Estimating unrestricted spatial interdependence in panel spatial autoregressive models with latent common factors

TL;DR

This paper introduces a Bayesian method for estimating panel spatial autoregressive models with latent factors, allowing flexible spatial linkages without prior structure assumptions, especially effective when cross-sectional units outnumber time periods.

Contribution

The paper presents a novel Bayesian estimation approach for panel spatial autoregressive models with latent factors, accommodating large N and small T without pre-specified spatial structures.

Findings

Method is computationally fast and accurate in simulations.

Estimated spatial weights and factors closely match true values.

Reveals country clusters and limited influence of latent shocks in EU GVA growth.

Abstract

We develop a new Bayesian approach to estimating panel spatial autoregressive models with a known number of latent common factors, where N, the number of cross-sectional units, is much larger than T, the number of time periods. Without imposing any a priori structures on the spatial linkages between variables, we let the data speak for themselves. Extensive Monte Carlo studies show that our method is super-fast and our estimated spatial weights matrices and common factors strongly resemble their true counterparts. As an illustration, we examine the spatial interdependence of regional gross value added (GVA) growth rates across the European Union (EU). In addition to revealing the clear presence of predominant country-level clusters, our results indicate that only a small portion of the variation in the data is explained by the latent shocks that are uncorrelated with the explanatory variables.