Do constraints improve portfolio performance?
The discrete-time dynamic investment model, using only historical data in various asset-allocation settings, often produces significant abnormal returns. However, the model does not choose the diversified portfolios that theory suggests it should. Therefore, in this paper, we compare the investment policies and returns of the model with and without constraints on the mix of risky assets. The constraints lead to appreciably more diversification and less realized risk, but only at the cost of less realized return. Visual comparisons of compound return—standard deviation plots and statistical comparisons of Jensen’s alpha suggest that the reduction in return is not worth the reduction in risk. For more risk-averse investors, ex post utility and certainty equivalent returns suggest that it is. The results, however, illustrate the problems associated with using ex post utility to measure performance.