Merger simulation: A simplified approach with new applications

Antitrust Law Journal

Volumn 69, Issue 3, 2001, Pages 883-912

  Epstein, Roy J ^a   Rubinfeld, Daniel L ^b,c  

^a Law and Economics Consulting Group   (United States)

^b UNIVERSITY OF CALIFORNIA   (United States)

^c US Department of Justice   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 0442279264     PISSN: 00036056     EISSN: None     Source Type: Journal    
DOI: None     Document Type: Article

References (58)

1. For a recent survey, see Jonathan B. Baker & Daniel L. Rubinfeld, Empirical Methods Used in Antitrust Litigation: A Review and Critique, 1 J. AM. L. & ECON. REV. 386 (1999).
2. The 1997 U.S. Department of Justice and Federal Trade Commission Horizontal Merger Guidelines and the 1995 U.S. Department of Justice and Federal Trade Commission Intellectual Property Guidelines have also emphasized the potential effects of a transaction on innovation, in general, and on the intensity of research and development efforts, in particular. For a general discussion and further references, see Daniel L. Rubinfeld & John Hoven, Innovation and Antitrust Enforcement, in DYNAMIC COMPETITION AND PUBLIC POLICY ch. 3 (Jerome Ellig ed., 2001). To our knowledge, merger simulation has yet to be applied to evaluate competitive issues that involve innovation markets explicitly.
3. Unilateral effects simulation can predict price increases or decreases for a merger involving firms in the same market, depending on efficiencies and changes in market structure, such as repositioning and divestitures.
4. See, e.g., Jerry Hausman, Gregory Leonard & J. Douglas Zona, Competitive Analysis with Differentiated Products, ANNALES D'ÉCONOMIE ET DE STATISTIQUE 34 (1994); Gregory J. Werden, Simulating the Effects of Differentiated Products Mergers: A Practical Alternative to Structural Merger Policy, 5 GEO. MASON L. REV. 363 (1997); Carl Shapiro, Mergers with Differentiated Products, ANTITRUST, Spring 1996, at 23; Gregory J. Werden, Simulating Unilateral Competitive Effects from Differentiated Products Mergers, ANTITRUST, Spring 1997, at 27; Jerry A. Hausman & Gregory K. Leonard, Economic Analysis of Differentiated Products Mergers Using Real World Data, 5 GEO. MASON L. REV. 321 (1997); see also Gregory J. Werden The Effects of Differentiated Products Mergers: A Practitioners' Guide, in STRATEGY AND POLICY IN THE FOOD SYSTEM: EMERGING ISSUES 95 (Julie A. Caswell & Ronald W. Cotterill eds., 1997).
5. See, e.g., Jerry Hausman, Gregory Leonard & J. Douglas Zona, Competitive Analysis with Differentiated Products, ANNALES D'ÉCONOMIE ET DE STATISTIQUE 34 (1994); Gregory J. Werden, Simulating the Effects of Differentiated Products Mergers: A Practical Alternative to Structural Merger Policy, 5 GEO. MASON L. REV. 363 (1997); Carl Shapiro, Mergers with Differentiated Products, ANTITRUST, Spring 1996, at 23; Gregory J. Werden, Simulating Unilateral Competitive Effects from Differentiated Products Mergers, ANTITRUST, Spring 1997, at 27; Jerry A. Hausman & Gregory K. Leonard, Economic Analysis of Differentiated Products Mergers Using Real World Data, 5 GEO. MASON L. REV. 321 (1997); see also Gregory J. Werden The Effects of Differentiated Products Mergers: A Practitioners' Guide, in STRATEGY AND POLICY IN THE FOOD SYSTEM: EMERGING ISSUES 95 (Julie A. Caswell & Ronald W. Cotterill eds., 1997).
