Hanqi Wang

Estimating a Rational Inattention Discrete Choice Model: A Semiparametric Approach

People’s choices depend on both the payoffs of the available options and what they learn about those payoffs before deciding. This paper develops methods that use aggregate choice data to distinguish unobserved payoff heterogeneity from costly information acquisition in a rational inattention discrete choice model. Under Shannon information costs, I establish conditions for semiparametric point identification using exogenous payoff and information-cost shifters, leaving the taste shock distribution and the information-cost function nonparametric. With two alternatives, I show that the probability of choosing one option, as a function of an exogenous payoff shifter, can be interpreted as a cumulative distribution function whose moments identify the payoff coefficients and the information-cost function. A standard deconvolution argument then identifies the taste shock distribution. I extend identification to multiple alternatives. I then propose an estimator based on observed market shares and establish its consistency. Because binding parameter constraints can lead to nonnormal limiting distributions, I construct confidence regions by test inversion without requiring asymptotic normality of the estimator. An application to Amazon Mechanical Turk illustrates the implications of information costs for workers’ responses to wages.

Presented at the Econometric Society North American Summer Meeting (Emory University, 2026) and the Canadian Economics Association Annual Conference (Simon Fraser University, 2026).

Directed Search with Rational Inattention

This paper develops a directed search model of matching between heterogeneous workers and firms, in which workers must acquire costly information about wages before deciding where to apply. Workers observe firm types but not realized wages and acquire information at a Shannon entropy cost that can vary across worker types. I define a rationally inattentive search equilibrium in which application choices determine congestion across submarkets, while congestion shapes job-finding probabilities and the returns to information. I first establish equilibrium existence when the wage distribution has finite support and job-finding probabilities depend continuously on workers’ application probabilities. The existence proof allows workers to have multiple optimal information-acquisition strategies. It exploits the fact that linear-fractional maps send convex sets to convex sets, together with workers’ optimality conditions, to apply Kakutani’s fixed-point theorem. I then establish conditions for identifying worker-type-specific information costs using twin job types with identical amenity values but different wages. Given known matching technology and equal job-finding probabilities for the twins, identification uses conditional and ex-ante choice shares to recover information costs. The identifying relationship links the twins’ difference in expected wage income to the difference between their conditional and ex-ante log choice odds. Together, these results provide an equilibrium foundation and an empirical strategy for measuring attention costs in job search.