COMPSTAT 2022: Start Registration
View Submission - COMPSTAT2022
Title: Time-varying multilayer networks in a Bayesian spatial autoregressive model Authors:  Michele Costola - Ca' Foscari University of Venice (Italy) [presenting]
Matteo Iacopini - Vrije Universiteit Amsterdam (Netherlands)
Casper Wichers - Vrije Universiteit Amsterdam (Netherlands)
Abstract: The impact of political relationships on financial stock markets is investigated. Political relationships among countries are inherently dynamic and characterized by multiple types (positive or negative) and varying intensity (strong or weak). To account for these data features, we propose a new spatial autoregressive model (SAR) by introducing time-varying multilayer networks. The model also incorporates country-specific network exposure and stochastic volatility to account for known stylized facts on financial data, together with a layer-specific parameter to capture the relative importance of each layer. We adopt a Bayesian approach to inference and design a new Markov Chain Monte Carlo algorithm based on Metropolis-Hastings and slice sampling to draw from the joint posterior distribution. The proposed method is used to investigate the impact of international political relationships among the top-GDP countries on their stock index returns. We find evidence of different network impacts across layers and countries.