Note: This speech has been revised and is published in the Antitrust Bulletin, vol. 42, Spring 1997, pp. 177-196.
After a century of interpreting the Sherman Act, antitrust has learned how to do some things quite well. We understand what is wrong with naked horizontal price fixing and market division, and we have developed strong tools to deal with those problems. We can generally prevent mergers to monopoly and we are pretty good at preventing other mergers that increase market power. At least, we can get these results in the kinds of industries that antitrust grew up with, in which firms make and sell substantially similar products and compete in terms of current production and prices, rather than innovation.
But I am not here to talk about the successes. I want to explore the problems antitrust has had and will have even more in the future when products are differentiated. I will focus on two types of differentiation in particular. The first appears most commonly in consumer products industries, in which a large number of brands are available and the products vary in physical attributes and images. Think of products like soft drinks and automo- biles as examples. Here the differentiation is spatial, to use the economist s metaphor of products located at various positions in an abstract characteristics space. The second type of differentiation emphasizes that firms in these industries often intro- duce new products over time. There was a time, after all, when Coke and Pepsi did not sell sugar-free or caffeine-free colas, and when General Motors and Ford did not sell subcompact cars. I have in mind product innovation when I speak of goods differentiated over time. It is particularly appropriate that I address intertemporal differentiation because the Commission s recent Hearings on Global and Innovation-Based Competition focused heavily on the antitrust challenges connected with innovation.
In the differentiated product settings, antitrust has had trouble isolating anticompetitive harm and devising a procompetitive remedy within the conventional rule-of-reason, market-defin- ition paradigm. I will argue that the problems appear in legal doctrine, but their roots, and their possible resolutions, rest in economic analysis. As usual, my remarks today reflect my own views,and are not necessarily those of the Commissionor of any Commissioner.
Economics of product differentiation
Product differentiation across space is of particular interest in merger analysis. The 1992 Merger Guidelines contain a section devoted to analyzing the unilateral competitive effects of mergers when products are differentiated.2 Carl Shapiro, my Justice Department counterpart, has recently spoken on this topic, and my remarks are both complimentary of, and complementary to, his analysis.3
Although spatial product differentiation appears in multiple guises, I will focus primarily on the most obvious situation, in which producers respond to different consumer tastes by offering different sets of product attributes or projecting different product images and personalities.4 For example, the Commission has over the years examined mergers in the soft drink industry. Soft drinks are characterized by variety in flavor, packaging, and image that together produce distinct brand identities.5
Differentiation itself is not unambiguously good or bad. Buyers typically benefit from the availability of a wide variety of product offerings to serve their differing preferences. Yet differentiation can also facilitate the exercise of market power. The producer of a differentiated product often enjoys a localized monopoly and may be able to charge a higher price than it otherwise could.
The investments firms make in differentiation, moreover, can change market structure in ways that might, to first appearances, lead us to suspect that markets are not competitive. When the firms in an industry make large investments in brand reputations and their new product pipeline, we often observe that the number of firms is low and that prices are well above seller marginal costs.6 Also, firms may find it possible, and profitable, to discriminate in price or to sell different products to different segments of the market.7
Yet it is too quick to conclude that such an industry necessarily presents an antitrust problem. These features of industry structure high concentration, high price-cost margins, and price discrimination are also consistent with an industry setting we should reasonably call competitive: one characterized both by free entry and significant accommodation of entry by incumbents. Under such circumstances, the high price-cost margins merely cover the high fixed costs,8 and the entry constraint limits the ability of firms to use price discrimination to exploit identifiable groups of customers with inelastic demand by the constraint.9 But if incumbent firms are able to commit to aggressive post-entry competition, rather than accommodating entry, they can exploit such market structures to achieve prices above what competition would permit.10 Investments in product differentiation could be one way of doing so.11
As this recital suggests, ease of entry plays a critical role in preserving competition in many differentiated-product markets. Interfirm rivalry is often a less significant competitive force than in homogeneous product industries because compe- tition is localized.12 I will return to this point when I discuss the antitrust issues that arise from the introduction of new products.
Mergers where spatial differentiation is extreme
Extreme product differentiation the situation that arises when firms sell a wide range of styles, flavors, colors and similar characteristics, so there are no "gaps" in the chain of substitutes frequently befuddles the antitrust analysis of mergers.13 The law is often preoccupied with defining a single market so that proxies for market power, usually market share, can decide the ultimate questions. The output of this method, the impossibly gerrymandered left-handed one-eyed man with a limp product market, often appears in after-dinner speeches as comic relief. And the resulting market shares often mislead about the potential for the merger to harm competition in the localized regions in which differentiated product sellers may be able to raise price. Where products are highly differentiated, the competition policy concern is more likely to be about single-firm effects than about inter-firm coordination. The gains from coordinating an increase in the prices of distant substitutes are likely to be low, compared to the gains from manipulating competition among the closest substitutes. Thus the concern is less about what might happen in the outer reaches of a market, than about what might happen among the merging firms own products.