6. See, e.g., Jerry Hausman, Gregory Leonard & J. Douglas Zona, Competitive Analysis with Differentiated Products, ANNALES D'ÉCONOMIE ET DE STATISTIQUE 34 (1994); Gregory J. Werden, Simulating the Effects of Differentiated Products Mergers: A Practical Alternative to Structural Merger Policy, 5 GEO. MASON L. REV. 363 (1997); Carl Shapiro, Mergers with Differentiated Products, ANTITRUST, Spring 1996, at 23; Gregory J. Werden, Simulating Unilateral Competitive Effects from Differentiated Products Mergers, ANTITRUST, Spring 1997, at 27; Jerry A. Hausman & Gregory K. Leonard, Economic Analysis of Differentiated Products Mergers Using Real World Data, 5 GEO. MASON L. REV. 321 (1997); see also Gregory J. Werden The Effects of Differentiated Products Mergers: A Practitioners' Guide, in STRATEGY AND POLICY IN THE FOOD SYSTEM: EMERGING ISSUES 95 (Julie A. Caswell & Ronald W. Cotterill eds., 1997).
7. See, e.g., Jerry Hausman, Gregory Leonard & J. Douglas Zona, Competitive Analysis with Differentiated Products, ANNALES D'ÉCONOMIE ET DE STATISTIQUE 34 (1994); Gregory J. Werden, Simulating the Effects of Differentiated Products Mergers: A Practical Alternative to Structural Merger Policy, 5 GEO. MASON L. REV. 363 (1997); Carl Shapiro, Mergers with Differentiated Products, ANTITRUST, Spring 1996, at 23; Gregory J. Werden, Simulating Unilateral Competitive Effects from Differentiated Products Mergers, ANTITRUST, Spring 1997, at 27; Jerry A. Hausman & Gregory K. Leonard, Economic Analysis of Differentiated Products Mergers Using Real World Data, 5 GEO. MASON L. REV. 321 (1997); see also Gregory J. Werden The Effects of Differentiated Products Mergers: A Practitioners' Guide, in STRATEGY AND POLICY IN THE FOOD SYSTEM: EMERGING ISSUES 95 (Julie A. Caswell & Ronald W. Cotterill eds., 1997).
8. See, e.g., Jerry Hausman, Gregory Leonard & J. Douglas Zona, Competitive Analysis with Differentiated Products, ANNALES D'ÉCONOMIE ET DE STATISTIQUE 34 (1994); Gregory J. Werden, Simulating the Effects of Differentiated Products Mergers: A Practical Alternative to Structural Merger Policy, 5 GEO. MASON L. REV. 363 (1997); Carl Shapiro, Mergers with Differentiated Products, ANTITRUST, Spring 1996, at 23; Gregory J. Werden, Simulating Unilateral Competitive Effects from Differentiated Products Mergers, ANTITRUST, Spring 1997, at 27; Jerry A. Hausman & Gregory K. Leonard, Economic Analysis of Differentiated Products Mergers Using Real World Data, 5 GEO. MASON L. REV. 321 (1997); see also Gregory J. Werden The Effects of Differentiated Products Mergers: A Practitioners' Guide, in STRATEGY AND POLICY IN THE FOOD SYSTEM: EMERGING ISSUES 95 (Julie A. Caswell & Ronald W. Cotterill eds., 1997).
9. See, e.g., Jerry Hausman, Gregory Leonard & J. Douglas Zona, Competitive Analysis with Differentiated Products, ANNALES D'ÉCONOMIE ET DE STATISTIQUE 34 (1994); Gregory J. Werden, Simulating the Effects of Differentiated Products Mergers: A Practical Alternative to Structural Merger Policy, 5 GEO. MASON L. REV. 363 (1997); Carl Shapiro, Mergers with Differentiated Products, ANTITRUST, Spring 1996, at 23; Gregory J. Werden, Simulating Unilateral Competitive Effects from Differentiated Products Mergers, ANTITRUST, Spring 1997, at 27; Jerry A. Hausman & Gregory K. Leonard, Economic Analysis of Differentiated Products Mergers Using Real World Data, 5 GEO. MASON L. REV. 321 (1997); see also Gregory J. Werden The Effects of Differentiated Products Mergers: A Practitioners' Guide, in STRATEGY AND POLICY IN THE FOOD SYSTEM: EMERGING ISSUES 95 (Julie A. Caswell & Ronald W. Cotterill eds., 1997).
