Conditional dynamics and the multi-horizon risk-return trade-off

We propose testing asset-pricing models using multi-horizon returns (MHR). MHR effectively generate a new set of test assets that are endogenous to the model and that identify a broad set of possible conditional misspecifications. We apply MHR-based testing to prominent linear factor models and show that these models typically do a poor job of pricing longer-horizon returns, with pricing errors of similar magnitude as the risk premiums they were designed to explain. We trace the errors to the conditional factor dynamics. Explicitly incorporating factor timing in the models often makes mispricing worse, posing a challenge for future research.