On predicting the probability of default of credit organizations in the Russian Federation

Abstract

Objective: to form a model for predicting the default of credit organizations under the modern conditions of the banking sector functioning.Methods: unidimensional analysis of variance, regression analysis of binary choice models.Results: in the modern economy, the banking system stability largely affects not only the financial sector, but also the economic and investment climate in the country. Understanding of the banks’ influence on the economy necessitates the formation of appropriate effective forecasting systems that allow identifying problem banks before revoking their licenses is necessary. The existing methodology of the Bank of Russia is characterized by subjectivity and inaccuracy of assessment. The analysis of studies on predicting bank defaults showed various approaches to the methodology of assessing the probability of credit institutions’ bankruptcy, though they have a number of shortcomings. Based on the selection of key factors affecting the bank’s financial stability, the logistic regression model for predicting bankruptcy of banks was formed. The methodology proposed in this article includes five predictors, selected on the basis of the improved methodology for selecting logit regression variables, and complements the existing methodologies.Scientific novelty: a methodology for assessing the probability of commercial banks’ bankruptcy in the Russian Federation was developed, which includes five key predictors for assessing the bank’s financial stability: return on assets, unit weight of liquid assets in the balance sheet currency, unit weight of the loan portfolio in the balance sheet currency, share of loans to the real sector in the balance sheet currency, and share of long-term placements in the loan portfolio. The logistic regression model of binary choice proposed in the paper allows distinguishing financially stable credit organizations from problem banks with a forecasting horizon of five months and a classification accuracy of 88,33 %.Practical significance: the relatively high classification accuracy of the model allows its use by the Bank of Russia in controlling the credit organizations functioning, as well as directly by the credit organization’s management, in order to assess the organization’s financial stability and to predict the default probability, as well as to form the bank’s development strategy.