Merger analysis and the treatment of uncertainty: Should we expect better?

Antitrust Law Journal

Volumn 74, Issue 3, 2007, Pages 537-574

  Katz, Michael L ^a,b   Shelanski, Howard A ^b  

^a Department of Economics ^*

^b UNIVERSITY OF CALIFORNIA   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 39449135584     PISSN: 00036056     EISSN: None     Source Type: Journal    
DOI: None     Document Type: Review

References (105)

1. From 1996 to 2005 the Antitrust Division of the U.S. Department of Justice conducted an average of 210 merger investigations each year, which is more than all of the Division's other civil and criminal investigations combined. U.S. Dep't. of Justice, Antitrust Division, Workload Statistics 1996-2005 (2006). Merger investigations constitute a similarly important part of the Federal Trade Commission's work. FED. TRADE COMM'N, CHAIRMAN'S REPORT 2007 at 5-10, available at http://www.ftc.gov/os/2007/04/Chairmans Report2007.pdf.
2. For purposes of this article, we are taking as given that the statutory objective of competition policy is to promote consumer welfare, .see, e.g., FTC v. University Health, Inc., 938 F.2d 1206, 1222-23 (11th Cir. 1991);
3. United States v. United Tote, Inc., 768 F. Supp. 1064, 1084-85 (D. Del. 1991), notwithstanding strong economic arguments for the use of a total surplus standard.
4. See, e.g., Oliver E. Williamson Economies as an Antitrust Defense: The Welfare Tradeoffs, 58 AM. ECON. REV. 1372 (1968).
5. For a survey and assessment of economic arguments in favor of applying a consumer welfare standard in antitrust analysis, see Joseph Farrell & Michael L. Katz, The Economics of Welfare Standards in Antitrust, 2 COMPETITION POL'Y INT'L 3 (2006).
6. An obvious problem with this view is that something has to happen in the future, so not all probabilities can be zero. Thus, by assigning zero probabilities to the lowprobability events, the agencies and courts either implicitly assume the alternative will happen with certainty or are forming irrational beliefs.
7. For a discussion of recent enforcement actions involving innovative industries, see Michael L. Katz & Howard A. Shelanksi, Mergers and Innovation, 74 ANTITRUST L.J. 1 (2007).
8. In his analysis of how different responses to uncertainty can help to explain international differences in enforcement decisions (notably differences between the United States and European Union), Ken Heyer also advocates the application of decision theory to merger analysis and other antitrust actions. Because we address only the application of decision theory to merger analysis, we share his perspective but not his focus and emphases. Ken Heyer, A World of Uncertainty: Economics and the Globalization of Antitrust, 72 ANTITRUST L.J. 375 (2005).
9. Several other authors have recommended the use of decision theory to assess the burden of proof and antitrust policy rules targeted at particular types of conduct (e.g., whether to have a high or low standard of proof for finding tying to be anticompetitive). That work tends to focus on the relative likelihood of false positives and false negatives for a given rule, rather than estimating the consumer welfare effects of a particular conduct in a specific case. Although the fundamental tools are similar, the divergent applications raise different considerations. See, e.g., Richard A. Posner, An Economic Approach to Legal Procedure and Judicial Administration, 2J. LEGAL STUD. 399 (1973);
10. Frederick Beckner III & Steven C. Salop, Decision Theory and Antitrust Rules, 67 ANTITRUST L.J. 41 (1999) ;
11. David S. Evans & A. Jorge Padilla, Designing Antitrust Rules for Assessing Unilateral Practices: A NeoChicago Approach (AEI-Brookings Joint Center for Regulatory Studies, Related Publication 04-20 (2004));
12. Keith N. Hylton & Michael Salinger, Tying Law and Policy: A Decision Theoretic Approach (Boston University School of Law, Working Paper No. 01-04, 2004).
13. Merger analysis is also used to assess alternative remedies, such as partial divestitures. Ultimately, the questions asked are the same: would the proposed transaction-in its originally proposed form or subject to modification-harm consumers?
14. 15 U.S.C. § 18. Mergers can also be challenged under Section 1 of the Sherman Act, 15 U.S.C. § 1, which bars "[e]very contract, combination in the form of trust or otherwise, or conspiracy in restraint of trade or commerce among the several States, or with foreign nations." In addition, mergers can be challenged by state enforcers under state antitrust laws or by private parties, the latter of which often face antitrust-injury and standing hurdles and which may be motivated by very different considerations than government enforcers. Our focus here is on federal enforcement.
