Strategic Effects of Horizontal Merger Synergies on. Rivals, Customers, and Suppliers

Transcription

1 Strategic Effects of Horizontal Merger Synergies on Rivals, Customers, and Suppliers Abstract We examine the impact of merger-related operating synergies on firms that have output or input market linkages with companies engaged in horizontal mergers. Using a unique dataset of synergies forecasts by management, we find empirical support for a model that produces unambiguous predictions about the effects of merger synergies on firms directly or indirectly linked to merging firms. Our evidence shows for the first time that horizontal merger synergies have strategic effects on the valuation and operating performance of merging firms rivals, customers, and suppliers. Overall, our results highlight the importance of inter-firm linkages for firms strategies, valuations, and performance. Keywords: horizontal mergers, synergies, competitors, customers, suppliers. JEL Classification Numbers: G34, L13.

2 It was Galileo s longing for a mechanical proof of the motion of the earth which misled him into formulating a wrong theory of the tides. Albert Einstein, foreword to Dialogue Concerning the Two Chief World Systems by Galileo Galilei, Introduction Mergers and acquisitions may affect not only the merging firms, but also firms that are strategically linked with the merged entity: firms competing with the merged company in product markets (rivals) as well as those operating along its supply chain (corporate customers and corporate suppliers). There are two main channels through which horizontal mergers would affect rivals, corporate customers, and suppliers. The first channel is through operating synergies that arise from the merger. These can take the form of operating efficiencies (e.g., economies of scale) and/or cost savings (e.g., elimination of duplicate functions and facilities). The synergy hypothesis (e.g., Bradley, Desai and Kim (1988), Kaplan (2000), Maksimovic and Phillips (2001), Jovanovic and Rousseau (2002), and Lambrecht (2004)) implies that merger-related gains give the merged entity a competitive advantage over its rivals, reducing their profits and, thus, valuations. Moreover, synergies change the merging firms supply function and demand for inputs, and in turn affect the profits and valuations of corporate customers and suppliers, respectively. The second channel is through changes in the merged entity s industry structure. According to the market power (collusion) hypothesis, horizontal mergers reduce the intensity of competition in the merged firm s industry and raise economic rents for the industry as a whole (e.g., Stigler (1964)). In addition, increased collusion among industry participants raises prices for the industry output, hurting firms in downstream industries (customers), and reduces the demand for inputs, hurting firms in upstream industries (suppliers). In this paper we examine the implications of the synergy hypothesis by investigating the effects of operating synergies on merging firms product market rivals, customers, and suppliers. To guide our empirical tests, we begin by developing a simple model of oligopolistic competition that characterizes the effects of two types of merger-related synergies i.e., operating efficiencies and cost savings on merging firms rivals, customers, and suppliers. This model generates a set of unambiguous empirical predictions about the effects of the two types of synergies on firms operating along the merging firms supply chain, which guide our empirical investigation. The main contributions of our study are empirical. First, likely due to the lack of readily usable data, there is no existing study that tests directly the implications of the synergy hypothesis for 1

3 merging firms rivals, customers, and suppliers. We fill this gap in the literature by using a novel dataset of synergies forecasts disclosed by executives around merger announcements. 1 These forecasts reflect projections of anticipated operating efficiencies and/or cost savings that management discloses around merger announcements. Although these projections need not be informative, 2 the evidence indicates that they are economically relevant. Bauguess and Bernile (2011) show that projected merger gains are a significant determinant of investors and analysts reactions to mergers, as well as of the merged entities post-merger changes in operating performance. Second, we depart from the static definitions of industries used in most examinations of the effects of horizontal mergers on firms that operate along the merged entities supply chains (e.g., Shahrur (2005), Fan and Goyal (2006), Ahern (2012), and Ahern and Harford (2012)). This is important because standard industry classifications often do not correctly capture firms interactions in output markets, as they are based on production processes rather than products. Equally important, industry definitions and affiliations are seldom adjusted in response to the evolution of industries and to firms transitions across industries (e.g., Hoberg and Phillips (2011)). For these reasons, in most of our analysis, we employ Hoberg and Phillips (2010a, 2010b, 2011) Text-based Network Industry Classification (TNIC) to identify firms whose product offerings are similar and that are therefore likely to actually compete in product markets. Accordingly, we classify mergers between firms with related product offerings (i.e., firms in the same product network) as horizontal. Third, our main analysis relies on the identification of actual supply chain relations. This identification is based on firms disclosures of the identity of their (large) corporate customers, which are available in Compustat Customer Segment files. This is different from the more common approach based on industry-level input-output relations derived from the make and use tables provided by the Bureau of Economic Analysis, BEA henceforth (e.g., Shahrur (2005) and Ahern (2012)). In sum, unlike most existing studies, we focus on the effects of explicitly identified merger synergies on actual, rather than potential, rivals, customers, and suppliers. This unique experimental design produces several important results, which show for the first time that horizontal mergers have substantial strategic consequences. First, consistent with the model s predictions, we find that merger-related synergies have a statistically and economically significant impact on the operating performance and valuations of firms that are directly linked to the merging firms: their rivals, customers, and suppliers. Specifically, rivals announcement returns and post- 1 This dataset was first employed in Bauguess and Bernile (2011), and more recently in Bernile, Lyandres, and Zhdanov (2012) and Schneider and Spalt (2012). 2 Insiders may not provide truthful projections of merger synergies because of self-interested behavior (e.g., Jensen (1986), Morck, Shleifer and Vishny (1990) and Gorton, Kahl and Rosen (2005)), hubris (e.g., Roll (1986)), and market misvaluations (e.g., Shleifer and Vishny (2003) and Rhodes-Kropf and Vishwanathan (2004)). 2

