The 2010 U.S. Horizontal Merger Guidelines: From Hedgehog to Fox in Forty Years
By Carl Shapiro*
The U.S. Department of Justice and the Federal Trade Commission recently updated their Horizontal Merger Guidelines,1 which build upon and replace the 1992 Guidelines.2 The revised Guidelines are the product of an extensive team effort at the Agencies that took place over roughly a year, under the leadership of Assistant Attorney General Christine Varney and FTC Chairman Jon Leibowitz. The process for revising the Guidelines was lengthy, collaborative, and open: the Agencies posted a series of questions, inviting public comment on possible revisions; numerous useful public comments were received and reviewed; the Agencies sponsored five public workshops at which panelists discussed possible revisions to the Guidelines; subsequently, the FTC made public a draft of the proposed Guidelines, again inviting additional public comments; numerous thoughtful comments were again received and reviewed; and in response to those comments, the proposed Guidelines were further clarified.3 Inevitably, however, many of the questions raised in the public comments submitted in response to the proposed Guidelines are not explicitly addressed in the final Guidelines. In this article, I respond to some of those questions, especially the questions pertaining to the economic principles underlying the revised Guidelines. I also elaborate in greater detail on some of the points made in the Guidelines themselves.4
The 2010 Guidelines are best understood in historical context. They reflect the ongoing evolution of merger enforcement that has taken place since the DOJ first issued merger guidelines in 1968. The 2010 Guidelines rely heavily on earlier versions of the Guidelines, especially those released in 1982 and 1992, and on the 2006 Commentary on the Merger Guidelines. Many of the approaches in the 2010 Guidelines that some commentators have considered novel actually are contained in those earlier statements of merger enforcement policy.
Isaiah Berlin’s famous allusion to the different ways in which the Hedgehog and the Fox view the world is a useful model for how to think about the evolution of the Merger Guidelines. The hedgehog knows one big thing.5 Likewise, the 1968 Guidelines were based on one big idea: horizontal mergers that increase market concentration inherently are likely to lessen competition.6 By today’s standards, the 1968 Guidelines are rather shocking. For example, in a market in which the combined share of the four largest firms is at least 75 percent, they state that the Department “will ordinarily challenge” a merger if the acquiring firm’s share is at least 15 percent and the acquired firm’s share is at least 1 percent.7 Few would advocate such an enforcement stance today.
However, this focus on market concentration reflected unambiguous Supreme Court precedent. In Brown Shoe, the Court stated: “The dominant theme pervading congressional consideration of the 1950 amendments [to § 7 of the Clayton Act] was a fear of what was considered to be a rising tide of economic concentration in the American economy.”8 In Philadelphia National Bank, the court quoted this passage from Brown Shoe and then stated:
One cannot help but marvel at how far merger enforcement has moved over the past forty years, with no change in the substantive provisions of the Clayton Act and very little new guidance on horizontal mergers from the Supreme Court. But the Court has given a great deal of guidance in Sherman Act cases, moving away from simple rules and towards an approach emphasizing the practical reality of the market and the likely effects of the practice in question. As Justice Souter explained in California Dental, “What is required . . . is an enquiry meet for the case, looking to the circumstances, details, and logic of a restraint.”10
Returning to Berlin’s prototypes, the fox knows many things. Likewise, merger enforcement in recent years has become increasingly eclectic, reflecting the enormous diversity of industries in which the Agencies review mergers and the improved economic toolkit available. The Agencies and the courts look at a wide variety of evidence and use a wide variety of methods to determine whether mergers may substantially lessen competition. Based on decades of experience examining mergers, the Agencies recognize that each industry has unique features and each merger presents unique circumstances.
The transition of merger enforcement from hedgehog to fox can be traced through the various merger guidelines published from 1968 to 2010. At times, most notably in 1982, new guidelines have spurred changes in Agency enforcement practice. At other times, including 2010, new guidelines have primarily been an exercise in transparency, reflecting ongoing changes in Agency enforcement practice and advances in economic learning.
The 1982 Guidelines were a revolution. Five innovations formed the foundation on which all subsequent Merger Guidelines have been built:
While the 1982 Guidelines were a dramatic step forward in merger enforcement policy, they proved to be limited in some respects due to their heavy emphasis on what today we refer to as “coordinated effects,” and specifically the danger that the merger would increase the likelihood of collusion, either express or tacit.
The 1982 Guidelines were written with relatively homogeneous, industrial products in mind. Product differentiation was considered as a factor affecting the ease and profitability of collusion, that “will be taken into account only in relatively extreme cases.”14 This mindset reflected longstanding antitrust concerns about the performance of concentrated markets for basic industrial commodities. Antitrust attention was focused on markets of this type during the industrial age—the age of steel. The Sherman Act itself was motivated by concerns about collusion in markets for homogeneous products, which took the form of the 19th century trusts. The HHI thresholds were thus best suited to evaluate concerns about collusion in markets for homogeneous products. Indeed, in his classic 1964 article, George Stigler derived expressions involving the HHI from a model of collusion.15
The Guidelines were slightly revised in 1984, but the next major change arrived with the 1992 Guidelines, the first that were jointly issued by the DOJ and the FTC.16 The 1992 Guidelines increased the sophistication of the economic analysis and explained more fully how the Agencies evaluate various types of competitive effects. These changes reflected the accumulation of Agency experience and the advance of economic learning during the 1980s. Two innovations in the 1992 Guidelines stand out.
First, the most significant advance in the 1992 Guidelines was their introduction of “unilateral effects.” The earlier guidelines had focused almost exclusively on coordinated effects. They considered what we now call “unilateral effects” only via their “leading firm proviso,” which comprised a single paragraph in the 1984 Guidelines.17 In recent years, more DOJ investigations have involved unilateral effects than coordinated effects. The 2010 Guidelines build upon the treatment of unilateral effects in the 1992 Guidelines.
Second, the 1992 Guidelines introduced a more detailed and sophisticated analysis of entry. Entry analysis in the 1992 Guidelines is built upon the principle that entry must be “timely, likely, and sufficient” to deter or counteract the competitive effects of concern. The 2010 Guidelines retain this basic approach to the analysis of entry.18
The leading firm proviso in the 1984 Guidelines stated that “the Department is likely to challenge the merger of any firm with a market share of at least one percent with the leading firm in the market, provided the leading firm has a market share that is at least 35 percent.”19 The aim of the proviso was to prevent “mergers that may create or enhance the market power of a single dominant firm.”20
The 1992 Guidelines expanded on the leading firm proviso, developing the idea of unilateral effects, i.e., that eliminating competition between the merging firms could itself constitute a substantial lessening of competition, even without post-merger coordination between the merged firm and its remaining rivals.21 Critically, the 1992 Guidelines explained how such unilateral effects could be diagnosed in markets with differentiated products, where the adverse competitive effects of concern typically are not uniform throughout the relevant market. The introduction of unilateral effects in the 1992 Guidelines reflected and anticipated a shift in merger enforcement away from relatively homogeneous industrial commodities and towards more differentiated products. While the Guidelines necessarily apply to all industries, the 1992 Guidelines were a major step in the evolution of antitrust enforcement from the industrial age to the information age.
The next change to the Guidelines was the substantial revision and expansion in 1997 of the treatment of merger efficiencies.22 The 1997 changes reflect an appreciation that mergers can promote competition by enabling efficiencies, and that such efficiencies can be great enough to reduce or reverse adverse competitive effects that might arise in their absence. The 2010 Guidelines make very few changes to the treatment of efficiencies articulated in 1997.
The 2010 Guidelines reflect the ongoing trend in merger enforcement from hedgehog to fox that has continued since 1992.