15. See FTC v. University Health, Inc., 938 F.2d 1206, 1211 n.12 (11th Cir. 1991); FTC v. PPG Industries, Inc., 798 F.2d 1500, 1503 (D.C.Cir. 1986);
16. FTC v. Staples, Inc., 970 F. Supp. 1066, 1081-82 (D.D.C.1997).
17. U.S. Dep't of Justice & Fed. Trade Comm'n, Horizontal Merger Guidelines (1992, revised 1997), available at http://www.ftc.gov/bc/docs/ horizmer.htm [hereinafter Merger Guidelines].
18. Brown Shoe Co. v. United States, 370 U.S. 294 (1962). For a discussion of whether this stage of the analysis is helpful or necessary to the determination of the welfare effects of a merger, see infra Part III.B.
19. Merger Guidelines, supra note 9, § 4. See, e.g., Staples, 970 F. Supp. at 1088;
20. United States v. Rockford Mem'l Corp., 717 F. Supp. 1251, 1288-92 (N.D. 111. 1989).
21. The concentration thresholds set out in the Merger Guidelines, supra note 9, § 1.5, provide a highly incomplete description.
22. FED. TRADE COMM'N, ANTICIPATING THE 21ST CENTURY: COMPETITION POLICY IN THE NEW HIGH-T ECH, GLOBAL MARKETPLACE 27 (1996), available at http://www.ftc.gov/opp/ global/report/gc_v1.pdf, [hereinafter 1996 FTC Report].
23. Merger Guidelines, supra note 9, § 4.
24. Id. (emphasis added).
25. See, e.g., United States, v. Country Lake Foods, 754 F. Supp. 669, 680 n.13 (D. Minn. 1990);
26. United States v. Rockford Mem'l Corp., 717 F. Supp. 1251, 1289 (N.D. 111. 1989).
27. See David Balto, The Efficiency Defense in Merger Review: Progress or Stagnation, ANTITRUST, Fall 2001, at 74 n.39 and accompanying text.
28. 970 F. Supp. 1066, 1089 (D.D.C. 1997).
29. Id.
30. 246 F.3d 708, 721 (D.C. Cir. 2001).
31. Id. at 720.
32. Id.
33. Id. at 721.
34. See id. at 722 (finding the efficiencies not to be merger-specific).
35. See Balto, supra note 17.
36. The following nonmerger hypothetical illustrates our fundamental objection to the use of probability thresholds. Suppose you were diagnosed as having a disease that, absent treatment, would prove fatal within six months. Moreover, suppose that the only known course of treatment cost $1,000 and had no side effects but, through years of experience, was known to work in only one in ten patients. We would be very surprised if any reader would refuse to undertake the treatment on the grounds that the chance of success was so low that any benefits of being cured were too speculative to count.
37. Moreover, as discussed below, it is well known that a consumer who is risk-neutral with respect to income will be risk-loving with respect to prices.
38. Balto, supra note 17.
39. Michael L. Katz, The Role of Efficiency Considerations in Merger Control: What WeDo in the U.S., in EC MERGER CONTROL: A MAJOR REFORM IN PROGRESS (Götz Drauz & Michael Reynolds eds., 2003). This assumption has been called a "standard deduction" for merger efficiencies. Michael Salinger, Director, FTC Bureau of Economics, Presentation to ABA Antitrust Section Economics Committee Brown Bag: Four Questions About Horizontal Merger Enforcement 3 (Sept. 14, 2005), available at http://www.ftc.gov/speeches/ salinger.htm.
40. U.S. Dep't of Justice & Fed. Trade Comm'n, Commentary on the Horizontal Merger Guidelines (2006), available at http://www.usdoj.gov/ atr/public/guidelines/215247.htm [hereinafter Merger Guidelines Commentary].
41. See id. (citing recent cases involving the analysis of entry). This approach is sometimes rationalized on the grounds that there is little credible evidence regarding entry, but-as we discuss below-the lack of credible evidence is typically better handled by adjusting the probabilities to account for the lack of evidence rather than applying an arbitrary threshold for rounding probabilities to zero.
42. Specifically, for non-durable goods the agencies tend to look only two years forward, although the Merger Guidelines provide no rationale for this cutoff. Merger Guidelines, supra note 9, § 3.2. The agencies sometimes take a more sophisticated view, at least when deciding whether to file a case against a proposed merger, if not in court.
43. Id. at n.37.
44. We present a somewhat stylized depiction of current practice as a probability-threshold-oriented approach. We recognize that enforcement is not based exclusively on probabilities, but we think it fair to say that the current approach emphasizes likely events over unlikely ones and, thus, for purposes of our comparison present it as a probabilitythreshold-based framework.
45. Merger Guidelines, supra note 9, § 4 (emphasis added).
46. Id. at n.37. To the extent that future efficiencies are discounted to account for the time value of money, this approach is sound. Our concern is with the use of discounting to account for uncertainty.
