and

Abstract: Dynamic portfolio choice crucially depends on the predictability of re-turns. The existence and extent of return predictability is contested however. We consider a robust investor who hedges the adverse effects of uncertainty about pre-dictability but also exploits the benefits of predictability insofar it is significant. We show that robust dynamic portfolio choice does not exhibit the extreme risky invest-ment of traditional and Bayesian portfolios, is less volatile and features inter temporal hedging demands. We also identify the worst plausible form of predictability for the robust portfolio result. This worst case depends on the horizon, initial state, the port-folio, and the relations among the predictor variables. We argue that these worst case forms should be the focus of econometric research as their exclusion from plausibility promptly improves expected performance