10. In recent years the agencies have begun to look critically at remedies involving restructuring. See, e.g., Robert Pitofsky, Chairman, FTC, The Nature and Limits of Restructuring in Merger Review, Remarks at Cutting Edge Antitrust Conference (Feb. 17, 2000), available at http://www.ftc.gov; and Richard G. Parker & David A. Balto, The Evolving Approach to Merger Remedies, ANTITRUST REP., May 2000.
11. In recent years the agencies have begun to look critically at remedies involving restructuring. See, e.g., Robert Pitofsky, Chairman, FTC, The Nature and Limits of Restructuring in Merger Review, Remarks at Cutting Edge Antitrust Conference (Feb. 17, 2000), available at http://www.ftc.gov; and Richard G. Parker & David A. Balto, The Evolving Approach to Merger Remedies, ANTITRUST REP., May 2000.
12. For a lawyer's assessment of merger simulation, see James F. Rill, Practicing What They Preach: One Lawyer's View of Econometric Models in Differentiated Products Mergers, 5 GEO. MASON L. REV. 393 (1997).
13. For an overview of publicly available merger simulation tools, see http://www.antitrust. org/economics/mergers/ simulation.html.
14. For a basic introduction to the "Nash-Bertrand" equilibrium, see ROBERT S. PINDYCK & DANIEL L. RUBINFELD, MICROECONOMICS ch. 12 (5th ed. 2000); a more advanced presentation appears in JEAN TIROLE, THE THEORY OF INDUSTRIAL ORGANIZATION (1988).
15. For a basic introduction to the "Nash-Bertrand" equilibrium, see ROBERT S. PINDYCK & DANIEL L. RUBINFELD, MICROECONOMICS ch. 12 (5th ed. 2000); a more advanced presentation appears in JEAN TIROLE, THE THEORY OF INDUSTRIAL ORGANIZATION (1988).
16. In a general demand model there is no requirement that own-price elasticities be equal for the different brands or that cross-price elasticities take on particular values.
17. See Appendix equation (A1). Using the first-order conditions to estimate margins avoids the distortions associated with the inclusion and allocation of fixed costs in accounting data, a particular problem for multibrand firms. Moreover, relevant accounting data are likely only to be available for the brands sold by the merging parties. As a result, the FOC approach is particularly useful if one is to perform the simulation when there are more than two firms in the market and data sources are limited. We note, however, that the FOCs may yield negative margins, which are generally not consistent with the assumption that goods are substitutes. Because estimated margins depend on the price elasticities in the model, negative estimated margins could signal that the model is relying on inappropriate elasticities.
18. See Appendix equations (A2) and (A3) for the solution to the relevant optimization problem.
19. See Philip Crooke, Luke Froeb, Steven Tschantz & Gregory Werden, The Effects of Assumed Demand Form on Simulated Post Merger Equilibria, 15 REV. INDUS. ORG. 205 (1999).
20. For the original presentation of AIDS, see Angus Deaton & John Muellbauer, An Almost Ideal Demand System, 70 AM. ECON. REV. 312 (1980).
21. Calibrated-demand models based on other types of demand systems also require comparably strong structural assumptions.
22. Other demand models will also require a similar number of estimated coefficients.
23. In addition to imposing adding-up and homogeneity, the number of parameters can also be reduced significantly by specifying a demand model that results from a multilevel decision-making process. For an evaluation of this approach, see Daniel L. Rubinfeld, Market Definition with Differentiated Products: The Post/Nabisco Cereal Merger, 68 ANTITRUST L.J. 163, 173-76 (2000).
24. See Baker & Rubinfeld, supra note 1, for a survey of a variety of approaches to the calibration of demand systems, including auction models and conjoint survey methods.
