On the Post-merger Operating Performance of U.S. Companies

Transcription

1 On the Post-merger Operating Performance of U.S. Companies Bachelor Thesis in Corporate Finance Department of Banking and Finance University of Zurich Advised by Prof. Alexander F. Wagner, Ph.D. Ivan Petzev Submitted by Sascha Andreas Herzog Submission date January 31, 2013

2 Abstract This paper examines changes in post-merger operating performance of U.S. public firms involved in mergers and acquisitions between 1998 and 2008 and the relationship of these changes to the method of payment. I find that the abnormal operating performance of merging firms relative to control firms matched based on size, industry, and pre-merger performance increases significantly in the post-merger period relative to pre-merger levels and that the performance changes are related to the method of payment. Stock and mixed financed acquisitions exhibit significantly increased post-merger operating performance, while transactions financed solely with cash remain unaffected in terms of operating performance.

3 Contents 1. Introduction Hypotheses Development and Theoretical Background Summary of Prior Research Sample Selection and Research Design Sample & Descriptive Statistics Research Design The change and the intercept model Performance Measurement Benchmark Performance and Control Firm Characteristics The Relationship Between Post-merger Operating Performance Improvements and the Method of Payment Empirical Findings Empirical Results for the Changes in Abnormal Cash Flow Margins in the Change Model Analysis Empirical Results for the Relationship Between Post-merger Operating Performance Changes and the Method of Payment Summary Appendix A. Overview of the Relevant Prior Research Ordered by Their Findings Appendix B. List of all Mergers and Acquisitions in the Sample Appendix C. Change Model Analysis Excluding Extreme Performance Changes References... 48

4 1. Introduction During the last decade, renowned management consulting firms have not been reluctant to publish research reports which claim that the majority of the M&A deals are not profitable. For instance, The Boston Consulting Group recently reported that more than half of the M&A transactions among public firms destroy value for the acquiring firm in the long term. In 2003, they stated that even 64 % of the deals destroy value for the acquiring firm`s shareholders. McKinsey & Company regularly states that the fraction of M&A deals that fail is astonishingly high with failure rates reaching roughly 70 %. These reports have reinforced the conventional wisdom that the majority of all M&A transactions destroy value. But, what does it mean for M&A deals to fail or destroy value? What does academic research tell us about the success and profitability of M&A transactions? How do researchers measure the value created or destroyed by such transactions? The primary means for investigating the value creation or destruction of M&A transactions in academic research is the event study design. Many studies examine the abnormal stock market returns of companies involved in M&A deals surrounding the announcement of such transactions. The majority of these studies support the view that M&A transactions on average create value. In his review of evidence, Bruner (2002) lists a number of studies which show that combined returns, i.e., the weighted returns to acquiring and target firms` shareholders, are mostly positive and in about half of the studies statistically significant. However, these studies usually fail to provide evidence for a causal relationship between stock market gains and superior financial performance after the merger as Healy, Palepu, and Ruback (1992, p. 136) remark: These increases in equity values are typically attributed to some unmeasured source of real economic gains, such as synergy. But researchers have had very little success in relating the equity value gains to improvements in subsequent corporate performance. Therefore, the equity gains could also be due to capital market inefficiencies, arising simply from the creation of an overvalued security.. This calls for a closer examination of the post-merger operating performance. The present thesis updates the existing empirical research on post-merger operating performance changes by examining a recent sample of U.S. M&A transactions. Conceptually, my study is strongly influenced by the work of Ghosh (2001), Healy, Palepu, and Ruback (1992), and Barber and Lyon (1996). I examine the operating performance defined as operating cash flows divided by sales across the three years prior and subsequent to the merger completion year in a research design sometimes referred to as the change model. 1

5 Particularly, I investigate whether the operating performance changes in the post-merger period relative to pre-merger levels. To avoid biases introduced by economy-wide or sectorspecific factors, I adjust the operating performance of merging firms by the performance of corresponding control firms which are matched based on size, industry, and pre-merger performance. Furthermore, I examine the relationship between merger-induced changes in the operating performance and the method of payment 1. The whole empirical analysis is conducted with the statistical software package Stata 2. As hypothesized, I find that merging firms exhibit statically significant superior median abnormal operating performance (i.e., the performance compared to matched control firms) of 1.36 % in the first and 1.88 % in the second year after the transaction completion year. Relative to pre-merger levels, the abnormal operating performance significantly increases between 0.74 % and 0.97 % in the post-merger years. However, the empirical analysis yields results that contradict my second hypothesis that the changes in post-merger operating performance are stronger for cash transactions than for other payment methods. Contrary to the hypothesis, cash transactions do not exhibit significantly positive operating performance changes while stock and mixed consideration transactions do on average experience abnormal operating performance improvements between 2 % and 3.2 %. This study proceeds as follows. Section 2 presents hypotheses concerning the research questions at hand. Section 3 gives a comprehensive overview of the relevant previous research in this field. The sample used for the empirical analysis and the research design are described in section 4. Section 5 presents the results from the empirical analysis and section 6 summarizes the paper. 2. Hypotheses Development and Theoretical Background There exist many different motives to pursue mergers and acquisitions. Popular textbooks list among others accelerated expansion compared to organic growth, operating and financial synergy, diversification, stock market undervaluation of target companies, and tax considerations as potential motives for M&A transactions (Gaughan (2010)). In practice, 1 I distinguish between three different methods of payment. First, in stock transactions or stock-only transactions, the acquiring firm completely pays the consideration for the target firm with its own stock. Second, acquisitions can be financed solely with cash. I refer to them as cash transactions, independently of whether the acquiring firm has to increase its debt to raise enough cash to pay the total consideration for the target firm. Third, the acquiring firm pays for the target with a combination of both stock and cash, which I refer to as mixed consideration transactions. 2 The commands I use to edit the raw data, implement the sample selection criteria and matching procedure, and conduct the statistical hypothesis tests can be accessed in the documentation on the CD accompanying this paper. This ensures a simple reproduction of my results. 2

6 synergies are the most widely quoted rationale for conducting mergers and acquisitions (Damodaran (2002)). Various surveys of both scholars and practitioners support this view. Representatively, Mukherjee, Kiymaz, and Baker (2004) survey CFOs of companies involved in large mergers between 1990 and 2001 and find that synergy is the most important motive for acquisitions. The vast majority of the respondents in the survey declare operating economics as the prevalent source of the synergetic effect from the combination of their businesses. In contrast to these economic motives that eventually ought to materialize in one way or the other in improved financial performance of the combined entity, there exist motives for mergers and acquisitions that are not driven by managerial and economic prudence. Typical examples are managerial overconfidence made popular by Roll`s hubris theory (see Roll (1986) for more details on his hubris hypothesis) or a phenomenon known as empire building (see Volkart (2008)). Despite this wide range of different motives for M&A transactions, there is only one expedient way to assess the success of a M&A transaction from an economic standpoint, which is to investigate the value creation for shareholders associated with such transactions (Bruner (2002)). From this sober perspective, the question for a company`s management whether to participate in M&A activity is not much more than the evaluation of an investment opportunity for which standard corporate finance principles apply. In their seminal textbook, Koller, Goedhart, and Wessels (2010) outline a value creation framework for acquisitions in which they state: Acquisitions create value when the cash flows of the combined companies are greater than they would have otherwise been. If the acquirer doesn t pay too much for the acquisition, some of that value will accrue to the acquirer s shareholders. (Koller, Goedhart, and Wessels (2010, p. 446)). According to the authors, the value created for the acquiring firm`s shareholders (i.e., the difference between the value received from and the price paid for the target firm) can be expressed as 3 : ( ) (1) ( ) 3 Gaughan (2010) presents a similar formula, but he also incorporates the expense of the acquisition process which the operating performance improvements must compensate for in order to create value with the acquisition. These costs occur explicitly, for instance as professional fees for M&A lawyers, accountants, or investment bankers, or in the form of implicit costs, such as absorbed internal capacity due to the transaction or costs for the post-merger integration process. 3

7 The value the acquirer`s shareholders get from the acquisition is the intrinsic value of the target firm as a stand-alone company and the value of performance improvements (which mainly stem from realized operating synergies), minus the price they pay, which is composed of the current market valuation of the target firm and the acquisition premium (i.e., the difference between the current market valuation and the effective price paid for the target firm, which is often referred to as control premium ). Assuming that the stand-alone value of the target equals its market value, the formula (1) reduces to 4 : ( ) (2) The value created for the acquiring firm`s shareholders is the difference between the value of (operating) performance improvements and the acquisition premium. Thus, acquiring firms only create value with acquisitions if they manage to improve the operating performance enough to outweigh the acquisitions premium, which can be substantial. The historical control premium fluctuates roughly between 35 % and 50 % of the pre-acquisition market capitalization of the target firm in the years from 1978 through 2009, with peaks as high as 62 % shortly after the economic downturns in the early 2000`s and in the recent financial crisis, which marked the end of the fifth and sixth merger wave, respectively (Gaughan (2010)). Consistent with Gaughan (2010), Healy, Palepu, and Ruback (1997) find acquiring firms paying median acquisition premiums between 27 % and 45 % for different subsamples in a follow-up study on the their pivotal paper presented in section 3. Based on the facts and theoretical viewpoints presented so far in this section, it can be argued that the acquisition premium paid for a target firm creates pressure on the acquiring firm`s management to realize an often sizeable amount of synergies in order to create value with an acquisition (Gaughan (2010)) 5. However, performance improvements valued in the order of 50 % of the deal value are not impossible as Koller, Goedhart, and Wessels (2010) exemplarily show. A recent example of ambitious performance improvement expectations 4 The assumption that the market value of target firms equals its stand-alone value is not misguided as Koller, Goedhart, and Wessels (2005 p. 429) declare: Today`s market for corporate control is fairly efficient: Easy, good deals are hard to come by, if they exist at all.. 5 The focus of this study does not lie on the exact sources and forms of synergy, neither does the thesis examine whether the acquisition premiums are actually fully recovered in the post-merger period (see e.g., Healy, Palepu, and Ruback (1997) for a closer examination of this issue). I consciously abstract from the exact form of the synergies and I solely focus on operating performance changes associated with M&A transactions. For more details on synergies, see e.g., Gaughan (2010) and the related research. 4

