Comparative Credit Risk Assessment: Integrating Autoregressive and Bayesian Models

This work proposes an alternative approach to modeling Loss Given Default, which is one of the credit risk parameters used in calculating Basel capital requirements. Baseline linear autoregressive models are developed and then challenged by models that combine Bootstrap techniques with Bayesian probability theory. To do this, time series are used; a database of recovery rates as the variable to be explained and a database of economic (e.g., GDP) and financial variables (e.g., 3-month EURIBOR) as candidate explanatory variables to account for the levels of recovery rates observed over time.