Analyzing the Extreme Risk of a U.S Corporate Bond Portfolio, November 2009

Abstract

We use the Barra Extreme Risk (BxR) model to analyze a US dollar-denominated corporate bond portfolio consisting of 2142 distinct issues. As in the case of equities, we find that the BxR proprietary extreme risk forecasts, xShortfall and xVaR, are higher than value-at-risk and expected-shortfall forecasts generated by a conditionally normal model. Further, the impact on xShortfall of tilting the portfolio toward high-yield bonds is materially greater than the impact on volatility, and the discrepancy increases as quality declines. As a result, increasing the weight on investment-grade bonds while lowering the weight on high-yield bonds mitigates tail risk more than it mitigates volatility. This intuitive result reflects the high degree of sensitivity of high-yield bonds to extreme events.

Keywords: Barra, Extreme Risk, BxR Model, Corporate Bond Portfolio, Forecast, xShortfall, xVaR