Bringing a Natural Experiment into the Laboratory: the Measurement of Individual Risk Attitudes

Controlled laboratory experiments have become a generally accepted method for studying economic behavior, but there are two issues that regularly arise with such work. The first pertains to the ability to generalize experimental results outside the laboratory. While laboratory experiments are typically designed to mimic naturally occurring situations, ceteris paribus comparisons are rare. Using data from a promotional campaign by a bank and a matching laboratory experiment, we find similar patterns of risk taking behavior controlling for gender and age. The second issue pertains to the impact that the payment procedure in an experiment has on observed risk taking behavior. Specifically, we compare behavior on a risk taking task where that is the only task and payment is assured, where it is one of several similar tasks of which one will be randomly selected for payment, and where it is the only task but there is only a small probability of receiving payment. We find similar behavior across these three payment procedures.