Abstract: | In this paper I examine the out-of-sample performance of five mean-variance strategies using different models of expected returns within a U.K. industry asset-allocation framework between January 1970 and December 1991. The performance of the five strategies is evaluated with different measures. I find superior performance for the strategy that uses conditioning information to estimate expected returns. This consistently outperforms the two passive benchmarks and earns positive abnormal returns over the sample period regardless of how frequently the portfolio is revised and whether portfolio restrictions are imposed. |