Forecastable Default Risk Premia and Innovations
Traichal, Patrick A., Johnson, Steve A., Journal of Economics and Finance
We examine the generating process for default risk premia in short-term and long-term debt sectors of the U.S. economy over the recent period of January 1977 through December 1996. Using weekly aggregates reported by the Federal Reserve, this research finds that all univariate series examined have unit roots which suggests a "long memory" for innovations in the series. Models presented here demonstrate an adjusted R^sup 2^ of 59 to 97 percent in sample with similar hold-out sample results. Long-term investment grade corporate bonds exhibit substantial feedback such that lagged innovations have predictive power for nearby risk classes. In the money markets, the flow of information is from less risky assets to more risky assets. (JEL E43, E47)
Modeling risk premia is important from both an academic and practitioner perspective. Academically, trends are an important part of the business cycle decomposition used to measure economic activity (Beveridge and Nelson 1981). Bernanke (1990) uses various interest rate spreads to predict macroeconomic activity and concludes that the spread of commercial paper over T-bills is a reasonable tool. Keirn and Stambaugh (1986), in closely related work, find that small stock price levels significantly predict future bond returns. Commercial econometric modeling is often interested in forecasting rates, and the risk premium is a component of overall nominal rates (Fullerton 1993). As pointed out by Brenner, Harjes, and Kroner (1996), accurate forecasts of interest rate levels and volatility are crucial to pricing long-dated derivatives and effective hedging. Better predictive models for default risk premia can lead to better nominal rate prediction. Indirectly, examination of default risk premia provides evidence of the value of credit standing by documenting the interest cost associated with changing standing.
This paper seeks to address two distinct questions concerning modeling the underlying generating process for default risk premia. First, do the default risk premia of debt securities follow predictable trends such that a parsimonious model is useful in forecasting? Second, does knowledge of premia innovations in other risk classes help with predictability of same maturity risk classes? To answer these questions, this research examines the generating process of default risk premia at both ends of the maturity spectrum for the recent period of January 1977 through December 1996. This research extends the existing literature on default risk premia and interest rate modeling through the use of a robust time-series estimator, higher frequency (weekly) data, and by examining a larger set of debt classes.
Review of Literature
In one of the earliest attempts to explain the default risk premia on corporate bonds, Fisher (1959) examined the market yields on 366 fixed-income securities at five points in time from 1927 until 1953. This cross-sectional analysis defined default risk as a function of the coefficient of variation of firm income, duration of firm operations, the equity/debt ratio of the firm, and number of bonds outstanding. One finding of the study was that the elasticity of the risk premium with respect to each of the four variables is relatively stable over time (p. 218). This early result, risk premium stability, suggests that meaningful interest rate forecasting exercises are possible.
Cyclical variation of default risk spreads on corporate bonds was the focus of the research by Jaffee (1975). The author examined the level and variability of risk spreads on four classifications of corporate bonds over the fifteen-year period (1954-1969). The explanatory variables employed to explain the variation in these risk spreads were a measure of consumer sentiment, the unemployment rate, the growth rate of aggregate corporate retained earnings, the growth rate of fixed capital investment, the growth in national output, and the interest rate on Baa-rated corporate bonds. …