Revealed Preferences and Risk Aversion in Financial Markets - New Empirical Evidence from a Maximum Likelihood Approach

Topic and Objectives

It is well documented that the actual behavior of individual investors deviates from the prescriptions of orthodox theory. It is still unclear whether such deviations are mainly due to truly irrational investor behavior or whether they are rather the result of misspecification or wrong parameterization of preferences. For example, investors who are acting in line with prospect theory might rationally invest into stochastically dominated investments. Extant literature from that field concentrates on theoretical implications from asset pricing models or on experimental studies in a laboratory setting. Either approach suffers from various deficiencies. In a precursory study we applied state-of-the-art techniques from the laboratory to real data from the field in order to estimate parameterized preference functions. For each individual investor from our large sample we use maximum likelihood estimation on each round-trip. For many investors we have more than 100 of those round trips which improves the precision of our estimates. Our approach also naturally allows for out-of-sample testing and the results so far – although preliminary - are very promising. In this project we refined the methodology, applied it to a larger number of investors and merged the estimates on individual preferences with administrative data on demographics and, most importantly, with answers to a questionnaire that was filled out by our sample investors. The comprehensive investor survey was conducted in 2012 and contains questions on self-perceived risk aversion and a number of classic questions to measure the susceptibility to behavioral biases.

Key Findings

In the three working papers resulting from this project, the authors not only find that customized maximum likelihood methods adopted from experimental economics can be applied to financial data, but also identify Prospect Theory as the most common utility function prevalent in the trading patterns of their dataset. Based on an extended likelihood analysis, the authors find that approx. 44 % of investors’ trading behavior seem to be explained by preferences such as Prospect Theory. With regard to Prospect Theory parameters, their results indicate loss aversion parameters to be close to one, a high curvature of the Prospect value function and a moderate curvature of the probability weighting function. However, neither the occurrence of a particular utility model and its parameters nor the impact of preferences on trading behavior seem to correlate with observable sociodemographic characteristics.