How combining demand-side tools for immunization delivered tremendous impact
Nobel Laureate and J-PAL economists conducted a large-scale experiment in India to test how to increase childhood vaccination rates. Their research shows that a package of information-hub ambassadors, text reminders, and incentives raised rates up to 44%. Without incentives, rates still climbed by 26% while costing less than the status quo.
This site previews our May 21 webinar. Dr. Esther Duflo, Dr. Arun Chandrasekhar, and J-PAL team members will walk through their research, and Evidence Action will share how they're applying these findings through a pilot in Nigeria.
Vaccines are widely recognized as a "best buy" for global health, generating $52 for every dollar invested. But 21 million children remain unvaccinated or undervaccinated, allowing for morbidity and mortality from preventable diseases. This isn't only because of a lack of immunizations; over half of those who miss a shot say it's because they were unaware. Dr. Banerjee, Dr. Chandrasekhar, Dr. Duflo, and other researchers from The Abdul Latif Jameel Poverty Action Lab (J-PAL) ran a randomized experiment testing ways of increasing the demand for vaccinations.
Three tools that worked better together
Each component targets a different barrier to vaccination. their effect was limited — but combined, they amplified each other.
Caregivers were randomly assigned to receive text and voice call messages reminding them that their child was due to receive a specific vaccine. Researchers randomly varied the fraction of caregivers receiving reminders in the catchment area of each subcenter: either zero, 33, or 66%.
Researchers leveraged social networks to test an ambassador program in a subset of villages. The ambassadors received one text message and one voice call every month asking them to remind their friends, family, and other community members of the value of immunization and, in villages with incentives, remind them about the incentives.
Within each village, randomly selected individuals were asked to identify people with certain characteristics. The six people nominated most frequently in each village were recruited as ambassadors. Villages were randomly assigned to one of five different ambassador selection strategies:
- Information hubs: respondents were asked to identify who was good at relaying information.
- Trusted individuals: respondents were asked to identify who was generally trusted to provide good advice on health or agricultural questions.
- Trusted information hubs: respondents were asked to identify who was both trusted, and good at transmitting information.
- Random selection: six ambassadors selected randomly.
- Comparison group with no ambassadors.
Caregivers at a subset of randomly selected health centers received a small incentive through mobile credit each time they brought their children to get immunized. Caregivers received one of four incentive structures: High incentive, flat payment; High incentive, increasing payment; Low incentive, flat payment; Low incentive, increasing payment.
An important nuance on incentives
The combined package was most effective in poorer areas with low initial vaccination levels — in the most responsive villages, on-time measles immunization went from 13% to 61%. However, in some villages with higher baseline immunization, the full package with incentives actually had a negative effect. This is why targeted deployment matters — and why we'll explore it more in the webinar.
Two packages, tested at scale
Researchers used Treatment Variant Aggregation (TVA), a novel machine-learning technique, to evaluate 75 combinations of three interventions across nearly 300,000 children in Haryana, India. Two packages emerged as the most promising — one optimized for cost-effectiveness, the other for maximized coverage.
Reminders + ambassadors
Reminders + ambassadors + incentives
Why this matters
One of the most striking findings from this research isn't any single intervention — it's that delivering demand-side tools as an integrated package produced results that no component achieved alone. SMS reminders had no significant effect in isolation. Ambassadors moved the needle on their own, but combining all three tools unlocked dramatically larger gains. The study also used an innovative methodology — combining a large-scale RCT with machine learning to evaluate 75 policy bundles — to determine not just which packages work, but where and for whom. The workshop will explore both the case for bundled delivery and the methods behind these findings with the research team.
Adapting the evidence for Nigeria
Nigeria houses the highest number of unvaccinated (8.7M) and zero-dose children in the world (2.1M). UNICEF surveying shows that low demand is one of the most critical barriers to Nigeria's vaccination gap, indicating that of those who miss vaccine doses, 50% attribute it to awareness gaps, 40% to immunization not being a priority, and 12% to long distances among reasons for missed doses. Evidence Action and J-PAL are collaborating to adapt and pilot the intervention packages tested in Haryana, combining reminders, information hub ambassadors, and targeted incentives, in two Nigerian states.
Evidence Action has been delivering health interventions in Nigeria since 2016, successfully delivered the community ambassador model at scale before, and is already reaching pregnant women with SMS messages.
The workshop will also explore how Evidence Action is building on the Haryana study and broader evidence in the ecosystem to address Nigeria's underimmunization crisis.
Three papers, one study ecosystem
These papers are the source of the evidence that will be discussed in the workshop. For further reading, click to expand each for a brief summary.
The primary paper from the Haryana study. Researchers cross-randomized three interventions — incentives, SMS reminders, and community ambassadors — across 140 primary health centers and 755 subcenters serving ~295,000 children. This produced 75 unique treatment combinations.
Rather than reporting results for each combination individually, the authors developed a machine-learning method (Treatment Variant Aggregation) to identify which packages were most effective and most cost-effective. This is the methodological innovation that makes this study unusual — and what will be a focus of the workshop.
Read the full paper →| Location | 7 low-coverage districts in Haryana, India |
| Scale | 295,038 children; 471,608 vaccines administered |
| Timeline | 2016–2018 |
| Baseline full immunization | ~39% (parent-reported); <20% on-time measles |
| Outcome measure | Measles vaccine receipt (proxy for schedule completion) |
This paper establishes the theoretical and empirical foundation for the ambassador component. In two experiments — 213 villages in Karnataka and 521 villages in Haryana — researchers tested whether community-nominated "gossips" could spread information more effectively than randomly selected individuals, influential leaders, or trusted advisors.
The key finding: simply asking a few villagers who is good at spreading information was an easy, inexpensive, and reliable way to identify network-central individuals who then drove meaningful behavior change. In Haryana, villages with information hub ambassadors saw 22% more children vaccinated monthly. These nominated individuals were central in a network sense — not just popular or powerful.
Read the full paper →A 3ie evidence impact summary presenting the results of the RCT testing each of the three interventions — reminders, information-hub ambassadors, and incentives — and their combinations across Haryana. This summary translates the research findings for a policy and practitioner audience.
Read the impact summary →