Estimating Complementarity with Large Choice Sets: An Application to Mergers
RAND Journal of Economics, 56(4), 689-707.
Abstract
Standard discrete choice models assume that products are substitutes. This paper develops a GMM estimator that identifies demand complementarity with price endogeneity and large choice sets, applies it to chips and soda, and shows that accounting for complementarity can substantially change merger counterfactuals.
Main Finding
Accounting for chip-soda complementarity lowers predicted soda price increases from a PepsiCo/Frito-Lay and Dr Pepper merger by about 30% relative to a substitutes-only model; post-merger chip prices decrease.
Policy Relevance
Merger simulations in large differentiated-product markets can be misleading when they impose substitution patterns that rule out complementarity.