4 merger operating performance are negatively related to projected merger-related synergies, while the returns and operating performance of merging firms corporate customers are positively related to synergies. Moreover, synergies have a negative impact on corporate suppliers of merging firms, but only when synergies take the form of operating efficiencies as opposed to cost savings due to elimination of duplicate functions/facilities. Second, our paper is the first to examine the effects of mergers along the supply chains of merging firms and of their rivals separately. This analysis allows to further differentiate the strategic effects of synergies from the effects of market power and information revelation. Given that mergers can lead to the terminations of relationships with customers and suppliers (e.g., Karceski, Ongena and Smith (2005)), our model implies that switching costs would determine the effects of horizontal merger synergies on customers and suppliers of merging firms rivals. Consistent with these predictions, we find that customers and suppliers of merging firms rivals operating in industries with relatively high switching costs are hurt by merger synergies, while this is not the case in industries with relatively low switching costs. Overall, our study provides direct evidence that merger synergies-induced changes in firms operating strategies affect not only the merging firms, but also firms that strategically interact with them in output and input markets. These novel findings make a unique contribution to the existing literature on the effects of horizontal mergers on economically linked firms. This literature starts in the eighties with the analyses of stock returns of merging firms rivals around announcements of horizontal mergers and associated antitrust challenges (e.g., Eckbo (1983, 1985) and Stillman (1983)). 3 The general conclusion from these early studies is that horizontal mergers affect rivals mainly through an information channel, 4 with no evidence that would support the existence of merger-related strategic effects through either the synergy or the market power channels. More recently, Fee and Thomas (2004) and Shahrur (2005) examine the effects of horizontal mergers on firms operating along merging firms supply chains, in addition to the merged entity s rivals. These studies conclude that there is evidence consistent with improved production efficiencies (Fee and Thomas (2004)) and that the average acquisition... is driven by efficiency considerations (Shahrur (2005)). However, these studies are plagued with identification problems and in fact provide no evidence that changes in product market strategies due to horizontal merger synergies affect companies in the merging firms output and input markets. 3 In addition, Banerjee and Eckard (1998) examine announcement returns of rivals of firms involved in mega-mergers at the turn of twentieth century and Singal (1996) examines the effects of horizontal mergers in the airline industry on rival airlines. 4 The positive [merger] proposal-related performance shown by the rivals of the challenged mergers reflects discounted future cost-savings made possible by information signaled through the proposal announcement (Eckbo (1983)). 3

5 Fee and Thomas (2004) postulate their hypotheses (and thus base their conclusions) on the recognition that merger-related changes in market power should affect customers and suppliers, but ignore the fact that synergies also affect the strategic interactions between merging firms and their rivals, customers, and suppliers. 5 They find that mean returns of the merging firms customers and suppliers around merger announcements are insignificant, while mean combined returns of the merging firms are positive. They infer from this evidence that market power considerations are not significant and argue by exclusion that the merging firms positive returns must be due to synergies. In fact, however, merger synergies and market power have opposite effects on merging firms rivals, customers, and suppliers, even though both have positive effects on the merging firms. This identification problem prevents reaching conclusions regarding the strategic effects of horizontal mergers along the supply chain based on the observation of insignificant announcement returns to customers and suppliers. Similar to Fee and Thomas (2004), Shahrur (2005) documents close-to-zero mean returns for rivals, customers, and suppliers around horizontal merger announcements. In addition, he finds that mean returns of rivals, customers, and suppliers are positive when the merging firms combined returns are positive, and concludes that this is due to merger-related synergies. However, in theory, the strategic effect of merger synergies on rivals values is negative, contrary to the evidence reported in Shahrur (2005). Rather, the positive returns of rivals, customers, and suppliers are consistent with a positive information spillover effect that leads to the positive revaluation of those firms. 6 In the spirit of the opening quote to our paper, it is important to examine whether conclusions from past studies that are based on misidentified tests hold when using theoretically justified hypotheses, tests, and inferences. With this objective in mind, we analyze the strategic effects of horizontal merger synergies while designing our empirical tests based on a well-defined theory. In the next section, we present a stylized model of oligopolistic competition that characterizes the effects of two types of merger synergies (i.e. operating efficiencies and cost savings) on merging firms rivals, and on customers and suppliers of merging firms and of their rivals. In Section 3, we present the data, describe our empirical strategy, and discuss our measure of merger synergies based on management projections. In Section 4, we discuss the results of our empirical tests and, in Section 5, we summarize our analysis and conclude. Appendix A contains proofs and Appendices B-D present additional details about our empirical methods. 5 See Table 1 in Fee and Thomas (2004), which summarizes the empirical predictions following from various hypotheses. 6 In fact, another finding reported but not emphasized in Shahrur (2005) supports this inference: when the merging firms combined returns are negative, the returns of rivals, customers, and suppliers are also negative. 4