Many observers have noted specifically that the 2010 Guidelines place less weight on market shares and market concentration than did predecessors. This is a central example of the fox’s eclectic approach, tailoring the methods used to the case at hand and to the available evidence.
The 2010 Guidelines also follow a more integrated and less mechanistic approach. Section 0.2 from the 1992 Guidelines described a step-bystep approach followed by the Agencies: (1) market definition and concentration; (2) competitive effects; (3) entry; (4) efficiencies; and (5) failing firm defense. Even in 1992 the Agencies did not rigidly follow these steps, and by 2009 many witnesses observed at the hearings that they gave an inaccurate impression of Agency practice. The 2006 Commentary acknowledged as much, stating that “the Agencies do not apply the Guidelines as a linear, step-by-step progression that invariably starts with market definition and ends with efficiencies or failing assets.”24 There was a consensus at the hearings that new guidelines should reflect the movement away from the step-by-step approach described in the 1992 Guidelines to a more integrated approach that does not necessarily start with market definition or base predictions of competitive effects primarily on market concentration.25
The revised Guidelines emphasize that merger analysis ultimately is about competitive effects. The new Section 2, “Evidence of Adverse Competitive Effects,” provides guidance about the types of evidence the Agencies normally seek, and the sources of evidence the Agencies normally use, to inform their analysis of competitive effects. The section is placed near the front of the Guidelines because investigations usually start with the formulation of candidate theories of harm to competition and the exploration of evidence to support or reject those theories. In most cases, especially where market boundaries are unclear, DOJ staff will analyze evidence of possible harm before it has determined the scope of the relevant market. Indeed, the same piece of evidence may be relevant to competitive effects and to market definition, as emphasized in the 2006 Commentary.26 The 2010 Guidelines make a similar observation in Section 4: “Evidence of competitive effects can inform market definition, just as market definition can be informative regarding competitive effects.”
Thus, like the fox, the 2010 Guidelines embrace multiple methods. But this certainly does not mean they reject the use of market concentration to predict competitive effects, as can be seen in Sections 2.1.3 and 5. The 2010 Guidelines recognize that levels and changes in market concentration are more probative in some cases than others. In particular, as the revised Guidelines explain, the Agencies place considerable weight on HHI measures in cases involving coordinated effects.27 The statement that “merger analysis does not consist of uniform application of a single methodology” certainly also does not mean that the DOJ will dispense with identifying the relevant line of commerce and section of the country when going to court to challenge a merger.28 Instead, it means that predictions about competitive effects may rely on evidence other than market shares and market concentration. For this reason, the revised Guidelines state in Section 4: “The measurement of market shares and market concentration is not an end in itself, but is useful to the extent it illuminates the merger’s likely competitive effects.”29
Concern that the revised Guidelines, with their more flexible approach, provide less valuable guidance to the business community and increase the uncertainty faced by companies considering or undertaking horizontal mergers is unwarranted.
First, the revised Guidelines, by increasing transparency and providing more up-to-date guidance, should allow the business community to assess more accurately how the Agencies are likely to evaluate proposed horizontal mergers. The public hearings confirmed our internal assessment that actual practice had departed from the 1992 Guidelines. To a considerable degree, these departures were already reflected in the 2006 Commentary: “In some investigations, before having determined the relevant market boundaries, the Agencies may have evidence that more directly answers the ‘ultimate inquiry in merger analysis,’ i.e., ‘whether the merger is likely to create or enhance market power or facilitate its exercise.’”30
To respond to this discrepancy between the 1992 Guidelines and actual practice, both Assistant Attorney General Varney and Chairman Leibowitz stated their goal was to provide transparency by updating the Guidelines themselves, while referencing the 2006 Commentary as a useful supplement to the 2010 Guidelines. For example, Assistant Attorney General Varney explained in a speech in January 2010 that a major goal of revising the Guidelines was to provide greater transparency:
Second, the supposed simplicity and predictability based on market definition and market concentration was more apparent than real. Market definition is often disputed. In many merger investigations, such as the Staples or Whole Foods cases,32 the merging parties assert a broad market in which they argue that the post-merger HHI or the change in HHI is small, but the Agencies respond that the hypothetical monopolist test properly leads to a narrower market. Unfortunately, completely eliminating any uncertainty about the results of the hypothetical monopolist test is not possible. It is inherent in the need to measure “reasonable” interchangeability. Some of this uncertainty can be reduced, however, when one focuses on competitive effects rather than the line-drawing exercise of market definition.
Furthermore, placing greater weight on market concentration does not eliminate uncertainty. The 1992 Guidelines state: “Where the post-merger HHI exceeds 1800, it will be presumed that mergers producing an increase in the HHI of more than 100 points are likely to create or enhance market power or facilitate its exercise.”33 Merger enforcement data show that this presumption has frequently been overcome.34 Few would favor giving the business community greater certainty by making this presumption irrebuttable.
Third, the tradeoff between simple bright lines and accuracy is inherent in the antitrust review of proposed horizontal mergers. This fundamental tradeoff has been a consideration going back to Philadelphia National Bank and the 1968 Guidelines.35 The 1968 Guidelines are anything but flexible, but I doubt the business community would welcome a return to those Guidelines. Accounting for the real-world business conditions in which a merger takes place is worthwhile, even if doing so means that some simplicity must be sacrificed to achieve greater accuracy in merger enforcement. The second paragraph in the 1982 Guidelines states:
Lastly, of specific relevance to businesses considering mergers, the vast majority of mergers reported under the Hart-Scott-Rodino Act (HSR) do not trigger a second request for information from the Agencies. During the ten-year period from Fiscal Year 1999 through Fiscal Year 2008, the percentage of all HSR transactions involving a second request varied annually from a low of 2.1 percent to a high of 4.3 percent.37 The detailed analysis of competitive effects described in the Guidelines is most relevant to transactions that join together two substantial competitors among a few; these are well less than 5 percent of HSR transactions. Among those mergers, where the Agencies conduct a thorough investigation, experienced practitioners already know that “investigations are intensively fact-driven iterative processes.”38
In practice, economic analyses of mergers often focus on certain quantitative measures, such as prices, costs, market shares, or demand elasticities. But that does not indicate any tendency for DOJ investigations to favor quantitative evidence over qualitative evidence. In practice, a great deal of investigative time and effort is expended to develop qualitative evidence, e.g. by reviewing documents and conducting interviews, and such evidence typically is central to our evaluation of likely competitive effects. The concepts described in the Guidelines inform the gathering and interpretation of this evidence.39 The 2010 Guidelines, like all of their predecessors, provide a high-level economic framework within which investigative work takes place.
The biggest shift in merger enforcement between 1992 and 2010 has been the ascendency of unilateral effects as the theory of adverse competitive effects most often pursued by the Agencies. Prior to 1992, merger enforcement focused primarily on coordinated effects. In recent years, a sizeable majority of DOJ merger investigations have focused on unilateral effects. Along with this pronounced shift in practice has come considerable new economic learning about unilateral effects. This shift in practice and advance in learning regarding unilateral effects was one of the chief reasons we at the DOJ felt that the time had come to update the Guidelines.40
Section 6 in the 2010 Guidelines, “Unilateral Effects,” is broken into four parts. These parts describe the distinct modes of analysis that the Agencies use to investigate unilateral effects in different market settings. Sections 6.1 and 6.2 address pricing and bidding competition among suppliers of differentiated products; they are closely related descendents of Section 2.21 from the 1992 Guidelines. Section 6.3 addresses capacity and output for homogeneous products; this part descends from Section 2.22 from the 1992 Guidelines. Section 6.4 addresses innovation and product variety and is entirely new.
