UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Essays on the dynamics of mergers and acquisitions Moran-Villar, Pablo Christian 2013

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2013_fall_moranvillar_pablo.pdf.pdf [ 1.07MB ]
Metadata
JSON: 24-1.0073897.json
JSON-LD: 24-1.0073897-ld.json
RDF/XML (Pretty): 24-1.0073897-rdf.xml
RDF/JSON: 24-1.0073897-rdf.json
Turtle: 24-1.0073897-turtle.txt
N-Triples: 24-1.0073897-rdf-ntriples.txt
Original Record: 24-1.0073897-source.json
Full Text
24-1.0073897-fulltext.txt
Citation
24-1.0073897.ris

Full Text

Essays on the Dynamics of Mergers and Acquisitions by Pablo Christian Moran-Villar B.Sc., Universidad de Talca, 2000 M.Sc., Concordia University, 2003 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in The Faculty of Graduate Studies (Business Administration) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) June 2013 c© Pablo Christian Moran-Villar, 2013 Abstract In this thesis, I present three essays on the interaction of two dynamic aspects of mergers and acquisitions. First, merger activity follows waves within industries over time. Second, acquirers’ announcement returns are on average small and decline within merger waves. In the first essay, I develop a model of merger waves to study the interplay between merger timing and market anticipation of deal announcements. I show that the pattern of small and declining announcement returns for acquirers in merger waves is consistent with the notion that the market learns over time and is thus able to better anticipate deal announcements. This explanation contrasts with existing theories which attribute the declining pattern in announcement returns to a decline in deal quality. The model delivers several predictions about time-series and cross- section aspects of acquirers’ stock returns during merger wave episodes. In the second essay, I test a set of the model’s predictions. As a testing lab- oratory, I use four industries that underwent merger deregulations in the 1990s. Consistent with existing theories, high quality deals tend to be announced early in a merger wave. However, I show that this pattern in deal quality does not ex- plain the declining pattern in acquirers’ announcement effects. Consistent with the model’s predictions, I find that what matters for this pattern is the ‘unexpected’ portion of deal timing. I also find evidence of contagion effects on acquirers’ peers that is consistent with the information channel in the model. In the third essay, I study the drivers of merger waves by examining the alloca- tion of equity proceeds raised at times of high merger activity. My results indicate that firms do not systematically increase debt repayment or equity payout with eq- uity proceeds raised in high merger years. This pattern does not conform with the view that managers believe the stock is overvalued at the time of the equity issue. Instead, the observed pattern of proceeds allocation is consistent with the existence of time-varying adverse selection and investment lags. The evidence supports the idea that these frictions are important elements behind the dynamics of merger and acquisition activity. ii Preface This dissertation is original, unpublished, independent work by the author, Pablo Christian Moran-Villar. iii Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 Anticipation and Timing in Merger Waves: Theory . . . . . . . . . 7 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.2 Model Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.3 Acquisition Strategies . . . . . . . . . . . . . . . . . . . . . . . . 12 2.3.1 Complete Information . . . . . . . . . . . . . . . . . . . 12 2.3.2 Heterogeneous Information . . . . . . . . . . . . . . . . 14 2.4 Equilibrium Outcomes . . . . . . . . . . . . . . . . . . . . . . . 16 2.4.1 Timing of Announcements . . . . . . . . . . . . . . . . . 17 2.4.2 Announcement Effects . . . . . . . . . . . . . . . . . . . 18 2.5 Predictions and Related Empirical Evidence . . . . . . . . . . . . 23 2.5.1 Timing and Anticipation Affects . . . . . . . . . . . . . . 24 2.5.2 Contagion Effects . . . . . . . . . . . . . . . . . . . . . . 26 2.6 Final Remarks and Conclusion . . . . . . . . . . . . . . . . . . . 27 iv 3 Anticipation and Timing in Merger Waves: Evidence . . . . . . . . 36 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3.2 Data and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 39 3.2.1 Merger Waves . . . . . . . . . . . . . . . . . . . . . . . 39 3.2.2 Sample CARs . . . . . . . . . . . . . . . . . . . . . . . . 41 3.3 Timing Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 3.3.1 Total Timing Effect . . . . . . . . . . . . . . . . . . . . . 43 3.3.2 The Timing Surprise . . . . . . . . . . . . . . . . . . . . 45 3.4 Contagion Effects . . . . . . . . . . . . . . . . . . . . . . . . . . 49 3.4.1 Portfolio Contagion Effects . . . . . . . . . . . . . . . . 49 3.4.2 Individual Contagion Effects . . . . . . . . . . . . . . . . 51 3.5 Final Remarks and Conclusion . . . . . . . . . . . . . . . . . . . 56 4 Market Timing and Merger Waves . . . . . . . . . . . . . . . . . . . 70 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 4.2 Hypotheses Development . . . . . . . . . . . . . . . . . . . . . . 75 4.3 Sample and Data . . . . . . . . . . . . . . . . . . . . . . . . . . 78 4.3.1 M&A Descriptive Statistics . . . . . . . . . . . . . . . . 79 4.3.2 The Time Series of Merger Activity . . . . . . . . . . . . 80 4.4 Market Timing Test . . . . . . . . . . . . . . . . . . . . . . . . . 82 4.4.1 Regression Specification . . . . . . . . . . . . . . . . . . 82 4.4.2 Control Variables . . . . . . . . . . . . . . . . . . . . . . 84 4.4.3 Sorts on Financing Variables . . . . . . . . . . . . . . . . 86 4.4.4 Panel Data Results . . . . . . . . . . . . . . . . . . . . . 86 4.5 Dynamic Allocation of Issuance Proceeds . . . . . . . . . . . . . 88 4.5.1 The Core Predictions . . . . . . . . . . . . . . . . . . . . 88 4.5.2 Flow of Funds Regression . . . . . . . . . . . . . . . . . 90 4.5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 4.6 Final Remarks and Conclusion . . . . . . . . . . . . . . . . . . . 95 5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 v List of Tables Table 2.1 Timing of Announcements . . . . . . . . . . . . . . . . . . . 35 Table 3.1 Cumulative Abnormal Returns . . . . . . . . . . . . . . . . . 62 Table 3.2 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . 63 Table 3.3 Total Timing Effect . . . . . . . . . . . . . . . . . . . . . . . 64 Table 3.4 Timing Determinants . . . . . . . . . . . . . . . . . . . . . . 65 Table 3.5 Unexpected Timing Effect . . . . . . . . . . . . . . . . . . . . 66 Table 3.6 Portfolio Contagion Effect . . . . . . . . . . . . . . . . . . . . 67 Table 3.7 Individual Contagion Effects . . . . . . . . . . . . . . . . . . 68 Table 3.8 Bidder Follower Comovement . . . . . . . . . . . . . . . . . . 69 Table 4.1 Sample Deals . . . . . . . . . . . . . . . . . . . . . . . . . . 105 Table 4.2 Merger Intensity Descriptive Statistics . . . . . . . . . . . . . 106 Table 4.3 Bidder Likelihood by Financing Sorts . . . . . . . . . . . . . . 107 Table 4.4 Merger Intensity Regression . . . . . . . . . . . . . . . . . . . 108 Table 4.5 Use of Funds Predictions . . . . . . . . . . . . . . . . . . . . 109 Table 4.6 Use of Funds Regressions . . . . . . . . . . . . . . . . . . . . 110 Table 4.7 Use of Funds Regressions and Financial Constraints . . . . . . 111 vi List of Figures Figure 2.1 Time Line of Events . . . . . . . . . . . . . . . . . . . . . . 33 Figure 2.2 Timing Outcomes . . . . . . . . . . . . . . . . . . . . . . . . 34 Figure 2.3 Timing Distortions . . . . . . . . . . . . . . . . . . . . . . . 35 Figure 3.1 Merger Waves Plot . . . . . . . . . . . . . . . . . . . . . . . 60 Figure 3.2 Histogram of Bidder Cumulative Abnormal Returns . . . . . 61 Figure 4.1 Merger Intensity Plot . . . . . . . . . . . . . . . . . . . . . . 99 Figure 4.2 Detrended Merger Intensity . . . . . . . . . . . . . . . . . . 100 Figure 4.3 Intensity based on Value and Number of Deals . . . . . . . . 101 Figure 4.4 Use of Funds Dynamics . . . . . . . . . . . . . . . . . . . . 102 Figure 4.5 Use of Funds Dynamics by Financial Constraints . . . . . . . 103 Figure 4.6 Use of Funds Dynamics by Investment Horizon . . . . . . . . 104 vii Acknowledgments I am profoundly indebted to Lorenzo Garlappi and Ron Giammarino (my commit- tee co-chairs) for their encouragement and constant support during the completion of this thesis and the job market process. They generously shared their time and knowledge with me. I am deeply grateful to the members of my committee Kai Li, Russell Lundholm and Hernan Ortiz-Molina for their help and constructive comments. I thank Kai for writing a recommendation letter for my job market applications. I would like to sincerely thank the faculty members in the Finance Division at the Sauder School of Business. Each of them helped me in different ways throughout the PhD program. I thank my fellow PhD students for many long and valuable discussions. Milka Dimitrova and Gonzalo Morales deserve a special mention. I thank the University of British Columbia and the Sauder School of Business for providing financial support to complete the PhD program. I would not have completed my thesis without the relentless support of my family. I would like to thank my wife Paula and my daughter Catalina for helping me cope with difficult times and for completing my life. Pablo C. Morán Villar Vancouver, June 2013 viii Dedication To my wife Paula, my daughter Catalina and my parents Olivia and Pedro. ix Chapter 1 Introduction Financial decisions are inherently dynamic. In this thesis I take a closer look at two dynamic aspects of mergers and acquisitions (M&As). First, M&A activity follows waves over time within industries. Further, these waves of activity peak at times of high stock market valuation. Second, acquirers experience a small stock price reaction upon deal announcement on average. However, recent evidence suggests that this announcement stock price reaction shows a declining pattern along merger waves1. The interaction of these two empirical regularities is the unifying theme of this thesis. By studying the sources and consequences of these dynamic patterns, my work sheds light on whether M&A activity is an efficiency-enhancing activity and what motivates firms to participate in these types of transactions. The first two essays evolve around the idea that the existence of merger waves suggests commonality in deal motivation (i.e., a common driver). If there exists a common driver, it is reasonable to argue that a well-functioning stock market will partially anticipate who will be reacting and when. Consequently, the research questions behind the first two essays are: Does learning (anticipation) by the mar- ket explain the declining pattern in announcement returns of acquirers?; Could this learning process help explain the ‘poor’ performance of acquirers?. In the first essay I propose a dynamic model that seeks to understand the interplay between acquisition timing and market anticipation. The second essay performs a battery of tests for the basic predictions of the model. Based on my analysis and results, I argue that failure to properly capture market anticipation can vastly alter the inter- 1I will use stock price reaction, announcement effects and abnormal returns as synonymous here- after. Along the same lines, I will refer to M&As as mergers, acquisitions or takeovers when the differences in the type of transaction are not important. 1 pretation of existing evidence based on announcement effects. The third essay examines what are the underlying forces behind merger waves. I address this question by studying why merger activity is correlated with stock market valuation. To this end, I first show that merger activity and equity issues are positively correlated at the industry level. I then examine the dynamic alloca- tion of proceeds towards different uses from equity issued at times of high merger activity. The premise of this empirical exercise is that the use of equity proceeds is informative about what the manager believes about the stock price level. I ex- plore the predictions of three hypotheses and reject the notion that times of high merger activity are systematically associated with equity overvaluation. My results support hypotheses based on investment and financial frictions. The literature documenting the existence of merger waves and the relationship with the stock market in the U.S. starts with Nelson (1959). Golbe and White (1988) provide a more comprehensive account of the early U.S. evidence and Clarke and Ioannidis (1996) show that in the U.K., the stock market Granger causes both the number and the value of mergers. Jovanovic and Rousseau (2001) docu- ment that mergers and the stock market in the U.S. (both scaled by GDP) are corre- lated (ρ = 0.6). Mitchell and Mulherin (1996) report that the intensity of takeover activity varies across industries in a non-random fashion and Andrade, Mitchell and Stafford (2001) provide a broad discussion about the origins of merger waves. Harford (2005) and Rhodes-Kropf, Robinson, and Viswanathan (2005) study em- pirically the determinants of merger waves. Merger waves require both a motivation for deal initiation and a source of clus- tering. An influential strand of the empirical literature argues that M&As are the least expensive way to restructure after shocks that induce changes in the optimal industry structure (Mitchell and Mulherin, 1996; Andrade et al., 2001; Harford, 2005). Thus, rational managers react to these shocks to fundamentals by optimally reallocating assets. Merger activity varies over time as shocks to fundamentals ar- rive discretely over time. Cross-industry variation in merger activity results from different shocks affecting different industries over time. Jovanovic and Rousseau (2002) offer a theoretical framework rooted in the neoclassical approach to explain M&A waves. In the model, firms are heterogeneous in the marginal value of cap- ital. Thus, high-productivity firms have incentive to buy low-productivity firms. 2 Industry shocks of different sorts bring dispersion in productivity and thus merger waves. The second strand of the literature posits that merger waves are caused by waves of overvaluation in the equity market. The core prediction is that when a rational manager faces an irrationally high stock price, the manager finds it prof- itable to arbitrage this mispricing by undertaking stock swap acquisitions. This action benefits the acquiring shareholders at the expense of the selling sharehold- ers. At the heart of this explanation for merger waves lies the assumption of a rational manager making decisions in an irrational financial market. The ques- tion that remains unanswered is why target shareholders would accept overvalued shares as payment. Shleifer and Vishny (2003) argue that short horizon managers in overvalued targets are willing to accept (more) overvalued equity from bidders because it al- lows them to cash out. Rhodes-Kropf and Viswanathan (2004) propose a model in which information frictions induce target managers to accept overvalued equity. In particular, takeover synergy valuation errors are correlated with the overall valua- tion error. Thus, target managers cannot tell apart high bidder valuation stemming from overvaluation or from expected synergies. The overestimation of synergies becomes more severe during market valuation peaks, which makes stock offers more likely to be accepted at these times. As for acquirers’ performance during merger events, the typical results indicate that acquirers obtain a small stock price reaction. Leeth and Borg (2000) show that low announcement effects were observed even for the 1920s merger wave, despite the numerous economic and regulatory changes since then. Given this evidence, the issue of why bidders engage in takeovers has motivated a great deal of research. Studies on the cross section of bidder announcement effects have mainly relied on agency or managerial (behavioral) biasses to explain the low stock price reactions for bidders. For example, existing evidence indicates that bidder stock price reactions are negatively related to size (Moeller et al., 2004) and free cash flow (Harford, 1999), and positively associated with leverage (Maloney et al., 1993). These results are conventionally interpreted as supportive of the agency cost of free cash flow hypothesis (Jensen, 1986). More recently, Masulis, Wang and Xie (2007) find that acquirers with more 3 antitakeover provisions experience significantly lower announcement returns than those with fewer defenses. Datta, Iskandar-Datta and Raman (2001) show that equity-based compensation is positively associated with the bidder stock price performance around the acquisition announcement. For an extensive sample of takeovers, Eckbo, Betton and Thorburn (2008) find that the lowest announcement return for bidders corresponds to the subset of deals with large bidders, all-stock payments and public targets. On the other hand, the highest stock price reaction is found in deals with small bidders, all-stock and private targets. In terms of the time-series patterns of bidders’ announcement effects, Fuller, Netter and Stegemoller (2002) document a declining pattern in stock price reac- tions for serial bidders2. Klasa and Stegemoller (2007) find support for the idea that the best targets are acquired first, and that the declining pattern in announcement effects reflects a declining investment opportunity set. Billet and Quian (2008) attribute this declining pattern to a self-attribution bias in CEOs that leads to man- agerial overconfidence. Aktas, de Bodt and Roll (2009, 2011) argue that learning (which improves selection and valuation abilities) from deal to deal reduces the risk in each subsequent target. This lower integration risk leads bidders to bid more aggressively for each subsequent target. Bouwman, Fuller and Nain (2007) argue that different theories of merger waves imply different patterns of quality for deals announced inside a merger wave. They compare short-run and long-run abnormal returns for deals inside and outside high valuation times for the US and the UK. Their findings indicate that while short- term abnormal returns are higher in high valuation markets, the opposite pattern holds for long-run abnormal returns. Duchin and Schmidt (2013) also report that in-wave deals show a lower long-run abnormal return than that of out-wave deals. They report that merger waves are characterized by greater volatility and analysts’ forecasts dispersion. Duchin and Schmidt argue that merger wave times are times of weaker information environment and governance (i.e., in-wave deals are worse deals than out-wave deals). Carow, Heron and Saxton (2004) is the first paper documenting that bidders’ stock price reaction declines along merger waves. McNamara, Haleblian and Dykes 2The terms repeat, serial or frequent bidder refer to firms that acquire more than one firm in a relatively short period of time. 4 (2008) and Goel and Thakor (2010) provide more support for the same finding. These papers relate the timing of the deal to the quality of the acquisition, in which better deals are announced first. While McNamara et al. (2008) adhere to the first mover advantage story, Goel and Thakor (2010) argue that late deals are of lower quality as their motivation follows from managerial motives (envy). The current thesis makes several specific contributions to the literature. The first essay contributes in three fronts. First, to my knowledge, the model presented in the first essay is the first real options model that delivers predictions about conta- gion effects. This extension allows for the study of interactions among anticipation, contagion and investment timing in sequential investments in general. Second, competing theories about the declining pattern in bidders’ stock price reactions in merger waves attribute the effect to a declining pattern in deal quality (e.g., McNa- mara et al., 2008; Goel and Thakor, 2010). My model demonstrates that how the market learns over time is what matters for the stock price reactions and not the pattern of deal quality along the merger wave. In fact, I show that when the market has complete information about the synergy gain, a counterfactual empirical impli- cation emerges: Deal announcements lead to no stock price reactions even if the economic gain for bidders is strictly positive. Finally, the model shows how the market learning continues even if there are no announcements in the market. This prediction implies that any ad-hoc adjustment for the anticipation effect based on what the econometrician can observe is potentially incomplete. The second essay contributes by showing that the information channel that drives the action in the first essay’s model exists in the data. In particular, I show that the predictable portion of timing (including deal quality) does not drive the declining pattern in bidder announcement returns in a merger wave. Only the un- expected portion of timing drives the pattern in announcement returns. Second, consistent with previous empirical research, I find that contagion effects arise from more than one source. Consistent with the model’s predictions, I show that part of the observed contagion effect is driven by incomplete information about deal announcement timing of future deals. Finally, through the study of stock returns comovement, I provide evidence consistent with the notion that the market learns not only from actual announcements but also from non-announcements or inaction events by bidders. 5 Thus, my theoretical and empirical results in the first two essays are the first to systematically tie anticipation, contagion and deal timing in the M&A literature. The empirical literatures on contagion effects (e.g., Eckbo, 1983; Singal, 1996; Shahrur, 2005; Akdogu, 2009) and anticipation effects (e.g., Shipper and Thomp- son, 1983a; Asquith, Bruner and Mullis, 1983) have evolved to a large extent inde- pendently and away from the literature on merger waves. One of the main lessons in my model is that the existence of merger waves has implications for market an- ticipation and is potentially an important source of contagion effects. Overall, I show theoretically and then confirm empirically that the dynamics of market an- ticipation pose a potentially serious challenge for the interpretation of the bidder’s stock price reaction as a measure of actual economic gain in M&A deals. The last essay in this thesis contributes in two ways to existing literature on the drivers of merger waves. First, most studies arguing that windows of equity overvaluation drive waves of merger activity use some form of decomposition of the market-to-book ratio. To some extent, defining the ‘correct’ valuation involves arbitrary decisions by the researcher. In my empirical analysis, I focus on man- agerial decisions as a window into managers’ beliefs about the level of the stock price. In particular, I use the dynamics of proceeds allocation from equity issues to study the forces behind merger waves. Papers studying managerial decisions to uncover market timing motivation include Jenter (2005) who looks at equity issues and insider trading, Dittmar and Dittmar (2008) who study the time-series patterns of equity issues and repurchases and Jenter, Lewellen and Warner (2011) who show that firms are able to time the market through the sale of put options. As a second innovation, I incorporate in my analysis predictions implied by two alternative hypotheses of merger waves. Neither the proceeds allocation ap- proach described above nor these two alternative hypotheses have been explored in the existing literature on merger waves. In the last essay I find that, relative to equity issues in years with low merger activity, the allocation of proceeds raised in high merger years shows patterns that are inconsistent with the view that man- agers believe the stock is overvalued at times of high merger activity. Overall, the evidence I provide supports the notion that investment and financial frictions are important elements behind the dynamics of merger activity. 6 Chapter 2 Anticipation and Timing in Merger Waves: Theory If the deal enriches an acquirer’s shareholders, the statistics say, it is probably an accident ... it’s not that all deals fail. It’s just that timing appears to be everything. Deals at the very beginning of a merger cycle regularly succeed. It’s the rest that fall flat. — The New York Times. Mergers in a Time of Bears, Feb 26, 2008 2.1 Introduction Academic research on mergers and acquisitions consistently finds that acquirers’ stock price reaction upon deal announcement is small on average3. Although it is well understood that the stock price reaction to deal announcements only captures the unanticipated component of the total economic value of the event, the dynamics of market anticipation has not been widely studied. The implications of market anticipation can be potentially serious once we account for the fact that M&A activity clusters by industry and over time. In particular, the existence of these so- called “merger waves” suggests a source of deal interdependence (i.e., a common driving force) that has implications for what the market knows right before deal announcements. Indeed, one might conjecture that the pattern of timing mentioned in the above quote4 is consistent with a dynamic process in which the information held by the market about forthcoming deals improves along the merger wave. 3Eckbo, Betton and Thorburn (2008) provide a comprehensive account of existing research in the area. 4The quote refers to findings in McNamara, Haleblian and Johnson (2008). 7 In this chapter I offer a theoretical analysis of the interplay between merger announcement timing and market anticipation within a merger wave5. I propose a model of merger waves in a real options setup that delivers predictions about cross- section and time-series aspects of stock returns during merger wave episodes. The model borrows insights from Grenadier (1999) to study merger timing and the stock price reaction to merger announcements for acquiring firms and their rivals. The model predicts a time-series pattern of declining bidders’ stock price reactions along the merger wave. Although the model predicts that better (or less costly) deals are announced first in a merger wave, the declining pattern in the bidders’ announcement effects arises due to the fact that late deals are more anticipated by the market. Indeed, I show that with complete information there should not be any pattern in bidder abnormal returns, regardless how deal quality is sorted within a merger wave. In the model, the declining pattern in bidders’ stock price reactions reflects the dynamics of market learning and anticipation along the merger wave. Furthermore, the informational interdependence between deals in the model induces a cross- sectional pattern of stock price reactions for firms that will eventually announce a deal. In particular, the model delivers contagion effects on stock returns from early bidders’ announcements (the leader) to bidders in late deals (the follower). The model also suggests that the market learns not only from early bidders’ actual announcements, but also from early bidders’ inaction. In other words, lack of activity by early bidders can also be potentially informative to the market about the timing and value of follower deals. The model in this chapter is related to the theoretical literature that studies the timing of mergers and takeovers in a real option setup (Lambretch, 2004; Lam- brecht and Myers, 2007; Morellec and Hackbarth, 2008; Morellec and Zhdanov, 2005; Thijssen, 2008). Among these, only Morellec and Zhdanov (2005) feature announcement effects in acquisitions. However, in their model there is no deal interdependence, which is the key ingredient I use to obtain deal clustering and contagion effects. The current paper is also related Goel and Thakor (2010) and Toxvaerd (2008). In the first paper, a potentially idiosyncratic and value-increasing 5The first paper conjecturing a link between anticipation and merger waves is Mitchell and Mul- herin (1996). 8 deal induces a chain reaction of deals. The initial deal triggers an increase in CEO compensation, which triggers other CEO’s envy who try to increase their com- pensation by pursuing potentially value destroying but size enhancing deals. This model predicts a decreasing pattern in deal quality (and economic gains) along a merger wave just like my model does. In Toxvaerd (2008), the timing of deals balances the value of waiting and the cost of being preempted. This trade-off re- sults from the assumptions that there is imperfect knowledge about merger gains (the real option element) and also a relative scarcity of target firms (musical chair effect). The models in Goel and Thakor (2010) and Toxvaerd (2008) differ from mine in two key aspects. First, although both papers focus on the timing of deal an- nouncements, they do not focus on how the market learns from these announce- ments. In other words, these models do not address market anticipation and an- nouncement effects for bidders and rivals as I do in this chapter. Second, and more subtly, the channel of strategic interdependence between bidders is different. In Goel and Thakor (2010), agents interact directly through their preferences (CEOs care directly about what others do). In Toxvaerd (2008), bidders interact mostly through the constraints imposed by the actions of others (i.e., winner takes all). In my model the interaction between bidders is purely informational. In other words, my model does not feature payoff or congestion externality between bidders; all the interaction is through the formation of bidders’ beliefs about the synergy gains. 2.2 Model Setting The model is about corporate acquisitions (i.e., takeovers or mergers) in a setting where information about synergy gains is initially dispersed among bidders. In order to isolate the effect of interest, I assume away traditional sources of cross- sectional deal heterogeneity such as deal form or method of payment. In the model, the bidder’s objective function is to maximize the value of its acquisition opportu- nity in a setting with no agency problems and all-equity financing. I assume that time is continuous and indexed by t ∈ [t0,∞). The game starts at time t0 when two firms i = 1,2 in a given industry receive an acquisition opportu- nity (a real option). Time t0 could be understood as the arrival time of an industry 9 shock. In this setting, each potential bidder must decide when to proceed with the acquisition. I assume that there is a large set of potential target firms. Further, from the bidder’s point of view, any potential target firm is homogeneous in terms of the synergy gain. Bidders face no competition when buying a target. Thus, I refer to i as bidder, target or deal when no confusion arises. Figure 2.1 depicts the time line of the game. The first bidder will acquire its target at time t̃L, becoming the leader and the second bidder will become the follower, with acquisition time t̃F . These two acquisitions comprise a merger wave in the model. For simplicity, I assume that the arrival of the acquisition option and the identity of the option holders are common knowledge6. The pre-shock market value for the typical target firm in the industry is denoted by Vt and follows the diffusion process: dVt Vt = µdt+σdWt , Vt0 = v0 (2.1) where Wt is a standard Brownian motion with natural filtration denoted byFt . Pa- rameters µ and σ are assumed constant, with µ > 0 and σ > 0. I assume universal risk neutrality and a constant risk-free rate r that satisfies δ = r− µ > 0. In the model, if bidder i takes over any potential target firm, the new (ex-post) value of the target firm’s assets is given by (1+ θi)Vt , where θi is a synergy gain (defined below) captured by bidder i. At any feasible acquisition time t, the realized gain for bidder i from completing an acquisition is defined as: g(Vt ,θi) = θiVt −A (2.2) where A is an integration or search cost that is sunk once the deal is complete7. The value of θi is initially unknown to any deal participant, including bidder i. However, the information about this parameter is dispersed among bidders. The 6This assumption implies that at time t0 the market becomes aware of forthcoming deals and impounds the expected gains in the bidders’ stock price. This assumption could be relaxed by em- bedding the model into a larger structure (e.g., a regime model). However, the qualitative analysis that follows would remain unchanged. 