Implied Severity Density Estimation: An Extended Semiparametric Method to Compute Credit Value at Risk

Abstract

This paper focuses on estimating implied severity, which does not rely on historical data and can be used especially for low default companies. We perform an extended semiparametric estimation method based on a mixture start to estimate it. We carry out an empirical analysis and our results show that our method allows us to capture the observed multimodal behaviour of severity better than the commonly used single beta distribution assumption. Futhermore, we highlight the relevance of this modeling approach by focusing on its role for credit VaR.