Researchers in finance very often rely on highly persistent - nearly integrated - explanatory variables to predict returns. This paper proposes to stand up to the usual problem of persistent regressor bias, by detrending the highly auto-correlated predictors. We find that the statistical...