The FSA published the English version of "FSA Analytical Notes (2025.6) vol.2 ". 
full text
As financial institutions’ business environments and profit structures change, it is important to understand economic and market trends based on data, and to accurately grasp the business conditions of individual financial institutions and also the resilience and vulnerabilities of the financial system as a whole. From this perspective, the FSA has been focusing on the utilization of granular data, such as transaction-level bank loan data and financial data on individual corporations. Some case examples of data analyses using such granular data are published as a series of reports titled “FSA Analytical Notes.”
This issue contains two analyses on: 1) Analysis on Credit Risk Management Practices in Regional Banks - Investigation on Credit Risk Mitigation for Loans and Study on Rating Transition Prediction Models -, and 2) Understanding the Utilization of Credit Guarantee System.
This paper presents an analysis of credit risk mitigation (loan coverage) and an examination of a rating transition prediction model, using loan-by-loan data collected through the Common Data Platform. While this paper does not assess the appropriateness of loan coverage-given that it should vary depending on factors such as borrower size, characteristics, and purposes of funds-the analysis revealed a tendency for lower coverage ratios particularly among shared borrowers (borrowers with loans from multiple banks) and prefecture-wise cross-border loans (loans extended to borrowers outside the bank’s home region). In the verification of the rating transition prediction model, it was suggested that the model predicting downgrades from “needs attention or above” to “in danger of default or below” performs with relatively high accuracy using financial information alone, compared to models predicting other transition patterns.
This paper conducted an analysis of the utilization of the Ccredit Gguarantee Ssystem using loan-level data collected through the Common Data Platform. A machine learning approach was employed to identify the key factors influencing whether a loan is guaranteed. Among borrower-related factors, sales and capital ratio were found to have relatively large effects-borrowers with higher values for these indicators were less likely to utilize credit guarantees. While this analysis does not aim to assess the appropriateness of credit guarantee usage-given that such usage varies depending on borrower characteristics and other various factors- it did reveal that the tendency to use guarantees differs significantly based on whether a borrower is in excess liabilities, and difference across industries are also observed.
Enhancing the use of data in financial supervision and policy-making is a medium- to long-term agenda. The FSA will continue to build its data analysis capabilities and data infrastructure.
* Unless otherwise noted, the figures and tables in this report were prepared by the FSA.
