This paper examines time-series forecast errors of expected returns from conditional and unconditional asset pricing models for portfolio and individual firm equity returns. A new result that increases predictive precision concerning model specification and forecasting is introduced. Conditional versions of the models generally produce higher mean squared errors than unconditional versions for step ahead prediction. This holds for individual firm data when the instruments are firm specific. Mean square forecast error decompositions indicate that the asset pricing models produce relatively unbiased predictions, but the variance is severe enough to ruin the step ahead predictive ability beyond that of a constant benchmark.

This paper studies the estimation of asset pricing model regressions with conditional alphas and betas, focusing on the joint effects of data snooping and spurious regression. We find that the regressions are reasonably well specified for conditional betas, even in settings where simple predictive regressions are severely biased. However, there are biases in estimates of the conditional alphas. When time-varying alphas are suppressed and only time-varying betas are considered, the betas become biased. Previous studies overstate the significance of time-varying alphas.