6 2. A model of horizontal mergers with production efficiencies In this section we provide a simple illustration of the effects of horizontal mergers, which increase the operating efficiency of the merging firms and/or result in production cost savings, on the merging firms, on firms operating in the merging firms industries (product market rivals), as well as on firms operating in upstream and downstream industries (corporate suppliers and customers respectively). Assume that an industry consists of n firms, which compete à la Cournot in homogenous products. 7 The industry demand is given by p(q) = a bq = a b n q i, (1) where p(q) is the equilibrium output price, Q is the combined output of n firms, each of which produces q i. Firms production functions are of the Cobb-Douglas specification with two inputs (fixed input K and variable input I): i=1 q i = K α i I 1 2 i. (2) This production function exhibits decreasing (constant, increasing) returns to scale when α < 1 2 (α = 1 2, α > 1 2 ) and, therefore, α is a convenient measure of production efficiencies obtained by joining capital. The cost of employing one unit of variable input I by firm i is C(I i, K i ) = c K β i I i. (3) The cost of a unit of variable input is assumed to be a decreasing function of firm s capital (fixed input) in order to enable modeling of cost savings due to consolidation (e.g., elimination of duplicate functions and overhead costs), which are cited as one of the most important reasons for merging (e.g., Kaplan (2000)). Firm i s variable cost of producing q i units of output is, therefore, C(q i, K i ) = q2 i c K 2α+β i. (4) Maximizing each firm s profit, π i, given by π i = pq i C(q i, K i ), (5) 7 The assumption of homogenous products is made to simplify the exposition of the main comparative statics. All of the results hold for competition in heterogenous products. The results are also robust to assuming Bertrand competition in heterogenous products. 5

7 with respect to the firm s output quantity results in firms optimal output levels as functions of their rivals outputs. Solving the resulting system of n equations results in the following expressions for firm i s equilibrium quantity produced, qi, quantity of variable input used, I i, equilibrium profit earned, π i, and equilibrium output price, p : q i = I i = π i = (bγ + 1) a ( ), (6) b + 2c K 2α+β i a 2 ( ) 2, (7) Ki 2α (bγ + 1) 2 b + 2c K 2α+β i a 2 K β i (b + c K 2α+β i (bγ + 1) (b 2 + 2c K 2α+β i abn p = a (bγ + 1) ( b + ) ) 2, (8) ), (9) 2c K 2α+β i where Γ = n j=1 ( b + 1 2c K 2α+β j ), is common to all firms. In what follows, in order to simplify the exposition, we put more structure on the model and assume that all firms are identical and have the same level of capital, i.e. K i = K j i j. In addition, in order to simplify the algebra, we assume without loss of generality that K i = 1 i, 8 resulting in the variable cost of producing q i units of output equal to C(q i ) = q 2 i c. (10) Under these assumptions, firms first-order conditions are identical and the equilibrium is symmetric. It is characterized by the following equilibrium firm profits, π i,no merger, firm-level output, q i,no merger, total industry output, Q no merger, equilibrium output price, p no merger, firm-level required quantity of 8 This assumption is innocuous and all of the results hold for K i 1. 6

8 variable input, I i,no merger, and total industry-wide required quantity of variable input, I no merger: πi,no merger = qi,no merger = Q no merger = p no merger = Ii,no merger = I no merger = a 2 (b + c) (b(n + 1) + 2c) 2, (11) a b(n + 1) + 2c, (12) na b(n + 1) + 2c, (13) a(b + 2c) b(n + 1) + 2c, (14) a 2 (b(n + 1) + 2c) 2, (15) na 2 (b(n + 1) + 2c) 2. (16) Assume now that two of the n firms merge and combine capital. producing q i units of output becomes The merged firm s cost of where κ α = 2 2α 1 (i.e. C(q i ) = q 2 i c (2K i ) 2α+β = q2 i c 2 2α+β = q2 i c 2κ α κ β, (17) κ α > 1 (κ α < 1) in case of increasing (decreasing) returns to scale) and κ β = 2 β. The total cost of producing q i units of output in (17) is clearly decreasing in κ α as long as the production function exhibits increasing returns to scale (α > 1 2 ) and it is decreasing in κ β if the per-unit overhead cost is decreasing in the level of fixed input (β > 0). Solving for the merged firm s and n 2 stand-alone firms (rivals ) optimal output levels results in the following equilibrium merged firm s profit, π merged, each of the n 2 rivals profits, π i,st. alone, merging firms and each of their rivals equilibrium output levels, q merged and q i,st.alone respectively, total industry output, Q merger, equilibrium output price, p merger, and total quantity of input demanded by the merging firms, each of their rivals, and the industry as a whole, I merged, I i,st.alone, and I merger 7