Section 2.21 in the 1992 Guidelines dealing with pricing of differentiated products was a major advance over the leading firm proviso in the 1984 Guidelines. This section introduced into the Guidelines two important strands of research from the field of industrial organization economics: (1) pricing competition among suppliers of differentiated products, including the workhorse Bertrand model; and (2) bidding competition in procurement settings. These two strands have been separated in the 2010 Guidelines.
The basic economic principles articulated in Section 2.21 of the 1992 Guidelines are fundamental and should not be controversial.41The 2010 Guidelines rely heavily on these basic principles. This key passage from Section 2.21 of the 1992 Guidelines has been retained virtually unchanged:
The central role of diversion between the products sold by the merging firms is then stressed:
Economists have long measured diversion from one product to another using the cross-elasticity of demand between the two products, and elasticities have been used in antitrust for decades to measure “reasonable interchangeability.”44 By 1995, DOJ was using the term “diversion ratio,” to capture this same concept in a more intuitive way. The diversion ratio from Product 1 to Product 2 is defined as the percentage of unit sales lost by Product 1, when its price rises, that are captured by Product 2.45
Section 6.1 in the 2010 Guidelines, like Section 2.21 in the 1992 Guidelines, explains how the Agencies assess the impact of the merger on pricing competition. But the very same concepts can be applied to non-price competition. For example, one can examine how improvements in the quality of a product sold by one merging firm capture sales from a product sold by the other merging firm. The “quality” diversion ratio need not equal the normal (price) diversion ratio.
The focus on diversion in the 1992 Guidelines was impeccable in terms of the underlying economics. But it presented a conundrum: how could this approach be reconciled with the emphasis on market shares found in the case law and perpetuated in the 1992 Guidelines? In a path-breaking article, Robert Willig, one of the primary authors of the 1992 Guidelines, showed the way.46 First, Willig acknowledged the challenge: “On the face of it, this perspective appears to remove consideration of market shares from merger analysis since there are no obvious systematic relationships among market shares and cross-price derivatives of demand.”47But then Willig identified certain conditions under which “market shares can be accurate indicators of the competitive effect of a merger between producers of differentiated products.”48The required conditions were subsequently described in the 1992 Guidelines:
In modern parlance, these are the circumstances in which market shares yield good proxies for diversion ratios.50 In particular, as Willig demonstrates, the diversion ratio from Product 1 to Product 2 is proportional to S2/(1 – S1), where S2 and S2 are the market shares of Products 1 and 2.51 Connecting market shares and unilateral price effects in this way was a theoretical tour de force. But Willig was very careful to emphasize the limitations of this approach. "We shall see that the assumptions are unlikely to be valid in many areas of application where specific information can be developed about product characteristics and about consumer preferences for them. For such applications, merger analysis that focuses exclusively on market shares is likely to go awry."52 Furthermore, even under those special circumstances in which market shares are informative, even Willig, for all his theoretical prowess, could not relate the level of the HHI to diversion ratios.53
Consequently, the treatment in the 1992 Guidelines of unilateral price effects in markets with differentiated products suffered from a mismatch between the basic theory of differentiated product pricing competition, which emphasizes diversion, and the Guidelines' historical reliance on market shares and HHIs. As one commenter expressed it at the Stanford Workshop, the 1992 Guidelines were like a centaur: the head of differentiated products pricing was grafted onto the body of market definition and market concentration.54
This left the 1992 Guidelines in an uncomfortable state: the link they emphasized between market shares and unilateral price effects rested on a strong assumption about demand (i.e., markets shares are good proxies for diversion ratios) that often cannot be justified. Willig anticipated this difficulty, writing: “The analysis here also points to the strong need to develop information beyond shares in markets with differentiated products, particularly the relative proximity of the products of the merging firms in the space of salient characteristics.”55Indeed, this is just how practice has evolved since 1992: the DOJ looks at a wide variety of evidence to assess whether the products offered by the merging firms are close substitutes and to measure diversion ratios when possible, sometimes but not always starting with shares in plausibly defined markets.
Spurred by the 1992 Guidelines, and in parallel with major advances in practice, the economic literature relating to unilateral price effects, including the estimation of demand and full merger simulation, developed over the past eighteen years. Many Ph.D. theses have been written about estimating demand systems with differentiated products, and considerable strides have been made in developing simpler approaches that are feasible when data are limited.56I cannot possibly do justice to that literature here; in any event, it has been well surveyed quite recently.57 Suffice it to say that enormous strides have been made in theory and in practice.58
As economic learning and practice evolved, the emphasis on market shares found in Section 2.21 of the 1992 Guidelines became less helpful to achieve transparent and accurate merger enforcement using a unilateral-effects theory. For example, in a recently litigated case, the court, citing the relevant passage from the Guidelines, rejected the FTC’s attempt to invoke the 35 percent presumption quoted above.59 In that case at least, the court wanted more data to support the unilateral effects theory. DOJ economists routinely look for this type of evidence.
The 2010 Guidelines modestly update the treatment of unilateral price effects to reflect the substantial changes in economic learning and Agency practice since 1992.60Two aspects of that updating are of special significance: (1) reduced emphasis on market shares, and (2) introduction of the “value of diverted sales” as an indicator of upward pricing pressure.
Before turning to those two topics, it is worth highlighting that all of this analysis involving diversion takes as given the set of products being offered and thus does not account for the supply-side responses of repositioning and entry.61 Although a number of comments criticized the revised Guidelines for purportedly establishing unjustified presumptions about unilateral price effects based on diversion ratios and margins, the Guidelines explicitly state:
This language, however, led to criticism that the revised Guidelines take an overly skeptical approach to repositioning by treating it like entry. Yet the same basic approach can be found in the 1992 Guidelines.63 The 2006 Commentary observed that in practice repositioning has rarely been a significant factor:
The revised Guidelines recognize that the ease or difficulty of repositioning varies greatly across markets.65
1. Reduced Emphasis on Market Shares
The 2010 Guidelines do not explicitly link diversion ratios to market shares. This reflects experience gained over the years: while market shares are often a useful starting point for assessing diversion ratios, and can indeed be used as proxies for diversion ratios, the DOJ will normally look as well for more direct evidence of diversion ratios. The new language states:
The revised Guidelines go on to state:
This differs somewhat from the 1992 Guidelines, which stated:
The revised Guidelines reflect Agency practice, which involves assessing whether the price of any product sold by the merging firms is likely to increase significantly due to the merger. That depends heavily on diversion to products sold by the merging partner, not on any market-wide measure.69The central role of diversion between the merging parties is explained this way:
Some comments criticized this passage for purportedly downplaying the importance of competition from products offered by non-merging firms. However, that criticism is inapt: if products offered by non-merging firms are close substitutes for a product sold by a merging firm, diversion to those products will normally be high, necessarily depressing the diversion ratio to products sold by the other merging firm.71 This same point was explicitly made in the 2006 Commentary:
In a merger joining Products 1 and 2, significant unilateral effects for Product 1 can occur even if Product 2 is not the "closest substitute" overall to Product 1. What these effects require is that a significant percentage of the customers purchasing Product 1 consider Product 2 to be their next second choice. That percentage is captured by the diversion ratio.
DOJ puts far more weight on diversion ratios and margins (see below) than on the HHI level when diagnosing unilateral price effects. This has been the case for many years, and again the 2006 Commentary made clear that HHI levels are of limited predictive value for this purpose:
As noted below, the market shares of the merging firms, and the change in the HHI, are more informative in this context than the level of the HHI.
