Making text count: Identifying systemic risk spillover channels in the Chinese banking sector using annual reports text

Abstract

Using the annual report text information of Chinese listed banks from 2007 to 2020, this paper constructs multiple kinds of spatial weight matrices from the perspective of business similarity to identify the systemic risk spillover channels. Furthermore, the Bayesian posterior probability methodology proposed by Debarsy and LeSage (2018) is employed to assess the relative importance of each spillover channel. Besides, we discuss the risk spillovers of specific factors. The empirical results show that the loan type, loan region, investment industry, and income structure are all effective risk spillover channels, and the loan type channel is of the utmost importance. And also the spillover effects of bank-specific factors are identified and continuous. Our results are validated by robust analysis.