Forward-Looking Factor Models of Partial Default Correlations: Identification of Systemic Importance

Abstract

Using historic PD (probability of default) and POE (probability of other exits) as raw data, this paper has three primary goals. First, we simulate 12-month PDs for constituents of five particular industry communities using a revised version of the forward-looking default correlation model in Duan and Miao (2015). Secondly, we analyze industry networks as signaled by the simulated PDs using a methodology proposed by Lau et al. (2016), which only focuses on financial institutions. Finally, based on the constructed networks, we discuss the systemic importance of leading economies in the five sectors. Our results suggest that Japanese firms tend to be more connected than heavily capitalized, with the reverse asymmetry observed in their Chinese and American counterparts. This dynamic has become especially pronounced for firms in China since the financial crisis. Further, we observe that size does not imply higher connectivity. We also discuss the formation of relatively insulated domestic and bilateral communities in the five sectors.