Identification Results in Interaction Models with Nonparametric Payoffs and Weakly Consistent Beliefs

There exists a continuously growing literature focused on identification and estimation of static, strategic-interaction models. This proposal focuses on a simultaneous discrete choice model where a collection of actions are decided by expected-utility maximizing agents. The beliefs of are subjective probability functions over the space of others' choices. This project obtains bounds on the probabilities of agents that are valid: (i) Whether or not there is complete information in payoffs among agents; (ii) regardless of the specific features of the underlying equilibrium selection mechanism; (iii) for any collection of payoffs functions that satisfy the nonparametric restricts and (iv) whether or not beliefs are correct (e.g, as in Nash equilibrium) as long as they are weakly consistent. The papers produced by this proposal would be the first ones in the literature on econometric analysis of interaction models to explicitly aim for this range of robustness. Besides characterizing bounds for probabilities, this proposal also describes nonparametric bounds involving Nash equilibrium profiles. These include bounds for the probability that an arbitrary choice profile is a Nash equilibrium (either mixed or pure), as well as bounds for the probability that a given profile is selected by the agents given that it is a Nash equilibrium. These bounds are valid for any collection of payoff functions that satisfy the nonparametric assumptions. The nonparametric bounds for Nash equilibrium selection are also a contribution of this proposal to the literature. All probability bounds in the proposal satisfy sharpness conditions given the set of assumptions made. The papers resulting from this proposal would also be th first ones to characterize nonparametric bounds for equilibrium selection probabilities in discrete interaction models (both for Nash and the more general behavior).

Broader Impact: The proposal would create an advanced topics course at the Ph.D second-year level on the subject of Robust Inference in Structural Interactions-Based Microeconometric Models. The target audience for this course would be students interested in microeconometric models, empirical industrial organization, and more generally all those interested in applied microeconomics research with structural models in general. This course is intended to produce dissertations centered around robust inference (both in the parametric and/or behavioral sense) in structural models of economic interaction. The project organizes workshops and seminars, where leading authors in related fields present and discuss their most recent work and their connection to the goals and objectives of this proposal.