Exposure to Default: Estimation for a Credit Card Portfolio

Abstract:

This work estimates the exposure at default without using the credit conversion factor, a common mechanism used in the expected loss estimation literature and suggested by the Basel Committee. To achieve this objective, the probability distribution of this variable (exposure at default) has been identified, which is subsequently estimated in parts (EAD = 0 and EAD > 0) using generalized linear models (logit and GLM-Gamma). The results obtained are competitive with those found in the literature. This shows that the simultaneous estimation of parameters, as well as the separate estimation, give promising results. Additionally, the EAD > 0 case is contrasted with a MARS model whose performance is superior to GLM-Gamma. These models were applied to a data set of a credit card portfolio of a financial institution in Ecuador.

Año de publicación:

2022

Keywords:

• Cbkp_redit risk
• Machine learning
• Expected loss
• Exposure at default
• Generalized linear models
• Gamma distribution
• CREDIT RISK
• Machine Learning

Fuente:

scopus