8 associated with such transactions is Swatch`s acquisition of the Canadian jewelry and diamond mining company, Harry Winston. In January 2013, the Swiss watchmaker Swatch announced to acquire the jewelry and watch unit of Harry Winston for $ 750 million in cash and the assumption of $ 250 million of Harry Winston`s debt. Swatch`s CEO Nick Hayek was quoted to see the sales and net profit potential of the acquired unit at about $ 1.1 billion and $250 million in four to five years, while the figures for the financial year 2012 amount to $ 412 million and $ 19 million, respectively (Reuters (2013) and Harry Winston Diamond Corporation (2012)). Thus, I hypothesize that the abnormal post-merger operating performance of merging companies increases relative to pre-merger levels in an attempt of the acquiring firms` management teams to recover the acquisition premium paid for the target firms, and consequently, to create value for their shareholders 6. Furthermore, the realization of operating synergies is a widely cited rationale for mergers and acquisitions, so I would expect them to actually materialize in post-merger years. Even if the assumption of the efficiency of the market for corporate control does not hold and acquirers were able to buy a controlling interest in an undervalued firm, i.e., the market value of the target firm is below its standalone intrinsic value, acquirers still need to at least partly overcome the acquisition premium by operating performance improvements. This hypothesis is also consistent with the empirical finding that the target and acquiring firm`s combined market valuation often increase around the merger announcement (Bruner (2002)) 7. However, if M&A transactions are not followed by post-merger operating performance gains, this would bolster the notion that other motives than strictly economic play a major role in the M&A decision making process or to the proposition of managerial inability to realize post-merger synergetic performance gains. The second hypothesis concerns the relationship between post-merger performance increases and the method of payment. My hypothesis is that cash transactions perform better in terms of post-merger operating performance improvements relative to stock or mixed consideration transactions. The theoretical and empirical arguments in favor of this notion are manifold. 6 This requires that the acquiring firms in my sample also pay a considerable acquisition premium. More details on the empirical findings concerning the control premium in my sample of transactions can be found in section The relationship between announcement returns of merging firms and post-merger operating performance changes is emphasized by Andrade (2001, p. 114) who states: Operating performance studies attempt to identify the sources of gains from mergers and to determine whether the expected gains at announcement are ever actually realized. If mergers truly create value for shareholders, the gains should eventually show up in the firms cash flows.. 5

9 Linn and Switzer (2001) invoke the explanation offered by Fishman (1989) who develops a model of preemptive bidding which predicates that bidders offer cash when they have beneficial private information about the target firm indicating a high valuation, perhaps due to synergies, to discourage other bidders from competing for a target firm. Furthermore, acquirers often rely on unused debt capacity to raise cash to finance large transactions (Gaughan (2010)). Therefore, the superior post-merger operating performance of cash transactions may stem from the disciplinary effect of debt. Increased post-merger financial leverage leads to better investment decisions, and therefore to better post-merger operating performance (see Ghosh and Jain (2000) for more details and empirical evidence on leverage and M&A transactions). There are further reasons why cash transactions are expected to perform better. For instance, Ghosh (2001) referring to the empirical evidence of Loughran and Ritter (1997) argues that stock transactions signal a decline in future cash flow performance as this form of financing is often associated with a stock issue of the acquiring firm. Loughran and Ritter (1997) find that firms issuing stock in a seasoned offering underperform control firms in terms of operating performance. Another explanation is the signaling effect of stock transactions related to possible earnings management as outlined in section (Heron and Lie (2002)). 3. Summary of Prior Research While a vast body of research on abnormal stock market returns surrounding merger announcements exists, fewer studies examine the post-merger operating performance of merging firms (see Bruner (2002)). This section gives an overview of the relevant prior research on post-merger operating performance with a focus on U.S. studies and the methodologies employed by the different authors. I defer a thorough discussion of the different research designs which are commonly used to examine changes in operating performance accompanying M&A transactions to the next section. But, it is helpful to briefly outline the two main research designs utilized to analyze post-merger operating performance changes to better understand the results of previous studies with respect to the specific methodologies employed. First, researchers simply test changes in the operating profitability from pre-merger to post-merger years, usually adjusted for median industry or control firm performance. In the literature, this is commonly referred to as the change model. Second, some empirical studies employ a cross-sectional regression model to analyze industry-adjusted operating performance changes due to M&A transactions. This method is referred to as the intercept model. 6

10 Nevertheless, there exist other research designs beside the two prevalent methodologies. Mueller (1980), for instance, edits a collection of studies across seven countries (Belgium, Germany, France, the Netherlands, Sweden, UK, and the U.S.) that examine performance changes with a difference-in-differences approach similar to the change model. The edited studies of mergers that occur roughly between 1962 and 1972 find mixed, but usually insignificant results across the different countries. Other early studies of Ravenscraft and Scherer (1987) and Herman and Lowenstein (1988) utilize some kind of the change model. The former examine the profitability of firms acquired via tender offers between 1950 and 1977 using data from the U.S. Federal Trade Commission`s Line of Business surveys. They find that nine years on average after the transaction, 153 lines of business with a tender offer history have significantly lower returns on assets of 3.1 % compared to non-merging lines of business. But, when using other profitability measures such as operating income to sales and cash flows to sales, the statistical significance of the decreased post-merger performance does not sustain. The latter examine returns on equity and returns on total capital for a sample of 56 hostile tender offers initiated between 1975 and 1983 in the United States which eventually led to an acquisition. They find steadily increasing mean post-merger returns in the five years subsequent to the completion of the transactions for bids initiated between 1975 and But for bids initiated between 1981 and 1983, bidders experience decreased returns on equity in two years after the transaction. However, both studies suffer from several methodological drawbacks. For example, Ravenscraft and Scherer (1987) examine post-merger profitability from 1975 to 1977 while the transactions occur anywhere between 1950 and Further, their study focuses on accounting data of acquired lines of business, thereby ignoring that synergies from the transactions might as well be realized in other lines of business of the acquiring company. Besides of using incomplete data, Herman and Lowenstein (1988) utilize a very simple form of the change model in that they do not adjust their profitability measures by control group or median industry performance and therefore allow for general economy-wide and sectorspecific factors to affect their results. In an attempt to overcome the flaws of earlier studies, Healy, Palepu, and Ruback (1992) introduce several methodological improvements such as the systematic adjustment of the performance measures by a median industry benchmark, proper alignment of the analysis to the merger completion year by examining the five years prior and subsequent to the transaction year, and the use of a profitability measure that is immune to different methods of accounting for and financing of the transaction. In their empirical analysis, they examine the 7

11 50 largest U.S. mergers of public industrial firms completed between 1979 and mid Despite overall declining post-merger operating returns relative to pre-merger levels, they find that merging firms deliver superior post-merger operating returns relative to their nonmerging industry peers. In particular, their analysis reveals significant median superior operating returns of 2.8 % compared to industry benchmarks over the five years subsequent to the merger. Beside the methodological improvements mentioned above, they introduce the intercept model 9. Their intercept model estimates yield a significant 2.8 % increase in postmerger industry-adjusted cash flow returns relative to pre-merger levels. Many subsequent studies adopt the research design of the widely cited Healy, Palepu, and Ruback (1992) paper. For example, Andrade, Mitchell, and Stafford (2001) implement the intercept model with a cash flow to sales operating performance measure for a sample of roughly 2,000 mergers from 1973 to Consistent with Healy, Palepu, and Ruback (1992), their regression analysis yields a significant improvement in operating performance adjusted by a median industry performance of 1.07 % relative to pre-merger levels. Parrino and Harris (1999) and Moeller and Schlingemann (2005), too, replicate the research design of Healy, Palepu, and Ruback (1992). Parrino and Harris (1999) analyze a sample of 197 mergers among U.S. listed companies during the period from 1982 to 1987 using operating cash flows divided by the market value of assets as performance measure. They find evidence which is largely consistent with the replicated study for transactions after which the management is replaced within one year after the transaction. Moeller and Schlingemann (2005) analyze a sample of 296 acquisitions comprising the acquisition of 260 domestic and 36 foreign targets by U.S. public firms between 1985 and 1995 using both the change and the intercept model. In contrast to Healy, Palepu, and Ruback (1992) and Parrino and Harris (2005), their application of the change model yields significantly reduced industry-adjusted median operating returns in the five years subsequent to the transaction relative to the five years preceding the transaction for the cross-border sample and unchanged industry-adjusted returns for the domestic sample. Their intercept model analysis in the spirit of Healy, Palepu, and Ruback (1992) shows largely similar results, although with weaker significance. In another study based on the Healy, Palepu, and Ruback (1992) intercept 8 The alignment of the analysis with the merger completion year can be easily illustrated. Say, firm A announces to acquire firm B in the year 2005 and completes the transaction in the same year. The year 5 refers to the fifth year prior to the completion year, which is 2000 in this example. Year - 1 and year + 1 refer to the years 2004 and 2006, respectively. Hereafter, for my analysis of the three years prior and subsequent to the merger completion year, I will refer to certain years in the - / + X notation. I further detail the methodological improvement in section 4. 9 See section for a more detailed discussion of the change and the intercept model. 8

