Website CV Paper
Fields
Labor, Public, Econometrics
Job market paper
Inferring the Effects of Permanent Policies from Temporary Interventions
Experiments to evaluate policy interventions are temporary, but the policies they inform are meant to be sustained or permanent. When past choices shape future ones, the effect of an incentive builds over time, so a temporary experiment can end before the effect has fully built up and understate the response to a permanent policy. We develop and apply a framework that uses temporary experiments to measure the causes of observed persistence in choices and to predict responses to permanent policies. The framework is built around a general dynamic choice model in which choices may be persistent either because past choices shape future ones or because preferences and costs are serially correlated. We show that the model’s key features are identified from the path of treatment effects in a temporary randomized experiment. We apply the framework to the Seattle–Denver Income Maintenance Experiment, which was intended to inform a permanent negative income tax but paid most participants for an announced three year period. We find that taking the experimental findings at face value substantially understates the effect of a permanent policy on hours worked, and that the effect of past hours accounts for over 60 percent of the persistence in hours. In two additional applications, we also show that a temporary incentive to use less social media reveals most of what a permanent one would achieve, while a temporary incentive to exercise reveals only its beginning.