Loss given default (LGD) is a proportion of a credit exposure that is lost if the obligor defaults on a loan. Response variable LGD contains values between 0 and 1 including both 0 and 1, where 0 means that the balance is fully recovered while 1 means total loss of exposure at default. This article addresses two alternative semi parametric approaches for modelling loss given default, which is measured on the interval [0,1]. The class of models are very flexible and can accommodate skewness and bimodal characteristics of LGD data. The dependence of the predictors of each of the parameters (of the proposed model distribution for LGD) on explanatory variables can be additive P-splines, regression trees or neural network models. The proposed models are applied to a loss given default data set and compared with current popular models.
|Numero di pagine||6|
|Stato di pubblicazione||Published - 2016|