Improving Refugee Integration through Data-Driven Algorithmic Assignment
2017 CSS Fellowship
The world is currently experiencing the worst refugee crisis since World War II. The U.S. has long run the world’s largest refugee resettlement program, yet little empirical evidence exists on the efficacy of different approaches to integrating refugees within American society. Currently, the assignment of refugees across domestic resettlement locations is based primarily on local office capacity, and does not leverage the rich insights that historical data can provide regarding which refugees do better in which locations. To address this issue, we will develop a data-driven assignment algorithm that combines supervised machine learning and optimal matching to assign cohorts of refugee arrivals across resettlement locations in such a way that maximizes refugees’ employment outcomes. The algorithm will achieve this by identifying and leveraging synergies between refugee characteristics (such as age, gender, language skills, etc.) and resettlement locations, and will be tested via randomized controlled trials with partnering resettlement agencies.