Price Search Experiment

Structural econometric methods that assume agents have rational expectations are often criticized. Yet, little is known about the relative costs and benefits of adopting alternative empirical strategies. This paper compares three procedures for inference about a single structural parameter using data from a laboratory price search experiment. Our novel experimental design induces preferences up to the subjective rate of time preference, leaving unrestricted only this parameter and the decision rule that subjects use in solving the search task. We analyze the experimental data under the assumptions of both rational expectations and heuristic behavior, and we also draw inferences using a simple revealed preference analysis that does not require strong behavioral assumptions. We find that the revealed-preference analysis does not provide much information about the discount rate, while stronger behavioral assumptions provide sharper and statistically identical inferences about the population’s discount rate distribution. However, substantial differences in inference appear at the individual level. We compare the individual discount-rate estimates to an external measure of forward looking behavior obtained for each subject using an instrument validated in the psychology literature. The estimates obtained under heuristic behavior are statistically significantly positively correlated with our external measure of time preference, while the estimates obtained under rational expectations and the revealed-preference estimates are not.