Tom Engsted

We document that over the period 1953-2011 US bond returns are predictable in expansionary periods but unpredictable during recessions. This result holds in both in-sample and out-of-sample analyses and using both univariate regressions and combination forecasting techniques. A simulation study shows that our tests have power to reject unpredictability in both expansions and recessions. To judge the economic significance of the results we compute utility gains for a meanvariance investor who takes the predictability patterns into account and show that utility gains are positive in expansions but negative in recessions. The results are also consistent with tests showing that the expectations hypothesis of the term structure holds in recessions but not in expansions. However, the results for bonds are in sharp contrast to results for stocks showing that stock returns are predictable in recessions but not in expansions. Thus, our results indicate that there is not a common predictive pattern of stock and bond returns associated with the state of the economy.