Mergers with Differentiated Products: the - Aviv Nevo

RAND Journal of Economics Vol. 31, No. 3, Autumn 2000 pp. 395–421

Mergers with differentiated products: the case of the ready-to-eat cereal industry

Aviv Nevo*

Traditional merger analysis is difficult to implement when evaluating mergers in industries with differentiated products. I discuss an alternative, which consists of demand estimation and the use of a model of postmerger conduct to simulate the competitive effects of a merger. I estimate a brand-level demand system for ready-to-eat cereal using supermarket scanner data and use the estimates to (1) recover marginal costs, (2) simulate postmerger price equilibria, and (3) compute welfare effects, under a variety of assumptions. The methodology is applied to five mergers, two of which occurred and for which I compare predicted to actual outcomes.

1. Introduction

Traditional analysis of horizontal mergers is based primarily on industryconcentration measures. The market is defined and pre- and postmerger market shares of the relevant firms are used to compute pre- and postmerger concentration measures, which give rise to presumptions of illegality. Using this approach to evaluate mergers in industries with differentiated, or closely related but not identical, products is problematic. In many cases the product offerings make it difficult to define the relevant product (or geographic) market. Even if the relevant market can easily be defined, the computed concentration index provides a reasonable standard by which to judge the competitive effects of the merger only under strong assumptions.1 To deal with these challenges, a new methodology to evaluate mergers has been developed. 2 The basic idea consists of ‘‘front-end’’ estimation, in which demand functions and possibly supply relations are estimated, and a ‘‘back-end’’ analysis, in which the estimates are used to simulate the postmerger equilibrium. This article follows this

* University of California, Berkeley, and NBER; nevo@econ.berkeley.edu. I wish to...