When Incentives Matter Too Much: Explaining Significant Responses to Irrelevant Information

When economic agents make decisions on the basis of an information set containing both a continuous variable and a discrete signal based on that variable, theory suggests that the signal should have no bearing on behavior conditional on the variable itself. Numerous empirical studies, many based on the regression discontinuity design, have contradicted this basic prediction. We propose two models of behavior capable of rationalizing this observed behavior, one based on information acquisition costs and a second on learning and imperfect information. Using data on school responses to discrete signals embedded in North Carolina's school accountability system, we find patterns of results inconsistent with the first model but consistent with the second. These results imply that rational responses to policy interventions may take time to emerge; consequently evaluations based on short-term data may understate treatment effects.