Anticipation, Acquisitions and Bidder Returns By + Moon H. Song and ++ Ralph A. Walkling* +San Diego State University, San Diego, California, 92182 Phone: 619-594-5334; e-mail [email protected] ++Ohio State University, Faculty of Finance, College of Business, Columbus, Ohio, 43210 Phone: 614-292- 1580; e-mail: [email protected] * Corresponding author. The authors appreciate comments on earlier drafts by participants in finance workshops at the University of Delaware, University of Pittsburgh, University of South Florida and Ohio State University. We also appreciate the benefit of comments from and discussions with Rene Stulz Jean Helwege, Kewei Hou, Karl Diether, Simi Kedia, and Karen Wruck. We are grateful to Gilberto Loureiro, Carrie Pan and Rodolfo Martell for excellent research assistance Current version 01/26/2005 Anticipation, Acquisitions and Bidder Returns Abstract The announcement of an acquisition bid has the potential to signal information not only about the bidder and target but also about other firms in the bidder’s industry. This paper develops and tests the anticipation hypothesis as applied to bidding firm returns and the returns of their rivals. Simply stated, anticipated events exhibit price reactions at the time of anticipation. To the best of our knowledge, anticipation effects related to rivals of bidding firms have not been examined in the literature. Nevertheless, our results document strong support for anticipation hypothesis. First, the abnormal returns of bidding firms are significantly related to the degree of surprise surrounding the announcements. Second, at the time of an initial industry announcement, rival firms experience significant and differential price adjustments when they are subsequent bidders themselves. Third, these price adjustments are proportional to the abnormal returns earned by the initial industry bidder. Our results hold after controlling for variables typically associated with bidding firm returns. 2 Anticipation, Acquisitions and Bidder Returns 1. Introduction 1.1 Anticipation, acquisitions and Bidder Returns Every time a bidder announces an acquisition attempt it signals information to the market. A rational market adjusts prices to the degree that the information is unanticipated and is expected to impact particular firms. Early literature studying whether program bids were anticipated reaches alternate conclusions. Schipper and Thompson (1983) find that bidders announcing acquisition programs experience most of their abnormal returns at the time of the initial announcement. Malatesta and Thompson (1985) also document the importance of anticipation. Loderer and Martin, (1990) find that bidders making first acquisitions earn significantly larger returns. The result is even more pronounced for non-program bidders. Alternatively, Asquith, Bruner and Mullins (1983) find that bidding firms earn significant returns preceding each of their first four bids suggesting the absence of anticipation. More recently, Fuller, Netter, and Stegemoller (2002) analyze frequent bidders during the 1990’s and find that the order of the acquisitions does not affect acquirers’ excess returns. Similarly, Conn, Cosh, Guest, and Alan, (2004) analyze bidding by UK firms over the 1984-1998 period. They find that returns from multiple acquirers are similar to those of single acquirers. Bidders making multiple acquisitions experience wealth declines with each acquisition but only if their first acquisition is successful. Each of these results focuses on whether sequential bids by a specific firm are anticipated. But information about impending bids is also disseminated through industry shocks. Indeed, any bid by a firm has the potential to contain information pertinent to firms in its industry. Recent work in the asset pricing literature underscores this idea. Hou (2004), for example, finds that industries are “the primary channel for news dissemination in the equity market.” We note that information may impact all firms in an industry, or 3 if the information relates to bidding itself and the market is prescient about which rivals will be subsequent bidders, only the prices of those firms would adjust. There is little analysis in the literature of the extent to which the shock waves of a bid are reflected throughout the industry. Song and Walkling (2000) find that rivals of target firms experience abnormal returns to the extent that they are also likely to be targets. Moreover, the size of abnormal returns increases with the degree of market surprise about the initial industry acquisition. To the best of our knowledge, a similar analysis of anticipation in bidding industries does not exist. We simply don’t know if the impact of an acquisition attempt is reflected throughout a bidder’s industry. Nor do we know whether differential effects