9 respectively: π merged = π i,st. alone = q merged = q i,st.alone = a 2 κ α κ β (b + 2c) 2 (bκ α κ β c) (b(bn + 4c)κ α κ β + c(b(n 1) + 2c)) 2, (18) a 2 (b + c)(bκ α κ β + c) 2 (b(bn + 4c)κ α κ β + c(b(n 1) + 2c)) 2, (19) a(b + 2c)κ α κ β b(bn + 4c)κ α κ β + c(b(n 1) + 2c), (20) a(bκ α κ β + c) b(bn + 4c)κ α κ β + c(b(n 1) + 2c), (21) Q merger = a((b(n 1) + c)κ ακ β + c(n 2)) b(bn + 4c)κ α κ β + c(b(n 1) + 2c), (22) p merger = I merged = I i,st.alone = a(b + 2c)(bκ α κ β + c) b(bn + 4c)κ α κ β + c(b(n 1) + 2c), (23) 1 2 a2 (b + 2c) 2 κ α κ 2 β (b(bn + 4c)κ α κ β + c(b(n 1) + 2c)) 2, (24) a 2 (bκ α κ β + c) 2 (b(bn + 4c)κ α κ β + c(b(n 1) + 2c)) 2, (25) I merger = a2 ( 1 2 (b + 2c)2 κ α κ 2 β + (n 2)(bκ ακ β + c) 2 ) (b(bn + 4c)κ α κ β + c(b(n 1) + 2c)) 2. (26) In what follows we examine the differences between post-merger profits, output levels, input levels, and prices in (18)-(26) and the corresponding quantities in (11)-(16) to examine the effects of production efficiencies and cost savings on the merging firms, their product market rivals, and firms operating in downstream and upstream industries (customers and suppliers). We begin by examining the effects of a merger on the output of merging firms, of their product market rivals, and total industry output. Proposition 1. Following a horizontal merger, 1) the relative change in merging firms output, and in cost savings (κ β ); q merged 2q i,no merger, is increasing in operating efficiencies (κ α ) 2) the relative change in combined output of product market rivals of merging firms, decreasing in κ α and in κ β ; 3) the relative change in industry output, Q merger Q no merger, is increasing in κ α and in κ β. qi,st.alone qi,no merger Both operating efficiencies (κ α ) and cost savings (κ β ) increase the equilibrium output levels of the merging firms, ceteris paribus, because both types of synergies reduce the marginal costs of production for a given level of output. Industry rivals react optimally to the increased output of the merging firms by reducing their output, however this effect is second-order relative to the effect of the increased output by the merged entity, leading to an overall positive relation between both types of merger synergies and total industry output., is 8

10 A change in equilibrium industry output has a direct effect on firms operating in downstream industries (customers). As a whole, customers benefit (are hurt) if, for a given price for the merging firms industry output, the equilibrium supply of industry output increases (decreases), since it is used as an input in the customer industry. Thus, as follows from the third part of Proposition 1, the change in the overall value of firms in the customer industry is positively related to both types of merger synergies. However, the effects of operating efficiencies and cost savings may vary across customer firms. If the costs to a customer of switching from one supplier to another are high, then the effects of horizontal merger synergies on customers of the merging firms and those on customers of the merging firms rivals would be very different. In particular, customers of merging firms would benefit from higher merger synergies, while customers of merging firms rivals, whose equilibrium output is negatively related to merger synergies, would be hurt by these synergies. If, on the other hand, the switching costs in the customer industry are low then the effects of merger synergies would be similar for all firms operating in the customer industry. In particular, all firms that use the output of the merging firms and their rivals as their input, would benefit from both types of merger synergies. Next, we examine the effects of a horizontal merger on the total quantity of variable input employed by the merging firms, by their output market rivals, and by the merging firms industry as a whole. Proposition 2. Following a horizontal merger, 1) the relative change in the quantity of variable input employed by merging firms, I merged 2I i,no merger, is a) decreasing in κ α if the overall output of the merging firms has declined following the merger, and b) is increasing in κ β ; 2) the relative change in combined quantity of variable input employed by merging firms product market rivals, Ii,st.alone Ii,no merger, is decreasing in κ α and in κ β ; 3) the relative change in total industry quantity of variable input, I merger I no merger, is a) decreasing in κ α if the overall output of merging firms has declined following the merger, and b) is increasing in κ β if κ α is sufficiently low, and is decreasing in κ β if κ α is sufficiently high. The effects of the two types of merger synergies on the merging firms and their rivals demand for variable input are somewhat more subtle than the effects on the equilibrium industry output. The reason is that abstracting from market power considerations, the two types of merger synergies have two distinct effects on the quantity of variable input demanded by the industry firms. Operating efficiencies lead to higher equilibrium output of merged firms. However, a lower quantity of variable input is required to produce any given level of output. For a typical merger that antitrust authorities are concerned with, the overall output of the merging firms declines following the merger (e.g., Stigler (1950), Salant, Switzer and Reynolds (1983), Perry and Porter (1985)). It turns out that for such 9

