and average the measures over three years before and after the announcement/completion. We include NoInstitutions and InsOwnership in equations (??) and (??) separately to avoid multicollinearity. 16

18 We report the above regression results in Table?? based on four liquidity proxies with each capturing a different dimension of liquidity. For the trading discontinuity measure of liquidity (LM12), NoAnalysts in equation (??) has a negative coefficient of 0.065% (t = 2.80), meaning that when analyst following increases, trading frequency increases. This is consistent with our hypothesis that greater analyst coverage relates to increased liquidity. The slope coefficient on NoInstitutions is significant at (t = 2.20) and InstOwnership in equation (??) has a negative coefficient of (t = 3.17). These are again in line with our hypothesis that a larger shareholder base and more institutional holdings relates to higher liquidity. Pre-acquisition LM12 has a coefficient of (t = 6.26), indicating that everything else equal, an acquirer that has a lower LM12 continues to be more liquid after the transaction. The NASDAQ dummy has a coefficient of (t = 2.22), implying that an average NASDAQ acquirer has an LM12 that is greater than a non-nasdaq acquirer. An acquirer in a tender offer deal is more liquid post transaction since the tender offer dummy has a coefficient of (t = 2.31). The stock dominated offer dummy has a coefficient of (t = 2.73), meaning that an acquirer with stock-dominated financing has a post-merger LM lower than that of an acquirer with nonstock dominated financing. Results are similar in other regressions where we measure liquidity using RV12, DV12, or CS12. An increase in the number of analysts following relates to a lower post-merger RV12, a lower CS12, and a higher post-merger DV12 except in equation (??) where the coefficient on NoAnalysts is insignificant at conventional level. An increase in the number of institutional shareholders leads to a significant (at 1%) improvement in all but one liquidity measure (CS12). Similarly, an increase in institutional ownership leads to a significant (at 1%) improvement in all liquidity proxies except for DV12. The findings in this section show that acquirers experience significant liquidity improvements in postacquisition years and the commonly employed MV-B/M matching fails to match on acquirer liquidity. The evidence also suggests that acquirers liquidity gains are associated with an improved information environment. Because acquirers become more liquid after acquisitions and have persistently better liq- 17

19 uidity than their MV-B/M matches, in the next section, we turn to the question at the center of our study: how do acquirers perform in the long run when liquidity or liquidity risk is accounted for? 4. Acquiring firms long-run post-merger performance and liquidity In this section, we analyze acquiring firms long-term post-merger performance using both the buy-andhold abnormal return (BHAR) and the calendar time approaches. In terms of the calendar time approach, we use both the traditional monthly-rebalanced calendar time approach and the decomposed buy-andhold calendar time (DBHCT) approach, with a preference for the latter. To calculate the BHAR, we rely on MV-B/M matched and liquidity matched benchmarks. For the calendar time approach, we employ asset pricing models such as the FF3FM, the momentum-extended FF3FM of?, the Fama French fivefactor model (FF5FM), the Pastor Stambaugh liquidity-extended FF3FM (FFPS4FM), and the liquidityaugmented CAPM (LCAPM) of? to evaluate acquirer performance. Our assessment focuses on acquirers in mergers because acquirer low performance is mainly from this sample of transactions according to previous literature. For completeness, we also report results for tender offers and for the full sample pooling together both mergers and tender offers Acquirer performance based on MV-B/M matching and the traditional calendar time approach Table?? reports the acquirer BHAR measured over the three years after deal completion using a MV- B/M matched benchmark. Results in Panel A show that merger acquirers significantly underperform their MV-B/M matching firms by 10.46% (t = 3.08) over the three years post-merger. The BHARs for glamour acquirers and acquirers of stock-dominated offers are most pronounced at 19.46% (t = 3.66) and 15.74% (t = 2.82), consistent with the findings of? and?. Neutral acquirers, value acquirers and acquirers in non-stock dominated offers have insignificant long-run abnormal performance. For the full sample, Panel B presents similar results to Panel A. The BHARs are significantly negative at a 1% level for all acquirers, glamor acquirers and acquirers of stock-dominated offers, but insignificant for other acquirers. Turning to panel C (tender offers only), acquirers show no abnormal performance.? argue 18

