Duflo’s talk was based on research – related to the COVID-19 pandemic – using randomized control trials (RCTs) to make “messaging to poor people” more effective. “That’s at the core of a lot of potential strategies in developing countries,” she said.
Duflo has won acclaim for her pioneering use of randomized control trials, or RCTs, which subject randomly chosen subjects to different treatments and measure their response in a way aiming to reduce potential sources of bias. One of their strong points is a high-powered evidence-based assessment of whether designed impacts are being achieved – which FAO Director-General QU Dongyu hailed as in line with his insistence on moving “beyond simply reporting on outputs and results towards an emphasis on impact and outcomes.”
“We need proof” of what works in order to pursue the bold action the world needs to eradicate hunger and extreme poverty, Qu said, adding that RCTs have had a “dramatic impact on our daily work” and more broadly helped improve direct aid and public policy.
FAO’s Hand-in-Hand Initiative “welcomes randomized control trials to find the best solutions,” Qu said. “But it also aims at scaling up such solutions.”
Duflo, a professor at the Massachusetts Institute of Technology, won the 2019 Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel along with other RCT pioneers Abhijit Banerjee, and Michael Kremer, who also gave a lecture at FAO recently.
Her FAO lecture explored the very recent work she has led into how various people have responded to messaging aimed at optimizing actual behavior during the COVID-19 public health emergency, during which there has often been an information overload.
Simplicity and expertise
In 2020, Duflo and her team set up RCT-based experiments to test what kind of communication tactics were effective in conveying impactful information promoting physical distancing to people in West Bengal, India, and to Black and Hispanic people in the United States of America.
The studies focused on direct messaging via smart-phone applications, to test the efficacy of different contents and approaches. The idea was to ascertain whether such light-touch campaigns can have an impact.
In West Bengal, using a celebrity figure to spread reminder-like information strongly increased the frequency with which people reported symptoms to local health agencies, while having marginal influence on the use of masks, travel and hand washing frequency. The knowledge gap narrowed- information levels being quite robust already – but above all “it led to a behavioral nudge,” Duflo explained.
In the United States, the experiment sought to grasp whether minority communities preferred to receive information from a doctor from their own group. By and large the answer was negative.
Duflo’s team later scaled up up the U.S. experiment with Facebook to detect responses to public health messaging by medical professionals “at scale”, using a broad sampling of countes and mobility indicators available through that platform.
For Duflo, one broad takeaway is that having experts speaking directly to people in simple terms has the greatest impact. That said, celebrities are often more effective than officials appointed as experts, and leaders who emerge organically within their own communities are especially powerful, she noted.
She also emphasized how complex subjects – such as nutrition – are especially complicated in terms of breaking down what factors are the real behavior drivers. For example, poorer people may be more likely to wear masks because they are more likely to have to work even amid lockdowns.
RCTs are quite resource-intensive to conduct, and their findings can be marginal rather than revolutionary, but it is on the margin that reality on the ground changes. “If we learn how to properly scale up these randomize and control trials the impacts could be significant and also provide more information on how they will vary depending on the context where they are scaled up,” said Maximo Torero, Chief Economist of FAO, who moderated the event. Duflo noted how she is now working – using Bayesian methods and machine learning – on how to interpret the results of several randomized control trials done in various places – which would be of large value to FAO.