An Econometric Analysis of the Demand for RTE Cereal: Product Market Definition and Unilateral Market Power Effects

Abstract

This research report is the econometric analysis of product market definition and unilateral market power that the senior author presented as expert economic witness for the state of New York in State of New York v. Kraft General Foods et al. at trial in September 1994. It is the first, and to date as of September 1997, only full-scale attempt to present in a federal district court analysis of a merger?s impact using scanner generated brand level data and econometric techniques to estimate brand and category level elasticities of demand. The court rejected this analysis [State of New York v. Kraft General Foods, 926 F. supp. 358 (S.D. N.Y. 1995)]. We think that it should not have done so, and would like to make this work more readily available than the court record so that economists who are working on the analysis of market definition and market power can review it. Also, this study is the only public study to date that has had access to weekly brand level coupon distribution data by city, and Nielsen group rating points (GRP), which are weekly measures of ad exposure for a brand in local city market areas. Thus it provides own and cross brand coupon and advertising elasticities. The empirical results are generally as hypothesized; own advertising and couponing increase sales, competitor activities reduce sales. The genesis of this work during litigation is an interesting story for those who would attempt to do similar work on a merger case. The acquisition of Nabisco Shredded Wheat by Philip Morris via its Kraft General Foods subsidiary occurred in the fall of 1992. Robert Abrams, the New York Attorney General, announced his challenge of the merger in January 1993. For over a year litigation moved through a series of hearings and rulings. In the spring of 1994 and earlier, the state of New York asked via discovery for any scanner data on Ready-to-Eat (RTE) cereal that the defendants possessed. Defendants provided no data. In July 1994, in preparation for the trial that occurred in September/October 1994, the senior author of this report prepared ?Expert Report of Ronald W. Cotterill? [93 Civ. 0811 (KMW)]. Exhibit 6 of that report was an analysis of unilateral market power using brand level elasticities of demand from the business records of the defendants. It documented the exercise of unilateral market power between Post and Nabisco brands. The first section of that exhibit is reproduced as the foreword to this report because it nicely explains unilateral power analysis.1 Defendants responded to this statistical analysis by introducing on the eve of trial (August 29, 1994) in ?Expert Report of Daniel L. Rubinfeld? [their economist] a comprehensive scanner data set for the leading brands of breakfast cereal and an analysis to refute our work. Using the scanner data that finally became available, we had 18 days to produce the analysis presented in this research report. It was presented to the court as Exhibit C in Affidavit of Professor Ronald W. Cotterill, September 16, 1994 (State of New York v. Kraft General Foods et al. 93 Civ. 0811 [KMW]). One of defendant?s responses to this econometric study was to introduce data for four additional brands during trial and claim that their incorporation destroyed a key market definition result of this analysis. In fact, when those brands were included the general conclusions of this study did not change [trial transcript @ Vol. 7, p. 1249]. Defendants also presented other criticisms of this study at trial and we rebutted them. Here we are not presenting arguments made at trial surrounding this report, nor are we presenting any discussion from Judge Wood's opinion wherein she rejects this analysis. A full understanding of this econometric foray into court requires reading of the defendant?s economist expert report, especially the statistical analysis, the trial transcript, and the court?s opinion. The defendant?s economist?s report is briefly critiqued at the end of this report (pages 17-19). It is a classic example of ?garbage in garbage out? analysis. A future report by the senior author of this report will present a complete review.