Penalized Sieve Estimation of Structural Models

Abstract

Estimating structural models is an essential tool for economists. However, existing methods are often inefficient either computationally or statistically, depending on how equilibrium conditions are imposed. We propose a class of penalized sieve estimators that are consistent, asymptotic normal, and asymptotically efficient. Instead of solving the model repeatedly, we approximate the solution with a linear combination of basis functions and impose equilibrium conditions as a penalty in searching for the best fitting coefficients. We apply our method to an entry game between Walmart and Kmart.