12 approach, Harford (2005) finds evidence that mean post-merger industry-adjusted net income to sales and return on assets measures are significantly deteriorated relative to pre-merger levels and that merger waves do not influence these findings. However, Ghosh (2001) spotlights several methodological flaws of the Healy, Palepu, and Ruback (1992) paper by criticizing the use of a median industry performance benchmark, i.e., the median operating performance of firms operating in the same industry as the sample firms that are involved in M&A transactions, to adjust for economy-wide and sector-specific performance trends and the use of the intercept model. He shows that, if merging firms outperform median industry firms in terms of operating performance in pre-acquisition years, the intercept model as employed by Healy, Palepu, and Ruback (1992) is likely to result in biased estimates for abnormal post-merger performance improvements. Instead, Ghosh (2001) employs the research design prescribed by Barber and Lyon (1996) to improve the econometric approach for analyzing post-merger operating performance. Thus, he uses operating cash flows to sales as his primary performance measure and control firms matched based on pre-acquisition performance, size, and industry as a performance benchmark in a change model setting. His analysis of 315 acquisitions completed during the period from 1981 through 1995 exhibits that the post-merger mean and median abnormal operating performance does not significantly improve in the three years subsequent to the merger relative to pre-merger levels. But, for the sake of comparability, Ghosh (2001) implements the analysis of Healy, Palepu, and Ruback (1992) in both the change and intercept model, using a median industry benchmark to adjust for industry- and economy-wide trends. His implementation of Healy, Palepu, and Ruback (1992)`s analysis with the change model does not yield any significant change in the post-merger industry-adjusted operating performance relative to pre-merger levels, while his replication of their intercept model exhibits a significant improvement in operating cash flow returns of 2.4 % 10. Linn and Switzer (2001) examine the changes in post-merger operating performance of 413 tender offers for U.S. listed companies that lead to a cash, stock, and mixed consideration transaction between 1967 and They find an increased median adjusted operating performance of 1.81 % for their overall sample with a change model analysis. But, their analysis exhibits considerable differences between methods of payment subsamples. Only the cash transactions experience a statistically significant median abnormal operating 10 However, depending on the nature of the out- or underperformance (i.e., whether the performance differences are temporary or persistent across the merger period), there might also exist biases in the change model. See section 4.2 for a thorough discussion of the methodological problems and the possible sources of biases in both the change and the intercept model. 9

13 performance improvement of 3.14 %, while the abnormal operating performance improvements for stock and mixed consideration transactions are not statistically significant. The same analysis with mean operating performance yields virtually the same results. However, unlike Ghosh (2001), they use a median industry benchmark to adjust operating performance and use a performance measure based on operating cash flows and the market value of assets. They also use some kind of the intercept model which, however, aims at further analyzing the differences in changes of post-merger operating performance between different methods of payment and yields results consistent with their findings from the change model analysis. Heron and Lie (2002) study operating cash flows to sales measures in a change model setting similar to the one employed by Ghosh (2001). They analyze merging firms` post-merger performance relative to both a median industry benchmark and the performance of control firms matched based on industry and pre-merger performance. Their analysis of a sample of 859 acquisitions between 1985 and 1997 yields that the post-merger performance of merging firms is significantly stronger relative to median industry firms and matched control firms. However, only the abnormal post-merger operating performance relative to firms matched based on industry and pre-merger performance significantly improves from pre-merger levels, which is in conflict with the findings of Ghosh (2001). Using a similar change model setting as Heron and Lie (2002), Fee and Thomas (2004) do not find evidence for increased median post-merger operating performance measured as cash flows to sales relative to firms matched based on industry, asset size, and pre-merger operating performance for a sample of 554 horizontal mergers of U.S. listed companies from 1980 to 1997, which is consistent with the findings of Ghosh (2001). However, certain subsamples, for instance mergers in concentrated industries, yield significantly increased median performance in the order of about 1 % for the three years subsequent to relative to the year prior to the transaction. In summary, the evidence from previous research on changes in post-merger operating performance following M&A transactions is ambiguous. The studies can be broadly speaking divided into three categories: studies that find improved post-merger operating performance (for example, Healy, Palepu, and Ruback (1992) and Andrade, Mitchell, and Stafford (2001)), studies that report deteriorated post-merger performance (for instance, Harford (2005)), and papers that document insignificant changes in operating performance 10

14 relative to pre-merger levels (e.g., Fee and Thomas (2004) and in his core analysis Ghosh (2001)) 11. However, the findings from the previous studies seem to be sensitive to the research design that is utilized. The choice of the research design (i.e., change vs. intercept model), the performance benchmark, and the performance measure might affect the outcome of empirical tests of post-merger operating performance changes 12. Although an obvious pattern does not seem to exist, many studies using median industry performance as a performance benchmark in an intercept model approach reveal post-merger operating performance improvements (e.g., Parrino and Harris (1999), Healy, Palepu, and Ruback (1992), Andrade, Mitchell, and Stafford (2001), and Ghosh (2001) in his replication of Healy, Palepu, and Ruback (1992)`s analysis), with most of them using cash flows to market value of assets as the performance measure. The empirical findings for the relationship between the method of payment and post-merger operating performance improvements are similarly mixed as for the performance improvements in general. Besides of Linn and Switzer (2001), whose findings on operating performance changes with respect to different methods of payment are outlined above, this relationship is examined by other authors including Healy, Palepu, and Ruback (1992), Ghosh (2001), Powell and Stark (2005), and Heron and Lie (2002). Consistent with the findings of Linn and Switzer (2001), Ghosh (2001)`s analysis indicates that cash offers are positively related to post-merger performance improvements, while the other researchers do not find any significant relationship between the method of payment and post-merger performance changes. 4. Sample Selection and Research Design This section describes the sample and the research design used for the empirical analysis of this paper. Particularly, I explain how the sample is constructed and I provide descriptive statistics for the transactions included in the sample in section 4.1. Section 4.2 and its subsections introduce as outlined in the previous section general research designs more thoroughly and describe the exact research design for this paper, which is based on the work of Ghosh (2001) and follows the recommendations of Barber and Lyon (1996). 11 As the prior research is ordered by research designs in this section, see table AI in appendix A for a complete overview of the prior studies disposed by their findings. 12 This sensitivity is very well illustrated in the study of Powell and Stark (2005) who examine operating performance changes associated with M&A transactions for a sample of 191 takeovers in the UK between 1985 and The studies of Ghosh (2001) and Heron and Lie (2002) also indicate that the empirical results are sensitive to certain model choices. 11

15 4.1 Sample & Descriptive Statistics I obtain M&A transaction data from Thomson Reuters Corporation`s SDC Platinum database. All transactions in my sample meet the following criteria: 1. The transaction is announced between January 1, 1997 and December 31, 2008 and is completed between January 1, 1998 and December 31, Both the target and the acquiring firm are U.S. publicly traded firms, i.e., the merging firms both have their primary listing at the New York Stock Exchange (NYSE), the NASDAQ, or the former American Stock Exchange (AMEX) at the time of the announcement of the transaction Both the target and the acquiring firm have the relevant consolidated accounting data available in the Compustat North America database. 4. The transaction has a value of at least $100 million and is either a cash, stock, or mixed consideration transaction. 5. The transaction has a material impact on the acquiring firm`s operations. Therefore, mergers in which the target has net sales of less than 10 % of the acquiring firm are excluded. 6. By the transaction, the acquiring firm obtains a majority interest in the target firm. Transactions in which the acquiring firm already owns a majority interest in the target firm at the time of the announcement of the transaction are excluded. 7. Neither the target, nor the acquiring firm is a financial or strongly regulated company. Following Parrino and Harris (1999), I exclude transactions that involve firms with the primary SIC codes , , or The acquiring firm does not buy another company within the three years prior to the completion year. I only include transactions among companies listed at U.S. stock exchanges due to financial data availability concerns. For the empirical analysis, I need accounting data for the years - 3 to - 1 for the target and the acquiring firm and for the years + 1 to + 3 for the acquiring firm (which then is the combined entity) from the Compustat North America database. To ensure that the transaction has a material impact on the acquiring firm`s operations, I use the target and acquiring firms` reported net sales for the twelve months before the transaction. 13 The American Stock Exchange (AMEX) was itself taken over by NYSE Euronext in 2008 and was renamed to NYSE Alternext U.S. (NYSE Euronext (2008a) and NYSE Euronext (2008b)). The former AMEX is as NYSE MKT LLC now part of the NYSE Euronext Group which also operates the New York Stock Exchange (NYSE Euronext (2012)). My search of the SDC database reflects these changes so that no transaction is omitted. 12

16 Therefore, all transactions for which net sales data are not available in the SDC database for either the acquiring or the target firm are excluded. Furthermore, all firms with the primary SIC codes , , or are excluded. This is mainly due to different accounting practices of financial firms (SIC codes ) which make such companies difficult to compare to other firms and due to the fact that transactions in heavily regulated industries (SIC codes and ) might not solely be driven by competitive motives (Parrino and Harris (1999)). I also exclude transactions for which the acquiring firm has multiple transactions within the same four year period (including the merger completion years) in the transaction sample obtained from SDC. With that, I avoid that the pre- and postmerger operating performance analysis for a transaction is distorted by another transaction made by the same acquiring firm. Section provides more details on the definition of my control firm universe for the matching procedure between sample and control firms. I obtain consolidated accounting data for the fiscal years from 1995 to 2011 from Standard and Poor`s Compustat North America database. To minimize the potential omission of accounting data due to matching obstacles between the Compustat and SDC Platinum databases, I search the Compustat database using GVKEY, CUSIP, and ticker symbols as company identifiers 14. GVKEY codes serve as my primary company identifiers and I complement my accounting data with search queries using 9-digit CUSIP codes and ticker symbols. From the 756 transactions initially obtained from the SDC Platinum, I find 242 M&A transactions that meet the sample formation criteria outlined above. Table I lists the number of transactions by completion year and provides an overview of their distribution across the methods of payment as reported by SDC Platinum 15. The transactions are not evenly distributed across the sample years. The sample period captures the peak of the fifth and the sixth merger wave around the year 2000 and the years 2006 and 2007, respectively 16. This is reflected in the number of deals per year reported in Table I. 14 Standard and Poor`s Compustat database uses so called GVKEY codes as proprietary company identifiers which do not change over time. However, SDC Platinum only reports 6-digit CUSIP codes for merging firms. I obtain GVKEY codes from 6-digit CUSIP codes by using the supporting tools available on the research platform Wharton Research Data Services (WRDS). Particularly, I translate 6-digit CUSIP codes into PERMNO codes (that are used in the CRSP database to identify a particular security) by using the CRSP tools provided by WRDS. As CUSIP codes for companies can change over time, the CRSP tools allow for matching of historical CUSIP to PERMNO codes. I then use the PERMNO codes to get GVKEY codes with the WRDS tools. However, to make sure that I get as complete accounting data as possible, I run Compustat queries using ticker symbols and 9-digit CUSIP codes and complement the accounting data I obtain with GVKEY codes. 15 Table BI in appendix B lists all transactions in the sample and provides more information on deal characteristics. 16 See Gaughan (2010), Andrade, Mitchell, and Stafford (2001), and Harford (2005) for more details on merger waves and related research. 13

