Poverty alleviation programs in Kenya have a problem. Randomized controlled trials are showing that direct cash transfers to poor families are effective in helping them out of poverty, and hundreds of millions of dollars have been allocated for this purpose. The problem is to identify the families who would most benefit.
For this project, the Data Impact Lab is partnering with government agencies and NGOs to solve this problem. Using a recent, high-quality dataset of a sample of households from across Kenya and additional data that is available at high resolution across Kenya, we are modeling the 10 indicators of the Multidimensional Poverty Index (MPI). If successful, these models will create poverty estimates for the very small areas needed for targeting by cash transfer programs.
Our partners, including the NGO GiveDirectly, can put accurate small-area poverty estimates to almost immediate use, helping out of poverty the families the team’s models find. Many other countries have datasets very similar to what we will be using for Kenya, so if the team is successful, the methods it develops could benefit poverty targeting efforts more broadly.
This project is for advanced students. The team is working at the research frontiers of small area estimation, geostatistics, and new Bayesian methods for hierarchical, spatial data. Students must have the necessary math and statistics background to deeply understand the current state of the art, and potentially advance it.
The faculty involved with this project include
- Bill Behrman, ICME
- Lester Mackey, Statistics
Interested students should contact firstname.lastname@example.org.