Re:Match
What does “matching” mean?
The Matching
Data-driven approaches are being increasingly used to address governance challenges due to their effectiveness. Algorithmic matching is one such example.
The term “matching” refers to creating suitable, compatible pairs. In the Re:Match context, a suitable match is made between refugees and cities, after collecting the needs and preferences of both sides, and assigning them in a way that ensures their compatibility to the highest possible extent. For example, if a refugee has specific medical needs for heart disease, and has education and expertise in the IT industry, the algorithm will match the person to a specific city that fulfils, among other needs and preferences, these specific conditions and long-term integration possibilities.
The goal is to improve relocation and reception processes and pave the way into a future co-determined by those affected.
Why Re:Match?
While there exist innovative approaches for matching refugees to geographical locations, the focus rarely is on the individual preferences and competencies of the newcomers or the capacities of municipalities. Re:Match bridges this gap by centering its matching process around this critical data, which is then used to run the algorithm.
About the Algorithm
Together with our partner Pairity, city representatives and protection-seekers, we’ve co-designed and developed a specialized algorithm to implement this data-driven matching. This algorithm includes preference-ranking and optimises the collective welfare of Re:Match participants by assigning best possible matches given the above data.
Such preference-based algorithmic matching serves several important functions:
It centers the individual needs, preferences, and capacities on both sides, while generating the best possible matches for everyone across the cohort.
It ensures fairness without prioritizing a certain participant’s preferences ahead of the overall group welfare, while maximizing the returns from scarce municipal resources.
It allows for analysis of a large corpus of data.
Verified data serves as a baseline for analysing the quality of matches, program satisfaction, and beneficiary integration outcomes.
What kind of data is used?
The Re:Match algorithm finds the best direct match between participating cities and individuals/families within a given participant group (cohort) based on:
Personal data and individually weighted preferences of the protection seekers, like:
cultural and social factors critical for protection seekers overall welfare and integration,
factors determining economic integration, etc.
Municipal data on local conditions, services, and needs, such as:
educational, employment, and training opportunities,
available cultural services,
capacity to support medical vulnerabilities, etc.
The algorithm avoids group-level assumptions and biases about what newcomers need, since individuals with similar characteristics might hold different priorities and aspirations.
Practically, the algorithm simulates all potential assignments to identify the scenario achieving the highest overall quality of matches for an entire cohort. Computationally, a group of ten participant households and 6 cities requires simulating approx. 7 million match scenarios.
If you want to learn more about our algorithm and matching procedure, check out Chapter 4 of our interim evaluation report and our implementation guide on data-driven matching
A research-backed approach
Algorithm-based matching as a tool for migration management is gaining global traction. Interdisciplinary teams in several countries have explored various matching-based migration distribution strategies, including in the areas of resettlement and community sponsorship. A few examples include:
The GeoMatch project (Stanford University and ETH Zurich) has tested its matching tool, which is based on predicting probabilities for integration outcomes, in a Swiss resettlement program.
The Annie™ MOORE machine learning software has been employed by HIAS to recommend optimal placements of arriving refugees across hosting communities in the US.
The Match’In project (Universities of Hildesheim and Erlangen-Nuremberg) developed a tool for German federal states to match and distribute asylum seekers more effectively across municipalities.
Research has consistently underscored matching as an effective approach to migration. Re:Match builds on international experience in this area, while addressing specific gaps, by developing a holistic national and EU-level relocation tool.
As a practical approach, relocation via matching offers EU member states the opportunity to explore an efficient and humane refugee distribution mechanism as a long-term integration strategy.
