Dynamic Panel Threshold Spatial Durbin Model with an Application to the Relationship between Financial Development and Green Growth

Abstract This paper studies computation, estimation, inference and specification testing in dynamic panel threshold spatial Durbin (DPTSD) model with multiple thresholds. We first develop a new Markov chain Monte Carlo (MCMC) based algorithm to jointly estimate the threshold parameters and simultaneously construct the confidence intervals for the parameters, after suggesting a within-group spatial two-stage least squares estimator. We then construct test statistics for threshold effect and the number of thresholds. Monte Carlo experiments indicate that the proposed estimator and tests have desired performance in finite samples. We finally apply the DPTSD model to investigate the relationship between financial development and green growth, and find that the empirical results based on the DPTSD model are quite different from these based on the dynamic panel threshold model.