Pricing errors and estimates of risk premia in factor models

In this paper, we consider the effect of the theoretical pricing error in the arbitrage pricing model on estimates of risk premia implied by the model. Under arbitrage pricing, the pricing error satisfies a strong bounding condition where for an infinite set of assets, the sum of squared pricing errors is bounded. We characterize the pricing error in terms of orders of probability and estimate an expected returns model which allows for pricing errors less than order one in probability. The principal finding of the paper is that misspecification of the pricing error and misspecification of the factor structure has no effect on the bias or mean squared error of the dominant risk premium. This implies that an exact form of arbitrage pricing can be used to estimate risk premia.