Competitive analysis using a flexible demand specification

Journal of Competition Law and Economics

Volumn 1, Issue 2, 2005, Pages 279-301

  Hausman, Jerry A ^a   Leonard, Gregory K ^a,b  

^a MASSACHUSETTS INSTITUTE OF TECHNOLOGY   (United States)

^b NERA ECONOMIC CONSULTING   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 27744531327     PISSN: 17446414     EISSN: None     Source Type: Journal    
DOI: 10.1093/joclec/nhi008     Document Type: Article

References (60)

1. Prior to the mid-1980s, the available aggregate-level data (i.e., data aggregated over individuals to the level of a city or some other geographic region) typically allowed the estimation of own and cross price elasticities for only broad categories of goods, e.g., food, clothing, etc., or for commodities, e.g., wheat, corn, etc. The lack of aggregate-level data on the individual products within broad categories prevented the estimation of own and cross elasticities for these products (e.g., individual brands of bread or soap). The most common exception was where data on individual decisions was available from surveys and data on prices was publicly available. Examples are demand for various modes of transportation and demand for recreational sites.
2. See, e.g., Jerry Hausman, et al., 'Competitive Analysis with Differentiated Products', 34 ANNALES D'ECONOMIE ET DE STATISTIQUE (1994), at 159
3. Gregory Werden and Luke Froeb, 'The Effects of Mergers in Differentiated Products Industries: Logit Demand and Merger Policy', 10 J. LAW ECON. & ORGAN. (1994), at 407
4. Gregory Werden, et al., 'The Use of the Logit Model in Applied Industrial Organization', 3 INT. J. ECON. BUS. (1996), at 83
5. Jerry Hausman, 'Valuation of New Goods Under Perfect and Imperfect Competition', in Timothy F. Bresnahan and Robert J. Gordon (eds), The Economics of New Goods (1997);
6. Jerry Hausman and Gregory Leonard, 'Economic Analysis of Differentiated Products Mergers Using Real World Data', 7 GEO. MASON L. REV. (1997), at 321
7. Aviv Nevo, 'Mergers with Differentiated Products: The Case of Ready-to-Eat Cereal Industry', 31 RAND J. ECON. (2000), at 395
8. Aviv Nevo, 'Measuring Market Power in the Ready-to-Eat Cereal Industry', 69 ECONOMETRICA (2001), at 307
9. Jerry Hausman and Gregory Leonard, 'The Competitive Effects of a New Product Introduction: A Case Study', 50 J. IND. ECON. (2002), at 237.
10. Some of the issues we address, as well as other issues, are also discussed in Daniel Hosken, et al., 'Demand System Estimation and Its Application to Horizontal Merger Analysis', FTC Working Paper (2002).
11. Another consideration, particularly when estimating the price effects of a proposed merger, is the behavior of the demand system as prices move away from the point of approximation. Demand systems that yield the same elasticities at the point of approximation can predict substantially different post-merger price changes. See, e.g., Philip Crooke, et al., 'Effects of Assumed Demand Form on Simulated Postmerger Equilibria', Rev. Indus. Org. (1999), at 205.
12. See, e.g., Erwin Diewert, 'An Application of the Shephard Duality Theorem: A Generalized Leontief Production Function', 79 J. POL. ECON. (1971), at 481 for a definition of flexible functional forms;
13. see also Angus Deaton, 'Demand Analysis', in Zvi Griliches and Michael Intriligator (eds), Handbook of Econometrics, (1986) vol. 3
14. and Robert Pollak and Terrence Wales, Demand System Specification & Estimation (1992).
15. For example, the AIDS and log-log specifications discussed below are not nested within one another.
16. Slutsky symmetry requires that the compensated cross price derivative of product A with respect to product B equals the compensated cross price derivative of product B with respect to product A. Homogeneity of degree zero in prices and expenditure requires that demand for all products be unchanged if the prices of the products and total expenditure all increase by the same percentage. Finally, adding up requires that the sum of expenditures on the individual products equals total expenditure.
17. See Deaton, above n 5.
18. Angus Deaton and John Muellbauer, Economics and Consumer Behavior (1980), 148-59.
19. Angus Deaton and John Meullbauer, 'An Almost Ideal Demand System', 70 AM. ECON. REV. (1980), at 312.
20. We will often refer to Y as total expenditure on the category. Note that consumers' expenditure is equivalent to producers' revenue.
21. Deaton and Meullbauer, above n 10, at 316.
22. See Diewert, above n 5.
23. Deaton and Muellbauer, above n 10, at 312.
24. Pollak and Wales, above n 5.
25. Deaton and Muellbauer, supra note 10, at 312.
26. Deaton and Muellbauer, supra note 10, at 312.
27. See, e.g., Werden and Froeb, above n 2
28. Werden, et al., above n 2
29. Gregory Werden, et al., 'Economic Analysis of Lost Profits From Patent Infringement With and Without Noninfringing Substitutes', 27 AIPLA Q. J. (1997), at 305.
