An experimental analysis of the disposition effect: Who and when?

Abstract

The disposition effect (DE) is a common bias by which investors tend to sell winning assets too soon and hold losing assets too long. We complement the existing evidence in three directions. First, we check whether the DE is robust to realistic features such as transaction costs and competitive payment schemes. Second, by using a gender-balanced design, we check for gender differences. Third, we search for psychological correlates of the DE. We find that the DE is positive and significant in all our treatments. We do not find significant differences across treatments, although transaction costs significantly reduce the propensity to sell both winners and losers. We find somewhat larger DE in women, but this effect is only significant in the second half of the experiment. On the other hand, women are more reluctant to sell losing assets throughout the experiment. Finally, we find that the most significant psychological predictors of the DE are difficulty recognizing one's mistakes and optimism. Subjects scoring high in these traits are less likely to sell at a loss and therefore exhibit a larger DE. Our results provide further suggestive evidence of cognitive dissonance as an important determinant of the DE.