11 mergers the second effect dominates the first one, and the quantity of input employed by the merging firms is decreasing in operating efficiencies even as the output quantity is increasing in them. On the other hand, reduction of overhead costs ( κ β > 0) does not affect the quantity of variable input required to produce a given output but lowers the marginal cost of the variable input. Thus, higher κ β leads to higher output (and the quantity of variable input demanded) by the merging firms. The positive effect of κ α and κ β on the optimal output by merging firms results in the negative effect of the two types of synergies on the optimal output and, importantly, input levels of the merging firms rivals, as discussed in the second part of Proposition 2. Since the relations between κ α on one hand and the quantity of input employed by both the merging firms and their rivals on the other hand are negative in the case in which merging firms output declines post-merger, the relation between operating efficiencies and the total quantity of input employed by the merging firms industry is negative as well. For relatively low production efficiencies (κ α ), the effect of κ β on the quantity of input required by the merging firms dominates the effect of κ β on the quantity of input employed by the merging firms rivals, leading to the positive relation between κ β and the overall industry demand for variable input. However, for sufficiently high κ α, the merging firms production function is sufficiently efficient and its demand for variable input is low, leading to the negative effect of cost savings on the rivals demand for input to dominate the positive effect of κ β on the quantity of input employed by the merging firms. In this case, the relation between κ β and overall industry demand for variable input is negative. Note that the relations in Proposition 2 are conditional on the price of the variable input being constant. In an equilibrium that accounts for the optimal response of upstream firms (suppliers) to the merger, which is outside of the scope of our stylized model, decreased (increased) demand for variable input (which is the output of firms operating in the upstream industry) would reduce (raise) the equilibrium variable input price and mitigate the strength of the relations in Proposition 2. However, this effect is second-order and it cannot reverse the signs of the relations between κ α and κ β on one hand and the quantity of input employed by the firms in the merging firms industry on the other hand. As a whole, firms in upstream industries are hurt by operating efficiencies, while they may benefit from merger-related cost savings if operating efficiencies are relatively low. However, as in the case of downstream industries, the effects of operating efficiencies and cost savings may vary across supplier firms. If the costs to a supplier of switching from one firm in the merging firms industry to another are high, then suppliers of the merging firms are expected to be hurt by higher operating efficiencies and are expected to benefit from cost savings. The suppliers of merging firms rivals would be hurt 10

12 by both types of synergies. If, on the other hand, the switching costs in the supplier industry are low then all firms that belong to the supplier industry would be hurt by higher operating efficiencies and may benefit from higher cost savings associated with the merger if operating efficiencies are relatively low. Examining the effects of a merger on the equilibrium output price in the merging firms industry, given in (23), leads to the following result: Proposition 3. Following a horizontal merger, the relative change in the output price in the merging firms industry, p merger p no merger, is decreasing in operating efficiencies (κ α) and in cost savings (κ β ). This result is quite intuitive. Both operating efficiencies and cost savings are associated with higher equilibrium output in the merging firms industries, which immediately leads to negative relations between κ α and κ β on one hand and the equilibrium output price in the merging firms industry on the other hand. Examining the effects of the two types of synergies on the merging firm s equilibrium value (profit) in (18) leads to the following intuitive result: Proposition 4. Following a horizontal merger, the change in the combined value of merging firms, π merged 2π no merger, is increasing in operating efficiencies (κ α) and in cost savings (κ β ). The intuition is, of course, that both production efficiencies and cost savings reduce production costs for any given level of output. Reduced production costs increase output, as demonstrated in Proposition 1, and increase equilibrium combined profit of the merging firms. While the theoretical result in Proposition 4 is not surprising, the resulting empirical implication is useful as a test of the validity of our proxy for merger synergies that we discuss in the next section. The effects of κ α and κ β on the change in the values of merging firms product market rivals following the merger, leads to the following result: Proposition 5. πi,st.alone πi,no merger Following a horizontal merger, the change in the values of merging firms rivals,, is decreasing in κ α and in κ β. Both operating efficiencies and cost savings increase the combined output of the merging firms, hurting their product market rivals. Thus, the change in rivals values following a horizontal merger announcement is decreasing in both types of synergies. The following table summarizes the results in Propositions

13 Table 1: Summary of comparative statics Effect of efficiency (κ α ) Effect of cost savings (κ β ) Merging firms + + Competitors Customers + + Customers of rivals + for low customer switching costs for high customer switching costs + for low customer switching costs for high customer switching costs Suppliers + Suppliers of rivals + for low supplier switching costs for high supplier switching costs Industry total output + + Industry output price The comparative statics of the model lead to the following empirical predictions. Prediction 1 Merging Firms: The announcement returns of merging firms in horizontal deals and the post-merger change in their operating performance are expected to be positively related to merger synergies. Prediction 2 Merging Firms Rivals: The announcement returns of merging firms product market rivals in horizontal deals and the post-merger change in their operating performance are expected to be negatively related to merger synergies. Prediction 3 Merging Firms Customers: The announcement returns of merging firms corporate customers in horizontal deals and the post-merger change in their operating performance are expected to be positively related to merger synergies. Prediction 4 Customers of Merging Firms Rivals: The announcement returns of corporate customers of merging firms rivals in horizontal mergers and the post-merger change in their operating performance are expected to be negatively related to merger synergies in industries with high customer switching costs and may be positively related to merger synergies in industries with low customer switching costs. 12

