Estimating Credit Spread Risk Using Extreme Value Theory

Abstract

In 1998, many fixed-income investors grossly underestimated the extent of credit spread risk. The main reasons were a failure to take heavy tails into account when estimating risk, and a failure to incorporate a sufficiently long history of credit spreads into the statistic estimation process. In this article, the author provides a non-technical presentation that shows how longer-term daily data, combined with extreme value analysis, can be used to generate more accurate and more robust quintile estimates for credit spread shifts.