Abhijit Banerjee and Esther Duflo won the 2019 Nobel Prize in Economics for their “experimental approach in alleviating global poverty”. Their experimental approach encompassed a variety of novel methods to understand and analyse interventions and Randomised Control Trials (RCTs). Their research has been used by policy makers to make informed policy decisions to best help the marginalised.
What are RCTs?
To understand the effect of a policy, intervention, or medicine, decision makers try to measure the efficacy of the treatment. Do deworming pills given to children improve test scores? Does providing chlorinated water improve the health and economic outcomes of villages? These are some causal (read causal, i.e. caused by, not casual) questions researchers are interested in. The best way to analyse causal effects is to randomise the selection of people in the treatment and the control group (for example: children who are given deworming pills versus children who are not given the pills). This random selection of the two groups removes many statistical biases that might affect the results.
RCTs in India:
Many of the RCTs performed by Banerjee and Duflo were in India. They involved short- and long-term impact assessments of various interventions, policies, models, and treatments. We look at a few RCTs implemented in India:
Teacher absenteeism rates:
Troubled by the low attendance rates (or high absence rates) of public-school teachers in India, Duflo assessed the impact of financial incentives on the absence rates of teachers in Rajasthan. The study monitored teacher attendance by cameras, which was tied to a financial incentive if the attendance was high. From a baseline absence rate of 44%, teacher absenteeism in the treatment group fell by 21%, relative to the control group. High teacher attendance caused child test scores to improve too.
COVID-19 and health-seeking behaviour:
In the context of COVID-19, Banerjee tested the effect of sending messages via SMS that promoted health preserving behaviour. The results were very positive. By sending a short, 2.5-minute clip to 25 million randomly selected individuals in West Bengal, the intervention i) found a two-fold increase in symptom reporting to village health workers, ii) increased hand washing rates by 7%, and iii) increased mask-wearing by 2%. While mask-wearing rates increased only marginally, the spillover effects (wearing a mask stops the virus from infecting more people) were moderately high and positive.
Asset Transfers and the Notion of Poverty:
An RCT by Banerjee in West Bengal involving a productive asset transfer accompanied with training found large and persistent effects on monthly consumption and other variables. The treatment group reported 25% higher consumption levels relative to the control group, who did not receive the asset transfer and training. Implications of such RCTs are huge. The notion that the poor are lazy and unwilling to perform strenuous labour is falsified by this RCT. Often, what the poor lack are opportunities that are hard to come by, given their financial status. A small nudge, like the asset transfer, can cause large and positive effects on their well-being.
Salt fortification to reduce anaemia:
RCTs also help rule out less cost-effective interventions. Duflo and Banerjee evaluated an RCT which distributed fortified salt in 400 villages of Bihar, to reduce the prevalence of anaemia. However, this intervention found no statistically significant impact on health outcomes like anaemia, hemoglobin, etc. Thus, while RCTs help introduce novel methods of impacting the lives of the poor, they also help in ruling out in-effective measures. A policy maker might try other alternatives to reduce the prevalence of anaemia.
Are RCTs the gold standard?
Maybe. Extrapolating results from a regional RCT to national policies could present problems. Contextuality matters. A study that indicates positive gains for one region might present different, and rather adverse effects for another region. Nation wide effects might not be as prominent as regional results of a single RCT. The good part is that Banerjee and Duflo have a solution. Just perform more RCTs!
.jpg){kind=link}
