Estimating Unobserved Individual Heterogeneity through Pairwise Comparisons

This paper proposes a new methodology to study environments with unobserved agent heterogeneity. We focus on the settings where the heterogeneous factor takes values from an unknown finite set, and the economic model yields testable implications in the form of pairwise inequalities. The method produces a consistent classification of economic agents according to their unobserved types. The paper verifies that the method performs quite well in Monte Carlo simulations. We demonstrate empirical usefulness of the method by estimating the model of first-price auction characterized by both agent and auction level unobserved heterogeneity using the data from California highway procurement market.