Banerjee, Duflo & Kremer (2019): Fighting Poverty One Experiment at a Time

The Problem With Good Intentions

For decades, development economics was dominated by grand theories and grand debates. Should poor countries focus on free markets or state planning? Aid or trade? Education or infrastructure? Billions of dollars hung on these arguments — but nobody was testing the answers.

Meanwhile, on the ground, poverty remained stubbornly persistent. Over 700 million people lived on less than $1.90 a day. Children in poor countries died of diseases that cost pennies to prevent. Schools were built but teachers didn’t show up. Clinics were stocked but patients stayed away. Something was clearly not working — but what?

Abhijit Banerjee, Esther Duflo, and Michael Kremer proposed a radical idea: stop debating grand theories and start running experiments. Borrow the randomized controlled trial from medicine, apply it to development programs, and let the evidence decide what works. Their approach didn’t just provide better answers — it changed the questions development economists ask.

Kremer: The Pioneer in Kenya

In the mid-1990s, Michael Kremer was working in western Kenya, trying to figure out why schools in poor countries performed so badly. The conventional wisdom had many answers — too few textbooks, too few teachers, too little funding. But which one mattered most?

Kremer did something that development economists had rarely done before: he ran an experiment. Working with a Dutch NGO, he randomly assigned schools to receive different interventions — free textbooks, flip charts, extra teachers — and measured the results.

The findings were surprising and humbling:

• Free textbooks didn’t help — at least not for the average student. Textbooks were written for the best students and left most behind
• Flip charts didn’t help either
• What did help? Treating children for intestinal worms. Deworming was cheap (about 50 cents per child per year), dramatically improved school attendance, and had spillover benefits to untreated children nearby

This single experiment overturned years of assumptions. It showed that the binding constraint on education wasn’t what anyone expected — it was health. And it demonstrated the power of rigorous experimentation to reveal surprises that theory and intuition miss.

Banerjee and Duflo: Scaling the Method

Abhijit Banerjee and Esther Duflo, both at MIT, took Kremer’s experimental approach and built it into a systematic research program covering education, health, agriculture, credit, and governance across dozens of countries.

Their core insight: poverty is not one big problem — it’s many small problems. And each small problem can be studied with a carefully designed experiment.

How an RCT works in development:

• Identify a specific intervention — say, offering free bed nets to prevent malaria
• Randomly divide communities into a treatment group (gets the bed nets) and a control group (doesn’t — yet)
• Measure outcomes in both groups after a set period
• The difference in outcomes is the causal effect of the intervention — not contaminated by selection bias, pre-existing differences, or wishful thinking

Key findings from their experiments:

• Immunization: In Indian villages, providing a small incentive (a bag of lentils) at immunization camps boosted child immunization rates from 6% to 38%. The problem wasn’t ignorance or opposition — it was the hassle of getting to a distant clinic on a specific day
• Education: Hiring extra teaching assistants to help struggling students improved test scores far more than reducing class sizes or providing more textbooks. The problem wasn’t resources — it was that teaching wasn’t tailored to students’ actual level
• Microfinance: Microcredit — small loans to the poor — was widely celebrated as a silver bullet for poverty. Rigorous RCTs showed the reality was more modest: microcredit helped some people start small businesses, but didn’t produce the transformative effects its advocates claimed. No magic bullet
• Savings: The poor often don’t save — not because they don’t want to, but because saving is inconvenient and the temptation to spend is constant. Simple nudges — labeled savings accounts, commitment devices, automatic deposits — significantly increased savings rates
• Agriculture: Kenyan farmers knew that fertilizer increased yields but still didn’t use it. Why? Not ignorance, but procrastination. Offering a small discount at harvest time (when farmers had cash) dramatically increased fertilizer adoption — more than a larger subsidy offered later

J-PAL: The Infrastructure of Evidence

In 2003, Banerjee, Duflo, and Sendhil Mullainathan founded the Abdul Latif Jameel Poverty Action Lab (J-PAL) at MIT — a research center dedicated to running and coordinating RCTs in development economics.

J-PAL transformed the field:

• A network of nearly 200 affiliated researchers conducting over 1,000 RCTs in more than 80 countries
• Programs validated by J-PAL evidence have been scaled up to reach over 400 million people worldwide
• Governments in India, Indonesia, Peru, and dozens of other countries now use RCT evidence to design policy
• J-PAL created a model for “evidence-based policy” that influenced fields far beyond economics

Kremer also helped found the Development Innovation Ventures program at USAID and the Global Innovation Fund, creating pipelines to scale successful interventions from lab to policy.

What Experiments Revealed About Poverty

Beyond any single finding, the experimental approach revealed deep patterns about how the poor make decisions:

• The poor are rational but constrained: They make reasonable choices given their circumstances — but those circumstances include constant stress, limited information, and zero margin for error. A bad harvest or a sick child can undo years of progress
• Small barriers have huge effects: Charging even a tiny fee for bed nets, deworming pills, or water purification drops drastically reduces adoption. Free distribution works far better — not because the poor can’t afford pennies, but because any friction point becomes an excuse not to act when you’re overwhelmed by daily survival
• Default settings matter as much for the poor as for the rich: Just as Thaler’s nudges work in wealthy countries, default enrollment, automatic reminders, and simplification of procedures dramatically improve outcomes in poor countries
• Context is everything: An intervention that works in Kenya may fail in India. A program that works for women may not work for men. RCTs force humility — the only way to know is to test

The Limits of Experiments

The experimental approach is powerful but not unlimited. Critics have raised important concerns:

• External validity: Does a result in one village generalize to an entire country? To a different country? Scaling up a small experiment can change the dynamics entirely
• Big questions are hard to randomize: You can randomize bed nets, but you can’t randomize trade policy, institutions, or political systems. The biggest determinants of poverty may lie beyond the reach of RCTs
• Ethics: Withholding a potentially beneficial intervention from a control group raises moral questions — even if the evidence produced saves more lives in the long run

Despite these limitations, the experimental approach brought something that development economics desperately needed: discipline. Before Banerjee, Duflo, and Kremer, too much policy was based on ideology, anecdote, or the preferences of donors. After them, the standard is evidence.

Their 2019 Nobel Prize was awarded “for their experimental approach to alleviating global poverty.” Duflo, at 46, was the youngest-ever recipient of the economics Nobel and only the second woman to win (after Elinor Ostrom in 2009).

Explain It to a Child

Imagine you’re a doctor and lots of kids are sick, but you don’t know why. You could guess — maybe they need vitamins, maybe they need exercise, maybe they need cleaner water. But guessing wastes time and money. So instead, you try one thing at a time: give vitamins to half the kids and not to the other half, and see who gets better. Then try clean water the same way. Then exercise. That’s what Banerjee, Duflo, and Kremer did with poverty. Instead of arguing about big ideas like “should we give poor countries more money?”, they tested small, specific things — like giving free bed nets to stop malaria, or offering a bag of lentils to parents who bring their kids for shots. They found that often the cheapest, simplest solutions worked best — but you’d never know without actually testing. Their experiments helped over 400 million people get access to programs that were proven to work.