47. In our example, either there are no efficiencies or efficiencies are fully realized. In practice, there may be several different levels of efficiencies that might be attained with positive probability. The expected-payoff readily generalizes to any number of possibilities.
48. As we will discuss below, payoffs can be measured in ways that incorporate the decision maker's attitudes towards risks and process errors. We are taking the payoffs to be dollar consumer benefits in this example solely for expositional convenience.
49. We note that the use of econometric studies and formal simulation models can contribute to this problem if they are not properly interpreted. Specifically, there is sometimes a tendency to focus on the "point estimates" that come out of models, rather than to consider the full distributions of the predicted values.
50. Although it is counter-intuitive to many people, if one is making a prediction about a binary outcome (e.g., whether a coin toss results in heads or tails), each of the two probabilities tends to 50 percent, not zero, as uncertainty grows.
51. We discuss below why temporal cutoffs are not an appropriate way to incorporate uncertainty that increases over time.
52. Because this approach considers strategic interaction with other parties, it can be viewed as crossing over from decision theory to game theory. For a brief survey of formal models of the interaction between antitrust enforcement and the population of proposed mergers, see Farrell & Katz, supra note 2.
53. We also want to be clear that we do not consider current policy to be an example of this practice. We have in mind thresholds somewhere below 5 percent.
54. p.
55. 1996 FTC Report, supra note 13, at 32.
56. Id. at 32, 34.
57. Id.
58. See, e.g., DAVID M. KREPS, A COURSE IN MICROECONOMIC THEORY 82 (1990).
59. This fact is what underlies consumers' preference for risk with respect to price changes discussed supra note 27.
60. Another implication of this phenomenon is that merger review should integrate the efficiencies analysis and competitive effects analysis. Simply summing up separate, expected price changes might be an improvement over the probability-threshold approach of current merger enforcement. But calculation of an integrated expected-consumer-welfare effect would clearly be better than either. Gregory Werden offers a clever methodology that integrates the market power and efficiencies analyses and, in effect, allows one to determine if the price effects of the most likely efficiencies will outweigh the price effects of the most likely increase in market power. However, because Werden's computations are based on the most likely values of relevant variables, rather than computing the expected change in consumer welfare, it is subject to die criticisms of Part H.B above. Gregory J. Werden, A Robust Test for Consumer Welfare Enhancing Mergers Among Sellers of Differentiated Products, 44 J. INDUS. ECON. 409 (1996).
61. This issue arose in the dispute over Oracle's merger with PeopleSoft. United States, v. Oracle Corp., 331 F. Supp. 2d 1098, 1129-30 (N.D. Cal. 2004).
62. 2 =p(0.1×α-0.095×p).
63. Pedro P. Barros & Célia C. Cabral, Cost Savings, Uncertainty and the Efficiency Defence in Merger Evaluation (unpublished manuscript, Universidade Nova de Lisboa, June 8, 2001);
64. Philippe Choné & Laurent Linnemer, Assessing Horizontal Mergers Under Uncertain Efficiency Gains (unpublished manuscript, CREST-LEI Oct. 2006) (manuscripts on file with authors).
65. Rabah Amir, Effronsyni Diamantoudi, & Licun Xue, Merger Performance Under Uncertain Efficiency Gains (Working Paper, Fondazione Eni Enrico Mattei, May 2004) (on file with authors).
66. 15 U.S.C. § 18.
67. Merger Guidelines Commentary, supra note 30.
68. However, this statement can also be read as excluding efficiencies from consideration unless they clear thresholds for likelihood and magnitude.
69. FTC v. Staples, Inc., 970 F. Supp. 1066, 1089 (D.D.C. 1997).
70. United States v. Oracle Corp., 331 F. Supp. 2d 1098 (N.D. Cal. 2004).
71. Id. at 1175.
72. We are not offering an opinion on whether (a) is, in fact, a correct statement in this particular case.
73. See, e.g., George W. Stocking & Willard F. Mueller, The Cellophane Case and the New Competition, 45 AM. ECON. REV. 29, 44-48 (1955).
74. Merger Guidelines, supra note 9, § 1.0;
75. FTC v. Swedish Match, 131 F. Supp. 2d 151, 160 (D.D.C. 2000);
76. California v. Sutter Health Sys., 130 F. Supp. 2d 1109, 1120 (N.D. Cal. 2001).
77. See also Michael L. Katz & Carl Shapiro, Critical Loss: Let's Tell the Whole Story, ANTITRUST, Spring 2003, at 49.
78. Merger Guidelines, supra note 9, §§ 1.0 & 1.11.
79. We are assuming that the baseline price is greater than or equal to incremental cost.