7This payoff for the bidder can be understood as the reduced-form payoff. I discuss this issue at the end of this chapter. 10 ex-ante synergy parameter for bidder i is defined as: θ̃i = α+(1+κi)B̃i+ B̃ j where j denotes the ‘other’ bidder and κi ∈ (1 2 ,1 ) is an (observable) characteristic of bidder i, which without loss of generality is assumed to obey the ordering κ1 > κ2. B̃i is a random variable whose realization is bidder i’s private information about the profitability of the deals in the wave. Bidders’ private information (signals) is the realization of a discrete uniform distribution: Bi ∈ {−ε,0,ε} with probability 13 for all i. I use small caps (bi) to refer to a generic realization of a private signal, and the actual outcome (e.g., ε) to refer to a specific realization of the private signal. In this setting, the realized value of the synergy parameter for bidder i depends on the complete profile of bidder signals, θi = θ(b1,b2;κi) = α+(1+κi)bi+b j. In other words, the private information held by bidder i is relevant for all deals in the merger wave. I assume that although bidder j and the market can observe and learn from bidder i’s actions, bidder i’s realized payoff is not observed by any party other than bidder i. Thus, the acquisition time of bidder i only reveals i’s private signal bi, and not the true value of θi8. Along the same lines, in the model bidders are not allowed to exchange their private information directly, although bidders have incentive to do so. I assume that this exchange is forbidden, too costly or non-credible. Two additional initial conditions simplify the subsequent analysis. First, I as- sume that α > (2+κ1)ε . This assumption implies that θi is positive for any sig- nal profile. Thus, all acquisitions in a merger wave are carried out eventually. Second, to avoid equilibria in which acquisitions are optimally carried out at the beginning of the game, I impose conditions so that waiting is optimal at the be- ginning of the game. In particular, upon receiving its acquisition option and pri- vate information at time t0, bidder i’s best estimate of the synergy gain parame- ter is E[θi|bi] ≡ θ̂i,t0 = α + (1+ κi)bi. Thus, this initial condition requires that g(θ̂i,t0 ,v0) = θ̂i,t0v0−A < 0 for both bidders and for any realization bi. In summary, the timing of the game is as follows. First, at time t0 both bid- 8Even though bidder j will have complete information after observing bidder i’s timing, the market will still have uncertainty about bidder j’s private information. 11 ders receive an acquisition option. Second, the signal profile (b1,b2) is realized. The signal bi is privately observed by bidder i at time t0. Next, acquirer i picks the optimal acquisition time so as to maximize the value of the opportunity to ac- quire, given its information at that time and the equilibrium strategy of bidder j. Acquirers face no competition for the control of target firms. Even though the setting above resembles the typical asymmetric information game with observed actions, there are two key differences. On the one hand, the payoff function of bidder i only depends on bidder i’s actions. This lack of direct payoff externality implies that there is no incentive for bidder i to preempt other bidders, or to signal (or hide) private information from other players. On the other hand, the timing of moves is endogenous in the model; whether a bidder decides to be an early or a late mover is an equilibrium outcome. 2.3 Acquisition Strategies In this section I derive bidders’ equilibrium strategies and beliefs. I start with the complete information setup. As pointed out earlier, the interaction in the model comes from the expectation of θi. Further, all the heterogeneity in the model also comes through the same parameter. Thus, the complete information setup provides a benchmark case which can be extended to the setup with disperse information. 2.3.1 Complete Information Under complete information there is no strategic interaction between bidders in the model because each of them perfectly observes the value of θi. In other words, once the signal profile (b1,b2) is realized, it is revealed to all agents in the model, including the market. Under this condition, information is symmetric among deal participants. Players still face the uncertainty regarding the value of the target firm Vt . However, this is the standard decision theory problem (optimal stopping time) in real option models. The value of the investment opportunity (acquisition option) for bidder i at time t is given by: G(Vt ,θi)≡ sup τ>t E [ e−r(τ−t)g(Vτ ,θi)|Ft ] (2.3) 12 where τ is a (random) stopping time that is adapted to Ft . The optimal stopping time for bidder i is defined by a trigger strategy as follows: τ i ≡ inf[t ≥ t0 : Vt ≥V ∗(θi)] where V ∗(θi) is the optimal (non-strategic) acquisition trigger (free boundary). I summarize the relevant results in the next proposition. Proposition 1: If the target’s assets in place and the bidder’s gain are defined as in 2.1 and 2.2 respectively, and the synergy gain θi is known, the value of the acquisition option 2.3 and the optimal acquisition trigger are given by: G(Vt ,θi) =  Aβ−1 ( Vt V ∗(θi) )β for Vt <V ∗(θi) θiVt −A for Vt ≥V ∗(θi) (2.4) V ∗(θi) = ( β β −1 ) A θi with β > 1 (2.5) Proof: See Appendix A  Proposition 1 shows that the timing of acquisitions is driven entirely by het- erogeneity in κibi. It follows that, within a merger wave and under complete in- formation about synergy gains, deals are optimally sorted from the most profitable to the least profitable (quality) acquisition. Under the conditions in Proposition 1, acquisitions are strictly sequential in the model unless bi = 0 for i = 1,2. Complete information about the synergy gain brings a counterfactual empir- ical implication: Deal announcements have no stock price reactions, even if the economic gain for bidders is strictly positive. Intuitively, once the bidder receives its acquisition option at time t0, the value of that real option is immediately im- pounded into the market price of the corresponding bidder firm. Although this is an extreme form of anticipation, this claim is, in general, neither new nor sur- prising. Malatesta and Thompson (1985) show a similar effect in a discrete time stationary setup9. The next proposition summarizes this result in the current setup. 9In their model, a firm has infinite projects to announce, and the announcement decision at each period follows a Poisson distribution. In their model the timing of announcements is exogenous and, given the constant intensity of event arrivals, the unexpected portion of the economic gain is constant 13 Proposition 2: In the complete information setup described in Proposition 1, ac- quisition announcements are predictable. Thus, • Timing differences between deals follow strictly from heterogeneity in deal quality (κibi), • Bidder announcement returns are zero. Proof: See Appendix A  Proposition 2 highlights the need for incomplete information in order to deliver any pattern of announcement returns when takeover waves follow from common shocks that induce ownership reallocation. For example, Morellec and Zhdanov (2005) obtain announcement effects for acquisitions in a model that features in- complete information by outside investors but symmetric information among deal participants10. On a more applied side, Proposition 2 shows that any pattern of announcement returns along the merger wave does not need to be related to the true economic gain from a merger. At the extreme case of complete information, announcement returns are uninformative about deal quality or value creation. How- ever, measures of survival or duration are a good empirical proxy for deal hetero- geneity. 2.3.2 Heterogeneous Information The full setup of the game resembles a stochastic game in which the dynamics of the environment is driven by the value of the target firm (state variable) and the bidders’ actions up to time t. A behavior strategy for bidder i in this setting is a mapping from its information set into its action set at every time t. I restrict atten- tion to pure trigger Markov strategies11. Finally, both bidders must have rational beliefs about the signal of the other bidder at the start of each continuation game over time. 10As in the current model, in their paper the market’s ability to partially anticipate the deal and stock returns follows from the resolution of remaining uncertainty. In their model, there is no deal interdependence, which is the key ingredient I use to obtain deal clustering and contagion effects. 11This is not without loss of generality. However, if there exists a state vector that summarizes all payoff-relevant histories, this assumption is not restrictive in the current setup. 14 along the equilibrium path. The equilibrium concept is thus a Markovian Perfect Bayesian Equilibrium. In order to keep track of the acquisition history, I define two state variables. First, the indicator variable dit ≡ 1t≥τ i becomes 1 once bidder i bids at time τ i. The acquisition history at time t is then summarized by the vector dt = (d1t d 2 t ) ′. Given the initial assumptions, dt0 = (0 0) ′. Finally V̂t ≡ sup{Vs|t0 ≤ s ≤ t} denotes the running maximum of the pre-shock value of the target assets. Thus, the vector of public information is denoted as It = (dt V̂t Vt) ∈ It . Bidder i’s strategy is more formally defined as: σi(bi, It) : {−ε,0,ε}×It −→ Ait where the action set at time t is given by Ait = {Wait,Bid} while bidder i is still waiting to bid (t0 < t < τ it ), and Ait = /0 afterwards. Given the discrete structure of the signal space, player i and the market know when player j 6= i should optimally move if it is of a particular type. At these discrete points in the state space, bidder i learns something about bidder j’s type by observing his action Bid or Wait12. As the game is played in continuous time, this belief updating process might induce an immediate action from player j upon observing an action by bidder i. Thus, I allow for sequential actions at the same time t. In other words, a bidder bidding right after the other bidder is allowed to condition its decision on the observed action from the previous bidder, even though they are moving at the same time t. Given the lack of payoff externality in the model, this feature does not have a material effect on the solution. The trigger strategies were denoted in Proposition 1 as V ∗(θi). The fact that bidders do not directly interact through the payoffs allows me to keep the same strategy mappings in this section. However, I must recognize that the synergy gain parameter is now a conditional expectation. I define this conditional expectation as θ̂i,t ≡ E [θi|bi, It ], where the set of beliefs on b j carry all the interaction between bidders. The next proposition characterizes an equilibrium for the game. 12This feature, together with the heterogeneity in κ , simplifies substantially the solution of the game. Even though the state space of Vt is continuous, bidders only move at discrete and predictable points in the state space along the equilibrium path. The heterogeneity in κ ensures that these points do not coincide for different bidders and different types. 15 Proposition 3: Provided that bidder i is still waiting to acquire, an equilibrium for the incomplete information setup is characterized by the trigger strategy: ait = Bid if Vt ≥V ∗(θ̂i,t)Wait if Vt <V ∗(θ̂i,t) where: • V ∗(·) is defined as in Proposition 1, • θ̂i,t = α +(1+κi)bi +E[b j|It ] for the set of beliefs defined in the Appendix A. In equilibrium, the bidder with the highest interim quality (κibi) will be the leader in the merger wave. Proof: See Appendix A  Two properties of the acquisition threshold V ∗(θi) are of interest. First, it is strictly monotonic, decreasing and continuous in θi. This feature, together with the continuous time assumption, induces full revelation of bidder i’s private informa- tion upon deal announcement. Second, the threshold depends on the ratio Aθi . Then, heterogeneity in θi is observationally equivalent to heterogeneity in the integration cost A when there is complete information about the synergy gain (i.e., Proposition 1). This equivalence breaks down under incomplete information. While the invest- ment trigger V ∗(θi) is linear in A, it is a convex function of θi. This convexity is crucial in the current setup to distort the bidders’ optimal acquisition timing. Given these strategies and beliefs, I discuss the properties of this equilibrium in the next section. 2.4 Equilibrium Outcomes The model delivers several equilibrium configurations depending on the realized signal profile. In this section I focus on the pattern of acquisition timing and an- nouncement returns for different signal profiles. 16 2.4.1 Timing of Announcements In this section I discuss how the information friction in the model distorts the equi- librium timing with respect to that of complete information about the synergy gain. In the model, the second bidder will always bid with complete information about the synergy gain. Then, I can restrict attention to what happens to the optimal timing of the first bidder. Figure 2.2 shows the timing outcomes for different signal profiles. It is impor- tant to keep in mind that, conditional on a signal profile, the time between deals is still a random variable (with respect to Ft) in the model. Thus, the outcomes in Figure 2.2 should be interpreted path-wise. For example, assume that the realized signal profile is (b1,b2) = (0,ε). The figure shows that bidder 2 goes first at time t2, when its estimate of synergy gain is θ̂2,t2 = α + (1 2 +κ2 ) ε . Although bidder 1 is still waiting at time t2, bidder 2 learns at time t1 that bidder 1 is not of type ε . This non-announcement event allows bidder 2 to update its beliefs about the type of bidder 1, which changes θ̂2,t . Bidder 1 acquires with complete information at time t3 under this signal profile. The figure shows that imperfect information about synergy gains induces clus- tering for some signal profiles. Clustering is understood here as a situation in which both bidders optimally decide to acquire at the same time, even though they are heterogeneous. Under complete information, the only signal profile in which both bidders decide to acquire together is when both bidders are of (known) type bi = 0. This is the only case in which heterogeneity in κ does not affect the ac- quisition timing and both bidders optimally bid at the same time. However, under incomplete and disperse information, there is another signal profile (ε,ε) which results in clustered deals. Under this specific signal profile, both bidders differ in quality ex-ante and ex-post, but they both decide to acquire observationally at the same time. In particular, bidder 1 acquires first, and bidder 2 follows imme- diately after. Upon observing bidder 1, bidder 2 acquires right away because its belief that b1 = ε jumps to 1. Thus, bidder 2’s new estimate of the synergy gain is: θ̂2,t1 = α +(2+κ2)ε > θ̂1,t1 . Note that if the value of Vt reaches a level high enough to justify bidding with θ̂1,t1 = α +(1+κ1)ε for bidder 1, then the mono- tonicity of the strategy mapping implies that it is also high enough for bidder 2 to 17 acquire immediately after updating its beliefs. Figure 2.3 shows the timing outcome for four different signal profiles. In the figure, L and F stand for leader and follower respectively. Small and large caps denote the outcome with complete and disperse information respectively. As the figure shows, the strongest bidder ex-ante (i.e., largest κ) is not always the first to announce the acquisition. Even though in most signal profiles the announce- ments are sequential, disperse information induces both clustering and spreading of announcements depending on the realized signal profile. As the second bidder always bids with complete information in the model, I take a closer look at the timing distortion induced by incomplete information through the delay or rush for the leader in the merger wave. Table 2.1 shows all signal profiles with the corresponding timing. A check mark identifies the leader bidder in the table. The first thing to note is that the information friction in the model does not change the identity of the leader bidder. However, the timing is distorted. The last column in the Table 2.1 reveals whether the leader delays or rushes, with respect to the setup with complete information. The table also shows that the average difference between θi and θ̂i, condi- tional on bidder i being the leader, is zero, EB j [ θi− θ̂i|bi,bi > b j ] = 0. This ‘unbi- asedness’ is certainly not surprising given the Bayesian nature of bidders’ beliefs. Nonetheless, recall that the strategy mapping V ∗(θi) is convex in θi. This becomes important here as it implies that: EB j [V ∗(θi)]>V ∗ ( EB j [θi] ) In other words, in expectation the leader speeds up his acquisition time with respect to that under complete information discussed in the previous section. Flip- ping the coin, and given that the second bidder bids with complete information, the time between deals grows on average with disperse information. 2.4.2 Announcement Effects Announcement effects in the model are a consequence of the dynamic updating of synergy gains by the market every time a bidder picks an action. The market only observes the vector of public information It in the model. Denote the market’s es- 18 timate of the synergy parameter for bidder i at time t as θ̄i,t ≡ E [θi|It ]. In addition, the market does not take any action in the model. Thus, the relevant ‘times’ are still t1, t2, t3 and t4 in Figure 2.2. At each of these times, the market is able to refine its partition of the type space of at least one bidder. If at any of these times a bidder chooses Wait, there will be an informative event even though there is no acquisi- tion announcement by any bidder. These events will be called non-announcement or inaction events. The difference between announcement and inaction events is important because the econometrician can only observe announcement events. My discussion in this section focuses on what the econometrician can see when bidders announce acquisitions. In order to simplify the discussion, I define the (dollar value) abnormal return for bidder i at event time tk as: ARi,tk = G(Vtk , θ̄i,tk)−G(Vtk , θ̄i,t−k ) where G(·) is defined in Proposition 1 and t−k is the time right before the event time tk. Note that in general G(·) is a continuous function of Vt ; optimality (smooth pasting) implies that this is true at the optimal bidding threshold as well. Thus, jumps in G(·) that bring announcement returns come from the discrete updates in beliefs about θi once events occur. As a shorthand notation, define, for an arbitrary stopping time τ > t, the expected discount factor as D(τ, t)≡ E [e−r(τ−t)|Ft]. As in the case of timing in the previous section, the pattern of announcement and contagion returns depends on the signal profile. Thus, in what follows I group signal profiles according to the predicted pattern. I start with a signal profile that delivers relatively rich dynamics in announcement and contagion returns. Profile (b1,b2) = (ε,0). This profile features sequential acquisition, with an ob- servable delay between announcements (see Figure 2.2). Right before t1, the mar- ket’s estimate of the synergy gain for bidder i is θ̄i,t−1 = α . Given that under this signal profile bidder 1 is of type b1 = ε , the first acquisition is at time t1, when Vt first hits V ∗(θ̂1). After observing this acquisition, the market learns bidder 1’s type. Then, the updated synergy gain for bidder 1 is given by θ̄1,t1 = α+(1+κ1)ε . 19 Thus, the announcement effect (in dollar value) for bidder 1 is given by: AR1,t1 = G(Vt1 , θ̄1,t1)−G(Vt1 , θ̄1,t−1 ) = [ θ̄1,t1Vt1−A ]−D(τ1t−1 , t1)[θ1,t−1 V ∗(θ̄1,t−1 )−A]> 0 where τ1t−1 > t1 (a.s.) denotes the optimal acquisition time for bidder 1 expected by the market right before the event time t1 (i.e., considering θ̄1,t−1 = α). AR1,t1 would be the announcement return in a world with independent bidders and incomplete information about the synergy gain. However, in the current setup with interdepen- dent bidders, after observing that bidder 1 revealed its type and that bidder 2 did not react immediately, the market learns that b2 6= ε . This event brings ‘bad news’ for both bidders. Thus, immediately after t1, bidder 1 experiences a downward adjustment: AR1,t+1 = (θ̄1,t+1 − θ̄1,t1)Vt1 < 0 where θ̄1,t+1 = α + ( 1 2 + κ1)ε . The same bad news affects bidder 2. However, bidder 1 revealed its type, which is ‘good news’ for bidder 2. The market’s revised estimate (netting out both effects) of bidder 2’s synergy gain is θ̄2,t1 = α + ( 1 − κ2 ) ε2 > α = θ̄2,t−1 . Denoting the pre-event expected optimal time for bidder 2 as τ2t−1 and the optimal acquisition time for the revised level of synergy (post-event) for bidder 2 by τ2t1 , the contagion effect for bidder 2 at time t1 is: AR2,t1 = G(Vt1 , θ̄2,t1)−G(Vt1 , θ̄2,t−1 ) =D(τ2t1 , t1) ( θ̄2,t1V ∗(θ̄2,t1)−A )−D(τ2t−1 , t1)(θ̄2,t−1 V ∗(θ̄2,t−1 )−A)> 0 Note that AR2,t1 is likely to be small, and less than AR1,t1 +AR1,t+1 , for two reasons. First, there is a (stochastic) discounting effect. Note that τ2t1 < τ 2 t−1 for all ω ∈ Ω, which implies that D(τ2t1 , t1) > D(τ2t−1 , t1). Second, the jump in θ2 is positive but smaller than that for θ1. At the event time t1, the market is still uncertain about the bidder 2’s type. The estimate considers the possibility that bidder 2 is of the low type. However, the analysis of timing before showed that if bidder 2 is of type b2 = 0, then the optimal acquisition time, conditional on bidder 1 being an ε-type, is t3. Then, given the 20 signal profile under study here, at this time bidder 2 will bid and the market’s new estimate of the synergy gain will be θ̄2,t3 = α+ε . Thus, the stock price reaction to bidder 2’s own announcement at time t3 will be: AR2,t3 = G(Vt3 , θ̄2,t3)−G(Vt3 , θ̄2,t−3 ) = [ θ̄2,t3Vt3−A ]−D(τ2t−3 , t3)[θ̄2,t−3 V ∗(θ̄2,t−3 )−A]> 0 where τ2t−3 is the optimal timing that a bidder 2 with θ̄2,t − 3 = θ̄2,t1 = α +(1−κ2) ε2 would have. Even though AR2,t3 is positive in this case, it is still an attenuated version of the true economic gain for bidder 2. The attenuation bias is captured by: G(Vt3 , θ̄2,t3)−AR2,t3 =D(τ2t−3 , t3) [ θ̄2,t−3 V ∗(θ̄2,t−3 )−A ] > 0 A similar logic also indicates that the stock price reaction to bidder 1’s own announcement is an attenuated version of the expected synergy gain. Bidder 1’s economic gain is anticipated because the market knows with certainty the arrival of the option to acquire at time t0. The abnormal return in the model thus reflects only uncertainty about the timing of the acquisition. This effect is also a factor in bidder 2’s anticipation effect. However, the contagion effect from bidder 1’s announcement makes bidder 2’s abnormal return an even more attenuated version of the expected synergy gain. The fact that the market only observes actions, as opposed to outcomes, implies that the deal announcement by bidder 2 should also be informative about the gain for bidder 1. This forward interdependence brings a post-announcement stock price reaction for bidder 1. This effect is consistent with a continuation in abnormal returns after own announcements (akin to the post-earnings announcement drift). Recall that θ̄1,t−3 = θ̄1,t+1 = α+( 1 2 +κ1)ε , given the market’s belief about bidder 2. However, once bidder 2 reveals it is of type b2 = 0, the market updates the estimate for bidder 1 back to θ̄1,t3 = α +(1+κ1)ε . This implies a drift effect for bidder 1 of: AR1,t3 = (θ̄1,t3− θ̄1,t−3 )Vt3 > 0 at the time of bidder 2’s announcement. 21 Remaining Profiles: The other signal profiles offer less action that can be observed by the econometrician. Under the remaining profiles, the market is able to predict with more certainty the acquisition timing after key relevant non-announcement events. In particular, note that at t3 or at t4 in Figure 2.2 and depending on whether any bidder received a high signal bi = ε , the market will know the exact value of the synergy parameter for both deals in a merger wave. Complete predictability of θi implies that Proposition 2 applies. Under profiles (−ε,ε), (ε,−ε), (−ε,0), (0,−ε), (−ε,−ε), most (if not all) price adjustments are carried out during ‘non- announcement events’ that precede the actual deal announcement. Then, the an- nouncement brings little (or no) surprise to the market. Despite this lack of discrete observed action, the bidders’ pre-announcement stock price dynamics captures these effects on a more continuous basis. For exam- ple, suppose the signal profile is (−ε,−ε). The model predicts that both bidders will acquire after time t4 (see Figure 2.2). The market does not know this ex-ante. However, at time t1 the market realizes that bidder 1 is not of type ε . This brings a negative valuation effect for both bidders. At time t2, the market learns that bid- der 2 is not of type ε either, which brings another negative ‘innovation’ for both bidders. The process continues with another negative and common shock for both bidders at times t3 and t4. By the time of the actual announcements, the market knows perfectly that both bidders are of type bi =−ε , and thus the announcement effect is zero for both bid- ders despite having a deal with positive gain. Even though full anticipation implies no announcement return, it also implies a higher degree of co-movement between the two bidders’ stock returns during the pre-announcement period. More broadly, this increase in correlation between both bidders’ stock returns also implies a de- crease in the correlation between each bidder’s returns and the returns of other assets in the economy. These patterns of co-movement in bidders’ stock returns can be captured by the econometrician13. 13Early literature on event studies identified a related effect, which is that of event clustering on inference. The argument in that literature is about how cross-sectional dependence in returns affects the standard errors of abnormal returns (see Schipper and Thompson, 1983b). Kothari and Warner (2008) provide an updated discussion for this inference problem. The comovement effect I describe in the current chapter goes beyond the inference problem to argue that cross-sectional dependence can also arise due to learning by the market. 22 2.5 Predictions and Related Empirical Evidence In this section I discuss the predictions related to stock price information effects brought by the announcement of deals in a merger wave. These predictions in the model can be categorized in two groups. First, conditional on a merger wave having arrived, early deals should have a higher stock price reaction than late deals. Early deals are those announced by bidders with higher expected value creation. Higher expected gains make waiting more expensive for bidders, which induces them to be the leader in the merger wave. In addition to this decreasing pattern in perceived deal quality, early deals have a higher stock price reaction than late deals because early deals are less anticipated by the market. This effect leads to a declining pattern in abnormal returns along a merger wave. Second, given that the timing of announcements reveals bidders’ private infor- mation about a common value component, the market and the soon-to-be bidders update their beliefs about the synergy gains at the time of peer acquisition an- nouncements. This information spillover gives rise to two potential effects. On the one hand, there is a contagion effect that comes from the fact that early deals are informative to follower bidders and to what the market knows about follower deals. Thus, the model predicts that the stock price of follower bidders should react to the announcement of early deals, provided that early deals are not fully anticipated. As long as the market is still uncertain about the true value of the synergy gain, additional announcements will keep inducing these patterns of contagion returns. In contrast to these effects observed by the econometrician in the level of returns, the arrival of a merger wave should induce an increased level of co-movement in the rivals’s stock returns. This increase in the cross-correlation of returns follows from the fact that the market learns about the value of the synergy gain of both bidders when they decide to delay their acquisition. In the model, I described these informative times as non-announcement events. Given that neither the arrival of the merger wave nor the non-announcement events are observed by the econome- trician, empirical tests on second moments of bidders’ stock returns should com- plement tests on first moments of returns (contagion abnormal returns). 23 2.5.1 Timing and Anticipation Affects Carow et al. (2004) is to the best of my knowledge the first paper documenting em- pirically the timing effect (i.e. that bidders in early deals tend to outperform those in late deals). Their sample includes 520 U.S. deals in nine industries that com- prise 14 mergers waves14. They find that combined cumulative abnormal returns (CARs) are larger for early acquisitions in a merger wave. Further, early acquires also perform better than late acquirers. They argue that the declining pattern in CARs is consistent with the idea that early bidders are able to gain a first mover advantage by capitalizing on their superior information to identify and act upon an opportunity. McNamara et al. (2008) and Goel and Thakor (2010) also report that US bid- ders tend to realize larger CARs if they acquire early in a merger wave. Both papers use different procedures to identify waves. The first paper uses the procedure in Carow et al. (2004) and Harford (2005). Goel and Thakor (2010) use the aggregate price/earnings ratio in a procedure that identifies merger waves through periods of high market valuations. Both papers relate the timing of the deal to the quality of the acquisition, in which better deals are announced first. While McNamara et al. (2008) adhere to the first mover advantage story, Goel and Thakor (2010) argue that late deals are of lower quality as these follow from managerial motives (envy). A common feature in the three papers above is that the declining pattern in economic gains is empirically implemented as a declining pattern of announce- ment effects (i.e., CARs). In other words, any potential anticipation by the market is implicitly assumed to be independent of the timing of deals. Thus, on average, CARs capture proportionally the economic gains from the event. This interpre- tation of CARs is explicit in McNamara et al. (2008) as they state: “Our results show acquiring firms can benefit from early acquisitions ... acquirers moving early within an industry acquisition wave not only outperform those acquiring later, but on average achieve an economic gain.”, p.124. The model I present in this chapter 14The authors acknowledge that they are not aware of any method to identify mergers waves. Thus, they start with all 2-digit SIC industries with more than 30 deals between 1979 and 1998. Then, they keep an industry if it has at least 10 deals in a given year. For the industries remaining, they identify the peak year. The wave start is identified by working backward from the peak until the time the number of deals is one third of those in the peak time. The end of a wave is identified in the same fashion. 