These changes in practice had left many practitioners uncertain about whether and how the Agencies use HHIs in cases involving unilateral price effects for differentiated products. The revised Guidelines clarify the role of HHIs in such cases:
The express acknowledgement that HHI levels typically are not very helpful diagnostics in these cases has led to concerns that the valuable screening role played by the HHI thresholds since 1982 has been reduced or lost. In fact, the 2010 Guidelines recognize the importance of these HHI thresholds to help identify mergers that are "unlikely to have adverse competitive effects and ordinarily require no further analysis.”75 Indeed, the 2010 Guidelines not only retain HHI thresholds but raise them. DOJ continues to apply the HHI thresholds to all horizontal mergers.76Of course, HHIs can only be calculated after a relevant market has been defined, so uncertainty about the scope of the relevant market necessarily creates uncertainty about applicable levels and changes in the HHI. Below, I discuss market definition in cases involving differentiated products.
The combined shares of the merging firms, and the change in the HHI, can be useful and informative metrics in unilateral effects cases, and these measures are used by the Agencies. If diversion is proportionate to market share, the diversion from Product 1 to Product 2 is proportionate to S2/(1−S1), which can be approximated as S2×(1+S1) if S1 is not too large.77 Approximating the diversion ratio from Product 2 to Product 1 in the same way, and adding up the two diversion ratios, gives S1 + S2 + 2×S11×S2 which equals the combined share of the merging firms plus the change in the HHI. Unilateral price effects are unlikely if the change in the HHI is less than 100, which corresponds to a merger between firms with market shares of 5 percent and 10 percent.
Nonetheless, the revised Guidelines do not retain the presumption that the merging firms are significant direct competitors if their combined market share is at least 35 percent. This presumption was dropped, for four reasons. First, the 1992 Guidelines did not provide a specific basis for the 35 percent figure. Evidently, it was taken from the 35 percent figure used in the leading firm proviso since 1982. But that proviso was based on a very different model and theory: the dominant firm/competitive fringe model in a market for a homogeneous good. Second, as practice evolved, the 35 percent presumption was often invoked as a safe harbor, with merging parties frequently asserting that, according to the Guidelines, there could be no substantial unilateral price effects if their combined share of the relevant market was less than 35 percent. In fact, the 1992 Guidelines contain no such safe harbor.78 Nor would one be justified: a merger combining two products that are close substitutes can lead to substantial unilateral price increases for those products even if their combined market share is less than 35 percent. Third, the presumption could only properly be invoked if market shares are a reasonable proxy for diversion ratios. As discussed above, DOJ often uses market shares to assess diversion, and higher shares in a properly defined relevant market do generally go along with elevated concern about unilateral price effects. But we also look for more direct evidence of diversion. Fourth, as emphasized in this article, economic theory relates unilateral price effects with differentiated products more directly to diversion ratios and margins than to the combined market share of the merging firms.
2. The Value of Diverted Sales
The 2010 Guidelines introduce the “value of diverted sales” into the analysis of unilateral price effects with differentiated products:
The basic economics underlying the “value of diverted sales” concept are not new. Suppose that the merger brings under common ownership Product 1, formerly owned by Firm 1, and Product 2, formerly owned by Firm 2. One key question is whether the merger is likely to lead to a significant price elevation for Product 1?80 As stressed above, repositioning and entry are not considered at this point in the analysis, which takes as given the set of competing products offered by non-merging firms. One can also take as given the prices charged by non-merging rivals for their products. Holding these prices fixed typically will lead to an under-estimate of the magnitude of the post-merger price change.81
With these simplifications, the central question can be posed very specifically: “Taking as given all other products and their prices, is the profit-maximizing price for Product 1 significantly higher for a firm that owns both Product 1 and Product 2 than it was for Firm 1, which owns just Product 1?” The answer to this question depends entirely on (a) how the demand for these two products varies as their prices rise above pre-merger levels, and (b) their pre-merger margins.82
As discussed in more detail below, this is precisely the same question posed by the hypothetical monopolist test to see if Products 1 and 2 form a relevant market. This very tight connection between unilateral price effects with differentiated products and market definition was not clear in earlier Guidelines. The Guidelines now clarify this relationship by explaining in more detail how the hypothetical monopolist test works with differentiated products.83
As a first step to answering this question, it is instructive to simplify even further by holding fixed the price of Product 2 and asking how common ownership of Product 2 changes the pricing incentives for Product 1, starting at pre-merger prices. Studying these incentives requires far less information than estimating the profit-maximizing price increase for Product 1.
To see how common ownership changes incentives, it is a bit easier to think in terms of the incentives to sell more units of Product 1 (the reverse of raising the price of Product 1). Owning Product 2 creates a disincentive to sell more units of Product 1. Suppose that for every four extra units sold of Product 1 by lowering its price, one fewer unit of Product 2 is sold. This corresponds to a diversion ratio of 25 percent. The higher the diversion ratio, the greater the disincentive to sell units of Product 1 created by the merger. So far so good, as per the 1992 Guidelines. The logical—and unavoidable—next step is to ask how cannibalizing sales of Product 2 affects the merged firm’s profits from selling more units of Product 1. Lost unit sales of Product 2 only affect the merged firm’s profits to the extent that those sales were contributing to profits, i.e., to the extent that price exceeds marginal cost for Product 2. This directs our attention to the gap between price and marginal cost for Product 2. This is just arithmetic.84
Suppose that Products 1 and 2 each sell for $100,000, and the marginal cost of each is $60,000, so each unit sold contributes $40,000 towards covering fixed costs and earning profits. For every four extra units sold of Product 1, one unit of Product 2 is cannibalized, leading to a lost contribution of $40,000. Thus, every extra unit sold of Product 1 reduces Product 2’s contribution by $10,000. Combining the ownership of Products 1 and 2 thus creates a $10,000 per-unit disincentive to sell units of Product 1. In economic terms, the merged entity bears a $10,000 per-unit opportunity cost not borne by Firm 1.85
Moving beyond this specific numerical example, the per-unit opportunity cost of selling Product 1 that is borne (internalized) by the merged firm but not Firm 1 is equal to D12(P2 − C2), where D12 is the diversion rate from Product 1 to Product 2, P2 is the price of Product 2, and C2 is the marginal cost of Product 2. The opportunity cost is equal to the multiplicative product of the diversion ratio and the margin.86 Neither the diversion ratio nor the margin operates alone to generate upward pricing pressure.
These ideas are at least twenty years old, as the Willig reference shows, and are not new at DOJ. When I served as Deputy Assistant Attorney General for Economics in 1995 I wrote:
For example, the DOJ's 1997 challenge to the proposed merger between Vail Resorts and Ralston Resorts noted the central role of diversion ratios and margins in unilateral price effects:
The 2010 Guidelines move beyond diversion ratios, directing attention to the “value of diverted sales.” The “value of diverted sales” incentive measure is constructed from the multiplicative product of a diversion ratio and a margin.
Consider a small price increase on Product 1, which we denote byP1. Holding fixed all prices other than P1, this will cause the unit sales of Product 1 to fall by some amount, call it
The value of diverted sales is usefully measured in proportion to the reduction in unit sales of Product 1 resulting from the price increase, i.e., DX1. On this per-unit basis, the value of diverted sales is equal to
This equals the opportunity cost term, D12(P2 − C2) that emerged inexorably out of the basic logic of unilateral price effects. The next and final step in this line of reasoning is to scale this opportunity cost in proportion to the price of Product 1. This gives D12(P2 − C2) / P1, which is a gross upward pricing pressure index for Product 1.90 We label this very useful index as
The Guidelines now provide a condition under which unilateral price effects are unlikely:
This condition corresponds to a low value of the GUPPI. As noted above, the value of diverted sales is equal to VX2 (P2 − C2). The lost revenues attributable to the reduction in unit sales of Product 1 are given by LX1× P1. Measuring the value of diverted sales in proportion to the lost revenues gives
which equals GUPPI1. Denoting the relative margin on Product 2 as M2 = (P2 − C2) / P2, the gross upward pricing pressure index on Product 1 can be expressed as
If the two products have equal prices, this index becomes simply GUPPI1= D12M2.
