October 2013 /
Should development organizations act more like venture capitalists and less like providers of ‘stuff’ and ‘technical assistance’? This was the argument advanced by leading development economist Esther Duflo during her recent lecture on ‘Effective poverty reduction beyond the MDGs’, part of the Kapuscinski Development Lectures series (1). Duflo is a professor at the Massachusetts Institute of Technology and a co-founder of the Jameel Poverty Action Lab. Amongst other publications, Dulfo is well known for her book, co-authored with colleague Abhijit Banerjee, entitled “Poor Economics: A Radical Rethinking of the Way to Fight Global Poverty”. In this book Duflo and Banerjee argue that so many anti-poverty policies have failed over the years due to an inadequate understanding of poverty and a lack of willingness to systematically and coherently generate and learn from evidence. The book explores the economic lives of poor people, arguing that they have to be skilled economists just to survive, and yet face daily barriers such as access to information, education, political institutions, etc. In other words, just being poor changes one’s set of opportunities (2). Banerjee and Duflo suggest that: “It is not easy to escape from poverty, but a sense of possibility and a little bit of well-targeted help (a piece of information, a little nudge) can sometimes have surprisingly large effects. On the other hand, misplaced expectations, the lack of faith where it is needed, and seemingly minor hurdles can be devastating. A push on the right lever can make a huge difference, but it is often difficult to know where that lever is. Above all, it is clear that no single lever will solve every problem” (3).
Understanding more about these levers – as prompts to social change - is centre to the work of the Poverty Action Lab, in which Dulfo and colleagues subject social policy ideas to randomized control trials, much the same as the process used in testing drugs. Duflo has suggested that development policy has not been “any better than the medieval doctors and their leeches” (4) and advocates rather for decision-making based on measurable evidence of cause and effect. Randomized trials are proposed as a means to connect cause and effect, on a scale big enough to filter out “statistical noise”. These trials seek to answer questions such as: how can you incentivize teachers to show up for class? To conduct the experiment, one group of teachers receives the proposed incentive, while the second “control” group does not. The results are then compared to assess the effectiveness of the incentive. Duflo and colleagues have applied this method to studies of the effectiveness of interventions in the fields of microfinance, education, anti-malaria campaigns, women’s’ empowerment, etc. Significantly, Dulfo and colleagues believe that the method of randomized control trials can be used to generate evidence about the notoriously difficult to measure field of human behaviour and social change (5).
The argument that Duflo made about development organizations playing the role of “venture capitalists of development” builds on this earlier work. In the Kapuscinski lecture, Duflo suggests that making progress on eradicating poverty should be oriented towards a series of more modest objectives, moving slowly and accounting for complexity. According to Duflo, poverty is a permanent crisis and requires an imaginative and creative response, two features often lacking in policy. She suggests that development organizations move towards a culture of learning and experimentation, not being afraid to try new things and being tolerant of failure (but also, failing fast). Traditionally development cooperation has been oriented towards filling perceived gaps in partner countries, whether the gaps are assessed to be in terms of infrastructure, human capacities, etc. In some countries where aid is a large percentage of the budget, Duflo acknowledges, this approach still makes sense. But in many others, aid risks to become obsolete if it continues with this approach. Duflo suggests that development organizations instead become the venture capitalists of development innovation – providing risk capital that can finance investment in new technology and new policy. This is not only about developing new “gadgets”, but, for example, could also be about innovatively organizing a primary school system to deliver better results in terms of learning. In the venture capital model, innovations go through three steps: 1. basic research; 2. making the innovation market-ready and 3. scaling-up & diffusion. Venture capitalists fund the first two steps, expecting to make a profit on the third. As Duflo points out, many socially valuable innovations are not going to have a market, thus a purely market-driven model will not work. And national political actors are less able to take on the risk associated with innovating new ideas (and potential failures), as they have promises to their constituencies to fulfil. Rather, she suggests that development organizations step in to fund steps 1 and 2 in the development of new and innovative technologies or policies aimed at alleviating policy. This approach also implies the emphasis on experimentation and testing that runs through Duflo’s earlier work, as these are key to steps 1 and 2 of developing innovations.
In some ways, this is not a new approach to development (or to policy development more generally). There are many cases of initiatives that start at a very small scale and that are tested and developed and “the kinks worked out” before, if successful, being up-scaled. The short-span community bridges in Nepal, based on an improved version of the bridges that had traditionally been built, is but one emblematic example of a small-scale innovation supported by development actors now implemented nation-wide (6). This initiative was based on innovation, adaptation and “field testing” (though not randomized experiments). However, the approach suggested by Duflo and colleagues would imply that international development organizations stay out of the scaling-up part of the process, leaving this to national actors. This more experimentation-oriented approach to supporting development innovations would require a change in priorities for development organizations as they, no less than national political actors, are under pressure to succeed and to produce success stories. Though analyzing failures is (in theory) part and parcel of “lesson learning” exercises, it is still not a topic that is widely addressed (7).
The broader notion of development cooperation as experimentation, and the use of randomized control trials as a methodology for learning about the effectiveness of development initiatives, has been criticized from different angles. An important critique is related to the ethical implications of experimentation on subjects (“poor people”) who are, as argued above, engaged in a daily struggle for survival. Reportedly, Duflo has declined to participate in trials that might significantly disadvantage either the treatment or control group (4). Additional critiques include the observation that randomization-based evidence has not played a role in significant social changes of recent times (for example, the civil rights movement), as well as methodological concerns about the quality of data garnered from implementing randomized trials at such scales and with such a complexity of factors and the difficulty in generalizing the findings, even if based on perfect data (4).
Whether or not we decide to adopt models from venture capital, the work of Duflo and colleagues is inspiring in several respects: 1. We should not shy away from the difficultly of measuring change in social and political fields but rather be more creative in how this can be assessed, 2. There should be more space to seek creative solutions to even very small-scale issues affecting the lives of poor people, 3. Much of what we do is actually rather experimental in nature (a theory of change is but a theory (8), thus we should continuously emphasize the learning aspect of learning-by-doing.
Download this article as a PDF
(1) Esther Duflo (2013). Effective poverty reduction beyond the MDGs. September 24, 2013.
(2) Jane Carter (2013). So Being Poor Saps Mental Abilities? Poverty-Wellbeing.net, September 2013. https://www.shareweb.ch/site/Poverty-Wellbeing/currentpovertyissues/Documents/20130909%20So%20Being%20Poor%20Saps%20Mental%20Abilities%20-%20Jane%20Carter.pdf(3) Abhijit Banerjee, Esther Duflo (2012). Poor Economics: A Radical Rethinking of the Way to Fight Global Poverty. Public Affairs, March 27, 2012.
(4) Ian Parker (2010). The Poverty Lab: Transforming development economics, one experiment at a time. The New Yorker, May 17, 2010, p. 79.
(5) In a New Yorker profile of the Poverty Action Lab (see reference 4), Duflo recounts an exchange during a meeting at the World Bank in Delhi. Duflo’s colleague made a passing comment that randomized evaluation can be taught – it’s not nuclear physics. To this one of the other meeting participants responded: studying human beings is much more complicated than nuclear physics! (Parker 2010).(6) SDC (2006). Trail Bridges in Nepal: Partnership Results. Asia Brief.
(7) There are exceptions to this trend, for example the “failure reports” produced by Engineers Without Borders
(8) On this topic, please see the recent blog post & discussion at:
http://www.oxfamblogs.org/fp2p/?p=16325, see also https://www.shareweb.ch/site/Poverty-Wellbeing/addressing-poverty-in-practice/impact-hypotheses