Horizontal Acquisitions and Buying Power: A Product Market Analysis

Transcription

1 Horizontal Acquisitions and Buying Power: A Product Market Analysis by Sugato Bhattacharyya* University of Michigan Amrita Nain** McGill University This version: November 2008 (Preliminary draft. Not for general circulation) Abstract Horizontal mergers exert price pressure on dependent suppliers and adversely affect their performance. Downstream consolidation appears to be preceded by significant horizontal merger activity undertaken by dependent suppliers. Consistent with the theory of countervailing power, it is the suppliers undertaking performance-enhancing horizontal mergers prior to the downstream consolidation who experience the largest decline in prices after consolidation downstream. Together, our results suggest that downstream mergers are successful attempts to countervail upstream market power. The findings are broadly consistent with pervasive beliefs in the business community about market power effects of horizontal mergers. *Sugato Bhattacharyya, Ross School of Business, University of Michigan, 701 Tappan Street, Ann Arbor, MI Tel: (734) , sugato@umich.edu **Amrita Nain, Faculty of Management, McGill University, 1001 Sherbrooke Street West, Montreal, QC H3G 1Y2. Tel: (514) , amrita.nain@mcgill.ca The authors thank Amy Dittmar, E. Han Kim, Francine Lafontaine, Vikram Nanda, Husayn Shahrur, Linda Tesar, Bilal Zia, seminar participants at the 2007 Western Finance Association meetings, the Finance Brownbag series at the University of Michigan and the McGill Finance Seminar series for helpful comments.

2 1. Introduction apparel-company executives say they are bracing for store closures, cutbacks and thinner profit margins. The potential fallout reflects the huge negotiating power that a combined Federated-May would wield and the diminishing clout of suppliers. it could also accelerate consolidation among apparel suppliers, as they strive to get bigger to better face off against their giant customers. --- Wall Street Journal article 1 1. Introduction There is a long-standing debate in the economics and finance literatures on the motives for horizontal mergers. While managers of firms undertaking horizontal mergers usually cite expected improvements in productive efficiencies, i.e. synergies, as the key rationale behind such moves, anti-trust authorities frequently express concern that horizontal mergers may increase market power vis-à-vis customers and suppliers of the merging firms industry. The latter view is also often supported in discussions in the business press pertaining to specific deals, as evidenced by the quote above. Academic research has extensively examined the effect of horizontal acquisitions on market power vis-à-vis customers and arrived at conflicting conclusions. 2 In contrast, there has been relatively little empirical research on the effect of horizontal acquisitions on suppliers. Despite the paucity of research on how suppliers are affected by horizontal mergers, there are compelling reasons to look upstream rather than downstream for market power effects of horizontal mergers. First, if a horizontal merger is expected to increase selling power and result in higher prices for customers, anti-trust authorities are likely to block the acquisitions. Thus, the truly anti-competitive mergers may never be observed. In contrast, horizontal mergers that create buying power are less likely to be blocked. Buying power, whether collusive or unilateral, can force suppliers to become more competitive and more efficient. Thus, mergers that increase market power vis-à-vis suppliers may be viewed favorably by regulatory authorities, and therefore, are more likely to be observed in a randomly selected sample of horizontal mergers. 1 Combined Federated-May Could Stress Apparel Makers, The Wall Street Journal March 1 st, Focusing primarily on announcement returns, Eckbo (1983), Stillman (1983), Eckbo (1985) and Eckbo and Wier (1985), Fee and Thomas (2004) and Shahrur (2005) conclude that horizontal mergers do not create selling power vis-à-vis customers. Looking at product prices directly, Barton and Sherman (1984), Borenstein (1990), Kim and Singal (1993), Singal (1996), Akhavein, Berger and Humphrey (1997), and Prager and Hannan (1998) conclude that horizontal mergers create selling power. 1

3 Our second motivation for looking at the effect of horizontal mergers on supplier industries is that the industrial organization literature already documents the importance of being a large buyer. Buyer size and buyer industry concentration have long been correlated with lower seller profits. 3 Yet the upstream effects of a major corporate event a horizontal merger that can create large buyers and increase buyer industry concentration remain largely unexamined. 4 Thus, the objective of this paper is to ask one overarching question: do horizontal mergers create buying power? We answer this question by systematically examining the effect of horizontal mergers on product prices and profits in the supplier industry as well subsequent actions of firms in the supplier industry. We use a relatively large, cross-industry sample of horizontal acquisitions to examine whether horizontal mergers bring about a decline in the profits of supplier industries and whether such a decline can be attributed to a decline in prices at which supplier industries sell. We identify a number of industries that experienced a significant jump in horizontal merger activity in a specific quarter. Having identified these downstream merger events we then ask whether supplier industries more dependent on the downstream merging industry experience more adverse changes in profits and output prices after the event. We find that supplier industries that sell a larger fraction of their output to the downstream consolidating industry (i.e. more dependent supplier industries) have lower cash-flow margins following the downstream consolidation. The abnormal cash-flow margin of dependent supplier industries after the downstream consolidation is 4.2% lower than that of non-dependent supplier industries. Thus, we confirm Fee and Thomas s (2004) result that some supplier industries suffer declines in operating profits after a horizontal merger downstream. However, we recognize that a decline in supplier profit margins can result from either an increase in production costs or a decrease in supplier selling prices. If the deterioration in profit margins of dependent supplier industries is attributable to an increase in buying power of the downstream industry, we should observe a decline in the selling prices of dependent supplier industries. Using the Producer Price Index (PPI) as a measure of selling prices, we conduct a 3 See, for example, Schumacher (RES 1990) Lustgarten (RES 1975), Clevenger and Campbell (IO Review 1977), McGuckin and Chen (IO Review 1977) 4 Exceptions are Fee and Thomas (2004) and Shahrur (2005). Both these studies find some preliminary evidence that downstream mergers adversely affect suppliers in concentrated industries. 2