Summarizing, the revised Guidelines direct attention to the disincentive created by the merger to sell additional units of Product 1 if these cannibalize unit sales of Product 2. This disincentive is measured as an opportunity cost borne by the merged firm for selling Product 1. That opportunity cost, scaled in comparison to the price of Product 1, is equal to the multiplicative product of the diversion ratio to Product 2 and the margin on Product 2. Unilateral price effects for Product 1 are unlikely if this measure is small.
Focusing in this way on how the merger changes pricing incentives achieves two important goals. First, the treatment of unilateral price effects in the Guidelines now rests on a rock solid economic foundation.92 The economic principles used are extremely basic and robust: (a) firms account for opportunity costs (cannibalization) when pricing and promoting product lines containing substitute products, and (b) higher costs tend to lead to higher prices.93 Second, the Guidelines now identify circumstances under which unilateral price effects for a given product are unlikely: when the opportunity cost term for that product is small as a fraction of that product’s price. Because the gross upward pricing pressure index is so well grounded in basic economics, a quasi-safe-harbor based on this index does not suffer from the mismatch between the economic logic of unilateral price effects and a quasi-safe-harbor based on the HHI level.94
This approach also indicates how to incorporate efficiencies into the analysis. For example, merger-specific reductions in the marginal cost of Product 1 create an incentive to lower the price of Product 1. In particular, efficiencies create downward price pressure that can reduce or reverse the incentive to raise price just discussed. One of the attractive features of the revised Guidelines is that efficiencies can easily and naturally be integrated into the analysis. One can directly compare any merger-specific reduction in marginal cost for Product 1 with the opportunity cost due to cannibalization.95 A merger thus generates net upward pricing pressure for Product 1 if the opportunity cost exceeds the efficiencies for that product.96 The value of diverted sales measure used in the Guidelines, scaled as GUPPI, indicates how large the marginal cost savings must be on Product 1, measured as fraction of the price of Product 1, for there to be no net upward pricing pressure on Product 1, given the price of Product 2.
The value of diverted sales, taken alone, does not purport to quantify the magnitude of any post-merger price increase. Rather, as the Guidelines state, it “can serve as an indicator of the upward pricing pressure on the first product resulting from the merger.”97 This is an important distinction not appreciated in some comments. In Appendix A, I elaborate on this point. The value of diverted sales is a measure of the extra (opportunity) cost the merged firm bears in selling units of Product 1. Higher costs give the merged firm an incentive to raise the price of Product 1. But further analysis is needed to determine how that cost increase translates into a price increase. That depends upon the rate at which costs are passed-through to prices, which in turn depends upon the curvature of the demand curve.98 Pass-through rates are important but can be difficult to estimate empirically. If the elasticity of demand is constant for small price changes, the pass-through rate is greater than one. If unit sales are equally sensitive to small price increases and decreases, demand is linear and the pass-through rate is one-half. In the extreme, if demand were sharply kinked at pre-merger prices, meaning buyers are far more sensitive to price increases than price decreases, the pass-through rate would be low, and even a large incentive to raise price would not translate into a significant price increase. Kinks are implausible when demand comes from multiple diverse buyers; kinks also generally lack empirical support.99
The value of diverted sales is an excellent simple measure for diagnosing or scoring unilateral price effects, but it cannot capture the full richness of competition in real-world industries. Indeed, as stressed above, all of the quantitative methods discussed here must be used in conjunction with the broader set of qualitative evidence that the Agencies assemble during a merger investigation.
A thorough analysis often must do more than just quantifying how the merger changes pricing incentives. Further information about demand is needed, and additional analysis is required, to translate these incentives into predictions of post-merger price increases. To accomplish this, DOJ economists and economists working for merging parties often undertake merger simulation exercises. The revised Guidelines, for the first time, identify merger simulation as a methodology used by the Agencies. In some cases, the DOJ uses merger simulation methods to diagnose unilateral price effects.100 Before using the output of any merger simulation model to actually predict the magnitude of the post-merger price increase, DOJ economists check the model’s output for robustness and consistency with other evidence. We also consider repositioning, entry, and efficiencies.
The competition authorities in the United Kingdom have been using very closely related techniques for the past five years to diagnose unilateral price effects. In its analysis of the proposed acquisition by Somerfield of 115 stores from William Morrison Supermarkets, the UK Competition Commission (CC) computed “illustrative post-merger price rises” based on diversion ratios and margins.
Consistent with these cases, the September 2010 UK Merger Assessment Guidelines emphasize diversion ratios and margins and refers to the illustrative price rise methodology.103 Likewise, the European Commission’s Guidelines on the Assessment of Horizontal Mergers state: “High pre-merger margins may also make significant price increases more likely.”104
Some observers have questioned whether these techniques are practical, given the need to measure diversion ratios and margins, suggesting that they are far more complex than simply measuring HHIs.105 These concerns are easily answered.
First and foremost, DOJ economists and economists working for the merging parties have been measuring diversion and margins for many years. Margins are used in critical loss analysis and are an essential element of market definition under the hypothetical monopolist test, as discussed in more detail below. Diversion ratios have been central to unilateral effects cases since 1992. Yes, there are well-known pitfalls in measuring margins using accounting data, but DOJ economists are well aware of these pitfalls and skilled at overcoming them when the data permit.106 Second, as noted above, in addition to U.S. agency experience, the UK competition authorities have been using these techniques for the past five years.107 Third, the documents of merging parties can be informative regarding diversion ratios and margins. Firms often are keenly interested in identifying the rivals to which they lose business, or from which they can gain business. Businesses are far more likely to ask these questions in their day-to-day operations than they are to ask how customers would respond to a price increase by a hypothetical monopolist. Margins are also central to business decisions. Margins are an essential element of pricing decisions, and the return on a marketing campaign that attracts new customers depends directly on the price/ cost margins that will be earned on those customers. Indeed, in suitable cases, where reasonably reliable measurement of diversion ratios and margins is possible, these techniques can offer a lot. But they are not meant to displace other methods in situations where diversion ratios and margins cannot be measured with reasonable reliability.
This is a good point to address another common criticism of unilateral effects theory: the claim that unilateral effects models “always predict a price increase” and thus are unsuitable for merger enforcement. This assertion is incorrect. First, the criticism ignores efficiencies, repositioning and entry. Efficiencies generate downward pricing pressure that may outweigh the upward pricing pressure, particularly when repositioning and entry mitigate the upward pricing pressure. Second, the criticism erroneously assumes that the Agencies mechanically run a merger simulation model without examining other evidence or exercising judgment. In fact, the Agencies put real weight on these models only when they are reliable and consistent with other evidence. The Guidelines emphasize that the Agencies use qualitative and quantitative evidence together. If a merger simulation model “predicts” a tiny price increase, that may alleviate DOJ concerns—precisely because DOJ understands that these models typically generate at least some post-merger price increase in the absence of any efficiencies. The Guidelines reflect this by stating that unilateral price effects are unlikely if the value of diverted sales is proportionately small. The UK Competition Commission made this same point very nicely:
Although this criticism often is coupled with an apparent preference for HHI analysis, the same criticism could be made about economic models involving the HHI.
For all of these reasons, DOJ investigations mainly use the GUPPI and merger simulation models to provide an indication—not a precise prediction—of whether a merger is likely to cause significant unilateral price effects. Both methods are used in conjunction with other evidence.