If the usual market-definition based methods are unreliable in analyzing mergers among sellers of differentiated products, what other tools can be used? How are we to identify and challenge those mergers whose primary effect is to create or enhance localized monopolies rather than to increase consumer choice or reduce production costs? Let me give an example about the problems of relying on market definition and market shares to identi- fy market power among sellers of differentiated products based on my own research on the brewing industry during the 1970s.14 Despite having similar shares of the beer market, Coors had a valuable localized monopoly but Pabst did not. Moreover, the market share leader, Anheuser-Busch (the seller of Budweiser), had little market power after 1975. Combinations among these brewers could have made a Budweiser price increase profitable, because a combination with either brand would have removed a constraint on Anheuser-Busch s ability to raise price. But such mergers would not have led to an increase in the price of Pabst s or Coors s flagship brands, because their positions were different: the Coors brand was isolated, with no other brand constrain- ing it, while the Pabst brand was constrained by many others. Differentiation in brewing is not as extreme as it is in other industries; it is likely that even more striking distinctions could appear for other products.
When it comes to identifying market power in differentiated product industries, as this example suggests, it may not be helpful to commence the antitrust analysis by asking what is the relevant market? and then computing market shares. Rather, in examining the competitive effects of a transaction such as a merger, the first question should be, if competition between products of the merged firm declines and one product s price rises, where do buyers go?15 This question is critical because it helps identify situations in which localized competition among the products of the merging firms keeps prices low and thus identify the products for which the firms will have a strong incentive to raise price following the acquisition. That is, if a firm selling product A acquires product B, its incentive to raise the price of A will be greater, the larger the number of buyers of A who would react to a price increase by shifting to B.16 The concern in these combinations of differentiated products is what happens to prices if firms internalize what was previously important localized competition.
I recognize that consumer substitution patterns in response to a price increase are not the only factors important to the analysis of unilateral competitive effects. The merged firm s incentive to raise the price of product A is also greater, for example, the higher the price-cost margin on product B relative to product A, and the less rival firms selling other potential substitutes respond by competing more aggressively on price or modifying product characteristics to become even closer substitutes. And economists have developed a number of good techniques for integrating these factors,17 all of which have been or would be used by the FTC s Bureau of Economics in appropriate cases.18 My point here is simply that market definition is often not the best place to start in assessing the competitive consequences of mergers in industries characterized by extreme product differentiation.
Can legal doctrine follow these economic insights and move beyond the traditional reliance on market share proxies, to deal with the issue of market power more directly? The answer has been thought to depend on the statutory setting. The Clayton Act statutory language requires demonstration of competitive effects in a line of commerce and section of the country. Thus, in a case brought under that act, the plaintiff or the prosecutor must prove up something like a relevant market. Even so, the Clayton Act language does not demand the use of market-share proxies for measuring competitive effects. A merger can also be challenged under the Sherman Act, in which case the rule of reason would apply, calling for balancing competitive harms and benefits. But applying the rule of reason does not necessarily entail definition of a market and proof of market shares as the way to do that although often that is just what happens.
The possibility of observing and measuring market power more directly leads me to suggest a new notion for Clayton Act doctrine, something I think of as the res ipsa loquitur market definition. When a piano crashes onto the sidewalk, the law does not ask whether someone was negligent; instead, it goes right to the question of who. This approach could translate to antitrust. Suppose we know, directly, that a merger or other practice is harmful. That is, we can observe, or confidently predict, an increase in price or the exclusion of efficient competition. But suppose also that it is hard to draw lines around a market, because the array of differentiated products is broad and seamless. If we can show the harm, there must be a market in there somewhere. Just exactly where the market s boundaries are may not be very important, though. Nor may it matter much whether the market in which the harm occurs is large or small. All that should matter to the doctrine is that the market contain the transactions or parties that are causing or suffering the consumer injury.