20 that the t-statistic used for testing the average BHAR is inflated due to the cross correlations between individual BHARs. Our aim here is to confirm the results from previous studies testing acquirer posttransaction BHAR without adjusting for the cross correlations. As? point out, the cross correlation structure is difficult to estimate. We, therefore, primarily rely on the calendar time approach for our statistical inference because it accommodates the cross correlation issue (Fama, 1998; Mitchell and Stafford, 2000). Table?? reports acquirer post-transaction performance using the traditional monthly-rebalanced calendar time approach, where acquirer portfolio return in a month is the equally-weighted average of individual acquirers that complete at least one transaction in the previous three years. The benchmarks are the FF3FM commonly used in previous studies, the? momentum-extended FF3FM and the recently developed five-factor model of? (FF5FM). For the sake of brevity, we only report the alpha (i.e., the abnormal performance of the acquirer portfolio) for each model. For the sample including mergers only, Panel A shows that glamor acquirers underperform the FF3FM benchmark by 0.384% per month (t = 3.28), which translates into an underperformance of 13.82% over the three years after deal completion. This again conforms to the finding of? that glamor acquirers underperform in the three years after mergers. Acquirers from stock-dominated offers have a FF3FM alpha of 0.315% per month (t = 2.53), which translates to an underperformance of 11.34% over the three years after deal completion. This is in line with previous findings by? and? that stock acquirers significantly underperform in the three post-deal years. Adding the momentum factor to the FF3FM still shows limited power to explain the post-merger performance of glamor acquirers, which have a significantly negative alpha of 0.203% per month (t = 1.96). After adjusting for the five-factor model (FF5FM), acquirers show no abnormal performance. For the full sample containing both mergers and tender offers, the results in Panel B are qualitatively the same as those in panel A. For the sample including tender offers only, results in Panel C display no abnormal performance in any circumstances. Untabulated results show that tracking acquirer post-deal performance over five years yields qualitatively similar evidence. 19

21 In a nutshell, we confirm the previously documented underperformance of glamor acquirers and acquirers in stock-dominated offers relative to the FF3FM benchmark and the MV-B/M matching. To our knowledge the momentum-extended FF3FM and the FF5FM, which show improved power relative to the FF3FM in explaining the traditional monthly-rebalanced acquirer portfolio returns, are not used in previous literature. Next, we turn to our preferred decomposed buy-and-hold calendar time approach for our analysis of acquirer long-term post-acquisition performance Acquirer performance based on the decomposed buy-and-hold calendar time (DBHCT) approach As discussed earlier, the long-term buy-and-hold method and the traditional monthly-rebalanced calendar time approach applied in the previous subsection have their advantages and disadvantages. The former represents an implementable strategy mimicking a long-term investor s experience, but it suffers from a statistical-inference issue due to cross-correlations of acquirer long-term BHARs. The latter does not suffer from the cross-correlation problem, but monthly portfolio rebalancing does not correctly measure the return to an investor who holds the portfolio for a long period post event. In contrast, the DBHCT approach not only preserves the advantages of the traditional calendar time approach as well as the buyand-hold method, but also overcomes their disadvantages, i.e., it correctly tracks the long-term buy-andhold return and successfully avoids the issue of cross-correlated BHARs. In this subsection, we use the DBHCT approach to calculate acquirer portfolio holding-period monthly returns and then benchmark the monthly returns to the FF3FM, the momentum-extended FF3FM, the FF5FM, and liquidity based pricing models. We also perform a liquidity matching to assess acquirer performance Results with benchmarks of the FF3FM, the momentum-extended FF3FM and the FF5FM Table?? reports the performance (on a monthly basis) of the acquirer portfolio over the three years post transaction: raw excess return (relative to the risk-free rate) and alphas of the FF3FM, the momentumextended FF3FM and the FF5FM. For the sample containing mergers only, Panel A shows that the portfolio of glamour acquirers has a FF3FM alpha of 0.247% (t = 2.75) per month, which translates into an 20

