Estimation of loan portfolio risk on the basis of Markov chain model

Abstract

A change of shares of credits portfolio is described by Markov chain with discrete time. A credit state is determined on as an accessory to some group of credits depending on presence of indebtedness and its terms. We use a model with discrete time and fix the system state through identical time intervals - once a month. It is obvious that the matrix of transitive probabilities is known incompletely. Various approaches to the matrix estimation are studied and methods of forecast the portfolio risk are proposed. The portfolio risk is set as a share of problematic loans. We propose a method to calculate necessary reserves on the base of the considered model. © 2013 IFIP International Federation for Information Processing.