Beyond Basel2: Modeling loss given default through survival analysis

In the last years the majority of European Banking Groups has chosen to adopt the advance status under Basel2. This has required banks to develop statistical models for estimating Probability of Default, Loss Given Default and Exposure at Default, within a horizon time of 1 year. Such models make no attempt at describing the exact timing of default, in particular, beside an extensive academic and practitioner’s literature on PD, LGD studies are in a less advance status. One of the main reasons could be due to the difficulties in modeling and forecasting the danger rates. The aim of this paper is to show the results of the first application on an Italian Bank Retail portfolio of survival analysis technique for estimating LGD, by modeling the danger rate. Two issues arise from the forecasting of danger rates: dealing positions that change, or not, their status towards charge off and obtaining a certain level of accuracy across time, thus resulting more difficult than in simpler classification methods. This paper analyzes the use of a parametric survival model, where time is assumed to follow some distribution whose PDF can be expressed in terms of unknown parameters: hazard and shape.