Previous literature indicates that stock returns are predictable by several strongly autocorrelated forecasting variables, especially at longer horizons. It is suggested that this finding is spurious and follows from a neglected near unit root problem. Instead of the commonly used t-test, we propose a test that can be considered as a general test of whether the return can be predicted by any highly persistent variable. Using this test, no predictability is found for U.S. stock return data from the period 1928-1996. Simulation experiments show that the standard t-test clearly overrejects whereas our proposed test controls size much better.

The ability of yield spreads to predict changes in long-term interest rates implied by the expectations hypothesis is usually rejected. It is suggested that this rejection is often caused by high persistence in the spread when standard inference is employed. Instead, the asymptotically valid method of Cavanagh et al. (1995) is applied to monthly U.S. data from 1952:1-1991:2. The persistence of the spreads seems to have varied over time, and in subsample analysis, the expectations hypothesis cannot be rejected at the long end of the maturity spectrum.