14 Prediction 5 Merging Firms Suppliers: The announcement returns of merging firms corporate suppliers in horizontal mergers and the post-merger change in their operating performance are expected to be negatively related to merger synergies, if those take the form of operating efficiencies, and to be positively related to merger synergies, if those take the form of cost savings. Prediction 6 Suppliers of Merging Firms Rivals: The announcement returns of corporate suppliers of merging firms rivals in horizontal mergers and the post-merger change in their operating performance are expected to be negatively related to merger synergies in industries with high supplier switching costs and may be positively related to merger synergies in industries with low supplier switching costs. Prediction 7 Output Quantity: The change in industry total output following a horizontal merger announcement is expected to be positively related to merger synergies. Prediction 8 Output Price: The change in the output price in the merging firms industry following a horizontal merger announcement is expected to be negatively related to merger synergies. 3. Data The data for our empirical tests are from multiple sources. We obtain the sample of mergers and acquisitions from Thomson Reuters Securities Data Company (SDC). We retrieve insiders forecasts of merger-related synergies from a web-based search of all English-language news stories and press releases available on the Dow Jones Factiva database. 9 Using the Text-based Network Industry Classification (TNIC) in the Hoberg-Phillips data library, we identify firms competing in the same product markets. 10 We rely on the firm-level customer-supplier linkages developed in Cohen and Frazzini (2008) using the Compustat Customer Segment files to identify actual supply chain relations. 11 We construct merger announcement returns and post-merger changes in operating performance using data from the Center for Research in Securities Prices (CRSP) and Compustat Annual Industrial Files, respectively. We also rely on Compustat data to construct some of the explanatory variables for the propensity of insiders to disclose synergy projections. The determinants of the propensity to provide projections include: managers past propensity to provide earnings guidance, from Thomson Reuters First Call Historical Database (FSHD); analyst measures (e.g., following, dispersion, recommendations), from 9 Appendixes B and C provide further details regarding the web-based search. 10 We are grateful to Gerard Hoberg and Gordon Phillips for generously sharing their data (available at 11 We are grateful to Lauren Cohen and Andrea Frazzini for generously sharing their data. 13

15 Thomson Reuters I/B/E/S database; and institutional investor presence, from Thomson Reuters 13-F database. We obtain industry-level output quantity and cost data in the Industry Productivity and Costs (IP) files, and pricing data in the Producer Price Index (PC) files compiled by Bureau of Labor Statistics (BLS). Finally, we identify industry-level input-output relations using the annual make and use tables compiled by the Bureau of Economic Analysis (BEA). We discuss these data in more detail below Sample of mergers and acquisitions We collect an initial sample of mergers and acquisitions that includes proposed deals between companies listed on NYSE, NASDAQ, or AMEX, announced between January 1, 1990 and December 31, 2005, available in the Securities Data Company (SDC) database. We restrict the initial sample to acquisitions of assets, acquisitions of majority interest, or mergers in which a bidder holds less than 50% of target s common stock on the offer date and the deal has come to a resolution (i.e., completed, unconditional, or withdrawn). This yields an initial sample of 4,017 merger offers. Eighty-two deals are further eliminated for the purpose of our analysis, either because the proposed deal involves more than two companies (i.e. our sample excludes M&A offers by the same bidder that are less than two days apart) or because the observation is a later bid for the same target by the same bidder occurring less than 365 days after the earlier bid. The resulting sample comprises 3,935 unique M&A offers. The sample that we use in the empirical analysis is further restricted due to data availability from several other sources CRSP, Compustat, First Call, I/B/E/S, and Thomson Reuters 13-F filings databases. Most notably, due to the sparseness of management earnings guidance in the First Call database prior to 1994, we follow other studies and restrict the sample to deals announced after 1994 (e.g., Feng and McVay (2010)) i.e., In addition, we require each deal to have the necessary data to estimate the selection model of availability of insiders synergy projections, presented in Table A1 in the Appendix and discussed below, resulting in the final sample of 2,723 unique M&A offers Identifying product market linkages: horizontal mergers and rivals Horizontal mergers are the focus of our analysis. Most studies examining the effects of M&A s on rivals, customers, and suppliers relies on industry classifications based on SIC or NAICS codes (e.g., Shahrur (2005), Fan and Goyal (2006), Ahern (2012), and Ahern and Harford (2012)) to identify product and input market linkages. However, such classifications can be inaccurate for at least three reasons. First, SIC and NAICS classifications are based on production processes, not on products that firms supply. Second, these classifications are static, in the sense that they are rarely adjusted over 14

16 time in the face of evolving product markets and/or firms entering different industries. 12 Third, SIC and NAICS classifications impose transitivity, while it is possible that two competing firms may have different product market rivals. For these reasons, in our firm-level analysis, we adopt Hoberg and Phillips (2010a, 2010b, 2011) Text-based Network Industry Classification (TNIC) to identify firms whose product offerings are similar. The TNIC classification is firm-centric and uses text-based analysis of descriptions of firms products in annual 10-K statements. Firms that exhibit strong correlation in product market vocabulary are assigned to the same product network. 13 If two firms are in the same product network prior to a merger offer according to the TNIC scheme, we classify the deal between them as a horizontal merger, Horizontal Ho-Ph. By the same logic, we identify as rivals all firms that are in the bidder or target s pre-merger product networks. We also conduct a series of industry-level tests that (must) rely on traditional, static industry classifications. For the purpose of this analysis, we classify a merger as horizontal if the firms involved in the deal share the first four digits of their primary NAICS code, Horizontal NAICS4. By the same logic, we identify as rivals all other non merging firms that derive at least 80% of their sales from the corresponding 4-digit NAICS segment, as reported in Compustat Business Segment files Identifying supply chain relationships Our main goal is to examine the effects of horizontal merger synergies on merging firms rivals as well as on firms operating along the supply chains of the merging firms and of their rivals (i.e., corporate customers and suppliers of merging firms and of their rivals). Therefore, identifying customer-supplier relations is crucial for our tests. Following Fee and Thomas (2004), Cohen and Frazzini (2008), and Hertzel, Officer and Rogers (2008), we use the disclosure of significant corporate customers available on Compustat Segment Files to identify firm-level supplier-customer relations. SFAS 131 requires firms to report the identity of customers that account for at least 10 percent of the firm s sales. The sample of corporate customers is obtained by matching the principal customers on the Compustat Segment files with firms on the CRSP/Compustat tape. 14 Then, if firm i is a principal customer of firm j, we classify firm j as being a supplier of firm i. This procedure provides a set of actual firm-level supplier-customer relations. 12 As an example of the static nature of existing industry classification schemes, hundreds of new technology and webrelated firms were assigned to the large and nondescript SIC-based business services industry through the nineties (see Hoberg and Phillips (2011)). 13 See Hoberg and Phillips (2010a, 2010b, and 2011) for details on the computation of the degree of similarity of product descriptions and the resulting product networks. 14 The detailed description of the matching algorithm is available in Cohen and Frazzini (2008). 15

