Analysis of financial contagion among economic sectors through Dynamic Bayesian Networks

Abstract

Crises severely impact economies and may spread across regions or sectors in a process called contagion. Understanding this process is crucial for anticipating crises’ effects and implementing mitigative strategies. Specific economic sectors may be major crisis propagators: banking and insurance are often considered decisive in this context. This study employs Dynamic Bayesian Networks to model sectoral interdependencies within the U.S. economy, utilizing daily data from nine Dow Jones industrial indices over the period 2000–2020. As a secondary objective, we evaluate whether the insurance industry plays a central role in spreading crises. Several crisis periods are analyzed, from dot-com bubble to Covid-19 pandemic. The results reveal the subprime crisis, European debt crisis and the 2016 presidential election as the main contagious periods. The last analyzed period – Covid-19 pandemic – was divided in two phases, showing, on phase 1, an interconnected economic system with three main spreaders (Oil & Gas, Real Estate and Pharmaceutical) and, on phase 2, the same configuration of the post-subprime. Furthermore, although the insurance sector was somehow relevant during subprime crisis, it primarily acts as a contagion receptor, specially from the banking sector, with the correlation between them being the highest of all, reaching 0.49 at the last period.