Abstract

We propose a novel BLP model of demand for bundles that deals with multiple purchases. It allows for Hicksian complementarity among products that are jointly purchased, a feature that is ruled out by BLP models of single products. Besides, our model only relies on the availability of aggregate data at product level (e.g., sales quantities) that are widely available. This feature enables researchers to fast implement reliable policy evaluations without collecting individual-level choice data of bundles. We propose a two-step identification method that implies a straightforward GMM estimation procedure. We estimate the demand for Ready-To-Eat (RTE) cereal and milk in the US and investigate the policy implications of their complementarity. We find that RTE cereal and milk are overall complementary. As a result, policy change in one category may negatively affect the other, leading to new sources of welfare loss in classic policy simulations such as mergers and sugar tax.