17 Table I Distribution of Transactions by Deal Completion Year and Method of Payment The sample includes transactions among New York Stock Exchange (NYSE), American Stock Exchange (AMEX), or NASDAQ listed firms recorded in the SDC Platinum database with the relevant accounting data for both the target and the acquiring firm available on Compustat. Transactions must be announced between the years 1997 and 2008, inclusive, and completed between 1998 and 2008, inclusive, and have a transaction value of at least $ 100 million. A transaction is excluded if either the target or the acquiring firm has a primary SIC code of , , or and if the acquirer conducts multiple acquisitions within the same four year period. Furthermore, transactions in which the target firm has less than 10% of the acquiring firm`s net sales in the twelve months prior to the acquisition and transactions in which the acquiring firm already owns a majority interest in the target firm at the time of the transaction announcement are excluded. The payment method types describe how the acquiring firm pays the consideration for the acquisition of the target firm. This can either be done only with cash ("cash transactions"), only with stock of the acquiring firm ("stock transaction"), or with a combination thereof ("mixed consideration transaction"). All data displayed in this table and used to evaluate the sample selection criteria outlined above are obtained from the SDC Platinum database. Method of payment Year Deals in % Cash Stock Mixed Total in % of all transactions Although I exclude deals in the media, telecommunications, banking, insurance, and real estate sectors which account for a notable share of the merger activity in the fifth merger wave, there is a lot of deal activity before and in the early 2000s. 126 transactions (roughly half of all transactions in the sample) are completed between the years 1998 and 2001, inclusive. The sixth merger wave is not strongly pronounced in my merger sample, but the number of deals per year in my sample increases towards the end of the sample period. Over the entire sample period, a clear preference for one payment method does not seem to exist. About two fifths of all transactions are mixed consideration transactions. An example for mixed consideration transactions is Compaq`s acquisition of Digital Equipment in Compaq offered Digital Equipment`s shareholders own common shares and $ 30 in cash for every common share of Digital Equipment, totaling to an acquisition price of 14

18 roughly $ 9.6 billion (Compaq Computer Corporation (1998)). The remainder of the acquisitions is about equally often financed solely with cash or stock. However, while stock is the most popular payment method in the first years of the sample period, only a few transactions are stock transactions during the sixth merger wave. Given the all-time high in stock market valuations in 2007 and empirical evidence that acquiring firms are more likely to finance an acquisition with their own stock in times of high stock market valuations (see Shleifer and Vishny (2003) for an overview of the related research), this seems surprising at first glance since overvalued stock is often considered to be an attractive acquisition currency. Though, there are other effects at work such as increasing cash holdings of U.S. companies since the early nineties that encourage acquiring firms to use at least partly the low interest bearing excess cash on their balance sheet to finance an acquisition (Gaughan (2010)). Moreover, in times of high stock market valuations, it is not only the acquirer`s stock that becomes more expensive, both also the target firm`s stock. However, as my sample omits transactions involving firms from certain industries and as I impose other restrictive sample construction criteria, my sample is hardly representative for trends in the method of payment and the overall M&A transaction activity. Table II presents further descriptive statistics for the M&A sample. Panel A reports summary statistics on deal values as reported in SDC Platinum by the different methods of payment. On average, target firms in the sample are acquired for about $ 2.5 billion, however, the deal values across the transactions vary considerably. The largest deal in the sample is Procter and Gamble`s acquisition of Gillette in 2005, with a deal value of nearly $ 55 billion reported in SDC Platinum while the smallest deal barely exceeds the minimum deal value of $ 100 million as defined in the sample formation criteria. Across the different payment methods too, the mean and median deal values vary substantially. Mixed consideration deals have a mean deal value of $ 3.66 billion which is statistically significantly higher than the mean deal value of $ 1.42 billion of cash transactions and $ 2.08 billion of stock transactions on the 5 % significance level and on the 10 % level, respectively. This is confirmed by non-parametric tests for differences in medians, which are conducted out of concern that the mean transaction values are likely to be skewed upward by a few very large deals. As depicted in Panel B, I aggregate transactions into 48 industry categories as defined in Fama and French (1997) using target firms` primary SIC codes reported by SDC. About half of all transactions involve target firms from either the business services, electronic equipment, computers, petroleum and natural gas, or pharmaceuticals industry. 15

19 Table II Descriptive Statistics Panel A exhibits an overview of the total consideration paid by the acquiring firm for the target firm in $ million by the method of payment as reported in the SDC Platinum database. The payment methods refer to how the acquiring firm pays for the target firm. This can either be done only with cash ("cash transactions"), only with stock of the acquiring firm ("stock transaction"), or with a combination thereof ("mixed consideration transaction"). Panel B and C report the industry of the target and the acquiring firm. The industry definitions are based on primary SIC codes reported in the SDC database and on the Fama and French (1997) industry definitions. I report the five most frequent industries involved in the transaction, either as target firm industry (Panel B) or as acquiring firm industry (Panel C) and summarize firms from other industries under "Other". Furthermore, I report SIC codes for the five most frequent industries as defined in Fama and French (1997). Panel D exhibits whether the target and the acquiring firm are in the same industry, again, based on the Fama and French (1997) industry definitions. Panel E provides details on target and acquirer firm net sales using data recorded in the SDC Platinum database. Net sales are defined as the last reported net sales for the (last reported) twelve months prior to the transaction. Furthermore, based on the net sales data from SDC, Panel E reports details on the ratio of target to acquirer firm net sales. Method of payment Panel A: Total consideration paid for the target firm by method of payment Consideration paid for the target firm in $ million Deals in % Mean Median Standard Deviation Min. Max. Cash ' ' '565.0 Stock ' ' '906.8 Mixed ' ' ' '861.3 All transactions ' ' '906.8 Panel B: Frequency of deals by industry (target firm) Industry Deals in % SIC codes Business Services , 3993, , , 7397, 7399, , , Computers , , 3695, 7373 Electronic Equipment , , , , 3660, , 3699 Petroleum and Natural Gas , , Pharmaceutical Products Other All transactions Panel C: Frequency of deals by industry (acquiring firm) Industry Deals in % SIC codes Business Services , 3993, , , 7397, 7399, , , Electronic Equipment , , , , 3660, , 3699 Computers , , 3695, 7373 Petroleum and Natural Gas , , Pharmaceutical Products Other All transactions Panel D: Industry relatedness Industry relatedness Deals in % Firms are in the same Fama and French (1997) industry Firms are not in the same Fama and French (1997) industry All transactions

20 Table II (continued) Panel E: Net sales in the twelve month prior to the transaction Net sales in the last reported twelve months prior to the transaction Net sales Mean Median Standard Deviation Min. Max. Target firms 1' ' '393.0 Acquiring firms 3' ' ' '399.0 Ratio of target to acquirer firm net sales in % '604.2 The rest of the target firms involved in the transactions spread across 29 different industries (as I exclude transactions involving certain SIC codes, some industries are not represented in the sample). Panel C does the same for acquiring firms. The top five industries represented among the acquiring firms are the same as for the target firms, however in a slightly different order. Here too, about half of the transactions comprise acquirers from the top five industries, with the rest of the transactions conducted by firms scattered across 30 other industries. Thus, one could assume that most of the transactions are conducted among firms from the same industry. Panel D confirms this notion. Almost three quarters of all transactions are completed between firms from the same Fama and French (1997) industry. This is consistent with Andrade, Mitchell, and Stafford (2001), who find mergers and acquisition occurring in waves with a strong industry clustering driven by industry shocks that lead restructuring and consolidation in certain sectors. Panel E further describes the net sales of target and acquiring firms for the latest reported twelve months prior to the transaction. The acquirer firms are substantially larger in terms of net sales than the acquired firms. The average acquiring firms has sales of about $ 3.6 billion and target firms report mean net sales of about $1.5 billion in the twelve latest reported months prior to the acquisition. This difference is statistically significant on the 1 % level (with both a paired t-test for mean differences and a nonparametric Wilcoxon signed-rank test for median differences). Across all transactions, target firms report about 76 % of the acquirer`s net sales. However, the ratios of target firms` sales to acquiring firms` sales differ widely across the sample. For instance, in 2000 the software maker Veritas Software Corp. with net sales of approximately $ 600 million according to SDC Platinum acquired the hard drive producer Seagate Technology Inc. with a turnover in excess of $ 6.7 billion. However, deals with such an extreme ratio of target to acquirer net sales often involve special financing and deal structures (for example for tax purposes), and are sometimes (partly) backed by private equity investors as in the Veritas / Seagate deal (Silver Lake Partners (2000)). At the other end of the range, the target firm barely report more than the predefined 10 % of the acquiring firm`s sales. 17

21 While deriving my hypothesis about increasing merger-induced abnormal operating performance in section 2, I base my argumentation on a positive, often substantial control premium paid by the acquiring firm. For 234 mergers that have acquisition premium data available on SDC Platinum, I find that acquiring firms pay on average a control premium of % over the target firm`s closing price one day prior and a premium of % over the target firm`s closing price one week prior to the announcement date, which are both statistically different from zero at 1 % levels using a t-test. This is in line with the acquisition premiums reported by Gaughan (2010) and Healy, Palepu, and Ruback (1997). As argued in section 2, a control premium in this order of magnitude gives the management of an acquiring firm a strong incentive to increase the operating performance of the acquired firms or the combined entity, respectively. Furthermore, the abnormal combined stock market return for the target and the acquiring firm is on average 2.50 % between the closing price on the day prior to the closing price one day after the announcement date (day -1 to day +1 relative to the announcement date), which is statistically significantly greater than zero on the 1 % level using a one-sided t-test 17. This is consistent with the findings of most of the event study papers listed in Bruner (2002). Given that the capital markets are efficient, these stock market gains at the announcement date suggest that the capital markets expect increases in the future cash flow performance of the companies involved in the transactions (Andrade, Mitchell, and Stafford (2001)). 4.2 Research Design The change and the intercept model As outlined in the literature overview, two models are commonly used to analyze mergerinduced operating performance changes. First, the intercept model introduced by Healy, Palepu, and Ruback (1992) which estimates operating performance changes induced by M&A transactions as the intercept α of the following cross-sectional regression model: 17 To obtain the average combined stock market revaluation of the target and acquiring firm, I weight the announcement returns of the target and the acquiring firms by their respective market capitalizations on the day prior the transaction announcement date (based on stock closing prices and the number of fully diluted shares outstanding as reported in SDC Platinum). Consistent with commonly used research settings, I adjust the target and acquiring firms` share price reactions for the their expected returns for the days -1 to +1 relative to the transaction announcement date. I adjust the announcement returns by the returns of the S&P 500 Composite over the corresponding days (I obtain the times series data for the S&P 500 Composite closing prices from Thomson Reuter`s Datastream database). However, other researchers sometimes use other adjustment procedures, e.g., they use a CAPM approach to estimate expected returns for the target and acquiring firms. As this study is mainly on operating performance improvements, I refer to the studies listed in Bruner (2002) or the paper of Andrade, Mitchell, and Stafford (2001) for more details on the short-window event study approach for analyzing combined stock market returns surrounding the announcement of M&A transactions. 18

