Credit Risk [18]

Go to: Summary | Previous | Next   
Bullet points include: Given ratings transition matrix, Pi, and values of exposures conditional on future ratings, we have univariate distributions of exposure values for a single name (issuer / obligor). To infer distribution of total portfolio value requires correlations (or more generally joint distribution) of ratings changes for the different exposures. Commonly use ordered probit approach to introduce correlations. For k’th exposure assume latent random variable Rk drives its one-period ratings transitions. Standardise mean and variance of Rk (to 0 and 1). For k’th exposure rated i at time 0 rating at time 1 is j if Zi,j-1 < Rk < Zi,j

Contents | Prev | Next | ERM Lecture Series

Desktop view | Switch to Mobile