Forecasting U.S. Bond Returns:
A Practitioner’s Perspective

Abstract

This article presents a practitioner’s view of the evidence on the predictability of U.S. bond returns, using forward rates found in the academic literature. The author examines this evidence from a dual perspective: statistical and economic. He finds that the regressions of monthly returns of U.S. Treasury futures on the set of three forward rates, with one-, five- and ten-year expirations, are statistically viable. An implementation of these forecasts as an investment strategy shows that it is historically profitable on a risk-adjusted basis. Furthermore, the author demonstrates the importance of the dual approach to assessing predictability, by providing an example of a slightly modified version of the main model that is more intuitive, more parsimonious, and statistically more robust, yet it fails to exhibit better economic performance.

TOPICS: Exchanges/markets/clearinghouses, factor-based models, statistical methods