ALFRED GALICHON'S

MASTERCLASSES


Date TBD
Enzo Di Pasquale
(NYU)
Bundle Choice Estimation
Abstract:
We propose an empirical framework for discrete choice models where agents face a combinatorial optimization problem. We study set and point identification of the model and propose an estimator which is the solution of a convex program. We demonstrate that the estimation problem can be solved in polynomial time, provided the underlying combinatorial problems faced by the agents are themselves solvable in polynomial time. To compute the estimator, we propose two algorithmic approaches: the ellipsoid method and a row-generation technique. We evaluate the numerical performance of these methods through various examples, including cases with supermodular and gross substitutes valuations.