ID : 1034
Short sales  ; Portfolio investments ; Management science ; Forecasting ; Estimating techniques ; Comparative studies ; Bayesian analysis
Forecasting the mean returns vector and the covariance matrix is a key feature in implementing portfolio theory.  The performance of the Bayes-Stein method for forecasting these parameters for use in the Markowitz model (with and without short sales) is compared with 7 other estimation methods, and 3 alternative portfolio selection techniques.  This analysis represents the first large-scale empirical investigation of the usefulness of the Bayes-Stein approach using historical data.  The data are drawn from the London Stock Exchange.  In contrast to earlier studies, the relative performance of the Bayes-Stein is mixed.  While it produces reasonable estimates of the mean returns vector, there are superior methods, such as overall mean, for estimating the covariance matrix when short sales are permitted.  When short sales are prohibited, actual portfolio performance is clearly improved, although there is little to choose between the various estimation methods.
Estimation methods in portfolio selection and the effectiveness of short sales restrictions: UK evidence, Board, John L G; Sutcliffe, Charles M S, Management Science, 40:4, Apr 1994
*****
