Patient enrollment in medical trials: Selection bias in a randomized experiment

Self-selection can bias estimates of treatment effects from randomized experiments if one is interested in extrapolating results to the general population. This paper notes that there is an isomorphism between the Roy model for the sorting of workers into sectors and the decision of subjects to participate in randomized experiments. The main implication is that, as the probability of receiving active treatment rises, patients who are less optimistic about new treatment will begin to enroll and estimates of treatment effects will fall. This, in turn, implies that selection bias is positive. These findings are confirmed with data from trials of ulcer medications.