Bayesian SAR Model with Stochastic Volatility and Multiple Time-Varying Weights

Abstract A novel spatial autoregressive model with time-varying structural variance for panels of time series data is introduced. It incorporates multilayer networks and accounts for dynamic relationships, thus enabling the analysis of shock propagation through time-varying spillover effects. The proposed method outperforms alternative spatial model benchmarks in an empirical application investigating the impact of cooperative and conflictual geopolitical relationships on G7 stock markets. The results indicate that cooperative interactions have a greater influence on stock markets than conflictual ones, highlighting the collaborative nature of the G7. They also reveal heterogeneous network exposures and distinct patterns of direct and indirect spillover effects.