Several decades ago, the most common ways of thinking about problems of poor people in low-income countries involved ideas like the \”poverty trap\” and the \”dual economy.\” The \”poverty trap\” was the idea low-income countries were close to subsistence, so it was hard for them to save and make the investments that would lead to long-term growth. The \”dual economy\” idea was that low-income countries had both traditional and a modern parts of their economy, but the traditional part had large numbers of subsistence-level workers. Thus, if or when the modern part of the economy expanded, it could draw on this large pool of subsistence level workers and so there was no economic pressure for subsistence wages to rise. In either case, a common policy prescription was that low-income countries needed a big infusion of capital, probably from a source like the World Bank, to jump-start their economies into growth.
Kremer and his colleagues took a large number of schools that needed considerable support and randomly divided them into different groups. The schools in these groups all received extra resources, but in different forms and at different times. In one study, one group was given more textbooks, while another study examined free school meals. Because chance determined which school got what, there were no average differences between the different groups at the start of the experiment. The researchers could thus credibly link later differences in learning outcomes to the various forms of support. The experiments showed that neither more textbooks nor free school meals made any difference to learning outcomes. If the textbooks had any positive effect, it only applied to the very best pupils.
Later field experiments have shown that the primary problem in many low-income countries is not a lack of resources. Instead, the biggest problem is that teaching is not sufficiently adapted to the pupils’ needs. In the first of these experiments, Banerjee, Duflo et al. studied remedial tutoring programmes for pupils in two Indian cities. Schools in Mumbai and Vadodara were given access to new teaching assistants who would support children with special needs. These schools were ingeniously and randomly placed in different groups, allowing the researchers to credibly measure the effects of teaching assistants. The experiment clearly showed that help targeting the weakest pupils was an effective measure in the short and medium term.
Such experiments have been done in a wide range of contexts. For example, what about issues of improving health?
One important issue is whether medicine and healthcare should be charged for and, if so, what they should cost. A field experiment by Kremer and co-author investigated how the demand for deworming pills for parasitic infections was affected by price. They found that 75 per cent of parents gave their children these pills when the medicine was free, compared to 18 per cent when they cost less than a US dollar, which is still heavily subsidised. Subsequently, many similar experiments have found the same thing: poor people are extremely price-sensitive regarding investments in preventive healthcare. …
Low service quality is another explanation why poor families invest so little in preventive measures. One example is that staff at the health centres that are responsible for vaccinations are often absent from work. Banerjee, Duflo et al. investigated whether mobile vaccination clinics – where the care staff were always on site – could fix this problem. Vaccination rates tripled in the villages that were randomly selected to have access to these clinics, at 18 per cent compared to 6 per cent. This increased further, to 39 per cent, if families received a bag of lentils as a bonus when they vaccinated their children. Because the mobile clinic had a high level of fixed costs, the total cost per vaccination actually halved, despite the additional expense of the lentils.
How much do the lives of low-income people change from receiving access to credit? For example, does it change their consumption, or encourage them to start a business? If farmers had access to credit, would they be more likely to invest in fertilizer and expand their output?
As the body of experimental evidence accumulates, it begins to open windows on the lives of people in low-income countries, on issues of how they are actually making decisions and what constraints matter most to them. The old-style approach to development economics of sending money to low-income countries is replaced by policies aimed at specific outcomes: education, health, credit, use of technology. When it\’s fairly clear what really matters or what really helps, and the policies are expanded broadly, they can still be rolled out over a few yeas in a randomized way, which allows researchers to compare effects of those who experience the policies sooner to those who experienced them later. This approach to economic development has a deeply evidence-based practicality.
From Kremer, here are a couple of JEP papers focused on development topics not directly related to the experimental agenda: \”Pharmaceuticals and the Developing World\” (Fall 2002) and \”The New Role for the World Bank\” (written with Michael A. Clemens, Winter 2016).
- \”From Proof of Concept to Scalable Policies: Challenges and Solutions, with an Application,\” by Abhijit Banerjee, Rukmini Banerji, James Berry, Esther Duflo, Harini Kannan, Shobhini Mukerji, Marc Shotland and Michael Walton
- \”Experimentation at Scale,\” by Karthik Muralidharan and Paul Niehaus
- \”Scaling for Economists: Lessons from the Non-Adherence Problem in the Medical Literature,\” by Omar Al-Ubaydli, John A. List, Danielle LoRe and Dana Suskind