Credit loss modelling using beta distribution in a Bayesian approach

The Advanced Internal Rating Based (AIRB) approach is more and more frequently applied by banks. Bank analysts decide to use their own approach to calculate basic risk parameters such as Probability of Default (PD), Exposure at Default (EAD), and Loss Given Default (LGD). The problem of small samples in LGD estimation is always a challenge for researchers and analytics. The paper proposes the basic LGD model based on splitting recoveries into two classes of recoveries: close to 0 or close to 1, and based on that split the construction of the LGD model with the combination of two binary models. The main advantage of the paper is, however, addressing the unresolved cases incorporated in the LGD estimation process by using a Bayesian approach which assumes a beta distribution of further recoveries for unresolved cases. An additional advantage of the paper is that the proposed modelling approach for LGD is illustrated on real data for mortgage loans for one of the European banks.