22 (3) where IACFR post,i is the median combined industry-adjusted cash flow return of the five postmerger years of the merging companies involved in transaction i. IACFR pre,i denotes the corresponding median combined industry-adjusted cash flow return of the five year premerger period for the same companies involved in transaction i. The slope coefficient β measures the persistence of industry-adjusted operating returns between pre- and post-merger years. Healy, Palepu, and Ruback (1992) argue that the regression intercept α (i.e., the change in operating performance) is independent of the pre-merger performance. Other researchers employ slight variations of the intercept model by using different performance measures or by using median performances over shorter pre-and post-merger time periods. For example, Harford (2005) examines net income divided by sales instead of cash flow returns or Ghosh (2001) analyzes median abnormal performance over three year pre- and post-merger periods instead of five year periods. Second, the change model directly compares changes in operating performance between pre- and post-merger years. Usually, researchers test the difference in operating performance changes between the post- and pre-merger period relative to a control group matched based on some characteristics, such as industry, premerger performance, and / or size. The change model analysis can either be conducted for specific years (e.g., the change from year - 1 to year + 1, from year - 1 to year + 2, etc.) or for mean or median values taken from the entire pre- and post-merger period under consideration. However, Ghosh (2001) demonstrates that under certain assumptions the intercept model can result in biased estimates of merger-induced abnormal performance changes. Namely, if merging companies outperform the performance benchmark in pre-merger years due to permanent factors (which persist into the post-merger period) and / or due to temporary factors (which do not persist into the post-merger period), the intercept model is likely to yield biased estimates of the intercept term α. The existence of superior pre-merger performance of the sample firms relative to the performance benchmark solely due to permanent factors suffices to yield upward biased results from the intercept model while the results of the change model remain unaffected 18. The empirical findings of Ghosh (2001) using median industry performance and cash flow to assets might, interestingly, suggest 18 See Ghosh (2001) for the complete argumentation including the assumptions. 19

23 exactly the effects described above. For merging firms with superior performance relative to an median industry benchmark, he does not find any industry-adjusted operating performance changes with the change model, however, his replication of the analysis of Healy, Palepu, and Ruback (1992), i.e., the intercept model, yields significant positive industry-adjusted operating performance. Heron and Lie (2002), and Andrade, Mitchell, and Stafford (2001) also find significant superior operating performance for merging firms over median industry benchmarks in pre-merger years. The divergent findings for the change and the intercept model are consistent with the empirical results of Powell and Stark (2005). There are several reasons to believe that merging firms outperform median industry firms in pre-merger years. First, differences in operating performance in pre-merger years could be due to differences in size between the firms involved in M&A transactions and industry median firms. According to economic theory, the difference in size between large merging firms and their smaller industry median counterparts could cause systematic pre-merger performance differences between merging and industry median firms (Ghosh (2001)). Furthermore, companies are inclined to conduct acquisitions during periods of strong performance and therefore tend to outperform industry peers due to temporary factors (Ghosh (2001) and Powell and Stark (2005)). Independently of the persistence of the superior premerger operating performance into the post-merger period, these results suggest that the findings of the intercept model in the respective studies are likely to be biased. The change model is according to the argumentation of Ghosh (2001) and Powell and Stark (2005) less likely subject to biases due to pre-merger performance differences between sample and control firms. Even so, one should not be lulled into a false sense of security when using the change model rather than the intercept model. The analysis framework for the two different models of Ghosh (2001) suggests that temporary pre-merger performance differences between sample and control firms introduce possible biases to both, the change and the intercept model analysis. Generally, the test for changes in post-merger operating performance needs to carefully control for pre-merger performance and the choice of the performance benchmark thereby is crucial (Barber and Lyon (1996)) Performance Measurement I use the ratio of operating cash flows to sales as a measure of operating performance for merging firms and their non-merging counterparts. I follow Ghosh (2001) and Healy, Palepu, and Ruback (1992) and define operating cash flows as sales (Compustat item SALE), minus 20

24 cost of goods sold (Compustat item COGS) and selling, general, and administrative expense (Compustat item XSGA), plus depreciation and amortization (Compustat item DP) and goodwill amortization expense (Compustat item GDWLAM) 19. I scale the operating cash flows by sales (Compustat item SALE) of the corresponding fiscal year. There are good reasons for using operating cash flows instead of accounting earnings to calculate operating performance measures. According to Healy, Palepu, and Ruback (1992), operating cash flows remain unaffected by two factors that impact accounting returns and make them hardly comparable over time and cross-sectionally. First, accounting earnings are impacted by the method of payment which can be easily illustrated. Suppose that, in order to fully pay for an acquisition in cash, an acquiring firm has to raise a considerable amount of debt. While future accounting earnings decrease due to higher interest expense for the debt raised and lower interest income due to lower post-merger cash holdings, operating cash flows remain unaffected by the lower net interest income. In contrast, if the acquisition is fully financed with newly issued stock, this is (at least theoretically) neither reflected in the future accounting earnings nor in the future operating cash flows. Second, accounting earnings are affected by the method of accounting for the transaction, while operating cash flows are constant across different accounting methods. International accounting standards had known the longstanding distinction between the pooling accounting and purchase accounting method. In short, under the purchase method the assets and liabilities of a target company were restated at their fair values and the goodwill (i.e., the difference between the acquisition price and the fair value of the assets and liabilities of the target) was capitalized in the balance sheet of the acquiring firm and amortized over time. In contrast, the pooling method (sometimes also referred to as pooling of interests accounting ) did not allow for the revaluation of the acquired assets and liabilities and the capitalization of goodwill. Thus, accounting earnings used to be lower when the purchase rather than pooling accounting method was used due to higher depreciation expense and goodwill amortization under the purchase method, while operating cash flows are the same across the different accounting treatments of mergers and acquisitions. In the early 2000s, however, international accounting standard setters started to conceive new directives concerning accounting for mergers and acquisitions. Namely, the Financial Accounting Standards Board (FASB) released SFAS 141 Business Combinations, which became effective for business combinations initiated as of July 1, 2001 and the International 19 This definition of operating cash flows ignores cash flows associated with changes in the working capital, but for the sake of comparability I adopt the definition of Ghosh (2001) and Healy, Palepu, and Ruback (1992). 21

25 Accounting Standards Board (IASB) introduced IFRS 3 Business Combinations as a corresponding standard on merger accounting, which became effective for mergers agreed upon on or after March 31, 2004 (Bonham et al. (2004) and Delaney et al. (2001)). Together with updated rules on intangible assets and goodwill (SFAS 142 and IAS 38), and the impairment of assets (SFAS 144 and IAS 36), they require acquiring firms to use purchase accounting and capitalize goodwill, which is no longer amortized, but instead, subject to regular impairment tests (Hachmeister (2008)) 20. This change in paradigm in merger accounting that falls into the sample period makes it even more important to use a performance measure that is not impacted by the different methods of accounting. Besides the effect of accounting and financing decisions, earnings figures might also be subject to deliberate manipulation, especially around major corporate events such as initial public offerings and mergers (Powell and Stark (2005)) 21. I follow Ghosh (2001) and Barber and Lyon (1996) and I do not scale operating cash flows by the market value of assets as Healy, Palepu, and Ruback (1992), but with sales instead 22. This choice too, is mainly because of the two factors the method of accounting for and financing of mergers and acquisitions as explained above. Furthermore, there exists empirical evidence suggesting that the share price of merging firms declines systematically in the three to five years after the merger (see specific studies on long-term stock market performance of merging companies such as Agrawal, Jaffe, and Mandelker (1992)) 23. This indicates that operating performance ratios based on market value of assets as denominators are systematically driven upward in post-merger years, without that merging companies deliver improved cash flows (Ghosh (2001)). This is consistent with the observation in section 3 indicating that empirical studies using a performance benchmark based on market values of assets seem more likely to find improved post-merger operating cash flow performance. As Ghosh (2001), I compute cash flow margins for merging firms and control firms for the three years prior (years - 3 to - 1) and the three years subsequent to the merger completion year (years + 1 to + 3). I exclude the year in which the merger is completed due to different accounting and consolidation procedures associated with the purchase and the pooling 20 Generally, the book of Ballwieser, Beyer, and Zelger (2008) offers a thorough treatise of merger accounting under IFRS and US GAAP. I refer to their book for more information on the accounting treatment of mergers and acquisitions. 21 There is evidence that the earnings of acquiring firms are managed upward in pre-merger periods (see Erickson and Wang (1999) and Louis (2004)). 22 For simplicity, I refer to this performance measure as cash flow margin hereafter. 23 Bruner (2002), too, lists a few long-term stock market return studies on merger and acquisitions. 22

