We examine the asymptotic efficiency of using individual stocks or portfolios as base assets to test cross-sectional asset pricing models. The literature has argued that creating portfolios reduces idiosyncratic volatility and enables factor loadings, and consequently risk premia, to be estimated more precisely. We show analytically and find empirically that the more efficient estimates of betas from creating portfolios do not lead to lower asymptotic variances of factor risk premia estimates. Instead, the standard errors of factor risk premia estimates are determined by the cross-sectional distribution of factor loadings and residual risk. Creating portfolios shrinks the dispersion of betas and leads to higher asymptotic standard errors of risk premia estimates.