24 shows that this interpretation is problematic. In particular, the model demonstrates that timing and anticipation are not independent when bidders can optimally choose the acquisition time in the context of merger waves. The literature of deal anticipation goes back to at least Shipper and Thomp- son (1983a) and Asquith, Bruner and Mullis (1983) in the context of corporate acquisition programs15. These early contributions emphasize that, once a bidder announces an acquisition program, the stock price reaction to the actual acquisi- tion announcements should be zero on average (Schipper and Thompson, 1983a) or should decline quickly with each subsequent announcement (Asquith et al., 1983). Consistent with this idea, Becher (2009) finds that banks that become bidders in the 1990s have a large and significantly positive stock price reaction during the passage of the Riegle Neal Act in 1994. More recently, Cai et al. (2011) provide evidence consistent with the idea that anticipation and information spillovers matter for bidder announcement returns. In particular, they argue: “(I)f the market anticipates a bid announcement, actual announcement returns will not fully capture the wealth effects ... any anticipation effects have the potential to drastically alter the perceived wisdom.”, p.2. They find that, after a dormant period (12 months) with no deals in an industry, initial bidders (presumably less anticipated) have significantly positive abnormal returns. Interestingly, their findings indicate that, once this anticipation effect is added to subsequent bidders’ announcement effect, some well accepted results are reversed or attenuated. Somewhat surprisingly, the deal anticipation literature has evolved indepen- dently from the literature on mergers waves. Cai et al. (2011) emphasize that their definition of initial bidder needs not coincide with an early bidder in a merger wave. An initial bid might or might not be followed by a cluster of subsequent deals. In fact, they report that 67% of their initial bidders announce their acquisitions at least 12 months before the start of merger waves or 36 months after the end of merger 15The discussion in the current chapter is related to the effect of anticipation on the measurement of bidders’ economic gains. Edman et al. (2012) focus on the bidders’ incentive to launch tender offers when the target price includes –in addition to a discount for poor management– an anticipation effect. Betton et al. (2012) study bidders’ incentive to renegotiate acquisition terms when the target prices run up during deal negotiations. 25 waves16. 2.5.2 Contagion Effects Eckbo (1983) is the seminal paper on rivals’ reaction to merger announcements. His empirical design seeks to reject the view that the rivals’ positive reaction in horizontal deal announcements follows from the increased likelihood of collusion in the industry. Eckbo (1983) argues that if this increase in rival valuations fol- lows from the collusion hypothesis, then announcement of antitrust challenges to open deals should reverse rivals’ initial gains. His results reject the idea that ini- tial gains come from an increase in market power for industry rivals. This paper has motivated a great deal of subsequent research that attempts to identify firms’ interactions by looking at the rival firms’ reactions17. For example, Shahrur (2005) studies the reaction of rivals, buyers and suppliers in horizontal deals. The findings indicate that rivals and corporate consumers react positively during deal announcements. Interestingly, he finds that the reaction of buyers, rivals and suppliers is positive when the combined CAR is positive. Sin- gal (1996) argues that, in addition to the market power hypothesis postulated by Eckbo, deal announcements might also lead to competitive and information effects on rivals. In the first case, good news for merging firms brings bad news for rivals. In the second case, good news for merging firms is also good news for rivals. Song and Walkling (2000) argue that rivals’ positive reaction in horizontal deals follows from the increased probability that those rivals become targets in sub- sequent deals. Akhigbe et al. (2000) study the stock price reaction of targets and their rivals at the announcement of deal agreements and deal terminations. The au- thors use the rivals’ reaction at deal termination announcements to disentangle the market power, the competitive and the information effects. Their findings indicate that rivals react positively to both agreements and terminations, which is consistent with the conjecture that rivals’ initial reaction is related to a higher likelihood of rivals being future targets. 16Cai et al. (2011) define merger waves with the procedure proposed by Harford (2005) 17An incomplete list of studies on contagion effects in corporate events other than mergers or takeovers include Lang and Stulz (1992) for bankruptcy announcements, Xu et al. (2006) for ac- counting irregularities, Gande and Lewis (2009) for shareholder-initiated class action lawsuits, and Hsu et al. (2010) for initial public offerings. 26 Under the same classification of rivals reactions, Akdogu (2009) studies bid- ders’ rivals in non-horizontal deals. She argues that by focusing on bidders’ rivals that go shopping in a different industry, her design minimizes the effects coming from the market power and the acquisition probability hypotheses. In addition, she focuses on deals in the telecommunications industry, after the Telecommunications Act of 1996. She finds that bidders’ rivals react negatively to acquisition announce- ments. This effect is more negative for rivals that are ‘close’ to the bidder in terms of size. Further tests suggest that the negative reaction to rivals is consistent with bidders gaining a competitive advantage through acquisitions. As in the case of anticipation, the literature on contagion effects has evolved to a large extent away from the literature on merger waves. One of the main lessons in the model is that the existence of merger waves has implications for market anticipation and is potentially an important source of contagion effects. 2.6 Final Remarks and Conclusion Merger waves are a pervasive phenomenon. In this chapter, I argue that the degree of market anticipation is fundamentally linked to the timing of deal announcements in a merger wave. I propose a model of merger waves in a real options setup that captures the interplay between deal timing and anticipation. The theoretical anal- ysis implies two broad predictions regarding market learning. First, early bidders in a merger wave experience a higher stock price reaction than do late bidders. Second, the announcement of early deals provides a signal to the market about the timing and value of subsequent deals. Thus, the model makes predictions about cross-section and time-series aspects of stock returns during merger wave episodes. The empirical testing of these predictions requires to keep in mind, and poten- tially control for, some of the model’s limitations. First, I assume a reduced form for the bidder’s payoff in the acquisition (equation 2.2). A framework that deliv- ers this payoff is as follows. Denote the surplus of the acquisition as W (Vt ,θi) = (1+ θi)Vt −A. The bidder and the target split this surplus in a Nash Bargaining fashion: s∗ = argmax [sW (Vt ,θi)]ρ [(1− s)W (Vt ,θi)−Vt ]1−ρ where ρ is the bidder’s bargaining power. The optimal solution implies that the 27 bidder gets: g(Vt ,θi;ρ) = ρ(W (Vt ,θi)−Vt) = ρ(θiVt −A) Thus, the reduced-form payoff in the model can be understood as a re-scaled ver- sion of the bidder’s actual payoff. Under all-equity financing or perfect debt con- tracting, this share of the acquisition’s surplus is captured by the acquirer’s share- holders. Thus, all predictions in the model can be tested using bidders’ stock re- turns. Another potential concern is the assumption that there is no payoff externality when companies merge. It is plausible to argue otherwise as industry rivals also interact in the product market. Furthermore, acquirers also compete in the market for control of target firms. The purpose of the model in this chapter is to highlight how an information channel (i.e., learning by observing) can potentially deliver the time-series patterns of stock price reactions observed in the data. The model serves this purpose well and provides additional testable implications. Empirical evidence shows that a sizable number of bidders experience a neg- ative price reaction upon deal announcement. The model fails to deliver this out- come. In this sense, the model can be understood as a model of the the dynamics (i.e., the slope) of the stock price reactions along a merger wave. My theoretical analysis on the link between deal timing and anticipation raises challenges for the interpretation of stock price reactions during acquisition an- nouncements. This is especially true for late deals in a merger wave. In the model, the anticipation cannot be fully captured by the subset of announcement events that the researcher can observe because non-announcement events are also informative to the market. For example, in an attempt to account for the effect of anticipation on follower bidders, Cai et al. (2011) add the own deal announcement return and the follower bidders’ reaction to initial bidder announcements. My theoretical re- sults imply that this approach may not capture the full extent of deal anticipation if the market also learns by observing inaction. Capturing the full extent of deal an- ticipation appears to be fundamental for the understanding of bidder performance in particular, and for the economic efficiency of the market for corporate control in general. 28 A. Proofs of Propositions. Proof Proposition 1. Given the discrete nature of the decision problem, the value of the option to acquire held by the bidder can be represented recursively as fol- lows: G(Vt ,θi) = max { θiVt −A,e−rdtEt [ G(Vt+dt ]} , where the first term denotes the payoff the bidder gets when it acquires (termina- tion payoff), and the second term denotes the payoff the bidder gets when it waits an instant dt (continuation payoff). In the continuation region and after some ma- nipulation, we obtain the Bellman equation for this problem: rG(Vt ,θi) = dG(Vt). The application of the Itô’s Formula delivers an ODE for the option value G(V ). After imposing the value matching and smooth pasting conditions, it is not difficult to show that the option to acquire held by the bidder is given by the first equation in Proposition 1, where β > 1 is the positive solution of the polyno- mial 1/2σ2β (β−1)+(r−δ )β−r= 0 and V ∗i = ββ−1 Aθi corresponds to the optimal acquisition threshold. For more details, see Dixit and Pindyck (1994). The following result is also well known and simplifies the proof of Proposition 2. Lemma 1: Take Vt as defined in the main text, with Vt defined on the probability space (Ω,F ,P) with the filtrationFt . Fix any time t, and define aFt-measurable stopping time as τ = inf [ s≥ t : Vs ≥V ∗ ] . Then, the following holds: D(t,τ)≡ E[e−r(τ−t)|Ft ] = ( V ∗ Vt )β with β > 1 Proof: See chapter 9 in Dixit and Pindyck (1994). Proof Proposition 2. This proposition has two claims. The link between acquisi- tion timing and deal quality (claim 1) is a corollary of Proposition 1. In particular, recall that κ1 > κ2 by assumption. Thus, for a fixed level of θ = α+b1+b2, it can be verified that θi(bi) = θ +κibi satisfies the following ordering: θ1(ε)> θ2(ε)> θ1(0) = θ2(0)> θ2(−ε)> θ1(−ε) 29 where I replaced the outcome of B̃i for convenience. The first claim then follows from the fact that under complete information, heterogeneity in deal quality is de- fined as κibi and the fact that V ∗(θi) is monotone and continuous in θi. To prove the second claim, start by fixing an optimal acquisition trigger V ∗ with corresponding optimal stopping time τ , where I suppressed the subindex i for con- venience. First, note that market rationality implies that the expected discounted value of g(·) must be a martingale with respect toFt . Thus, at time t0 the bidder’s stock price impounds the expected net present value of the acquisition, even though there has been no takeover announcement yet. Next, define a pre-announcement event as τδ = inf{t ≥ t0 : Vt ≥V ∗−δ}, for a small δ ≥ 0. By the initial conditions assumed, I have that tδ > t0. As τ and τδ are similarly defined as optimal stoping times, it holds that τδ ≤ τ (a.s.). Predictability of the takeover announcement then follows from the fact that the sample path of Vt is (everywhere) continuous. Thus, τδ → τ as δ tends to 0. In other words, the takeover announcement is predictable by the increasing sequence of pre-acquisition events τδ indexed by δ . Using Lemma 1 in the definition of G(·), the unexpected portion of the gain at time τ is given as: ARτ(δ )≡ [ 1− ( 1− δ V ∗ )β] (θV ∗−A) where ARτ(δ )→ 0 (almost surely) as δ tends to 0. Thus, announcement returns are zero as claimed. Proof Proposition 3. The proof goes in sequence for a pair of signal profiles. I first show that waiting for bidder j’s private information is never a best response for bidder i if its type is bi = ε and its interim optimal acquisition time has arrived. Establishing this result allows me to start partitioning the bidders’ signal space. Conditional on both bidders still being active, a second round with bi = 0 allows me to keep partitioning the signal space. This sequential process generates the set of beliefs for both players along the equilibrium path. Denote by θ̂ bii,t ≡ E[θi|bi, It ] the estimate of the synergy gain for a bidder i with 30 type bi and public information It . At the beginning of the game, the estimate for an ε-type bidder is θ̂ εi,t = α +(1+κi)ε . Given κ1 > κ2, I have that θ̂ ε1,t > θ̂ ε 2,t for all It . Then, if both bidders receive a high signal, bidder 1 will have to ‘decide’ first whether to bid with partial information or to wait for bidder 2 to bid (and learn the additional piece of information). If bidder 1 decides to bid with partial information at time τ , then his optimal bidding trigger would be V ∗(θ̂ ε1,τ). In addition, if a bidder 1 with type ε is expected to bid without bidder 2’s information at τ , then the event V̂t ≥ V ∗(θ̂ ε1,τ) together with a1τ =Wait (i.e., d1τ = 0) reveal to remaining players that bidder 1 is not of type ε . In other words, this event allows bidder 2 and the market to refine their information about bidder 1’s type, by partitioning bidder 1’s signal space. Given that bidder 1 will have to ‘decide’ first if b1 = ε , I first show that in this case he would never optimally bid after bidder 2. Suppose that b1 = ε and that bidder 1’s strategy involves always waiting for bidder 2 to reveal his private information. Given bidder 1’s strategy, bidder 2’s optimal trigger point would be V ∗(θ̂2,t), with θ̂2,t = α +(1+ κ2)b2. By simple substitution, it can be seen that for all b2 ∈ B2, the estimates of the synergy gains satisfy θ̂1,t > θ̂2,t . Given the symmetry and monotonicity of the strategy, it follows that V ∗(θ̂1,t)<V ∗(θ̂2,t), for all b2 ∈ B2. Thus, given a high signal, waiting by bidder 1 is never a best response to bidder 2’s equilibrium strategy. Given the dominance argument established above, when a bidder 2 with type b2 = ε has to decide, he knows whether B1 ∈ {ε} or B1 ∈ {0,−ε}. Thus, if bidder 1 did not bid, then bidder 2’s estimate of the synergy parameter is θ̂2,t = α+(1+ κ2)ε+b1, with B1 ∈ {0,−ε}. Again, by direct substitution, it can be verified that θ̂2,t > θ̂1,t , which implies that V ∗(θ̂2,t)<V ∗(θ̂1,t). Thus, an ε-type bidder 2 would never optimally wait if it is the first to decide. Continuing in the same fashion, it can be verified that, provided that both bid- ders are still active, the bidder that gets to decide first (the one with the highest interim quality, κibi) will be the leader in the merger wave. At each step, the type space of at least one bidder is partitioned. The set of beliefs consistent with this process is given by: 31 Beliefs about Bidder 1: P(b1 = ε|It) =  1 3 if V̂t <V ∗(θ̂ ε1,t) 0 if V̂t ≥V ∗(θ̂ ε1,t) and d1t = 0 1 if V̂t ≥V ∗(θ̂ ε1,t) and d1t = 1 P(b1 = 0|It) =  1 3 if V̂t <V ∗(θ̂ ε1,t) 1 2 if V̂t ≥V ∗(θ̂ ε1,t) and d1t = 0 0 if V̂t ≥V ∗(θ̂ ε1,t) and d1t = 1 or if V̂t ≥V ∗(θ̂01,t) and d1t = 0 and d2t = 1 or if V̂t ≥V ∗(θ̂01,t) and d1t = 0 and d2t = 0 1 if V̂t ≥V ∗(θ̂01,t) and d1t = 1 and d2t = 1 or if V̂t ≥V ∗(θ̂01,t) and d1t = 1 and d2t = 0 P(b2 =−ε|It) = 1−P(b2 = 0|It)−P(b2 = ε|It) Beliefs about Bidder 2: P(b2 = ε|It) =  1 3 if V̂t <V ∗(θ̂ ε1,t) or if V̂t <V ∗(θ̂ ε2,t) and d 1 t = 0 0 if V̂t ≥V ∗(θ̂ ε1,t) and d1t = 1 and d2t = 0 or if V̂t ≥V ∗(θ̂ ε2,t) and d1t = 0 and d2t = 0 1 if V̂t ≥V ∗(θ̂ ε1,t) and d1t = 1 and d2t = 1 or if V̂t ≥V ∗(θ̂ ε2,t) and d1t = 0 and d1t = 1 P(b2 = 0|It) =  1 3 if V̂t <V ∗(θ̂ ε1,t) or if V̂t <V ∗(θ̂ ε2,t) and d 1 t = 0 1 2 if V̂t ≥V ∗(θ̂ ε1,t) and d1t = 1 and d2t = 0 or if V̂t ≥V ∗(θ̂ ε2,t) and d1t = 0 and d2t = 0 0 if V̂t ≥V ∗(θ̂ ε1,t) and d1t = 1 and d2t = 1 or if V̂t ≥V ∗(θ̂ ε2,t) and d1t = 0 and d2t = 1 or if V̂t ≥V ∗(θ̂02,t) and d1t = 1 and d2t = 0 or if V̂t ≥V ∗(θ̂02,t) and d1t = 0 and d2t = 0 1 if V̂t ≥V ∗(θ̂02,t) and d1t = 1 and d2t = 1 or if V̂t ≥V ∗(θ̂02,t) and d1t = 0 and d2t = 1 P(b2 =−ε|It) = 1−P(b2 = 0|It)−P(b2 = ε|It) It is just a matter of substitution to show that given this set of beliefs, the strategy mapping is sequentially rational. Thus, it holds that given beliefs and strategy by bidder j, bidder i is playing a best response timing. 32 B. Figures and Tables Figure 2.1: Time Line of Events . t0 t̃L t̃F 21 33 Figure 2.2: Timing Outcomes. The figure shows the different signal profiles and the respective timing outcomes. Each path ends with the corresponding realized signal profile (b1,b2). The state space for Vt moves from left to right. The numbers in parentheses correspond to the vector dt = (d1t d 2 t ). In the figure, t1 is the time at which a bidder 1 with type b1 = ε bids. t2 follows the same definition for bidder 2. Provided that only bidder i has already bid, t3 defines the optional bid time for a bidder j with type b j = 0. Conditional on both bidders still waiting, t4 defines the optimal bidding time for a bidder with type bi = 0. These reference times should be understood as realized times, conditional on a sample path of the value of the target assets. ◦ • • • • • • • • • • (ε,ε) (ε,0) (ε,−ε) (0,ε) (−ε,ε) (0,0) (0,0) (−ε,0) (0,−ε) (−ε,−ε) (1+κ1)ε ( 12 +κ2)ε ε − ε2(1 0) (1 1) ( 1 0 ) (0 1) (1 1) ( 1 0 ) (1 1) ( 0 1 ) (1 1) (1 1) (0 1) ( 0 1 ) ( 1 0 ) ( 0 0 ) t1 t2 t3 t4 34 Figure 2.3: Timing Distortions. The figure shows the contrast between the timing out- come under complete and heterogeneous information in small and capital letters respec- tively for four signal profiles. l f L-F L F fl l f FL L F fl (ε,ε) (ε,−ε) (0,ε) (−ε,ε) Bidder 1 is leader Bidder 2 is leader Vt Table 2.1: Timing of Announcements Profile Complete Information Incomplete Information Leader (b1,b2) θ1−α θ2−α θ̂1−α θ̂2−α Timing (ε,ε) (2+κ1)εX (2+κ2)ε (1+κ1)εX (2+κ2)ε Delay (0,ε) ε (1+κ2)εX ε ( 12 +κ2)εX Delay (ε,0) (1+κ1)εX ε (1+κ1)εX ε Same (−ε,ε) −κ1ε κ2εX −κ1ε ( 12 +κ2)εX Rush (ε,−ε) κ1εX −κ2ε (1+κ1)εX −κ2ε Rush (0,0) 0X 0X − ε2 X − ε2 X Delay (−ε,0) −(1+κ1)ε −εX −(1+κ1)ε − ε2 X Rush (0,−ε) −εX −(1+κ2)ε − ε2 X −(1+κ2)ε Rush (−ε,−ε) −(2+κ1)ε −(2+κ2)εX −(2+κ1)ε −(2+κ2)εX Same 35 Chapter 3 Anticipation and Timing in Merger Waves: Evidence 3.1 Introduction There seems to be a consensus in the academic and practitioner community that bidders perform poorly during corporate acquisitions. Most of this consensus is sustained by evidence about stock price reaction to deal announcements18. In the previous chapter I proposed a theoretical model and showed that market anticipa- tion poses a serious challenge for the interpretation of bidders’ stock price reac- tion as a measure of bidders’ actual economic gain. The model shows that, in the context of merger waves, the dynamics of market learning (anticipation) varies in predictable ways over time. The model posits that early deals in a merger wave are of higher quality and have higher announcement effects because they are less anticipated. This is the timing effect, which corresponds to the time-series predic- tions in the model. The model also predicts that news of the early deals provides information about firms that will eventually announce an acquisition. I refer to this pattern of cross-sectional predictions as the spillover (or contagion) effects. In this chapter I offer an empirical test of these two basic predictions. I use a sample of 1,425 deals from four industries that underwent a change in merger regulation in the 1990s: Commercial Banking, Telecommunications, Electricity 18Bruner (2004) argues that this conventional wisdom is poorly grounded in scientific evidence. For example, Moeller, Schlingermann and Stulz (2005) find that a few large deals made by serial bidders in the period 1998–2001 explain in part the bidders’ disappointing stock price reactions observed in large sample studies. 36 and Insurance. Although regulated industries are typically excluded in mergers research (Becher, Mulherin and Walking, 2012), these industries provide a good laboratory for my tests as deregulation provides an observable event for industry- wide shocks that potentially induce a wave of merger activity. After controlling for a set of characteristics that have been previously found to be associated with the cross section of bidders’ cumulative abnormal returns (CARs), I find evidence of a significant timing effect (i.e., CARs decline along the merger wave). Economically, my estimates imply a 1.6% difference in CARs between the first and the last deal in a merger wave. This magnitude is about five times the average CAR for bidders in the sample. Although this finding is con- sistent with evidence in McNamara et al. (2008) and Goel and Thakor (2010), I present additional evidence indicating that this outcome follows mainly from in- complete information about deal announcement timing. In particular, I rely on the theoretical model in the previous chapter to specify an empirical timing model that includes a proxy for deal quality. My results show that the predictable portion of timing does not have any explanatory power in the cross section of CARs. I show that the unexpected timing (timing surprise) drives the declining pattern of CARs along the merger wave. I then focus on the prediction that the announcement of early deals serves as a signal to the market about the announcement timing of follower deals (i.e., conta- gion effects)19. I begin by looking at whether or not portfolios of rival firms react at the time of early deal announcements. I find evidence of contagion effects in both Commercial Banking and Electricity. Although this finding supports one of the model’s prediction, it is also consistent with other explanations. For exam- ple, rivals might react if they are likely targets in future deals (Song and Walk- ing, 2000), or if early announcements provide a signal about industry conditions (Eckbo, 1983). Additional tests confirm that, after controlling for product market interactions (competitive and collusion effects) and for interactions in the market 19As discussed in the previous chapter, the literature on contagion effects is substantial (albeit unrelated to merger waves). The scope of this research goes from spill-over effects through product market interactions (Eckbo, 1983; Shahrur, 2005) to spill-over effects due to information effects (Singal, 1996; Akdogu, 2009). The literature of deal anticipation goes back to at least Shipper and Thompson (1983) and Asquith, Bruner and Mullis (1983) in the context of corporate acquisition programs. 37 for corporate control (role of rivals in subsequent merger activity), the contagion effect is still present and is presumably driven by information about the deal timing of future deals. An additional implication of the model is that the market learns by observing bidders’ inaction. I test this prediction by looking at the stock return comove- ment among bidders that delay their acquisition announcements and their industry peers. The comovement prediction says that firms in the portfolio of bidder fol- lowers should start comoving less with other assets in the economy and more with rival firms once the merger wave starts. I find evidence of comovement effects, even for industries in which my previous tests did not detect contagion effects. This finding suggests that, consistent with the model, the learning process by the market includes inaction events that the researcher cannot observe. This evidence highlights a limitation regarding how much a researcher can learn from a set of ob- servable events. For example, Cai et al. (2011) find that, after a dormant period (12 months) with no deals in an industry, initial bidders (presumably less anticipated) have significantly positive abnormal returns. Their findings indicate that, once this anticipation effect is added to the own announcement effect of subsequent bidders, some well-accepted results are reversed or attenuated. Although these results are in line with the main predictions in the model, my findings also suggest that capturing all value-relevant anticipation effects from the set of events the researcher observes might prove challenging. This chapter makes three contributions to the empirical literature on mergers and acquisitions. First, I show that the declining pattern in bidder announcement returns in a merger wave likely follows from incomplete information in the mar- ket about the deal announcement timing, as opposed to deal quality. Second, I verify that contagion effects have more than one source, and show that there is a contagion effect driven by information about deal announcement timing of future deals. Finally, I provide evidence consistent with the hypothesis that the market learns not only from actual announcements, but also from non-announcement or inaction events by bidders. Overall, my results on the link between deal timing and anticipation raise challenges for the interpretation of stock price reactions during acquisition announcements, especially for late deals. 38 3.2 Data and Methods The empirical design in this chapter follows closely the model in the previous chap- ter. The predictions in the model involve cross-section and time-series aspects of stock returns during merger wave episodes. As such, all predictions are conditional on the arrival of a merger wave state for an industry (i.e., the arrival of acquisition opportunities and signals for bidders in the model). Consequently, the first step is to identify the start and end of merger waves. I begin this section with the sample selection criteria and the approach used to identify merger waves. I then continue with the results regarding the time-series and cross-sectional predictions for bid- ders’ cumulative abnormal returns (CARs) in a merger wave. 3.2.1 Merger Waves The selection process starts with all completed deals available in SDC’s Mergers and Acquisitions Database that were announced between January 1, 1981 and De- cember 31, 2010. I require the deal to be classified in SDC as Merger, Acquisition of Majority Interest, Acquisition of Assets or Acquisition of Certain Assets20. I restrict attention to deals between a US bidder and a US target, and deal value of at least $10 million (in 2009 dollars). Following the rationale in the model, I focus on control transactions. I keep in the sample deals with a toehold lower than 50% and ownership after the transaction of more than 50%. After imposing these selection criteria, the sample comprises of 35,275 deals. To identify merger waves, I follow a two-step approach. First, I relate the ar- rival of a merger wave to an observed industry event that is likely to reshape the boundary of the firms in that industry. In the context of merger waves, Andrade, Mitchell and Stafford (2001) argue that “ ... deregulation is the ideal candidate for analysis. First, it creates new investment opportunities for the industry. Second, it potentially removes long-standing barriers to merging and consolidating, which might have kept the industry artificially dispersed. Finally, it is fairly well-defined in time and in terms of parties affected, so empirically we know where and when to 20Studies typically exclude assets sales. Based on the model I have no reason to exclude asset sales ex-ante. Furthermore, Netter et al. (2011) argue that SDC classifies as asset sales all transactions in which the consideration sought is not given. This condition induces a potential misclassification bias. 39 look.” I exploit these conditions in the empirical design and focus on four indus- tries that underwent merger deregulation in the US during the 1990s: Commercial Banking, Telecommunications, Electricity and Insurance. Focusing on deregulation provides a laboratory that more closely mimics the setup in the theoretical model developed in the previous chapter21. In particular, the four industries in my sample underwent regulatory changes that lifted an own- ership constraint on a relevant business dimension. Provided that this ownership constraint was binding before the approval of the regulatory change, it is likely that the removal of this constraint will induce consolidation or reorganization in the industry22. Appendix A provides a brief description of the respective regulation changes in the selected industries. Deals with both an acquirer and a target firm in these industries form a sample of 6,005 transactions. After I identify the arrival of a shock, I follow a data-driven procedure to iden- tify the timing of the start and end of a merger wave for each of these industries23. In each month, I compute a twelve-month forward moving average of the value of deals in the industry: MAVj,t = 1 12 t+11 ∑ s=t Vj,s where Vj,s is the sum of deal values for acquisitions in industry j during month s. I use twelve months to filter out intra-year seasonal patterns in merger activity (see Netter et al., 2011). I use a forward measure (instead of a centered one) to detect the start of a rising activity period at an early stage. I then compute a measure of ‘abnormal’ activity level for month t as follows: A j,t = MAVj,t − M̂AV j,t 21Although a data-driven procedure to identify merger waves is feasible (e.g., Harford (2005) or Rhodes-Kropf and Wiswanathan (2005)), existing procedures are silent on the nature of the resulting merger wave. It implies that it is up to the researcher to ex-post relate the spikes in merger activity to some potential trigger. 22I do not need the deregulation event to be exogenous with respect to merger decisions for my tests. I only use deregulation as an outcome that allows me to identify the arrival of shocks. 23The procedure is adapted from Helwege and Liang (2004). Similar procedures have been used in the IPO literature to identify hot periods of IPO activity (Chemmanur and He, 2011) or hot debt markets (Doukas et al., 2011) 40 where M̂AV j,t is the forecast from a linear regression of MAVj,t on time. This regression is run with all deals in the corresponding industry in the period 1981– 2010. Month t is part of a merger wave if both of the following conditions are met: a) A j,t belongs to the top 50% of de-trended activity for industry j, and b) A j,t+s, s = 1, ...,6, belong to the top 30% of de-trended activity for industry j. To the month that fails to satisfy either of these conditions, I add 12 months and define the end of a merger wave. This procedure is meant to capture months in which industries transition towards or out of high activity states in a persistent fashion. Figure 3.1 shows the evolution of deal value and the number of deals for the four industries in the sample. Shaded areas indicate the time span the procedure identifies as the merger wave that is closest to the date of the regulatory change in the respective industry. Overall, the procedure seems to correctly capture periods of high activity. After identifying the start and the end of a merger wave for each of the selected industries, I only keep in the sample those deals that are announced during the merger wave. This leaves 2,061 deals in the sample. Stock price availability in CRSP for bidder firms further reduces the sample size to 1,425 deals. The split of number of transactions by industry is as follows: 1,021 deals in Commercial Banking, 214 deals in Insurance, 735 deals in Telecommunications and 91 deals in Electricity. The empirical analysis is carried out on this sample. 3.2.2 Sample CARs Panel A of Table 3.1 provides descriptive statistics for CARs in the run-up [-41,-3] and announcement [-2,2] periods for acquirers and targets. CARs are computed with the market model using an estimation window of 200 trading days ending at day t =−42, where t = 0 is the day of the announcement24. The combined returns are computed as the value weighted CAR of bidders and targets, using as weights the market capitalization of each firm at day t =−42. The results in the sample are consistent with prior evidence showing that targets experience significantly large run-up and announcement returns. The average combined announcement CAR is 0.77% (t=3.39), with 25th and 75th percentiles of -2.04% and 2.88%, respectively. 24Results are qualitatively unchanged if I do not use the market model and define abnormal returns as ARit = Rit −Rmt , where Rit is the daily return of the event firm and Rmt is the return on the CRSP value-weighted index. 