30. See, e.g., Daniel McFadden, 'Econometric Models of Probabilistic Choice', in Charles Manski and Daniel McFadden (eds), Structural Analysis of Discrete Data With Econometric Applications (1981), 222-23.
31. Jerry Hausman, 'Project Independence Report: An Appraisal of U.S. Energy Needs Up to 1985', 6 BELL J. ECON. (1975), at 517
32. McFadden, above n 19, at 222
33. Hausman and Leonard, above n 2, at 322.
34. For ease of exposition, equation (4) assumes that each consumer inelastically purchases one unit from the category. The logit model can be generalized to allow for an 'outside alternative', i.e., a choice for consumers not to purchase any product in the category. However, the conclusions regarding equal cross price elasticities continue to hold in the logit model with an outside alternative.
35. See, e.g., Steve Berry, 'Estimating Discrete Choice Models of Product Differentiation', 25 RAND J. ECON. (1994), at 242.
36. Steve Berry, et al., 'Automobile Prices In Market Equilibrium', 63 ECONOMETRICA (1995), at 841. A similar model was first proposed and estimated in Nicholas S. Cardell, 'Extensions of Multinomial Logit: The Hedonic Demand Model, the Non-Independent Logit Model, and the Ranked Logit Model' (Ph.D. thesis at Harvard University 1989).
37. Nevo, above n 2.
38. Indeed, imposition of the adding-up restriction is problematic, see Deaton and Muellbauer, above n 9, at 17.
39. See, e.g., Pollak and Wales, above n 5, at 53-59
40. Deaton, above n 8, at 1788-93.
41. It is not necessary that the consumer actually go through this thought process for the two-stage budgeting methodology to be an appropriate way of modeling the consumer's decision problem.
42. See, e.g., Charles Blackorby and Anthony Shorrocks, Collected Works of W. M. Gorman, Volume 1: Separability and Aggregation (1995).
43. Use of multi-stage budgeting does impose restrictions on the demand system. However, these restrictions can be tested, see, e.g., Hausman, et al., above n 2.
44. Many of the disadvantages of the log-log-specification discussed above are not present when only a single product (or in this case, category) is addressed.
45. Often total expenditure is not available by city. In that case, another income measure can be used instead, such as personal disposable income.
46. Jerry Hausman and William Taylor, 'Panel Data and Unobservable Individual Effects', 49 ECONOMETRICA (1981), at 1377.
47. Hausman and Leonard, above ns 2, at 250.
48. One must be careful not to choose an alternative demand system specification that forces all cross price elasticities to be positive, e.g., the logit demand system. These restrictions may well be inconsistent with the data.
49. Hosken, et al., above n 3, at 7.
50. See, e.g., Igal Hendel and Aviv Nevo, 'Sales and Consumer Inventory', Unpublished Paper (2002).
51. Time-aggregation provides another test of inventorying behavior. If the estimated elasticities are significantly lower in the time-aggregated model, inventorying behavior might be present. Hosken, et al., above n 3, report that they have found significant differences in elasticities after aggregating over time.
52. Due to confidentiality reasons, we cannot disclose the exact product category.
53. The hypothetical example we are about to use bears no relationship whatsoever to any actual events.
54. One might think that patent infringement litigation rarely involves products for which scanner data would be available. However, we are aware of patent infringement cases involving products in a number of supermarket, drug store, and mass merchandiser categories including lipstick, diapers, and contact lens cleaners. In addition, IMS provides data similar to scannder data that are available for pharmaceuticals and medical devices, which are frequently the subject of patent infringement litigation.
55. Werden, et al., above n 18, discuss a similar approach to estimating lost profits damages in patent infringement cases. The plaintiff may be entitled to reasonable royalty damages in addition to lost profits damages. We focus on lost profits damages only in this example.
56. A defendant might have available to it a non-infringing alternative that it could have sold in the 'but for' world. If it would have been profitable for the defendant to do so, the appropriate 'but for' world would be one where the defendant would have offered that non-infringing alternative product. Damages would depend in part upon how much lower consumer demand for the non-infringing product would have been. For our example, we assume that there was no such product available to the defendant.
57. State Industries, Inc. v Mor-Flo Industries, Inc., 883 F.2d 1573 (Fed. Cir. 1989).
58. Hausman, above n 2,
59. and Hausman and Leonard, above n 2.
60. Specifically, we use the Nash-Bertrand model of price competition. This model is widely used in competitive analyses of differentiated products industries, see the citations in above n 2.