Lingo

Development Economics & RCTs

Using randomized experiments to discover what actually reduces poverty.

What is it?

Development RCTs apply the methods of clinical trials to development policy. Researchers randomly assign villages, schools, or households to receive an intervention, then compare outcomes against a control group. This approach, championed by Abhijit Banerjee, Esther Duflo, and Michael Kremer (Nobel Prize 2019), shifted development economics from grand theories toward evidence about what actually works.

Why randomized trials

Clean identification: Random assignment means treated and control groups are balanced on both observable and unobservable characteristics, so any difference in outcomes is attributable to the intervention—not to selection.

Credible answers to specific questions: Instead of debating whether “education improves income,” an RCT asks precisely: does providing free deworming pills improve school attendance?

Iterative improvement: Results feed back into program design. First test the medication, then test how to get people to take it, then test how to scale the delivery.

Landmark findings

Deworming: Kremer’s school deworming program in Kenya found that deworming tablets dramatically improved attendance cheaply—spurring mass treatment programs across the developing world.

Micro-credit: Banerjee and Duflo found that extending micro-credit to poor households had modest effects on business creation but did not transform lasting poverty—tempering earlier hype.

Teaching at the right level: RCTs showed that tailoring instruction to students’ actual level (rather than the curriculum) dramatically improved learning, influencing policy across several countries.

Commitment savings: Devices that lock up savings helped people save more—consistent with present bias—guiding products like commitment savings accounts.

Criticisms and limits

External validity: A result in one village may not hold elsewhere or at scale. RCTs answer localized questions well but generalizing requires judgment beyond the experiment.

LATE and heterogeneity: RCTs estimate average effects for the studied population, not for everyone. Effects often vary across subgroups, making blanket conclusions risky.

Ethics: Randomly denying a potentially helpful intervention to a control group raises moral questions, though researchers argue ignorance is itself costly.

Narrow focus: Critics contend RCTs favor small, measurable interventions while sidelining big structural questions about institutions, power, and inequality.

References

  1. Nobel Prize: 2019 Economic Sciences— Nobel citation for Banerjee, Duflo, and Kremer for their experimental approach to alleviating global poverty.
  2. Wikipedia: Randomized Controlled Trial— Explanation of RCT methodology, design principles, and applications in development economics.
  3. J-PAL: Poverty Action Lab— MIT-led network conducting randomized evaluations of anti-poverty programs worldwide.

Examples