CREST Internal Seminar in Microeconomics : 

Abstract: The behavior of mix and match is pervasive, while applied researchers often estimate demand model of single products, ruling out complementarity among products. The model of demand for bundles typically relies on the availability of bundle-level choice data. In some situations, however, researchers may only observe product-level market shares, e.g., aggregate sales data in retailing, vote shares of elections on election day. This paper studies a random coefficients discrete choice model of bundles when only product-level market shares are available. Similar to BLP model for single products, I use a two-step identification and estimation strategy. First, I invert product-level market shares to mean utilities of products using a novel demand inverse to deal with complementarity. This demand inverse can be implemented by Jacobian based algorithms in estimation. Second, I use instrument variables to deal with endogenous prices. I provide constructive completeness conditions for identification and form GMM conditions in estimation. Finally, I examine the practical performance of the methods in the context of the ready-to-eat (RTE) cereal industry in the USA. Aligned with recent findings in the literature, the estimation results suggest substantial complementarity among different RTE cereal brands. Moreover, ignoring complementarity may result in misleading counterfactual simulations.

CREST Internal Seminar in Microeconomics : 

Abstract: The behavior of mix and match is pervasive, while applied researchers often estimate demand model of single products, ruling out complementarity among products. The model of demand for bundles typically relies on the availability of bundle-level choice data. In some situations, however, researchers may only observe product-level market shares, e.g., aggregate sales data in retailing, vote shares of elections on election day. This paper studies a random coefficients discrete choice model of bundles when only product-level market shares are available. Similar to BLP model for single products, I use a two-step identification and estimation strategy. First, I invert product-level market shares to mean utilities of products using a novel demand inverse to deal with complementarity. This demand inverse can be implemented by Jacobian based algorithms in estimation. Second, I use instrument variables to deal with endogenous prices. I provide constructive completeness conditions for identification and form GMM conditions in estimation. Finally, I examine the practical performance of the methods in the context of the ready-to-eat (RTE) cereal industry in the USA. Aligned with recent findings in the literature, the estimation results suggest substantial complementarity among different RTE cereal brands. Moreover, ignoring complementarity may result in misleading counterfactual simulations.