Bank risk aggregation based on dual perspectives of bank managers and credit rating agencies

Abstract

In this paper, the risk disclosure text information from the dual perspectives of bank managers and credit rating agencies is innovatively incorporated into the bank risk aggregation, which to some extent avoids the deviation of bank risk assessment caused by the risk perception from a single perspective, and measures the overall bank risk more comprehensively. In the empirical analysis part, this paper extracted 342 risk profit and loss data, 511 text risk information in 10-K forms and 356 available bank rating reports from 134 listed commercial banks in the United States. The Sent-LDA model was used to identify and compare risk factors from different perspectives. In addition, by constructing a total risk-adjusted index, this paper studies the effectiveness of textual risk disclosure from the dual perspectives of bank management and credit rating agencies in the overall risk measurement of commercial banks. The empirical results show that the text contains the incremental information of risk assessment, and the inclusion of double-perspective text risk information into the risk aggregation will amplify or reduce the overall risk of banks calculated only based on financial numerical data. In practice, it will be more effective to measure the overall risk of banks by considering the textual risk disclosure from dual perspectives.