Using AI and digital data to target cash transfers in Togo

Policy Context

The COVID-19 pandemic has devastated many low- and middle-income countries (LMICs), spurring the first increase in global extreme poverty in over two decades. The pandemic further highlighted challenges with traditional approaches to humanitarian response, which rely on expensive and slow manual process to enroll and pay beneficiaries.

Study Design

Beginning in April 2020, CEGA co-Director Joshua Blumenstock began providing direct technical support to the government of Togo in response to the COVID-19 pandemic. In collaboration with Dean Karlan and Chris Udry from Northwestern University, Blumenstock’s team at CEGA and the Data-Intensive Development Lab is using mobile phone and satellite data to help identify the country’s most vulnerable individuals and disburse cash transfers to them through the government’s flagship social protection program, Novissi.

Through Novissi, over 600,000 Togolese individuals have already received unconditional cash transfers of roughly $20/month (around a third of the minimum wage in Togo), delivered via mobile money directly to the beneficiary’s mobile phone. The research team led by Blumenstock is helping the government identify those people with the greatest need for this humanitarian support.

The targeting technology that is driving this expansion is based on a series of research studies published by Blumenstock (1, 2, 3). It relies on using supervised learning methods, with make use of traditional household surveys as well as non-traditional digital data from satellites and mobile phones, to estimate the wealth of small villages and individual mobile subscribers.