are reflected in firms that will be subsequent bidders themselves (as opposed to firms that are rivals but do not subsequently bid). Such an analysis is particularly difficult to test in the case of bidding firms because the magnitude of announcement period returns is quite small. For this same reason, the existence of an anticipation effect is important to document; any measurable effect is likely to be a sizable component of bidder returns. Indeed, after recognizing anticipation, our results are consistent with the interpretation that bidding activity is in general a wealth increasing activity. This is in contrast to the negative or zero abnormal returns documented in the literature. This research develops and tests the “Bidder Anticipation Hypothesis.” Simply stated, anticipated bids will experience smaller market reaction than bids that are a surprise. Unlike tests of program announcements for a specific firm, we develop our tests within and across industries. Our premise is that acquisition bids preceded by a long industry dormant period (without acquisition bids) are less likely to be anticipated. This is the same technique used to measure surprise for rivals of targets in Song and Walkling (2000). In addition to addressing the existence of anticipation, this research has the potential to reveal the type of information anticipated. As mentioned, bidding firm announcements could signal industry wide information with effects pervasive across all firms in the industry. If the information is specific to firms 4 that will be subsequent bidders or subsequent targets in an industry and if the market is able to anticipate these conditions, we would expect differential returns for those specific firms. Using a sample of 3393 acquisitions from 1985 to 2001, we find strong support for the bidder anticipation hypothesis. Abnormal returns to bidders are significantly related to the length of the dormant period. The first bidder in an industry after a minimum twelve month dormant period experiences significantly positive abnormal returns averaging 0.79%. In contrast, the average abnormal return to all bidders is an insignificant 0.04% typical of the literature. In industries with less than a twelve month dormant period between acquisitions, the average abnormal return to the next bidder is an insignificant - 0.13%. Moreover, abnormal returns to subsequent bidders earned at the time of their industry’s first bid, adjust in sign, magnitude and proportion to the abnormal returns of the initial industry bidder. Note that our approach is distinct from methodologies examining clustering and merger waves. Initial bids do not have to be followed by subsequent bids in their industry. Indeed, in pursuing alternate explanations for some of our results, we exploit the fact that clustering does not necessarily follow dormant periods. Our results also hold across after controlling for form of payment, organizational form, target attitude, target nationality, multiple bidders, offer outcome, merger waves, and other attributes associated in the literature with bidder returns. The remainder of the paper is organized as follows: Section 2 reviews the relevant literature and provides a development of the ‘anticipation hypothesis’. Section 3 describes our methodology. Univariate and multivariate results are presented in Section 4. Section 5 concludes. 2. Background and hypotheses 2.1 Bidder Returns For over three decades, the financial literature has been intrigued with understanding the magnitude and factors affecting abnormal returns to bidding firms. Jensen and Ruback (1983), Jarrell, Brickley and Netter (1988), Jarrell, and Poulsen (1989) and Andrade, Mitchell and Stafford (2001) summarize a large 5 body of evidence spanning four decades and report that the unconditional acquisition period return to bidders is generally zero or slightly negative.1 For example, Andrade, Mitchell, and Stafford report that bidder abnormal returns for acquisitions spanning the 1973-98 period average -0.7%. Bidding returns are also negative in each of the three sub periods analyzed. In contrast, Bradley and Sundaram (2004) find significant superior performance for bidders acquiring US firms during the 1990s. However, their measure of performance includes run-up in the pre- announcement period. Their conclusion is that performance (the run-up) drives acquisitions, not the other way around. Returns in the announcement and in the post-announcement period “arrests, and even reverses” pre-announcement run-up but not enough to eliminate it. Interestingly, negative announcement returns are only found in the subset of large, stock-financed acquisitions of publicly traded targets. Roll (1986), Jensen (2003), and Shleifer and Vishny (2003) argue the influence of hubris and overvaluation as explanations for bidder