We report on an experiment that isolates key components of pay-for-performance contracts: the effect of financial incentives beyond priming; spillovers from incomplete contracts; and the persistence of incentive effects after contracts end. First conducting a lab-in-the-field experiment with 1,359 maternity care workers in Nigeria, we prime them with a checklist of salient clinical actions and randomize them within 690 facilities to receive either no incentives, rewards, or penalties tied to care for hypothetical patients. Relative to priming alone, both rewards and penalties increase performance on paid actions by 7-8 percentage points (pp; 20 percent) on incentivized actions and generate spillovers of 4 pp (14-15 percent) on unincentivized actions. In the second stage, we observe care for real patients in a cross-randomized sub-sample of 400 participants conducted within 72 hours after the lab experiment. Exposure to an incentivized arm improves real-world performance by 16 to 18 pp (21-26 percent) on actions previously incentivized in the lab stage relative to workers not yet exposed to the experiment. These exploratory effects are significant at the 10 percent level after multiple-testing adjustment, providing suggestive evidence of short-run persistence. Together, these findings offer guidance for designing pay-for-performance contracts to increase their effectiveness and cost-effectiveness.
Replication data are available here.
This paper was first published in January 2024 as “Pay-for-Performance Contracts in the Lab and the Real World: Evidence from Nigeria.” The original version is available here.
