|Ph.D Student||Ert Eyal|
|Subject||Do People Exhibit Loss Aversion in Decision under|
Risk and Uncertainty?
|Department||Department of Industrial Engineering and Management||Supervisors||Professor Eldad Yechiam|
|Professor Ido Erev|
|Full Thesis text|
One of the best known implications of prospect theory (Kahneman & Tversky, 1979) is the assertion that “losses loom larger than gains.” Surprisingly, however, direct experimental evidence of this assertion is hard to find. The current dissertation contains a direct empirical evaluation of the loss aversion assertion in two main environments where it was suggested to drive behavior. The evaluation starts with empirical examination of loss aversion in description-based decisions (one shot decisions under risk as studied by Kahneman & Tversky, 1979). In several studies that evaluated such decisions no evidence was found for loss aversion. Moreover, an estimation of the Prospect theory model on the current data suggests that the loss aversion assumption is unnecessary to capture the main trends. The results also document a strong format effect: the tendency of people to reject risky gambles is observed only when these gambles are suggested as alternatives to the status quo.
The second part of the dissertation focuses on decisions from experience (repeated choice tasks under uncertainty as studied by Thaler, Tversky, Kahneman, & Schwartz, 1997). The analysis suggests that Thaler et al's (1997) abstraction of loss aversion is oversimplified. Specifically it neglects other plausible reference point behavioral regularities (e.g., diminishing sensitivity to payoff increments) that seem to facilitate behavior in this environment. Indeed, previous findings that were interpreted as supporting evidence for loss aversion are better captured with assumption of a strong diminishing sensitivity to the payoffs magnitude. Another oversimplification involve Thaler et al's (1997) abstraction of the stock market that reflects choice among different assets (stocks and bonds) as a binary choice between two classes. The current analysis refutes this assumption by showing that the generalization from binary choice to multiple alternatives is not trivial and might lead to systematic biases. The dissertation concludes with evaluating the apparent inconsistency between the current results and the assertion that the individuals are consistent in the relative weight they assign to gains and losses. The results suggest that individuals may exhibit consistency in their sensitivity to gains and losses only when they are given an opportunity to eliminate the risk of losing.