Connectedness, Structure, and Performance of Different Financial Networks: Evidence from Financial Institutions in China

Abstract

Financial networks are powerful tools for studying risk contagion and multiple financial networks with unique constructed ways are available. Based on stock returns from financial firms in China, we compare six commonly used financial networks to see whether they can reach similar conclusions for risk contagion. Generally, we find: different networks show similarity around the conditional mean or median and differences in the tails; non-data-driven networks underperform evaluating by the quantile goodness of fit; two LASSO networks show advantages in some aspects. Therefore, financial networks should be carefully selected when studying risk contagion, or else unrobust conclusions may be drawn.

Keywords: Dynamic Network Quantile Regression Model, Quantile Connectedness, Risk Contagion, Systemic Risk Network, Tail Dependence