4 number of tests to examine whether supplier selling prices are adversely affected by a consolidation in the downstream industry. First, we examine selling prices in dependent supplier industries during the three years preceding the downstream merger. Controlling for changes in input prices and demand shocks faced by the supplier industry, we find that prior to downstream consolidation, changes in the PPI of dependent and non-dependent supplier industries are statistically indistinguishable. Second, we examine the PPI of supplier industries during the three years following downstream consolidation. We find that dependent supplier industries experienced significantly larger declines in PPI than do non-dependent suppliers during the three years following downstream consolidation. The decline in prices of the dependent supplier industries is worse by about 0.1% per month relative to non-dependent industries. This can translate to a difference of up to 3.6% in the prices of dependent and non-dependent supplier industries over the three-years following downstream. Third, we demonstrate that our results are robust to alternative regression methods. For example, we combine the six-year period surrounding the downstream merger and use a difference-in-differences methodology to simultaneously control for the effect of supplier dependence and the post-merger period on supplier prices. We find, again, that dependent suppliers perform significantly worse compared to non-dependent suppliers, but only in the post-merger period. Finally, we address the concern that prices in highly dependent supplier industries may trend downwards regardless of whether a downstream consolidation occurs or not. We create random event dates and use these random dates as break points when analyzing supplier selling prices. We find that there is no difference in the selling prices of dependent suppliers before and after the random date. Based on this battery of tests, we conclude that decline in supplier selling prices can be attributed to the downstream merger or to possibly unobservable events that triggered the downstream merger. The decline in selling prices of supplier industries after downstream consolidation is consistent with the hypothesis that horizontal mergers create buying power. However, this outcome is also consistent with a merger-induced improvement in efficiency. If the downstream horizontal merger creates production efficiencies, the merged firms can produce the same amount of output with fewer inputs. If the demand for inputs goes down, the price of inputs will decline as well. Given the existence of such straightforward efficiency explanations for the observed decline in supplier selling prices, the remaining part of the paper focuses on linking the decline in supplier prices to factors that are clearly related to market power. To this end, we draw 3

5 on existing theory of countervailing acquisitions. Galbraith s (1952) theory of countervailing power argues that an increase in seller concentration induces buyers to grow large in order to neutralize the market power of sellers. In Snyder (1996,1998) mergers between buyers can intensify competition among colluding sellers, thereby allowing larger buyers to obtain lower prices. Countervailing acquisition theory implies that suppliers who build up industry concentration may experience the greatest loss of market power after downstream consolidation. In our sample, this notion of countervailing power would be supported if the suppliers who experienced the largest price decline after the downstream merger happened to be the ones that built up industry concentration prior to the downstream event. Thus, the second phase of our empirical analysis asks whether more dependent suppliers, who appear to experience more adverse price changes after downstream consolidation, were themselves engaged in significantly greater horizontal merger activity prior to the downstream merger. We find evidence that more dependent supplier industries were indeed engaged in significantly greater horizontal merger activity prior to the large downstream merger. Moreover, supplier industries that experienced more horizontal mergers enjoyed higher profit margins in the period just preceding the downstream consolidation. These results suggest that the large downstream merger may be an attempt to countervail market power created by upstream consolidation. If countervailing theory holds, then it must be the case that supplier industries that engaged in more horizontal merger activity prior to the downstream merger (thereby increasing industry concentration) subsequently experienced larger price declines after the downstream merger. The third part of our analysis tests this hypothesis by examining the relation between supplier merger activity prior to the downstream merger and supplier price changes after the downstream merger. We find that suppliers that engaged in more horizontal mergers in the three years prior to downstream consolidation experience larger price drops after downstream consolidation. No such relation exists between non-horizontal merger activity by suppliers prior to the downstream consolidation and supplier prices after downstream consolidation. This result continues to hold if we use a measure of change in industry concentration instead of horizontal merger activity. Suppliers that experienced increases in industry concentration prior to the downstream merger experienced larger price declines after the downstream merger. This correlation between horizontal mergers (or change in supplier concentration) prior to the downstream merger and supplier selling prices after the downstream merger provides further 4