The substantial majority of merger investigations at the DOJ involve firms that sell intermediate goods: the customers of the merging firms are themselves businesses, not final consumers. Indeed, in many cases the buyers are themselves large firms; below, I discuss powerful buyers. In the majority of cases I have worked on as Economics Deputy, the merging firms negotiate prices (and other terms and conditions) with their customers. As Section 2.2.2 points out, testimony from well-informed customers can be especially important in these cases.
Section 6.2 in the revised Guidelines, “Bargaining and Auctions,” addresses these very common situations. This section draws heavily from the 2006 Commentary, which contains separate sections on “Unilateral Effects Relating to Auctions”109 and “Unilateral Effects Relating to Bargaining,”110 including numerous examples.
Price discrimination is quite common in these settings. Suppliers often have considerable information about individual customers, including information about customers’ needs or options, customers’ switching costs, and the costs of serving different customers. DOJ often investigates to determine whether certain types of customers, or certain individual customers, are likely to be harmed by a merger. Section 3 in the revised Guidelines, “Targeted Customers and Price Discrimination,” has been added to reflect the importance of these situations in practice.
The Agencies analyze unilateral effects in bargaining and auction situations using similar approaches to those just discussed for differentiated products.111 To see the connection, consider a situation in which suppliers submit sealed bids to win a particular piece of business. The customer picks the most attractive bid, accounting for price, other terms and conditions, and differences among the suppliers in the products and services they offer, their reputation, etc. As a matter of formal economics, this is very similar to the situation just discussed, where suppliers set prices and each of many customers each picks his or her preferred product. In the bidding setting, each supplier tries to judge the relationship between its bid and the probability it will win the business. In the consumer products setting, each supplier tries to judge the relationship between its price and the number of consumers who pick its product. Either way, the supplier sees a negative relationship between its price and the number of units it expects to sell.
The details of unilateral effects analysis depend on the auction format. For example, in the classic English (open-outcry) procurement auction to provide specified goods or services, bidders publicly offer lower and lower prices to provide the required goods or services until the bidding stops. The equilibrium outcome for such an auction is for the bidder with the lowest cost to win at a price equal to the cost of the next most efficient bidder. In this setting, a merger between the two lowest-cost bidders will lead to a price increase, with the size of the price increase equaling the difference in cost between the less efficient merging firm and the next most efficient bidder. This opens up the distinct possibility that a merger will harm some customers—those for whom the merging firms are the two lowest cost suppliers—but not others.
Section 6.2 identifies the key factors that the Agencies consider in bidding and auction settings:
The first of these elements is the bidding analog of the diversion ratio. DOJ economists, and economists for merging parties, have long been working with win/loss data and other bidding and auction data to assess how often one merging firm is the runner-up when the other merging firm wins the business. We also routinely try to assess the second element—the magnitude of the advantage the runner-up merging firm has over rival suppliers. This advantage is likely to be larger, the more highly differentiated are the goods and services offered by the various suppliers. High margins tend to go along with such differentiation.
Customers sometimes structure their procurements in multiple rounds, down-selecting to just two or three suppliers for the final round. This is especially common when the procurement process itself is costly, e.g., because the suppliers must work closely with the customer to understand its needs and to prepare customized bids. In these circumstances, the frequency with which the merging firms met each other as finalists tends to be quite important to our analysis. Normally, when the merging firms are finalists, the customer benefits from competition between them. In that circumstance, we typically seek to determine whether replacing one of the merging firms with another supplier as a finalist would leave that customer in a less favorable negotiating position. Merging firms often claim that certain non-merging firms can and will offer an equally good alternative to customers. Customer evidence can be especially valuable in assessing this claim. There can be some tension between this claim and the presence of significant supplier differentiation. We may test this claim with evidence from procurement events in which the merging firms competed as finalists against these non-merging firms. If they really do offer very close substitutes, one would expect to see relatively low margins in those bidding situations.
The 1992 Guidelines have been widely criticized for putting an undue focus on pricing competition and giving short shrift to innovation. While this is not an entirely fair characterization, arguably the 1992 Guidelines gave the impression that the Agencies did not pay sufficient attention to competition in product quality, service, or innovation. There was a consensus that new Guidelines should do more to acknowledge the importance of non-price competition, especially innovation competition, and to explain how the Agencies incorporate non-price competition into their merger analysis.
The revised Guidelines place far greater emphasis on non-price competition. For expositional reasons, this was done “globally” in the introduction:
The Agencies are well aware of the importance of non-price competition, and especially the enormous importance over the long run of innovation competition in generating consumer benefits. At DOJ, we routinely consider non-price aspects of competition, including service, product quality, and innovation. In some cases, such as over-the-air radio and various Internet-based services and content, the product is free to consumers so competition to attract consumers takes place entirely on non-price dimensions.
Section 6.4, “Innovation and Product Variety,” explains in general terms how the Agencies evaluate whether a merger is likely to significantly harm customers by retarding innovation or reducing product variety. The analysis of innovation comes in two parts.
The first part looks at the shorter-term impact of the merger on the introduction of new products. This part focuses on whether new products being developed by one merging firm will cannibalize significant profits from products sold by the other merging firm. This analysis is much like that in Sections 6.1 and 6.2, in that it focuses on diversion and cannibalization of profits, but the business decisions here involve product introduction, not pricing.
The second part considers the longer-term impact of the merger on innovation. This usually involves looking beyond the products currently being offered, and perhaps even those being developed. This part of the analysis focuses more on the firms’ R&D plans and capabilities. Longer-term effects on innovation can be hard to assess, because of the inherent uncertainty associated with R&D, because of the difficulty of evaluating an organization’s innovation capabilities, and because these effects are more distant in the future. However, they can be very important, due to the critical role of innovation in generating long-term consumer benefits.
The revised Guidelines also add language in Section 10, “Efficiencies,” to clarify that the Agencies recognize and account for the possibility that a merger may generate innovation efficiencies.
Section 6.4 also addresses product variety. The analysis of product variety is very similar to the treatment of shorter-term innovation effects just described. The focus here, however, is on the withdrawal of existing products rather than the cancellation or delay of new products. A very similar approach, focusing on diversion and cannibalization of profits, is applied: “An anticompetitive incentive to eliminate a product as a result of the merger is greater and more likely, the larger is the share of profits from that product coming at the expense of profits from products sold by the merger partner.”115 This passage explains how one can distinguish between reductions of product variety that are “largely due to a loss of competitive incentives attributable to the merger”116 and those that are not anticompetitive. Anticompetitive reductions in product variety may well be accompanied by a price increase on the remaining product.
Market definition plays two roles in the Guidelines. First, market definition specifies the line of commerce and section of the country in which the competitive concern arises. Second, market definition allows the Agencies to identify market participants and measure market shares, which can be informative regarding the merger’s likely competitive effects. The Guidelines retain the basic hypothetical monopolist test used since 1982 to define relevant markets.
The 2010 Guidelines explain more fully (a) how the exercise of defining markets and measuring concentration relates to the ultimate question of whether the merger may substantially lessen competition; (b) why using market concentration measures based on broader groups of substitutes than required by the HMT can be misleading; (c) how the Agencies evaluate and perform critical loss analysis; and (d) how the Agencies define price discrimination markets, including geographic markets based on the locations of customers.
The HMT provides a well-defined and coherent method for delineating the relevant market. The test can be employed even in situations where there is no clear break in the chain of substitutes and where customers differ greatly in their willingness to substitute more distant products in response to a price increase. As Section 4.1.1 of the Guidelines states: “The Agencies use the hypothetical monopolist test to identify a set of products that are reasonably interchangeable with a product sold by one of the merging firms.”117 The HMT plays a very specific role in the Guidelines, Section 4:
A group of products can form a relevant market under the HMT even if there is significant substitution between that group of products and other products: “As a result, properly defined antitrust markets often exclude some substitutes to which some customers might turn in the face of a price increase even if such substitutes provide alternatives for those customers.”119 “Groups of products may satisfy the hypothetical monopolist test without including the full range of substitutes from which customers choose.”120
These statements follow from the economic logic of the HMT. They do not reflect any change in how the Agencies define relevant markets. For example, the 2006 Commentary states:
Some comments have suggested that the Guidelines now point to narrower markets than did the 1992 Guidelines. This is incorrect: the basic HMT remains unchanged. If anything, the opposite is true, since the “smallest market principle” has been relaxed, as I explain next.
