OPTIMUM VOLUME OF BANK RESERVE: FORECASTING OF OVERDUE CREDIT INDEBTEDNESS USING COPULA MODELS

Abstract

The article propose to consider the possibility of RLUF-copulas application for the creation of joint distributions of overdue credit indebtedness ranks with macroeconomic indicators for the purpose of indebtedness forecasting and also for the definition of optimum volumes of reserve requirements for the corresponding losses. In this research the comparative analysis of multivariate distributions of RLUF-copula estimation with such classical copulas, as FGM-copula, Frank's copula and Gauss's copula is made. In the article the method of maximum likelihood is used for receiving estimates of model parameters. In case of RLUF-copula Bayesian estimates of parameters are received using the Metropolis algorithm with random volatility. Forecasting of bank reserve volumes for all received models is executed in the form of random sample generation by the means of the algorithm of acceptance-deviation for the creation of the corresponding sample of joint distribution using the copula density function. As the result of playing of hundred possible scenarios of indebtedness volumes is obtained the 95 % confidence level for the possible volume of credit indebtedness which can fully act as the optimum volume of reserve requirements for the corresponding credit losses.

Keywords

• банковский резерв
• кредитный риск
• просроченная кредитная задолженность
• копулярная модель
• прогнозирование
• bank reserve
• credit risk
• overdue credit indebtedness
• copula model
• forecasting