41 These magnitudes are in line with previous studies. For example, Eckbo, Betton and Thorburn (2008) report combined announcement CARs of 1.06% (t=14.61) for a sample of 4,803 deals. Bidders’ mean and median announcement CAR are small and statistically in- significant. This result is again consistent with prior evidence25. Although the 25th and 75th percentiles of bidder announcement CAR are about±3%, Figure 3.2 shows that it is not uncommon to have CARs outside the range of ±10%. Eckbo et al. (2008) find that the two key drivers of bidder announcement CARs are the target’s status and bidder size. While large bidders acquiring public targets repre- sent the worst case for bidders (-2.21%), small bidders acquiring private targets in stock perform the best (6.46%) in their large sample. In Panel B of Table 3.1 I show that, consistent with a prediction from the model, deal timing generates dispersion in bidder announcement CARs. I define a timing measure as Ti = (τi− t0)/(T − t0), where t0 and T are the start and end date of the merger wave and τi is the announcement date of deal i. The table shows descriptive statistics by timing quintile26. Although these results are unconditional in the sense that known cross-sectional determinants have not been controlled for, the table suggests a monotonic relationship between timing and bidder announcement CARs in the sample. I further explore this prediction in the next section. 3.3 Timing Effects This section focuses on the time-series predictions in the model. These predictions relate mostly to the evolution of acquirers’ CARs along a merger wave. In par- ticular, the model predicts that early deals in a merger wave are of better quality and therefore generate a larger value gain for bidders. Importantly, if the deal gain and timing are fully predictable by the market, the stock price reaction should be zero at the announcement. In the model, however, CARs decline along the merger wave because the timing of early deals is not fully anticipated by the market due to 25Eckbo et al. (2008) provide large sample evidence and summarize several previous studies in their Table 6. 26As an alternative approach, I also defined timing Ti with the sequence of deals in a wave. I first index deals from 1 to n by announcement date within a merger wave and then divided each deal’s index by the median index. Results in this chapter remain qualitatively unchanged with this approach. 42 incomplete information about the synergy gain. 3.3.1 Total Timing Effect First, I show that CARs decline along the merger wave. I capture this total timing effect through the coefficient α in the following regression: CARi j = αTi j +β ′1Bi+β ′ 2Di+FE j + εi j (3.1) where Ti j ∈ [0,1] is the measure of timing for deal i in industry j, and Bi and Di are vectors of bidder and deal/target characteristics. The inclusion of these variables is important as deal announcement timing might be correlated with bidder and deal characteristics that have been previously found to be associated with bidder CARs. For example, Eckbo et al. (2008) find that bidder size and target status are key drivers of bidder announcement CARs. Fuller et al. (2002) find that serial or repeat bidder CARs decline with each subsequent acquisition announcement. With this in mind, I add as control variables bidder size, target status and the number of acquisitions the bidder (and the target) performed in the previous three years. I include industry fixed effects to control for unobserved time-invariant indus- try heterogeneity. For example, in the model different signal profiles may bring merger wave episodes with different patterns of timing. This will impact the real- ized timing. Furthermore, growth opportunities or the competitive effect of merg- ers may differ across industries. As an alternative to using industry fixed effects, I also use the number of firms in the industry at the beginning of the year. This is a crude proxy for the effect of mergers, entry and exit in the industry dynamics. A complete description of all variables used can be found in the Appendix. The corresponding descriptive statistics are provided in Table 3.2. All variables have been winsorized at the top/bottom 1% by industry. Table 3.3 provides different specifications for regression 3.1. In all specifica- tions, the timing measure is statistically and economically significant. The estimate implies that the first bidder in a merger wave experiences an announcement effect that is 1.6% higher than that of the last bidder. This magnitude is over five times larger than the mean CAR and about a third of the interquartile range reported in Table 3.1. A result that stands out in Table 3.3 is that the coefficient on timing Ti j 43 is remarkably stable across specifications. This result suggests that the effect of timing on bidder CARs is fairly orthogonal to other cross-sectional characteristics previously reported as important for explaining the cross section of bidder CARs27 Several characteristics in Table 3.3 are not statistically significant. For exam- ple, after controlling for the stock price run-up, Tobin’s Q is typically used as a proxy for growth opportunities. The indicator variables for method of payment and acquisition of assets are also statistically insignificant. One possible explana- tion for these results follows from the inclusion of industry fixed effects. Given the relatively short time span I study, industry fixed effects remove differences in the level of growth opportunities across industries. Furthermore, the type of acquisi- tion and the method of pay may follow an industry pattern in the sample. For exam- ple, deals in the banking industry are less likely to be asset sales, but more likely to be paid for in stock. Thus, even though these controls still serve the required purpose here, the interpretation might be compromised by poor identification due to lack of sufficient within industry variation. Finally, whether the bidder is a serial bidder does not seem to be systematically related to bidder CARs. Consistent with evidence first reported by Phalippou, Xu and Zhao (2012), bidders acquiring firms with multiple acquisitions in the last three years perform worse than the typical bidder. The timing effect I report here is consistent with two prior studies which at- tribute the timing effect to a decreasing pattern in deal quality. In a sample of 3,194 deals spanning twelve 4-digit SIC industries, McNamara et al. (2008) find that early bidders experience a positive and significantly higher stock price reac- tion than followers. They attribute this finding to an early-mover advantage in that acting early allows a bidder to choose from a pool of better target firms. In a sam- ple of 5,417 deals, Goel and Thakor (2010) also report that early bidders in a high valuation wave tend to have larger CARs than followers. They provide a model that predicts this outcome as a consequence of envy among managers. When man- agers care about their relative compensation and pay is tied to firm size, a small set of potentially value-increasing deals can trigger a wave of value-destroying deals motivated by managerial motives. 27For completeness, I also run regression 3.1 without the industry fixed effects. In this pooled regression, the coefficient on timing is -1.755 (t=-2.64). 44 In both papers, bidders’ economic gains are empirically measured with CARs. One of the lessons from the model developed in the previous chapter is that with complete information about the synergy gain, the market is able to predict the timing of deal announcements (Proposition 2). Under this condition, there is no stock price reaction to deal announcements, regardless of what pattern deal quality follows along the merger wave. In other words, the declining pattern in CARs I find here, and those reported by McNamara et al. (2008) and Goel and Thakor (2010), must arise due to incomplete information. In my model, the observed timing effect follows from a declining pattern in information dispersion along the merger wave. In the next section, I elaborate on this idea to show that unexpected deal timing drives the declining pattern in bidders’ CARs within merger waves. 3.3.2 The Timing Surprise A measure of timing surprise requires measuring the market’s expectations (be- liefs). Given that market beliefs are not observable to the researcher, I use the theoretical model I developed in the previous chapter to specify an empirical tim- ing forecast. I define the timing surprise as Si = Ti− T̂i. The basic premise of this test is that, as T̂i is based on public information at the time of the announcement, it is predictable. Therefore, T̂i should not be systematically related to CARs and Si should drive most of the timing effect. The empirical timing model is specified as follows: Ti j = β ′1ROi+β ′ 2Di+FE j + εi j (3.2) where Ti j is the realized (actual) measure of timing for deal i, ROi is a vector of proxies for the optimal exercise timing, and Di is a vector of deal characteristics. In the real option model in the previous chapter, the optimal acquisition trigger28 increases in the level of volatility and drift of the underlying asset (target pre- shock value). Furthermore, deals with high quality and low irreversibility should 28The optimal acquisition threshold is presented in Propositions 1 and 3 in the previous chapter. It is given by: V ∗(θ̂i,t) = ββ−1 A θ̂i,t . It is not difficult to show that β is a decreasing function of both µ and σ . Thus, timing Ti should increase in µ and σ as both increase the level of the acquisition trigger. A reflects the level of sunk costs (irreversibility) and θ̂i,t represents the expected synergy gain, measured with respect to bidder i’s information set. 45 be announced early, ceteris paribus. These four characteristics form the vector ROi. Due to data limitations with target firms, which are mostly private firms, I use bidder firm data to measure these four characteristics. Although it is arguably a crude approximation, bidders and targets in the sample belong to the same 3- digit SIC classification. I discuss at the end of this section some robustness tests on the main results in this section. I use bidders’ growth in assets in the year before the deal announcement to proxy for growth in target’s pre-shock assets (µ). I proxy for targets’ asset volatility with a measure of idiosyncratic volatility of bidders’ stock returns. The degree of reversibility is measured with the exit value of bidders’ assets. I follow Berger, Ofek and Swary (1996) to define the exit value with the index: (Cash+0.72Receivables+0.55Inventories+0.54PPE), scaled by total assets29. A measure of deal quality, which is based on managerial information (mostly private), is more challenging to obtain. Healy, Palepu and Ruback (1992) relate accounting returns in the pre and post acquisition periods in a cross-sectional re- gression to measure acquisition-driven changes in performance. I follow a slightly different approach, which captures the same rationale. In particular, I proxy for deal quality with the change in bidders’ operating income between year t− 1 and year t + 2, where year t is the announcement year. I scale this change in income by total assets in year t − 1 and denote this measure as dROA. In the regression equation 3.2, I control for the level of ROA at year t−1. Thus, controlling for the initial level of ROA, a higher deal quality should induce a higher dROA. Although this measure of realized deal quality is potentially a noisy proxy for expected deal quality, it captures the idea that, on average, bidders’ performance should improve after successful acquisitions. This measure of deal quality relies on the assumption that managers have unbiased expectations about deal quality. Thus, the average realized deal quality should capture interim expected deal quality30. 29Two of the four industries I study in this chapter are the financial sector. This measure of reversibility might be problematic for these two industries. I provide later some robustness tests on this issue. 30The assumption of managerial unbiasedness could be relaxed, for example, to include distortions in the levels of beliefs about synergy gains (e.g., managerial optimism). As long as all managers in the cross section over estimate synergies by the same amount, the correlation between realized and 46 Finally, the regression includes industry fixed effects to control for cross indus- try unobserved heterogeneity. In the model, a source of cross industry differences in timing comes from different signal profiles for different industries. Note that by including industry fixed effects, the identification of coefficients in regression 3.2 comes from within industry variation. Table 3.4 shows the results from estimating equation 3.2. The empirical model seems to capture a fair amount of variation in timing across deals: R2s are about 80%. Given that I want to use this empirical model to forecast timing, a fairly high R2 is good news. Real option variables are all significant at conventional levels. The signs are in line with the real option interpretation for all variables except for asset growth. A plausible explanation is that acquisitions allow low growth bidders to acquire high growth targets. The results also suggest that timing increases in deal size and decreases in relative size (deal size/bidder size). It is important to note that the coefficient on the proxy for deal quality indicates that higher quality deals are announced earlier within an industry wave. This is consistent with the predictions in the model, but it is also consistent with the story in McNamara et al. (2008) and Goel and Thakor (2010). I use the above regression as an auxiliary model to measure the expected tim- ing T̂i and the corresponding timing surprise Si = Ti− T̂i. I re-estimate regression 3.1 using this decomposition31. In the model, the predictable portion of timing should not have any explanatory power for the cross section of CARs. In Table 3.5, I re-estimate models 5 and 6 from Table 3.3 with the timing decomposition. The coefficients in all control variables are qualitatively similar in magnitude and significance to those reported in Table 3.3. It should be highlighted that the timing forecast includes observable factors that drive timing, in particular a proxy for deal quality. In addition, industry fixed effects control for time-invariant unobserved factors (i.e., the signal profile for the merger wave episode). As hypothesized, the timing effect is driven by the unexpected portion of tim- ing. This result is consistent with the idea that the stock price reaction to an an- expected synergy gains should not change. 31The use of auxiliary models to measure unobservable, though estimable, variables (e.g., expec- tations or unexpected components) was popular in the late 1970s. Murphy and Topel (1985) provide early references and discuss the effects of this approach on the resulting estimation and inference. 47 nouncement corresponds to the unanticipated component of the total economic ef- fect of the event. Such a finding is consistent with the predictions in my model, in that early deals are less anticipated by the market than late deals. A potential concern that could compromise the validity of this conclusion is that I use an auxiliary model to measure expected timing. Firstly, as discussed by Pagan (1984) and Murphy and Topel (1985), the use of generated regressors assumes that there is no sampling error in the estimation of the auxiliary model. In order to account for this effect on the inference in Table 3.4, I constructed a 95% confidence interval for T̂i and Si through bootstrap32. The 2.5% and 97.5% percentiles for T̂i and Si, respectively, in the bootstrapped sample are [-6.355,2.693] and [-2.893,-0.204]. This results is consistent with the results reported in Table 3.5. Secondly, the model of timing in regression 3.2 uses proxies for the determi- nants of the optimal acquisition time. All these proxies could be questioned on different grounds. For example, the measure of reversibility (Exit Value) could be problematic for financial firms. As a robustness check, I verified that the qualitative results in Table 3.5 remain the same as I remove Asset Growth and Exit Value (one at a time) from regression 3.2. This is indeed the case. Although the coefficients have minor changes, T̂i and Si remain statistically insignificant and significant re- spectively. In summary, my results in this section establish that a time-varying degree of anticipation along the merger wave (and not a declining pattern in deal quality) drives the declining pattern in CARs along the merger wave. The observation that partial deal anticipation can severely distort the estimation of bidders’ gains is not new in the empirical literature (see for example Schipper and Thompson, 1983a; Becher, 2009; Cai et al., 2011). Nonetheless, my theoretical and empirical results are, to my knowledge, the first to systematically tie anticipation and deal timing in the context of merger waves. 32I obtained 1000 bootstrap samples by resampling (with replacement) from the original sample. In my resampling I preserved the same number of observations by year and industry. 48 3.4 Contagion Effects This section focuses on the cross-sectional predictions of the model. In the model, announcements of early deals serve as signals for the market about the announce- ment timing of followers. This spill-over effect can potentially be captured as jumps (i.e., CARs) in the stock returns of bidder followers during early (leader) deal announcements. Further, the theoretical model suggests that non-announcement windows might also be potentially value-relevant for the market. 3.4.1 Portfolio Contagion Effects I begin by testing contagion effects from early (leader) deals to late (follower) deals. For this purpose, all deals in a merger wave are sorted into five groups based on the sequence of deal announcements. Group P1 contains the earliest 20% of the deals, and P5 contains the latest 20% of the deals. I designate the first two groups (40% of deals) as the leaders portfolio, and the remaining three groups as the followers portfolio. Serial bidders (i.e., those with more than two acquisitions in a merger wave) are removed from the followers portfolio, but they are included in the leaders portfolio. In addition to measuring the contagion effects on follower bidders in the merger wave, I measure the contagion effects on future targets and on a set of firms that I define as Pure Rivals. In order to identify the set of Pure Rivals, I start with all securities in CRSP traded on the NYSE, Nasdaq or Amex as common shares (i.e., share codes 10 or 11). The securities must also belong to one of the 3-digit SIC classifictions that identify the four industries under study. Pure Rivals are those firms in this set that were not involved in acquisitions (as bidder or targets) in the period starting one year before and ending one year after the merger wave identified for the corresponding industry. Based on the above design, early bidders’ rivals are grouped in roles as: Bidder Follower, Target Follower and Pure Rival. Each of these groups forms a portfolio. If the market is not able to predict rivals’ roles in the future, any contagion effect should be observed uniformly in each portfolio of rivals. After forming these portfolios, I identify the announcement date of all deals in the leaders portfolio. The three portfolios of rival firms should react to the 49 announcement of leader deals if these announcements are informative about the rivals’ value. I estimate the following regression for each rival portfolio p: Rpt = αp+βpRmt +δ+p D + t +δ − p D − t +δ ∗ p D ∗ t + εpt (3.3) where Rpt is the daily return of an equally-weighted portfolio of rival firms. The regression is estimated with daily returns from 12 months before the start of the merger wave until 2 days after the last announcement in the leaders portfolio. Rmt corresponds to the daily stock return of the value-weighted index available in CRSP. Although the model does not feature negative stock price reactions, these are empirically observed. Thus, I measure the contagion effect by including three indicator variables D+t , D − t and D∗t where the superscript identifies whether the stock price reaction to the leader’s announcement was positive, negative or miss- ing (no data in CRSP), respectively. The indicator variable equals 1 during the 5-day window centered at leader announcement dates with the respective sign and 0 otherwise. I allow these three indicators to overlap if two or more deals with differing stock price reactions are observed in the same days33. Table 3.6 shows the estimates of δ+p , δ−p and δ ∗p . The results for the Commer- cial Banking and the Electricity industries are consistent with informational effects. Contagion through competitive effects induces a negative correlation between the leaders’ stock price reaction and that of the rivals. The leader becoming stronger (weaker) is bad (good) news for rivals if competitive effect drive the contagion effects. Table 3.6 does not show such an effect. Although the results in Table 3.6 are consistent with the spill-over effect in the model, they are also consistent with two alternative explanations. First, Song and Walkling (2000) argue that rivals’ reactions in horizontal deals follow from the increased probability of rivals’ being a future acquisition target. In Table 3.6, for the two industries with clear and significant contagion effects, the difference between the contagion on Target and Bidder Followers is small. If the market has some predictive power regarding the role Followers play in future deals, this find- ing does not seem to support the ‘acquisition probability hypothesis’ of Song and 33Allowing this overlap weakens the test as the identification comes mainly from those days in which only one indicator equals 1, but the overlap avoids wrong attribution of effects. I also run the regressions with one dummy at a time and the results are materially the same. 50 Walking (2000). Second, a positive reaction for rivals is also consistent with the ‘collusion hypothesis’ suggested by Eckbo (1983). Once again, for the two indus- tries with clear and significant contagion effects in Table 3.6, Target and Bidder Followers do not seem to have a significantly different reaction from that of Pure Rivals. Furthermore, the collusion hypothesis suggests that rivals should always experience a positive reaction to mergers, independent of the stock price reaction to the leader announcement34. Overall, if the market is able to anticipate rivals’ roles in future deals, the evidence in Table 3.7 provides at best weak support for the two alternative explanations for the contagion effects35. The lack of results for the Insurance and Telecommunication industries in Table 3.6 could be due to at least three potential causes. First, there may be no contagion effect for these industries. Second, the approach used lacks power to uncover an effect in theses industries. For example, there could be (negative) competitive ef- fects on rivals that cancel out any (positive) contagion effects. The model suggests a third reason. The test is designed to detect information effects on rivals when leaders announce a deal. However, the model also suggests contagion effects dur- ing non-announcement events. In the next section, I address the last two concerns with more specific tests. 3.4.2 Individual Contagion Effects The results in the previous section are broadly consistent with contagion effects for two of the four industries under study. The challenge for these results in the context of the model is two-fold. First, contagion effects might also come from news unre- lated to deal value (contrary to what is assumed in the model). For example, lead- ers’ announcement might also signal information about industry conditions. Sec- ond, the model predicts that announcement events and non-announcement events by eventual bidder leaders are informative about the followers’ synergy gains. Thus, it implies that by focusing on early (leader) announcements exclusively, em- 34Under the collusion hypothesis, merger announcements are good news for rivals because mergers reduce the number of (symmetric) firms in the industry. In an oligopolistic setting (with Cournot competition), a merger thus increases the market power for all remaining firms. This effect is known as the free-rider effect. 35I also performed the test in Table 3.6 with leaders defined as the first 20% of deals, and followers as all remaining firms. The unreported results are qualitatively similar to those in Table 3.6. 51 pirical tests miss the informativeness of inaction by leader bidders. To alleviate both concerns, I provide two additional tests in this section. Re- garding the first concern, I design a test that measures the contagion effect after controlling for potential interactions in the product and takeover markets. I argue that this conditional test captures the contagion effect about deal value. I address the second concern by considering second moments in stock returns for Bidder Followers. In particular, I look at the pattern of comovement (covariance) of bid- der followers with industry peers and a market-wide index. The model implies that the transition from the no-wave state to the wave state will bring more comovement between bidder followers and industry peers and less comovement with the rest of the assets in the economy. The first test relies on a pair-wise regression between leader bidder CARs and rivals’ CARs. A test at the individual firm level allows me to control for hetero- geneity across rival firms. The specification is as follows: CARi jk = α0+α1CARk +α2Dik +α3Dk +α4Driv+α5Dbid +FE j + εi jk (3.4) where CARi jk is the CAR of rival i in industry j when leading bidder k announces a deal. Driv = 1 if rival i is a Pure Rival and Dbid = 1 if rival i is a Bidder Follower. Results are measured against the group of Target Rivals (the missing category). These indicator variables control for any differential effect that the role of rivals in subsequent deals might play in the contagion effects. Dk = 1 if bidder leader k announces a deal in a date in which multiple leaders made an announcement. This indicator variable controls for the information effect coming from leader an- nouncements other than that of bidder leader k. In the specification above, Dik = 1 if rival i and leader k are ‘close’ in the product market spectrum in the year before the announcement. I use the product market clusters for each firm (Text-based Network Industry Classifications, TNIC) developed by Hoberg and Phillips (2010)36. The TNIC data are available for the period 1996 - 2010. Whenever data on prior years is needed, I use the peer firms in 1996 as representative of those years. I also include interaction terms by allowing 36The data and more details are available at: http://www.rhsmith.umd.edu/industrydata/index.html. I thank both authors for making this data available. 52 α1 to vary with Dik and Dk. Finally, since CARk is a noisy measure, I create 10 deciles (by industry) based on leader CARk. In the regression 3.4 above, I replace the actual CARk by the median CAR value of the respective decile37. The information effect related to deal quality should be captured by the coef- ficient of CARk in regression 3.4 because the regression is controlling for all con- founding effects discussed previously. For example, Dik controls for any potential competitive or collusion effects. Note that the identification of the coefficient on Dik follows from variation in CARi jk across rivals that are close or far in the product market spectrum. Along the same lines, Driv and Dbid control for interactions in the market for control. For example, Pure Rivals are firms that are, for some reason, excluded from merger activity in the industry ex-post. Thus, differential spill-over effects on these firms speak to signaling effects about industry conditions in gen- eral. Table 3.7 presents the results of this test. Consistent with the results in the previous section, only Commercial Banking and Electricity exhibit a significant contagion effect. As expected, the effect is positive when the information effect comes from deal quality. This result lends support to the idea that early deals act as signals for the market about the quality of subsequent deals in a merger wave. As the specification controls for competitive effects, I can discard the conjecture that the lack of results in the Insurance and Telecommunication industries arises due to competitive effects cancelling information effects. Although the interpretation of the control variables is not my main focus, the results in the Banking industry provide an interesting result. First, the residual con- tagion effect, which I argue corresponds to deal quality, is economically small (5 basis points per 1% increase in leader CAR). Second, this industry features a fairly strong competitive effect as identified by the coefficient on Dik. Finally, the coeffi- cient on Driv is positive and strongly significant, which suggests that participating in the market for control is associated with lower contagion effects (for bidders and targets). The last two findings imply that the decision to participate in the mar- ket for control might be related to the strength of competitive effects coming from product market interactions. 37Results are unchanged if I use the average or the rank of deciles. 53 The second test to address the aforementioned concerns relies on second mo- ments of stock returns to capture contagion effects. The model in Section 2 also predicts contagion during non-announcement events, which results in stock return co-movement prior to announcements. To test this prediction, I follow the ap- proach proposed by Barberis, Shleifer and Wurgler (2005). In particular, I run the following regression for each firm i in the portfolio of Bidder Followers: Rit = β i0+β i 1Rpt +β i 2Rpt−1+β i 3Rmt +β i 4Rmt−1+δ i pDtRpt +δ i mDtRmt + εit (3.5) where Rit is the daily return of a Bidder Follower, Rmt is the daily return of the value-weighted index in CRSP and Rpt refers to the daily return for the portfolio of Followers (bidders or targets) or Pure Rivals. When estimating the coefficient for Bidder (Follower) i, I compute Rpt excluding bidder i. The estimation is performed with daily returns that start 24 months before the first leader announcement and end at the last announcement of the leader group. Dt is an indicator variable that equals 1 every trading day between the first and last announcement of the leader group and 0 otherwise. I include lags to control for microstructure frictions and for the effect of ‘slow diffusion of common information’ (e.g., big firms leading small firms). The comovement prediction says that firms in the portfolio of Bidder Follower should start comoving more with rival firms and comoving less with other assets in the economy. In other words, δ ip should be positive and δ im should be negative. The prediction regarding a lower comovement with the market is a necessary condition to support the contagion effect. Whether or not the test uncovers more comovement with the three different portfolios of rival firms depends on the firms’ role in these portfolios. For example, Pure Rivals capture only competitive effects and there could be less comovement between Bidder Followers and this portfolio. Bidders have potentially both deal and industry information spill-overs. Table 3.8 reports the proportions of δ ip and δ im that are positive and negative, 54 and the average effect across Bidder Followers by industry: δp = 1/n n ∑ i=1 δ ip δm = 1/n n ∑ i=1 δ im With a couple of exceptions, the results in Table 3.8 show that Bidder Follow- ers start comoving significantly less with the market-wide index when the merger wave starts. This results holds even for the Insurance and Telecommunications in- dustries. Such a decoupling from the rest of the economy suggests that an industry- specific factor drives more of the innovations in stock returns during the initial stage in a merger wave. The decrease in comovement is particularly strong when the portfolio of rivals is that of Target Followers, a result that suggests that at least part of these cross-correlated innovations are related to deal quality information. The results on the increase in co-movement with rivals are less clear, but still present in most industries. For example, Bidder Followers in the Insurance and Electricity industries significantly increase the co-movement with the portfolio of Target Followers and Pure Rivals. Uniformly, the change in co-movement with rivals is insignificant when I use the portfolio of Bidder Followers in the regression. Such a result is consistent with the idea that competitive effects play a role in co- movement as well. Overall, the results of the above tests are consistent with the presence of conta- gion effects at the industry level. The fact that the results are more clear for target firms indicates that deal information is relevant in this spill-over process. Finally, the fact that here I detect information contagion for the Insurance and Telecommu- nication industries–two industries for which I found no contagion in previous tests– suggests that the learning process by the market includes not only announcement events, but also non-announcement events (i.e., bidders’ inaction is also informa- tive for the market). Such an effect is largely important as non-announcement events are not observable by the researcher. 