Problems in the application of portfolio selection models

This paper discusses a number of problems which arise in the implementation of portfolio selection models. It is suggested that the effect of errors and biases in expected return forecasts can be reduced by using these forecasts to modify a prior distribution which leads to minimal trading activity. Efficient diversification across industry groups is hampered by the difficulties of predicting covariances. The use of a selection model through time raises the issues of revising forecasts and of the relationships upon which the appropriate investment horizon and portfolio turnover depend. Last, consideration is given to the conflicts which may exist between management objectives and the mean-variance criterion used by most models.