Horses for courses: Mean-variance for asset allocation and 1/N for stock selection

For various organizational reasons, large investors typically split their portfolio decision into two stages - asset allocation and stock selection. We hypothesise that mean-variance models are superior to equal weighting for allocation, while the reverse applies selection, as estimation errors less of a problem when used than confirm this hypothesis US data using Bayes-Stein with no short sales variance based constraints. Robustness checks four other types model (Black-Litterman three different reference portfolios, minimum variance, Bayes diffuse prior Markowitz), wide range parameter settings support our conclusions. also replicate core results Japanese data, additional replications Fama-French 5, 10, 12 17 industry portfolios equities from seven countries. In contrast previous results, but consistent empirical we show analytically superiority over 1/N is increased assets have lower cross-sectional idiosyncratic volatility, which in simulation analysis calibrated data.