The basic HMT dates back to the 1982 Guidelines. The implementation of the test has been slightly modified over the intervening twenty-eight years, during which time we have learned a great deal about the operation of the test, both in theory and in practice. That process continues in 2010.
As noted above, the Guidelines were updated in 1992 to better handle markets with differentiated products. As part of that updating, the 1992 Guidelines explicitly directed attention to the profit-maximizing price increases on the various products controlled by the hypothetical monopolist, recognizing that these price increases typically will not be uniform.124
The 1992 Guidelines implement the HMT using a specific, iterative algorithm.125 Products are added to the candidate market in the order of “next best substitutes” and the exercise is halted once the test is satisfied. “The Agency generally will consider the relevant product market to be the smallest group of products that satisfies this test.”126 The algorithm has much to commend it, but it suffers from a theoretical problem and a practical problem. The theoretical problem is that the “smallest market principle,” can fail to detect a merger as horizontal in some cases where the merging firms sell substitute products and their merger would likely lead to a substantial lessening of competition.127 The practical problem is that one may not be able to identify the “next best substitute” at each stage of the algorithm, yet the outcome of the iterative algorithm can be sensitive to this determination.128 As a result, while the iterative test in the 1992 Guidelines provides a very useful conceptual framework, in practice the Agencies often are unable to implement the test as stated.
Recognizing these difficulties, the revised Guidelines retain the HMT but take a more flexible approach to its implementation. The iterative procedure no longer appears. The smallest market principle is softened, and the scope of its use is explained in Section 4.1.1:
The 2010 Guidelines are more explicit than their predecessors about the role played by price/cost margins in the HMT. Section 4.1.3, “Implementing the Hypothetical Monopolist Test,” begins:
The revised Guidelines have not changed the role of profit margins in the HMT. The central role played by these margins follows from the economic logic inherent in the test. The 2010 Guidelines explain the role of profit margins in a way that reflects Agency experience and practice since 1992 along with advances in economic learning during that time.
The HMT asks a very specific economic question: would a profit-maximizing monopolist controlling a group of products raise the price of at least one of those products by at least a SSNIP? As noted above, the answer to this question depends entirely on (a) how the demand for these products varies as their prices rise above pre-merger levels; and
In principle, one can perform the HMT by estimating the demand for the products in the candidate market, measuring pre-merger margins, and then computing the profit-maximizing prices.132 DOJ economists and the economists consulting for the merging parties routinely devote considerable effort to estimating demand, using whatever reliable and relevant data are available. However, we often lack sufficient data to reliably and robustly estimate the demand system, making it necessary to follow approaches that are less stringent in terms of their data or modeling requirements. Furthermore, since we are often trying, at least initially, to screen mergers based on market concentration, it is highly desirable to have relatively simple methods of defining the relevant market that do not require the econometric estimation of an entire demand system. Fortunately, we have learned a great deal over the past twenty years about how to exploit the information contained in pre-merger prices, costs, and diversion ratios to perform the HMT without full estimation of the demand system.133
By focusing on how the pricing incentives facing the hypothetical monopolist differ from the pricing incentives of firms independently owning and controlling the relevant products prior to the merger, the HMT can be grounded in reality. Focusing on the change in incentives is a major and very sensible and practical simplification. The Clayton Act standard—whether the merger may substantially lessen competition—is explicitly focused on the change resulting from the merger. The unifying theme of the Guidelines since 1982 has also been about the change: whether the merger will enhance market power. And the HMT itself asks about whether the hypothetical monopolist will raise prices by at least a SSNIP, which again looks at a change from pre-merger conditions.
The hypothetical monopolist’s pricing incentives differ from those of the pre-merger firms because the hypothetical monopolist owns a larger group of substitute products. The hypothetical monopolist does not lose sales when the price of one product is elevated and customers shift away from that product to other products it owns. Therefore, in considering how the hypothetical monopolist’s incentive to raise the price of one product differs from the pre-merger incentives of the firm controlling that product, a key question is what percentage of the unit sales lost, when that product’s price rises, are recaptured by other products controlled by the hypothetical monopolist. This percentage is defined in Section 4.1.3 of the Guidelines as the recapture percentage, “with a higher recapture percentage making a price increase more profitable for the hypothetical monopolist.” In some cases, the Agencies can glean information about the recapture percentage even if they lack sufficient data to estimate the entire demand system. For example, if the price of one product was raised in the past (or if supplies of that product were disrupted or limited), one may be able to track how customers of that product shifted to other products. The recapture percentage is closely related to the cross-elasticity of demand that has been central to market definition for decades.134
The hypothetical monopolist’s incentive to raise the price on any one product under its control depends on the recapture percentage associated with that product and on the margins it receives on the sales recaptured by the other products it owns.135 Kevin Murphy and Bob Topel put it this way: “A larger fraction of sales diverted to other firms in the market or a larger profit margin on these sales will make the incentive to increase price greater for the hypothetical monopolist.”136 Applying this fundamental economic logic, the Guidelines state: “The higher the pre-merger margin, the smaller the recapture percentage necessary for the candidate market to satisfy the hypothetical monopolist test.”137 This is the same basic economic logic we saw above in the evaluation of unilateral price effects.
With linear demand, if each firm selling one of a symmetric group of differentiated products is setting its pre-merger price independently, that group of products forms a relevant market if the recapture percentage for any one product is at least as large as 2S/ (M + 2S ), where S is the size of the SSNIP and M is the pre-merger margin.138 In this special case, Appendix A shows that a symmetric pair of products satisfies the HMT if GUPPI is at least 10 percent. This highlights the tight connection between unilateral effects and market definition.
Merging parties sometimes conduct a “critical loss analysis,” typically to support their claim that a certain candidate market in which they have large shares is too narrow to satisfy the HMT. Critical loss analysis relies heavily on price/cost margins. The Guidelines now explain how the Agencies evaluate and properly conduct critical loss analysis. Since critical loss analyses have long been presented to the Agencies by merging parties, this explanation is overdue.
Most critical loss analyses presented to the Agencies use the “breakeven” approach.139 The Guidelines note that “this ‘breakeven’ analysis differs from the profit-maximizing analysis called for by the hypothetical monopolist test” since 1984.140 Breakeven analysis compares the “critical loss” with the “predicted loss.” The Agencies and others have been aware for some time of a fundamental flaw appearing in a number of breakeven critical loss analyses they receive. The flaw arises when the predicted loss is not reconciled with the pre-merger margins. Michael Katz, writing when he was Economics Deputy at the DOJ in 2002, described this flaw in some detail in his discussion of the Sungard case.141 FTC economists were equally aware of the flaw; additional cases are described by Daniel O’Brien and Abraham Wickelgren.142 The UK Competition Commission is also aware of this flaw:
The same flaw appeared more recently in the Whole Foods case.144 The revised Guidelines alert practitioners to this flaw and explain how the Agencies evaluate breakeven critical loss analysis: “Higher pre-merger margins thus indicate a smaller predicted loss as well as a smaller critical loss.”145 See Appendix A for further details.
The revised Guidelines add a separate section on targeted customers and price discrimination. This section sets forth the two basic conditions necessary for price discrimination to be feasible: differential pricing and limited arbitrage.146 The basic principles explained here have been well understood by economists for roughly one hundred years. They can be found in the Guidelines going back to 1982 and are not controversial.