There is ample legal authority for dispensing with the market definition exercise in rule-of-reason cases.19 If the harm is shown directly, there is no point in requiring that it also be shown by proxy.20 Proof of power to maintain price above competitive levels or to exclude competition the legal formulations that describe market or monopoly power in Sherman Act cases should establish the element of monopoly power directly. It would certainly suffice in a non-merger case brought under the Sherman Act or the Federal Trade Commission Act. It ought to be possible to devise some formal adaptation so these same principles would be available in merger cases brought under Section 7 of the Clayton Act especially given that the Section 7, Section 1 and Section 5 analyses of mergers are widely thought to have converged.21
Establishing the concept of a res ipsa loquitur market will undoubtedly need to overcome some forensic hurdles.22 A likely complaint will be that the res ipsa market comes perilously close to the kind of ad hockery that gave submarkets a bad name. It need not, if the idea is applied responsibly. After all, the point of the res ipsa approach is really to bypass formal market definition. What would end up identified as the focus of competition concern might not, by conventional standards, be called the market, but would be given that label as a kind of legal fiction so we could get on with the more important task of applying economic analysis to identify competition problems.
I can make the argument more concrete. To return to my brewing example, if we somehow knew for example based in part on the kind of econometric evidence previously cited that the acquisition of Pabst by Anheuser-Busch would give Anheuser-Busch the incentive and ability to raise the price of its flagship brand, Budweiser, why should it still be necessary to prove whether light beer, foreign imports, beer from microbreweries, non-alcoholic beer, wine, or soft drinks are in the relevant market in order to prevail in a challenge? It is not sufficient to answer that under such circumstances the relevant market, properly defined under the Merger Guidelines would include only Budweiser and Pabst Blue Ribbon, making this transaction challengeable as a merger to monopoly, because it is difficult to imagine a judge, employing the received approach of defining the market before reviewing the competitive effects evidence and fearful of criticism for gerrymandering, defining a market that includes Budweiser but not Miller.23
The legal problem, like the economic problem, is not one of conception but of proof. Within a market, differentiation can lead to localized competition problems that depend little on the precise location of the market s boundaries. Ideally, those problems would be demonstrated by quantitative economic evidence about the extent of market power exercised. But if the data needed for this kind of analysis are hard to come by, proof of particular, localized effects may rely on other evidence. This other evidence may include the parties admissions in their planning and marketing documents identifying their principal competitors or claiming credit for their plans anticipated exclusionary or price elevation effects. It may also include the testimony of customers or rivals. This is the same kind of evidence that is typically adduced in performing the standard exercise of defining relevant markets. That is no accident, for the kind of analysis I am proposing would be concerned about identifying the products that constrain the exercise of market power by the merging firms, just as the familiar market definition exercise is concerned. In an industry with extreme product differentiation, though, the output of the standard market definition exercise can be both unnecessary and distracting.
Let me anticipate and address some possible concerns about the res ipsa approach to market definition in the kind of markets where it would work best those with few gaps in the chain of substitutes, so that any market definition would appear arbitrary. First, this approach does not turn the unilateral theory of anticompetitive effects for differentiated products, set forth in the 1992 Merger Guidelines, into a per se theory of liability. On weighing the evidence, we might discover that significant competition is not internalized by the merger of the seller of brand A and the seller of brand B much as the econometric evidence discussed above suggested that a Coors-Pabst merger would not have harmed competition during the 1970s. Moreover, the merger will not have anticompetitive effects if it induces timely, likely and sufficient repositioning of rivals products or entry. The brewing industry demonstrates that repositioning and entry possibilities cannot be ignored, even when they would require substantial sunk investments in brand reputation: Miller s marketing investments in the early 1970s, repositioning its flagship brand through the Miller Time advertising campaign and introducing Miller Lite, appear to have competed away Budweiser's previous exercise of market power.24 Also, efficiencies matter. Firms may demonstrate fixed and variable cost savings from the transaction. In addition, the competitive effects analysis should recognize that, if the merger lowers the marginal costs of producing and selling brands A or B, prices could fall on net even if some localized competition is removed.
Second, I recognize that the enforcement agencies must engage in market definition, as the basis for computing market shares and HHIs, in order to determine whether a transaction falls into the Merger Guidelines safe harbors.25 I do note, however, that for mergers among sellers of differentiated products, market share measures may not achieve two important purpos- es of a safe harbor.26 One purpose is to remove from review transactions where the potential for the exercise of market power is limited. Yet market shares are often poor predictors of market power in settings where product differentiation is extreme, and thus may not reliably serve that purpose. Another purpose is to implement a presumption about efficiencies, and here again market share measures may serve the purpose poorly. High market shares for brands of differentiated products may well be correlated with better products or lower costs. But brand capital is often less transferable to other uses than physical capital is, so there may be less scope for efficiencies in combinations among branded products than in those among homogeneous goods. Until we develop better ways to implement the reasonable presumption that mergers generate some efficiencies in the differentiated products case, however, the enforcement agencies will continue to rely upon the safe harbors built into the Merger Guidelines to inform their exercise of prosecutorial discretion.
