"Decentralizing Nonlinear Econometric Models"
Abstract: Many nonlinear econometrics models pose substantial challenges with respect to estimation. One approach to overcoming these challenges is to decompose the estimation problem into a set of sub-problems for which tractable estimation approaches exist. This paper provides a unified treatment of identification and estimation through decentralization. We clarify the type of structure the researcher can exploit to make computation tractable and reliable, provide concrete estimation algorithms, and develop a unifying asymptotic framework including bootstrap validity results. In addition, we propose new and computationally attractive estimators for the instrumental variable regression (IVQR) model of Chernozhukov and Hansen (2005).