55 3.5 Final Remarks and Conclusion In the previous chapter I argued that, in the context of merger waves, the potential for market anticipation has implications for what a researcher can learn from ac- quirers’ stock price reactions. The claim that anticipation induces an attenuation bias in stock price reactions is not new. What is new in the model is that I combine this idea with another well established empirical fact: M&A activity follows waves by industry over time. In this chapter I take the qualitative predictions in the previous chapter’s model to perform a test. Three sets of findings are in line with the model’s predictions. First, I show that the timing effect holds after controlling for deal quality and timing determinants. I find that only the unexpected portion of timing drives the timing ef- fect. Second, I find that rivals’ stock price reacts to early bidders’ announcements. The correlation between rival and early bidder announcement is positive, even af- ter controlling for interactions in the product and the control markets. Finally, the results indicate that rivals’ stock returns start to comove more among themselves and less with other stocks (an index) during the early stage of a merger wave. I argue that this contagion effect reflects learning by the market about the timing and value of future deals when early deals are announced. My tests also indicate that inaction by early bidders may also serve an informational purpose. By confirming the predictions in the model, the evidence in this chapter sup- ports the argument that acquirers’ poor performance in M&As can be partially rationalized by a measurement problem. The failure to properly capture market anticipation can vastly alter the interpretation of existing evidence. The natural caveat of this conclusion is that I use a specific context as a testing laboratory. Further research providing out-of-sample tests and improved methods to measure bidders’ economic gains would enhance our understanding of what drives corpo- rate acquisition decisions. 56 Variable Description Bidder Characteristics CAR[-2,2] Cumulative abnormal return over the window [−2,2] for bidder firms as measured by the market model. I use the value- weighted index in CRSP and an estimation window of 200 days ending at day -42, relative to the announcement date. Run-up (one year) Buy-and-hold abnormal return computed with 12 months of daily returns ending a month before the deal announcement. Abnormal returns are computed as the daily stock return minus the value-weighted index in CRSP. Tobin’s Q Market value of total assets over book value of total assets for the bidder firm: (at− ceq+ csho∗ prcc f )/at. Acquirer Size (Asize) Log of bidder’s total assets from Compustat measured the year before the deal announcement. # Acqs in last 3 years (AcqB) Number of acquisitions performed by the bidder in the last three years (excludes current deal). Relative Size (RelSize) SDC deal value over market capitalization of the bidder at day -42. Deal/Target Characteristics Timing The timing measure for deal i is defined as: Ti = (τi− t0)/(T − t0), where t0 and T are the start and end date of the merger wave and τi is the announcement date of deal i. Deal Value (Dvalue) SDC deal value. In regressions, I use the log of deal value. # Acqs in last 3 years (AcqT) Number of acquisitions performed by the target in the last three years. Public Target Indicator variable that equals 1 when the target is a public firm, and 0 otherwise. Stock Deal Indicator variable that equals 1 when the method of payment is at least 95% stock, and 0 otherwise. Cash Deal Indicator variable that equals 1 when the method of payment is at least 95% cash, and 0 otherwise. Acquisition of Assets Indicator variable that equals 1 when the SDC deal form is Acq. of Assets or Acq. Cert. Asts., and 0 otherwise. Real Option Proxies (Bidder) Asset Growth Growth in assets is computed with Compustat data as: at(t−1)/at(t−2)−1, where t is the announcement year. Volatility Standard deviation of abnormal returns computed with 12 months of daily returns ending one month before the deal announce- ment. Abnormal returns are computed as the stock return minus the value-weighted index in CRSP. Exit Value Exit value is computed as: (che+ 0.72rect + 0.55invt + 0.54ppent)/at, where all variables are from Compustat for the fiscal year before the announcement. I follow Berger, Ofek and Swary (1996) to define the exit value with this index. ROA Return on assets is computed as: oibd p/at from Compustat for the fiscal year before the announcement. dROA Change in ROA is computed as: (oibd p(t+2)−oibd p(t−1))/at(t−1). Others NFirms This is the number of public firms in the industry at the end of the previous year. I use all firms in the intersection of CRSP and Compustat. I assign the industry based on the historical SIC code in Compustat. NFirms corresponds to the number of securities traded in NYSE, NASDAQ and AMEX with share code 10 or 11. Dik Indicator variable that equals 1 when bidder leader k belongs to the cluster of product market competitors of rival i as defined by Hoberg and Phillips (2010), and 0 otherwise. Dk Indicator variable that equals 1 when bidder leader k announces in a day with multiple merger announcements, and 0 otherwise. Driv Indicator variable that equals 1 when the rival of a leader bidder k is a Pure Rival (neither Follower Bidders nor Follower Target), and 0 otherwise. Dbid Indicator variable that equals 1 when the rival of a leader bidder k is a Follower Bidder, and 0 otherwise. 57 A. Regulatory Changes in the Sample A key condition for an industry to be selected in this study is for the industry to experience a regulatory change that lifts an ownership constraint on a relevant business dimension. Provided that this ownership constraint is binding before the approval of the regulatory change, it is likely that the lifting of this constraint will induce consolidation or reorganization in the industry. Next, I provide a brief de- scription of the regulatory changes in the four selected industries. 1. Commercial Banking: The Riegle-Neal Act of 1994 permitted interstate ac- quisitions in the banking industry. Before this regulatory change, states were able to block cross-state ownership of banks through the application of the Douglas Amendment to the 1956 Bank Holding Company Act, which pro- hibited bank holdings from owning banks outside their headquarter state (see Straham (2003) and Becher (2009) for more details). 2. Insurance and Trade: The Gramm-Leach-Bliley Act of 1999 allowed com- mercial banks, investment banks, security firms and insurance companies to consolidate (i.e., enable one-stop shopping). Under the Glass-Steagall Act of 1933, this holding structure was not allowed. Provided that the consumer values a one-stop experience, the deregulation across the product market spectrum should induce consolidation between (commercial and investment) banks and insurance companies. 3. Telecommunication: The Telecommunication Act of 1996 allowed for media cross-ownership, overturning ownership restrictions coming from the Fed- eral Communications Commission (FCC) under the Communications Act of 1934. In particular, the underlying rationale for cross media ownership restrictions was the belief that the public interest would be better served be- cause different owners of media in a region should bring a more diverse spectrum of viewpoints to the community. Akdogu (2009) also uses this industry as a laboratory to uncover the source of rivals’ reactions. 4. Electricity: The Federal Energy Regulatory Commission (FERC) issued Or- der 888 and 889 in 1996. With the idea of promoting competition in the US 58 electricity market, the Orders mandated the unbundling of wholesale gener- ation, transmission and distribution of electricity. In principle, the Energy Policy Act (EPA) of 1992 allowed fair access to the electric transmission system for new independently owned generation firms, thus lowering entry barriers. However, after frequent complaints from independent generators about unfair treatment, the spirit of the EPAct gained popularity with the introduction of the Order 888 (see Becher-Blease et al. (2008) for more de- tails). It should be highlighted that the electric sector had always been reg- ulated. As such, little was known about the natural structure of the industry in a non-regulated environment (Douglas, et al., 2009). I operationalize the notion of industry by using 3-digit SIC codes. The four industries are characterized as follows: 1. Commercial Banking: (602) Commercial Banks, (603) Saving Institutions, (606) Credit Unions, (671) Holding offices. 2. Insurance and Trade: (672) Investment Offices, (673) Trusts, (679) Miscella- neous Investing, (631) Life Insurance, (632) Accident and Health Insurance, (633) Fire, Marine, Casualty Insurance, (635) Surety Insurance, (636) Title Insurance, (641) Insurance Agents, Brokers and Service. 3. Telecommunications: (481) Telephone Communications, (482) Telegraph And Other Message Communications, (483) Radio And Television Broad- casting Stations, (484) Cable And Other Pay Television Services, (489) Com- munications Services, Not Elsewhere Classified. 4. Electricity: (491) Electric Services. 59 B. Figures and Tables Figure 3.1: Merger Waves. Shaded areas indicate the period identified as the merger wave closest to the regulatory change. The vertical (black) line indicates the time of the regulatory change. The solid (green) line shows a centered 12-month moving average of the deal value for a given industry. The segmented (red) line shows the corresponding count for the number of deals (multiplied by 1,000). M & A  A ct iv ity  (D ea l V al ue ) 0 10000 20000 30000 8 1 8 2 8 3 8 4 8 5 8 6 8 7 8 8 8 9 9 0 9 1 9 2 9 3 9 4 9 5 9 6 9 7 9 8 9 9 0 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 1 0 1 1 Commercial Banking Merger Wave M & A  A ct iv ity  (D ea l V al ue ) 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 8 1 8 2 8 3 8 4 8 5 8 6 8 7 8 8 8 9 9 0 9 1 9 2 9 3 9 4 9 5 9 6 9 7 9 8 9 9 0 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 1 0 1 1 Insurance Merger Wave M & A  A ct iv ity  (D ea l V al ue ) 0 10000 20000 30000 8 1 8 2 8 3 8 4 8 5 8 6 8 7 8 8 8 9 9 0 9 1 9 2 9 3 9 4 9 5 9 6 9 7 9 8 9 9 0 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 1 0 1 1 Telecommunication Merger Wave M & A  A ct iv ity  (D ea l V al ue ) 0 1000 2000 3000 4000 8 1 8 2 8 3 8 4 8 5 8 6 8 7 8 8 8 9 9 0 9 1 9 2 9 3 9 4 9 5 9 6 9 7 9 8 9 9 0 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 1 0 1 1 Electric Sector Merger Wave 60 Figure 3.2: Bidder Cumulative Abnormal Returns. This figure shows the histogram of bidders’ 5-day announcement returns by industry. CARs are computed with CRSP return data and the market model. I use 200 trading days ending at day t = −42, where t = 0 is the day of the announcement. 61 Table 3.1: Cumulative Abnormal Returns. Panel A shows descriptive statistics of cu- mulative abnormal returns (CARs). CARs are computed with the market model and 200 trading days ending at day t = −42, where t = 0 is the day of the announcement. The combined abnormal return is computed using the CARs for bidder and target, and their corresponding market capitalizations at day -42. The t-stat corresponds to the null hypoth- esis that the average CAR is zero (standard errors are computed with the cross section of CARs). Panel B shows descriptive statistics of bidders’ stock price reaction by announce- ment timing. Timing for bidder i is computed as Ti = τi−t0 T−t0 , where t0 and T are the start and end date of the merger wave and τi is the announcement date of deal i. Panel A: Sample CARs Acquirer Target Combined CAR[-51,-3] CAR[-2,2] CAR[-51,-3] CAR[-2,2] CAR[-51,-3] CAR[-2,2] Mean -0.37 0.31 3.71 14.44 0.29 0.77 t-stat -0.94 1.50 5.31 21.37 0.61 3.39 25th Pctl -7.25 -3.13 -4.88 4.10 -5.14 -2.04 Median -0.48 -0.23 2.46 12.07 0.36 0.25 75th pctl 6.71 2.88 11.08 22.20 5.86 2.88 N 1435 1435 448 448 448 448 Panel B: Bidder CAR[-2,2] by Announcement Timing Ti Mean t-stat 25th pctl Median 75th pctl N Early 1.10 3.61 -1.95 0.31 3.59 286 2 0.37 1.11 -2.04 -0.07 2.40 288 3 0.59 1.47 -2.88 -0.16 3.18 286 4 -0.47 -1.01 -4.45 -1.29 2.39 289 Late -0.52 -1.21 -4.16 -0.56 3.01 286 62 Table 3.2: Descriptive Statistics. Runup TobinQ BSize DValue RelSize AGrowth Sigma Exit ROA dROA AcqB AcqT Cash Stock Assets Public Mean 0.119 1.419 8.015 4.601 0.333 0.716 0.352 0.439 4.464 4.997 2.436 0.170 0.151 0.502 0.320 0.452 Std Dev 0.311 0.848 1.781 1.584 0.610 2.560 0.174 0.170 6.696 16.605 3.125 0.738 0.358 0.500 0.467 0.498 25th -0.046 1.050 6.616 3.401 0.042 0.078 0.224 0.352 2.496 1.288 0 0 0 0 0 0 Median 0.103 1.122 7.962 4.245 0.136 0.182 0.313 0.502 3.007 2.454 1 0 0 1 0 0 75th 0.265 1.406 9.413 5.407 0.358 0.435 0.418 0.554 5.762 5.477 4 0 0 1 1 1 N 1378 1101 1122 1435 1435 1113 1378 1060 1081 1056 1435 1435 1435 1435 1435 1435 Runup 1 TobinQ 0.29 1 BSize -0.12 -0.27 1 DValue 0.03 0.08 0.43 1 RelSize -0.04 -0.02 -0.24 0.36 1 AGrowth 0.09 0.18 -0.20 -0.06 0.01 1 Sigma 0.24 0.38 -0.53 -0.13 0.19 0.21 1 Exit -0.08 -0.14 0.14 -0.13 -0.12 -0.08 -0.26 1 ROA -0.02 0.12 0.11 0.20 0.02 -0.18 -0.16 -0.29 1 dROA 0.11 0.01 -0.02 0.15 0.06 0.00 -0.05 -0.25 0.44 1 AcqB -0.01 -0.05 0.40 0.04 -0.17 -0.03 -0.17 -0.01 -0.03 0.03 1 AcqT -0.01 -0.01 0.23 0.38 0.11 -0.04 -0.09 0.02 0.03 0.07 0.11 1 Cash -0.02 0.09 -0.07 -0.09 -0.05 0.00 0.05 -0.22 0.09 -0.02 -0.05 -0.08 1 Stock -0.05 -0.27 0.19 -0.06 -0.10 -0.14 -0.28 0.43 -0.19 -0.10 0.08 0.09 -0.42 1 Assets 0.10 0.32 -0.22 -0.11 -0.01 0.16 0.37 -0.48 0.16 0.02 0.02 -0.14 0.31 -0.61 1 Public -0.10 -0.22 0.31 0.31 0.04 -0.13 -0.33 0.27 -0.05 0.00 -0.02 0.24 -0.18 0.38 -0.61 1 63 Table 3.3: Total Timing Effect. This table shows regressions of bidders’ announcement returns on the timing measure and bidder and deal characteristics. Timing for bidder i is computed as (τi− t0)/(T − t0), where t0 and T are the start and end date of the merger wave and τi is the announcement date of deal i. Remaining control variables are described in the Appendix. The standard errors are clustered by Industry× Year. T-stats are reported below the estimates. Dependent Variable: CAR[-2,2] 1 2 3 4 5 6 Timing (Ti) -1.577 -2.335 -1.622 -1.269 -1.931 -1.571 -3.24 -3.76 -2.52 -2.14 -2.84 -2.79 Bidder Characteristics Runup (one year) -1.365 -1.502 -1.419 -1.88 -2.09 -1.95 Tobin’s Q 0.086 0.164 0.240 0.10 0.26 0.34 Asize -0.530 -0.283 -0.265 -4.46 -1.33 -1.19 # Acqs in last 3 years (AcqB) 0.079 0.037 0.054 1.58 0.68 0.95 Relative Size (RelSize) 0.242 0.639 0.702 0.39 0.90 0.98 Deal/Target Characteristics Deal Value (Dvalue) -0.333 -0.044 -0.091 -1.83 -0.14 -0.29 # Acqs in last 3 years (AcqT) -0.737 -0.782 -0.758 -3.62 -3.56 -3.73 Public Target -1.170 -1.262 -1.260 -3.32 -2.93 -2.89 Stock Deal -0.335 -0.163 0.000 -0.72 -0.36 0.00 Cash Deal 0.015 -0.472 -0.528 0.02 -0.75 -0.83 Acquisition of Assets -0.900 0.119 0.208 -1.37 0.22 0.37 Other Number of Firms in Industry -0.004 -0.002 -7.72 -1.21 Intercept 3.991 4.955 8.09 2.57 Industry FE YES NO YES YES NO YES N.Obs. 1435 1435 1086 1435 1086 1086 N.Clusters 24 24 24 24 24 24 Adj. R-Sq 4.5% 3.8% 5.9% 7.3% 7.7% 8.2% 64 Table 3.4: Timing Determinants. This table shows regressions of Timing on the proxies for real options exercise and deal characteristics. All variables are described in the Ap- pendix. The standard errors are clustered by Industry × Year. T-stats are reported below the estimates. Dependent Variable: Timing (Ti) 1 2 3 Real Option Trigger Proxies (Bidder) Asset Growth -0.012 -0.009 -2.41 -1.86 Volatility 0.741 0.908 4.37 4.90 Exit Value -0.225 -0.255 -2.59 -3.26 ROA 0.005 0.005 2.32 2.29 dROA -0.001 -0.001 -1.810 -2.350 Deal Characteristics Relative Size (RelSize) -0.035 -0.073 -2.12 -4.20 Deal Value (Dvalue) 0.031 0.048 3.50 5.27 Stock Deal 0.059 0.037 1.74 1.09 Acquisition of Assets 0.072 0.038 3.25 1.58 Industry FE YES YES YES N.Obs. 1003 1003 1003 N.Clusters 24 24 24 Adj. R-Sq 80.1% 77.4% 81.6% 65 Table 3.5: Unexpected Timing Effect. This table shows regressions of bidders’ an- nouncement returns on bidder’s and deal’s characteristics as defined in the Appendix. Columns 1 and 2 re-estimate columns 6 and 5 in Table 3.3 by replacing the timing measure by the Predicted Timing, and the Timing Surprise. The Predicted Timing is the fitted value from the corresponding model in Table 3.4. Timing Surprise is computed as the actual Timing minus the Predicted Timing. The standard errors are clustered by Industry × Year. T-stats are reported below the estimates. Dependent Variable: CAR[-2,2] 1 2 Predicted Timing (T̂i) -1.669 -2.292 -0.57 -0.63 Timing Surprise (Si) -1.542 -1.833 -2.62 -2.32 Bidder Characteristics Runup (one year) -1.440 -1.471 -2.08 -2.16 Tobin’s Q 0.264 0.168 0.36 0.25 Asize -0.381 -0.390 -1.34 -1.51 # Acqs in last 3 years (AcqB) 0.059 0.032 1.07 0.61 Relative Size (RelSize) 0.572 0.427 0.84 0.59 Deal/Target Characteristics Deal Value (Dvalue) -0.056 0.427 -0.14 0.59 # Acqs in last 3 years (AcqT) -0.763 -0.762 -3.78 -3.54 Public Target -1.187 -1.270 -2.66 -2.78 Stock Deal -0.342 -0.300 -0.64 -0.47 Cash Deal -0.509 -0.317 -0.70 -0.47 Acquisition of Assets -0.154 -0.182 -0.27 -0.32 Other Number of Firms in Industry -0.002 -1.12 Intercept 5.842 3.15 Industry FE YES NO N.Obs. 998 998 N.Clusters 24 24 Adj. R-Sq 8.8% 7.7% 66 Table 3.6: Portfolio Contagion Effect. This table shows the Followers’ and Pure Rivals’ stock price reactions in ‘leader’ event time. Results are shown conditional on the nature of the stock price reaction of leader k. For each industry, I first divide all deals in a merger wave in five groups: P1,...,P5. P1 has the earliest 20% of deals and P5 has the latest 20% of deals in the merger wave. Deals in P1 and P2 are designated as leader deals. Deals in P3, P4 and P5 comprise the set of followers. Pure Rivals are firms that are neither bidders nor targets in the merger wave. In order to measure the contagion effects, I create three indicator variables that equal 1 in the 5-day window surrounding the announcement of each leader bidder (k) depending on the type of the leader CAR (positive, negative or missing), and 0 otherwise. The number of Leader Bidders of each type is reported in N =. I use the market model to estimate the average daily stock price reaction: Rpt = αp +βpRmt + δ+p D+t + δ−p D − t + δ ∗p D∗t + εpt . Portfolio p refers to the portfolio of Followers (Bidders or Targets) or Pure Rivals. This regression is run starting one year before the start of a merger wave and ending a day after the last leader’s announcement. The standard errors are the Newey-West HAC standard errors (Bartlett Kernel and 5 lags). T-stats are reported below the estimates. Estimates (δ s) are in percentage points. Commercial Banking Insurance Telecommunications Electricity CARk > 0 CARk < 0 Missing CARk > 0 CARk < 0 Missing CARk > 0 CARk < 0 Missing CARk > 0 CARk < 0 Missing Portfolio p N=136 N=210 N=87 N=47 N=19 N=24 N=107 N=71 N=127 N=8 N=8 N=20 Bidder Followers 0.055 -0.082 0.047 0.013 0.069 0.028 -0.019 -0.039 0.100 0.236 -0.072 0.056 2.35 -3.70 2.05 0.43 1.83 0.87 -0.33 -0.58 1.61 1.74 -0.68 0.50 Target Followers 0.064 -0.096 0.040 0.053 0.030 0.032 0.148 -0.142 0.064 0.304 0.027 0.137 2.05 -3.03 1.35 1.46 0.63 0.70 1.78 -1.46 0.73 2.05 0.32 1.43 Pure Rivals 0.023 -0.061 0.021 -0.009 0.001 -0.035 -0.046 -0.132 0.070 0.229 -0.033 0.029 1.01 -2.51 0.88 -0.20 0.03 -0.68 -0.71 -1.86 1.05 2.05 -0.44 0.35 t-stat (Bidder-Rival) 1.58 -0.94 1.29 0.41 1.21 1.20 0.35 1.16 0.39 0.07 -0.47 0.38 t-stat (Target-Rival) 1.72 -1.41 0.84 1.23 0.48 1.31 1.98 -0.09 -0.06 0.61 0.61 0.10 67 Table 3.7: Individual Contagion Effects. This table shows the contagion effects from a test at the individual firm level. CARk is the stock price reaction of a leader bidder (in group P1 or P2) deal announcement. Given that CARk is noisy, I sort CARk by industry in 10 groups. The regression below replaces the actual CARk by the median of the respective decile. CARi jk is the stock price reaction to leader bidder announcement k of rival i in industry j. The regression is CARi jk =α1CARk+α2Dik+α3Dk+α4Driv+α5Dbid+FE j+ ε , where Driv = 1 if the rival is Pure Rival, Dbid = 1 if rival is a Follower Bidder, Dik = 1 if i and k are product market peers in the year prior to the announcement, Dk = 1 if the announcement k is in a date with multiple leader announcements. I also include interaction terms by allowing α1 to vary with Dik and Dk. I run the regression with industry fixed effects when data is pooled. See the appendix for variable definition. The standard errors are clustered by leader k announcement. T-stats are reported below the estimates. Dependent Variable: CARi jk All Banking Insurance Telecom Electricity CARk 0.050 0.111 0.062 0.018 0.403 2.64 3.56 1.27 0.78 5.14 CARk×Dik 0.015 -0.010 -0.031 -0.025 0.099 0.89 -0.53 -0.73 -0.67 0.96 CARk×Dk 0.000 -0.050 -0.110 0.054 -0.485 0.00 -1.24 -1.66 0.95 -3.58 Dik -0.296 -0.335 -0.105 0.310 0.190 -4.90 -5.21 -0.52 1.12 0.41 Dk -0.063 -0.188 1.376 -0.788 2.501 -0.58 -1.51 3.80 -2.15 5.85 Driv 0.236 0.259 0.192 0.057 -0.475 6.67 7.31 1.53 0.26 -1.15 Dbid 0.040 0.008 0.074 0.258 -0.462 1.19 0.24 0.66 1.35 -1.60 Intercept 0.467 -0.285 -0.061 0.793 4.54 -1.11 -0.28 1.92 FE Industry YES NO NO NO NO N.Obs. 169059 136998 17191 14358 512 N.Clusters 347 347 66 183 16 Adj. R-Sq 0.48% 0.41% 0.56% 0.22% 29.67% 68 Table 3.8: Bidder Follower Comovement. This table reports the estimates and the t-stats for a test of co-movement for the set of Bidder Followers in each industry. I first divide all deals in a merger wave in five groups: P1,...,P5. Deals in P1 and P2 are designated as leader deals. Deals in P3, P4 and P5 comprise the set of followers. Pure Rivals are firms that are neither bidders nor targets in the merger wave. I then create an indicator variable Dt that equals 1 every trading day between the first and last announcement of the leader group (P1 and P2). The estimation is done with daily returns that start 24 months before the first leader announcement and end at the last announcement of the leader group (P2). I estimate the regression: Rit = β i0 + β i 1Rpt + β i 2Rpt−1 + β i 3Rmt + β i 4Rmt−1 + δ i pDtRpt + δ imDtRmt + εit . In this regression, portfolio p refers to the portfolio of Followers (Bidders or Targets) or Pure Rivals. When estimating the coefficients for Bidder (Follower) i, I compute Rpt excluding bidder i. The table reports the cross-sectional average by industry and t-stat of δ ip and δ im. Bidder Follower Target Follower Pure Rival δp % > 0 δm % < 0 δp % > 0 δm % < 0 δp % > 0 δm % < 0 N Banking 0.094 60% -0.099 58% 0.141 65% -0.114 64% -0.107 35% -0.037 54% 72 1.69 -2.73 1.96 -2.78 -1.85 -0.88 Insurance -0.003 51% -0.047 47% 0.121 63% -0.097 53% 0.109 65% -0.089 57% 49 -0.05 -1.12 2.74 -2.15 1.99 -1.82 Telecom 0.076 65% -0.244 65% 0.119 55% -0.255 75% -0.074 40% -0.140 75% 20 1.43 -2.19 1.46 -2.22 -1.23 -1.70 Electricity 0.044 67% -0.099 73% 0.199 80% -0.224 87% 0.150 73% -0.281 87% 15 0.84 -2.55 4.32 -5.37 3.40 -5.26 69 Chapter 4 Market Timing and Merger Waves 4.1 Introduction Merger and acquisition (M&A) activity follows waves over time. These so-called ‘merger waves’ tend to peak at times of high stock market valuation38. Previ- ous research argues that this empirical regularity is driven by the action of rational managers who exploit windows of equity overvaluation by acquiring relatively less overvalued firms. Two alternative, albeit unexplored, views provide plausible ex- planations for this market timing behavior of acquirers. First, when the level of adverse selection in financial markets is low, equity prices are high and mergers are potentially less costly to negotiate. Second, M&As are irreversible investments that take time to complete (i.e., a real option with time-to-build). Under this view, stock prices increase as firms approach the optimal acquisition time. Uncertain merger completion time (i.e., an investment friction) induces a preference for eq- uity financing when financial distress is costly. In this chapter, I investigate the relative merits of the overvaluation, the ad- verse selection and the investment frictions hypotheses to explain why merger ac- tivity and stock market valuations are correlated in the time series. Answering this question sheds light on why merger activity follows waves over time. I begin by documenting that times of high merger activity are systematically associated with a preference for equity financing at the industry level. This finding is consistent with 38Golbe and White (1988), and references therein, discuss early evidence on U.S. merger waves at the aggregate level and the correlation with stock market valuation. Mitchell and Mulherin (1996) report that the intensity of takeover activity varies across industries in a non-random fashion. An- drade, Mitchell and Stafford (2001), Harford (2005) and Rhodes-Kropf, Robinson, and Viswanathan (2005) study more systematically the determinants of merger waves. 70 the idea that a ‘market timing’ force is behind the time-series behavior of M&A activity. Following Kim and Weisbach (2008), I next argue that the allocation of pro- ceeds from equity issues is informative about what managers believe about the stock price level at that time. Thus, I study the explanatory power of each hypothe- sis by exploiting differing predictions about the use of proceeds from equity issues at times of high merger activity. Overall, I fail to find support for the predictions implied by the overvaluation hypothesis. My results indicate that the correlation between stock market valuations and merger activity arises from managers’ attempt to minimize the costs of adverse selection and financial distress. In particular, I find evidence consistent with the adverse selection hypothesis in a sub-sample of finan- cially constrained firms. More broadly, the evidence is consistent with predictions implied by the investment frictions hypothesis. At the core of my empirical design is the idea that, if managers are rational agents who maximize existing shareholders’ wealth, investment and financing de- cisions must be optimal with respect to the managers’ information set. Thus, de- spite the unobserved nature of merger wave drivers, managerial decisions at times of high merger activity should reveal managers’ beliefs about equity prices. Con- sistent with this idea, I argue that if booms in merger activity arise from an industry- wide market timing motive, then an increase in merger activity at the industry level should be systematically related to an overweighing of equity in the composition of external finance39. In my analysis I find that, after controlling for the average level of external finance in the industry, the average share of equity in new exter- nal finance is positively and significantly related to merger activity at the industry level. This result is robust to several specifications and holds after controlling for industry conditions. Although this finding provides evidence to support that the time-series behav- ior of both merger activity and equity issues at the industry level are driven by a market-timing motive, this test is not informative about which of the aforemen- tioned hypotheses is more prevalent in the data. To this end, I track for three years the allocation of equity issue proceeds, conditioning on whether the issuance was 39Baker and Wurgler (2000) argue that the equity share in new issues provides information about the underlying cause of equity issues. 