Antitrust s confusion about extremely differentiated products in merger analysis comes because market definition can be difficult and uninformative in such industries. That confusion can be reduced by the use of economic tools to identify harm directly, rather than relying on the proxy of market share. That greater sensitivity will help us balance better the unquestioned but hard-to-measure value of variety and choice against what we are willing to pay for that variety in localized market power.
Can investment in new products harm competition?
Buyers can benefit from the new products that result from innovation,27 an activity that can be thought of as product differentiation over time. And while the results of research and development may be unpredictable, firms can influence the types of new products they produce and the probability of R&D success through the extent and direction of their R&D and new product marketing expenditures.
We are all well aware how seller investments in developing and marketing new products benefit buyers. But when it is an already leading firm that engages in aggressive innovation, in a market structure that is already highly concentrated, antitrust sensitivities can be aroused. Does the R&D and marketing investment represent competition on the merits, the creation of new and better products that benefit consumers? Or does it merely amount to a strategy of discomfiting and discouraging competitive threats from fringe firm innovations, without benefiting consumers?
An illustration familiar to antitrust practitioners is presented by the Kodak litigation of the 1970s.28 That case explored what obligations an innovating monopolist might have toward its fringe firm competitors. Other antitrust cases that have wrestled with leading firm innovation include the FTC s Xerox case in the mid-1970 s,29 which ordered compulsory licensing of xerography technology, and more recently, of course, the Justice Department s Microsoft case,30 which imposes some limits on the kinds of contracts Microsoft can enter. The succession of technologies and products in these industries photography, xerography, and software demonstrates a now seemingly familiar pattern. The leading firm introduces successive waves of new products or technological systems.31 These products may be improvements on previous ones or replacements for them. They may be incompatible with the previous versions or with those of the fringe firms, thus making old or competing products obsolete. Fringe firms are reduced to reacting to the leading firm s strategy.32
We must proceed with caution in applying the antitrust laws in this area. On the one hand, leading firms aggressively innovating can, in principle, harm competition. Because ease of entry plays a critical role in preserving competition in many differentiated-product markets, it is appropriate to worry that research and development investments could be an instrument of strategic entry deterrence. On the other hand, we must be wary of drawing that conclusion, because investments in innovation often have payoffs in better products, lower costs, or higher quality. No one wants to kill the goose that may be laying golden eggs even if we fear the goose may be eating its rivals young. We do not want a system of antitrust enforcement that reduces ex ante incentives to innovate below the efficient level.
Our fundamental task here is finding ways to challenge those R&D and marketing strategies involving new products that deter post-innovation competition by rivals without producing benefits to buyers, and it is a tough assignment.33 One reason it is difficult is that firms will often not know how their R&D efforts will work out. Thus, both the buyer benefits and the rival deterrent effects of R&D must be assessed from an ex ante perspective, focusing on what would be reasonable to expect at the time of innovation rather than by asking whether the R&D actually benefitted buyers or harmed competitors. Antitrust s likely approach to this task was suggested by the Berkey panel, which explained that Kodak s introduction of a new film format and simultaneous withdrawal of an old format, forcing photographers to buy photofinishing from Kodak rather than Berkey, could have supplied the bad act necessary to support a charge of monopolization if the new format not been better or cheaper than the old.34
Despite the difficulty of this task, antitrust has, on the whole, responsibly undertaken the assignment of distinguishing procompetitive innovation from monopolizing exclusion. It would be difficult to sustain the charge that antitrust routinely penalizes success. For example, the government ultimately withdrew its monopolization case against IBM, and virtually all of Berkey s private monopolization verdict against Kodak was reversed on appeal.35
Similarly, the antitrust standards applied by the courts tend to resolve unclear cases in favor of the innovative leading firms. The courts have admittedly signaled a willingness to target spurious R&D or new product marketing, that is, research and development and marketing expenditures that appear purely, or even largely, aimed at preempting rivals, without giving consumers much benefit.36 For example, this principle is consistent with the Berkey decision s treatment of dominant firm innovation.37 But courts hesitate to apply this principle when defen- dants proffer a legitimate business justification. In particular, courts are reluctant to find that a firm misused its monopoly power through the introduction of any innovation that lowers cost, improves quality or performance, or is otherwise desirable to consumers, even if the innovation creates incompatibilities or otherwise raises costs to rivals.38 And while the essential facilities and monopoly leveraging doctrines could help plaintiffs challenging leading firm practices, the courts may now be limiting their application.39
These doctrinal developments will likely channel some future litigation over the alleged exclusionary practices of innovative leading firms into disputes about alternatives. As the late Professor Areeda put it, an anticompetitive restraint can be redeemed only if it is reasonably necessary to a legitimate objective, and to be reasonably necessary, the restraint must promote that objective significantly better than the available less restrictive alternatives. 40 Areeda goes on to ask, Can that objective be achieved as well without restraining competition so much? 41 That question could appear directly if the litigation arises under Sherman Act Section 1 and the rule of 41 Id. reason applies,42 or indirectly through the determination of whether a monopolist has engaged in a bad act sufficient to support a Sherman Act Section 2 claim. Less restrictive alternative analysis is neither a panacea for plaintiffs nor a shield for defendants; we must ask whether the alternatives were reasonably available at the time the restraint was imposed, from an ex ante perspective.43
In sum, firm investments in the development and marketing of new products differentiation over time can harm competition notwithstanding the fact that new products are introduced. The more the investments appear aimed at creating aggressive post- entry competition rather than providing buyers with valuable new consumption alternatives, the easier it is to conclude that they are methods of weakening or removing the constraint on pricing imposed by the threat of entry. But the stakes are high in distinguishing such competitively malignant investments from welfare-enhancing R&D and new product marketing expenditures, because the costs of error can be large when we are dealing with innovation.44 In consequence, we must proceed with caution and care in applying the antitrust laws in this area, especially when the R&D or new product marketing investments at issue produce demonstrable benefits to consumers.
Product differentiation over space and across time presents special challenges for antitrust law. These cases involve difficult analytic tasks, novel fact patterns, and high stakes. These are the sort of enforcement actions well-suited for investigation and review by an expert deliberative body like the Federal Trade Commission. And I believe the Commission and its staff would be ready to take on such challenges.
1 This text expands upon the remarks delivered to the Antitrust and Trade Regulation Committee of the Association of the Bar of the City of New York, February 6, 1996. The author is grateful to Stephen Calkins, Steven Salop, Gary Roberts, and Michael Wise for comments and assistance.
2 U.S. Dept. of Justice and Federal Trade Commission, 1992 Horizontal Merger Guidelines, 2.21, 4 Trade Reg. Rep. (CCH) 13,104.
3 Carl Shapiro, Mergers with Differentiated Products (speech to the American Bar Association & International Bar Association program, The Merger Review Process in the U.S. and Abroad, Washington, D.C., November 9, 1995).
4 Differentiation can be thought of as spatial in at least two other settings. First, buyers may care about seller reputation, so that the plain fact of a seller s identity could differentiate its products from those of other sellers. This can appear in the form of buyers requiring sellers to become qualified to sell to them. Second, shopping or transportation cost may matter, so that differentiation appears on the basis of seller or producer location.
5 See, e.g., FTC v. Coca-Cola Co., 641 F. Supp. 1128 (D.D.C. 1986) (preliminary injunction), vacated mem. 829 F.2d 191 (D.C. Cir. 1987).
6 JOHN SUTTON, SUNK COSTS AND MARKET STRUCTURE: PRICE COMPETITION, ADVERTISING AND THE EVOLUTION OF CONCENTRATION (1991); see Timothy F. Bresnahan, Sutton s Sunk Costs and Market Structure: Price Competition, Advertising, and the Evolution of Concentration, 23 RAND J. ECON. 137 (1992).
7 Economic price discrimination, that is different price-cost margins on sales of the same goods to different buyers, is not necessarily the same as discrimination under the Robinson-Patman Act.
8 An incumbent s average prices will not exceed a new entrant s marginal cost by enough to induce entry. Note that what have been fixed costs for incumbents are variable (marginal) costs for the prospective entrant. Once the new firm enters, it expects to earn a profit contribution (revenues less short-run post-entry variable costs) that covers its fixed costs.
9 The literature on contestable markets emphasizes that high concentration is not necessarily inconsistent with competitive outcomes in free-entry settings. See WILLIAM J. BAUMOL ET AL., CONTESTABLE MARKETS AND THE THEORY OF INDUSTRY STRUCTURE (1982). Free-entry markets may also exhibit price discrimination. See Severin Borenstein, Price Discrimination in Free-Entry Markets, 16 RAND J. ECON. 380 (1985); T. Holmes, The Effects of Third-Degree Price Discrimination in Oligopoly, 79 AM. ECON. REV. 9 (1989). With free entry, an entrant expecting to mimic the price discrimination of incumbents would most likely generate the greatest contribution margins (price less marginal cost, relative to price) from those identifiable groups of buyers3 with the most inelastic demand. The aggregate contribution (price less marginal cost, summed over all groups of buyers) would be capped by the constraint that average revenue not exceed entrant's marginal cost by enough to induce entry.
10 See generally Richard J. Gilbert, Mobility Barriers and the Value of Incumbency, in 1 HANDBOOK OF INDUSTRIAL ORGANIZATION 475 (R. Schmalensee & R. Willig, eds. 1989); Steven C. Salop, Strategic Entry Deterrence, 69 AM. ECON. REV. 335 (Papers and Proceedings 1979). The greater the sunk investments required for entry, and the greater the incumbents commitment to a rapid and aggressive response, the more entry is deterred by the threat of post-entry competition.
11 Gilbert, supra note 10, at 503-06; Richard Schmalensee, Entry Deterrence in the Ready-to-Eat Breakfast Cereal Industry, 9 BELL J. ECON. 305 (1978). Sutton emphasizes that firms have an incentive to make fixed investments in product differentiation or R&D merely to exploit the integer effect the increment to prices that arises on average even with free entry when a market that previously could have held, for example, 5 1/2 firms (of at least minimum viable scale, given the fixed costs of entry) turns into one that fits only, say, 3 1/2 firms. See SUTTON, FIXED COSTS, supra note 6; cf. N. Gregory Mankiw & Michael D. Whinston, Free Entry and Social Inefficiency, 17 RAND J. ECON. 48 (1986) (product variety combined with the integer effect will lead to less than the socially optimal number of firms, unless compensated for by sufficient incumbent accommodation of entry). Strategic investments that rely merely on the increased price-depressing effect of a tighter fit to engineer higher prices are probably less troublesome competitively, on average, than strategic investments that achieve higher prices by leading incumbent firms to respond more rapidly or aggressively to entry (and thus change the nature of the post-entry oligopoly interaction).
12 Frequently, incumbent firms find that price-cutting does not allow them to increase sales long enough to compensate for the lost profit margin on their current output. (If so, a price above incumbent marginal cost may be 4sustainable.) Price-cutting is typically a less profitable strategy the more products are differentiated and the more rapidly rivals can respond by matching the price cut.
13 Markets with firms differentiated by location can also be highly differentiated, raising similar difficulties for merger analysis. Examples may include supermarkets, hospitals and drug stores.
14 Jonathan B. Baker and Timothy F. Bresnahan, The Gains for Merger or Collusion in Product-Differentiated Industries, 33 J. IND. ECON. 427 (1985); Jonathan B. Baker and Timothy F. Bresnahan, Estimating the Residual Demand Curve Facing a Single Firm, 6 INT L J. IND. ORG. 283 (1988) see also Jonathan Baker and Timothy F. Bresnahan, Empirical Methods of Identifying and Measuring5 Market Power, 61 ANTITRUST L.J. 3 (1992).
15 Shapiro asks a similar question in seeking to estimate what he calls the diversion ratio, that is, the fraction of the sales lost by the product whose price is increased that would be captured by the other party s product. Shapiro, supra note 3.
16 Of course, if producers of A and B combined, it would also be important to find out how buyers of B would respond if its price increased. Perhaps the merged firm would have an incentive to raise B s price too, and perhaps not.
17 In principle, if we knew all the own- and cross-elasticities of demand and supply (and how those elasticities changed with output), all the entry possibilities, and the nature of oligopoly interaction in the industry, we could compute post-merger price changes directly. But we never have all this information. The different methods for identifying when a firm might raise price unilaterally post-merger emphasize different approaches to the absence of information. Broadly speaking, estimation methods try to identify what can be determined about buyer substitution patterns from the available 6historical data on prices and outputs, despite the limitations of those data. Baker & Bresnahan, The Gains for Merger or Collusion in Product-Differentiated Industries, supra note 14; Jerry Hausman et al., Competitive Analysis with Differenciated Products, 34 ANNALES D ECON. STAT. 159 (1994). The simulation methods tend instead to assume buyer substitution patterns and work out their implications. Shapiro, supra note 3; Gregory Werden and Luke Froeb, Simulation as an Alternative to Structural Merger Policy in Differentiated Products Industries, Economic Analysis Group Discussion Paper EAG 95-2 (September 1995); Robert Willig, Merger Analysis, Industrial Organization Theory, and Merger Guidelines, BROOKINGS PAPERS ON ECONOMIC ACTIVITY: MICROECONOMICS, 281, 299-305 (1991) (characterizing scenarios where market shares are indicators of the competitive effect of a merger between producers of differentiated products). Thus, simulation techniques are well-suited to the creation of benchmarks, to be applied where data, particularly for estimating demand cross-elasticities, are not available. Both approaches have the advantage of minimizing the significance of market definition: the estimation methods do not employ market shares and simulation methods are often not strongly sensitive to market definition even when they use shares. The techniques also differ in their ambition. The Baker & Bresnahan and Willig methods have the limited goal of identifying products for which the merged firm would have a strong incentive to raise price. In contrast, the Shapiro Froeb & Werden, and Hausman et al. approaches seek in addition to compute the post-merger price for all the affected products. That added task is generally highly sensitive to assumptions about the way the elasticity of demand varies as output changes a matter about which it is often difficult to gather information or engage in informed speculation.