80. See United States v. SunGard Data Sys., 172 F. Supp. 2d 172, 182-88 (D.D.C. 2001) (example of how difficult it can be to determine where along a continuum of products a clean market boundary can be drawn).
81. United States v. Oracle Corp., 331 F. Supp. 2d 1098 (N.D. Cal. Sept. 9, 2004).
82. Id. at 41.
83. FTC v. Arch Coal Inc., 2004 U.S. Dist. LEXIS 15996 (D.D.C. Aug. 16, 2004).
84. Jonathan B. Baker, Product Differentiation: Contemporary Empirical Merger Analysis, 5 GEO. MASON L. REV. 347, 351 (1997).
85. Janusz A. Ordover & Daniel M. Wall, Understanding Econometric Methods of Market Definition, ANTITRUST, Summer 1989, at 20.
86. Indeed, the very question asked by the hypothetical monopolist test raises issues about the separation of market definition and competitive effects analysis. Under the Merger Guidelines' approach, the answer to the following question provides the basis of market definition: Would a hypothetical monopolist with control and ownership of a particular set of products be able to raise price profitably in a significant way, holding the prices of other products constant? But why not make predictions about what actual suppliers would do rather than focus on a hypothetical monopolist? Specifically, why not ask directly whether the merging parties would find it profitable to raise price by a significant amount post-merger? If one possesses the answer to that question, then the answer to the hypothetical monopolist question is superfluous. One distinction is that the second question does not entail holding other prices fixed. But in unilateral effects cases, an assumption along somewhat similar lines is made.
87. They would, however, have to establish significant expected harm to consumer welfare.
88. This kind of issue is particularly likely to arise in markets with spatial differentiation, where each supplier competes with its neighbors, which then has effects on their neighbors, but the decay of this chaining process-and thus where a Merger Guidelines-style relevant market ends-is difficult to determine.
89. One might wonder why the plaintiff would not simply include product 3 in the market on the grounds that the defendant would never attack the market definition for being overinclusive, In cases where the marginal product is clearly irrelevant, such an approach would make sense. In cases where the effect of the marginal product is ambiguous but potentially undermines plaintiffs theory of harm, the plaintiff would want to plead the narrower market in the hope of eliminating consideration of product 3 's potential effects. Moreover, in our experience, the government in practice often fails to take the broader approach, possibly in the belief that there is always some gain to the plaintiff from establishing a narrower market.
90. FTC v. Indiana Fed'n of Dentista, 476 U.S. 447, 458 (1986) (quoting 7 PHILLIP AREEDA, ANTITRUST LAW ¶ 1511, at 429 (1986)).
91. Todd v. Exxon Corp., 275 F.3d 191, 206 (2d Cir. 2001).
92. Andrew I. Gavil, Copperweld 2000: The Vanishing Gap Between Sections 1 and 2 of the Sherman Act, 68 ANTITRUST L.J. 87, 99 (2000).
93. Similarly, the effects of myopia might run in either direction. For example, the agencies and courts might fail to recognize that there will be significantly fewer competitors in the future, or they might fail to recognize significant new entry that will be occurring as the result of ongoing innovation.
94. Section 13(b) of the Federal Trade Commission Act, 15 U.S.C. § 53(b).
95. See, e.g., FTC v. Staples, Inc., 970 F. Supp. 1066, 1071 (D.D.C. 1997).
96. Id.
97. Id at 1072.
98. Id. at 1071.
99. Id. at 1092 (emphasis added).
100. It is important to note that the Commission does not have to show irreparable harm to prevail in its request for a preliminary injunction. FTC v. H.J. Heinz, 246 F.3d 708, 714 (D.C. Cir. 2001);
101. Staples, 970 F. Supp at 1071 n.2.
102. See, e.g., Heinz, 246 F.3d at 726 (quoting FTC v. Dean Foods Co., 384 U.S. 597 (1966)).
103. Id. at 727.
104. Recent work in economics and psychology suggests that people systematically misestimate probabilities and exhibit other biases in making judgments under uncertainty. An interesting avenue for future research is to incorporate recognition of established biases into the analytical and litigation process. For example, one might want explicitly to impose a rule that various conditional probabilities have to be logically consistent with one another. For an accessible survey of the many different biases, see MAX H. BAZERMAX, JUDGMENT IN MANAGERIAL DECISION MAKING (5th ed. 2002).
105. See, in particular, the discussion id. at 25 and 26 (observing that many people ascribe a higher probability to event A than event B even when the occurrence of event A is conditional on event B's taking place).