A timely Nobel win for policy-relevant research

Professors Esther Duflo (MIT), Abhijit Banerjee (MIT), and Michael Kremer (Harvard University) personify the data-driven, evidence-based approach that now comprises much of Economics. They have spearheaded the use of randomised controlled trials (RCTs) in development research since the early 2000s, rapidly advancing our knowledge of what works to fight poverty with tailored interventions for specific local contexts. The breadth of their impact on poverty alleviation and policy is rivalled closely by how short a time that impact has come in. Before this year, the average age of Nobel Prize winners in Economics was 67 and rising steadily. Duflo, Banerjee, and Kremer are all less than 60 years old. Duflo is only the second woman to win the prize in Economics, and at 46 is the youngest ever winner. Professor Lawrence Katz (Harvard University) put it succinctly in his reaction to their win, “This is probably the first 21st century prize in Economics.”

The widespread adoption of RCTs and other methods based in actual or “natural” experiments to analyse big data has transformed Economics, which detractors still accuse of being mired in complex mathematical theory that has limited relevance to the real world. In contrast, RCTs are relatively simple to conceptualise, but can deliver accurate results on the effectiveness of policy interventions when executed correctly. The methodology is common in medicine, where researchers randomly assign a drug to a set of patients, and later compare the health outcomes of these patients to those of a set of untreated control patients to determine the effectiveness of the treatment. The random assignment of the treatment greatly increases the chances of correctly accepting or rejecting its effectiveness by reducing the risk of unobserved patient characteristics contaminating the results.

Duflo, Banerjee and Kremer have expertly demonstrated the accessibility and efficacy of the methodology in uncovering the determinants of poverty in developing countries. This has led a significant number of young economists to spend their time designing and testing budget-friendly policy measures that governments can use to bring meaningful improvement to people’s lives.

The trio have contributed numerous solutions to the fight against poverty, tackling myriad hurdles to economic development in many countries, such as inefficient allocation of microfinance and low child immunisation rates. Improving children’s education is a particular goal they have collaborated on multiple times. Duflo and Kremer have shown that tracking pupil performance in Kenya increased learning by helping to target teacher effort to differing student levels. Banerjee and Duflo have also worked to bring this approach to the Indian education system. They have conducted large-scale experiments with four state governments and the education NGO Pratham to develop education models that have delivered remarkable learning improvements for children in Haryana and Uttar Pradesh and are poised to scale that up nationally.

They have previously run experiments in India to address teacher absenteeism in government schools and assess the impact of providing information on schooling to local communities on child learning. Perhaps the most interesting contribution Banerjee and Duflo have made recently to Indian development is in police reforms, using a unique experiment with the Rajasthan police that employs measures from transfer freezes to decoy police station visits to improve police effectiveness and public satisfaction with law enforcement.

There are, of course, limitations to the questions in development that RCTs can answer. As Prof Pranab Bardhan (University of California, Berkeley) has pointed out, one cannot randomly assign a monetary policy move such as an interest rate cut across economies to study the impact on national welfare measures such as GDP, for instance. There are clearly also limitations to how far results from an RCT implemented in very specific local settings can be generalised to other scenarios, even in the same country, if there are significant cultural and social differences across sub-national geographic regions and populations.
Recently Prof Angus Deaton (Princeton University), himself a winner of the Nobel Prize in Economics in 2015, raised several technical issues with RCTs as a methodology and stressed the need to research not just “what works” but “why things work”. These are valid concerns for economists and policymakers considering RCT as a research tool.

Another criticism levelled at RCTs is that they do not address larger structural determinants of poverty such as unequal distributions of power, discrimination and government corruption that constitute the overarching social construct within which poor people reside. The use of experiments by economists and other social scientists, including Duflo and Banerjee, is currently providing valuable insights into removing these structural issues as well. Duflo has previously relied on quasi-experimentally assigned quotas for women in elected Gram Panchayat councils in India to write extensively on the welfare and policy impacts of women’s representation in government. Banerjee has recently run RCTs to examine how providing politicians’ performance reports to voters in Indian slums changes who they choose to elect to power.

Professor Leonard Wantchekon (Princeton University), among others, has used RCTs to analyse how initiatives such as information provision to voters can fight corrupt political behaviour such as vote-buying and favour-granting. Findings from such work are valuable in addressing structural barriers to development, even if they come from localised experiments with individuals, as these barriers are possibly the result of individual behavioural choices aggregated across populations.

The decision to award the Nobel Prize in Economics to Professors Banerjee, Duflo and Kremer is timely given the pressing need to find concrete policy solutions to current development issues such as climate change and the rising risk of global economic slowdown. Policies that incorporate individuals’ behavioural preferences have a good chance of success, and experimental research methods can play a pivotal role in testing, tailoring, and scaling up effective interventions that tackle these problems.

(The writer is a lecturer in Economics at the University of Manchester, UK)