Market definition with differentiated products: The Post/Nabisco cereal merger

Antitrust Law Journal

Volumn 68, Issue 1, 2000, Pages 163-182

  Rubinfeld, Daniel L ^a  

^a UNIVERSITY OF CALIFORNIA   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 0034399167     PISSN: 00036056     EISSN: None     Source Type: Journal    
DOI: None     Document Type: Article

References (42)

1. New York v. Kraft Gen. Foods, Inc., 926 F. Supp. 321 (S.D.N.Y. 1995).
2. Both Nielsen and IRI use scanning devices to record supermarket sales data for a national random sample covering a range of metropolitan and rural areas. In the IRI Infoscan Data Base, for example, samples are drawn from a supermarket universe that includes stores with annual sales greater than $2 million (which accounts for 82% of U.S. grocery sales). Scanner data was employed, for example, by Jerry Hausman in support of Kodak's petition for modification or termination of a 1921 consent decree restricting Kodak's ability to sell private-label film. Professor Hausman estimated that the cross-price demand elasticities between Kodak and Fuji film were high, and used that evidence to support his opinion that Kodak did not have market power. The court agreed, relying in part on Hausman's testimony. United States v. Eastman Kodak Co., 853 F. Supp. 1454 (W.D.N.Y. 1994), aff'd, 63 F.3d 95 (2d Cir. 1995).
3. See generally Jonathan B. Baker & Timothy F. Bresnahan, Empirical Methods of Identifying and Measuring Market Power, 61 ANTITRUST L.J. 3 (1992); Jonathan B. Baker & Daniel L. Rubinfeld, Empirical Methods Used in Antitrust Litigation: A Review and Critique, 1 J. AM. L. & ECON. ASS'N 386 (1999).
4. See generally Jonathan B. Baker & Timothy F. Bresnahan, Empirical Methods of Identifying and Measuring Market Power, 61 ANTITRUST L.J. 3 (1992); Jonathan B. Baker & Daniel L. Rubinfeld, Empirical Methods Used in Antitrust Litigation: A Review and Critique, 1 J. AM. L. & ECON. ASS'N 386 (1999).
5. David A. Kaye & David A. Freedman, Reference Guide on Statistics, in REFERENCE MANUAL ON SCIENTIFIC EVIDENCE 331-414 (1994); Daniel L. Rubinfeld, Reference Guide on Multiple Regression, in REFERENCE MANUAL ON SCIENTIFIC EVIDENCE 415-470 (1994) (rev. ed. forthcoming). These materials seek to explain empirical methods and highlight issues a court should consider in evaluating the admissibility and probative value of statistical evidence. Admissibility issues are often framed as a question of the interpretation of judicial rules intended to exclude "junk science." See generally Daubert v. Merrell Dow Pharm., Inc., 506 U.S. 914 (1992); Kumho Tire Co. v. Carmichael, 526 U.S. 137 (1999).
6. David A. Kaye & David A. Freedman, Reference Guide on Statistics, in REFERENCE MANUAL ON SCIENTIFIC EVIDENCE 331-414 (1994); Daniel L. Rubinfeld, Reference Guide on Multiple Regression, in REFERENCE MANUAL ON SCIENTIFIC EVIDENCE 415-470 (1994) (rev. ed. forthcoming). These materials seek to explain empirical methods and highlight issues a court should consider in evaluating the admissibility and probative value of statistical evidence. Admissibility issues are often framed as a question of the interpretation of judicial rules intended to exclude "junk science." See generally Daubert v. Merrell Dow Pharm., Inc., 506 U.S. 914 (1992); Kumho Tire Co. v. Carmichael, 526 U.S. 137 (1999).
7. U.S. Department of Justice & Federal Trade Commission, Horizontal Merger Guidelines (1992, as amended 1997), reprinted in 4 Trade Reg. Rep. (CCH) ¶ 13,104. A strong case can be made that the Guidelines' analysis of market definition also is applicable in monopolization cases, albeit with a different treatment of the hypothetical monopolist test. In United States v. Microsoft Corp., for example, both governments' (the United States and the States) testifying economic witnesses, Frederick Warren-Boulton and Franklin Fisher, defined the relevant market as all Intel-based PC desktop operating systems, based on Guidelines principles. Microsoft's economic expert, Richard Schmalensee, argued that a market definition (whether Guidelines-based or not) was inappropriate in this context. United States v. Microsoft Corp., No. 98-1232 (D.D.C. filed May 18, 1998).
8. U.S. Department of Justice & Federal Trade Commission, Horizontal Merger Guidelines (1992, as amended 1997), reprinted in 4 Trade Reg. Rep. (CCH) ¶ 13,104. A strong case can be made that the Guidelines' analysis of market definition also is applicable in monopolization cases, albeit with a different treatment of the hypothetical monopolist test. In United States v. Microsoft Corp., for example, both governments' (the United States and the States) testifying economic witnesses, Frederick Warren-Boulton and Franklin Fisher, defined the relevant market as all Intel-based PC desktop operating systems, based on Guidelines principles. Microsoft's economic expert, Richard Schmalensee, argued that a market definition (whether Guidelines-based or not) was inappropriate in this context. United States v. Microsoft Corp., No. 98-1232 (D.D.C. filed May 18, 1998).
9. Kraft, 926 F. Supp. at 326.
10. Id. at 328-29 (citation omitted). The study also suggests that one in four adults eats Kid cereals (after all, they were once kids), whereas 50% of children eat both Adult and Kid cereals. Id. at 329.
