Real-Time Impact Evaluation

Reliable estimates of post-program outcomes are essential to impact evaluation. In low- and middle-income countries (LMICs), such outcomes are traditionally measured through surveys. However, a new paradigm has emerged for estimating living standards based on the application of machine learning algorithms to digital data from mobile phones (e.g. Blumenstock et al., 2015, Aiken et al., 2022b), satellites (e.g. Jean et al., 2016, Yeh et al., 2020), and other non-traditional sources (e.g. Fatehkia et al., 2020, Sheehan et al., 2019). Prior work indicates that the resulting estimates are quite accurate, and can be produced much more rapidly and cheaply than traditional surveys.

In a series of related projects based in Afghanistan, Togo, Haiti, and Malawi, we study whether the impact of humanitarian interventions and government policies can be reliably estimated using “digital trace” data  from mobile phones and satellites. If possible, this could open up new opportunities for low-cost program monitoring and impact evaluation. Initial results are described below; contact us for more details on ongoing projects!