PDF Ebook Default correlation: empirical evidence
A lot of work has been carried out over the last years regarding the diversification effect within a portfolio of credit instruments. Most of this research has considered correlation as a good proxy for dependence. The purpose of this paper is not to question this choice but rather to provide some empirical results on the behaviour of correlation.
Several methodologies have been developed to proxy for transition correlation, but very little has yet been done to benchmark their output with empirical findings. In this paper, we investigate the properties of default correlation using Standard & Poor’s historical database.
We first review various ways of estimating bivariate transition probabilities and correlations from actual default data, without relying on any specific default generating model. We study the performance of these measures on a fictitious bond portfolio and show that no single measure outperforms the others for all levels of correlation and sample size. In the remainder of the paper, we use the measure we think best suited to the size of our database to calculate empirical bivariate transition and correlation matrices per class of risk and per industry.
We then test the hypothesis that default correlation can efficiently be extracted from equity return correlation. Indeed over the past few years, it has become market practice to use a factor model of credit risk relying on equity correlation as proxy for asset correlation. This approach assumes that latent factors (such as firms’ assets) drive the individual obligors’ probabilities of default and, ultimately, default correlations among obligors.
Posted in :