returns. Of particular interest, Shleifer and Vishny illustrate cases where bidders earning negative short run returns could still be maximizing value; the alternative to the acquisition may have even decreased value even more. The trading of overvalued equity for lower valued assets improves the position of bidding shareholders. Other measurable factors that emerge in analyses of bidder returns include form of payment [Travlos (1987), Huang and Walkling (1987), and Wansley, Lane and Yang (1983)], organizational form and nationality of the target (e.g., public or private; domestic or foreign), [Faccio, Mara, McConnell, John J. and Stolin (2004), Fuller, Netter, and Stegemoller (2002) and Moeller, Schlingemann and Stulz (2003a) and Moeller, Schlingemann and Stulz (2003b)]. 1 A partial list of the literature examining the conditional returns to bidders includes Travlos (1986), Lang, Stulz and Walkling (1991), Hubbard and Palia (1995), Eckbo and Thorburn (2000), Fuller, Netter and Stegemoller (2002), and Moeller Schlingemann and Stulz (2004) . 6 2.2 Anticipation effects As discussed in the introduction, the recognition of anticipation effects in acquisitions stems from the work of Schipper and Thompson (1983), Asquith, Bruner and Mullins (1983) and Malatesta and Thompson (1985). They argue that anticipated effects will be impounded in stock prices. A recent application finds anticipation effects within industry groupings. Song and Walkling (2000) note that rivals of targeted firms experience contemporaneous positive abnormal returns to the extent they are likely to be targeted themselves.2 While not a formal analysis of anticipation effects, the early work of Palepu (1986) recognizes the possibility of their existence including a dummy variable set equal to one if an acquisition occurred in a firms industry in the previous 12 months. Two recent theoretical papers also develop ideas that can be related to anticipation. Jovanovic and Braguinsky (2002) present a model in which the negative market reaction to a bidder’s acquisition announcement stems from a signaling effect. Specifically, they presume that the bidder’s announcement of external acquisition signals that internal investment opportunities are poor. Their theoretical work does not address the issue of anticipation. Nevertheless, positive or negative industry information signaled by the first announced bid in an industry should be reflected in the stock prices of competing firms. 2.3 Merger waves and dormant periods The concept of merger anticipation effects is related to (but distinct from) evidence that acquisitions cluster by industry. Mitchell and Mulherin (1996) report this phenomenon for target firms. At the industry level 50% of the targets they examined over the 1982-89 period are concentrated in 25% of the years. Andrade and Stafford (2001) provide preliminary evidence that such clustering occurs for bidders. Harford (2003) reports results consistent with industry merger waves occurring as an efficient response to industry specific shocks. On average, mergers occurring during waves are associated with significantly positive wealth gains; mergers in the same industry, but outside of the wave period do not create wealth. In 2 The contagion literature also recognizes the anticipation effect across many venues. 7 theoretical work, Rhodes-Kropf and Viswanathan (2003) suggest that periods of over or undervaluation lead to merger waves. Rhodes-Kropf, Robinson, and Viswanathan (forthcoming) find empirical support for this proposition. These ideas are also reminiscent of Gort (1969) who suggests that exogenous shocks to an industry provide opportunities for consolidation and expansion. The occurrence of acquisition activity, after a long period without such activity, could signal an industry wide revaluation of targets (as in Song and Walkling, 2000) or bidders, as examined in this paper. Akbulut and Matsusaka (2003) examine diversifying acquisitions reporting that the means and medians of combined target and bidder returns are significantly positive during waves and insignificantly different from zero outside of the waves. Mean and median bidding firm returns are insignificantly negative both within and outside of waves.3 We take a different approach, following Song and Walkling (2000) by using the length of dormant periods (without bidding activity) within an industry as a proxy for the degree of surprise about an acquisition. While there is likely to be some positive relation between clustering and dormant periods, one need not be associated with the other. Clustering of bidding within an industry may be preceded by a long dormant period, but this need not be the case. Moreover, a long dormant period can be followed by an intense cluster of activity (i.e., an industry wave) but again, this need not be the case. In a subsequent section we utilize the distinction between dormant periods and waves to provide evidence on an alternate explanation for abnormal returns to initial bidders. 