Abstract:
Standard economics models require that financial incentives improve performance, while leading theories in psychology allow for the opposite. Experimental results are mixed, and so far have not been corrected for publication bias and model uncertainty. We collect 1,568 economics estimates together with 46 factors capturing the context in which the estimates were obtained. We use novel nonlinear techniques to correct for publication bias and em- ploy Bayesian model averaging to account for model uncertainty. The corrected estimates are zero or tiny across contexts of field experiments, including differences in performance measurement, task definition, reward size and framing, motivation beyond money, subject pool, and estimation technique. Laboratory experiments produce statistically significant estimates on average after correction for publication bias, but even there the effect is weak. Experimental economics evidence is inconsistent with standard economics models.