Estimation of Probability of Default in Low Default Portfolios

Abstract: Estimation of probability of default (PD) is a fundamental part of credit risk modeling, and estimation of PD in low default portfolios is a common issue for banks and financial institutions. The Basel Committee on Banking Supervision requires banks and financial institutions to add an additional margin of conservatism to its PD estimates in the case of insufficient data, as in low default portfolios with few default observations. In addition, the Basel regulations also require banks to report PD estimates on grade level. The purpose of this thesis is to study methods for estimation of probability of default in low default portfolios. In order to fulfill this purpose, two different models for estimation of probability of default in low default portfolios are considered. These are the Benjamin, Cathcart and Ryan (BCR) approach and a Bayesian approach. Because these models estimate PD on a portfolio level, different methods for allocation of portfolio PDs to rating grades are also considered. Lastly, methods to assign portfolio PDs to grade level for a portfolio consisting of several subportfolios are compared.