Multilayer Interdependencies in the Banking System of Japan: Correlation Dynamics and Determinants

Abstract

In this paper, we analyse the multilayer architecture of the correlation dynamics among Japanese banks over the period from 1983 to 2012. This period encompasses nearly all significant events in the Japanese economy over the past few decades. The two layers we consider represent the loan portfolio correlations stemming from the credit market and the stock return correlations arising from the equity market, respectively. We find that in these different market layers, the correlations can have different structures and they may imply different signals of systematic and non-systematic risk components in the banking system. In addition, our results show that in spite of a certain degree of persistence, the interactions among banks in each layer also evolve over time. Furthermore, utilizing the characteristics and the balance sheet information of individual banks, we find that the similarities in this kind of fundamental information can only partially explain the formation of correlations in each layer. In addition, overall, no explanatory variables based on quantitative or qualitative similarities can predict future correlations better than contemporaneous ones.

Keywords: Loan Correlations, Stock Return Correlations, Systemic Risk, Principal Component Analysis, Correlation Filtered Methods, Determinants of Correlations