17 To identify NAICS industry-level customer-supplier relations, we rely on the benchmark inputoutput (IO) tables constructed by the Bureau of Economic Analysis of the U.S. Department of Commerce. This approach is quite common in empirical studies of the effects of corporate events on firms customers and suppliers (e.g., Shahrur (2005), Fan and Goyal (2006), and Ahern (2011) and Ahern and Harford (2011) for the case of M&As, Fan and Lang (2000) for the case of corporate diversification, and Chang and Tsai (2009) for the case of strategic alliances). There are 61 four-digit NAICS industries in the sample. For each pair of industries, i and j, we compute the following ratios, which we call customer and supplier shares, Cust share i,j = Sales i,j, and Supp share i,j = Sales i Sales i,j P urchases j, where Sales i,j is the dollar amount of output of industry i sold to industry j, and Sales i (P urchases j ) is the total dollar value of sales (purchases) of industry i (j). We rely on these ratios to identify each industry s most important customer and supplier industries. Following Shahrur (2005), we define industry j as the most important customer of industry i if Cust share i,j > 5% and Cust share i,j > Cust share i,k k j. Similarly, we define industry i as the most important supplier of industry j if Supp share i,j > 5% and Supp share i,j > Supp share i,k k j. As for the NAICS-based rivals, when constructing industry portfolios of customers and suppliers, we require that the customer/supplier firm derives at least 80% of its sales from the corresponding 4-digit NAICS segment, as reported in Compustat Business Segment files Projections of merger synergies Numerous studies highlight the importance of synergies in merger decisions (see Kaplan (2000) for an extensive review of the synergy motive for mergers). Most commonly, researchers attempt to infer the existence of merger synergies from short-term merger announcement returns, and postmerger long-term stock returns and operating performance. However, drawing inferences regarding the existence and magnitude of merger synergies from such tests is difficult at best, because synergies are not the sole driver of changes in merging firms values and operating performance. The release of ancillary firm- and industry-level information associated with merger announcements will affect valuations and returns. Moreover, changes in industry structure will also impact valuations and operating performance measures (e.g., Bradley, Desai and Kim (1988), Kaplan (2000), Maksimovic and Phillips (2001), Houston, James and Ryngaert (2001), and Jovanovic and Rousseau (2002)). Therefore, using these contaminated measures to draw inferences about the strategic effects of merger synergies on rivals, customers, and suppliers poses a serious inference problem. We overcome this inference problem by employing a direct measure of merger synergies, which was 15 Our industry-level empirical results hold for alternative cutoffs (70, 90, 100%) of important customer and supplier industries. These results are available upon request. 16

18 first used in Houston, James and Ryngaert (2001) in a small sample of bank mergers and later extended in Bauguess and Bernile (2011) to a large M&A sample. This measure is based on projections of synergistic gains from proposed mergers, which merging firms insiders disclose upon merger announcements. Using close to one thousand horizontal and non-horizontal deals, Bauguess and Bernile show that these projected gains explain the merging firms deal announcement returns as well as their post-merger stock and operating performance. In this paper, we use the projections disclosed in approximately five hundred horizontal mergers to examine the strategic effects of synergies on the merging firms rivals, customers, and suppliers. For each deal, we collect news stories and press releases published between the offer announcement and its ultimate resolution. When available, we retrieve any merger-related forecasts of cost savings and operating efficiency gains provided by the merging firms management. The news stories and press releases are from all English-language sources on Factiva and satisfy the criteria specified in Appendix B. Figure 1 shows the annual frequency of mergers in our sample period (leftmost bar in each year), the number of Horizontal Ho-Ph or NAICS4 deals (middle and rightmost bars in each year, respectively), and the overall fraction of mergers with synergies projections as well as the fraction among horizontal mergers. These latter fractions are computed on an equal-weighted or on a value-weighted basis. Insert Figure 1 here Figure 1 is consistent with the notion that mergers occur in waves (e.g., Mitchell and Mulherin (1997), Harford (2005), and Rhodes-Kropf, Robinson and Viswanathan (2005)). The aggregate volume measured by the number of deals rises during the second half of the nineties, reaching its peak toward the end of the decade, followed by a notable decline. Over fifty percent of mergers are classified as horizontal according to the TNIC scheme, while sixty percent are between firms in the same four-digit NAICS industry. Synergies forecasts are disclosed in 785 cases out of 2,723 examined. This represents 28.8% of the sample on an equal-weighted basis. The evolution of this proportion for all mergers and horizontal mergers is depicted by the solid lines in Figure 1. Importantly, synergy projections are available in mergers that amount to over two-thirds of aggregate assets targeted by merger offers, as depicted by the dashed lines in Figure 1. Thus, while we lose a large fraction of the sample due to unavailability of projections, important mergers involving relatively large firms that can have significant consequences along the supply chain are typically included in our sample. To estimate the value of projected synergies, we use the discounted cash flow method, similar to prior studies (e.g., Bauguess and Bernile (2011), Gilson, Hotchkiss and Ruback (2000), Houston, James and Ryngaert (2001), and Kaplan and Ruback (1995)). As described in detail in Appendix D, 17

