A Bayesian Multilevel Model with Time-Varying Spatial Autoregressive Coefficient

• In Event: Networks

Abstract

Spatial Modeling is one of the most widely-applied statistical tools for International Relations scholars to analyze interdependence of states and inferring their general level of connectivity. However, the existing spatial literature of spatial modeling all assume time-invariant spatial
autoregressive coefficient even when applied to Time-Series Cross-Sectional data. The assumption of constant general connectivity of the system is methodologically restrictive and substantively unrealistic. We relax this restriction and specify a spatial-time hierarchical
model allowing the magnitude and direction of the spatial autoregressive coefficient to change over time. We suggest several specification options for capturing the dynamic process of the spatial autoregressive coefficient, including local smoothing with a random walk process and a stationary autoregressive process. We develop simulation algorithms using Markov Chain Monte Carlo for model estimation. Finally, we demonstrate the methodological and substantive gains of the proposed methods by applying the model and MCMC algorithm to a study of the evolution of the WTO trade system as an international bargaining forum.

Authors

• Xun Pang, Tsinghua University

• Licheng Liu, Tsinghua University