The role of divestitures in horizontal mergers: Evidence from product and stock markets Abstract In this first large-sample study of merger-related divestitures, we find that divestitures both reduce the market power and affect the competitive efficiency of merging firms. Postmerger output prices are lower when assets are divested, particularly when sold outside the industry. Consistent with this, stock price reactions of customer firms are more positive if merging firms divest assets outside the industry. In contrast, stock price reactions of the acquirer and rivals suggest that firms are more concerned with maintaining a competitive edge. We also find that the results vary based on industry competitiveness and buyer size.
1. Introduction In the last three decades, more than 25% of mergers and acquisitions (M&A) in the United States were between rivals in the same industry. These deals, known as horizontal mergers, constituted almost half of the total dollar value of merger activity. 1 The Federal Trade Commission (FTC) and the Department of Justice (DOJ) routinely scrutinize horizontal mergers to ascertain whether a merger has the potential to reduce competition within the industry. Both regulatory bodies state that if a merger is deemed anticompetitive, the most common remedial measure pursued is divestiture. 2 Consistent with that statement, Kühn (2010) observes that antitrust policy increasingly imposes structural remedies rather than a categorical veto of a merger. Kühn also notes that in anticipation of regulatory objections, firms often voluntarily divest assets. Regulators expect that divestiture will limit the market power of the merging companies. This notion is intuitive and supported by nascent industrial organization theory (Vergé, 2010; Vasconcelos, 2010). Yet, there is little empirical evidence on the impact and effectiveness of merger-related divestitures. We try to fill this void in the literature. Examining the impact of merger-related divestitures can also shed new light on the market power potential of horizontal mergers, particularly given the lack of agreement in existing literature regarding this issue. Studies relying on stock-return analysis conclude that horizontal mergers are motivated by efficiency considerations and do not create market power (see Eckbo, 1983; Stillman, 1983; Eckbo, 1985; Eckbo and Wier, 1985; Fee and Thomas, 2004; Shahrur, 2005). Since this research does not account for divestiture, it cannot determine whether 1 Calculated using Thompson Reuter s SDC Platinum database as total transaction value of all mergers and acquisitions in which acquirer and target operate in the same four-digit SIC code divided by the total transaction value of all mergers and acquisitions. 2 See Antitrust Division Policy Guide to Merger Remedies provided by the DOJ and Negotiating Merger Remedies provided by the Bureau of Competition of the FTC. 1
anticompetitive potential exists but is mitigated through mandated or voluntary divestitures. A few studies that examine post-merger output prices conclude that horizontal mergers result in higher output prices due to a reduction in competition (see Barton and Sherman, 1984; Borenstein, 1990; Kim and Singal, 1993; Singal, 1996; Prager and Hannan, 1998). Again, these papers do not consider the role of divestitures. This paper s primary objective is to provide the first large-sample analysis of divestitures that accompany horizontal mergers. The first question we ask is whether divestitures limit the price-increasing effect of horizontal mergers as predicted by nascent industrial organization theory. We examine post-merger product prices as well as the merger announcement returns of the acquirer, rivals and customers conditional on the occurrence of a divestiture. The second question we address is whether the effectiveness of divestitures as a merger remedy depends on who buys the divested assets. Recent theory by Vergé, (2010) and Vasconcelos (2010) predicts that divestitures are more effective in limiting the price-increasing impact of mergers if assets are sold to an existing, well-established rival. An alternative argument is that selling assets to a firm outside the industry brings a new player into the market and infuses more competition in the industry. We conduct the first large-sample analysis to test these two competing arguments. The third question we ask is whether the efficacy of divestitures as a remedy for mergerinduced market power depends on industry conditions. In highly competitive markets, horizontal mergers are not expected to harm customers interests and divestitures will probably not be considered necessary to preserve competition. Anti-trust authorities look at a variety of indicators of the state of competitiveness in a market when deciding whether to challenge a 2
merger. However, there has been no systematic empirical effort to understand which industry characteristics indicate that customers of the merging firms will benefit from a divestiture. Our analysis of merger-related divestitures contains two sets of tests. In industry-level tests, we examine the impact of divestitures on post-merger output prices. In firm-level tests, we examine the effect of divestitures on merger announcement returns. In all our tests, we use two theoretical viewpoints to help outline the empirical predictions: the Market Power Hypothesis and the Competitive Efficiencies Hypothesis. The two hypotheses are not mutually exclusive; market power and efficiency gains coexist in horizontal mergers. Our tests capture the net effects of the two forces on acquirers, rivals, and customers. Under the Market Power Hypothesis, horizontal mergers lead to lower industry competition and higher output prices. Divestitures, however, can reduce the size of the merged firm and limit the merged firm s ability to exercise monopoly power. Thus, the market power argument predicts that divestitures mitigate the upward pressure on output prices. Under the Competitive Efficiencies Hypothesis, horizontal mergers enhance economies of scale and give the combined firm a competitive advantage. If merger-driven scales economies are passed on to customers, prices will decline following horizontal mergers. However, divestitures can limit scale economies and potentially thwart this post-merger price decline. The two hypotheses also have contrasting predictions if assets are sold to an existing rival instead of a firm outside the industry. Under the Market Power Hypothesis, if the divested assets are purchased by a firm outside the industry (i.e. by a new entrant), the larger number of players makes the marketplace more competitive, and cooperative collusion harder to sustain. Thus, divestitures are more effective in keeping output prices down if assets are sold to firms outside the industry. An alternative argument, presented by Verge (2010) and Vasconcelos (2010) is that 3
new entrants cannot be cost efficient enough to compete effectively with the large merged firm. These papers, which fall under our Competitive Efficiencies Hypothesis, argue that divestitures will be more effective in keeping output prices low if assets are sold to existing players in the market. We test these predictions on a sample of 1,153 large horizontal mergers that occurred in manufacturing industries between 1980 and 2010. Industry-level analysis shows that postmerger output prices are lower by 2.2 percentage points if the acquirer or target divests assets compared to mergers without accompanying divestitures. However, if a larger fraction of divested assets is purchased by existing rivals, post-merger output prices decline to a smaller degree. A onestandard-deviation increase in the fraction of assets sold to existing rivals increases the postmerger prices by 0.9 percentage points. These differences are economically significant given that the average post-merger percentage decline in output prices is 3.8%. When examining firm-level merger announcement returns, we find that customer cumulative abnormal returns (CARs) are higher if a divestiture occurs but this positive effect is dampened if existing rivals buy a larger fraction of divested assets. In terms of economic significance, a divestiture increases a customer firm s three-day return by 0.7 percentage point on average; a one-standard-deviation increases in the fraction of assets sold to existing rivals however, decreases the return by 0.6 percentage point, largely eliminating the benefits of the divestitures. Thus, customer CARs appear to move in line with output prices, and both industrylevel analysis and customer CARs are supportive of the Market Power Hypothesis. To gauge the net effect of market power and competitive efficiencies on industry peers, we look at the stock price reactions of acquirers and rival firms. We find that acquirer returns are lower when the merger is accompanied by a divestiture, especially when firms in the same 4
industry purchase divested assets. This result is noteworthy given our finding that output prices are higher when firms in the same industry purchase divested assets. It suggests that, for an acquiring firm, concerns about an existing rival growing bigger outweigh any benefits from gaining pricing power relative to customer firms. Similarly, for rivals, we find that concerns over relative competitive efficiencies dominate the advantage of increased market power. Rivals CARs are significantly higher when the merged firm divests assets even though such divestitures result in lower output prices. Moreover, rivals CARs are lower if a larger fraction of divested assets is bought by existing competitors rather than by new entrants. Finally, we explore how our results depend on industry characteristics and the size of the buyer (of divested assets). In highly competitive industries, one firm s output decision has little impact on output prices in the industry. Thus, mergers and divestitures are unlikely to have a significant effect on pricing power. In line with this, we find that customer stock-price reaction to merger-related divestitures is positive in concentrated industries but insignificant in competitive industries. Likewise, in concentrated industries, customers react negatively if an existing competitor buys the divested assets, while in competitive industries customers are indifferent to whether an existing competitor or an outsider purchases divested assets. In addition, the size of the firm buying the divested assets matters. Although we have limited data on the size of the firm buying divested assets, our tests show that customers react more negatively to mergers in concentrated industries if assets divested by merging firms are bought up by large competitors. Our paper has three key contributions. First, we are the first to demonstrate that divestitures accompanying horizontal mergers reduce the market-power impact of the mergers, 5
which is precisely what antitrust authorities hope to achieve. Second, we show that divestitures are more effective in reducing the market-power impact of horizontal mergers if the assets are sold to firms outside the industry. Third, we find that merger-related divestitures are most beneficial for customers when industry conditions are conducive to the exercise of market power, for example, in more concentrated industries. Our paper also helps to reconcile the seemingly conflicting findings of existing product market studies, which conclude that horizontal mergers create market power, and of stock market studies, which conclude that mergers are primarily motivated by cost efficiencies. Our empirical analysis shows that horizontal mergers are accompanied by changes in both market power and competitive efficiency. The market power implications of mergers and accompanying divestitures are visible in output prices and customer stock price reactions. The impact of mergers and divestitures on firms relative competitive efficiencies is reflected in the stock price reactions of acquirers and rival firms. The rest of the paper is organized as follows: Section 2 develops the hypotheses. Section 3 describes the sample. Section 4 presents the output price analysis. Section 5 presents announcement returns. Section 6 discusses additional cross-sectional tests and Section 7 concludes. Robustness issues are presented in the appendix. 2. Hypothesis development Merging firms often sell assets in the period preceding or following a merger, either mandated by regulatory authorities or voluntarily. Selling assets reduces the size of the firm and can affect the scale of output produced by the firm. The quantity produced, in turn, has implications for output prices as well as cost economies. This section focuses on how output 6
prices and stock returns might be expected to change due to divestitures under the Market Power and Competitive Efficiencies Hypotheses. Competitive efficiencies and market power can coexist. Our tests aim to tease out the presence of market power effects while accounting for the existence of efficiency gains. All predictions discussed in this section are summarized in Table 1. 2.1. Predictions for industry output prices The Market Power Hypothesis contends that horizontal mergers can result in greater pricing power either through a unilateral reduction in output (e.g., Perry and Porter, 1985; Perry, 1984) or by increasing the opportunities for cooperative collusion (e.g., Green and Porter, 1984; Tirole, 1988; Vasconcelos, 2005). Whether unilateral or cooperative, a reduction in output results in higher prices for customers. Antitrust authorities often require merging firms to divest assets in order to mitigate the market power impact of horizontal mergers. The structural remedy guides of the DOJ and FTC note that if a merger is deemed potentially anticompetitive, the merging firms must sell clearly identifiable assets that can be immediately deployed by the purchasing firm to preserve long-term competition in the industry. Thus, the Market Power Hypothesis posits that postmerger output prices are expected to be lower, ceteris paribus, if a merger is accompanied by a divestiture than if not. However, horizontal mergers can also create economies of scale and eliminate redundancies. Divestitures might reduce the merging firms ability to realize economies of scale and pass on cost cuts to customers. Under this Competitive Efficiencies Hypothesis, horizontal mergers with accompanying divestitures may be followed by higher selling prices than mergers that have none. 3 3 We do not attempt to distinguish market power from the so-called scale-decreasing efficient mergers. Firms in industries experiencing a long-term decline in demand or suffering from excess capacity may undertake mergers and divestitures with the intention of consolidating and downsizing. In such cases, closing idle plants and reducing output is an efficient response to long-term industry conditions (see Andrade and Stafford, 2004; Eckbo, 1983; 7
Next, we discuss the role of the purchasing firm, i.e., whether the divested assets are bought by a firm in the same industry or by a firm outside the industry. The logic of the market power hypothesis predicts that divesting assets to an outsider is more effective in limiting the price impact of horizontal mergers. Selling assets to firms outside the industry effectively brings new players into the market and makes the industry more competitive. With more players in the market, it is more difficult for firms either to either exercise unilateral market power or collude to restrict output. In contrast, relying on cost-efficiency arguments, Vasconcelos (2010) and Verge (2010) predict that divestitures limit the price-increasing impact of horizontal mergers only if the assets are sold to a rival within the same industry. Divesting assets to other firms in the same industry allows existing rivals to become larger and more cost efficient. These larger rivals can compete more effectively with the merged firm than a new entrant can. 2.2. Firm-level predictions We conduct a series of firm-level tests to see how merger-related divestitures affect the welfare of the acquirers, their rivals, and customer firms. Specifically, we examine the various parties stock price reactions to merger announcements. Our main interest is in determining whether the stock price reactions depend (a) on the occurrence of a divestiture and (b) on whether the assets are divested to an existing rival rather than to a firm in a different industry. 2.2.1. Customers We begin by discussing customer firms because the impact of a change in selling prices on the welfare of customer firms is straightforward. If output prices are higher (lower) following a merger, customer firms should experience negative (positive) stock returns. As discussed Shahrur, 2005). In the context of industrial organization theory, any decline in the number of firms that results in lower industry output and higher selling prices is considered to be a unilateral decline in the competitiveness of the industry. By this metric, scale-decreasing efficient mergers reduce competition and create pricing power. 8
above, under the Market Power Hypothesis, a divestiture limits the merged firm s monopoly power. Therefore, horizontal mergers accompanied by divestitures should elicit more positive stock-price reactions from customers compared to mergers that have no accompanying divestitures. In contrast, under the Competitive Efficiencies Hypothesis, a divestiture can limit economies of scale. If merged firms are believed to routinely pass on the benefits of scale economies to customers, then mergers accompanied by divestitures will lead to more negative stock price reactions from customers. What about the identity of the firm buying the divested assets? Under the Market Power Hypothesis, output prices are higher if an existing rival buys the divested assets. This implies a negative stock price reaction for customers if divested assets are bought by an existing rival rather than by a firm outside the industry. The opposite is predicted under the Competitive Efficiencies Hypothesis. Output prices are lower; therefore, customer returns will be higher if an existing rival buys the assets. 2.2.2. Acquirers When considering the stock-price reactions of acquirers, we take into account changes in market power relative to customers, the acquiring firms ability to realize cost cuts, and one additional factor: whether a rival firm becomes larger due to a divestiture. If horizontal mergers are motivated primarily by market power, then acquirer returns will be lower if some assets are divested since the divestiture will limit the acquirer s ability to exercise unilateral market power. However, a divestiture may hurt the acquirer s ability to realize cost efficiency. Thus, both the Competitive Efficiencies and Market Power Hypotheses predict that acquirer returns will be lower for mergers with divestitures than for mergers without. 9
We also consider how acquirer returns depend on the buyer of the divested assets. Under the Market Power Hypothesis, divesting assets to a buyer outside the industry brings a new player to the market. A larger number of firms in an industry limits unilateral market power and reduces the sustainability of collusive agreements. Therefore, acquirer returns will be higher if assets are divested within the industry because doing so keeps the number of firms in the industry lower and makes collusion more sustainable. Under the Competitive Efficiencies Hypothesis, acquirer returns will be lower if assets are divested to a firm within the industry because selling assets to existing rivals results in a larger, more cost-efficient rival firm that can compete aggressively with the merged firm. 2.2.3. Rivals In their discussion of horizontal mergers, Spiegel and Tookes (2003) argue that rivals benefit from fewer competitors but are hurt if the combined firm is a much stronger competitor than the standalone firms. 4 Our analysis of divestitures allows for the possibility that both pricing power and productive efficiencies are at play and tests the relative importance of each for rival firms. To do so, we focus on the rival firms that do not purchase the assets divested by the merging firms. Under the Market Power Hypothesis, industry output is lower and output prices are higher after the merger than before. A divestiture reduces the pricing power that can be created via the merger. This has negative implications for the stock returns of rival firms, since they also lose out on the possibility of selling at higher prices. Under the Competitive Efficiencies Hypothesis, a divestiture reduces the size of the merged firm and limits its ability to realize 4 For example, when AT&T announced its bid for T-Mobile in 2011, Sprint-Nextel urged the United States government to block the acquisition by saying that the transaction would hurt competition. When the DOJ moved to block the acquisition, Sprint-Nextel s share price went up by 8%. See U.S. Moves to Block AT&T Merger with T- Mobile, New York Times, Aug 31, 2011, and Spring Shares Surge on AT&T Setback, New York Times, Aug 31, 2011. 10
economies of scale. This is good news for a rival firm because it has to contend with a somewhat smaller merged entity. Thus, if competitive efficiencies relative to the merged firm are more important for rivals than market power relative to customers, then stock prices of rivals would react more positively to mergers accompanied by divestitures. Finally, we consider how rivals stock prices react if a firm outside the industry purchases the divested assets. Under the Market Power Hypothesis, divesting assets to a buyer outside the industry brings a new player in the market and increases competition. Therefore, rival returns will be higher if assets are divested within the industry because doing so keeps the number of firms in the industry lower and prices higher. In contrast, under the Competitive Efficiencies Hypothesis, if the merged firms sell assets within the industry, rivals who do not purchase the divested assets now have two larger firms to compete with: the combined firm resulting from the merger and the rival that purchases the divested assets. This will cause the stock returns of rival firms to be lower. 3. Sample description We obtain from the SDC Platinum database all completed mergers and acquisitions announced between 1980 and 2010 in which more than 50% of the targets shares were acquired. From these, we select all horizontal mergers, defined as deals in which the acquirer and target operate in the same four-digit SIC code. If the acquirer or target operates in multiple segments, we use the three largest segments of each firm to determine whether the firms operate in the same industry segment. To identify horizontal mergers likely to have significant bearing on product markets, we calculate quarterly horizontal merger activity in each industry as the total transaction value of all mergers announced in that quarter divided by industry total assets (based 11
on Compustat firms) in that quarter. If merger activity involves more than 5% of an industry s total assets, we classify that industry as having experienced a horizontal merger event and retain all horizontal mergers contributing to that event. If an industry experiences more than one merger event within a three-year period, we keep only the first event in the sample. Doing so gives us clear pre-merger and post-merger periods over which to study the change in output prices. Our sample has 1,153 horizontal mergers that contribute to 813 horizontal merger events in 306 distinct four-digit SIC codes. Table 2 summarizes the number of merger events and the number of deals in our sample, organized by two-digit SIC code. Next, we identify which of these horizontal mergers are accompanied by divestitures. Our primary analysis considers divestitures during the two years before or two years following merger announcement year. 5,6 Our results are robust if we select divestitures that occur within one year before or one year after the merger announcement year. The last row of Table 2 shows that of the 813 industry merger events, 340 (i.e. 42%) are associated with at least one divestiture by either the acquirer or the target. At the deal-level, of the 1,153 mergers, 434 (i.e., 38%) are accompanied by at least one divestiture. Table 3 reports the frequency of divestitures and the percentage of events and deals that involve sale of assets to a firm in the same industry. Panel A reports summary statistics at the industry-event level. Our first main explanatory variable is Divestiture Dummy, which takes a 5 We select a two-year horizon to identify divestitures because news searches reveal that even divestitures with a long time gap can be directly related to the horizontal merger. For example, in January 1997, Raytheon bought the defense business of Texas Instruments. More than 18 months later, Raytheon sold two business units to DRS Technologies as part of its agreement with the Justice Department. 6 Our sample includes divestitures for which there is no direct media coverage linking the divestiture to the merger. It is possible that these divestitures are due to motives unrelated to the merger. (For existing evidence on various motives for divestitures, see Jain, 1985; Berger and Ofek, 1995; Comment and Jarrell, 1995; and Lang, Poulson and Stulz, 1994.) Nonetheless, selling assets reduces the size of the firm and can affect the scale of output produced by the firm. The quantity produced, in turn, has implications for output prices as well as for cost economies. If these divestitures truly have no impact on market power or competitive efficiency, including them biases us against finding evidence in support of our hypotheses. 12