Price discrimination is frequently an important factor in DOJ merger investigations. The majority of our mergers involve intermediate goods and services. In these markets, prices typically are negotiated and price discrimination is common. For example, manufacturers may negotiate lower prices with larger customers than with smaller customers, and these price differences may constitute price discrimination, i.e., they may not merely reflect lower costs of supplying the larger customers. In other settings, established customers with high costs of switching away from their incumbent supplier may pay higher prices than new customers. In yet other settings, prices vary across customers based on their locations in a manner not merely reflecting transportation costs. This is relevant for geographic markets based on customer location.147
This new section was placed relatively early in the Guidelines because the basic principles of price discrimination articulated here are used throughout the Guidelines. They are relevant to market definition. For that purpose, we usually are asking whether the hypothetical monopolist can engage in price discrimination. They are also relevant to competitive effects. When considering unilateral effects, we often ask whether the merged firm can engage in price discrimination. In some cases, we ask whether the merged firm can raise prices to certain customers by ending discrimination that had been in their favor. When considering coordinated effects, we may ask whether a group of coordinating firms could engage in price discrimination.
In fact, DOJ investigations often begin by asking whether there are particular types of customers who are most likely to be harmed by the merger. We often find that some types of customers are more vulnerable than others to adverse competitive effects. We look for pre-existing price discrimination and we consider the possibility of post-merger price discrimination.
The Guidelines address the danger that mergers may harm some customers more than others, or some customers but not others, usually by making a discriminatory price increase profitable. But this observation should not be taken to imply any hostility to price discrimination as a stand-alone form of business conduct.148 For many years, economists have studied the effects of price discrimination, usually by comparing price discrimination with uniform pricing. These studies are directly relevant to the evaluation of regulations that limit or prohibit price discrimination.149 But the comparison of uniform pricing and price discrimination is not directly relevant for the analysis of horizontal mergers, and the Guidelines do not undertake any such comparison. Nor do the Guidelines address the issue of whether or when price discrimination by a firm indicates that the firm has significant market power under the antitrust laws. The Guidelines are focused on whether the merger is likely to enhance market power. Price discrimination is highly relevant to this question if the merger may enhance market power over some customers but not others.
The revised Guidelines add a discussion of “Powerful Buyers” in Section 8. In this respect, they follow the lead of the European Commission’s 2004 Horizontal Merger Guidelines, which include a discussion of “Countervailing Buyer Power.”150
Many DOJ merger investigations involve intermediate goods markets, where the customers of the merging firms are themselves sizeable enterprises. Merging parties often argue that their customers are large and powerful and will not be vulnerable to adverse competitive effects. This section explains how the Agencies evaluate “power buyer” arguments and how merger analysis is influenced by the presence of large or powerful buyers.
Three basic economic themes underlie this section. First, whatever leverage buyers have in negotiations must ultimately rest on the alternatives available to them. In some cases, larger buyers are better placed than small buyers to vertically integrate upstream or sponsor entry, or to shift a greater portion of their business to price cutters. Options such as these can give larger buyers additional bargaining leverage.151 In contrast, mere size alone, without options, does not normally create bargaining leverage, although it can imply large gains from trade.
Second, the Agencies are interested in the impact of the merger on all buyers, not just powerful buyers. The Guidelines state: “Furthermore, even if some powerful buyers could protect themselves, the Agencies also consider whether market power can be exercised against other buyers.”152 The 2006 Commentary sounded a similar message:
In some cases, the actions of powerful buyers can protect more vulnerable customers, e.g., when the lumpy sales of the large buyers disrupt coordination and engender price wars. However, this is not always the case, particularly when the concerns involve unilateral effects. If powerful buyers are protected and other buyers are not, there may be a price discrimination market in which those other buyers are the targeted customers.
Third, the Agencies focus on how the merger will change bargaining leverage. “The Agencies examine the choices available to powerful buyers and how those choices likely would change due to the merger. Normally, a merger that eliminates a supplier whose presence contributed significantly to a buyer’s negotiating leverage will harm that buyer.”154
These principles can be applied to situations in which a large buyer purchases other products from the merging firms in addition to the products over which they compete. Merging parties sometimes assert that the merged firm would be foolish to try to raise price to such a powerful buyer, because that buyer would retaliate by shifting its purchases of these other products away from the merged firm. While buyers of this type do have an extra tool at their disposal, and may indeed be able as a consequence to negotiate lower prices than other buyers, such buyers will normally still be harmed if the merger eliminates a supplier whose presence contributed significantly to their negotiating leverage.
The 2010 Guidelines provide updated and more accurate guidance regarding merger enforcement at the DOJ and the FTC than did the 1992 Guidelines, which they replace.
UNILATERAL PRICE EFFECTS: THE ROLE OF DIVERSION RATIOS AND MARGINS
The Guidelines identify diversion ratios, margins, and the value of diverted sales as objects that the Agencies seek to measure to diagnose unilateral price effects in markets with differentiated products. The Guidelines also, for the first time, list merger simulation as one of the tools used by the Agencies to “quantify the unilateral price effect resulting from a merger.”
Merger simulation, in its full-blown form, involves estimating the demand system for a set of differentiated products, backing out or directly measuring marginal costs, computing the post-merger equilibrium, and then comparing pre-merger and post-merger prices. In principle, this is the “gold standard,” since it involves predicting post-merger price increases based on detailed demand and cost information, under some maintained assumption about oligopolistic behavior, usually independent (Bertrand) pricing behavior. However, the data required for full merger simulation are often not available, the predictions of merger simulation models may not be robust, and merger simulation techniques can be opaque to non-specialists. Therefore, less demanding and less sophisticated methods are often needed.
One way to achieve substantial simplification and increased transparency is to focus just on the demand for the products sold by the merging firms, holding fixed the prices of competing products sold by non-merging parties. As noted in the text, doing so will normally generate smaller price effects than the full model; but the simplification is considerable and the price effects coming out of the full model often differ very little from those of the simplified model.
With this major simplification, we rephrase the key question posed in the text: “Taking as given all other products and their prices, how much higher are the merged firm’s profit-maximizing prices for Products 1 and 2 than the pre-merger prices of those products?” If we can answer this question, we can derive a useful diagnostic measure of tendency of the merger to raise the price of these products. Technically, this diagnostic consists of the post-merger price increases for Products 1 and 2, holding other prices constant, and assuming that there is no repositioning or entry and no efficiencies. This diagnostic measure is not a “prediction” of the post-merger price increases. The diagnostic measure provides a relatively simple way of ranking or scoring mergers by their tendency to raise price.155 Predicting post-merger price increases requires further analysis.
One good diagnostic measure relies on the fact that one can treat GUPPI1 as a post-merger opportunity cost for Product 1, and then apply a default pass-through rate to those costs, holding fixed the price of Product 2.156 Basing the default pass-through rate on linear demand gives a pass-through rate of 50 percent; this figure is within the general range of pass-through rates that are estimated empirically. With a default pass-through rate of 50 percent, the indicated price increase, measured as a fraction of the price of Product 1, is
With equal, prices this becomes
Using this method, a 10 percent value of GUPPI1 translates into an indicated price increase of 5 percent.
This approach has been criticized for holding fixed the price of Product 2 when calculating the indicated price increase for Product 1.157 Instead, one can specify the demand system at prices just above pre-merger levels and calculate the indicated post-merger prices for that demand system. A larger indicated price increase, with a somewhat different ranking, is obtained by simultaneously considering a price increase for Product 2. Yet again, Willig led the way, working with a linear demand system for Products 1 and 2. He writes: