An Econometric Model of Credit Spreads with Rebalancing, ARCH and Jump Effects

Abstract

In this paper, we examine the dynamic behavior of credit spreads on corporate bond portfolios. We propose an econometric model of credit spreads that incorporates portfolio rebalancing, the near unit root property of spreads, the autocorrelation in spread changes, the ARCH conditional heteroscedasticity, jumps, and lagged market factors. In particular, our model is the first that takes into account explicitly the impact of rebalancing and yields estimates of the absorbing bounds on credit spreads induced by such rebalancing. We apply our model to nine Merrill Lynch daily series of option-adjusted spreads with ratings from AAA to C for the period January 1997 through August 2002. We find no evidence of mean reversion in these credit-spread series over our sample period. However, we find ample evidence of both the ARCH effect and jumps in the data especially in the investment-grade credit spread indices. Incorporating jumps into the ARCH type conditional variance results in significant improvements in model diagnostic tests. We also find that while log spread variations depend on both the lagged Russell 2000 index return and lagged changes in the slope of the yield curve, the time-varying jump intensity of log credit spreads is correlated with the lagged stock market volatility. Finally, our results indicate the ARCH-jump specification outperforms the ARCH specification in the out-of-sample, one-step-ahead forecast of credit spreads.