18 Recent cases illustrate the use of economic evidence about the extent of competition between particular differentiated, branded products. These include the state of New York s unsuccessful challenge to the cereal industry merger and the Antitrust Division s successful challenge to some aspects of a recent bread baking merger. New York v. Kraft General Foods, Inc., Nabisco Cereal, Inc., et. al., Civil No. 93-0811 (KMW) (S.D.N.Y. February 22, 1995); U.S. v. Interstate Bakeries Corp. & Continental Baking Co., Civil Action No. 95C-4195 (N.D. Ill., complaint filed July 20, 1995). In these cases, high-quality data about consumer responses, taken from supermarket scanners, made detailed econometric analysis possible. Similarly, at the FTC, the staff has used detailed Nielsen scanner data to refine its analysis of likely competitive effects in mergers among soft drink producers. See Lawrence White7 & John Kwoka, THE ANTITRUST REVOLUTION at 85 n. 14 (1989).
19 See Richard Schmalensee, Another Look at Market Power, 95 HARV. L. REV. 1789 (1982); cf. Thomas G. Krattenmaker et al., Monopoly Power and Market Power in Antitrust Law, 76 GEORGETOWN L.J. 241 (1987) (suggests ways that the approach to market definition should vary with the competitive effect alleged).
20 FTC v. Indiana Fed n of Dentists, 476 U.S. 447, 460-61 (1986); NCAA8 v. Board of Regents, 468 U.S. 85, 109-110 (1984).
21 PHILLIP AREEDA & HERBERT HOVENKAMP, 2 ANTITRUST LAW 304, 397f (rev. ed. 1995).
22 Until experience has clarified how to apply this analytic approach, it would not be appropriate to employ it to criminal cases brought under the Sherman Act. Criminal cases are exceedingly rare in merger enforcement, however. 923 Also, more technically, if the Budweiser price would not quite rise five percent, the Guidelines would require broadening the product market in any case.
24 See Baker & Bresnahan, The Gains for Merger or Collusion in Product- Differentiated Industries, supra note 14.10
25 In addition to the familiar safe harbors involving the HHI, the unilateral competitive effects sections of the Guidelines also make reference to a safe harbor that applies when the merging firms in aggregate account for less than 35 percent of the market. As applied to firms distinguished primarily by differentiated products, however, the language employed in the Guidelines appears to apply the 35 percent safe harbor only when market shares are employed as a benchmark for measuring likely competitive effects ( 2.2), and not when other data is used to make that determination. Some agency officials have nevertheless described the 35 percent safe harbor as applying regardless of how the degree of competition among products is estimated. See, e.g., interview with Assistant Attorney General James Rill, 61 ANTITRUST L. J. 229, 238 (1992); Paul Denis, Practical Approaches: An Insider s Look at the New Horizontal Merger Guidelines, ANTITRUST, Summer 1992, at 9.
26 The problems noted below are less acute when products are relatively11 homogeneous.
27 Although this discussion focuses on new products, it is important to recognize that innovation may also benefit buyers by allowing firms to improve the quality of existing products or to lower their production costs.
28 Berkey Photo, Inc. v. Eastman Kodak Co., 603 F.2d 263 (2d Cir. 1979).
29 In re Xerox Corp., 86 F.T.C. 364 (1975).
30 U.S. v. Microsoft Corp., 1995-2 Trade Cas. (CCH) 71,096 (D.D.C. 1995) (consent decree). 12 31 For a discussion of how one leading firm manages its product portfolio to keep successive waves of technology coming to market, see MICHAEL A. CUSUMANO & RICHARD W. SELBY, MICROSOFT SECRETS (1995).
32 Yet leading firms need not be aggressive innovators. They may instead accommodate fringe innovations and investment. That is what the Big 3 automakers did in the 1960s and 1970s in response to product and process innovations from Europe and Japan. Jonathan B. Baker, Fringe Firms and Incentives to Innovate, 63 ANTITRUST L.J. 621 (1995). The reasons leading firms might accommodate rather than retaliate include fear that they might cannibalize their other product investments and a belief that the fringe firm threat will be confined and will not spread. See id., at 637-39. If leading firms are not aggressive, they are unlikely to get into antitrust trouble for suppressing innovation. Of course, they may still get into other kinds of antitrust trouble, for example, by price fixing.