19 this amounts to discounting the projected annual gains from operating efficiencies and cost savings using an estimate of the merged entity s cost of capital. Our empirical tests are based on the value of synergies obtained in perpetuity scaled by pre-merger value of the merging firms. However, our inferences are qualitatively similar when we assume a ten- or five-year horizon for the merger-related gains. Insert Table 2 here Table 2 reports summary statistics of the value of projected synergies as a fraction of pre-merger market capitalization of the merging firms, assuming the operating gains are realized in perpetuity (first three columns) or over a ten-year horizon (last three columns). The mean (median) value of synergies in perpetuity is approximately 14% (7%) of the combined value of merging firms in horizontal deals. There is substantial variation in estimated projected synergies, as indicated by the spread between the first and third synergy quartiles (3% and 15% respectively). The mean (median) synergies computed over a ten-year horizon are over 7% (close to 4%) Effect of synergies on the performance of the merged entity Our main empirical tests rest on the premise that insiders projections of merger-related gains are economically meaningful i.e., informative about synergies, and therefore would explain the merged entity post-merger strategic behavior in product and input markets. Consistent with this premise, Bauguess and Bernile (2011) find that insiders synergy forecasts indeed explain the merged entity s returns and post-merger operating performance. Before proceeding with our main tests, we perform a similar analysis for the horizontal mergers in our sample based on Hoberg-Phillips TNIC scheme and find results consistent with Bauguess and Bernile. Specifically, we estimate regression models where the dependent variable is either the merging firms value-weighted mean cumulative abnormal return from 20 trading days prior to 20 trading days after the announcement, or the difference between the merged entity s average industry-adjusted EBITDA-to-sales ratio during the three years after successful deal completion and its ratio in the year prior to completion. 16 The main independent variable is the value of the projected merger-related gains scaled by the merging firms combined market capitalization prior to the merger announcement. We also control for other factors that may explain the performance of the merging firms. Important for the purpose of our study, mergers may result in increased market power, which would in turn also affect the merging firms rivals, customers, and suppliers. Successful horizontal mergers increase 16 The results are robust to estimating cumulative announcement returns over windows of different lengths and to computing the change in return on sales over various horizons. 18

20 industry concentration and may lead to higher rents for the merging firms industry (e.g., Salant, Switzer and Reynolds (1983) and Perry and Porter (1985) among many others). Similar to Eckbo (1983), Fee and Thomas (2004), and Shahrur (2005), we include the industry s Herfindahl index as an explanatory variable, because the market power effect should be stronger in more concentrated industries. In addition, we include the merging firms combined market share (i.e. merging firms combined sales divided by sales of all firms belonging to bidder s and/or target s TNIC industries), because the market power effect should increase with the relative importance of the merging firms in their product markets. We also include the relative size of the target-to-bidder assets, because mergers of firms that are more similar in size should be associated with larger market power increases. In our regression models, we also account for the endogeneity of insiders decision to provide synergies forecasts by applying Heckman (1976) correction. In particular, we first estimate a probit regression in which the dependent variable is an indicator equal to one when projections are disclosed. This model includes variables expected to influence the decision to disclose synergy forecasts: the premerger propensity of bidder and target to make voluntary disclosures (earnings guidance indicators); the number of analysts following the merging firms; the number of institutional investors holding the merging firms stocks; deal characteristics; bidder s and target s accounting performance measures; industry characteristics; and measures of industry concentration. For brevity, we remand to Bauguess and Bernile (2011) for a detailed discussion of these variables. Table A1 in the Appendix reports estimates of this probit model. We then include the inverse Mills ratio from this probit regression in the models that relate merger announcement returns or post-merger operating performance to the value of projected synergies. Table 3 reports the resulting second-stage regression estimates. Insert Table 3 here For our purposes, the most important finding in Table 3 is that synergy projections are strongly and significantly related both to the merger announcement returns and to the post-merger changes in operating performance of merging firms. A one standard deviation increase in the scaled value of synergies is associated with an average increase of three percentage points in the combined cumulative abnormal return, ceteris paribus. Similarly, a one standard deviation increase in synergies is associated with an average four percentage points increase in return-on-sales. None of the market-power-related determinants of the market s reaction to merger announcement and post-merger change in operating performance is consistently statistically and economically significant. Overall, consistent with the first empirical prediction of our model, the evidence shows that the synergies (as forecasted by insiders) are a strong determinant of horizontal mergers performance. In 19