11. Id. at 333 (citations omitted).
12. According to the court: Although it is conceivable that data concerning purchases by households that contain children and adults is not probative of whether individual cereal eaters would switch from one cereal to another (e.g., adults who prefer plain cereals may buy pre-sweet cereals only for their children and not for themselves), . . . There is evidence that household purchasers make purchase decisions themselves, and are not mere "order-takers" for others in the households. Id. at 325.
13. See Jonathan B. Baker, Contemporary Empirical Merger Analysis, 5 GEO. MASON L. REV. 347, 352-55 (1997). Moreover, firms selling branded consumer products may not compete solely on price. Other marketing variables, including advertising and promotion and possibly some physical product characteristics, may affect buyer substitution patterns. If such variables are important in buyer decision making, they should be considered in the analysis of demand, although we do not address their modeling and measurement here.
14. Kraft, 926 F. Supp. at 327.
15. In some antitrust applications, residual demand functions (rather than the more traditional Marshallian - structural - demand functions) are estimated. However, because the Merger Guidelines rely on Marshallian elasticities for market definition, I will not pursue the estimation of residual demand elasticities further here. In Kraft both experts chose to rely on direct rather residual demand elasticities. (A residual demand function omits the quantity variables for products other than the one of interest and adds a group of cost-shift variables thought to affect the price and output of the omitted products.)
16. One exception may arise in time series analyses, including those involving scanner data, in which the variables are observed over short time intervals (e.g., weekly). In some industries, prices may be set sufficiently in advance of consumer purchases so that they are predetermined, thus avoiding simultaneity problems in measuring the demand elasticities with respect to short term price promotions. But the use of high frequency data may require careful modeling of the relationship between consumer inventorying and seller promotions in order to recover the demand elasticities with respect to intermediate term price variation. See Baker, supra note 10, at 352-55.
17. The analysis of the price effects in such situations can be both theoretically and empirically demanding. The theoretical issues are complex because the increased prices of the merging firms' products can induce increases in rivals' prices, and because the analysis of these effects can be sensitive to assumptions about the form of demand functions (e.g., whether linear or nonlinear), demand symmetries (e.g., whether or not there are symmetries in the system of demand equations), the nature of the interaction among firms (e.g., Bertrand or Cournot competition), and the possible presence of economies of scale and scope. In cases involving consumer products, the analysis may also be sensitive to the assumptions that are made (or not made) about the timing of consumer purchases. See, e.g., Raymond Deneckere & Carl Davidson, Incentives to Form Coalitions with Bertrand Competition, 16 RAND J. ECON. 473, 476-86 (1985).
18. This instrumental variables approach can be an attractive method for exploiting scanner data containing prices in multiple cities when it is difficult to observe cost-shift variables in the same frequency as the price data if the nationwide component reflects mainly nationwide variation in product cost rather than demand. Indeed, this approach may have generated reasonable estimates in one data set involving breakfast cereals. But when the method is employed in other industries (or other time periods in the same industry), where the critical assumption cannot be tested, it would be useful to have some basis for believing that shocks to the nationwide component of prices in the industry and time period under study mainly result from variation in cost, not demand. See Baker & Rubinfeld, supra note 3, for further discussion.
19. Specification issues related to the omission of advertising and promotion when buyers hold inventories, and the time period over which the demand elasticity is measured, are considered in Baker, supra note 10, at 347-61, and are not discussed further here.
20. It is generally not a good idea to assume that cross-price elasticities are zero. However, it may be reasonable to impose symmetry among the elements of the Slutsky matrix that underlies the determination of demand substitution, but one would normally expect cross-price elasticities to be asymmetric.
21. Separability of demand and multi-stage budgeting are not equivalent. Weak demand separability is necessary and sufficient for the last (lowest level) of the budgeting process. For the higher stages, the assumption that expenditures can be allocated among groups depends on stronger separability assumptions or on other restrictions on preferences. See generally Aviv Nevo, Mergers with Differentiated Products: The Case of the Ready-to-Eat Cereal Industry (unpublished Ph.D. dissertation 1997) (on file with author); ANGUS DEATON & JOHN MUELLBAUER, ECONOMICS AND CONSUMER BEHAVIOR ch. 16 (1980).