6 credence to the hypothesis that the downstream merger successfully counteracted upstream market power. Our interpretation could be challenged with the argument that horizontal merger activity in supplier industries is an attempt to prevent an expected decline in industry demand conditions. That is, it could be argued that the relation between horizontal merger activity in the supplier industry and subsequent decline in supplier prices exists regardless of the downstream merger event. However, we demonstrate that high horizontal merger activity in the supplier industry is not always followed by a decline in prices. For our sample of supplier industries, we generate random event dates and show that there is no relation between horizontal merger activity by suppliers prior to the random event date and subsequent selling price changes in the supplier industry. The downstream consolidating merger is key to the relation between horizontal merger activity by suppliers and subsequent decline in supplier prices. Together these results provide a comprehensive picture in support of the hypothesis that horizontal mergers create market power. The downstream merger events identified at the beginning of our analysis appear to be a response to upstream consolidation. Moreover, the decline in supplier selling prices after the downstream merger event reflects loss of market power built up prior to the downstream merger. Our tests confirm the Fee and Thomas (2004) finding that supplier operating performance deteriorates after downstream consolidation. Our results are also consistent with Shahrur s (2005) finding that more concentrated suppliers have poorer announcement returns when a downstream merger occurs. However, our paper is the first to provide direct evidence that supplier selling prices decline after downstream merger and that this decline is attributable to a shift in market power in favor of the downstream merging industry. Our paper is also the first to demonstrate that horizontal acquisitions successfully counteract upstream selling power and suggests that such consolidating acquisitions can propagate along industries that share product market relationships. Finally, in addition to contributing to our understanding of the market power effects of horizontal acquisitions, this paper adds to existing evidence that merger activity is determined by industry level factors (see Mitchell and Mulherin (1996) and Andrade, Mitchell and Stafford (2001)). The paper is organized as follows. Section 2 motivates the empirical tests. Section 3 contains methodology, data sources and results. Section 4 addresses issues of robustness. Section 5 concludes. 5

7 2. Hypothesis Development Merging firms can exercise buying power in different ways. They can, for example, pool purchases to obtain quantity discounts from suppliers, or reduce profit margins by squeezing suppliers. Insofar as these actions promote greater efficiency on the part of suppliers, Fee and Thomas (2004) label these as evidence of efficiency-increasing buying power. A merged firm may also exercise buying power by restricting purchases to monopsony levels causing input prices to fall below marginal cost (see Robinson (1933)). In the presence of sunk costs, such price decreases may be sustained in the short-run but will come at a cost to efficiency. In this paper, we are agnostic about the welfare consequences resulting from the possible use of market power and focus entirely on the possible exercise of market power alone 5. If horizontal mergers do, indeed, create market power, we expect this effect to show up in the operating performance of supplier industries. Since merging firms will be able to exercise buying power more effectively when the supplier industries are more dependent, we hypothesize that dependent supplier industries will suffer a greater decline in performance after downstream consolidation than those less dependent on the downstream industry. Since such effects may take some time to show up in the data, we hypothesize that such performance effects will be evident over a three-year horizon. We choose this horizon because acquisitions of significant size often take six months to a year to be consummated. Thus, the first hypothesis we test is: Hypothesis 1: More dependent suppliers experience greater adverse changes in cash flow margins in the three years subsequent to an announcement of downstream consolidation. Declines in operating performance, while consistent with market power enhancement, are not definitive evidence of the exercise of buying power. Other unrelated factors like increases in production costs or wages may also account for a drop in profitability of supplier industries. If, however, the decline in performance is related to enhanced downstream market power, we should expect such market power to also show up in the form of diminished selling prices in the supplier industry. Therefore, our second hypothesis is as follows: 5 Firms are known to publicly justify mergers with the cost savings that would arise due to increased buying power. A proposed merger between Staples and Office Depot in 1997 was blocked by anti-trust authorities on the grounds that it would lead to higher prices for consumers. Staples countered with the argument that the merger would allow it to lower selling prices because of the greater purchasing power the transaction would bring. See Office Depot Staples Deal is Blocked, The Wall Street Journal, July 1 st,

8 Hypothesis 2: More dependent supplier industries experience larger declines in selling prices subsequent to downstream consolidation. While price declines in supplier industries is highly suggestive with regard to the creation of market power when consolidation downstream is mainly predicated on taking advantage of cost-savings, horizontal acquisitions that improve productive efficiencies could also adversely affect supplier prices and cash flows. For example, if consolidation enables the merging firms to produce the same output with a lower use of an input, the demand for this input will fall. An observed decline in supplier selling prices could, then, be explained without relying on enhanced market power downstream. Since, production efficiencies and market power creation can coexist when horizontal acquisitions occur, distinguishing clearly between these two causes behind price declines in supplier industries is almost impossible. As a result, we draw on the existing theory of countervailing acquisitions to motivate further confirmatory tests which rely on the dynamics of the consolidation process in order to distinguish between these two sources of price declines upstream. Galbraith (1952) presents a theory of countervailing power in which buyers have an incentive to grow larger in order neutralize upstream market power. Galbraith contends that in the typical modern market of a few sellers, the active restraint to hold prices close to marginal cost is provided not be competitors but by strong buyers. Ellison and Snyder (2001), Stole and Zwiebel (1996) and Chipty and Snyder (1999) examine buyer bargaining power relative to a single seller. Their models present reasonable conditions under which large buyers are charged lower prices. Another variety of models allow for multiple sellers whose ability to collude depends on the characteristics of the buyer. For example, Snyder (1996) presents a dynamic theory of countervailing power in which large buyers are shown to obtain lower prices from colluding sellers. In Snyder (1996, 1998) mergers between buyers intensify competition among colluding sellers. Moreover, Snyder (1996) shows that mergers in the buyer industry increase profits for all buyers, not just those of the merging firms, at the expense of the seller. The notion that large buyers have an advantage in obtaining price concessions from sellers has been verified 7