25. This approach has long been used in other settings involving economics and law when data are limited. For example, in State Industries Inc. v. Mor-Flo Industries, Inc., 883 F.2d 1573 (Fed. Cir. 1989), one of the leading decisions in the patent damages area, the assumption is that the patent holder suffers lost sales equal to its market share applied to the infringer's sales (the remaining infringing sales would have been made by the other firms in the market in proportion to their respective shares). For a recent analysis of this decision see Roy J. Epstein, State Industries and Economics: Rethinking Patent Infringement Damages, 9 FED. CIR. B.J. 367 (2000).
26. Earlier discussions of proportionality in the context of merger analysis include Robert D. Willig, Merger Analysis, Industrial Organization Theory, and Merger Guidelines, in BROOKINGS PAPERS ON ECONOMIC ACTIVITY: MICROECONOMICS 299 (M. Baily & C. Winston eds., 1991); Shapiro, supra note 4.
27. See Horizontal Merger Guidelines ¶ 2.211.
28. Our discussion of PCAIDS focuses on implementation with aggregate market share information. However, the method is also applicable as a set of restrictions that could be imposed when estimating standard AIDS with scanner data. We show in the Appendix that PCAIDS and its extensions to non-proportionality satisfy Slutsky symmetry, an important theoretical property for demand systems.
29. 11, which confirms that adding-up is satisfied.
30. The assumption of proportionality is equivalent to the assumption of "Irrelevance of Independent Alternatives" (IIA) that underlies the logit model. Unlike the logit model, however, the PCAIDS post-merger elasticities are not constrained by IIA.
31. For an extensive discussion of the range of empirical methods that can be used to obtain estimates of demand elasticities, see Baker & Rubinfeld, supra note 1, Section 3.
32. Suppose the prices of all cereals rose by 10%. Because many consumers, particularly children, are likely to continue eating the same quantities of cereal for breakfast (some, of course, will not and consumption of cereal for other purposes, such as snacks, may fall), ready-to-eat demand is not likely to be highly price-sensitive. On the other hand, a 10% increase for a single brand, such as corn flakes, with no change in competitors' prices, will be more price-sensitive, because it will likely result in substantial switching to other products within the cereal category.
33. This follows from the rule of thumb for pricing by a monopolist. See, e.g., PINDYCK & RUBINFELD, supra note 8, ch. 11.
34. The PCAIDS coefficients satisfy adding-up and homogeneity and are symmetric, as required.
35. Cf. Horizontal Merger Guidelines ¶ 2.211: "The market shares of the merging firms' products may understate the competitive effect of concern, when, for example, the products of the merging firms are relatively more similar in their various attributes to one another than to other products in the relevant market. On the other hand, the market shares alone may overstate the competitive effects of concern when, for example, the relevant products are less similar in their attributes to one another than to other products in the relevant market."
36. A statistical test procedure is described in Jerry A. Hausman & Daniel McFadden, Specification Tests for the Multinomial Logit Model, 52 ECONOMETRICA 1219 (1984). One recent AIDS analysis of a grocery item using scanner data indicates that proportionality is reasonable but it does not formally test the hypothesis. See David A. Weiskopf, Assessment of the Relationship Between Various Types of Estimation Bias and the Simulated Economic Impact of Certain Anti-Competitive Scenarios at 55 and Table B2 (unpublished Ph.D. dissertation, Vanderbilt University 1999).
37. A statistical test procedure is described in Jerry A. Hausman & Daniel McFadden, Specification Tests for the Multinomial Logit Model, 52 ECONOMETRICA 1219 (1984). One recent AIDS analysis of a grocery item using scanner data indicates that proportionality is reasonable but it does not formally test the hypothesis. See David A. Weiskopf, Assessment of the Relationship Between Various Types of Estimation Bias and the Simulated Economic Impact of Certain Anti-Competitive Scenarios at 55 and Table B2 (unpublished Ph.D. dissertation, Vanderbilt University 1999).
38. In econometric terms, coefficients estimated with the PCAIDS restrictions could have lower mean square error, i.e., the reduced variance of the estimates may more than balance any bias that is introduced. See, e.g., PINDYCK & RUBINFELD, supra note 8, at 29-32.
39. For a discussion of estimation problems with nested logits, see Gregory J. Werden & Luke M. Froeb, The Effects of Mergers in Differentiated Products Industries: Logit Demand and Merger Policy, 10 J.L. ECON. & ORG. 407, 420 (1994).