value of one if a merger event is associated with at least one divestiture and zero otherwise. Panel A of Table 3 shows that this variable has a mean value of 0.42 (which corresponds to 340 out of 813 merger events). For 91 of the 340 events with divestitures, the divested assets were bought by at least one firm within the same four-digit SIC code. Our second main explanatory variable, Pct_SameInd, captures the fraction of divested assets sold to firms within the same industry. To calculate Pct_SameInd at the industry-event level, we first count the total number of divestitures accompanying all the mergers in a merger event. Next, we count the number of divestitures in which the buyer operates in the same four-digit SIC code as the merging firms and divide this number by the total number of divestitures in the event. On average, 15% of divestitures per event involve sale of assets to firms within the same industry. Pct_SameInd is calculated conditional on the occurrence of a divestiture and has a missing value if a merger event has no accompanying divestitures. We also calculate an unconditional version of this variable, called Pct_SameInd_All, which is set equal to Pct_SameInd if a divestiture occurs and is set equal to zero if it does not. The three main explanatory variables are similarly defined for each deal (as opposed to each industry event). The deal-level statistics are shown in Panel B of Table 3. Of the 1,153 M&A deals, 434 (i.e. 38%) are associated with at least one divestiture. Pct_SameInd at the deal level is defined as follows: For each deal, we count the number of accompanying divestitures in which the buyer operates in the same industry as the merging firms and divide this number by the total number of divestitures per deal. Panel B shows that 18% of divestitures per deal involve sale of assets to firms within the same industry. Finally, we calculate a deal-level measure of Pct_SameInd_All, which is set equal to zero when there is no divestiture and equal to Pct_SameInd when there is. 13
4. Output price analysis This section describes our industry-level variables and presents an analysis of change in output prices conditional on divestitures. 4.1. Industry-level variables For industry-level output price, we rely on the monthly producer price index (PPI) obtained from the Bureau of Labor Statistics (BLS). The PPI measures the average selling prices received by domestic producers for their output. 7 We deflate the PPI using the consumer price index to adjust for inflation and use the natural log of this deflated series as a measure of the real producer price index (RPPI). We calculate the change in RPPI (ΔRPPI) as the average RPPI during the three years following the industry merger event less the average RPPI during the three years prior to the merger event. Since small changes in the natural log of a variable can be interpreted as percentage changes, ΔRPPI measures the percentage change in real PPI. In the multivariate analysis of ΔRPPI, we follow Bhattacharyya and Nain (2011) and control for input prices and demand conditions. Higher input prices and higher demand are expected to put upward pressure on selling prices. To obtain input prices, we first use the benchmark input-output tables provided by the Bureau of Economics Analysis to identify, for each industry, the two upstream industries that supply the largest fraction of that industry s input. Then we calculate ΔRPPI for the largest and the second largest supplier industries. To capture changes in overall demand conditions in the economy, we use changes in Total Industrial 7 The Producer Price Index Series reflect price movements for the net output of goods-producing sectors of the US economy. To the extent possible, prices used in constructing the indexes are the actual revenue or net transaction prices producers receive for sales of their outputs. Scientific (probability) sampling techniques are used to select reporting establishments, products, and transactions for all types and volumes of output. The PPI measures changes in prices received by domestic producers; imported products are not priced in the survey. More details can be found in chapter 14 of the BLS Handbook of Methods. 14
Production (ΔTP), which is obtained from the Federal Reserve Board. The production index measures real output and is expressed as a percentage of output in the base year 2007. The production index is also converted to natural logs. Finally, we control for pre-merger industry concentration. An industry that is highly competitive prior to the merger event is likely to benefit more (in terms of enhanced pricing power) from industry consolidation than an industry already heavily concentrated before the merger. Therefore, we use the four-firm concentration ratio obtained from the Census Bureau as a control variable (CR4). The four-firm concentration ratio is the percentage of Value of Shipments accounted for by the four largest firms in the industry and includes all establishments in the industry, both privately held and public. Panel A of Table 4 summarizes industry-level variables and compares mean values before and after the merger event. Out of 813 merger events, we can calculate ΔRPPI for 570 events. For the remaining industry events, SIC codes of the merging industries cannot be matched with the SIC codes in the PPI data or sufficient PPI data are not available. It is evident from Table 4 that horizontal merger events are accompanied by significant changes in the industry. The average output price in the merging industry (RPPI) is significantly lower after the merger than before. Input prices, Supplier 1 RPPI and Supplier 2 RPPI, are also lower. Total industrial production is higher after merger events. 