9 by a number of empirical studies. 6 Although industry consolidation is an important process giving rise to larger buyers, these studies have not, however, looked at the role of horizontal mergers in imposing such price competition in the upstream industry. Drawing on the theory of countervailing power, horizontal mergers in a downstream industry can be viewed as a response to the growth of market power in upstream industries. Extending the same logic in symmetric fashion, we can also hypothesize that the most vulnerable supplier firms would be the most likely to try and countervail downstream market power by resorting to horizontal mergers themselves. As a result, industry consolidation can be viewed as a dynamic process that results from industries trying to countervail the enhancement of market power, whether downstream or upstream. Our next two hypotheses formalize these ideas to generate empirical tests for countervailing power. Suppose, then, that downstream consolidation is, indeed, an attempt to countervail upstream market power. Then, it must be that downstream consolidation was triggered by attempts to create market power upstream. Since it is likely that dependent suppliers are subject to greater pricing pressure by their downstream counterparts, it should be more likely that we would find attempts at consolidation among such suppliers. In addition, we should expect to find that such consolidation succeeded in enhancing immediate performance among such suppliers, at least in the short-run. These observations lead to our third hypothesis: Hypothesis 3: Prior to consolidation downstream, horizontal merger activity should be more concentrated in dependent supplier industries. Moreover, such consolidation should be accompanied by improved short-term performance. It should be noted, however, that consolidation upstream, just like consolidation downstream, could arise also from opportunities to save on costs, in addition to obtaining enhanced market power. If this were the case, we would expect short-term effects to show up more in operating performance than in price levels. However, even if the impact on price levels observed by an outside observer were low, the information available to downstream firms is likely to be of much higher quality. Specifically, any enhanced market power upstream would give rise to incentives 6 See, for example, Lustgarten (1975), McGuckin and Chen (1976), Boulding and Staelin (1990), and Schumacher (1991) 8

10 on the part of downstream firms to consolidate and countervail the emergence of such pricing power. Moreover, we should then expect that price declines in supplier industries, if any, to be disproportionately concentrated in precisely those supplier industries that engaged in consolidation activity prior to significant downstream consolidation. This logic leads to our fourth hypothesis. Hypothesis 4: Supplier industries engaging in greater horizontal merger activity prior to downstream consolidation experience larger price declines after downstream consolidation. Taken together, our four hypotheses comprise a holistic description of price and profit consequences of horizontal merger activity as it may contribute to the creation of market power. In particular, price and performance effects as summarized in the first two hypotheses constitute a smoking gun that argues for delving in to corroborative evidence. In turn, using the hypothesis of countervailing power, the third and fourth hypotheses serve to directly relate the creation of market power downstream as a strategic response to the creation of market power upstream. However, we are agnostic about whether, in turn, upstream consolidation is also provoked by downstream consolidation. 3. Data Sources, Methodology and Results 3.1 Data Construction We begin by constructing a sample of industries that experienced an identifiable increase in horizontal merger activity in order to obtain distinct pre- and post-merger periods. We obtain from SDC Platinum all acquisitions announced between 1984 and 2003 that meet the following criteria: (i) the target and acquirer both were U.S.-based, (ii) the target and acquirer shared the same primary four-digit SIC code, (iii) the announced acquisition was eventually completed, and (iv) the acquirer bought more than fifty percent of the target s stock. From SDC Platinum we also obtain the transaction value associated with each merger, that is, the total value of consideration paid by the acquirer, excluding fees and expenses. For each four-digit SIC code in the merger sample, we measure quarterly acquisition activity as the total transaction value of all 9

