Tradable Factor Risk Premia and Oracle Tests of Asset Pricing Models.

We isolate the core economic assumptions that lead to misspecification and identification failures in conventional approaches to estimating factor risk premia. To overcome these challenges, we introduce Tradable Factor Risk Premia (TFRP), defined as the negative covariance between a factor and the projection of the stochastic discount factor onto the span of asset returns. TFRP are point-identified, invariant to the presence of other factors, and admit a natural interpretation via mimicking portfolios. They also facilitate Oracle estimation and inference -- behaving as if weak or irrelevant factors were known in advance. Empirically, we conduct the first large-scale investigation of the identification problem across the full factor zoo, demonstrating that identification failure is both pervasive and consequential.