Helping Design Togo’s Flagship Cash Transfer Program

Machine learning, combined with digital data, can help identify poor for cash transfer delivery

In response to the COVID-19 pandemic, CEGA co-Director Joshua Blumenstock helped design Togo’s flagship social protection program, Novissi, using insights from machine learning.

Blumenstock’s team at the Data-Intensive Development Lab, in collaboration with Emily Aiken (UC Berkeley), Suzanne Bellue (University of Mannheim), and Dean Karlan and Chris Udry (Northwestern), used mobile phone and satellite data, combined with machine learning, to help identify those Togolese citizens with the greatest need for humanitarian support. The government and the NGO GiveDirectly then transferred cash aid, delivered via mobile money, to the individual’s identified by Blumenstock’s team.

The team first began by applying machine learning algorithms to high-resolution satellite imagery to develop micro-estimates of the relative wealth of every 2.4 km by 2.4 km region in Togo. This initial step relied on nationally representative survey data from 6,171 Togolese citizens collected before the start of the Covid-19 pandemic in 2018 and 2019. Using these data, the team built an algorithm that could predict relative wealth in Togo at a more granular level than the original survey: the 2.4km grid predictions were an improvement over the prefecture-level representative sample collected. The team leveraged the models to create initial poverty maps of rural Togo which helped inform the subsequent 2020 phone-based survey sampling strategy and provided the information needed for the Togolese government to geographically target Novissi to the 100 poorest cantons.

The research team buttressed this initial geographic targeting with an additional phone-based survey of a representative sample of 8,915 individual cell phone subscribers in September 2020. The sampling strategy inferred these individuals lived in rural cantons eligible for Novissi from their mobile phone data. As planned, this second survey provided researchers with a more accurate picture of canton-level individual variation in wealth and consumption; this provided the empirical base to train and validate the ML algorithms using cell phone data to predict an individual’s wealth from their mobile use.  The research team matched both surveys’ responses to a separate CDR dataset obtained from Togo’s two mobile network operators.

The research team developed algorithms optimized using high-dimensional CDR data to predict an individual’s wealth, using the survey data collected to validate its estimates. Blumenstock et al found that these algorithms performed quite well; they generated estimates that correlated strongly with survey and satellite-based estimates of wealth at the canton and prefecture level. Indeed, researchers calculated that this phone-based approach improved the precision of the social assistance program targeting by 42% relative to a naive geographic targeting of the 100 poorest cantons in Togo. Our government partners in the Togolese government then used the algorithm developed by the research team to calibrate Novissi in its second phase and scale up this phone-based approach in Togo. By incorporating this innovative targeting approach, the Togolese government provided unconditional cash transfers of $20 per month to 154,238 of its citizens in most need between December 2020 and April 2021 (around a third of the income of a Togolese earning minimum wage).