3 Additional research on merger waves in contained in Gugler, Mueller, and Burcin (2004). 8 2.4 Hypotheses Four hypotheses stem from our discussion of anticipation: H : In wealth creating (decreasing) acquisitions the magnitude of bidder returns will increase 1 (decrease) with the degree of surprise (i.e., length of the dormant period) preceding the announcement of a bid.4 H : Abnormal returns to rivals that are subsequent bidders will be different from firms that do 2 not subsequently bid. H : Abnormal returns to rivals will experience a contemporaneous and proportional adjustment 3 to the abnormal return earned by the initial industry bidder. H : Abnormal returns to rivals who are subsequent bidders will experience a contemporaneous 4 and proportional adjustment to the abnormal return earned by the initial industry bidder. The magnitude of this adjustment will be different from those of non-bidding rivals. The first hypothesis also suggests that the magnitude of returns for the initial bid following a dormant period will be greater than for subsequent bidders. Note that this first hypothesis is agnostic about the sign of bidder returns. It relates to the magnitude of returns conditional on the acquisition being wealth creating or wealth destroying to the bidding firm. We first test this hypothesis ignoring the conditional aspects, thus mixing positive and negative acquisitions and biasing our results towards insignificance. The second hypothesis tests whether the market differentiates subsequent bidders from non-bidding rivals at the time of an initial bid. Hypotheses three and four test whether the magnitude of any adjustment by rival firms is related to the abnormal return of the initial bidder. Hypothesis three is similar in spirit to hypothesis four but does not require the market to correctly recognize subsequent bidders at the time of an initial industry bid. The null hypotheses associated with each of the above statements would imply the lack of anticipation across firms in an industry. But even if our hypotheses are correct, there are at least four 4 For simplicity, we use the phrase wealth increasing (decreasing) to mean acquisitions with positive (negative) abnormal returns. We fully appreciate the insight of Shleifer and Vishny (2003) that a negative abnormal return does not necessarily mean the bidder was destroying wealth. The alternative may have been worse. 9 reasons why our empirical tests may find these hypotheses to be false. First, we may be incorrect in our assumption that the appearance of an initial industry bid signals subsequent bidding activity in the same industry. Second, our measure of industry could be inadequate. Third, mixing wealth increasing and wealth decreasing acquisitions biases tests of the first hypothesis toward insignificance. Fourth, our measure of surprise, the dormant period within an industry, could be flawed or diminished by confounding effects. Alternatively, investors may anticipate bids through other means (e.g., a predictive model) without reference to dormant periods. Related to this, Boone and Mulherin (2004) report that while only 13% of the firms in their sample had multiple public bidders about half of these firms disclose private negotiations with other potential bidders in SEC documents. Our classification measures rely on public announcements. Private negotiations are often reported with a lag or may not be reported at all. Moreover, selling firms may solicit bids from multiple parties resulting in the spread of private information in the industry even before the first public announcement. The spread of private information prior to public announcements can result in price adjustments that bias against our finding significant abnormal returns at the time of the public announcement. All of these reasons work against our finding significant results. 3. Research design 3.1 Sample selection We use the SDC database to identify both domestic and international acquisition bids above $10 million by US bidders over the period 1/1/1985 through 12/31/2001. This produces an initial sample of 14,564 deals. We delete financial firms and utilities (SIC codes 6000 through 6999 and 4900 through 4999) and cases where CRSP SIC codes or CRSP returns are not available to calculate abnormal returns as well as 249 cases where multiple bids occur within an industry on the same day.5 To avoid spurious results from industries with little acquisition activity, we also delete bidders in four-digit CRSP industries containing less 5 It is problematic to determine which of these bids occurred first. Retaining these cases does not alter our results. 10