Risk-averse placement optimization in refugee resettlement

Operational decision-making for refugee resettlement is an emerging area of research. Previous studies have relied on employment likelihood as a key measure of refugee integration, and used it as a quality score to guide optimal placement decisions of refugees into communities. Motivated by the inherent uncertainty in these point estimates, we propose risk-averse placement optimization models and examine the extent to which uncertainty can be appropriately incorporated in light of the risk preferences of the decision-makers. Building on classical mean-risk approaches that emphasize aggregate trade-offs, we develop alternative optimization formulations that incorporate family-level risk around the employment likelihood of individual refugee families, achieving stronger risk reduction while maintaining expected outcomes. The varying levels of risk reduction observed across families provides further motivation to pursue fairness in allocation, so that particularly vulnerable refugees with the lowest employment likelihoods are not disproportionally disadvantaged. Computational experiments on actual refugee data confirm the effectiveness of the proposed optimization approaches, offering practical advances over current placement approaches based solely on expected outcomes. Our research offers important managerial insights into operational improvements in refugee resettlement, demonstrating the value of risk-aware approaches that empower decision makers to optimize placement decisions while mitigating outcome uncertainties for refugees.