Institute for Advanced Study
Abstract：In structural models with parameter-dependent support, maximum likelihood problems are nonregular. We reformulate these problems by parameterdependent transformations and propose modified likelihood estimators that have regular asymptotic properties. We then describe several applications to search models, auction models and frontier production functions and demonstrate the performance of our method through Monte Carlo simulations. Lastly, we apply the method to a Vuong non-nested test of additive v.s. multiplicative separable auction-specific heterogeneity in Michigan Department of Transportation procurements.
Key Words: Boundary model; Job search; Auction; Frontier production function; Sutton’s bound test.