71 done in a high or a low M&A activity year. The rationale for this exercise is that the change in the allocation of proceeds of equity issued at times of high versus low merger activity should be informative about the motivation for increasing both equity issues and merger activity at the same time. I focus on five uses of funds in my analysis: Increase in cash holdings, debt payments, capital expenditures, cash acquisitions and equity payouts (dividends and share repurchases). The results for the overall sample indicate that the transition from low to high M&A activity brings a temporal incentive to allocate more of the equity proceeds towards increasing cash holdings. Out of a dollar of proceeds from equity issued in low merger times, about 30 cents are allocated to increasing cash during the same year. This figure jumps to 42 cents when equity is issued in high merger years. Despite its significance, this difference vanishes almost entirely after three years. Capital expenditure also increases by 12 cents in the year of the equity issuance, but then it remains fairly flat. Out of a dollar of equity issued in low (high) M&A times, 27 (44) cents end up in capital expenditure three years after the equity issuance. Debt repayment significantly grows for two consecutive years after an equity issuance in high merger times. By the end of the third year, the increase in allocation towards debt payment is 28 cents out of a dollar of equity proceeds. Finally, equity payout shows a statistically significant increase by the end of the third year. However, the change brought by the transition from low to high merger activity is only 2 cents in the sample. These results are not consistent with the predictions implied by the overvalua- tion hypothesis. According to that hypothesis, firms should use the equity proceeds to return cash to shareholders. However, after three years of issuance, firms have not allocated a substantially larger fraction of the proceeds raised in high merger times towards increasing dividends or repurchasing shares. The observed increase in equity payout is very modest to be considered of first order importance. Further- more, firms do not seem to substitute old debt for new (presumably less expensive) equity issued during high merger times. On the one hand, debt repayment does not seem to start in the year of the equity issuance. On the other hand, the fast depletion of cash savings does not correspond one-to-one to the increase in debt repayment over time. At a conceptual level, both types of substitutions (old equity and old debt for 72 newly issued equity) constitute a pure arbitrage that should be more easily exe- cuted by a financially unconstrained firm (Stein, 1986). However, I also fail to find evidence consistent with these predictions in a sub-sample of less financially constrained firms. Finally, at the time of issuance, firms tilt the equity proceed allocation towards a permanent increase in investment. Although overvaluation motives, together with financial constraints, predict such an increase in investment, I find that the increase in investment is driven by less financially constrained firms (i.e., the opposite). The adverse selection hypothesis predicts that firms time equity issuance so as to lower issuance costs and to avoid future underinvestment. Under this view, ad- ditional financial slack (i.e., cash saving or debt capacity) should have a permanent nature. The fast cash flow depletion I observe in the data seems to be at odds with this prediction. Furthermore, the cash depletion does not correspond to an increase in capital expenditures or cash acquisitions over time. Although these findings lend no support to the view that firms issue equity to save at times of low adverse selection, the picture changes when I look at the sub-sample of firms which are more financially constrained. In this sub-sample, there is no significant increase in investment, but there is a permanent increase in cash holdings. The cash saving from equity proceeds increases from 29 cents in low merger times to 44 cents in high merger times. The increase in cash holdings remains after three years in this sub-sample, which provides evidence consistent with a precautionary motive of eq- uity issuance at times of high merger activity. Thus, although the adverse selection hypothesis is unlikely to provide a general explanation for the correlation between merger activity and stock market valuations, it provides a plausible explanation for firms that face financial constraints. The overall set of findings is consistent with the investment friction hypothesis in several respects. First, the increase in allocation of equity proceeds towards in- vestment occurs mainly at the year of the equity issue. This evidence is stronger for less financially constrained firms which increase by 30 cents the allocation towards investment when moving from low to high merger times. The results suggest that firms with existing needs for capital issue equity. Furthermore, the increase in debt repayment shows an upward pattern over time only when there is an increase in the allocation to investment. This observation is consistent with the idea that initially 73 firms finance investment with equity (to avoid distress costs), and then they start refinancing as cash flows from investment arrive. This refinancing should allow for some equity payout if the firm does not need the extra cash. A small increase in equity payout three years after the issuance is consistent with a partial reversal in the initial equity issuance. These results are magnified when I focus on firms classified as less financially constrained and with short horizon. My measure of investment horizon is based on the investment turnover of institutional investors in the firm. The fact that the results are stronger for the firms with higher turnover suggests that institutional investors might be able to successfully chase investment opportunities. In summary, I do not find evidence to support the claim that acquirers’ timing behavior arises from a ‘window of opportunity’ due to overvalued equity. My evi- dence suggests that periods of high M&A activity are associated with times of low adverse selection, and that more financially constrained firms exploit these times to hoard cash. The results in this chapter are thus related to the literature arguing that time-varying adverse selection is a first-order effect in equity issuance deci- sions (Alti and Sulaeman, 2012; Autore and Kovacs, 2010; and Kim and Wisbach, 2008). The evidence presented here is more generally consistent with the idea that times of high merger activity are associated with the exercise of real options which are optimally financed with equity issues due to time-to-build. Thus, this work also relates to prior research arguing that real investment frictions are necessary for capital structure theories to match the observed leverage dynamics (DeAngelo, DeAngelo and Stulz, 2010; Pierce, Tsyplakov and Wang, 2012; and Tsyplakov, 2008). More specifically, the current research contributes to two related literatures in mergers and acquisitions. First, it contributes to the literature on the drivers of M&A waves (Harford, 2005; Rhodes-Kropf et al., 2005). Second, the chapter contributes to the empirical literature arguing that overvaluation drives individual merger outcomes (Ang and Cheng, 2006; Dong et al., 2006). I claim that the literature has omitted time-varying adverse selection and investment frictions as plausible drivers of individual M&A outcomes and M&A waves. On the methodological front, this chapter is related to several prior studies. In particular, my work in this chapter is related to Dasgupta, Noe and Wang (2011) 74 who study the dynamic allocation of internal cash flows to different uses. I use their method to study the dynamic allocation of proceeds from equity issues. In this vein, this chapter also relates to Kim and Wisbach (2008) who study the use of proceeds from IPOs and SEOs to explore the motivation for equity issues. While they study the allocation of proceeds to learn about the motivation for equity issues, I focus on the change in the proceeds’ allocation between high and low M&A years to understand what drives merger waves. Butler, Cornaggia, Grullon and Weston (2011) exploit the information in the level and composition of financing to study the new issues puzzle (i.e., the inverse relationship between security issuance and future stock returns). I use the same rationale to explore the relevance of market timing in the context of merger waves. Finally, as Alti and Sulaeman (2012) do in the context of equity issues, I argue that prior literature has focused exclusively on ‘valuation-driven variables’ to study market timing behavior. In my empirical analysis I avoid using valuation-driven variables to construct proxies for the unobserved drivers of merger waves. I try to identify what drives merger activity by looking at managerial actions. This choice is critical to isolate managers’ information set in my empirical design. 4.2 Hypotheses Development In this section I outline the three hypotheses I explore in this chapter. I refer to them as the overvaluation, adverse selection and investment frictions hypotheses. The idea that equity overvaluation drives merger activity is typically justified by the empirical observation that peaks in merger activity are systematically re- lated to times of high stock market valuations. This notion follows closely the evidence that equity issuances are performed when stock prices are high. For ex- ample, Baker and Wurgler (2002) use the long-run underperformance of issuers to argue that equity overvaluation explains the market timing behavior of equity issuers. Nonetheless, recent theoretical work and empirical evidence show that the long-term underperformance of equity issuers can be rationalized without equity mispricing (i.e., irrational markets). For example, Carlson, Fisher and Giammarino (2006) argue that standard match- ing on observable variables typically used in long-run event studies fails to capture 75 the dynamics of risk and return during SEO episodes. In a real options context, they show that the riskiness of a firm’s assets decreases as firms convert real options into assets in place. This change in fundamentals induces a problem with the ‘bench- mark’ return for the issuing firm. Carlson, Fisher and Giammarino (2010) provide evidence consistent with the idea that firms’ riskiness changes in predictable ways during SEO episodes. Bessembinder and Zhang (2012) empirically study this im- perfect control-firm matching in the context of IPOs, SEOs, mergers and dividends. I extend this rationale on equity issues to analyze the time-series behavior of merger activity. In particular, I argue that the outcomes usually used to support the equity overvaluation hypothesis are, in many respects, observationally equivalent to those predicted by two alternative hypotheses. These alternative hypotheses exchange the assumption of irrational markets for real investment frictions and financing frictions. The first hypothesis relies on time-varying adverse selection in the equity mar- kets. In the context of equity issues, Lucas and McDonald (1990) provide a dy- namic version of Myers and Majluf (1984) and show that managers will issue eq- uity after a stock price run-up. Korajczyk, Lucas and McDonald (1991, 1992) show that managers, aware that future investments might arrive at times of high adverse selection, will rationally time the equity market by issuing equity at times of low adverse selection. Theoretical work in the context of merger negotiation also suggests that low levels of asymmetric information about the stand-alone value of merging firms may lower the cost of negotiating a deal. Brusco, Lopomo, Robinson and Viswanathan (2007) and Gartner and Schmutzler (2009) use mechanism design to study merger negotiations when the merging parties have private information not only about the synergy gain but also about the stand-alone values. Both papers show that private information about stand-alone values prevents efficient implementation (i.e., merg- ers take place if and only if they increase joint profits). Furthermore, Gartner and Schmutzler (2009) show that private information about stand-alone values makes it very difficult to avoid ex-post regret for merging firms40. Thus, time-varying 40Both papers use a mechanism-design approach, in which agents report their type and outcomes are made contingent on these reports. Brusco et al. (2007) use Bayesian implementation and Gartner and Schmutzler (2009) use ex-post implementation. 76 adverse selection can potentially affect the incentive to participate in merger trans- actions. Overall, the adverse selection hypothesis due to asymmetric information pre- dicts that when the information asymmetry in stock markets is low, equity prices are high (i.e., firms are less undervalued) and mergers are potentially less costly to negotiate. Chemmanur, Paeglis and Simonyan (2009) provide empirical evidence consistent with the notion that the method of payment is affected by asymmetric information on both sides of a merger deal41. The second alternative hypothesis follows from the consequences of real in- vestment frictions for the optimal financing of investments. Tsyplakov (2008) en- riches a standard dynamic trade-off model of capital structure with investment irre- versibility and time-to-build. In the model, the stock price increases (i.e., a run-up) as the firm approaches the optimal investment time. Furthermore, equity financing is preferred because the project is risky and generates no cash flows to service debt until the project is complete. In order words, uncertain completion time and the potential cost of financial distress induce a more conservative leverage ratio (i.e., a preference for equity financing) at the time of investment. Huang, Pierce and Tsyplakov (2012) argue that merger decisions are also char- acterized by irreversibility and uncertain synergy realization times. This would help rationalize the observation that stock deals are prevalent even among acquirers with substantial cash holdings or debt capacity. In their model, bidders’ leverage choice at the time of the announcement is decreasing in the anticipated duration of the post-merger integration period. Leverage will gradually increase as the inte- gration stage moves closer to completion. Their evidence on 200 mergers supports both claims. Overall, the investment friction hypothesis predicts that merger decision times should be characterized by a preference for equity financing. The initial conser- vative leverage ratio is then adjusted upwards as expected synergy gains are ma- terialized over time. Thus, a pattern of refinancing should follow merger decision times. Baker and Wurgler (2002) place the overvaluation and the adverse selection 41Early theoretical work relating the method of payment to asymmetric information includes Fish- man (1989), Hansen (1987) and Eckbo, Giammarino and Heinkel (1990). 77 motive under the roof of market timing42. In a slight abuse of the name, I will also place the real investment frictions hypothesis under the roof of market timing. Overall, the market timing hypothesis can be stated as follows: H0 : The time series of M&A activity is unrelated to market timing motives, HA : The time series of M&A activity is positively related to market timing motives. Each version of the market timing explanation has a different source. Following the previous discussion, I define three alternative hypotheses as follows: HA1 : The time series of M&A activity is positively related to overval- uation in the equity market, HA2 : The time series of M&A activity is negatively related to the level of adverse selection in the market, HA3 : The time series of M&A activity is positively related to the extent and relevance of irreversibility and time-to-build in merger decisions. In the empirical analysis I proceed in two steps. I first focus on the main hypothesis (H0 vs HA). Then, I try to differentiate among the three alternative hypotheses. In the next section, I describe my sample and data, and then proceed to the empirical analysis of the hypotheses stated in this section. 4.3 Sample and Data The selection process for mergers and acquisitions starts with all domestic deals (U.S. target) available in SDC’s Mergers and Acquisitions Database that were an- nounced between January 1, 1983 and December 31, 2010. I require the deal to be complete and classified by SDC as: Merger or Acquisition of Majority Interest. From this general M&A sample, I only keep deals with U.S. public bidders and 42Recent literature has tended to refer to equity market timing as that motivated exclusively by overvaluation. Baker and Wurgler (2002) define more generally equity market timing as “... the practice of issuing shares at high prices ... the intention is to exploit temporary fluctuations in the cost of equity relative to the cost of other forms of capital.”. 78 deal value (in 2009 dollars) of at least $10 million. My focus on deals with public bidders follows from the evidence in Netter, Stegemoller and Wintoki (2012) and Maksimovic, Phillips and Yang (2013). The first paper finds that M&A waves are less pronounced when small deals with a private acquirer are removed from the sample. Their findings show that M&A activity by public bidders follows a more wave-like behavior. Maksimovic et al. (2013) find that public firms are more likely to be bidders, especially during times of high merger activity. In addition, I require a toehold lower than 50% and ownership after the deal greater than 50%. I assign each bidder to one of the Fama-French 12-industry classification based on SDC SIC codes. I delete all deals with bidders in Financials, Utilities and Other (in FF12 industries). The sample comprises 6,119 deals. The construction of accounting variables for my tests starts with the intersec- tion of CRSP and Compustat. A firm is required to be traded on NYSE, Nasdaq or Amex (i.e., exchange code 1, 2 or 3) as a U.S. common stock (i.e., share code 10 or 11). Firms’ assignment to the Fama-French 12-industry classification is done by using the historical SIC code in Compustat when available. If that is not available, I use the SIC header in Compustat. Given that I use data from the Cash Flow State- ment, I also require firms to have the format code (scf ) available in Compustat. I exclude firms with negative sales or total assets below $10 million. To mitigate backfilling bias, I exclude firms with less than three years of data in Compustat. These requirements leave a base sample of 134,419 firm-year observations for a total of 12,203 unique firms for the period 1983–2010. 4.3.1 M&A Descriptive Statistics Table 4.1 displays a summary of the sample of deals by decade, industry and method of payment. Industries are ranked by decade and deal value. Consistent with previous evidence, the industry ranking varies over decades. The less-than- perfect correlation between rankings says something about the origins of merger activity. Note that if some industries were, by nature, more prone to acquisitions, then the ranking should be relatively stable over time. The time-variation in the ranking suggests that some industry phenomenon drives M&A activity over time (Andrade et al., 2001). 79 In addition, the 1990s were characterized by a larger proportion of pure stock deals. This is true for both number and value of deals. On the other hand, cash payment was more prevalent in the 1980s. This observation is consistent with idea that the type of deals in the two decades were different in terms of who knows what when closing a deal, and the role played by different theories of merger waves. The observation that the rankings based on number or value of deals differ is perhaps not surprising given that industries differ in many ways that affect the optimal operation scale. However, the difference points out to a critical decision when studying the time series of merger activity: The use of number of deals or the dollar value of deals. Figure 4.1 shows the time series of four industries in the sample. The plot shows the time series of number of mergers, number of stock mergers and dollar value of mergers. Looking at all industries (including the five unreported plots), merger activity defined by deal value tends to be more volatile and less regular than that defined by number of deals. It seems that a few large deals can easily drive these spikes. 4.3.2 The Time Series of Merger Activity In the subsequent analysis I use the time series of merger intensity by industry. To compute merger intensity, I first count the number of mergers in every quarter- industry combination. I then scale the number of mergers in each quarter-industry by the number of firms in CRSP at the end of the previous quarter in that industry. The number of firms in each industry-quarter corresponds to all U.S. firms in the Compustat-CRSP intersection with non-missing end-of-quarter return, traded on NYSE, Amex or Nasdaq and share codes 10 or 11. Annual intensity is obtained by summing the four quarterly intensities. Figure 4.1 shows the intensity for four FF12 industries. The four industries correspond to the two with the highest and the two with the lowest merger activity level in the sample. Fig 4.2 shows the evolution of the nine de-trended intensities in the nine industries I consider. Even though there is consensus in the literature that M&A activity for public firms comes in waves, there is less consensus on a workable definition of these waves, and thus on how to model the times series of merger activity. This lack of consensus seems to follow from inconsistencies in what defines a deal and a 80 wave, and in data collection procedures. For example, Town (1992) study four US aggregate M&A series that cover different time spans and data collection ap- proaches. The best ARIMA model for each series implies fairly different time series dynamics for, technically, the same series43. He advances that the data col- lection procedures matter for the properties of the time series. This is consistent with evidence reported in Netter et al. (2012). They find that the inclusion of small and private bidders impacts not only the wave-like behavior but also the timing of merger waves. Table 4.2 provides summary statistics of merger intensity by industry. As out- lined by Lowry (2003) in the context of IPOs, scaling by the number of firms in the industry deals with potential non-stationarity and controls for differences in in- dustry size. Even though merger intensity is truncated at zero, there seems to be enough time series variation in the data for all industries. The coefficient of vari- ation goes from 0.41 (Consumer Durables) to 0.72 (Business Equipment). More- over, the maximum merger intensity in each industry is not larger than the sample mean plus three standard deviations. This observation is important as it suggests that a linear process might be a fair description of the time series of deal intensity. In non-tabulated results, I find that the average first order serial correlation in the sample is 0.6, ranging from 0.1 (Chemicals) to 0.8 (Business Equipment). The average second order serial correlation is 0.4, which is often insignificant at con- ventional levels at the industry level. In additional non-tabulated results, I find some degree of co-movement in the industry intensity series. I used principal com- ponents to extract the factor structure from the nine industry time series. The first factor explains about 60% of the variance, and the first three factors explain about 85% of the variance. This evidence suggests that there is a fairly strong common driver of merger intensity across industries. The last two columns in Table 4.2 show the correlation between the time series of intensity defined by number and value of deals. Consistent with the observations in Table 4.1 and Figure 4.1, the correlation is less than perfect. This is particularly true for Consumer Nondurables and Healthcare. Figure 4.3 shows the scatter plot and the histogram for the intensity of merger activity defined by number and value 43Golbe and White (1988) provide an interesting analysis of the inconsistencies in M&A series from different sources in the pre-SDC period. 81 of deals for these industries. The figure confirms that the low correlation comes from outliers (i.e., extremely large deals). Overall, the histogram of intensity based on number of deals seems better behaved. This is one of the reasons I focus on the merger intensity defined by the number of deals. Beyond the data issues just described, the basic motivation for this choice is also that the informativeness each deal has for my test does not depend on the deal’s size. In other words, each deal will receive equal weight in the tests44. 4.4 Market Timing Test In this section I ask whether a market timing motive can help explain the time series behavior of merger intensity. I take advantage of the panel structure of the data to identify this effect. In order to construct a proxy for the market timing motive, I use the patterns of financing decisions at the industry level. The premise of this design is that managers have information (i.e., a signal) about the arrival and the nature of the unobservable timing motive. If managers are rational value maximizers, corporate choices must be optimal with respect to the managers’ information sets. This condition implies internal consistency across managerial decisions45. 4.4.1 Regression Specification The market timing hypothesis is tested with the following regression model: I jt = α j +β1ES jt +β2XFjt +β ′3Z jt +β ′ 4Xt + ε jt (4.1) where I jt is the deal intensity in industry j at year t, ES jt is the industry equity share in external finance, XFjt is the industry level of external finance, Z jt is a vector of industry-level control variables and Xt is a vector of macro-level control 44I am still imposing a $10 million (in 2009 dollar) deal value cutoff for inclusion in the sample. Therefore, the evidence is not completely driven by very small deals. 45Other papers have focused on investor-based measures to capture market timing and the effects on individual merger outcomes. For example, Ben-David, Drake and Roulstone (2011) use the short interest to construct a measure of overvaluation. Their measure is correlated with merger decisions, stock mergers and post merger returns. Khan, Kogan and Serafeim (2012) use price pressure from purchases by mutual funds with large capital inflows to uncover demand shocks (overvalued equity). They find that the probability of equity issues, insider sales and stock deals increases significantly in the four quarters following the price pressure. 82 variables. In order to control for potential specification problems coming from serial correlation in the residuals, I assume an AR(1) structure for the error term and estimate the regression with maximum likelihood. I define the ES jt starting from the measure at the individual firm level. For firm i in industry j at year t, the equity share is defined as: ESi jt = EIssuei jt EIssuei jt +DIssuei jt where EIssue and DIssue are the gross equity and debt issues in year t of firm i in industry j. The equity share at the industry level ES jt is defined as the (industry) average of ESi jt . External finance is defined as: XFi jt = NDebti jt +NEquityi jt TotalAssetsi jt where NDebt and NEquity are net debt and equity respectively, both measured in year t for firm i in industry j. In both measures, net refers to net of payout (i.e, debt retirement, dividends, repurchases). Financing flows for these measures (and the proxies below) come from the Statement of Cash Flows in Compustat. A detailed description is presented in the Appendix at the end of the chapter. The market timing hypothesis of merger waves predicts that firms will tilt their external financing towards equity at times of high merger activity. Note that changes in external finance (XF) likely come from changes in firms’ demand for funds (i.e., growth opportunities). Thus, specification 4.1 implies that, controlling for the level of external finance, ES jt should provide additional information about the timing motive. If the same timing motive drives both merger intensity and financing patterns over time, then β1 in equation 4.1 should be positive. The use of industry-wide financing flows mitigates in part the concern that both merger and financing decisions are driven by firm-specific fundamentals. As an additional safeguard for potential endogeneity, I also use financing flows after excluding firms that merge in a window around the measurement year. As part of my empirical design, I deliberately avoid using stock prices, analysts forecast or institutional trades to proxy for unobservable variables as these need not be consistent with managers’ information set. 83 Although the information in financing flows has not been used in the M&A context to uncover market timing incentives, it has been used in the empirical asset pricing literature. For example, Baker and Wurgler (2000) find that the equity share in new issues is a strong predictor of US aggregate returns. In their tests, the equity share outperforms other known stock return predictors. Butler et al. (2011) use the information in the level and composition of financing to study the new issues puzzle in asset pricing (i.e., the inverse relationship between security issuance and future stock returns). In a corporate finance application of the same notion, Ivashina and Becker (2011) study the substitution between bank and public debt, conditioning on the level of new debt issuance. Like these papers, I claim that conditional on the level of external finance, the composition of financing elicits decision makers’ private information. 4.4.2 Control Variables I control for two sets of variables in equation 4.1. As discussed previously, an influ- ential strand of the literature argues that M&A waves follow from industry shocks. In addition, industry conditions and the business cycle may affect both M&A activ- ity and financing flows for reasons other than the market timing motive. I discuss below the related control variables. A description of all variables is provided in the Appendix. The industry fixed effect in equation 4.1 captures any (potentially unobserved) time-invariant industry heterogeneity such as long-run growth opportunities or fric- tions coming from the nature of assets and technology in the industry. This is important as, for example, industries with more growth options are more likely to experience more acquisitions and are also more likely to issue external finance (and potentially may have a different preference for equity financing). In general, the specification forces identification of β1 with the within industry variation of explanatory variables. This approach is consistent with the idea that merger waves are a time-series phenomenon and that my tests target the time series variation in industry merger intensity. I also include industry-specific time-varying variables that might drive both mergers and financing choices at the industry level, for reasons other than the tim- 84 ing effect of interest. Given that I use the yearly average of accounting variables on the right-hand-side of equation 4.1, changes in the industry composition over time might induce correlation between ES and merger activity. Thus, the vector Z in- cludes a measure of asset tangibility to control for potential changes in the industry composition in the sample. The vector Z also includes three proxies for industry shocks. A more detailed description of each of these proxies is in the Appendix. Rosseau’s (2002) Q-theory approach to merger waves states that exogenous industry shocks bring an increase in dispersion in Tobin’s Q. I use the industry cross-sectional standard deviation of the investment-to-capital ratio (vIK) to con- trol for this effect46. In addition, I follow Harford (2005) who proposes two addi- tional proxies related to the neoclassical hypothesis of merger waves. I extract the first principal component from the median of the absolute year-to-year change in six performance accounting metrics for each firm in industry j: cash flow margin, asset turnover, capital expenditure, employee growth, return on assets and sales growth. This factor is denoted as EcS in my tests below. I also include a regula- tory indicator variable as in Harford (2005). This variable is denoted as DReg and equals 1 in years with a regulatory event in the industry. The vector Xt includes the cyclical component of real Gross Domestic Product (GDP) and a time trend47. The time trend absorbs secular trends in merger activity and/or industry characteristics (e.g., financial development, technology, etc.). The inclusion of GDP follows from existing evidence indicating a pro-cyclical behavior of mergers, investment and financing decisions48. 46I use the dispersion in the IK ratio instead of dispersion in the market-to-book ratio because the latter embeds, in addition to the true Tobin’s Q, any potential mispricing in the stock market. Through its first order condition, the Q-Theory of investment provides, a relation between investment rates and the expected marginal/average value of capital (Tobin’s Q). 47As a robustness, I change these time-varying macro controls for year fixed effects. Year fixed effects might absorb variation of interest, as again merger waves are a time-series phenomenon. Thus, year fixed effects are not part of the main specification. 48Becketti (1986) finds that good capital market conditions (measured by a stock index and T- bond yields) and the business cycle (measured by change in GDP) are related to increases in merger activity. Makaew (2011) studies cross-border deals and finds that these deals also show wave-like behavior, and a high correlation with business cycle. Indeed, these deals tend to occur when both the target’s and the bidder’s country are booming. Erel, Julio, Kim and Weisbach (2012) find that equity and bond issuances are pro-cyclical for non-investment grade issuers. Pro-cyclical behavior in cross-sectional standard deviation of IK is reported in Bachmann and Bayer (2011). 85 4.4.3 Sorts on Financing Variables In this section I begin by looking at the relationship between the likelihood of becoming an acquirer and the financing variables defined above. This analysis is important in order to establish the existence of correlation between the merger activity and the preference for equity financing. In Table 4.3, I sort the Compustat universe into bins based on external finance (XF) and Equity Share (ES) at the firm level. For variable definitions see the Appendix. I further condition on the level of merger activity based on industry de-trended intensity. Figure 4.2 shows the time series of de-trended intensity. In all panels of Table 4.3 we see that, conditional on the level of external fi- nance, the likelihood of a firm becoming a bidder increases as the relative impor- tance of equity increases, relative to debt issues. For the overall sample, this effect is larger and more significant for levels of external finance around zero. In Panel B I focus on stock deals, and the evidence shows that the behavior of stock deals conforms more closely to what a timing behavior would predict. For a firm with positive external finance, the likelihood of becoming a bidder increases from es- sentially zero for a firm issuing exclusively debt to about 5% for a firm issuing exclusively equity. In Panel C I focus attention on stock deals announced in years with high merger intensity. The market timing motive predicts that the relationship between merger and equity financing should be stronger when industry merger activity is high. The results in Panel C confirms this prediction. The likelihood of becoming a bidder increases to 7.5% when a firm issues exclusively equity. 