Website CV Paper
Fields
Theory, Behavioral & Experimental
Job market paper
Dynamic Pilot Design
This paper studies the optimal dynamic pilot design consisting of joint choice of pilot treatment, pilot duration, and deployment policy. In the model, adaptive treatment affects pilot participants' welfare and generates information about a payoff-relevant state. In pilot participants are treated gradually, however commitment enables faster deployment under a fixed policy, creating an opportunity cost of continued experimentation. I show that the optimal pilot intensity increases with confidence and, holding beliefs fixed, declines over time. Pilot duration is determined by a confidence threshold for deployment, which also declines but retains a minimum standard of evidence. This declining threshold creates a selection effect: among pilots that lead to deployment, longer duration predicts lower terminal confidence and expected welfare per recipient. Pilot treatment is more aggressive than immediate deployment would warrant at the same belief, and intensity falls discontinuously at deployment as its informational return disappears. In the large-population limit, increasingly intensive pilots resolve uncertainty in vanishing time, while welfare per person converges to its full-information level.