11 horizontal acquisitions announced in a quarter as a proportion of industry total assets. We classify an industry as having experienced a merger event in a given quarter when the following conditions hold: (i) the quarterly acquisition activity in the current quarter is greater than 10% and (ii) the quarterly acquisition activity in any of the previous 12 quarters did not exceed 2.5%. The first condition ensures that the selected industries experienced significant consolidation in a particular quarter, while the second condition ensures that we have a clean preevent period during which there was little horizontal merger activity. This definition enables us to identify 259 four-digit SIC codes that experienced at least one merger event between 1980 and Next, we use benchmark Input-Output (I-O) tables provided by the Bureau of Economic Analysis to match each merging industry with its major supplier industries. The benchmark I-O tables, compiled every five years at the detailed 6-digit NAICS or 4-digit SIC level, show how approximately 500 industries provide inputs to, and use output from, each other. We use the 1992 benchmark tables to match suppliers to industries that consolidated in 1994 or earlier and use the 1997 tables to match suppliers to industries consolidating in 1995 or after. This process allows us to find suppliers for 141 merging industries. 7 These 141 industries serve as our starting sample of merging industries. Table I Panel A lists these merging industries, along with the number of mergers that contribute to the merger event and the ratio of merger transaction value to industry total assets. The list includes three industries for which the ratio of merger transaction value to industry total (book) assets is greater than 10. In Section 4 of the paper, we discuss robustness of our results to an alternative sample selection method. Using the IO tables, we also calculate the fraction, f mj, of supplier industry j s output sold to the merging industry m. Higher values of f mj indicate that the supplier industry j is more dependent on the merging industry. For each merging industry, we identify up to ten supplier industries with the highest values of f mj. As there are 141 merging industries, this allows for, at the most, 1,410 merger industry-supplier industry pairs. We are able to obtain data on f mj for 1,155 merger industry supplier industry pairs. By choosing to work with as many as 10 suppliers per merger industry, we include industries that sell a very small fraction of their output to the merging industry and, therefore, are unlikely to be affected by the merger. This allows our 7 Some of the merging industries experience more than one horizontal merger event and, therefore, appear more than once in the sample of 141 industries. Moreover, industries in our merger event sample can share customersupplier relationships. We discuss the robustness of our results to these issues in Section 4. 10

12 cross-sectional tests to have greater power in detecting any relation between supplier dependence and price/profit changes experienced by the supplier industry. We define Dependent Suppliers as those with values of f mj in the top quintile of the distribution. Remaining suppliers are classified as non-dependent. 8 Table I Panel B provides the distribution of f mj for dependent and non-dependent suppliers. Dependent suppliers provide, on average, 15.6% of their output to the merging industries. Non-dependent suppliers provide, on average, 1.7% of their output to the merging industry. 3.2 Supplier Industry Operating Performance Pursuant to our first hypothesis, we examine first the impact of downstream consolidation on supplier industry operating performance. We measure the operating performance of an industry by the cash flow-to-sales ratio of the median firm in the industry: As in Fee and Thomas (2004), the cash flow-to-sales ratio of a firm is the ratio of operating income (Compustat item 13) to sales (Compustat item 12).We then define an industry s abnormal operating performance by the deviation of its operating performance from that of median industry in the economy. To examine the effect of downstream mergers on supplier industries, we take the average abnormal operating performance of a supplier over the three years following downstream consolidation and compare it with the average abnormal operating performance during the three years prior to downstream consolidation. In Table II A, we present a univariate analysis. 9 We see that, on average, supplier industries in our sample have negative abnormal performance both before and after downstream consolidation. However, the negative abnormal performance of suppliers is entirely driven by the top dependence quintile for both periods. In contrast, abnormal performance of non-dependent suppliers is statistically indistinguishable from zero in both periods. In addition, dependent suppliers suffer a significant drop in abnormal operating performance after a downstream merger event, while non-dependent suppliers experience no 8 Our findings are robust to changes in the f mj cut-off used to classify suppliers as being dependent. For example, our results continue to hold if we define dependent suppliers as those in the top quartile of f mj. Moreover, instead of defining a binary variable to capture supplier dependence, we have also used the continuous variable f mj as a measure of supplier dependence (not shown). Our results continue to hold, albeit with smaller statistical significance. 9 The 322 observations used in the profit margin analysis comprise 98 unique supplier industries. We are unable to obtain profit margin data for the remaining supplier industries primarily because Compustat SIC codes are often aggregated at the two or three digit SIC level. 11

13 change. These results are supportive of our first hypothesis that downstream consolidation adversely affects the operating performance of dependent suppliers. In Table III we present a multivariate analysis of supplier operating performance. The regressions are estimated using ordinary least squares with robust standard errors clustered at the two-digit SIC level. The dependent variable is the abnormal operating performance measure described above. The key explanatory variable is supplier dependence captured by a dummy variable that equals 1 for suppliers in the top quintile of f mj and zero otherwise. Control variables are derived from previous research on the determinants of industry profitability. 10 We take the natural logarithm of the number of firms in an industry as a direct measure of competition within the industry. Since high capital requirement are likely to function as a barrier to entry, we use capital intensity and capital expenditures as additional control variables for estimating industry profitability. For each four-digit SIC, Capital Intensity is calculated as industry total assets (Compustat item #6) divided by industry total sales (Compustat item #12). Capital Expenditures are calculated as an industry s total capital expenditures (Compustat item #128) divided by industry total assets. While Capital Intensity provides a scaled measure of the total capital stock in an industry at a point in time, Capital Expenditures provide a scaled measure of the annual capital investment required in an industry. Finally, we use Advertising Intensity to proxy for product differentiation, where this is calculated as industry total advertising expense (Compustat item # 45) divided by industry total sales. The first column of Table III has, as the dependent variable, the supplier industry abnormal operating performance averaged over the three years preceding the downstream merger. The coefficient on the dependence dummy is insignificantly different from zero. This indicates that, controlling for general factors affecting industry profit margins, profitability of dependent suppliers statistically indistinguishable from that of non-dependent suppliers prior to the downstream merger. The dependent variable in the second column is the supplier industry abnormal operating performance averaged over the three years following the downstream merger. The coefficient on the dependence dummy is now negative and significant at the 95% confidence level. The magnitude of the coefficient indicates that the abnormal cash-flow margin of dependent supplier industries is 4.2% lower than that of non-dependent supplier industries. We note that the signs on 10 See for example, Schumacher(1991) 12

