Research Summary
Research Summary
An Uncomfortable Predictability Paradox
Description
In predictive regressions, we test the null hypothesis that a predictor has no information about expected returns, i.e. beta equals zero. However, the literature neglects to recognize that we are testing a joint hypothesis. The maintained hypothesis-predictability exists or it does not-matters. If we maintain that predictability exists, a null that beta equals zero implies that the error term inherits the persistence properties of expected returns. After adjusting the standard errors accordingly, I find that every study claiming statistical evidence for predictability is overturned. Further, I show that we will not have a sufficiently long time series to reliably claim predictive power for any of the posited predictors until the year 2104. The paradox is that if we posit a world without predictability, then there is strong evidence contradicting that hypothesis. However, if we assume predictability does exist, then finding predictors becomes a fool's errand-at least for now.