A1679
Title: Profile GMM estimation of spatial dynamic panels with endogenous weight matrix and interactive fixed effects
Authors: Chen Yahui - Xiamen University (China)
Xiaoyi Han - The Wang Yanan Institute for Studies in Economics, Xiamen University (China)
Liangjun Su - Tsinghua University (China)
Jiajun Zhang - Shanghai University of International Business and Economics (China) [presenting]
Abstract: The estimation and inference of spatial dynamic panel data (SDPD) models with interactive fixed effects (IFEs) is studied, where the time-varying (TV) spatial weight matrix is constructed from multidimensional socio-economic variables such as TV trade or mobility flows across regions. The TV spatial weight matrix is modeled using a three-dimensional (3D) panel with a multi-level factor structure, and the possible endogeneity of the spatial weight matrix is captured through the correlation between the error terms in the SDPD outcome equation and the local factors in the 3D panel. Using a control function approach, a two-stage estimation procedure is proposed. First, the local factors are estimated from the 3D panel and plugged into the SDPD equation to control for the endogeneity of the spatial weight matrix. Then, a nuclear norm regularization (NNR)-based profile GMM method is considered, followed by post-NNR GMM iterations. The asymptotic properties of these estimators are rigorously established, and the most efficient GMM estimator (GMME) with the best moment conditions is derived. A test for the endogeneity of the spatial weight matrix is also considered. Monte Carlo simulations demonstrate that the estimators and test statistic perform well in finite samples. The models and estimation methods are applied to study virus transmission through the human mobility network across U.S. states.
