Portfolio Size Revisited

Abstract

Using a sophisticated sampling technique, the authors compare randomly constructed stock portfolios with portfolios using the underlying population and evaluate them with 18 different measures. The randomization included portfolio size and portfolio start date to eliminate timing bias from the analysis. By comparing the 18 statistics across the portfolios, average portfolio sizes to reproduce the population characteristics were computed. The optimal portfolio size depended greatly on the criterion being used to judge the adequacy of diversification.

TOPICS: Security analysis and valuation, portfolio construction, performance measurement