social · humanitarian
matched by likeness, missed by need
German volunteer tandems pair refugees with local mentors to build friendships, but in a randomised evaluation only 85 of 234 refugees offered a place were paired, and those with greater needs were less likely to get a mentor
Source: Krieger et al., “Mentoring Programs Support the Integration of Refugees” Open ↗source: primaryLast updated: 2026-11-17
Problem statement
After the 2015–2016 arrivals, German civil society built mentoring programmes that pair a refugee with a local volunteer. The aim is friendship and everyday contact, which are hard to get in a new country. One such programme, Start with a Friend (SwaF), took part in an unusual test. In 2017, interviewers for the national IAB-BAMF-SOEP Refugee Survey offered it to 733 adult refugees in the 14 cities where it was active. 465 said they were interested, and these were randomly split into a programme group (234) and a control group (231). The pairing step is where the offer thinned out. Contact details reached the programme for 215 of the 234. Volunteers met 127 of them, and 85 were paired with a mentor. Those who were paired were younger, more often unmarried and childless, and spoke better German than those who were not. A later analysis of the same programme finds that "refugees with greater needs are less likely to be matched with a mentor" (abstract). The unsolved problem is a way of forming refugee–local tandems that reaches refugees with families, little German or older age, and does not mainly pair people who were already easier to connect.
Why this matters
This kind of programme operates at scale. The DIW report says that the federal "Menschen stärken Menschen" programme, which funds SwaF, had set up more than 100,000 partnerships for refugees by December 2019. Where people were paired, the results were good. In the DIW analysis, the share of refugees with good German rose to 60% in the programme group against 34% in the control group (a gain of 13 against 4 percentage points). The programme group also took part in more leisure activities a year later. The journal paper finds positive effects on social connectedness and housing satisfaction among the 30 most intensively mentored refugees, and less robust effects on German. Most mentors saw their partner as a friend and wanted to keep volunteering. If pairing selects the refugees who were already doing better, these gains do not reach those who most need them. The DIW authors warn that if only refugees with certain characteristics are interested, such programmes "may not be able to reduce existing inequalities within the refugee population."
What’s been tried and why it hasn’t worked
(1) Open offer plus matching on fit. SwaF tries to pair refugees with mentors who suit each other. The DIW authors say this indirectly applies research showing that friendships form more easily between people who are alike in visible traits such as age. Its mentors were mostly German citizens (92%), 73% female and young, 83% had the Abitur (the university-entrance school certificate), and a third were students or trainees. The DIW authors suggest this may be why pairing came more easily for younger refugees with better German. The DIW authors write that "it may be more difficult to find matching partners for older refugees or those with poorer German language skills." The journal paper adds that its results "may not necessarily apply to, for example, refugees with very low language skills as they mostly did not receive a local mentor." (2) A fixed time commitment. SwaF suggests tandems meet weekly for two to three hours for six months. The authors suggest that the under-representation of married refugees and parents among those paired could be related to this time commitment, which could be harder to meet for refugees with families, who often have young children who are also less likely to attend day care. (3) Who asks to join. Refugee women expressed interest less often than men. Interested refugees had arrived more recently and were more likely to live in collective accommodation. The 2025 follow-up finds "limited evidence of self-selection based on needs" at enrolment but "more pronounced selection ... during matching with mentors" (abstract). (4) What the evaluation could measure. Because few in the programme group were paired and not all answered the 2018 follow-up survey, the comparison was restricted to 30 refugees in the programme group and 130 in the control group. In the journal paper, 54 of the 85 paired refugees were re-interviewed, and 30 of them had met their mentor regularly for at least four months; the authors say this is too few for reliable sub-group analysis. No measurable effect on employment was found. The authors suggest the group may be too small or the year too short, and note that the programme aims at friendship, not jobs. The journal paper reports that statistical imbalances between the groups arose during enrolment, matching and panel attrition, which it addresses with propensity-score weighting. The 2025 paper concludes that volunteer mentoring "complements, but cannot replace, more professionalized refugee support programs."
What would unlock progress
A pairing method that treats the hard cases as the main design target instead of leftovers. That means formats that fit around childcare (for example, tandems that include children, or a shorter or more flexible commitment), activities that work with little shared language, and matching rules that weigh need as well as similarity, plus a way to recruit volunteers who fit those refugees. The evaluation design (a randomised offer inside a national panel survey) already exists. A redesigned pairing step could be tested in it against today's similarity matching.
How we’d know it’s being solved
(a) Conversion: the share of refugees offered a place who end up in a working tandem, against 85 of 234 (36%) here. (b) Equity of conversion: pairing rates for parents of young children, women, older refugees and those with limited German, compared with the rest. The gap reported here should close. (c) Persistence: tandems still meeting at six months. (d) Outcomes for those who are paired: social connectedness and German skills, measured as in the SOEP evaluation. The comparison should be by intention-to-treat, so that pairing failures count against the programme. Beware the proxy: "partnerships set up", the headline count, says nothing about who was left unpaired.