22 underperformance of 8.89% over the three years after deal completion. The portfolio of acquirers with stock-dominated financing has a FF3FM alpha of 0.239% (t = 2.25) per month. In contrast to the results based on the monthly-rebalanced portfolio reported in Table??, the momentum-extended FF3FM gives abnormal performance for both glamor acquirers and acquirers of stock dominated offers. The momentum-extended FF3FM alpha of the glamor-acquirer portfolio is 0.262% (t = 2.86) per month, which is equivalent to an underperformance of 9.43% over the three post-deal years. The acquirer portfolio of stock dominated offers has a momentum-extended FF3FM alpha of 0.259% (t = 2.40) per month. When we pool mergers and tender offers, Table??, Panel B shows that glamor acquirers remain underperformers against both the FF3FM and the momentum-extended FF3FM. On the other hand, acquirers of stock-dominated offers have an insignificant alpha using either model. Turning to tender offers only, Panel C indicates that acquirers yield no significant abnormal performance in any circumstances. Similar to the Table?? results, the FF5FM also accounts for the acquirer post-merger returns calculated with the DBHCT approach. As we explain below, the ability of the FF5FM to explain acquirer post-transaction performance appears to be its improved power to partially capture the liquidity premium Results of liquidity-based asset pricing models At the center of our study, we hypothesise that lower exposure to liquidity risk accounts for acquirers long-term low performance. Our early results (see Table??) indicate that MV-B/M matching is not able to provide an appropriate liquidity benchmark. In this section, we investigate whether acquirer post-merger underperformance documented in the previous literature disappears once liquidity risk is accounted for. We employ two liquidity risk based pricing models, namely, Liu s (2006) liquidity-augmented CAPM (LCAPM) and Pastor and Stambaugh s (2003) liquidity-extended FF3FM (FFPS4FM). With the LCAPM, we estimate the following time-series regression: R pt R f t = α + β m (R mt R f t ) + β l LIQ t + ε pt, (6) 21

23 where R pt is the month-t return of the acquirer portfolio, R f t is the risk-free rate for month t, R mt is the month-t return of the market portfolio proxied by the value-weighted NYSE/AMEX/NASDAQ/ARCA index, and LIQ t is the month-t value of Liu s (2006) liquidity risk factor. With the FFPS4FM, we estimate the following time-series regression: R pt R f t = α + β m (R mt R f t ) + β s SMB t + β h HML t + β PS PSvwf t + ε pt, (7) where SMB t is the month-t value of the Fama French size factor, HML t is the month-t value of the Fama French book-to-market factor, and PSvwf t is the month-t value of the Pastor Stambaugh (PS) traded liquidity factor. Table?? reports the parameter estimates of regressions (??) and (??). For the sample including mergers only, Panel A shows that none of the acquirer portfolios has a significant LCAPM alpha. The glamor acquirer portfolio has a LCAPM alpha of virtually zero (t = 0.13). Similarly, the acquirer portfolio for stock-dominated offers has a LCAPM alpha close to zero (t = 0.14). Both glamor acquirers and acquirers of stock-dominated offers load negatively on the liquidity factor. The evidence indicates that glamor acquirers and acquirers of stock-dominated offers have lower exposure to liquidity risk, and their low post-merger performance documented in the literature is due to the neglect of liquidity risk. Consistently, both glamor acquirers and acquirers of stock-dominated offers load negatively on the PS liquidity factor. 9 Adding the PS factor to the FF3FM also explains acquirer performance better than the FF3FM. The alpha of the four-factor model (??) is 0.211% per month (t = 2.34) for glamor acquirers and it is 0.184% per month (t = 1.75) for stock-dominated offers. With the pooled sample of mergers and tender offers, Table??, Panel B shows that the liquidity factor loading mirrors that observed in Panel A. Namely, glamor acquirers and acquirers of stock-dominated offers have significantly negative liquidity factor loadings in models (??) and (??). Consistent with the results in Panel A, the LCAPM explains 9 We also extend the FF3FM by the PS non-traded liquidity factor and by the? non-traded liquidity factor constructed with the variable component of price impact. Untabulated results show low or negative loadings of acquirers on the two non-traded liquidity factors. For glamor acquirers, the loading on the PS non-traded factor is (t = 0.92) and it is (t = 0.04) on the Sadka factor. For acquirers with stock-dominated financing, the loading on the PS non-traded factor is (t = 0.80) and it is (t = 2.06) on the Sadka factor. 22