Aid organizations can deliver aid more efficiently to those who need it most by using mobile phone data. This technique has been used in Togo to distribute covid-19 emergency aid.
The Covid pandemic is estimated to have pushed more than 100 million people into extreme poverty, especially in low-wage countries. In response, governments and aid organizations worldwide have stepped up financial aid to 1.5 billion people since early 2020. But determining who needs that help most is difficult, especially in countries that do not centrally track household income.
Mobile Phone Data
To tackle this problem, an American research group developed a technique where an artificially intelligent computer system (AI) can estimate poverty in a small area, and even of individuals, based on mobile phone data. The researchers tested this technique in Togo in 2020, shortly after the pandemic hit this country. They cooperate with Togo’s Ministry of Digital Economy and GiveDirectlya non-profit organization that sends money to people living in poverty.
ALSO READ
Blowing without turning
“The idea behind our approach is that rich people use phones differently than poor people,” write two of the researchers a short analysis of their research. For example, rich people tend to make international calls more often. They also call themselves more often, instead of receiving phone calls. And they buy larger data bundles all at once. Poorer people mainly have shorter and more local conversations.
So the phone calls and text messages from the poorest people follow different patterns. AI algorithms can learn to recognize those differences. This allows them to estimate whether a particular mobile subscriber is rich or poor.
To train their AI systems, the researchers first collected information about the living conditions of several thousand households. They did this through telephone surveys. They compared that information with telephone use data they received from telephone companies. The AI learned to recognize user patterns of people living on less than $1.25 (€1.13) a day.
Testing in Togo
The next step was to see if this technique could provide financial aid to the poorest people. To do this, the researchers conducted a test in which the AI analyzed phone data from the two main mobile networks in Togo. In November 2020, the government and GiveDirectly paid out the first amounts based on the AI’s estimates.
The technique proved successful. It ensures that more money reaches the poorest. In the analysis, the researchers write: “To date, the program has provided nearly $10 million to approximately 137,000 of the poorest citizens in the country.”
The system also works better than other methods, such as providing financial aid only to the poorest provinces. With that method, only 33 percent of people living on less than $1.25 a day benefit, the researchers calculated. The AI approach ensured that 47 percent of that group benefited.
The researchers are now conducting follow-up research to assess the impact the money had on the recipients. They also emphasize the importance of finding ways to allow people to participate without a phone. In Togo, about 85 percent of households have at least one telephone. It is unclear how many people missed out on help because they did not have a mobile phone.