The Markowitz portfolio model requires estimates of the rate of return and variance in return of the securities which are available to form the portfolio. The estimate of the return has been conventionally calculated from the mean of past returns. Stein in 1955 suggested that the mean as an estimator was in admissible when several values were being simultaneously estimated. This study applies the James and Stein estimation method to the determination of the expected rates of return for an internationally diversified portfolio.
The estimate of the variance is subject to error because of serial correlation in the national capital market index. A method to correct the international market covariance matrix is investigated.
Portfolios are formed using data collected from. 12 countries covering the national capital market index for the period 1975-86. The performance of ex ante generated portfolios using different strategies is compared on the basis of their Sharpe performance measure and the stochastic dominance of the compound return.
It was found that portfolios formed using James- Stein estimation outperformed those using the conventional mean. Portfolios formed using a minimum variance criteria outperformed the traditionally formed, the equally weighted, and the James-Stein portfolios. Analysis showed that end of month market indices had low serial covariance, and correction did not improve the portfolio performance.