14 the control variables are as expected: Cash-flow margins are lower in more competitive industries; Margins are higher in industries with greater barriers to entry; and margins are higher in industries with greater product differentiation. 3.3 Supplier Industry Selling Prices The results in the previous sub-section, while establishing deterioration in the performance of dependent suppliers subsequent to downstream consolidation, do not say anything about why the deterioration occurs. Deterioration in operating margins can occur due to a rise in costs or a fall in selling prices. If horizontal mergers create buying power for the merging firms industry, we should observe a decline in supplier selling prices. Therefore, we now test Hypothesis 2 - do the selling prices of more dependent suppliers decline more after a significant downstream consolidation? An advantage of examining selling prices instead of cashflow margins is that supplier cash-flow margins are available only for a subset of 4-digit supplier SIC codes (see Footnote 9). Since product price data compiled by the Bureau of Labor Statistics are available for a larger sample of supplier industries, we can both enlarge our sample of supplier industries and answer the question of market power creation more directly. Existing product market studies have examined the effects of horizontal mergers on the selling prices of the merging industry itself using an event-study type approach. In these studies, changes in prices charged by merging firms have been compared with changes in prices charged by a control group of firms in the same industry. The control group is expected to account for other factors that affect price changes in an industry, like demand conditions, changes in input prices, etc. For example, Kim and Singal (1993) compare airfare changes on routes affected by airline mergers with airfare changes on routes not affected by the mergers. By examining price changes relative to a control group within the same industry, Kim and Singal are able to account for other factors that affect airfares. In our industry-level study, we wish to examine the effect of horizontal mergers on prices at which supplier industries sell. The control group approach would require that, for each supplier industry, we are able to identify another industry that experiences similar changes in demand conditions and factor prices, is not affected by a downstream merger, and is not an upstream or downstream industry to the supplier industry. This is, clearly, a tall 13

15 order. As a result, the control group approach is not suitable for our study. Instead, we explicitly account for changes in input prices and demand shocks that a supplier industry may face. For each supplier of a merging industry, we obtain the Producer Price Index (PPI) from the Bureau of Labor Statistics (BLS). The PPI series allows us to measure the change over time in the selling prices received by domestic producers of goods and services. We adjust the PPI series for inflation using the GDP price deflator, and call it real Producer Price Index (RPPI). Table IV presents summary statistics of supplier industry RPPI. The table shows the average value of RPPI for suppliers as a whole over the three years before the downstream merger, the average value three years after the downstream merger as well as the difference between the two values. The difference is negative and significant at the 99% confidence level suggesting that supplier industries, on average, experience a decline in selling prices after the downstream merger. Next, we split the sample into two groups dependent suppliers and non-dependent suppliers. Table IV shows that the most dependent suppliers had significantly lower prices than the remaining suppliers both before and after the downstream consolidation. Although both groups experience a decline in prices after the downstream merger, the difference-in-differences test in the last column shows that the fall in prices is significantly larger for the dependent suppliers. These univariate results support the notion that downstream consolidation affects dependent suppliers more adversely as compared with non-dependent suppliers. However, the finding that dependent suppliers have lower prices both before and after the downstream merger requires that, in the multivariate analysis, we control for the possibility that more dependent supplier industries are fundamentally different from less dependent suppliers in terms of the average price levels over time. Moreover, since univariate tests show that both groups of suppliers experience significant price drops after the downstream merger, we need to control for the possibility that, due to exogenous factors, price levels after a downstream merger are lower for all industries. Our multivariate analysis begins with the following regression model estimated using pooled OLS with Newey-West standard errors. Δrppi jt = α 0 +α 1 D j +α 2 Δrppi _ inp 1 jt +α 3 Δrppi _ inp 2 jt +α 4 Δwage jt +α 5 Δtp t +ε jt (1) 14

