Paper available here.
Authors: William Perraudin and Siyi Zhou
This paper analyses the dynamics of credit ratings. We devise statistical techniques for estimating both intra- and inter-industry correlations between factors driving ratings changes. The analysis is performed unconditionally and then conditional on de-trended GDP.
The unconditional estimates we produce may be used in credit portfolio modelling. In this activity, common challenge are (i) to describe the dependence of ratings across multiple geographical regions and sectors in a parsimonious but convincing way and then (ii) to calibrate that dependence using data. The approaches described here involve use of tensor product factor correlation matrices.
The conditional estimates we generate may be used for macro stress testing in which the credit quality of a portfolio is simulated conditional on a hypothesised future path of real output. The techniques are also highly applicable to the modelling problems that banks currently face in implementing forecasts of their provisions as required by the new IFRS 9 accounting standard.
We illustrate our approaches by applying them to historical data on agency ratings. The sector-region factor structure we assume represents a realistically complex example, comparable to those employed by large international banks with global, multi-sector corporate credit exposures.