40. For an analysis of curvature of alternative demand models, see Crooke et al., supra note 12.
41. We are aware of very few studies that directly compare AIDS and logit using real-world data. A recent article that uses grocery scanner data on white pan bread sales indicates AIDS fit the data significantly better than logit. See Atanu Saha & Peter Simon, Predicting the Price Effect of Mergers with Polynomial Logit Demand, 7 INT'L J. ECON. Bus. 149, 154 (2000).
42. The informative discussion at http://www.antitrust.org/mergers/economics/ simulation.html concludes that "much progress has been made using the linear and nested logit demand specifications. . . . However, more progress can be made, by simulating the effects of mergers within the context of more flexible functional forms, like the AIDS model."
43. Shapiro, supra note 4; Carl Shapiro, Mergers with Differentiated Products, Address Before the ABA and IBA (Nov. 9, 1995), available at http://www.usdoj/atr/public/ speeches/shapiro.spc.txt.
44. Jerry A. Hausman & Gregory K. Leonard, Economic Analysis of Differentiated Products Mergers Using Real World Data, 5 GEO. MASON L. REV. 321 (1997).
45. See FTC v. H.J. Heinz Co., 246 F.3d 708 (D.C. Cir. 2001).
46. The use of composite goods or firms is common in merger simulation because, when appropriate, it greatly diminishes the number of parameters in the model and simplifies the analysis.
47. The elasticity was calculated as the negative of the ratio of sales ($9,407,949) to gross profit ($3,619,424). At the profit-maximizing price for a firm, the negative of its markup of price over cost as a proportion of price equals the inverse of its elasticity. See H.J. Heinz Company Form 10-K for fiscal year ended May 3, 2000, Consolidated Statements of Income, available at http://www.edgar-online.com.
48. Heinz Co., 246 F.3d at 721.
49. We understand (from personal communication) that Jonathan Baker testified (on behalf of Beech-Nut and Heinz) to an expected 15% reduction in marginal cost for the gains passed through to the Beech-Nut brand. According to Baker, these gains would come from a price reduction; the gains to Heinz buyers would come from getting a brand that is 15% higher in quality (at the same price as their old brand, according to the merging parties).
50. See Saha & Simon, supra note 40.
51. The Horizontal Merger Guidelines ¶ 2.12 n.23 indicates that the same analysis applies to both cases.
52. See id. ¶¶ 1.0 & 3.0.
53. We wish to thank Kraft Foods for providing us with the breakfast cereal data.
54. We use the notions of Kids and Adult nests for illustrative purposes only. We believe, nevertheless, that the relevant market for antitrust purposes is all ready-to-eat cereals. See New York v. Kraft Gen. Foods, Inc., 926 F. Supp. 321, 356 (S.D.N.Y. 1995).
55. The Horizontal Merger Guidelines ¶¶ 2.211 and 2.22 leave open the possibility of finding significant unilateral effects when the merging firms have combined market shares of less than 35%, indicating that this criterion is not equal in importance to the HHI safe harbor. For simplicity, however, we will refer to the 35% standard as a safe harbor and investigate its properties.
56. The simulations used an industry elasticity of -1, a brand elasticity of -3 for the first merger partner, and a third firm with a 65% share. There were no efficiencies or nests.
57. The HHI simulations used an industry elasticity of -1, a brand elasticity of -3 for the first merger partner, and merging parties ranging from equal 5% shares to 24% and 1% shares, and a third firm with the residual share.
58. Other researchers who advocate simulation have found little support for the 35% rule and have concluded that the existing HHI criterion "makes sense only if one believes either that mergers are likely to generate no efficiencies or that only consumer welfare should be considered in merger cases." See Gregory J. Werden & Luke M. Froeb, Simulation as an Alternative to Structural Merger Policy in Differentiated Products Industries, in THE ECONOMICS OF THE ANTITRUST PROCESS 77 (Malcolm B. Coate & Andrew N. Kleit eds., 1996).