Where to start
Facility-free door: using the published pipeline (733 → 465 → 234 → 215 → 127 → 85) and the reported differences between paired and unpaired refugees, build a simple model of the matching step. Compare similarity-first, need-first and mixed rules for a given pool of volunteers like those described (young, mostly female, highly educated), and show what each rule does to pairing rates for parents and for people with limited German. The underlying IAB-BAMF-SOEP microdata are not public; they are released through a SOEP data-use agreement. The model can rest on the published figures alone. Make door: design a tandem activity kit for pairs who share little language and who may bring young children. It could be a set of structured, low-talk activities (a city walk, a shared cooking or repair task, a bus-route practice), with picture prompts and a format that works with a child present. Conversation-card decks and phrasebooks already exist. The team's kit differs because it carries the meeting itself and makes room for children, rather than teaching vocabulary. Test it first with adult pairs who share little language, reached through a campus language-exchange or buddy programme for international students, then with adult newcomers and volunteers reached through a local refugee-support volunteer group or a class for adult language learners, in supervised sessions. Keep the child-present format as a mock-up in these first rounds. Access line: sessions with refugees, and any session with children, need ethics review. What carries over to the German programme: whether the activities carry a meeting with little shared talk, and whether the format fits around a child. What does not: campus pairs may share English as a fallback; students do not face the childcare load, collective accommodation or recent arrival the sources describe; and SwaF's matching step and weekly six-month format are not tested. So a campus result shows the kit can carry a meeting, not that it reaches the refugees who went unpaired. Partner door (design the trial, don't run it): with a mentoring organisation, design a randomised comparison of the current pairing rules and format against a redesigned version. Pre-register the share of offered refugees in a working tandem at six months, and the pairing gap for parents, as primary outcomes. Access note: work with refugees, and any trial, needs ethics review and a partner organisation. Relevant skills: service and matching design, product design, social statistics, migration studies.
Genes
Related problems
Stuck the same way, in other fields.
field: food & the living worldasbestos-free, says the paperworkEvery internationally trading ship of 500 gross tons or more must certify a hazardous-materials inventory by 26 June 2030 — but the inventory rests on supplier "asbestos-free" declarations, IMO itself says asbestos is still found on ships despite the ban, and in most materials only a laboratory can tellbuilt without the people who'd live with itsource: primaryfield: the built worldthe battery eats the cargo spaceHeavy-duty long-haul truck battery weight displaces cargo capacitybuilt without the people who'd live with itsource: secondaryfield: food & the living worldfresh at dawn, half price by duskIn small-scale fisheries the fish is rarely thrown away — it is sold for half price as it warms, and no one can objectively grade the quality being lostbuilt without the people who'd live with itsource: primaryfield: the built worldthe accident reports that never comeRoughly half of recent airline accidents have no published final report, even though the treaty rules ask States to publish one within a year where possiblebuilt without the people who'd live with itsource: primarySources
- Krieger M, Jaschke P, Kroh M, Legewie N, Löbel L-M. "Mentoring Programs Support the Integration of Refugees." DIW Weekly Report 10(49):457–465, 2020, doi:10.18723/diw_dwr:2020-49-1, accessed 2026-11-17. Link ↗
- German original: Krieger M, Jaschke P, Kroh M, Legewie N, Löbel L-M. "Mentorenprogramme fördern die Integration von Geflüchteten." DIW Wochenbericht 87(49):905–914, 2020, doi:10.18723/diw_wb:2020-49-1 (Wayback copy, snapshot 2026-03-10), accessed 2026-11-17. Link ↗
- Jaschke P, Löbel L-M, Krieger M, Legewie N, Kroh M, Jacobsen J, Schacht D. "Mentoring as a grassroots effort for integrating refugees – evidence from a randomised field experiment." Journal of Ethnic and Migration Studies 48(17):4085–4105, 2022, doi:10.1080/1369183x.2022.2058918 (full text via the Wayback copy, snapshot 2022-04-18), accessed 2026-11-17. Link ↗
- Legewie N, Jaschke P, Krieger M, Kroh M, Löbel L-M. "Support for everyone or selection of some? Self-selection and assignment into a large-scale refugee mentoring program in Germany." European Sociological Review 42(1):87–102, 2026 (online 2025-08-25), doi:10.1093/esr/jcaf033 (full text via library access to), accessed 2026-11-17. Link ↗
Revision notes
This brief gets checked against its sources and contains the following revisions. Future revisions get added here.
Worth knowing
- the gains for refugees who were paired come from comparing 30 regularly mentored refugees with 130 in the control group
- the journal paper gives 446 participants in its text and 465 in its table without explaining the gap; the page uses the institute report's figures
Full check
Sources read
- Krieger et al., DIW Wochenbericht on mentoring programmes for refugees (2020): full text (the German original)
- Krieger et al., DIW Weekly Report, the authors' English version (2020): full text
- Jaschke et al., randomised field experiment on refugee mentoring (Journal of Ethnic and Migration Studies, 2022): full text
- Legewie et al., self-selection and matching in a refugee mentoring programme (European Sociological Review, 2026): full text
What changed
- Stated more strongly than the source: the authors' warning that such programmes may not reduce inequalities is now tied, as in the report, to the case where only some kinds of refugees are interested; the link between a fixed weekly time commitment and fewer parents being paired is now given as the authors' suggestion, not a finding.
- Title: the full title said only 85 of 234 were 'ever paired'; no source says the rest never were (the programme had a waiting list), so it now says 'were paired' (display title unchanged).
- Entry points: the make door said campus pairs often share English as a fallback; no source says how often, so it now says they may.
Left open
None noted in this check.
Separate check of specific claims: A later check changed no fact on the page.
Earlier full check: Two statements were brought back to the reports' wording (why younger refugees with better German were paired more easily is the authors' suggestion; interested refugees were more likely, not usually, to live in shared housing), and translation apps, which no source names, were removed from the make door.
Separate check of specific claims: The make door now names its test users, adds an ethics line for sessions with refugees or children, and says what a campus test can and cannot show.
Earlier full check: The German original and the journal paper were read in full, the description of how the programme matches pairs was corrected, and the journal paper's caution about refugees with very little German was added.