Policy Context

The absence of reliable and comparable cost data for global development interventions has long constrained the ability of funders, practitioners, and researchers to make decisions based on cost-effectiveness. As a result, programs are very often designed—and resources allocated—based on rough guesses about return on investment, which are often inaccurate. CEGA is addressing this problem by making new tools and resources available to help researchers and practitioners proactively generate cost evidence when conducting impact evaluations. However, this strategy is limited when the existing impact evidence is decently good, but the cost data are sparse or unavailable. There is a clear need for a trustworthy cost prediction tool to fill these knowledge gaps.

In collaboration with the Global Poverty Research Lab (GPRL) at Northwestern University, CEGA demonstrated the model’s ability to predict costs per participant across varied settings by leveraging costing vignettes and global datasets. By reconciling incomplete information, the tool unlocks new value from decades of existing research.

Study Design

CEGA is leveraging our cost model tool built with GPRL to develop a pioneering cost prediction tool to support modeling of GiveWell’s mass campaigns. These include programs that deliver insecticide-treated nets (ITN), seasonal malaria chemoprevention (SMC), and vitamin A supplementation (VAS), across a wide range of low- and middle-income country settings.

CEGA will develop model variants tailored to different delivery methods, such as door-to-door distribution versus fixed-post clinics. This expansion will allow GiveWell grantmakers to predict the costs of scaling proven interventions in new contexts, shifting from retrospective analysis to predictive decision-making.

Results and Policy Lessons

This infrastructure ensures that “Smart Buys” in global development are backed by standardized, credible cost estimates, helping to use resources even more efficiently to deliver impact.

Results forthcoming.