For the development community, the focus on ‘data’ has been very much on open data: making public where aid dollars are being spent. This is no small task, and I welcome the rise of platforms and initiatives such as The World Bank’s Mapping for Results, DFID’s Project Map, aidinfo and the International Aid Transparency Initiative. Transparency about aid is very important - it raises public awareness of development work, it enhances accountability among both the givers and receivers of aid, and it can drive out waste, bureaucracy and corruption.
But we can do much more with data. Big business already gets this: companies from Tesco to Facebook have been using the data they collect to gain valuable insight on their users and drive efficiency for years. It’s time for governments and the third sector to catch up. In many cases these groups, such as microfinance organisations, local government and community health centres, already collect plenty of data, but don’t make much use of it.
Big data is a powerful tool to help design policy that really works, and bust myths by revealing what doesn’t. Esther Duflo and Abhijit V Banerjee give a good example of this in Poor Economics. They use data on 18 countries to show that “government and international institutions need to completely rethink food policy”. The prevailing wisdom says that we must provide food grains to the very poorest to protect them from starvation: Egypt, for example, spent 3.8 billion in food subsidies in 2008-9. Yet the data reveals that the poor are not desperately striving for more calories. Food makes up only 45-77% of expenditure among rural extremely poor and 52-74% among urban extremely poor. Nor is the rest of their household budget dedicated to necessities: alcohol, tobacco and festivals comprise a large part of the spend. A survey from India confirms this: the number of people who consider that they do not have enough food fell from 17% in 1983 to 2% in 2004. And yet children growing up in these families persistently show stunted growth from a lack of nutrition. From the data the real problem emerges: people are not literally starving, but their diets are not nearly sufficiently nutritious. Thus the best role for governments is not to provide more staples like rice, noodles or wheat, but to provide or subsidize more nutritious foods. (Naturally this does not apply in natural or man-made emergencies).
Big data can be invaluable in improving public service delivery, as well as design. One major challenge for healthcare in the developing world is ensuring that limited supplies of life-saving medicines are distributed to the health facilities where they are needed. Demand for drugs, such as anti-malarials, is not entirely predictable – to ensure that the right distribution is achieved, you need to use real time data. A pilot programme called SMS for Life did just this to improve the distribution of malaria drugs at a health facility level in rural Tanzania, and prevent ‘stock-outs’. Their real innovation was getting front-line workers from every clinic to send an SMS with their stock count each week. Once senior coordinating staff had access to these figures, they were able to accurately target re-stocking of the clinics. The results were dramatic: the proportion of health facilities with no stock of one or more antimalarial medicine fell from 78% to 26%, and in one of the three districts, stock-outs were eliminated by week 8 of the pilot with virtually no stock-outs thereafter.
Randomized Controlled Trials (RCTs) are another powerful way to use data for development. These have long been the gold-standard for evidence in medicine and are gaining traction among NGOs and academics. RCTs can be used to prove the efficacy of the nutrition-based food policy set out above. For example, the Work and Iron Status Evaluation study in Indonesia provided randomly chosen adults with regular iron supplementation from fish sauce. For a self-employed male, the yearly gain in earnings as a result of being able to work harder from improved nutrition was $46 USD PPP, while a year’s supply of the sauce cost just $7 USD PPP. RCTs have also been used to measure the impact of introducing microfinance into a community in India, or to determine the effect on girl’s education of introducing community-based schools in rural Afghanistan.
So yes, open aid data is important, but it’s time that governments and NGOs started doing more to harness the insights of big data and randomized controlled trials. There are many moves in the right direction. One new organisation to watch is DataKind (originally Data Without Borders) which aims to bring data scientists together with NGOs who could use their services. The World Bank have also recently launched Global Findex, a store of data about people’s use of financial products, to inform financial inclusion policy. The data comes from a survey which covered at least 1,000 adults in each of 148 economies using randomly selected, nationally representative samples. I hope that these initiatives mark the start of a new of era of D4D - data for development.