16 In equation (1) above, rppi is the RPPI of supplier industry j in logs. The dummy variable, D, identifies suppliers who are highly dependent on the downstream merging industry: D equals 1 if the fraction of supplier output, f mj, sold to the downstream merging industry lies in 1 2 the top quintile and 0 otherwise. The control variables rppi _ inp jt and rppi _ inp jt represent the RPPI of supplier industry j s two primary inputs, again in logs. To get these input prices for the supplier industry, we use the benchmark Input-Output tables to identify two inputs that comprise the largest fraction of the supplier industry s total input usage. 11 To this end, we first calculate the weights, w ji, that represent the fraction of supplier industry j s input provided by industry i. We take the two industries, i, with the two highest values of w ji as the main contributors to input prices for the supplier industry j. Price data for these inputs are obtained from the Bureau of Labor Statistics. We also control for wages in the supplier industry. The variable wage represents log differences of average hourly earnings of production workers compiled by the BLS. Hourly earnings are available only for production workers in the mining and manufacturing industries and are often provided only at the three-digit SIC level. When wage data are available only at the three-digit level, we apply them to all four-digit industries within the three-digit SIC. Finally, total industrial production is used to control for demand conditions in the economy. The industrial production index, tp, obtained from the Federal Reserve Board measures log of the real output of the manufacturing, mining, and electric and gas utilities industries. The regression includes a time trend, industry dummies at the two-digit SIC level, and year dummies. We first estimate equation (1) for all supplier industries over the 36 months preceding the relevant downstream consolidation and then separately over the 36 months following the downstream consolidation. We ignore the quarter in which the downstream event occurred. For example, if a downstream industry consolidated in the second quarter of 1999, our pre-merger sample for supplier industry prices runs from April 1996 to March 1999 and the post-merger sample period runs from July 1999 June Estimates of these two regressions are provided in Columns 1 and 2 of Table V. Column 1 shows that the coefficient on the dependence dummy, D, is statistically insignificant in the period prior to the downstream merger. Once factor prices and demand conditions are controlled for, we see that price changes in dependent supplier 11 It is important that the input prices used as control variables in this regression are unaffected by events occurring in the merging industry. To reduce to possibility of endogenous input prices, we ensure that the industries that provide the main inputs of the supplier industry have no product market relation with the downstream merging industry. 15

17 industries prior to downstream consolidation are not significantly different from price changes in the remaining supplier industries. However, in the post-merger sample presented in Column 2, we see that the dependence dummy is significantly negative at the 99 percent confidence level. Therefore, after the downstream merger, the most dependent suppliers experience adverse price changes relative to the non-dependent supplier industries. The magnitude of the regression coefficient in Column 2 suggests that the decline in prices for dependent supplier industries is worse by about 0.1% per month relative to non-dependent supplier industries. The analysis above shows that, in the post-merger period, dependent suppliers experience adverse price conditions that did not exist prior to the downstream merger. However it does not tell us whether the coefficient on the dependence dummy is different between the two regressions in Columns 1 and 2. Therefore, we conduct a difference-in-differences analysis. We combine the pre- and post-merger periods and estimate the following regression using the full 72-month panel. Δrppi jt = α 0 +α 1 D j +α 2 PM jt +α 3 D j PM jt +α 4 Δrppi _ inp 1 2 jt +α 5 Δrppi _ inp jt +α 6 Δwage jt +α 7 Δtp t +ε jt (2) The dummy variable D, as defined earlier for Equation 1, captures supplier dependence. The coefficient α 1 captures whether more dependent suppliers are different from less dependent suppliers in terms of the average price levels over time. For any supplier industry merger industry pair, the post-merger dummy variable, PM, equals 1 after the downstream consolidation and 0 before. This coefficient, α 2, on this dummy captures whether price levels after a downstream merger are on average lower for all suppliers. It also controls for exogenous shocks that might affect prices in supplier industries as well as trigger mergers in the downstream industry. The coefficient of primary interest is, however, α 3 : If more dependent suppliers suffer larger declines in selling prices due to the downstream merger, the coefficient on the interaction of D and PM should be negative. All other variables are as described earlier for Equation 1. Column 3 of Table V presents the estimation results for Equation (2). The coefficient on the interaction of the Dependence dummy and the Post-Merger dummy, α 3, is negative and significant at the 95 percent confidence level. This means that selling prices of suppliers who are dependent on the merging industry decline after consolidation in the downstream industry and 16

18 the decline is larger for dependent suppliers. Since the two dummy variables are also included separately as control variables to account for level effects, it is the interaction term that isolates the differential impact of a downstream merger on upstream prices of more dependent suppliers. The coefficients on the Post-Merger dummy and the Dependence dummy are statistically indistinguishable from zero. Thus, there is no evidence that dependent suppliers always (i.e. at any point in time) have lower prices, or that price levels are generally lower after the downstream merger for all suppliers. Dependent suppliers have significantly lower prices relative to remaining suppliers only after the downstream consolidation. As expected, the results on input prices, wages and total production confirm that higher input prices do pass through to output prices and that demand shocks also translate into higher prices. To demonstrate that this outcome is not sensitive to the regression methodology used, we employ an alternative to the difference-in-differences method described in above. We run the following cross sectional regression using OLS with robust standard errors clustered at the 2- digit SIC level. Δ ln RPPI j =α 0 +α 1 D j +α 2 Δ ln RPPI _ INP j 1 +α 3 Δ ln RPPI _ INP j 2 +α 4 Δ lnwage j +α 5 Δ lntp +ε j (3) In Equation (3) above, ΔlnRPPI j is the supplier j s average log RPPI over the three years after the downstream merger minus the average log RPPI over the three years prior to the downstream merger. For control variables, we calculate the change in average input prices, wages and total production in the same manner. Again the explanatory variable of interest is the dependence dummy D. Results are presented in Column 4 of Table V. As expected, the coefficient on the dummy variable D is negative and statistically significant: Dependent suppliers experience the largest declines in prices after the downstream merger. One remaining concern is that prices in highly dependent supplier industries may trend downwards over time even in the absence of downstream consolidation. To address this concern, we conduct an experiment. For each supplier in our sample, we generate a random eventquarter between 1984 and 2003 by drawing from a uniform distribution. Then we repeat a crosssectional regression similar to the one shown Equation (3) but with one key difference we use the randomly generated date as the break point in place of the actual downstream merger event. That is, the change in supplier prices is calculated as the average price three years after the 17

