MCMC-based credit rating aggregation algorithm to tackle data insufficiency

Abstract

This paper investigates how credit rating aggregation might lead to a more efficient estimation of key portfolio risk management metrics: expected credit losses (ECL) and risk‐weighted assets (RWA). The proposed technique for credit rating aggregation is based on the Markov Chain Monte‐Carlo methodology and leads to a statistically smaller variance of ECL and RWA than the naïve and distribution‐based alternatives. This conclusion holds for three public datasets and four simulated studies. The paper results might be helpful for portfolios that suffer from data insufficiency or rely on external ratings for credit risk assessment: portfolios of international companies, interbank loans, and sovereign debt.