8 As expected, industry concentration is higher after horizontal merger events than before. These significant differences present a challenge for the industry-level analysis. Horizontal mergers are likely endogenously determined by changes in industry factors, thus making it difficult to assess market power simply by comparing output prices before and after the merger. A downward trend in output prices may trigger horizontal mergers that help shore up 8 The decline in supplier prices after the merger is consistent with Bhattacharyya and Nain s (2011) finding that horizontal mergers enhance buying power versus supplier industries. 15
prices relative to the counterfactual (i.e., relative to what prices might have been in the absence of the merger). However, post-merger output prices may still be lower than pre-merger prices. Thus, a decline in prices is not an indicator that pricing power was not created. Similarly, an increase in prices after a horizontal merger is not in itself evidence of an increase in market power since industry demand factors that cause an upward trend in prices may also drive merger activity. Our analysis focuses on postmerger changes in output prices conditional on the occurrence of divestitures. Comparing mergers with divestitures to mergers without divestitures should mitigate the concern that mergers are endogenous to industry conditions. 4.2. Output price results Panel B of Table 4 compares ΔRPPI for the subsample of merger events with accompanying divestitures and the subsample without. The producer price index drops by 3.6% for the subsample with divestitures and by 2% for the subsample without. 9 The difference is statistically significant at the 5% level. Table 4 also presents a comparison of ΔRPPI conditional on the industry of the buying firm. RPPI declines by 4.54% if all buyers of divested assets are outside the merging industry (i.e., Pct_SameInd=0) and declines by 1.15% if at least one buyer operates in the same four-digit SIC as the merging firms. The difference is statistically significant at the 5% level. Overall, the univariate comparison indicates that post-merger prices are lower when a horizontal merger is accompanied by a divestiture, particularly when the assets are bought by firms outside the industry. The univariate results are thus consistent with the Market Power Hypothesis. 9 Percentage changes and differences in naturals logs are almost identical up to a +/- 5% change and very similar up to a +/- 10%. 16
We confirm these results in a multivariate setting. We regress ΔRPPI on Divestiture Dummy, controlling for other factors that can affect the output prices described in Section 4.1. The regression results are presented in Table 5. Requiring data on supplier industry prices reduces the sample size to less than 200 industry events. Columns 1 3 of Table 5 include all merger events those with divestitures and those without. The coefficient on Divestiture Dummy is negative and statistically significant in all three regressions, which is consistent with the univariate results in Panel B of Table 4. In Column 2, the coefficient on Pct_Same_All is positive and statistically significant, indicating that output prices are higher when divested assets are purchased by firms within the industry. In Column 3, we also include the premerger industry concentration as a control variable and find that Pct_SameInd_All remains positive and statistically significant. Returning to the hypothesis summarized in Panel A of Table 1, the negative coefficients on Divestiture Dummy and the positive coefficients on Pct_SameInd_All are consistent with the Market Power Hypothesis. In terms of economic significance, Column 3 suggests that output prices are 2.2 percentage points lower for mergers with divestitures than for mergers with no divestitures. In addition, a one-standard-deviation increase in Pct_SameInd_All (0.21) increases the postmerger prices by 0.86 percentage points (=0.21*0.041, the latter being the regression coefficient). In comparison, the average postmerger decline in RPPI in our sample is 3.8% for the regression sample. In Columns 4 and 5, we restrict the sample to merger events that were associated with at least one divestiture and use Pct_SameInd as the primary explanatory variable. We see that in both columns, the coefficient on Pct_SameInd is positive, indicating that post-merger output prices are higher when more assets are divested within the industry. The result in Column 5 17
suggests that a one-standard-deviation increase in Pct_SameInd (0.28) increases post-merger prices by 1.12 percentage points (=0.28*0.040, the latter being the regression coefficient). Antitrust authorities routinely require merging firms to divest assets with the expectation that the divestitures will limit the market-power potential of the horizontal merger. The evidence presented here is the first large-sample evidence indicating that divestitures do indeed mitigate the price-increasing impact of horizontal mergers. It is also the first evidence suggesting that divestitures are more effective in curbing postmerger increases in output prices if assets are sold to firms outside the industry than to existing rivals. 5. Announcement returns analysis In this section, we examine whether returns to acquiring firms, their rivals, and their customers at the announcement of a merger depend on whether any assets are divested and whether the buyer is an existing rival or an outsider. We first describe firm-level variables and then present results of the firm-level analysis. 5.1. Firm-level variables We identify customer firms using the Benchmark Input-Output tables compiled by the Bureau of Economic Analysis. Specifically, for each merger, we rely on the Benchmark Use Tables to identify the merging industry s most dependent customers. For each customer industry (i.e., an industry using the merging firm s commodity), we follow Shahrur (2005) and calculate a ratio called the Input Coefficient that captures the fraction of the customer industry s total input usage provided by the merging industry. We identify the most dependent customer industry for 18
