Pay-for-performance (P4P) has been shown to be effective in improving job performance in both private and public organizations (e.g., Bandiera et al. 2008, Gertler and Vermeersch 2013, Muralidharan and Sundararaman 2011, Banerjee et al 2008, Duflo et al 2012). The existing literature has however studied the effect of P4P within a single tier of the organizational hierarchy, mostly focusing on lower-tier workers and rarely paying attention to their supervisors and/or managers. The literature has also ignored the fact that incentivizing one layer of the organization (either the lower- or upper-tier) might affect effort across both layers of the hierarchy, either through effort complementarities or fairness concerns (Tirole 1992, Holstrom 2017). This project aims to assess the effectiveness of incentives by considering the entire organization hierarchy (lower- and higher-tiers) rather than focusing on one layer only.
We plan to test this in the context of a large national public institution: the Community Health Program in Sierra Leone. The organization is structured around health units, each composed of one (or occasionally two) peer supervisors (PS) and an average of 9 community health workers (CHW) per PS. CHWs are frontline workers in charge of visiting households in their villages and provide them basic health services: inform them about health, treat/diagnose diseases and provide them ante- or post-natal services. Each CHW is trained, supervised and advised by the peer supervisor (PS) in their corresponding health unit.
The experiment takes place in a subsample of 372 health units spread in 6 districts across Sierra Leone. In each unit, we randomize: (a) the introduction of performance bonus, and (b) whom the bonus is paid to, i.e., only to the CHWs, only to the PS, to both, or to none. The bonus is disbursed every month during the intervention and is proportional to the number of services provided by the CHWs in that month. More specifically, the 372 health units are divided in 4 groups of roughly equal size:
· Control: PSs and CHWs just receive their base wage (100,000 SLL for CHWs and 150,000 SLL for SSL) and no bonus.
· Treatment 1 - Bottom-tier Incentive: On top of the base wage, each CHW receives a piece rate bonus of SLL 2,000 for each patient service performed by herself. The PSs receive no performance bonus.
· Treatment 2 - Top-tier Incentive: On top of the base wage, each PS receives a piece rate bonus of SLL 2,000 for each patient service performed by a CHW under her supervision. The CHWs receive no performance bonus.
· Treatment 3 - Group Incentive: On top of the base wage, each PS receives a piece rate bonus of SLL 1,000 for each patient service performed by a CHW under her supervision and each CHW receives a piece rate bonus of SLL 1,000 for each patient service performed by herself.
We collect information on the number of services provided by each CHW through a newly implemented reporting system: each time a CHW provides a service to a patient, she sends a text message with the name and contact number of the patient to a toll-free number, along with the type of service provided. The information is automatically uploaded to a server and provides a live database of CHW activity, from which the performance bonuses are calculated. The information provided by the CHWs is double checked by a monitoring team, composed of phone operators and monitors in the field, who call/visit a random subsample of the population and make sure CHWs are neither over- nor under-reporting. To ensure truthful reporting in all treatments (including those that do not provide any incentive to the CHW), all CHWs are offered a fixed bonus of SLL 10,000 conditional on truthful reporting.
Our main outcome variables for the study are measures that proxy the effort of CHWs and their supervisors, as well as health outcomes of the population. To measure CHW effort/performance, we will administer an endline household survey that will ask a random sample of the households: how many services they received from the CHW, the quality of those services and --more generally-- their health knowledge and health outcomes. We will complement this information with the number of services provided by the CHW per month (as measured with the number of text messages sent), and the self-reported number of hours spent on CHW activities. PSs’ effort will be measured by surveying CHWs at endline on the number of visits they received from the PS, the amount of time spent by the PS during the visits, and the level of knowledge of the CHW (to proxy effort in mentoring and advising), as well as self-reported hours spent on PS activities.