4.4.4 Panel Data Results Table 4.4 shows the results of the panel regression of merger intensity in equation 4.1. Column (1) shows the main specification. For completeness I also show the results when different sets of controls are included in the regression. All right- hand-side variables in equation 4.1 are expressed in units of standard deviation to aid interpretation. Consistent with the prediction of the market timing hypothesis, the coefficient on ES is positive and statistically significant in all specifications in Table 4.4. Ex- 86 cept for the deregulation indicator, none of the time-varying industry controls are significant in column (1). To some extent, this is not surprising as the regression includes the average level of external finance XF . It is plausible to argue that this variable captures the demand for funds, which subsumes the determinants of merger activity. In unreported results, I find that when ES and XF are excluded from regression 4.1, GDP and vIK become significant at conventional levels. In column (2) I show that the financing variables, together with the industry fixed effects, explain 89% of the total variation in industry merger intensity. In column (3), I replace all time-varying macro variables with year fixed effects. The evidence regarding the relevance of equity share remains qualitatively unchanged. The aggregation process to construct ES and XF involves taking the average of firm specific ratios. In column (4) I report the results when ES and XF are con- structed with the industry-level variables (i.e., the ratio of the averages instead of the average of the ratios). Although the coefficient on ES is smaller, the statistical significance remains. Even though I am using the average industry figures in the main specification, there could be some remaining concern about simultaneity bias (e.g. peer effects). In column (5) I present the results from a regression in which XF and ES are computed using non-merging firms. The sample of non-merging firms includes firms that, for a given benchmark year t, are neither bidders nor targets in the 5- year window [t−2, t +2]. The use of the non-merging sample alleviates concerns that unobserved firm-specific factors, other than market timing motives, drive both the acquiring and financing patterns of interest. A final comment refers to the effect of serial correlation in the error term. The main specification shows that the autoregressive term in the residual is highly sig- nificant. The Durbin Watson test is close to 2, which indicates that the remaining serial correlation is not significant. In column (6) I re-estimate the regression in column (1) with OLS. Although the qualitative evidence regarding the timing ef- fects remains the same, the size of the coefficients changes substantially. I discard these coefficients as they are likely biased due to the combination of serial correla- tion in the residuals and persistent regressors. Overall, the evidence from this test provides support for the market timing prediction that merger activity is positively associated with an incentive to tilt fi- 87 nancing decisions towards equity financing. The main specification in 4.1 shows that an increase in one standard deviation in ES is associated with an increase in merger activity of about 0.9% on average. This is a sizable change in merger in- tensity; it represents about 20% of the average intensity across industries in Table 4.2. 4.5 Dynamic Allocation of Issuance Proceeds The second test is related to the dynamic relationship between the issuance of eq- uity and the use of proceeds, conditional on the level of merger activity. Having shown that market timing is a relevant motive in the time series of merger activity, in this section I try to disentangle the three alternative versions of the market timing hypothesis described above. The core insight for this test rests on two premises. First, the time-series correlation between equity issuance and the use of proceeds is informative about the motive for the equity issue (Kim and Weisbach, 2008). Sec- ond, I argue that the inter-temporal allocation of proceeds should change between high and low M&A activity years in a way that is informative about the drivers of M&A activity over time. 4.5.1 The Core Predictions The investment friction hypothesis predicts that times of high merger activity are characterized by an increase in investment financed with a conservative leverage ratio. This hypothesis implies that equity issues at time t should be positively and contemporaneously correlated with investment. Firms issue equity because they need to invest. As firms start refinancing in future years, this hypothesis pre- dicts that equity issues at time t should be positively correlated with future debt repayments and equity payout. This correlation should increase over time as the firm approaches the project’s completion time. The reversal in equity financing is expected to be small if firms prefer to retain the future refinancing proceeds for investment (i.e., if the firm is growing). The adverse selection hypothesis predicts that when the information asymme- try in stock markets is temporally low, firms expecting the arrival of investment opportunities issue equity to increase cash holdings or to increase debt capacity. 88 This increase in financial slack serves a precautionary role so as to avoid underin- vestment in the future. This hypothesis implies that at times of high merger activity, firms should issue equity and allocate most proceeds to increasing financial slack. Thus, the predictions involve a positive correlation between equity issuance and increases in cash holdings at time t. Alternatively, the firm can use the proceeds to re-pay debt. This saving is expected to be depleted as investment opportunities ar- rive in the future. Thus, equity issues at time t should be positively correlated with future investment, with an increasing pattern over time. Financially constrained firms should be more prone to this precautionary saving behavior. Future equity payout is not expected to change under this hypothesis. The overvaluation hypothesis states that managers facing overvalued equity can exploit the mispricing by substituting existing debt and equity financing for (risk adjusted) less expensive capital. This substitution is desirable if managers care about existing shareholders, as the arbitrage-refinancing implies a redistribution of wealth from new to existing capital providers. The overvaluation hypothesis thus predicts that high merger activity times are characterized by more of such arbitrage activity by managers. By the nature of mispricing (i.e., systematically unrelated to fundamentals), overvalued equity will not systematically be available when firms desire to increase investment. Thus, a financially unconstrained firm is not expected on average to increase real investment under this hypothesis. In summary, this hypothesis predicts a positive contemporaneous correlation between equity issues and debt repayment. Alternatively, as proceeds are excess cash, this hypothesis predicts that equity issues at time t should be positively correlated with equity payout in future times. These predictions are expected to hold more clearly for financially unconstrained firms. Up to this point, the predictions relate equity issues to their driving force through the use of proceeds. I link these predictions to the underlying forces be- hind merger waves by conditioning on the level of merger activity at the time of the equity issue. For example, if the overvaluation explanation is the actual driver of both merger waves and equity issues, then a dollar of equity issued in high merger time should be used more intensively to hoard cash. In other words, the change in the allocation of equity proceeds between low and high merger states should be informative about what drives M&A activity. Table 4.5 summarizes the predictions 89 about the allocation in high merger times relative to low merger times. Column t describes what should happen at the time of the issuance. Column t + τ describes what happens over time. 4.5.2 Flow of Funds Regression In order to obtain the correlation between equity issuance and the use of funds at different future times, I start from the accounting identity in the Statement of Cash Flows: ∆Cash = Financing Activities+ Investing Activitties+Operating Activities (4.2) I re-arrange the items in this identity based on whether they are sources or uses of funds. Thus, for every firm i in year t and industry j, the following holds: Equityi jt =∆Cashi jt+DPayouti jt+EPayouti jt+Capexi jt+CAcqi jt+OtherUsesi jt (4.3) where OtherUses captures all accounting flows that are not in the other variables49. I delete all firm-year observations for which the cash flow identity 4.2 is violated by more than 1% of total assets. After this, any discrepancy is added to OtherUses in 4.3. Each variable at right hand side of 4.3 corresponds to one of the five uses of funds in Table 4.5. More details and Compustat codes are presented in the Appendix. This particular decomposition allows me to study the contemporaneous and dynamic allocation of equity proceeds through the following system of equations: ∆Cashi jt = α j +β11Equityi jt +β12Equityi jt−1+β13Equityi jt−2+ γ ′1Xi jt + ε1i jt (4.4) DPayouti jt = α j +β21Equityi jt +β22Equityi jt−1+β23Equityi jt−2+ γ ′2Xi jt + ε2i jt (4.5) EPayouti jt = α j +β31Equityi jt +β32Equityi jt−1+β33Equityi jt−2+ γ ′3Xi jt + ε3i jt (4.6) Capexi jt = α j +β41Equityi jt +β42Equityi jt−1+β43Equityi jt−2+ γ ′4Xi jt + ε4i jt (4.7) CAcqi jt = α j +β51Equityi jt +β52Equityi jt−1+β53Equityi jt−2+ γ ′5Xi jt + ε5i jt (4.8) OUsesi jt = α j +β61Equityi jt +β62Equityi jt−1+β63Equityi jt−2+ γ ′6Xi jt + ε6i jt (4.9) 49I also add to OtherUses the foreign currency translation adjustments (exre in Compustat) 90 Following Dasgupta et al. (2011), all variables other than Xi jt are scaled by total assets at time t − 3. The vector Xi jt includes leverage and Tobin’s Q, both measured at time t−3. The Appendix describes these variables. This system of equations 4.4–4.9 is special in two respects. First, these regres- sions have no causal interpretation. The only purpose is to compute the average firm’s allocation of cash coming from equity issues. Second, given the identity in 4.3, it holds that ∑6l=1βl1 = 1 and ∑ 6 l=1βlk = 0, for k = 2,3. Along the same lines, the sum across equations of γ for each control variable must also be zero. More importantly, ∑6l=1 εli jt = 0 for each i jt, which implies that the system is singular (i.e., the variance-covariance matrix for the system is singular) Following Dasgupta et al. (2011), this system is estimated with OLS on an equation-by-equation basis. This approach sidesteps the singularity of the variance- covariance matrix50. Furthermore, given that all equations have the same explana- tory variables, there should not be loss in efficiency with respect to a system (i.e, seemingly unrelated regression) estimation (see, Greene, p. 343) This dynamic system allows me to explore how equity proceeds are allocated contemporaneously and over time. For example, β11 captures how much (in cents) out of a dollar of equity proceeds is allocated to increasing cash in the same year as the equity issuance. The dynamic effect is captured by the sum of coefficients on lags. For example, β41+β42+β43 measures the three-year [t, t +2] cumulative allocation (in cents) towards increasing capital expenditures for every dollar of equity issued at time t. I depart from Dasgupta et al., (2011) in one important respect. They focus on the allocation of operating cash flows to different uses. Thus, they focus on the question: Where does an extra dollar of operating cash flows go? However, my test in this section addresses a different question: Does the allocation of equity proceeds change during high (versus low) merger intensity times? If so, are these 50The singularity of the variance-covariance matrix is also common in the estimation of demand systems (see Barnett and Serletis (2008) for a recent survey on demand systems). Barten (1969) shows that a solution for this singularity is to drop one equation and estimate the remaining system with full information maximum likelihood (FIML). As a robustness check, I also estimated the sys- tem 4.4–4.9 with FIML dropping different equations. The point estimates I obtained for βlk are very similar to what I report below (i.e., equation-by-equation estimation). If anything, the results with FIML are statistically more significant. 91 changes in proceeds allocations consistent with the predictions implied by any of the hypothesis under study? Thus, I enrich the specification with an interaction term. In particular, I allow βlk, k = 1,2,3 in equation l to vary linearly and deterministically with the level of M&A activity: βl1 = γ01+ γ11H jt βl2 = γ02+ γ12H jt−1 βl3 = γ03+ γ13H jt−2 where the indicator variable H jt equals 1 during the years in which the intensity of merger activity in industry j is high and 0 otherwise. To obtain this indicator, I first de-trend the merger intensity in each industry. H jt = 1 in year t if the residual of this regression at year t is in the top third of all residuals, and H jt = 0 otherwise. The inclusion of H jt creates a series of interaction terms that are the focus on interest. In each regression I also include the level effect of H jt to capture any potential direct effect on the allocation of cash. In order to refine the prediction above, I will analyze sub-samples based on financial constraints and investor horizon. As a measure of investor horizon, I use the measure developed in Gaspar, Massa and Matos (2005). This measure is based on the average turnover of institutional investors (13-F filing) for firm i (see Gaspar et al. (2005) for details). Firms with high turnover investors are classified as having shorter horizon. As a proxy for financial constraints, I use the Size-Age (SA) index developed by Hadlock and Pierce (2010). The SA index is based on firm i’s size and age. In general, larger and older firms belong to the set of less financially constrained firms. In particular, the index is computed as −0.737Size+ 0.043Size2− 0.040Age, where Age is the number of years the firm has been in Compustat with non-missing stock price and Size is the log of total assets (adjusted for inflation to 2004 dollars). Size and Age are capped at log(4.5b.) and 37 years, respectively. 92 4.5.3 Results In this section I explore the use of equity funds raised at times of high merger activity, relative to those raised in low merger states. Thus, my discussion will focus exclusively on the coefficients associated with the interaction terms. In Table 4.6 I look at the results from the entire sample of firms. Figure 4.4 plots the initial effects and the evolution over time for all uses of funds in Table 4.6. For a dollar of equity proceeds raised at time t, cash holdings increase by 11.5 cents. By the end of year t+2, this extra initial accumulation of cash is almost fully depleted. The change in allocation of equity proceeds at time t towards capital expenditures reaches 12.1 cents in year t; an additional 4.7 cents investment is completed the year after the equity issuance. By the end of year t+2, the effect on Capex reaches 16.2 cents. Debt payment also shows an increase in years t + 1 and t + 2, with a cumulative effect of 27.6 cents by the end of t + 2. Equity payout shows a very modest but statistically significant increase of 1.4 cents at t+2. Although these findings could in principle be consistent with the three hypothe- ses, I highlight what tilts the balance in favor or against any hypothesis. For exam- ple, the adverse selection motivation predicts an increasing pattern of investment, while the overvaluation hypothesis predicts no investment (unless the firm is fi- nancially constrained). Contrary to these predictions, capex significantly increases upfront in the data, with a fairly flat pattern in the following years (see Figure 4.4). Overall, the allocation pattern on capex is more consistent with the investment frictions hypothesis (see Table 4.5). The increasing pattern on debt payment is also consistent with the investment frictions hypothesis. In looking at predictions violations, we see that equity payout increases in year t+2. Although this is predicted by the overvaluation hypothesis, the magnitude of the increase is too small to claim it is a first order effect. Also, in contrast to the overvaluation hypothesis, debt repayment does not increase in year t. At this point, it seems that the typical firm is not using equity proceeds raised in high merger markets to refinance old financial capital. The significant upfront accumulation of cash holdings is predicted by the adverse selection hypothesis. However, the depletion of cash over time is very fast. Furthermore, this cash depletion does not seem to be used for investment. These aspects put in question the predictions from 93 the adverse selection hypothesis. A plausible explanation for the lack of support for the predictions from the overvaluation and adverse selection hypotheses in Table 4.5 is that financial con- straints cloud the results. For example, a financially constrained firm is already investing less than the desired level, so any cash inflow, even if unrelated to pro- ductivity, could profitably be allocated to real investment. Consequently, I split the sample in two groups based on the SA index of Hadlock and Pierce (2010). I re-estimate the system of equations with these two sub-samples. The results are presented in Table 4.7. The basic tenet of this exercise is two-fold. First, if the overvaluation hypoth- esis is true, the sample of less financially constrained firms should perform the substitution of financial capital more clearly. Second, if the adverse selection hy- pothesis is true, the sample of more financially constrained firms should perform the expected precautionary saving more actively. Table 4.7 (and Figure 4.5) shows that, for financially more constrained firms, there is a significant and permanent increase in cash holdings. The allocation of equity proceeds towards investment is not significant. Both findings are in line with the adverse selection hypothesis. As discussed above, the combination of overval- ued equity and financial constraints could result in an increase in investment. Table 4.7 does not show such a pattern. Predictions from the overvaluation hypothesis should shine in the sample of less financially constrained firms. Table 4.7 does not show firms increasing debt payment upfront, or increasing substantially equity payout over time. Even though the equity payout is larger in this sub-sample, it is still small (about 3 cents) to argue in favor of the overvaluation hypothesis. Interestingly, the support for the in- vestment friction hypothesis becomes stronger in the sub-sample of less financially constrained firms. For this subset of firms, the results match nicely what Table 4.5 describes for the investment frictions hypothesis. A final concern relates to the potential of a ‘catering’ explanation for the in- crease in the allocation to investment in high merger times (Polk and Sapienza, 2009). The catering argument says that managers can increase stock prices beyond fundamentals by catering to the type of investments investors (irrationally) want. Although it is true that catering could explain the positive correlation between eq- 94 uity issues and investment, the catering story would not predict that firms will use funds from equity issues to increase debt payments or equity payout over time. If investors provide funds for a specific type of investment they like, all proceeds should be allocated to investment. In order to explore this argument in some more detail, I split the sample of less financially constrained firms in two groups based on investors’ horizon. I mea- sure investment horizon with the measure developed by Gaspar, Massa and Matos (2005). This measure is based on the average turnover of institutional investors in a given firm. Firms with a high average investor turnover are classified as having shorter horizon. Figure 4.6 plots the results. The results found for the group of less financially constrained firms become stronger once I consider the sub-sample with shorter-horizon firms. In particular, this subset of firms increases even more equity and debt payout over time. This finding poses a challenge for the catering story and strengthens the support for the investment friction hypothesis. 4.6 Final Remarks and Conclusion A merger wave requires both a deal motivation and a source of deal clustering. On this regard, the current consensus in the literature seems to be that industry shocks and waves of equity overvaluation explain the behavior of merger activity over time. The fact that merger activity peaks at times of high stock market valuations seems to support the view that equity overvaluation drives merger waves. In this chapter I argue that explanations based on time-varying adverse selection and real investment frictions provide alternative and equally plausible explanations for this outcome. I start by documenting that the time-series behavior of merger activity is pos- itively associated with a preference for equity financing at the industry level. In order to disentangle the three hypotheses I explore in this chapter, I exploit their differing predictions regarding the use of proceeds from equity issued at times of high merger activity. I fail to find support for the predictions implied by the over- valuation hypothesis. The adverse selection hypothesis finds support in a subset of firms that face tighter financial constraints. The investment frictions hypothe- sis provides a more consistent explanation for the set of findings, in particular for 95 the subset of financially less constrained firms. Overall, the evidence I provide supports the notion that investment and financial frictions are important elements behind the dynamics of merger activity. The findings I document in this chapter are in line with previous research. For example, Eisfeldt and Muir (2013) document a positive correlation between the issuance of external finance and the accumulation of liquid assets. They argue that this positive correlation is consistent with time-varying costs of external finance or shadow cost of liquidity. Dittmar and Dittmar (2008) study waves in corporate financing events. They find a positive correlation between repurchases and equity issues or mergers, suggesting that equity overvaluation is not the driver of corporate financing events. While both repurchases and issues tend to increase over expan- sionary periods and decrease over contractionary periods, growth in issues tends to occur in earlier stages of the cycle than does growth in repurchases. DeAngelo et al. (2010) also find mild support for the overvaluation explanation for equity issues. They show that firms issue equity when equity prices are high, but only when firms need to meet near-term cash needs (e.g., investment). Similarly, Alti and Sulaeman (2012) document that high stock returns are more likely to trigger equity issues at times of strong institutional demand. This evidence lends further support to my conclusion that financial and real frictions are important drivers of merger waves. Sharpening our understanding of whether equity overvaluation or frictions ex- plain the time series of merger activity raises important questions. For example, questions about the economics of deal initiation and the channels through which an initial trigger propagates over time and across industries. In particular, my re- sults speak to the long-standing question as to whether merger transactions enhance economic efficiency or they are purely redistributive events with potentially dead- weight costs. Along the same lines, different explanations for the observed market timing behavior in merger activity imply different outcomes for deals carried out in a merger wave. The results in this chapter is also relevant for the interpretation of existing evidence, mostly based on announcement stock price reactions and long-run bid- ders’ performance. For example, if acquirers’ overvaluation drives merger waves, the inference about value creation based on stock returns around or after deal an- 96 nouncements has limited value as it confounds the economic value of the deal and the correction in bidder overvaluation. In contrast, if frictions drive the dynamics of mergers over time, then any long-run abnormal return must come from a poorly defined benchmark. These are important venues for future research. 97 Panel A: Merger Intensity Regression Merger Intensity (I) Number of mergers in every quarter-industry combination over the number of firms in CRSP for the same industry at the end of the previ- ous quarter. Annual intensity is the sum of the four quarterly intensities. External Finance (XF) Net Equity Issuance (inflow-outflow) plus Net Debt Issuance over total assets: XF(t) = sstk(t)−dv(t)−prstkc(t)+dltis(t)+dlcch(t)−xint(t)at(t−1) . Equity Share (ES) Percentage of (gross) equity issuance over (gross) total issuance: ES(t) = sstk(t) sstk(t)+dltis(t)+max(dlcch(t),0) . IK Volatility (vIK) The (cross section) standard deviation of the IK ratio for the industry- year combination. The IK ratio is defined as IK(t) = capx(t)ppent(t−1) . Economic Shock (EcS) Following Harford (2005), this is the first principal component from the median of the absolute year-to-year change in six accounting ratios: CFmg(t) = ib(t)+d p(t)sale(t) ; Aturn(t) = sale(t) at(t−1) ; Capex(t) = capx(t) at(t−1) ; HL = ∆emp(t) emp(t−1) ; ROA = ib(t) at(t−1) . I exclude Research and Development due to a large number of missing values. Tangibility (Tang) I follow Berger, Ofek and Swary (1996) to construct an asset tangibil- ity measure. It measures the expected asset liquidation value with the index: Tang(t) = che(t)+0.72rect(t)+0.55invt(t)+0.54ppent(t)at(t) . Deregulation (DReg) This is an indicator that equals 1 if the industry has a deregulatory event in year t−1 and 0 otherwise. I follow Harford (2005) for this classifi- cation (Table 2, p. 538). GDP This is the cyclical component of GDP. I use the annual real (log) GDP from Fred and HP filter (with penalty parameter 6.25) to extract the cyclical component. Time Time is a time trend from 1 (1983) to 28 (2010). Panel B: Use of Funds Regression Equity sstk, scaled by total assets lagged three years. Cash Change chech, scaled by total assets lagged three years. Debt Payout dltr−min(dlcch,0), scaled by total assets lagged three years. Equity Payout prstkc+dv, scaled by total assets lagged three years. Capex capx, scaled by total assets lagged three years. Cash Acquisitions aqc, scaled by total assets lagged three years. OtherUses Everything else that completes the cash flow identity. Under for- mat code 7, OtherUses = ivch− siv− ivstch− sppe− ivaco− OANCF + dltis + max(dlcch,0) + exre+ f iao. Missing values are set to zero. I delete all firm-year observations for which the cash flow identity is violated by more than 1% of total assets. After this, any discrepancy is added to OtherUses. This variable is also scaled by total assets lagged three years. For format codes 1, 2 and 3, variables are defined consistently following the Compustat manual. Tobin’s Q at−ceq+csho∗prcc fat , the market value of total assets over the book value of total assets. Leverage dlc+dlttat , the book value of financial debt over the book value of total assets. 98 A. Figures and Tables Figure 4.1: Merger Intensity Plot. The figure shows the intensity of merger activity. The number (value) of mergers in every quarter-industry combination is scaled by the number (market capitalization) of firms in CRSP at the end of the previous quarter. An- nual intensity is obtained by adding the four quarterly intensities. The figure plots the intensity of mergers (in number and value) and the intensity of stock mergers for two of the most active (Business Equipment and Telecommunications) and two of the most passive (Consumer Durables and Chemicals) industries in the sample. Industry definition follows the Fama-French 12-industry classification, based on 4-digit SIC codes. 99 Figure 4.2: Detrended Merger Intensity. The figure shows the intensity of merger ac- tivity after linear de-trending. The number of deals in every quarter-industry combination is scaled by the number of firms in CRSP at the end of the previous quarter. Annual inten- sity is obtained by adding the four quarterly intensities. The analysis includes 9 industries. The industries excluded are: Financials (8), Utilities (11) and Others (12). 100 Figure 4.3: Intensity based on Value and Number of Deals. The figure shows the scatter plot and histogram of merger intensity defined with number (I) or value of mergers (P) for two industries: Health Care and Consumer Durables. These industries have the lowest correlation between both measures of intensity. The number of deals in every quarter-industry combination is scaled by the number of firms in CRSP at the end of the previous quarter. Annual intensity I is obtained by adding the four quarterly intensities. Intensity based on values is defined similarly with deal value (from SDC) and market capitalization (from CRSP) 101 Figure 4.4: Use of Funds Dynamics. The figure plots the average dynamics of equity proceeds allocation in high and low merger times. I assign years into High and Low by sorting (within industry) the de-trended Merger Intensity in three groups. Years in the top intensity group are designated High and years in the medium and low intensity groups are designated Low. The plots are obtained by adding the coefficients in Table 4.6. The first five plots correspond to the five uses of funds. The last plot shows the change (High minus Low) over time in the use of proceeds for each of these five uses. 102 Figure 4.5: Use of Funds Dynamics by Financial Constraints. The figure plots the average dynamics of equity proceeds allocation in high and low merger times for sub-samples based on financial constraints (SA Index). The plots are obtained by adding the coefficients in Table 4.7. The figure shows the five uses of equity proceeds, where the first plot for each use of funds corresponds to the sample of less financially constrained firms. 103 Figure 4.6: Use of Funds Dynamics by Investment Horizon. The figure plots the average dynamics of equity proceed allocation in high and low merger times for sub-samples based on investment horizon within the sub-sample of less financially constrained firms. The figure shows side by side the five uses of equity proceeds, where the first plot for each use of funds corresponds to the sub-sample of longer investment horizon (higher turnover) firms. 