each merger, provided at least 5% of its input is supplied by the merging industry. 10 We then use Compustat to select all publicly traded firms that operate in the customer industry. We identify rivals as firms in the same four-digit SIC code as the acquirer and target (but not including the merging firms), provided stock-price data for the rival are available in CRSP. We exclude rivals that buy divested assets, thereby focusing on nonbuying rivals only. On average, we are able to identify 24 nonbuying rivals for each merger. 11 We calculate cumulative abnormal returns (CARs) around merger announcements using the market model. Market model parameters are estimated during the period of trading days [ 271, 21], relative to the announcement date of the merger. Cumulative abnormal returns are calculated for the three trading days centered on the announcement date. Since the returns to rival firms for a given acquisition can be correlated, for each acquisition we use the valueweighted portfolio return of all rival firms. Similarly, we use the value-weighted portfolio return of customer firms. Our main results hold if we use the equal-weighted portfolio return for rivals and customers. Our sample of 1,153 horizontal mergers consists of both public and private acquirers. We are able to calculate acquirer CARs for 644 deals, rival CARs for 985 deals, and customer CARs for 726 deals. Table 6 reports that the mean (median) acquirer announcement return is 0.90% (0.43%), significant at the 1% (5%) level. The mean (median) rival portfolios return is 0.04% ( 0.04%), which is not significantly different from zero. The mean (median) customer portfolios return is 0.21% ( 0.26%), significant at the 10% level. These average announcement returns are consistent with existing studies that generally find nonnegative 10 In a robustness check, we identify the three (instead of one) most dependent customer industries for each merger. The results are qualitatively similar. 11 In the appendix, we discuss the subset of rival firms that purchase divested assets. 19
average returns for rivals and positive returns for the combined merging firms (e.g., Eckbo, 1983; Fee and Thomas, 2004; Shahrur, 2005). Table 6 also presents summary statistics of deal-level variables that are expected to affect CARs and are used as control variables. The control variables are described in the appendix. 5.2. Customer CARs We consider the stock-price reactions of customer firms first since these are the most directly linked to the output price analysis in Section 4.2. In scenarios that lead to higher output prices, customer stock returns should be negative, and vice-versa. Table 7 presents regressions of customer CARs on our main explanatory variables: Divestiture Dummy, Pct_SameInd_All (if the full sample is used), and Pct_SameInd (if only the sample with accompanying divestitures is used). Panel A presents OLS regressions, and Panel B presents endogeneity-corrected regression models. In the first four columns of Panel A, we use the full sample including deals with and without accompanying divestitures. The coefficient on Divestiture Dummy is positive and statistically significant in three of the four regressions, indicating that customer CARs are higher if a horizontal merger is accompanied by a divestiture. This is consistent with our finding that output prices are lower when the merging firms divest assets. The result indicates that customer CARs are higher by 0.4-0.7 percentage points for mergers accompanied by divestitures. The coefficient on Pct_SameInd_All is negative in columns 2 through 4, indicating that customer CARs are lower if a greater fraction of divested assets is bought by firms within the industry. This is also consistent with our previous result that output prices are higher when assets are divested to firms within the industry rather than to firms outside the industry. 20
In Columns 5 7, we restrict the sample to deals accompanied by divestitures. Again, we see that the coefficient on Pct_SameInd is significantly negative. Taking the regression in Column 7 for example, the coefficient on Pct_SameInd is 0.016 (t-statistic of 2.92). Economically, an increase in Pct_SameInd by one standard deviation (0.35) is accompanied by a decline in customer CARs of 0.56 percentage points (= 0.016*0.35). Based on the summary of predictions presented in Panel B of Table 1, we conclude that customer CAR results are in line with the Market Power Hypothesis. We recognize that divestitures are not exogenous events and are likely to depend on deal characteristics and industry conditions. We address the endogeneity of divestitures with a Heckman model (or treatment effect model). In the first-stage regression, we model Divestiture Dummy as a function of the acquirer s and target s size relative to the industry (AT Sales Ratio), industry concentration, and a dummy variable that equals one if the merger was challenged by antitrust authorities and zero otherwise. We hand-collect data on antitrust challenges by the FTC and DOJ by searching media sources for news coverage of the acquirer and the target with keywords such as antitrust, federal trade commission, justice, Hart-Scott-Rodino and so forth. Only 45 of the 1,153 deals in our sample (i.e., 4%) were explicitly challenged by the FTC or DOJ. 12 Column 1 of Panel B presents the first-stage regression coefficients. As expected, divestitures are more likely when the merging firms have larger market share and when antitrust authorities challenge the deal. Industry concentration, however, is statistically nonsignificant. Columns 2 5 present the second-stage regression results of a treatment effect model, in which the full sample is used and Divestiture Dummy is included in the regressions. Columns 6 12 Fee and Thomas (2004) find that 7% of the deals in their sample are challenged. Fee and Thomas require both the acquirer and the target to be public firms whereas our initial sample does not. On the other hand, we include only those merger-events that constitute 5% of total industry assets. If we also require both the acquirer and the target to be public, the ratio of challenged deals is 11%. 21
