Building on the Shanken (1992) estimator, we develop a new methodology for estimating and testing beta-pricing models when a large number of assets N is available but the number of time-series observations is small. We show empirically that our large N framework can change substantially common...