26 method and possible one-time merger costs (Healy, Palepu, and Ruback (1992)). For the premerger years, I aggregate operating cash flows and sales figures of the acquiring and the target firm to obtain pro forma pre-merger cash flow margins of the combined entity. This is equivalent to the sales-weighted average cash flow margins of the acquiring and the target firms that Andrade, Mitchell, and Stafford (2001) use Benchmark Performance and Control Firm Characteristics As outlined in earlier sections, some of the cash flow margin variation of merging firms in the observation period might be due to external factors such as economy-wide or sectorspecific trends in operating profitability. This is also true for other major corporate events such as equity offerings or divestitures. In order to isolate the performance changes due to such major corporate events, researchers often compare the post-event operating performance of sample firms relative to their expected performance in absence of such an event. They usually do so by constructing a performance benchmark that serves as a proxy for the performance of companies involved in corporate events as if they were not involved in these events. The performance benchmark which tracks changes in the macro environment is calculated with performance data of companies not affected by the event in focus. This procedure is sometimes referred to as developing a model of expected performance (Barber and Lyon (1996)). There are different ways to construct a performance benchmark for sample firms involved in M&A transactions. First, researchers form a control group of non-merging firms in a matching procedure based on some pre-merger characteristic such as performance, industry or size. Second, researchers use median operating performance of same industry firms as a performance benchmark 24. To prevent the introduction of potential biases as described in section 4.2.1, I follow the prescriptions of Barber and Lyon (1996) and choose a benchmark that incorporates premerger performance of same two-digit SIC code firms with a similar size as the sample firms to form a control group with appropriate pre-merger characteristics. The assumption behind the use of control firms matched based on pre-merger performance is that they exhibit similar pre-merger performance characteristics i.e. a similar proportion of permanent and temporary abnormal performance components and similar mean-reversion patterns as the 24 The difference in the operating performance of merging firms relative to control firms matched based on some pre-merger characteristic described as the first method of constructing benchmark performance is referred to as abnormal performance a term I introduce in the introduction and use throughout the whole study. I refer to the performance relative to a median industry benchmark as industry-adjusted performance. 23

27 sample firms (Ghosh (2001)). Thus, I follow the first approach for developing a model of expected performance as described above. More precisely, I use a performance benchmark constructed similarly as in Ghosh (2001) and Fee and Thomas (2004). The matching procedure for the benchmark construction is based on Loughran and Ritter (1997) and is described in the following. To begin with, I form a control firm universe with companies that serve as matching candidates for the target and acquiring firms in my transaction sample. The control firm universe includes all firms tracked by the Compustat North America database with the same possible SIC codes as the firms included in the merger sample, i.e., all firms excluding those with SIC codes , , or From this set of potential control firms, I exclude companies that are a target or an acquiring firm in the merger sample. The exclusion of the sample firms from the control firm universe follows Fee and Thomas (2004). These steps yield a control firm universe of approximately 12`150 companies covered in the Compustat North America database. For every transaction in the sample, I identify firms from the control firm universe having the same two-digit SIC code with performance data available for the years - 3 to - 1 and for the years + 1 to + 3 for both the target and the acquiring firm. Then, I filter for control firms having sales between 25 % and 200 %, and cash flow margins between 90 % and 110 % in the year prior to the merger completion year (i.e., year - 1) for both the target and the acquiring firm. From this set of potential control firms, I select a firm for both the target and the acquirer company with the operating cash flow margin closest to that of my sample firms in the year - 1. If there are no control firms available that meet these criteria, I ease the matching criteria and allow for matching between sample firms and control companies based on the same one-digit SIC codes. Yet, if then still no control firm is available for matching, I entirely drop the requirement of the joint industry classification and match firms solely based on size and pre-merger performance. I aggregate operating cash flows and sales of the control firms in the three year pre- and postmerger period in the same manner as I do to obtain pro forma pre-merger operating margins for the sample firms in the pre-merger years. Given the pro-forma pre-merger cash flow margins of the sample firms, the post-merger performance of the merged entity, and the corresponding operating margins for the control firms, I compute abnormal operating cash flow margins for the pre-merger years - 3 to - 1 the post-merger years + 1 to + 3 by subtracting the operating margins for the control firms from the corresponding margins of the sample firms. 24

28 4.2.4 The Relationship Between Post-merger Operating Performance Improvements and the Method of Payment As hypothesized in section 2, there likely exists a relationship between the method of payment and post-merger operating performance changes. To examine this relationship, I estimate the following regression model which is virtually the same as found in Ghosh (2001): ( ) ( ) (4) where β 1, β 2, and β 3 measure the merger-induced abnormal cash flow margin improvements for cash, stock, and mixed consideration transactions. For every acquisition i in the merger sample for which the matching procedure yields control firms for both the target and acquiring firm, I calculate Δ(Cash flow margin). The term Δ(Cash flow margin) is equal to the difference between PACFM post and PACFM pre, i.e., the difference between post-merger abnormal cash flow margins and corresponding pre-merger figures. Thus, Δ(Cash flow margin) i denotes the merger-induced abnormal cash flow margin changes in the post-merger period relative to pre-merger levels for every transaction i 25. In the regression model with the regression equation (4), these changes in abnormal operating margins are regressed on the dummy variables CASH, STOCK, and MIXED. The dummy variables are 1 if the transaction i has the corresponding payment method, i.e., is a cash, stock, or a mixed consideration transaction and 0 otherwise. As in Ghosh (2001), the regression model does not have an intercept term. Furthermore, I estimate the regression model with additionally introducing the industry relatedness dummy SAMEINDUSTRY, which takes a value of 1 if both the target and the acquiring firm in transaction i have the same Fama and French (1997) industry classification and 0, otherwise. This industry relatedness dummy corresponds to the analysis depicted in panel D of table II presenting descriptive statistics on the transaction sample in section 4.1. This regression model is expressed by the regression equation (5): ( ) ( ) (5) 25 The terms PACFM pre and PACFM post are equal to the terms (MRG i - MAT i ) pre and (MRG i - MAT i ) post in table III from section 5.1. Thus, the term Δ(Cash flow margin) i from the regression equation (4) is equal to the corresponding terms found in table III in section

29 where β 4 additionally measures the impact of the industry relatedness of the target and the acquiring firm on abnormal operating performance changes. As the term (MRG i - MAT i ) pre is calculated in two different ways in the change model as depicted in table III, I conduct four regression model estimations in total. First, I use the abnormal cash flow margin changes from the year - 1 to the median of the years + 1, + 2, and + 3 to estimate the regression equations (4) and (5) which are denoted as Regression (1) and Regression (2) in table IV in section 5.2, which shows the empirical results for this analysis. Further, Regression (3) and Regression (4) are obtained by regressing the changes in abnormal cash flow margins from the median of the years - 1, - 2, and - 3 to the median of the cash flow margins in the years + 1, + 2, and + 3. The results from the regression models are presented in table IV in section Empirical Findings I report the results from the empirical analysis presented in the previous sections. Furthermore, I discuss the findings and reconcile them with the hypotheses formulated in section 2. If my empirical findings contradict the hypotheses, I provide possible alternative explanations for the observed phenomena. Section 5.1 presents the results from the change model analysis and section 5.2 exhibits the findings from the regression models. 5.1 Empirical Results for the Changes in Abnormal Cash Flow Margins in the Change Model Analysis The matching algorithm described in section yields 208 control firms for the target firms and 200 control firms for the acquiring firms in my sample with matching based on size, pre-merger performance, and two-digit SIC codes. When I ease the industry restriction and allow for matching between companies with the same one-digit SIC codes, I get further 26 and 36 control firms for the merger sample target and acquiring firms, respectively. Another 5 control firms for target firms and 4 control firms for acquiring firms are found by matching without any industry restriction. For 3 target and 2 acquiring firms, the matching algorithm does not deliver a control firm. As the failure of the matching for the target and acquiring firms does not overlap in a single transaction, totally 237 acquisitions are available for the change model analysis 26. Compared to Fee and Thomas (2004), I get more matches (roughly 84 % compared to 71 % of all sample firms) based on two-digit SIC codes, while 26 In the list of all sample transactions in table BI in appendix B, transactions for which the matching algorithm does not yield a control firm for either target or the acquiring firm are marked. 26

30 they go a step further by easing the matching criteria allowing for matching only based on pre-merger performance (i.e., without size restriction), so that all firms for which they seek a control firm are matched eventually. Table III exhibits the results from the change model analysis for the 237 transactions for which the matching was successful for both the target and the acquiring firm. Panel A separately shows the development of the mean and median combined cash flow margins of merging firms in the sample and corresponding matched firms. In the pre-merger years - 3 to - 1, merging and matched firms have very similar median operating cash flow margins. Merging firms experience a median combined cash flow margin of % in year - 3, % in year - 2, and % in year - 1, while matched firms have corresponding median cash flow margins of %, %, and %. This is reflected by the mean and median differences in combined cash flow margins between merging firms and matched firms across all transactions in the column Difference (MRG i MAT i ) which do not differ significantly in pre-merger years. This is not surprising as the merging and control firms are matched based on pre-merger performance. In the post-merger years, the mean and median combined operating margins of merging and matched firm begin to differ. In the year + 1, merging firms have a mean cash flow margin of % and matched firms have a cash flow margin of %. In the year + 2 the figures are % versus %, and in year % versus %. This is also very well reflected in the mean abnormal operating margins of the merging firms between merging firms and matched firms. The mean difference in the operating performance between merging and matched firms is 7.45 % in year + 1 and 6.54 % in year + 2. The mean difference for the year + 1 is statistically significant at the 1 % level using a matched pairs two-sided t-test and significant at the 5 % level in year + 2. The median difference in operating performance in the years + 1 and + 2 is 1.36 % and 1.88 %, which are both statistically significant at the 1 % significance level using the non-parametric matchedpairs Wilcoxon signed-rank test. The mean and median difference in operating performance between merging and matched firms in year + 3 are also positive, however not statistically significant. However, the mean cash flow margins in the post-merger years and the mean changes in operating performance from pre-merger to post-merger levels are misleading as the mean values are likely skewed upward by some mergers in the sample with extreme cash flow performance in the post-merger period. 27

