Esther Duflo and Randomized Controlled Trials in Development Economics

Esther Duflo revolutionized development economics by building a body of rigorous field experiments to answer a deceptively simple question: which interventions actually lift people out of poverty? Rather than relying on anecdote, intuition, or correlational data, Esther Duflo and randomized controlled trials became the gold standard for testing policy, unlocking billions in better-targeted aid and spurring hundreds of nonprofits and governments to measure impact with real data.

Why Development Economics Needed Randomized Trials

Before Duflo’s work, development economics relied heavily on intuition, anecdotal evidence, and observational data. Policymakers and donors assumed that if a country was poor, the fix was obvious: build schools, distribute cash, improve healthcare access. The problem was that intentions and outcomes diverged widely. Money allocated to schools did not always raise enrollment or learning. Free healthcare clinics in some regions saw tepid uptake. The question “does this program work?” was rarely answered with rigor.

Observational data in developing countries is messy. If a school is built in a wealthy village and enrollment rises, you cannot assume the causation—the wealthy village might have improved school outcomes anyway because of economic growth. Duflo recognized that the only way to isolate a program’s effect is to compare identical groups, where one receives the intervention and the other does not, assigned by chance. That is a randomized controlled trial (RCT).

This was not new methodology—pharmaceutical companies had run RCTs for decades—but applying it to development policy was unconventional and initially met skepticism. Duflo had to demonstrate that RCTs were feasible, ethical, and revelatory in the messy real world of villages, schools, and clinics across Africa, South Asia, and Latin America.

The Methodology: How Duflo Set Up Field Experiments

A Duflo-style RCT works like this. You identify a community or a region with a specific problem—low school enrollment, poor public health, lack of entrepreneurship. You recruit willing communities (often partnering with local governments or NGOs) and randomly assign them to treatment or control groups. The treatment group receives the intervention; the control group does not (or receives a delayed version for ethical reasons).

Before the program starts, researchers collect baseline data on both groups—enrollment rates, income, health markers, whatever the program aims to improve. Months or years later, after the program has run, they collect the same data again. The difference in outcomes between treatment and control is the program’s causal effect.

Duflo’s genius was in the scale and scope of this methodology. She did not run one small experiment; she coordinated dozens, often in partnership with the same communities over years, testing variations. Did conditional cash transfers (paying families to send kids to school) work better if the payment went to mothers or fathers? Did a classroom size reduction improve learning, or did teachers simply teach the same material less carefully? Did microcredit actually help the poor start businesses, or did borrowers use it to smooth consumption?

The answers surprised many. Conditional cash transfers did work—but paying mothers was far more effective than paying fathers at raising school completion and child health. Classroom size reductions had modest effects. Microcredit helped borrowers, but it was not a poverty-elimination tool as enthusiasts claimed.

Landmark Studies and Findings

Duflo’s early work tested schooling and health interventions in India and Kenya. One study examined whether deworming campaigns in schools improved attendance and learning. The result: deworming was highly cost-effective, raising attendance by 7% and improving test scores. Moreover, the cost per child per year was less than a dollar. This finding shifted aid allocation; organizations that had dismissed parasitic infections as a secondary issue suddenly prioritized deworming.

Another celebrated experiment tested whether providing free school meals versus take-home rations changed enrollment and completion. The answer varied by region and family income, but the experiments revealed that take-home rations were often more effective, because they reduced the opportunity cost for families and had persistent effects after the program ended.

Duflo also examined the “role model effect” in education. Does having female teachers encourage girls to attend and persist in school? Field trials in India showed the answer was yes—in some contexts, particularly in rural areas, exposing girls to female teachers raised enrollment and ambitions.

One of her most influential findings challenged a sacred assumption about poverty: that the poor simply needed more resources. By testing a bundled set of interventions (cash, health screenings, business training, savings groups), Duflo found that no single intervention reliably lifted families out of poverty, but a combination of modest, well-designed programs, sustained over time, did move the needle. This suggested that poverty is not a simple shortage of money—it is a dynamic trap of low expectations, thin social networks, and fragmented incentives.

The Founding of J-PAL

In 2003, Duflo co-founded the Abdul Latif Jameel Poverty Action Lab (J-PAL) at MIT with economist Abhijit Banerjee and others. J-PAL became the institutional engine for scaling RCT methodology. By 2019, J-PAL had run over 600 randomized evaluations across 80+ countries. The lab trained a new generation of development economists in experimental design and helped governments and nonprofits commission rigorous evaluations of their own programs.

J-PAL’s structure was deliberate. Rather than tell countries what to do, J-PAL partnered with local organizations and governments to test their own ideas. This built local capacity and ensured that findings were actionable within local political and institutional constraints. It also reduced the “white savior” dynamic that had plagued development work—the researchers were working with partners, not imposing solutions.

The Nobel Prize and Broader Impact

Duflo, along with Banerjee and fellow economist Michael Kremer, won the 2019 Nobel Prize in Economic Sciences “for their experimental approach to alleviating global poverty.” The prize recognition validated a quiet revolution: rigorous experimentation was now the gold standard in development policy.

Her work influenced donors and governments worldwide. The World Bank, UK aid (DFID, now the Foreign Commonwealth Office), and dozens of other institutions began funding and commissioning RCTs. Nonprofits that had distributed aid based on belief now had to prove impact. Foundations shifted capital toward tested interventions. This shift from hope to evidence has likely improved the lives of hundreds of millions, by concentrating resources on programs that work.

Limitations and Critiques

RCTs are not a panacea. Critics note that field experiments often test small-scale, short-term programs in specific contexts. Does a deworming campaign that works in rural Kenya scale to urban areas, or retain its effect if rolled out nationally? Can findings from one region generalize to another? Duflo herself has acknowledged these trade-offs: RCTs excel at isolating a causal effect but sometimes struggle with external validity and systemic questions.

Additionally, RCTs can be expensive and slow, taking years to yield results while urgent policy questions demand faster answers. And some outcomes—institutional change, political accountability, cultural shifts—are harder to measure in a traditional RCT framework.

Legacy

Duflo’s work transformed how development is conceived and funded. She demonstrated that poor communities are not passive recipients of charity but active participants whose decisions and constraints can be studied rigorously. Her insistence on evidence has raised the bar for aid organizations and policy, and her mentorship of younger economists has ensured that experimental methods spread globally.

See also