Macroeconomic Announcement Effects on the Covariance Structure of Bond Returns

Abstract

This paper contains a multivariate analysis of the effects of macroeconomic news on the U.S. bond market. In particular, we consider releases of Employment Situation and Producer Price Index (PPI) reports (released monthly on pre-announced dates) and the excess returns of 2, 3, 5, 7, 10, and 30 year Treasury bonds using daily data from 1983 to 1998. We extend the work of Jones, Lamont and Lumsdaine (1998) who analyze macroeconomic announcement effects on bond returns in a univariate GARCH framework. Our main interest is the impact of macroeconomic news on the conditional variances, covariances, and correlations of bond returns. To this end, we apply the factor-ARCH model of Engle, Ng and Rothschild (1990) using the principal components as substitutes for the real common factors. One factor is shown to be adequate, accordingly, the first principal component is assumed to evolve according to a GARCH(1,1) model extended to account for differences between announcement and non-announcement days. In this way, we document that the conditional variance of the common factor is deterministically larger on announcement days by 164%. Moreover, announcement shocks are found to be merely transitory and do not tend to persist. Hence, macroeconomic news cannot explain the observed high degree of persistency of the bond market. In addition, we do not find any leverage effects of announcement shocks. The conditional variances of bond excess returns are significantly larger on announcement days and increase with maturity. Likewise, the conditional covariances are larger on announcement days and increase with the maturity of either bond. Finally, the conditional correlations are found to be larger on announcement days, i.e. the co-movement of bond excess returns is stronger on announcement days than on other days.