33 Antitrust s difficulties in doing so are often exacerbated when the investment in research and development does not merely lead to new products, but also sets an industry standard. Then the locus of competition may become the opportunity to create the next standard as much as the products within a standard. Yet a standard-setter with intellectual property protection might deter competition or control access to the industry, thus inhibiting existing rivals employing the existing standard as well as potential rivals with a potentially better one. But standards often also amplify the consumer benefits from innovation, and intellectual property protection may be necessary to induce firms to invest in developing and promoting potential standards. And if, as often happens, adopting the standard facilitates the growth of network externalities, the stakes are raised even higher, for the scope of the potential harm to competition and the range of the potential efficiency13 benefits are likely both further increased.
34 Berkey v. Kodak, supra note 28, at 287 n.39. The court also concluded that Berkey might have recovered for Kodak s refusal to package its new film for camera formats other than the ones Kodak already made, if Berkey had shown it suffered damages as a result. Id., at 290. The judgment on that count was reversed because Berkey presented no evidence that it was injured in this way. The court also held that the leading firm s conspiracy with others to prevent disclosure of its innovation could be actionable, but as a violation of Section 1, not Section 2, of the Sherman Act. Id., at 302.
35 Admittedly, reaching these results can cost time and money.
36 The possibility of non-price predation through the development and marketing of new products is well established in the economics literature. Janusz A. Ordover & Robert D. Willig, An Economic Definition of Predatory Product Innovation, 91 YALE L.J. 8, 22-52 (1981); Joseph Farrell & Garth Saloner, Installed Base and Compatibility: Innovation, Pre-Announcements and Predation, 76 AM. ECON. REV. 940 (1986); Janusz A. Ordover & Garth Saloner, Predation, Monopolization, and Antitrust, 1 HANDBOOK OF INDUSTRIAL 14ORGANIZATION 537, 563 (Richard Schmalensee & Robert D. Willig, eds., 1989); see generally Thomas Krattenmaker & Steven Salop, Anticompetitive Exclusion: Raising Rivals Costs to Achieve Power over Price, 96 YALE L.J. 209 (1986).
37 See Berkey v. Kodak, supra note 28; In re IBM Peripheral EDP Devices Antitrust Litig., 481 F. Supp. 965, 1002-03 (N.D. Cal. 1979), aff d sub nom. Transamerica Computer Co. v. IBM, 698 F.2d 1377, 1383 (9th Cir.), cert. denied, 464 U.S. 955 (1983); Litton Sys. v. AT&T, 700 F.2d 785 (2d Cir. 1983), cert. denied, 464 U.S. 1073 (1984).
38 Berkey v. Kodak, supra note 28; Northeastern Tel. Co. v. AT&T, 651 F.2d 76, 93 (2d Cir. 1981); California Computer Prods., Inc. v. IBM, 613 F.2d 727, 744 (9th Cir. 1979).
39 Compare Phillip Areeda, Essential Facilities: An Epithet in Need of Limiting Principles, 58 ANTITRUST L.J. 841 (1990) (emphasizing potential breadth) with William Kovacic, Antitrust Analysis of Joint Ventures and Teaming Arrangements Involving Government Contractors, 58 ANTITRUST L.J. 1059, 1104 n.242 (1990) (recent narrowing). Compare Berkey Photo v. Kodak, supra note 24, at 276 (monopoly leveraging can be demonstrated if a monopolist merely obtains a competitive advantage in a second market, without attempting to monopolize it) with Alaska Airlines v. U.S., 948 F.2d 536 (9th Cir. 1991) (leveraging charge requires threatened or actual monopoly in second market) and Fineman v. Armstrong World Industries, 1992 Trade Cas. (CCH) 70,010 (3d Cir. 1992) (same).
40 Phillip E. Areeda, 7 ANTITRUST LAW 1505 (1986); cf. U.S. Department15 of Justice and Federal Trade Commission, Antitrust Guidelines for the Licensing of Intellectual Property, 4.2 (1995).
42 The alleged exclusionary practice could involve an agreement between the leading firm and a supplier, for example, thus permitting a Section 1 charge.
43 ERNEST GELLHORN & WILLIAM E. KOVACIC, ANTITRUST LAW & ECONOMICS IN A NUTSHELL 259 (1994).
44 Enforcement is even more difficult in an oligopoly setting, where none of the leading firms is a Section 2 monopolist and the firms may adopt entry- and innovation-deterring practices without entering into a demonstrable Sherman Act Section 1 agreement. This situation seems appropriate for the application of Section 5 of the Federal Trade Commission Act, either through enforcement actions or perhaps even through some form of competition rulemaking, to clarify what kinds of strategic investments in deterring 16competition are to be discouraged. Cf. Jonathan Baker, Two Sherman Act Section 1 Dilemmas: Parallel Pricing, the Oligopoly Problem, and Contemporary Economic Theory, 38 ANTITRUST BULL. 143 (1993).