UCLA

In this dissertation we describe conditions for nonparametric identification and methods for estimating dynamic simultaneous equation models. These models have two distinct sources of endogeneity: lagged dependent variables that are related to autocorrelated unobservable variables and endogeneity through a simultaneous equations structure. Until now, nonparametric estimation has been limited to models with either one or the other. In the first chapter we show that the structural functions in such models are identified with panel data under assumptions commonly made in nonparametric econometrics. We do so by borrowing intuition from existing literature on dynamic panel models. In the second chapter of the dissertation we describe conditions needed for consistent and asymptotically normal nonparametric estimation of dynamic simultaneous equations models. In the third chapter we nonparametrically estimate dynamic demand functions for airline travel using recent data.