The relationship between risk attitudes and heuristics in search tasks : A laboratory experiment ∗

The existing evidence from laboratory experiments suggests that relatively simple heuristics describe observed search behavior better than the optimal stopping rule derived under risk neutrality. Such behavior could be generated by two entirely different classes of decision rules: (i) rules that are optimal conditional on individual utility functions that depart from risk neutrality or (ii) heuristics that derive from limited cognitive processing capacities and satisfycing. In this paper, we develop and test search models that depart from the standard assumption of risk neutrality in order to distinguish these two possibilities. In our experiment, we present subjects not only with a standard search task, but also with a series of lottery tasks that serve to elicit the shape of their utility functions. For each subject, we fit utility functions with both constant absolute and constant relative risk aversion. We find that neither specification yields a search model that is consistent with observed behavior. Our data suggest, however, that loss aversion is important for explaining observed search behavior.