31 Table III Cash Flow Margins Relative to Matched Firms Operating cash flows are defined as sales, minus cost of goods sold and selling, general, and administrative expense, plus depreciation & amortization and goodwill amortization expense. The operating cash flows are divided by the sales which results in the cash flow margins. Firms from the control firms universe, which includes all firms tracked in the Compustat North America database except for the firms involved in the transactions in the sample and firms with the SIC codes , , or , are matched to sample firms based on the cash flow margins and size (in terms of sales) in the year - 1, and based on their respective industries. The industry restriction is eased from two-digit SIC codes to one-digit SIC codes and completely dropped each time the matching procedure on the higher level does not yield a control firm for a sample firm. The matching procedure is based on Barber and Lyon (1996) and on the matching algorithm described in Loughran and Ritter (1997). Cash flow margins are reported for the combined sample and matched firms based on aggregated operating cash flows and sales. Particularly, pro-forma operating cash flow margin figures of merging firms in the years - 3 to - 1 are created by aggregating acquiring and target firms operating cash flow and sales data. The post-merger operating cash flow margins of the merged entity in the years + 1 to + 3 are equal to the corresponding figures of the acquiring firm. Pro-forma operating margins of matched firms for the pre- and post-merger years are formed by aggregating the data of the two matched firms in the same manner as for the pro-forma pre-merger figures of the merging firms. The accouting data are obtained from Standard and Poor`s Compustat North America database. Each transaction (the analyzed sample for this table includes 237 transactions for which the matching procedure yields control firms for both the target and the acquiring firm) is denoted by the subscript i. Panel A: Mean and median cash flow margins differences across the years - 3 to - 1 and + 1 to + 3 Years relative to the transaction completion year Merged firms (MRG) Matched firms (MAT) Difference (MRG i - MAT i ) Mean Median Mean Median Mean Median % 18.27% 20.04% 18.32% -0.10% 0.04% % 18.46% 21.41% 18.59% -1.50% -0.14% % 20.18% 23.38% 20.09% -0.01% 0.07% % 19.95% 20.79% 18.43% 7.45% *** 1.36% *** % 19.84% 21.42% 18.83% 6.54% ** 1.88% *** % 19.55% 21.89% 19.11% 1.73% 0.30% Panel B: Mean and median abnormal cash flow margin changes from year - 1 relative to the median of the years + 1 to + 3 Basis for change model comparison (MRG i - MAT i ) pre : year - 1 (MRG i - MAT i ) post : median of years + 1, +2, and + 3 Δ(Cash flow margin) i = (MRG i - MAT i ) post - (MRG i - MAT i ) pre Difference (MRG i - MAT i ) Mean Median -0.01% 0.07% 6.39% *** 1.38% *** 6.40% *** 0.97% *** Panel C: Mean and median abnormal cash flow margin changes from the median of years - 3 to - 1 relative the median of the years + 1 to + 3 Difference (MRG i - MAT i ) Basis for change model comparison Mean Median (MRG i - MAT i ) pre : median of year - 3, - 2, and - 1 (MRG i - MAT i ) post : median of years + 1, + 2, and + 3 Δ(Cash flow margin) i = (MRG i - MAT i ) post - (MRG i - MAT i ) pre -0.47% -0.06% 6.39% *** 1.38% *** 6.86% *** 0.74% *** * denotes significance at the 10 % level for a two-tailed test. ** denotes significance at the 5 % level for a two-tailed test. *** denotes significance at the 1 % level for a two-tailed test. Mean differences in operating performance between merging firms and matched firms are tested with a t-test, median differences with a Wilcoxon signed-rank test (both for matched-pairs) A brief look at the accounting data compiled from Compustat suggests that some merging firms especially from the software and IT sector make considerable writedowns that inflate operating cash flows as calculated using the definitions of Ghosh (2001) and Healy, Palepu, and Ruback (1992). This, however, does not come as a surprise because the sample period spans over the dot-com bubble which was accompanied by the fifth merger wave and followed by the recession in 2001 (Gaughan (2010)). I report the same chart as table III in 28

32 appendix C (table CI), however, I conduct the change model analysis excluding extreme observations in terms of changes in cash flow margins. I define cash flow margin changes, i.e., Δ(Cash flow margin) i, in a transaction i as extreme if they are outside of the range from the first quartile minus 1.5 times the interquartile range (defined as the third minus the first quartile) to the third quartile plus 1.5 times the interquartile range 27. Qualitatively, though, the results from the change model analysis in appendix C do not change and the results remain the same (although with slightly decreased significance levels). Panel B and panel C are the core of my change model analysis. They show the mean and median changes in abnormal combined cash flow margins between pre- and post-merger years denoted by Δ(Cash flow margin) i across all transactions i. As outlined in section 4.2.4, the term Δ(Cash flow margin) i is either defined as the changes from the combined abnormal cash flow margin of the year - 1 to the median abnormal combined cash flow margin of the years + 1, + 2, and + 3 (panel B) or as the change from the median abnormal combined cash flow margins of the years - 3, - 2, and - 1 to the median abnormal combined cash flow margin of the years + 1, + 2, and + 3 (panel C). Panel B reports a median change in combined abnormal operating performance of 0.97 % that is statistically significant on the 1 % significance level (the corresponding mean change in operating performance is 6.40 % which is also significant on the 1 % level). Panel C reports similar, highly significant median and mean performance changes of 0.74 % and 6.86 %, respectively. As for the hypothesis tests in panel A, I use matched-pairs t-tests for hypothesis tests concerning mean changes in operating performance and non-parametric matched-pairs Wilcoxon signed-rank tests to assess the statistical significance of the median changes 28. The results regarding post-merger operating performance changes are consistent with the findings of Healy, Palepu, and Ruback (1992), Parrino and Harris (1999), and Andrade, Mitchell, and Stafford (2001) who, however, all use the intercept model with a median industry benchmark to determine industry-adjusted post-merger operating performance 27 I use this rather simple approach to detect extreme values because I am reluctant to use standard regression diagnostics to detect outliers in the regression of the changes in cash flow margin changes on the payment method reported in the following section because the regression only uses dummy variables. 28 The use of medians and non-parametric hypothesis tests follows the prescription of Barber and Lyon (1996) and mitigates concerns over skewed or biased data. Other authors such as Ghosh (2001), Fee and Thomas (2004), and Heron and Lie (2002) also use non-parametric Wilcoxon signed-rank tests to assess the statistical significance of median changes in operating performance. Generally, non-parametric tests make minimal assumptions about the distribution of the underlying population of the sample. Furthermore, non-parametric tests are in contrast to parametric tests (like, e.g., t-tests) not concerned with parameters (or defining quantities of a distribution) such as the mean and the variance of distributions (DeFusco et al. (2007)). Due to concerns that not all of the underlying performance data are normally distributed and because Barber and Lyon (1996) find that non-parametric hypothesis tests are uniformly more powerful for analyzing operating performances, I use Wilcoxon signed-rank tests. 29

33 changes due to M&A transactions. Referring to section 4.2.1, the results of these studies might be skewed upward due to methodological issues as explained by Ghosh (2001), who finds positive, however not statistically significant changes in the same order of magnitude in a research setting replicated in this study. Using the change model with an operating profitability measure based on cash flows and the market values of assets, Linn and Switzer (2001) find similar evidence as presented in this paper. But as explained in section 4.2.2, their measure of operating performance is likely to skew post-merger operating performance upward. The empirical results of my study can be best reconciled with the findings of Heron and Lie (2002) who find similar results in virtually the same research setting. However, my empirical findings contradict to Moeller and Schlingemann (2005) and Harford (2005) who report significant negative post-merger operating performance changes. All in all, it seems that this study and the paper of Heron and Lie (2002) provide the strongest evidence that the operating performance of merging firms increases in the post-merger period as both studies systematically account for the factors that are commonly identified as potential sources of biases in the most current research in this field (especially for the factors identified by Ghosh (2001)). The results also support the hypothesis that M&A transactions lead to enhanced combined post-merger operating profitability stated in section 2. Moreover, I also estimate the Pearson and Spearman rank correlation between non-extreme post-merger abnormal operating performance changes and abnormal announcement returns for acquiring firms following Moeller and Schlingemann (2005). I do not find a statistically significant relationship between acquiring firms` announcement returns and post-merger performance changes which contradicts the findings of Moeller and Schlingemann (2005). This indicates that the markets are not statistically reliably successful (or unsuccessful if the correlation coefficients were significantly negative) in correctly assessing future cash flow performance changes at the announcement date of the transactions in my sample. 5.2 Empirical Results for the Relationship Between Post-merger Operating Performance Changes and the Method of Payment As outlined in the previous section, the analysis of the mean operating performance changes in the change model setting is treacherous because of outliners that skew mean operating performance changes upward. The same pitfall applies to the regression analysis presented in section While I analyze operating performance changes based on median values and Wilcoxon signed-rank test statistics, I exclude transactions that I consider as outliners in appendix C in the corresponding regression analysis in this section to avoid biased estimates 30

34 of the influence of the method of payment and industry relatedness on the post-merger abnormal operating performance changes. The empirical results on the relationship between the changes in abnormal post-merger operating performance and the method of payment as depicted in table IV do not support my hypothesis formulated in section 2. The regressions (1) (4) uniformly show that stock and mixed consideration transactions perform better than cash transactions, which is in contrast to the hypothesis that cash transactions are likely to perform best in terms of post-merger operating performance improvements. I find that stock and mixed consideration transactions lead to an improvement of abnormal combined cash flow margins in the order of 2.01 % to 3.19 % depending on the exact model employed. Regression (1) in table IV uses the cash flow margin changes from year - 1 to the median performance of the years + 1, + 2, and + 3 (Δ(Cash flow margin) i from panel B in table III) that are regressed on the method of payment dummies CASH, STOCK, and MIXED. Particularly, stock transactions are associated with an abnormal operating performance improvement of 2.53 %, while firms involved in mixed consideration transactions see their combined abnormal operating margin improved by 2.08 %, which are both statistically significant on the 1 % level using a two-tailed t-test. Regression (2) further incorporates the industry relatedness dummy SAMEINDUSTRY, which is 1 if both the target and the acquiring firm operate in the same Fama and French (1997) industry and 0, otherwise. The introduction of the industry relatedness dummy in regression (2) does not dramatically change the estimates of the coefficients for the dummies already included in regression (1) which are 3.18 % for stock transactions and 2.72 % for mixed consideration transactions, respectively. The coefficient for the stock transactions remains statistically significant at the 1 % level, the estimate for the performance change of mixed consideration transactions is significantly different from zero at the 5 % level. Thus, the operating profitability changes for stock and mixed consideration transactions even increase compared to regression (1). The coefficient for the cash transactions also increases and break-evens, though, remains statistically insignificant. The regressions (3) and (4) which replicate the analysis from the regressions (1) and (2) with the cash flow margin changes of the median abnormal cash flow margins of the pre-merger years - 3, - 2, and - 1 to the median abnormal cash flow margin of the post-merger years + 1, + 2, and + 3 (i.e., Δ(Cash flow margin) i from panel C in table III) yield estimates that are virtually identical. Another surprising result is that mergers among firms from the same industry do not lead to improved abnormal post-merger performance due to allegedly easily realizable synergies. On the contrary, merging firms from the same Fama and French (1997) 31

