Power to the People: Evidence from a Randomized Field Experiment on Community-Based Monitoring in Uganda

Publication Details

Quarterly Journal of Economics, May 2009, v. 124, iss. 2, pp. 735-69. Available From:

Link to Source
Martina Bjorkman and Jakob Svensson
Sub-Saharan Africa (includes East and West Africa)
Public Sector Management
Equity Focus
None specified
Evaluation design
Difference-in Difference (DID), Instrumental Variables (IV), Randomised Control Trials (RCT)
Journal Article


Ineffective monitoring and weak systems of accountability are some of the reasons for the high rate of absenteeism among health workers and the poor health outcomes in developing countries. This study investigates whether making community members aware of the status of health service delivery in their community, educating them on individuals’ rights to health services and crystallizing the community’s expectations and demands improves health outcomes and monitoring and accountability of service providers.
The study was conducted among 50 rural public dispensaries and households in their catchment areas, defined as areas within a 5-kilometre radius, in nine districts in Uganda. The dispensaries were stratified by district and population size before 25 dispensaries were randomly allocated to the treatment group.
The intervention consisted of a series of public meetings that culminated in the adoption of a shared action plan by community members and health workers. The action plan identified key issues related to health and health services, specified goals, delegated responsibilities to achieve the goals, created a timeline and identified ways for the community to monitor progress. After the initial meetings, communities were left to themselves to implement the action plan. Six month later, health workers and community members met again to track progress.
Information on community monitoring, health workers’ behaviour, utilization of health facilities and health outcomes was collected directly from the internal records of the dispensaries and visual checks by enumerators. A stratified random sample of 5,000 households was surveyed for information on monitoring, utilization of health facilities, health outcomes, perceptions of service delivery, under-five mortality and infant weight. The causal impact of the intervention was measured using a regression equation that controlled for pre-intervention covariates and district fixed effects. For outcomes for which pre- and post-intervention data were available, the impact was measured using difference-in-differences (DID) specification. To examine if community-based monitoring was responsible for the improvement in health utilization and health outcomes, intensity of monitoring was instrumented by district-by-treatment interactions.

Main findings

Health facilities in the treatment group were 32 percent more likely to have a suggestion box, 16 percent more likely to institute a waiting card system and 27 percent more likely to post information on free services. The performance of health staff was discussed 13 percent more frequently in local council meetings in treatment communities, and households in those communities were 4 percent more likely to be informed about the roles and responsibilities of the Health Unit Management Committee, the governing body linking the community and the health facility.
The rate of absenteeism of health workers declined by 13 percent in the treatment communities, and the proportion of households receiving information about the perils of self-treatment and the importance of family planning increased by 7 percent and 6 percent, respectively. Clinics in the treatment group were better managed, and there seemed to be less leakage of drugs from the facilities. Qualified evidence also indicates an increase in the use of medical equipment by health workers and a reduction in patient waiting time. Finally, newborns and young children in treatment communities were more likely to have received required dose of vaccines and vitamin A supplements.
The use of outpatient services and the number of deliveries were, respectively, 20 percent and 58 percent higher in the treatment facilities. Households in treatment communities were less likely to self-treat or rely on traditional healers and were more likely to access antenatal care and family planning services.
In treatment communities, there was a 10 percent decrease in pregnancies, a 33 percent reduction in the under-five mortality rate and an increase in weight-for-age z-score of 0.14. Using instrumental variable regression, and by systematically ruling out alternative explanations, the authors establish a credible case for the role of monitoring in the improvement in outcomes.

Scroll to Top