Dynamic Refugee Placement in Government Assisted Refugees’ Settlement in Canada

This project develops a dynamic placement model to improve the settlement of government-assisted refugees (GARs) in Canada. Each year, thousands of GARs are designated to communities across the country through the Resettlement Assistance Program, with outcomes such as employment, education, and retention strongly influenced by initial placement. While current methods optimize placement based on available capacity and immediate outcomes, they do not account for the arrival of future refugees or long-term impacts. This study introduces a model that combines machine learning predictions with dynamic optimization to balance present and future placement decisions. Using publicly available IRCC resettlement data and Statistics Canada’s Longitudinal Immigration Database, the research will validate the model and provide evidence to inform refugee resettlement policy, aiming to improve socioeconomic outcomes and reduce pressures associated with secondary migration.

1. How decision outcomes, such as employment rate, can be incorporated in an optimization model for placement of refugees in Canada?

2. How can a dynamic placement model create a balance between decision outcomes and the lost employment opportunities for upcoming refugees caused by consuming communities’ capacities for the current refugees?

A supervised machine learning model will be developed to predict expected employment outcomes within three months of arrival, using refugee characteristics such as age, gender, and language knowledge. These predictions will feed into a linear integer programming model that optimizes placements given community constraints. To address future arrivals, dual variables will be used to estimate the opportunity cost of consuming community capacity. Model verification and validation will use IRCC’s Resettled Refugees datasets and Statistics Canada’s Longitudinal Immigration Database.

The project is ongoing. A literature review has been completed, identifying significant gaps in existing refugee placement models. Work is now progressing on model development.

Expected completion date: December 2026

Refugee Resettlement, Dynamic Optimization, Machine Learning, Employment Outcomes, Secondary Migration