Performance Analysis: We used a multi-criteria performance analysis of Neighborhood Support Teams (NSTs), community-powered teams mobilized to provide essential support and resources to refugees resettled to the US under private sponsorship pathways. In partnership with Ascentria Care Alliance, a regional multi-state nonprofit coordinating these efforts in the New England area: (1) We conducted a first-of-its-kind empirical analysis of NST operations through interviewing and surveying to collect high-quality data. (2) We crafted a well-suited Data Envelopment Analysis (DEA) informed by contextual operational dynamics, providing a rigorous assessment of NSTs' relative efficiency and marking its first application in community-led resettlement efforts. (3) To address the challenges of model specifications in DEA, we further applied logistic regression with statistical testing in an iterative algorithmic process to systematically select variables for DEA. (4) We offered best practices to improve NSTs' effectiveness and sustain their long-term welfare while discussing our findings' broader societal and ethical implications.
Preference-based matching: Around a quarter of the population in Ukraine was displaced by the 2022 Russian invasion, creating the largest refugee crisis in Europe since World War II. Of the more than 6 million people who crossed international borders, around 110,000 have been granted temporary asylum in the United States (US) as part of the Uniting for Ukraine (U4U) program. If there is substantial diversity in refugee preferences over communities, then preference elicitation is valuable as it can help to improve the relocation process using refugees’ private information. On the other hand, respecting community priorities can lead to better integration outcomes and efficient use of local resources. RUTH (Refugees Uniting Through HIAS ) is a first-of-a-kind algorithmically-driven platform that matches refugees to communities by explicitly considering the preferences of eligible Ukrainian beneficiaries with the priorities of US sponsors. Working directly with HIAS, an international refugee agency that resettles refugees in the United States, we incorporate several staff-informed features that include additional fairness factors important in the context of relocating refugees but retain the desirable strategy-proofness, efficiency, and envy-free properties of the original procedure.
Outcome-Based Matching: Past refugee placement and outcomes data have been used to estimate future refugee employment likelihoods, which are then integrated as match quality scores into decision models to recommend optimal placement decisions of refugees into communities by refugee resettlement agencies. As these scores are point estimates, uncertainty can lead to suboptimal placement recommendations, thereby putting vulnerable refugees at a disadvantage when accessing scarce employment opportunities. For the first time, we introduce family-level risk measures to incorporate the uncertainty around employment likelihood estimations of individual refugee families into placement optimization. We further explore the fairness of refugee placement decisions to promote the employment outcomes of particularly vulnerable refugees with the lowest employment likelihoods. Computational results on actual refugee data demonstrate the effectiveness of the proposed optimization framework that embeds family-level risk aversion, generating optimal placement recommendations with high expected employment and less uncertainty.
Timely access to essential supplies is critical in emergency response, underscoring the need for efficient logistical solutions. Motivated by the potential of intelligent logistics and drone delivery in both remote and metropolitan settings, we developed two integer programming models to optimize last-mile deliveries. First, we extended the vehicle routing model to a multi-echelon supply chain involving third-party logistics providers, warehouses, customers, and charging stations. In this model, drones begin their routes from a third-party provider, pick up packages from a warehouse, stop at charging stations when needed, and deliver them to customer locations. Building on this approach for metropolitan areas, we formulated a last-mile delivery system where drones operate in coordination with public transportation to fulfill customer orders. Leveraging public transport vehicles allows drones to charge their batteries on the move or hop on to only get closer to customers, reducing the number of drones required to meet demand. Our extensive numerical analysis demonstrates significant potential for energy savings in drone-enabled last-mile delivery being coupled with the public transportation network.