Key points are not available for this paper at this time.
A novel spatial dynamic panel data (SDPD) model is proposed, incorporating interactive fixed effects and time-varying endogenous spatial weight matrices. The framework captures evolving spatial spillovers and latent heterogeneity without imposing restrictions on the relationship between unobserved factors in the outcome and auxiliary equations. A quasi-maximum likelihood estimator (QMLE) with finite-sample bias correction is developed, and its asymptotic distribution is established using perturbation expansions of linear operators. Monte Carlo simulations show that the estimator performs robustly across various factor configurations, endogeneity levels, and degrees of spatial dependence. An empirical study on U.S. state-level reverse mortgage activity from 2001Q1 to 2013Q4 demonstrates the model’s effectiveness in capturing dynamic spatial interactions in regional financial behavior.
Baltagi et al. (Wed,) studied this question.