Three Field Experiments on Incentives for Health Workers

Abstract

The economic study of incentives in firms has traditionally focused on one type of incentive—pecuniary—and one causal mechanism—the direct effect of incentives on effort. This dissertation uses three randomized field experiments to explore non-traditional incentives, and non-traditional incentive effects, in the setting of health care delivery.

The first experiment (jointly authored with Nava Ashraf and Oriana Bandiera) addresses an under-appreciated phenomenon: incentives affect not only the effort of agents on the job, but also the selection of agents into the job. We collaborate with the Government of Zambia to experimentally vary the salience of career incentives in a newly created health worker position when recruiting agents nationally. We find that making career incentives salient at the recruitment stage attracts health workers who are more effective at delivering health services, with administrative data showing an improvement in institutional deliveries, child health visits, and immunization rates in the treatment areas. While career incentives attract agents who differ on observables (e.g., they have higher skills and career ambitions), 91% of the performance gap is due to unobservables. The results highlight the importance of incentive design at the recruitment stage for attracting high performers who cannot be identified on observables alone.

The second and third experiments examine the use of non-pecuniary incentives in health care. The second experiment (jointly authored with Nava Ashraf and Oriana Bandiera) studies non-monetary awards. Awards may affect behavior through several mechanisms: by conferring employer recognition, by enhancing social visibility, and by facilitating social comparison. In a nationwide health worker training program in Zambia, we design a field experiment to unbundle these mechanisms. We find that employer recognition and social visibility increase performance, while social comparison reduces it, especially for low-ability trainees. These effects appear when treatments are announced and persist through training. The findings are consistent with a model of optimal expectations in which low-ability individuals exert low effort in order to avoid unfavorable information about their relative ability. The results highlight the importance of anticipating the distributional consequences of incentives in settings in which the performance of each worker affects social welfare.

The third experiment turns from extrinsic incentives (such as career opportunities and non-monetary awards) to "intrinsic incentives"—that is, incentives that make work more intrinsically rewarding. In the context of a rural health worker program in India, I develop and test a novel, mobile phone-based self-tracking app designed to increase agents' intrinsic returns to effort. At nine months of follow-up, the self-tracking app leads to a 27% increase in performance as measured by the main job task (home visits). Moreover, the app is most effective when it leverages pre-existing intrinsic motivation: it produces a 46% increase in performance in the top tercile of intrinsically motivated workers, but no improvement in the bottom tercile. Evidence from survey and performance data indicates that the treatment effect is mediated primarily by making effort more intrinsically rewarding, and not by other mechanisms such as providing implicit extrinsic incentives. The results suggest the potential for wider use of intrinsic incentives that may increase performance at low cost, when agents are intrinsically motivated.