22. Separability of demand and multi-stage budgeting are not equivalent. Weak demand separability is necessary and sufficient for the last (lowest level) of the budgeting process. For the higher stages, the assumption that expenditures can be allocated among groups depends on stronger separability assumptions or on other restrictions on preferences. See generally Aviv Nevo, Mergers with Differentiated Products: The Case of the Ready-to-Eat Cereal Industry (unpublished Ph.D. dissertation 1997) (on file with author); ANGUS DEATON & JOHN MUELLBAUER, ECONOMICS AND CONSUMER BEHAVIOR ch. 16 (1980).
23. According to the court, "[h]is [plaintiffs economic expert] model was biased in favor of finding very low cross-price elasticities between 'adult' and 'kid' cereals." Kraft, 926 F. Supp. at 334.
24. In the appendix, infra, I describe how cross-price elasticities between individual products in different segments can be calculated. The price elasticity of demand for Adult cereals could be calculated either in a single equation in which the "price" of Adult cereals is a function of the quantity of Adult cereals and other variables, or it can be calculated as a weighted sum of the price elasticities calculated on a product-by-product basis. Either approach necessitates aggregation and relies on important assumptions. (Note, in particular, that the "price" of Adult cereals involves aggregation, because it is typically calculated as the ratio of total revenues to total units sold.)
25. The appropriate analysis, that relies on the Lerner condition, asks whether the increased profits from the sale of a higher-priced product to inframarginal (non-switching) consumers outweighs the reduced profits associated with the lost margins of marginal (switching) customers.
26. See Gregory J. Werden, Market Delineation Under the Merger Guidelines: A Tenth Anniversary Retrospective, 38 ANTITRUST BULL. 517 (1993).
27. See Gregory J. Werden, Demand Elasticities in Antitrust Analysis, 66 ANTITRUST L.J. 363 (1998). The court-appointed expert made this point clearly when he based his view that the correct market was all RTE cereals on the fact that "(1) there is no clear break in the chain of substitutes among RTE cereals, and (2) the chain of substitutes for RTE cereals is not along straight, unidirectional lines radiating out in a uniform and orderly manner." Kraft, 926 F. Supp. at 335.
28. According to the court: These [manufacturing] processes are not dedicated to producing cereals in only one marketing segment; . . . If prices for cereals in the "adult" market were to increase significantly relative to the prices of other cereals, Kellogg, General Mills, Post and other manufacturers could increase production of "adult" cereals relatively quickly and easily by redirecting the use of some portion of the capacity currently used to product "kid" cereals. Kraft, 926 F. Supp. at 332 (citation omitted).
29. Id. at 333.
30. According to Judge Wood: In a differentiated product market such as the RTE cereal market, the decision whether to include a product in the market is inevitably somewhat arbitrary, because not all products in a relevant market compete equally with all other products. In such a market, any market definition is likely to exclude some products . . . that are included in the market. Whether such errors of inclusion or exclusion are significant depends in part on the relevant cross-price elasticities of demand between the individual products. Id. at 334. The presence of a chain of substitutes should not, by itself, be seen as dispositive with respect to the market definition question. A Guidelines analysis could support a narrower market definition, even without a clear break in the chain.
31. Id. at 333 (citation omitted).
32. Id. at 361.
33. Id. at 333. This opinion was supported by the court-appointed expert Alfred Kahn, who said: own-price elasticity, in a sense you don't have to know what went into it in setting your profit-maximizing price. If you really know what the price elasticity is, it tells you what the profit-maximizing price is. . . . On the other hand, if you want to know what constitutes a market - and I think that is what Professor Rubinfeld is getting at - then it is necessary to have some sense of the extent to which you are losing if you raise your price because you are losing to substitute, and what are those substitutes. Trial Transcript, vol. 5, at 934, Kraft (No. 93 Civ. 0811).
34. 1992 Merger Guidelines, supra note 5, § 2.21.
35. See, e.g., Werden, supra note 23, at 363-409.
36. See generally Baker & Rubinfeld, supra note 3.
37. The court did find reliable defendant's expert econometric analysis that found insubstantial unilateral effects. Kraft, 926 F. Supp. at 35. According to the court, the "precise amounts of the cross-price elasticities that Professor Rubinfeld found varied depending on the assumptions he made regarding advertising and marketing variables, but his calculations consistently showed that any unilateral effects are likely to be very small." Id. at 356.
38. Id. at 352.
39. Id.
40. Id. at 358 n.9 (citation omitted).
41. 418 U.S. 656, 669 (1974).
42. 641 F. Supp. 1128, 1132 (D.D.C. 1986)