In the summer of 2009, mobile phones were ringing all across Rwanda. In addition to the millions of calls from family, friends, and business associates, about 1,000 Rwandans received a call from Joshua Blumenstock and his colleagues. These researchers were studying wealth and poverty by conducting a survey of a random sample of people from a database of 1.5 million customers of Rwanda’s largest mobile phone provider. Blumenstock and colleagues asked the randomly selected people if they wanted to participate in a survey, explained the nature of the research to them, and then asked a series of questions about their demographic, social, and economic characteristics.
Everything I have said so far makes this sound like a traditional social science survey. But what comes next is not traditional—at least not yet. In addition to the survey data, Blumenstock and colleagues also had the complete call records for all 1.5 million people. Combining these two sources of data, they used the survey data to train a machine learning model to predict a person’s wealth based on their call records. Next, they used this model to estimate the wealth of all 1.5 million customers in the database. They also estimated the places of residence of all 1.5 million customers using the geographic information embedded in the call records. Putting all of this together—the estimated wealth and the estimated place of residence—they were able to produce high-resolution maps of the geographic distribution of wealth in Rwanda. In particular, they could produce an estimated wealth for each of Rwanda’s 2,148 cells, the smallest administrative unit in the country.
Unfortunately, it was impossible to validate the accuracy these estimates because nobody had ever produced estimates for such small geographic areas in Rwanda. But when Blumenstock and colleagues aggregated their estimates to Rwanda’s 30 districts, they found that their estimates were very similar to estimates from the Demographic and Health Survey, which is widely considered to be the gold standard of surveys in developing countries. Although these two approaches produced similar estimates in this case, the approach of Blumenstock and colleagues was about 10 times faster and 50 times cheaper than the traditional Demographic and Health Surveys. These dramatically faster and lower cost estimates create new possibilities for researchers, governments, and companies (Blumenstock, Cadamuro, and On 2015).
This study is kind of like a Rorschach inkblot test: what people see depends on their background. Many social scientists see a new measurement tool that can be used to test theories about economic development. Many data scientists see a cool new machine learning problem. Many business people see a powerful approach for unlocking value in the big data that they have already collected. Many privacy advocates see a scary reminder that we live in a time of mass surveillance. And finally, many policy makers see a way that new technology can help create a better world. In fact, this study is all of those things, and because it has this mix of characteristics, I see it as a window into the future of social research.