Econometric Analysisof Collusive Behaviorin a Soft‐Drink Market

Abstract

This paper proposes an empirical methodology for studying various (implicit or explicit) collusive behaviors on two strategic variables, which are price and advertising, in a differentiated market dominated by a duopoly. In addition to Nash or Stackelberg behaviors, we consider collusion on both variables, collusion on one variable and competition on the other, etc. Using data on the Coca‐Cola and Pepsi‐Cola markets from 1968 to 1986, full information maximum likelihood estimation of cost and demand functions are obtained allowing for various collusive behaviors. The collusive hypothesis is not rejected, and the best form of collusive behavior is selected via nonnested testing procedures. Using the best model, Lerner indices are computed for both duopolists to provide summary measures of market power. Finally, our approach is contrasted with the conjectural variation approach and is shown to give superior results.

Product Differentiation and Mergers in the Carbonated Soft Drink Industry

“…In Table ( 17), I report the parameter estimates for two speci…cations of the aggregate DCM. In the second speci…cation, I add random coe¢cients on some of the product attributes to be more consistent with the proposed model.…”

“…Adding a random disturbance term directly in (6) gives the standard random utility DCM (Manski and McFadden 1981): 17 The recent popularity of aggregate models for which the underlying consumer behavior re ‡ects (6) makes them an interesting comparison model. By comparing the measures of market power and the e¤ects of mergers in the proposed model to those of the aggregate DCM, I compare the results of modeling multiple-discreteness explicitly as opposed to imposing single-unit purchase behavior, which eases aggregation.…”

“On a Level with Dentists?” Reflections on the Evolution of Industrial Organization

“…A number of authors, including Hendricks et al (2003) have tested behavioral hypotheses in auctions, and the literature on detecting collusion continues to grow, though not at a rapid pace (Porter (2005). Finally, Gasmi et al (1992) use non-nested hypothesis testing to evaluate price and advertising behavior by This is of course not a complete list of recent studies concerned with estimating conduct parameters or testing hypotheses about conduct, but it does not omit large numbers of such studies.…”

Assessing firm behavior in carve-out markets: Evidence on the impact of carve-out policy

“…The DOT must weigh these potential costs and benefits when deciding to impose a carve-out. 3 Gasmi, Laffont and Vuong (1992) similarly use non-nested likelihood ratio tests to examine cooperative behavior of Coca-Cola and Pepsi in the soft drink market. For a more complete survey of applications of this type of statistical test, see Kadiyali, Sudhir and Rao (2001).…”