Measuring Interconnectedness between Financial Institutions with Bayesian Time-Varying VARS

In this paper we propose a time-varying parameter framework to estimate the dynamic network of financial spillovers. In a series of simulation exercises, we show that our framework performs better than the classical approach based on Granger causality testing over rolling windows. We apply it to all financial stocks listed in the S&P 500 and uncover a gradual decrease in interconnectedness after the crisis, which is not observable using the rolling window approach. We show that this is because the rolling window results are highly sensitive to crisis observations.