8 present the second-stage regression results of a Heckman model in which only mergers with accompanying divestitures are included. For the full-sample tests (Columns 2 5), Divestiture Dummy is significant in only one regression specification. However, the coefficient on Pct_SameInd_All remains significantly negative in all three regressions. For the tests based on the subsample of deals with divestitures (Columns 6 8), Pct_SameInd is also consistently negative and statistically significant. Although Divestiture Dummy becomes mostly insignificant after correcting for endogeneity, customer CARs move in lockstep with output prices with regards to who buys the divested assets. Output prices are higher when assets are divested within the industry and, consequently, customer CARs are lower. This finding supports the Market Power Hypothesis. 5.3. Acquirer and rival CARs Table 8 presents results for acquirers and Table 9 for nonbuying rivals. In Panel A of Table 8, we present OLS regressions of acquirer CARs. In Columns 1 4, we use the full sample, including deals. The coefficient on Divestiture Dummy is negative and statistically significant in two of the four regressions. Where significant, this result suggests that divestiture leads to lower acquirer CAR by 1.0 1.2 percentage points. As is evident from the summary predictions in Table 1, the lower stock price reaction of acquirers to the existence of a divestiture is consistent with both the Market Power Hypothesis and the Competitive Efficiencies Hypothesis. The coefficient on Pct_SameInd_All is negative in all regressions of Columns 2-4. When we restrict the sample to the subset of deals in which at least one divestiture occurred (Columns 5-7), the coefficient on Pct_SameInd is also negative. Taking the regression in Column 7, for example, the coefficient on Pct_SameInd is 0.032 (t-statistic of 2.41). Economically, an 22
increase in Pct_SameInd by one standard deviation (0.28) is accompanied by a decline in acquirer CARs of 0.89 percentage points (= 0.032*0.28). A negative coefficient on Pct_SameInd and Pct_SameInd_All supports the Competitive Efficiencies Hypothesis. That is, the acquirer is more concerned about an existing rival becoming larger and potentially more cost-efficient than it is about gaining market power relative to customers. In Panel B of Table 8, we control for the endogeneity of the divestiture decision by using a treatment effect or Heckman model. As in the case of customer CARs, the decision to divest is modeled in the first stage regression. 13 We focus on the second stage results. Columns 2 5 present the results of a treatment effect model based on the full sample. The coefficient on Divestiture Dummy is negative as predicted by both the Market Power and Competitive Efficiencies Hypotheses. Pct_SameInd_All is again negative. In Columns 6-8, we present a Heckman model for the subsample of mergers accompanied by a divestiture. Pct_SameInd is negative, meaning that acquirer stock returns are more negative when assets are purchased by a firm within the industry. This suggests that the disadvantage of competing with a larger rival overshadows the pricing power created vis-à-vis customers. 14 Thus, the coefficients on Pct_SameInd and Pct_SameInd_All are supportive of the Competitive Efficiencies Hypothesis. In Table 9, we present results for rivals. Panel A of Table 9 presents OLS regressions and Panel B presents treatment effect and Heckman self-selection models. We see that the coefficient on Divestiture Dummy is always positive in both panels and statistically significant after 13 The first-stage results in Column 1 are slightly different from those in Column 1 of Table 7 because the sample of nonmissing acquirer CARs is different from the sample of nonmissing customers CARs. 14 A reasonable question that arises here is why an acquirer would sell assets to an existing rival instead of to a firm outside the industry if doing the former creates a more difficult competitive environment and more negative stock returns. Previous research suggests that typically only one potential buyer actively negotiates to purchase divested assets (Jain, 1985; Sicherman and Pettway, 1987). Thus, once an acquirer decides to divest assets, it may not have much flexibility in whom it sells to. 23
controlling for endogeneity in Panel B. This is supportive of the Competitive Efficiencies Hypothesis. The benefits of limiting scale economies and other cost efficiencies of the merging firm appear to be more important for the rival than any pricing power that the merger might bring to the industry. Table 9 also shows a negative coefficient on Pct_SameInd and Pct_SameInd_All. Taking the regression in Column 7 of Panel A, for example, the coefficient on Pct_SameInd is 0.011 (tstatistic of 2.10). Economically, an increase in Pct_SameInd of one standard deviation (0.30) is associated with a decline in rival CARs of 0.33 percentage point (= 0.011*0.30). Thus, rival firms who do not buy the assets divested by the merging firms react negatively when an existing competitor buys the divested assets. The disadvantage of having to compete with larger, potentially more cost-effective rivals outweighs any benefits that might arise from industry market power. A key finding of prior studies that support the Competitive Efficiencies Hypothesis rather than the Market Power Hypothesis is that rivals stock prices react positively when a merger is challenged by regulators (Eckbo, 1983; Stillman, 1983; Eckbo and Wier, 1985; Fee and Thomas, 2004). This suggests that instead of worrying that industry participants will lose market power vis-à-vis customers, rival firms are relieved that a larger competitor might not be created. Our results are consistent with these prior studies in suggesting that rivals are more concerned about competition among industry peers than about selling power versus customers. However, none of these findings rule out the possibility that market power is created. 6. Subsample tests 24