Asset pricing models: Implications for expected returns and portfolio selection

When a risk factor is missing from an asset pricing model, the resulting mi spricing is embedded within the residual covariance matrix. Exploiting this phenomenon leads to expected return estimates that are more stable and pre cise than estimates delivered by standard methods. Portfolio selection can also be improved. At an extreme, optimal portfolio weights are proportional to expected returns when no factors are observable. We find that such port folios perform well in simulations and in out-of-sample comparisons.