This paper attempts to improve the forecast performance of single variable predictive regressions used in the equity premium prediction literature through Bayesian priors derived from consumption-based asset pricing models. To implement these model-based priors, I develop a Bayesian procedure which is rooted in the macroeconometrics literature. The priors are derived from four asset pricing models: Habit Formation, Habit Formation Term Structure, Long Run Risk, and Prospect Theory. The model-based priors can substantially increase the explanatory power of the single variable predictive regressions. Further, they help to assess consumption-based asset pricing models in a novel way. ∗I thank Tarun Ramadorai and Kevin Sheppard for excellent supervision. I thank Andrew Patton, Narayan Naik, Dimitri Vayanos, Mungo Wilson, and seminar participants at the Säıd Business School and the Oxford-Man Institute of Quantitative Finance for useful comments. Financial support from the David Walton Memorial Fund is thankfully acknowledged. †Mathias Kruttli is at the Department of Economics, University of Oxford, and the Oxford-Man Institute of Quantitative Finance. Email: firstname.lastname@example.org.