Experience-Based Learning in Repeated Portfolio Decisions in the Laboratory

12.12.2019 12:00 – 13:15

INSTITUTE OF MANAGEMENT / ABSTRACT

According to classical portfolio theory, people should take the average return, the variance and the correlation between assets into account when deciding how to allocate savings in a portfolio. Early experimental evidence showed that people are not very good in allocating efficient portfolios and in particular that they neglect correlations. With the recent interest in experience-based learning compared to information acquisition from descriptive statistics, we experimentally investigate whether people can improve portfolio allocation decisions via learning. In two experiments, people made 90 allocation decisions between two risky and one riskless asset with feedback. We varied the correlation between the two risky assets and found that people react sensibly to the manipulation. Although people are not constructing efficient portfolios, they get better with feedback. In particular, they decrease investment in the risk-free asset more in the negative correlation condition compared to the positive correlation condition over time, indicating that they understand the effect of the negative correlation on portfolio variability. Finally, we model behavior with a new reinforcement learning model that takes learning about asset variability and correlation into account and show that such a model predicts behavior better than standard reinforcement learning models.

Lieu

Bâtiment: Uni Mail

Boulevard du Pont-d'Arve 40
1205 Geneva

Room: M 3250, 3rd floor

entrée libre