Experimental Evidence on Portfolio Size and Diversification: Your Mileage May Vary ... A Lot

Abstract

This paper reports on the results of an experiment in which MBA student participants select securities at random for the purpose of reproducing the familiar exponentially declining relationship between portfolio volatility and number of securities. We find that the overall set of participants reproduces the pattern, but there is considerable variation within the participants. Somewhat similar results are even achieved with randomly selected securities, but the exponential pattern is observed more frequently. We examine potential differences in how humans ostensibly select random securities from how a random process selects securities. We find that the human participants tend to select larger companies with lower risk and that as a marginal security is added, that security is more correlated with the existing portfolio than when all securities are selected randomly. These results suggest that the process of achieving diversification when human beings select securities apparently at random is hindered by biases. It also suggests that there is wide variation from participant to participant, and therefore from investor to investor, in the relationship between diversification and the number of securities.

Keywords: diversification, portfolio size, number of securities, experimental finance

JEL Classification: G11