104 Table 4.1: Sample Deals. The sample includes all domestic deals (US target) available in SDC that were announced between January 1, 1983 and December 31, 2010. A stock deal is defined as a deal in which at least 95% of the payment was in stock. A cash deal is defined similarly. I exclude firms in the financial, utility and other industries. Panels are sorted in descending order on acquisition value. Mergers Cash Deals Stock Deals Industry (FF12) Number US$ b. Number US$ b. Number US$ b. Panel A: Overall Sample Business Equipment 2389 1156.0 527 227.0 1200 720.4 Telecommunication 482 1039.0 69 28.4 159 372.2 Healthcare 858 688.5 181 113.6 347 271.9 Energy 422 505.8 62 38.3 90 140.4 Manufacturing 712 363.8 238 72.8 136 94.9 Wholesale and Retail 589 291.3 146 44.4 191 64.3 NonDurables 361 211.9 104 31.1 62 14.2 Chemicals 151 207.7 50 20.1 25 87.9 Durables 155 63.4 60 15.3 24 3.0 Panel B: 1983 - 1989 Manufacturing 169 56.5 77 33.4 19 6.3 Chemicals 43 36.8 17 5.7 6 1.8 NonDurables 73 31.9 20 5.7 8 1.0 Energy 62 30.4 4 2.5 7 0.5 Healthcare 69 26.4 10 0.6 17 14.0 Wholesale and Retail 89 20.5 28 8.3 15 4.1 Telecommunication 63 19.9 11 2.1 16 3.8 Business Equipment 102 18.8 27 4.0 38 4.8 Durables 40 16.9 20 8.2 2 0.8 Panel C: 1990 - 1999 Telecommunication 264 700.1 28 14.4 99 288.1 Business Equipment 1018 290.5 120 33.5 698 197.5 Healthcare 388 220.3 47 25.7 219 150.0 Manufacturing 318 179.0 77 13.1 84 70.8 Wholesale and Retail 312 139.9 57 10.6 125 37.5 Energy 157 135.9 29 7.4 40 99.9 Chemicals 62 81.6 22 11.9 9 28.0 NonDurables 140 38.6 30 7.8 41 10.7 Durables 63 26.2 19 4.4 15 1.1 Panel D: 2000 - 2010 Business Equipment 1269 847.0 380 189.5 464 518.1 Healthcare 401 441.8 124 87.3 111 107.9 Energy 203 339.5 29 28.4 43 40.0 Telecommunication 155 319.1 30 12.0 44 80.3 NonDurables 148 141.5 54 17.5 13 2.6 Wholesale and Retail 188 130.9 61 25.6 51 22.7 Manufacturing 225 128.3 84 26.3 33 17.8 Chemicals 46 89.2 11 2.6 10 58.0 Durables 52 20.2 21 2.7 7 1.1 105 Table 4.2: Merger Intensity Descriptive Statistics. The table shows (time series) de- scriptive statistics for merger intensity by industry. Annual merger intensity is the sum of quarterly merger intensities. The quarterly merger intensity is defined as the number of deals in every quarter-industry combination, scaled by the number of firms in CRSP at the end of the previous quarter. The analysis includes 9 industries. The industries excluded are: Financial (8), Utilities (11) and Others (12). The last two columns show the time se- ries correlation between intensity defined with number of deals (IN) and intensity defined with deal values (IV ). The asterisk denotes significance at 5% or less. Industry (FF12) Merger Annual Intensity (%) Corr(IN , IV ) Mean St. Dev. Max. Min. Spearman Pearson Consumer NonDurables 4.4 1.8 9.1 0.3 -0.11 0.04 Consumer Durables 4.0 2.2 9.3 0.0 0.68* 0.60* Manufacturing 4.2 1.9 7.7 1.5 0.56* 0.45* Energy 7.2 4.5 17.8 0.8 0.47* 0.35* Chemicals 4.4 2.5 9.7 0.0 0.60* 0.29 Business Equipment 7.9 5.7 22.2 1.1 0.73* 0.79* Telecommunication 11.1 6.0 28.1 2.4 0.67* 0.63* Wholesale and Retail 3.6 2.0 7.6 1.2 0.46* 0.60* Healthcare 5.4 2.4 9.7 1.2 0.27 0.11 106 Table 4.3: Bidder Likelihood by Financing Sorts. The table shows the proportion of firms in a given bin that become bidders (at least once) during year t. ES refers to the Eq- uity Share defined as equity issuance over debt plus equity issuance. XF refers to external finance defined as Net Debt plus Net Equity over (lagged) total assets. Sorts on XF and ES are independent. Panels A uses the whole sample and Panel B uses only stock deals. Panel C uses stock deals in years classified as High Merger years. In order to assign years into High/Low merger years, I first sort (within industry) the de-trended Merger Intensity and assign each year to a High, Medium or Low group. t-stats are computed with standard errors assuming maximum variance (i.e., p = q = 0.5) and the number of observations in each cell. Panel A: Overall Sample Equity Share XF > 0 ES = 1 0.5 < ES < 1 0 < ES < 0.5 ES = 0 H−L tstat External Finance H 6.9 5.9 11.5 5.1 1.8 0.92 L 7.2 6.4 6.5 2.1 5.1 3.87 H−L -0.3 -0.5 5.0 3.0 tstat -0.27 -0.39 6.46 1.44 Equity Share XF < 0 ES = 1 0.5 < ES < 1 0 < ES < 0.5 ES = 0 H−L tstat External Finance H 3.5 5.4 5.4 1.2 2.3 2.35 L 3.3 5.2 4.4 1.5 1.8 1.84 H−L 0.2 0.2 1.0 -0.3 tstat 0.20 0.13 1.40 -0.31 Panel B: Stock Deals Equity Share XF > 0 ES = 1 0.5 < ES < 1 0 < ES < 0.5 ES = 0 H−L tstat External Finance H 4.6 3.5 2.6 0.4 4.2 2.13 L 4.1 3.6 1.9 0.5 3.6 2.73 H−L 0.5 -0.1 0.7 -0.1 tstat 0.45 -0.08 0.90 -0.05 Equity Share XF < 0 ES = 1 0.5 < ES < 1 0 < ES < 0.5 ES = 0 H−L tstat External Finance H 1.4 1.8 1.3 0.2 1.2 1.23 L 0.9 1.5 1 0.2 0.7 0.71 H−L 0.5 0.3 0.3 0.0 tstat 0.50 0.19 0.42 0.00 Panel C: Stock Deals in High Merger years Equity Share XF > 0 ES = 1 0.5 < ES < 1 0 < ES < 0.5 ES = 0 H−L tstat External Finance H 7.5 5.6 4.4 0.8 6.7 1.99 L 7.7 5.8 2.8 0.4 7.3 3.44 H−L -0.2 -0.2 1.6 0.4 tstat -0.12 -0.11 1.32 0.11 Equity Share XF < 0 ES = 1 0.5 < ES < 1 0 < ES < 0.5 ES = 0 H−L tstat External Finance H 2.8 3.9 2.1 0.2 2.6 1.49 L 1.1 2.9 1.6 0.2 0.9 0.51 H−L 1.7 1.0 0.5 0.0 tstat 0.98 0.36 0.39 0.00 107 Table 4.4: Merger Intensity Regression. The table shows a panel regression of Merger Intensity with industry fixed effects. The regression has 252 observations: 9 industries and 28 years. All variables are defined in the Appendix. ES refers to Equity Share defined as equity issuance over debt plus equity issuance. XF refers to external finance defined as Net Debt plus Net Equity. Net means net of payout. Equity payout includes dividends and repurchases. Columns 1-5 use maximum likelihood estimation with an AR(1) error term to adjust for serial correlation in the residuals. The Durbin Watson (DW) test for serial correlation in the residuals is also reported. Column (5) uses only the sample of non-merging firms to compute the left hand side variables. Non-Merging Firms is the sample of firms that, in the 5-year window [t− 2, t + 2], are neither bidders nor targets in an acquisition. T-stats are presented below the estimates. Dependent Variable Merger Intensity All Firms Non-Deal All Firms (1) (2) (3) (4) (5) (6) (7) Financing Variables Equity Share (ES) 0.922 1.006 1.110 0.682 0.867 1.262 2.037 3.77 4.73 3.45 3.64 3.56 5.31 2.37 External Finance (XF) 1.032 1.079 0.660 0.910 0.830 1.571 0.842 5.02 5.78 2.78 5.26 4.02 4.57 1.60 Control Variables IK Volatility (vIK) 0.083 -0.016 0.389 0.131 -0.120 0.289 0.47 -0.09 2.37 0.75 -0.46 9.41 Economic Shock (EcS) 0.165 0.014 -0.032 0.124 0.344 0.038 0.78 0.06 -0.15 0.58 1.41 0.15 Tangibility (Tang) -0.350 0.051 0.015 -0.027 -0.565 -0.135 -1.07 0.15 0.04 -0.08 -2.77 -0.42 Deregulation (DReg) 2.142 2.165 2.325 2.066 1.836 2.799 2.18 2.13 2.34 2.09 1.70 5.38 GDP 0.201 0.404 0.198 0.235 1.20 2.40 1.17 1.51 Time Trend 0.314 1.688 0.663 0.058 0.70 4.20 1.46 0.16 Error term AR(1) -0.499 -0.518 -0.463 -0.519 -0.514 -8.39 -9.17 -7.33 -8.69 -8.69 DW(1) 2.00 2.04 2.04 2.03 1.99 DW(2) 2.01 1.95 1.96 1.93 2.06 Industry FE YES YES YES YES YES YES NO Year FE NO NO YES NO NO NO YES Estimation Method ML ML ML ML ML OLS OLS R-Square 0.89 0.89 0.91 0.89 0.89 0.87 0.59 108 Table 4.5: Use of Funds Predictions. Use of Proceeds for Equity Issued at t Equity Overvaluation Adverse Selection Investment Frictions t t+ τ t t+ τ t t+ τ Cap. Expenditures Increasing Positive Flat Cash Acquisitions Increasing Positive Flat Debt Payment Positive Flat Positive Decreasing Increasing Cash Increase Positive Decreasing Positive Decreasing Equity Payout Increasing Increasing 109 Table 4.6: Use of Funds Regressions. The table shows a set of six regressions of different uses of funds on Equity Issue. In order to incorporate dynamics, I add two lags of Equity Issue. I use an interaction term to capture the change in the propensity to use the proceeds in different uses. H is an indicator variable that equals 1 in years with High Merger activity. In order to find the High Merger years, I first sort (within industry) on the de-trended Merger Intensity and assign each year to a High, Medium or Low group. H = 1 for the top third (High group) and 0 otherwise. All variables are defined in the Appendix. All regressions include industry and year fixed effects. Standard errors are clustered by year. T-stats are presented below the estimates. EquityIssue EquityIssue×H H TobinQ Leverage t t−1 t−2 t t−1 t−2 t t−1 t−2 t−3 t−3 R−Sq. N Cash Change 29.2 -16.6 0.9 11.5 0.6 -10.9 -0.6 -1.1 0.7 1.2 -1.8 0.22 57,076 8.93 -3.55 0.29 2.34 0.10 -2.98 -0.92 -1.61 1.48 3.22 -1.94 Debt Payout 11.8 0.3 1.6 9.4 8.9 9.3 1.8 0.6 0.5 -1.9 28.0 0.04 57,084 2.17 0.18 0.81 0.95 1.68 1.79 1.73 0.84 0.78 -3.17 7.67 Equity Payout -0.1 0.0 0.1 0.3 0.4 1.4 0.2 0.0 0.1 0.5 -0.2 0.11 57,011 -0.26 0.12 0.41 0.94 0.93 2.61 1.56 0.24 0.94 3.95 -0.35 Capex 10.3 8.2 8.9 12.1 4.7 -0.6 1.2 0.5 0.1 0.1 4.3 0.32 56,581 2.44 5.17 4.69 1.81 1.85 -0.23 1.70 1.21 0.27 0.61 4.24 Cash Acqs. 14.0 2.5 0.2 5.3 1.1 2.1 1.7 1.3 1.7 -1.4 10.6 0.09 54,933 1.66 0.98 0.11 0.51 0.31 1.17 1.83 2.33 3.73 -4.42 7.79 Other Uses 35.1 5.7 -11.4 -37.7 -14.7 -1.4 -4.3 -1.3 -3.1 1.0 -40.7 0.08 57,084 2.17 0.90 -2.53 -2.01 -1.37 -0.18 -2.31 -0.94 -2.76 1.80 -7.42 110 Table 4.7: Use of Funds Regressions and Financial Constraints. The table shows a set of six regressions of different uses of funds on Equity Issue for two sub-samples based on a measure of financial constraints. I use the SA index proposed by Hadlock and Pierce (2010) to measure financial constraints. I compute this measure every year for all firms and assign firms into two groups (More and Less financially constrained) by year using the Compustat universe. Panel A: Less Financially Constrained EquityIssue EquityIssue×H H TobinQ Leverage t t−1 t−2 t t−1 t−2 t t−1 t−2 t−3 t−3 R−Sq. N Cash Change 30.0 -11.5 19.8 9.5 -4.6 -31.8 -0.4 -0.6 0.1 1.4 0.1 0.21 36,131 6.98 -1.12 1.90 1.55 -0.33 -2.79 -0.56 -0.74 0.23 2.09 0.08 Debt Payout 12.7 4.1 1.9 10.4 10.4 29.9 1.9 0.5 1.4 -1.8 40.3 0.05 36,138 1.34 1.34 0.40 0.73 1.06 1.80 1.43 0.58 1.45 -1.99 10.02 Equity Payout 0.2 0.8 -0.2 0.1 -0.6 3.2 0.3 0.2 0.3 0.7 0.6 0.15 36,102 0.40 1.03 -0.41 0.27 -0.70 2.42 1.74 1.42 1.63 2.47 0.69 Capex 6.5 12.1 22.6 17.7 5.0 -1.5 1.0 0.3 0.5 0.2 7.0 0.35 35,774 1.79 3.03 3.47 2.32 0.86 -0.15 1.29 0.49 0.84 0.94 4.68 Cash Acqs. 5.3 6.3 3.6 12.7 -2.2 2.5 1.7 1.2 2.6 -1.4 16.8 0.09 34,354 2.00 1.28 1.13 1.67 -0.30 0.52 2.00 1.82 4.22 -3.13 12.51 Other Uses 45.3 -11.7 -46.7 -49.6 -5.3 -3.3 -4.3 -1.7 -4.7 0.3 -64.3 0.09 36,138 2.90 -0.79 -3.55 -2.46 -0.23 -0.13 -2.16 -0.92 -2.95 0.46 -12.38 Panel B: More Financially Constrained EquityIssue EquityIssue×H H TobinQ Leverage t t−1 t−2 t t−1 t−2 t t−1 t−2 t−3 t−3 R−Sq. N Cash Change 28.6 -21.6 -4.7 15.0 6.0 -4.3 -1.4 -2.2 2.7 1.2 -3.9 0.27 20,443 5.48 -6.28 -2.43 1.78 1.28 -1.42 -1.29 -2.36 3.28 4.25 -4.54 Debt Payout 10.2 -0.6 1.9 -3.2 2.2 0.2 2.1 1.7 -0.4 -1.4 15.5 0.13 20,444 1.67 -0.70 0.93 -0.52 1.94 0.09 1.97 1.98 -0.37 -6.53 4.98 Equity Payout -0.2 -0.1 0.7 0.3 0.9 0.4 -0.1 -0.4 0.0 0.3 -1.0 0.07 20,407 -0.49 -0.72 1.95 0.57 1.74 0.74 -0.27 -2.02 0.18 5.37 -2.62 Capex 13.5 6.0 4.5 -3.5 1.7 -1.4 2.5 1.2 0.3 0.2 0.4 0.33 20,305 1.93 4.21 3.58 -0.50 1.01 -1.17 2.88 2.16 0.49 0.90 0.68 Cash Acqs. 22.1 -0.5 -1.5 0.0 3.6 2.2 1.2 1.5 0.4 -1.2 4.6 0.12 20,092 1.47 -0.18 -0.67 0.00 1.46 1.11 0.61 1.84 0.76 -2.33 3.20 Other Uses 25.9 16.8 -1.0 -7.7 -14.4 2.8 -4.4 -1.7 -3.0 0.9 -15.5 0.14 20,444 0.96 3.22 -0.21 -0.28 -2.41 0.62 -1.34 -1.00 -2.04 1.21 -4.00 111 Chapter 5 Conclusion In this thesis I study the interaction of two empirical regularities to shed light on what motivates firms to participate in merger and acquisition transactions and whether merger activity is an efficiency-enhancing activity. In the first essay I argue that the degree of market anticipation of M&As is linked in predictable ways to the timing of deal announcements in a merger wave. Although the consequences of anticipation are known to the literature, the inno- vation in the current work comes from combining this idea with another well- established empirical fact: M&A activity clusters by industry over time. I ar- gue that the existence of merger waves suggests a commonality in deal motivation which has implications for deal anticipation. In other words, I argue that merger waves are an important source of market anticipation. In the second essay I show that the key predictions implied by the information channel in the model are supported in the data. In particular, I find that the differ- ence in announcement returns between the first and the last deal in a merger wave is on average 1.6%, with a declining pattern. This pattern holds even after con- trolling for deal quality and several other determinants of acquirers’ announcement effects. Further, I find evidence of contagion effects from early bidders in a merger wave to other industry peers. This evidence holds after controlling for other poten- tial sources of contagion effects (e.g., product market). My evidence also indicates that the market potentially learns even when no announcement takes place. The first two essays systematically tie market anticipation, contagion and deal timing in the M&A context. The joint contribution of both essays indicates that the small and declining pattern in acquirers’ abnormal returns along merger waves can be partially rationalized by a measurement problem. More generally, the link 112 between market anticipation and deal timing in the context of merger waves raises a challenge for interpreting announcement returns as a measure of actual economic gain for acquirers. In the third and final essay, I address a different, albeit related, question: Why is M&A activity correlated with stock market valuations?. As one of the novelties in this chapter, I address this question by examining the dynamic allocation of equity proceeds conditional on the level of merger activity. The fact that merger activity peaks at times of high stock market valuations appears to support the view that equity overvaluation drives the time series patterns of M&A activity. However, my empirical results in the last essay indicate that, at times of high merger activity, managers do not behave as if they believe the stock is overvalued. My findings point to a direction unexplored in previous research. In particular, I find that equity raised at times of high merger activity is allocated in a way that suggests other motivations for equity issues at times of merger activity. In particular, managers behave as if they are minimizing the costs associated with frictions such as time- to-build and adverse selection. Thus, the second innovation in my third essay is to put forward and provide evidence for these two alternative explanations for the dynamics of merger waves. In line with the theme in the first two essays, I argue in the third essay that a better understanding of the underlying forces behind the clustering in M&A ac- tivity is also relevant for the interpretation of existing evidence based on (short and long run) event studies. For example, if acquirers’ equity overvaluation drives merger activity, the inference about value creation based on stock returns around or after deal announcements has limited value as it confounds the economic value of the deal and the correction in bidder overvaluation. In contrast, if frictions drive the dynamics of mergers over time, then any long-run abnormal returns observed in the data must come from a poorly defined performance benchmark. In other words, different explanations for merger waves imply measurement problems of a different nature. Further, different explanations for the observed market timing be- havior in merger activity imply different outcomes for deals conducted inside and outside a merger wave. Overall, my results in this thesis suggest that relying exclusively on stock re- turns to study acquirers’ performance is problematic. Recent papers have made 113 progress in this direction. For example, in order to uncover the magnitude of bid- ders’ gain in stock deals, Savor and Lu (2009) study the stock price reaction to deals that fail for exogenous reasons. Barraclough, Robinson, Smith, and Whaley (2013) use the nonlinearity of call options written on acquirer and target firms to disen- tangle the different sources of price changes during deal announcements. By the nature of the design, these studies are based on small samples. Circumventing the measurement problems in acquirers’ performance during merger announcements in general and during merger waves in particular is an interesting venue for future research. 114 Bibliography [1] Alti, A., J. Sulaeman, 2012. When Do High Stock Returns Trigger Equity Issues?. Journal of Financial Economics 103(1), 61-87. [2] Andrade, G., M. Mitchell and E. Stafford, 2001. New Evidence and Perspec- tives on Mergers. Journal of Economic Perspectives 15, 103–120. [3] Ang, J. and Y. Cheng, 2006. Direct Evidence on the Market-Driven Acquisi- tion Theory. Journal of Financial Research 29, 199–216. [4] Akdogu, E., 2009. Gaining a Competitive Edge Through Acquisitions: Evi- dence from the Telecommunications Industry. Journal of Corporate Finance 15, 99–112. [5] Akhigbe, A., S. Borde and A.M. Whyte, 2000. The Source of Gains to Targets and Their Industry Rivals: Evidence Based on Terminated Merger Proposals. Financial Management 29(4), 101–118. [6] Aktas, N., E. de Bodt and R. Roll, 2009. Learning, Hubris, and Corporate Serial Acquisitions. Journal of Corporate Finance 15(5), 543–561. [7] Aktas, N., E. de Bodt and R. Roll, 2011. Serial Acquirer Bidding: An Em- pirical Test of the Learning Hypothesis. Journal of Corporate Finance 17(1), 18–32. [8] Autore, D. and T. Kovacs, 2010. Equity Issues and Temporal Variation in Information Asymmetry. Journal of Banking & Finance 34, 12–23. [9] Asquith, P., R. Bruner and D. Mullins, 1983. The Gains to Bidding Firms from Merger. Journal of Financial Economics 11, 121–139. 115 [10] Bachmann, R. and C. Bayer, 2011. Investment Dispersion and the Business Cycle. NBER Working Paper 16861 [11] Baker, M. and J. Wurgler, 2000. The Equity Share in New Issues and Aggre- gate Stock Returns. The Journal of Finance 55(5), 2219–2257. [12] Baker, M. and J. Wurgler, 2002. Market Timing and Capital Structure. The Journal of Finance 57(1), 1–32. [13] Barberis, N., A. Shleifer, and J. Wurgler, 2005. Journal of Financial Eco- nomics 75, 283–317. [14] Barnett, W. and A. Serletis, 2008. Consumer Preferences and Demand Sys- tems. Journal of Econometrics 147(2), 210–224. [15] Barraclough, K., D. Robinson, T. Smith and R. Whaley. 2013. Using Option Prices to Infer Overpayments and Synergies in M&A Transactions. Review of Financial Studies 26(3), 695–722. [16] Barten, A. 1969. Maximum Likelihood Estimation of a Complete System of Demand Equations. European Economic Review 1, 7–73. [17] Becher, D., 2009. Bidder Returns and Merger Anticipation: Evidence from Banking Deregulation. Journal of Corporate Finance 15, 85-98. [18] Becher, D., H. Mulherin and R. Walking, 2012. Sources of Gains in Corporate Mergers: Refined Tests from a Neglected Industry. Journal of Financial and Quantitative Analysis 47(1), pp 57–89. [19] Becketti, S., 1986. Corporate Mergers and the Business Cycle. Economic Re- view 71, 13–26. [20] Ben-David, I., M. Drake and D. Roulstone, 2011. Acquirer Valuation and Ac- quisition Decisions: Identifying Mispricing Using Short Interest. Available at SSRN. [21] Berger, P., E. Ofek and I. Swary, 1996. Investor Valuation of the Abandon- ment Option. Journal of Financial Economics 42, 257287. 116 [22] Bessembinder, H. and F. Zhang, 2012. Firm Characteristics and Long-Run Stock Returns after Corporate Events. Available at SSRN. [23] Betton, S., E. Espen, T. Rex and K. Thorburn, 2012. Merger Negotiations with Stock Market Feedback. Available at SSRN. [24] Betton, S., E. Eckbo, and K. Thorburn, 2008. Corporate Takeovers. Hand- book of Corporate Finance: Empirical Corporate Finance 2 (Ch.15), 291– 430. [25] Billett, M. and Y. Qian, 2008. Are Overconfident CEOs Born or Made? Ev- idence of Self-Attribution Bias from Frequent Acquirers. Management Sci- ence 54, 1037–1051. [26] Bruner, R., 2004. Where M&A Pays and where it Strays: A Survey of the Research. Journal of Applied Corporate Finance 16, 63–76. [27] Bouwman, C., K. Fuller and A. Nain, 2008. Market Valuation and Acquisition Quality: Empirical Evidence. Review of Financial Studies 22(2), 633–679. [28] Bris, A. and C. Cabolis, 2008. The Value of Investor Protection: Firm Evi- dence from Cross-Border Mergers. Review of Financial Studies 21, 605-648. [29] Brusco, S., G. Lopomo, D. Robinson and S. Viswanathan, 2007. Efficient Mechanisms For Mergers And Acquisitions. International Economic Review 48(3), 995–1035. [30] Butler, A., J. Cornaggia, G. Grullon and J. Weston, 2011. Corporate Financ- ing Decisions, Managerial Market Timing, and Real Investment. Journal of Financial Economics 101(3), 666–683. [31] Cai, J., M. Song, and R. Walkling, 2011. Anticipation, Acquisitions, and Bid- der Returns: Industry Shocks and the Transfer of Information across Rivals. Review of Financial Studies 24(7), 2242-2285. [32] Carlson, M., A. Fisher and R. Giammarino, 2006. Corporate Investment and Asset Price Dynamics: Implications for SEO Event Studies and Long-run Performance. Journal of Finance 61, 1009–1034. 117 [33] Carlson, M., A. Fisher and R. Giammarino, 2010. SEO Risk Dynamics. Re- view of Financial Studies 23, 4026–4077. [34] Carow, K., R. Heron and T. Saxton, 2004. Do Early Birds Get the Re- turns? An Empirical Investigation of Early-Mover Advantages in Acquisi- tions. Strategic Management Journal 25, 563–585. [35] Chemmanur, T., I. Paeglis and K. Simonyan, 2009. The Medium of Exchange in Acquisitions: Does the Private Information of both Acquirer and Target Matter?. Journal of Corporate Finance 15(5), 523–542. [36] Chemmanur, T. and J. He, 2011. IPO Waves, Product Market Competition, and the Going Public Decision: Theory and Evidence. Journal of Financial Economics 101(2), 382–412. [37] Clarke R. and C. Ioannidis, 1996. On the Relationship Between Aggregate Merger Activity and the Stock Market: Some Further Empirical Evidence. Economics Letters 53(3), 349–356. [38] Dasgupta, S., T. Noe and Z. Wang, 2011. Where Did All the Dollars Go? The Effect of Cash Flows on Capital and Asset Structure. Journal of Financial and Quantitative Analysis 46(5), 1259–1294. [39] Datta, S., M. Iskandar-Datta and K. Raman, 2001. Executive Compensation and Corporate Acquisition Decisions. The Journal of Finance 56, 2299-2336. [40] DeAngelo, H., L. DeAngelo and R. Stulz, 2010. Seasoned Equity Offerings, Market Timing, and the Corporate Lifecycle. Journal of Financial Economics 95(3), 275–295. [41] Dittmar, A. and R. Dittmar, 2008. The Timing of Financing Decisions: An Examination of the Correlation in Financing Waves. Journal of Financial Economics 90(1), 59–83. [42] Dixit, A., and R. Pindyck, 1994. Investment Under Uncertainty. Princeton, NJ: Princeton University Press. 118 [43] Dong, M., D. Hirshleifer, S. Richardson and D. Teoh, 2006. Does Investor Misvaluation Drive the Takeover Market? Journal of Finance 61, 725–762. [44] Doukas, J., J. Guo, and B. Zhou, 2011. Hot Debt Markets and Capital Struc- ture. European Financial Management 17(1), 46–99. [45] Duchin, R. and B. Schmidt, 2013. Riding the Merger Wave: Uncertainty, Reduced Monitoring, and Bad Acquisitions. Journal of Financial Economics 107(1), 69–88. [46] Eckbo, E., 1983. Horizontal Mergers, Collusion, and Stockholder Wealth. Journal of Financial Economics 11, 241–273. [47] Eckbo, E., R. Giammarino and R. Heinkel, 1990. Asymmetric Information and the Medium of Exchange in Takeovers: Theory and Tests. Review of Financial Studies 3(4), 651–675. [48] Edman, A., I. Goldstein and W. Jiang, 2012. The Real Effects of Financial Markets: The Impact of Prices on Takeovers. The Journal of Finance 67(3), 933–971. [49] Eisfeldt, A and T. Muir, 2013. Aggregate Issuance and Savings Waves. Avail- able at SSRN. [50] Erel, I., B. Julio, W. Kim and M. Weisbach, 2012. Macroeconomic Condi- tions and Capital Raising. Review of Financial Studies 25(2), 341–376. [51] Fishman, M. 1989. Preemptive Bidding and the Role of the Medium of Ex- change in Acquisitions. The Journal of Finance 44(1), 41–57. [52] Fuller, K., J. Netter, and M. Stegemoller, 2002. What Do Returns to Acquir- ing Firms Tell us? Evidence from Firms that Make Many Acquisitions. The Journal of Finance 57, 1763–1793. [53] Gande, A. and C. Lewis, 2009. Shareholder-Initiated Class Action Lawsuits: Shareholder Wealth Effects and Industry Spillovers. Journal of Financial and Quantitative Analysis 44, 823–850. 119 [54] Gaspar, J., M. Massa and P. Matos, 2005. Shareholder Investment Horizons and the Market for Corporate Control. Journal of Financial Economics 76(1), 135–165. [55] Gartner, D. and A. Schmutzler, 2009. Merger Negotiations and Ex-Post Re- gret. Journal of Economic Theory 144(4), 1636–1664. [56] Golbe, D. and L. White, 1988. A Time-Series Analysis of Mergers and Ac- quisitions in the U.S. Economy. NBER Chapters, in: Corporate Takeovers: Causes and Consequences, 265-310. [57] Goel, A. and A. Thakor, 2010. Do Envious CEOs Cause Merger Waves?. Review of Financial Studies 23(2), 487–517. [58] Grenadier, S., 1999. Information Revelation Through Option Exercise. Re- view of Financial Studies 12, 95-129. [59] Greene, W., 2002. Econometric Analysis 5th edition, Prentice Hall. [60] Hadlock, C. and J. Pierce, 2010. New Evidence on Measuring Financial Con- straints: Moving Beyond the KZ Index. The Review of Financial Studies 23(5), 1909–1940. [61] Hansen, R., 1987. A Theory for the Choice of Exchange Medium in Mergers and Acquisitions. Journal of Business 60(1), 75–95. [62] Harford, J., 1999. Corporate Cash Reserves and Acquisitions. The Journal of Finance 54, 1969–1997. [63] Harford, J., 2005. What Drives Mergers Waves?. Journal of Financial Eco- nomics 77, 529–560. [64] Healy, P., K. Palepu and R. Ruback, 1992. Does Corporate Performance Im- prove after Mergers?. Journal of Financial Economics 31, 135–175. [65] Helwege, J. and N. Liang, 2004. Initial Public Offerings in Hot and Cold Markets. Journal of Financial and Quantitative Analysis 39(3), 541–569. 120 [66] Hoberg, G. and G. Phillips, 2010. Product Market Synergies and Competition in Mergers and Acquisitions: A Text-Based Analysis. Review of Financial Studies 23(10), 3773–3811. [67] Huang, J., J. Pierce and S. Tsyplakov, 2012. Post-Merger Integration Dura- tion and Leverage Dynamics of Mergers: Theory and Evidence. Available at SSRN. [68] Hsu, H, A. Reed and J. Rocholl, 2010. The New Game in Town: Competitive Effects of IPOs. The Journal of Finance 65, 495–528. [69] Ivashina, V. and B. Becker, 2011. Cyclicality of Credit Supply: Firm Level Evidence. Available at SSRN. [70] Jensen, M., 1986. Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers. American Economic Review 76, 323–329. [71] Jenter, D., 2005. Market Timing and Managerial Portfolio Decisions. The Journal of Finance 60, 1903–1949. [72] Jenter, D., K. Lewellen and J. Warner, 2011. Security Issue Timing: What Do Managers Know, and When Do They Know It?. The Journal of Finance 66, 413–443. [73] Jovanovic, B. and P. Rousseau, 2001. Mergers and Technological Change: 1885–1998. Vanderbilt University Department of Economics Working Papers 0116. [74] Jovanovic, B. and P. Rousseau, 2002. The Q-Theory of Mergers. American Economic Review 92(2), 198–204. [75] Khan, M., L. Kogan and G. Serafeim, 2012. Mutual Fund Trading Pressure: Firm-Level Stock Price Impact and Timing of SEOs. The Journal of Finance 67, 1371–1395. [76] Klasa, S. and M. Stegemoller, 2007. Takeover Activity as a Response to Time- Varying Changes in Investment Opportunity Sets: Evidence from Takeover Sequences. Financial Management 36, 19–43. 121 [77] Kim, W. and M. Weisbach, 2008. Motivations for Public Equity Offers: An International Perspective. Journal of Financial Economics 87(2), 281–307. [78] Korajczyk, R., D. Lucas and R. McDonald, 1991. The Effect of Information Releases on the Pricing and Timing of Equity Issues. Review of Financial Studies 4(4), 685–708. [79] Korajczyk, R., D. Lucas and R. McDonald, 1992. Equity Issues with Time- Varying Asymmetric Information. The Journal of Financial and Quantitative Analysis 27(3), 397–417 [80] Kothari, S. and J. Warner, 2008. The Econometrics of Event Studies. Hand- book of Corporate Finance: Empirical Corporate Finance 1 (Ch.1), 3–36. [81] Lang, L., R. Stulz and R. Walkling, 1991. A Test of the Free Cash Flow Hypothesis: The Case of Bidder Returns. Journal of Financial Economics 29, 315–336. [82] Lambrecht, B., 2004. The Timing and Terms of Mergers Motivated by Economies of Scale. Journal of Financial Economics 72, 41–62. [83] Lambrecht, B. and S. Myers, 2007. A Theory of Takeovers and Disinvest- ment. The Journal of Finance 62, 809–845. [84] Leeth, J. and J.R. Borg, 2000. The Impact of Takeovers on Shareholder Wealth during the 1920s Merger Wave. Journal of Financial and Quantitative Analysis 35(2), 217–238. [85] Lowry, M., 2003. Why Does IPO Volume Fluctuate so Much?. Journal of Financial Economics 67, 3–40. [86] Lucas, D. and R. McDonald, 1990. Equity Issues and Stock Price Dynamics. Journal of Finance 45(4), 1019–43. [87] Malatesta, P. and R. Thompson, 1985. Partially Anticipated Events: A Model of Stock Price Reaction with an Application to Corporate Acquisitions. Jour- nal of Financial Economics 14, 237-250. 122 [88] Malcolm, B. and J. Wurgler, 2000. The Equity Share in New Issues and Ag- gregate Stock Returns. Journal of Finance 55(5), 2219–2257. [89] Maloney, M., R. McCormick and M. Mitchell, 1993. Managerial Decision Making and Capital Structure. The Journal of Business 66, 189–217. [90] Maksimovic, V., G. Phillips and L. Yang, 2013. Private and Public Merger Waves. Journal of Finance forthcoming. [91] Makaew, T., 2011. Waves of International Mergers and Acquisitions. Avail- able at SSRN. [92] Masulis, R., C. Wang, F. Xie, 2007. Corporate Governance and Acquirer Re- turns. The Journal of Finance 62, 1851–1889. [93] McNamara, G., J. Haleblian and B. Dykes, 2008. The Performance Impli- cations of Participating in an Acquisition Wave: Early Mover Advantage, Bandwagon Effects, and Moderating Influence of Industry Chratacteristics and Acquirer Tactics. Academy of Management Journal 51, 113–130. [94] Moeller, S., F. Schlingemann and R. Stulz, 2004. Firm Size and the Gains from Acquisitions. Journal of Financial Economics 73, 201–228. [95] Moeller, S., F. Schlingermann and R. Stulz. 2005. Wealth Destruction on a Massive Scale? A Study of Acquiring-Firm Returns in the Recent Merger Wave. The Journal of Finance 60, 757–782. [96] Morellec, E. and A. Zhdanov, 2005. The Dynamics of Mergers and Acquisi- tions. Journal of Financial Economics 77(3), 649–672. [97] Morellec, E. and D. Hackbarth, 2008. Stock Returns in Mergers and Acqui- sitions. Journal of Finance 63, 1203–1242. [98] Mitchell, M. and H. Mulherin, 1996. The Impact of Industry Shocks on Takeover and Restructuring Activity. Journal of Financial Economics 41, 193–229. 123 [99] Murphy, K. and R. Topel, 1985. Estimation and Inference in Two-Step Econo- metric Models. Journal of Business & Economic Statistics 3(4), 370–379. [100] Myers, S. and N. Majluf, 1984. Corporate Financing and Investment Deci- sions when Firms Have Information that Investors Do Not Have. Journal of Financial Economics 13, 187–221. [101] Nelson, R., 1959. Merger Movements in American Industry, 1895–1956. Princeton University Press for NBER, 1959. [102] Netter, J., M. Stegemoller, and M. Wintoki, 2011. Implications of Data Screens on Merger and Acquisition Analysis: A Large Sample Study of Mergers and Acquisitions from 1992 to 2009. Review of Financial Studies 24(7), 2316–2357. [103] Pagan, A., 1984. Econometric Issues in the Analysis of Regressions with Generated Regressors. International Economic Review 25(1), 221–247. [104] Phalippou, L., F. Xu, and H. Zhao, 2012. Hunting the Hunters: New Evi- dence on the Drivers of Acquirers Announcement Returns in M&As. Avail- able at SSRN. [105] Pierce, J., S. Tsyplakov and J. Wang, 2012. Time-to-Build and Leverage Dynamics. Available at SSRN. [106] Polk, C. and P. Sapienza, 2009. The Stock Market and Corporate Investment: A Test of Catering Theory. Review of Financial Studies 22, 187–217. [107] Rhodes-Kropf, M., and S. Viswanathan, 2004. Market Valuation and Merger Waves. Journal of Finance 59, 2685–2718. [108] Rhodes-Kropf, M., D. Robinson, and S. Viswanathan, 2005. Valuation Waves and Merger Activity: The Empirical Evidence. Journal of Financial Economics 77, 561–603. [109] Savor, P. and Q. Lu. 2009. Do Stock Mergers Create Value for Acquirers?. The Journal of Finance 64, 1061–1097. 124 [110] Shahrur, H., 2005. Industry Structure and Horizontal Takeovers: Analysis of Wealth Effects on Rivals, Suppliers, and Corporate Customers. Journal of Financial Economics 76, 61–98. [111] Schipper, K., and R. Thompson, 1983a. Evidence on the Capitalized Value of Merger Activity for Acquiring Firms. Journal of Financial Economics 11, 85–119. [112] Schipper, K., and R. Thompson, 1983b. The Impact of Merger-Related Reg- ulations in the Shareholders of Acquiring Firms. Journal of Accounting Re- search 21(1), 184–221. [113] Shleifer, A. and R. Vishny, 2003. Stock Market Driven Acquisitions. Journal of Financial Economics 70(3), 295–311. [114] Singal, V., 1996. Airline Mergers and Competition: An Integration of Stock and product Price Effects. Journal of Business 69(2), 233–268. [115] Song, M. and R. Walkling, 2000. Abnormal Returns to Rivals of Acquisi- tion Targets: A Test of the “Acquisition Probability Hypothesis”. Journal of Financial Economics 55, 143–172. [116] Stein, J., 1996. Rational Capital Budgeting in an Irrational World. The Jour- nal of Business 69(4), 429–55. [117] Strarford, D., T. Garrett and R. Russell, 2009. Disallowances and Overcap- italization in the U.S. Electric Utility Industry. Federal Reserve Bank of St. Louis Review 91(1), 23–32. [118] Straham, P., 2003. The Real Effects of U.S. Banking Deregulation. Federal Reserve Bank of St. Louis July/August 111-116. [119] Thijssen, J., 2008. Optimal and Strategic Timing of Mergers and Acquisi- tions Motivated by Synergies and Risk Diversification. Journal of Economic Dynamics & Control 32, 1701-1720. 125 [120] Town, R., 1992. Merger Waves and the Structure of Merger and Acquisition Time-Series. Journal of Applied Econometrics 7, S83–S100. [121] Toxvaerd, F., 2008. Strategic Merger Waves: A Theory of Musical Chairs. Journal of Economic Theory 140(1), 1-26. [122] Tsyplakov, S. 2008. Investment Frictions and Leverage Dynamics. Journal of Financial Economics 89(3), 423–443. [123] Xu, T., M. Najand and D. Ziegenfuss, 2006. Intra-Industry Effects of Earn- ings Restatements Due to Accounting Irregularities. Journal of Business Fi- nance & Accounting 33, 696–714. 126

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0073897/manifest

Comment

Related Items