19 randomly selected quarter minus the average price three years before the random quarter. Changes in control variables are calculated in a similar manner. If prices in highly dependent supplier industries were to naturally trend downwards, the dependence dummy should be significantly negative in this randomized sample as well. Column 5 of Table V presents the results. We note that the coefficient on the dependence dummy is no longer statistically significant. This finding reinforces the significance of the downstream merger event as a structural break in the prices of dependent supplier industries. 3.4 Explaining the Decline in Prices: Countervailing Acquisitions While the tests in Section 3.3 clearly establish a decline in the prices of dependent supplier industries in the period following downstream consolidation, important questions remain unaddressed. Firstly, the decline in prices may not necessarily be due to the exercise of buying power by the merging industry. If the downstream merger creates production efficiencies that allow the merging industry to produce the same amount of output with lower input usage, the demand for inputs and, therefore, the price of inputs would decline. This type of an efficiency story could just as easily explain the decline in supplier prices after downstream consolidation. Secondly, can we be certain that the downstream merger is not itself endogenous to events occurring in the upstream industries? In this section, we address these issues by taking Hypothesis 3 and Hypothesis 4 to the data. In the process, we make the argument that the downstream consolidation is not an exogenous event but is an integral part of the shifts in market power between supplier and customer industries. Countervailing theory suggests that suppliers that have built up industry concentration would be the most adversely affected by a major consolidation downstream. In Section 3 above, we found that suppliers who are the most dependent on the downstream merging industry experienced the most adverse changes in selling prices after the downstream consolidation. This finding would be consistent with countervailing theory if the most dependent supplier industries had experienced an increase in industry concentration, possibly through horizontal mergers, in the period preceding the major downstream consolidation. Therefore, our third hypothesis states that dependent supplier industries engaged in significantly greater horizontal merger activity prior to the downstream merger event. To test this hypothesis, we examine merger activity in 18

20 supplier industries over the three years preceding the downstream merger. Using data from SDC Platinum, we first obtain the number of horizontal acquisitions announced in each of the 10 most dependent supplier industries during the three years preceding the downstream consolidation. As before, horizontal mergers are defined as deals where the acquirer and target operate in the same primary 4-digit SIC code. We scale the number of horizontal acquisitions announced in each supplier industry by the total number of firms in that industry. This gives us a measure of horizontal merger activity in each supplier industry. We also create an index of non-horizontal merger activity and unrelated merger activity in a similar manner. Non-horizontal acquisitions are defined as deals where an acquirer in the supplier industry buys a target firm that does not share the same 4-digit SIC code. However, the acquirer and target could share the same primary 3-digit, 2-digit or 1-digit SIC code. Unrelated acquisitions are defined as deals where an acquirer in the supplier industry buys a target firm that does not even share the same 1-digit SIC code. To test whether dependent suppliers were engaged in consolidating merger activity during the three years preceding the major downstream event, we regress these measures of supplier merger activity on the dependence dummy, D. The following equation is estimated using OLS with robust standard errors clustered at the 2-digit SIC level. MergerActivity j = β 0 +β 1 D j + β 2 Shock j + β 3 Energy j + β 4 RnD j +ε j (4) The dependent variable MergerActivity measures merger activity in the supplier industry during the three years preceding downstream consolidation. We use the three different measures of merger activity horizontal merger activity, non-horizontal merger activity or unrelated merger activity as dependent variables in three separate regressions. The right-hand side variables remain unchanged in the three regressions. The main explanatory variable is the Dependence dummy, D, which, as before, takes a value of one for suppliers in the top quintile of supplier dependence and zero otherwise. The control variables in the regressions are based on Mitchell and Mulherin s (1996) study of the inter-industry patterns in the rate of takeovers and restructurings. These include industry sales shocks, employment shocks, energy dependence and research and development (R&D) expenditures. 12 Energy dependence, Energy, is calculated as 12 Deregulation is also considered an important determinant of merger activity (see for example, Mitchell and Mulherin (1996) and Andrade, Mitchell and Stafford (2001)). Andrade, Mitchell and Stafford (2001) classify the 19