Embedded Predictor Selection for Default Risk Calculation: A Southeast Asian Industry Study

Default risk estimation is a core business of rating agencies. Banks and other financial institutions need to scale the default risk of their counterparties, identifying predictors that significantly contribute to default probability insight into fundamentals of credit risk analysis. Default analysis and predictor selection are two related issues, but many existing approaches address them separately. A unified procedure is employed, a regularization approach based on the GLM model, which allows simultaneously selecting the default predictors and optimizes all the parameters within the model. For this purpose Lasso and elastic-net penalty functions are employed as regularization terms. The methods are applied to predict default of companies from the industry sector in Southeast Asian countries. The relevant default predictors over the countries reveal that credit risk analysis is sample specific. The empirical result shows that the proposed method has a very high accuracy of default prediction.