35 industry classification see their combined abnormal cash flow margins deteriorate in the postmerger period relative to pre-merger levels. However, these results are not statistically different from zero using two-tailed t-tests in the regressions (2) and (4). Table IV Regression of Post-merger Operating Performance Changes on the Method of Payment (Adjusted Sample) Operating cash flows are defined as sales, minus cost of goods sold and selling, general, and administrative expense, plus depreciation & amortization and goodwill amortization expense. The operating cash flows are divided by the sales which results in the cash flow margins. Firms from the control firms universe, which includes all firms tracked in the Compustat North America database except for the firms involved in the transactions in the sample and firms with the SIC codes , , or , are matched to sample firms based on the cash flow margins and size (in terms of sales) in the year - 1, and based on their respective industries. The industry restriction is eased from two-digit SIC codes to one-digit SIC codes and completely dropped each time the matching procedure on the higher level does not yield a control firm for a sample firm. The matching procedure is based on Barber and Lyon (1996) and on the matching algorithm described in Loughran and Ritter (1997). Cash flow margins are reported for the combined sample and matched firms based on aggregated operating cash flows and sales. Particularly, proforma operating cash flow margin figures of merging firms in the years - 3 to - 1 are created by aggregating acquiring and target firms operating cash flow and sales data. The post-merger operating cash flow margins of the merged entity in the years + 1 to + 3 are equal to the corresponding figures of the acquiring firm. Pro-forma operating margins of matched firms for the pre- and post-merger years are formed by aggregating the data of the two matched firms in the same manner as for the pro-forma pre-merger figures of the merging firms. The accouting data are obtained from Standard and Poor`s Compustat North America database. Each transaction (the analyzed sample for this table includes 237 transactions for which the matching procedure yields control firms for both the target and the acquiring firm) is denoted by the subscript i. However, transactions in which the change in the abnormal operating performance from the comparison of year - 1 to the median cash flow margin of the years + 1, + 2, and + 3 is considered as extreme are exluded for the regression (1) and (2). Transactions in which the change in the abnormal operating performance from the median cash flow margin of the years - 3, - 2, and - 1 to the median cash flow margin of the years + 1, + 2, and + 3 is extreme are excluded from the regression (3) and (4). For all regressions I analyze 208 transactions (it is a mere coincidence that exactly the same number of transactions are exluded). Cash flow margin changes are considered as extreme if they fall out of the range between the first quartile minus 1.5 times the inter quartile range (third minus first quartile) and the third quartile plus 1.5 times the inter quartile range. The OLS regression analysis is conducted without an intercept term. The dummy variables CASH, STOCK, and MIXED are 1 if a transaction i has the corresponding payment method, i.e., is a cash, stock, or a mixed consideration transaction and 0 otherwise. The industry relatedness dummy SAMEINDUSTRY takes the value of 1 if both the target and the acquiring firm in a transaction i have the same Fama and French (1997) industry classification and 0, otherwise. Dummy variable Regression (1) Regression (2) Regression (3) Regression (4) CASH -0.06% 0.51% -0.44% 0.38% ( ) ( 0.49 ) ( ) ( 0.31 ) STOCK 2.53% *** 3.18% *** 2.22% ** 3.19% ** ( 2.75 ) ( 2.63 ) ( 2.12 ) ( 2.26 ) MIXED 2.08% *** 2.72% ** 2.01% ** 2.93% ** ( 2.93 ) ( 2.59 ) ( 2.35 ) ( 2.36 ) SAMEINDUSTRY % % - ( ) - ( ) R # Observations The numbers in parentheses denote t-values * denotes significance at the 10 % level for a two-tailed test. ** denotes significance at the 5 % level for a two-tailed test. *** denotes significance at the 1 % level for a two-tailed test. These findings are inconsistent with the results of other authors. The empirical analysis does not suggest a statistically significant positive relationship between cash transactions and increases in post-merger abnormal operating performance which is found by Ghosh (2001) and Linn and Switzer (2001). Neither do the findings agree with the findings of Healy, Palepu, and Ruback (1992) and Heron and Lie (2002), who do not find any relationship 32

36 between the method of financing an acquisition and the post-merger performance changes associated with such transactions. There are some reasons that help explain the discrepancy between the hypothesis of the existence of a positive relationship between post-merger changes in abnormal operating performance and the method of payment from section 2. The availability of record high cash holdings of U.S. companies presented in section 4.1, which are often referred to as acquisition war chests by practitioners and the financial press, are likely to play a role. Abundant cash on acquiring firms` balance sheets make the use of debt to finance cash transactions unnecessary and therefore the disciplinary function of debt does not take effect. Rather, these war chests which are part of what is known as financial slack in the corporate finance literature, tempt managers to conduct hasty acquisitions (often, as an answer to investors requiring high returns on the invested capital). These hastily conducted acquisitions often do not follow a strict acquisition strategy and therefore fail to create value for acquiring firms (Volkart (2008)). Another possible explanation of the findings differing from the hypothesis are systematic differences in the size and therefore, according to economic theory, systematic differences in operating performance of merging firms across the different payment methods as encountered in the argumentation against the use of median industry firms as performance benchmark in section The descriptive statistics in section 4.1 show that the mean deal values as a proxy for the firm size are larger for stock and mixed consideration transactions than for cash transactions (however, only the median deal values for mixed consideration transactions differ significantly from the respective median deal values of the other payment methods). Besides the increasing returns to scale, larger firms are likely to be monitored more closely by stock analysts and lenders, which puts their management under pressure to deliver on operating performance improvement expectations and therefore, to push through more rigid cost cutting and performance improvement programs. Furthermore, bigger merging firms are likely to offer a larger potential for the realization of synergies due to more global geographic overlapping of their operations. Moreover, some of the premises used to derive the hypothesis of superior performance changes of cash acquisition over other payment methods do not necessarily apply for the acquisitions in my sample. Besides the model introduced by Fishman (1989), I base my argumentation for deriving the hypothesis on empirical findings of other studies, such as the presence of earnings management to create an overvalued stock as an acquisition currency as presented by Erickson and Wang (1997) or the operating underperformance of firms issuing stock in secondary offerings reported by Loughran and Ritter (1997). 33

37 However, to determine the exact explanation for this empirical phenomenon, further examinations such as a test of the presence of earnings management would be needed, which however, go far beyond the scope of this paper. 6. Summary This paper examines whether the post-merger abnormal operating performance changes due to mergers and acquisitions by analyzing a sample of 242 transactions among U.S. public companies completed between 1998 and 2008, inclusive. I hypothesize that the post-merger abnormal cash flow margins increase relative to premerger levels because of considerable acquisition premiums paid for the target firms. More precisely, based on a simple theoretical equation found in Koller, Goedhart, and Wessels (2010), which states that an acquiring firm only creates value with an acquisition if the value of the future operating performance improvements of the combined entity outweigh the acquisition premium paid over the current market capitalization of the target firm, I expect that the managers of public acquiring firms are under considerable pressure to enhance the operating profitability of the combined entity. Furthermore, operating synergies are among the mostly cited rationales for conducting acquisitions, therefore I would expect them to actually materialize in the post-merger period. My second hypothesis concerns the relationship between abnormal post-merger operating improvements and the method of payment. I expect stronger increases in post-merger operating performance for cash transactions compared to stock and mixed consideration transactions. I derive this hypothesis partly from empirical findings. There is empirical evidence that firms which issue stock in secondary offerings (which is often necessary to fully finance an acquisition with stock) underperform control firms in terms of operating performance in post-issuance years and that companies manage their earnings in the period prior to mergers and acquisitions to inflate the value of their own stock, which leads to inferior post-merger performance. There also exist theoretical models on preemptive bidding in tender offers which indicate that bidders tend to offer cash when they give the target firm a high valuation. Finally, this hypothesis is supported by the disciplinary effect of debt which is often raised to finance cash transactions. The prior research in this field does not give a clear indication whether the operating performance of merging firms increases in the post-merger period. Rather, the findings seem to be sensitive to the exact research design utilized in the different studies. I base my study on the work of Ghosh (2001) who criticizes several methodological issues of the influential 34

38 Healy, Palepu, and Ruback (1992) study. I follow Ghosh (2001) by using a research design which is often referred to as the change model which directly compares pre- to post-merger operating performance, measuring the operating performance as operating cash flows divided by sales. I adjust the operating performance of merging firms by the corresponding performance of control firms matched based on size, industry, and pre-merger performance. I find my first hypothesis confirmed using the research design and methodologies prescribed in Ghosh (2001) and Barber and Lyon (1996). Merging firms experience significantly stronger operating performance relative to their non-merging counterparts in the two years subsequent to the merger completion year. The change in mean and median abnormal operating performance surrounding the transaction is statistically significant positive. However, my empirical findings are in contrast with my second hypothesis regarding the relationship between post-merger operating performance improvements and the method of payment Cash transactions do not exhibit significantly positive post-merger abnormal operating performance changes while stock and mixed consideration transactions do on average experience abnormal operating performance improvements between roughly 2 % and 3.2 %. I attribute this difference between my findings and my expectations to three possible factors. Due to record high cash holdings in recent years, U.S. acquiring firms are less likely rely on debt financing to conduct acquisitions, in contrary, they tend to conduct less rigorously reasoned acquisitions putting the low interest bearing excess cash on their balance sheets to work. Another explanation may lie in the notion that the premises on which I derive my hypothesis do not necessarily have to hold for my merger sample. However, to evaluate these first two explanations, further empirical analyzes such as tests for the presence of earnings management would be necessary, which though, go beyond the scope of the present thesis. Finally, the surprising findings may stem from the systematic size differences between firms conducting acquisitions using the different payment methods. 35

39 Appendix A. Overview of the Relevant Prior Research Ordered by Their Findings 36

40 Appendix A 37

41 Appendix A 38

42 Appendix B. List of all Mergers and Acquisitions in the Sample 39

43 Appendix B 40

44 Appendix B 41

45 Appendix B 42

46 Appendix B 43

47 Appendix B 44

48 Appendix B 45

49 Appendix B 46

50 Appendix C. Change Model Analysis Excluding Extreme Performance Changes 47