Disentangling Persistence from Predictability in Asset Pricing

We find persistent predictors known to bias predictive regressions for stock re- turns to matter also for asset pricing models. For example, deciding whether hedge funds offer an expansion of the investment opportunity set depends, among other things, on the persistence levels of predictors used to create managed returns. Us- ing simulations to disentangle the effects of persistence from predictability, we find highly persistent predictors to bias asset pricing models and tests even if managed portfolios and conditioning information are used optimally. Our framework enables us to construct tests that are robust in the presence of persistent predictors, and we find it to be more difficult to construct such robust tests for linear than for the non-linear ways of utilizing conditioning information.

• View at SSRN