The Hidden Labor of Relocation
What skilled migrants do before the job offer exists — mapped through official data from ILO, Eurostat, IND & the EU AI Act.
Overview
This data story makes visible the unpaid, uncounted work that skilled migrants do before a job offer exists. While relocation is typically shown as a clean sequence — visa approved, job secured, life begins — the real journey starts months or years earlier, in a parallel research project that runs alongside the job search itself.
Using the Netherlands as the focal case — one of the most sought-after destinations for skilled migrants in Europe — the story maps the gap between what hiring systems are designed to process and what international candidates are actually doing. Official data from the ILO, Eurostat, European Commission, and IND reveals that this gap is not a personal struggle: it is a structural feature of how hiring and immigration systems were built, for whom, and on what assumptions.
The story also examines how AI has entered this system — scaling both its efficiency and its inherited patterns — and how the EU AI Act’s 2026 compliance deadline creates a concrete opportunity to recalibrate. Relocation is framed throughout as a user journey: one that has clear friction points, a mappable service blueprint, and significant room for intentional design.
Methodology
The story draws exclusively on published, official data sources — no primary polling was conducted. Sources were selected for geographic specificity (Netherlands and EU), recency (2024–2026), and institutional authority.
Data was gathered from six institutional sources: the International Labour Organization (ILO), Eurostat’s Migrant Integration Statistics, the European Commission’s residence permit dataset, the IND public register of recognised sponsors, the OECD International Migration Outlook, and CaixaBank Research’s EU labour market analysis.
AI hiring bias data was sourced from three independent research bodies: ResumeBuilder (2025 employer survey, n=948), CoverSentry (2026 aggregated survey data), Gartner (July 2025, n=2,918), the University of Washington / Brookings Institution (3M+ resume comparisons), and Stanford HAI’s replication of GPT detector bias research (2023, confirmed 2025).
Legal compliance data was drawn from the EU AI Act (Regulation 2024/1689), the DLA Piper GENIE briefing on the AI Omnibus (April 2026), California Civil Rights Council regulations (effective October 2025), and the Mobley v. Workday NDCA class certification filing (May 2025).
Relocation was mapped as a nine-section narrative journey. A funnel visualisation models the application attrition pattern for international candidates based on published callback rate research. An emotional resilience chart models the momentum arc of a 12-month skilled migrant job search. A service blueprint diagrams the three-layer hiring system: candidate actions, employer actions, and system infrastructure. An EU AI Act compliance timeline visualises the regulatory sequence from 2024 to 2027. Visualisations were built as a self-contained responsive HTML page with inline SVG and Chart.js, with no external data dependencies.
Findings
- 167.7 million international migrant workers are active globally — 4.7% of the entire world labour force — yet non-EU citizens in the EU face an unemployment rate of 12.3%, nearly 2.5× higher than nationals at 5.1%. (ILO 2024; Eurostat 2024)
- 50%+ of all new EU jobs created between 2019–2024 were filled by non-EU workers, despite non-EU citizens making up just 6.6% of the EU labour force — structural demand that the hiring system does not yet reflect in process design. (CaixaBank Research 2025)
- 39.6% of non-EU citizens in the EU are overqualified for the jobs they hold — a “brain waste” rate nearly double that of EU nationals at 21%, and one that has only narrowed by 6 percentage points since 2014. (Eurostat 2024)
- In the Netherlands, a Highly Skilled Migrant applicant must secure employment with one of 9,000+ IND-recognised sponsors, cross-referenced against every job posting, before a visa application can be made. (IND Public Register 2026)
- A meta-analysis of 123 resume studies across 18 countries found ethnic discrimination in hiring in 95%+ of cases; a King’s College London study of 12,000+ applications found callback rates of ~27% for local-sounding names versus ~11% for non-local names — on identical resumes. (KCL / The Conversation 2021–2024)
- By 2025, 83% of companies use AI to screen resumes, up from 48% the prior year; 50% of candidates are now assessed by AI before any human reviews the application. 67% of companies acknowledge their AI tools could introduce bias. (ResumeBuilder 2025; CoverSentry 2025)
- Non-native English writing is flagged as AI-generated by leading AI detectors 61% of the time — versus ~2% for native English writing — because second-language sentence structures match AI output patterns. (Stanford HAI 2023, replicated 2025)
- Only 26% of job seekers trust AI hiring systems to evaluate them fairly. (Gartner, July 2025)
- The Mobley v. Workday class action (certified May 2025) disclosed that Workday’s AI tools have rejected approximately 1.1 billion applications since 2020. (NDCA 2025)
- The EU AI Act classifies AI hiring tools as high-risk systems. Full compliance obligations — mandatory bias audits, human oversight, candidate transparency — take effect 2 August 2026, with penalties up to €35M or 7% of global turnover from 2027. (EU AI Act / DLA Piper 2026)
Takeaways
The hidden labor of relocation is real, measurable, and currently invisible to the systems that assess candidates. A skilled migrant preparing to apply to roles in the Netherlands performs approximately four times as many pre-application tasks as a local candidate — cross-referencing sponsor lists, verifying salary thresholds, completing language certification, rebuilding portfolios for local norms — none of which appears on a resume or in an ATS record.
The data reveals a system design problem, not a talent problem. The 2.5× unemployment gap between non-EU citizens and nationals, the 39.6% overqualification rate, and the 27% vs 11% callback rate disparity are not random: they are outputs of hiring processes built around assumptions of locality, native language, and name recognition that have been inherited rather than examined.
AI has scaled this system without correcting it. The 61% AI-detector false-positive rate for non-native writing, the name-preference patterns in AI resume screening, and the 50% pre-human-review rejection rate represent a new layer of friction — one that is now subject to mandatory regulatory correction under the EU AI Act from August 2026.
The opportunity is clear: the friction points are identifiable, the regulatory framework is in place, and the demand for international talent is structurally real. Employers and system designers who treat this as a design brief — rather than a compliance task — will access the talent pool their competitors are systematically excluding.
Data Sources
Ratings use the ODON Open Data Maturity Model (ODMM).
-
ILO — Global Estimates on International Migrants in the Labour Force (December 2024)
Legal L4 Published under ILO open data terms — free reuse with attribution. L4.
Technical T4 Structured dataset with full documentation, direct download, and statistical methodology notes. T4.
-
Eurostat — Migrant Integration Statistics: Labour Market Indicators (2024)
Legal L4 Free reuse with attribution under Eurostat open data policy. L4.
Technical T4 Available via Eurostat API in SDMX, CSV and TSV with full metadata. T4.
-
Eurostat — Migrant Integration Statistics: Over-qualification (2024)
Legal L4 Free reuse with attribution. L4.
Technical T4 Structured download available via Eurostat data browser with API access. T4.
-
European Commission — First Residence Permits by Reason, Citizenship and Age (2024)
Legal L4 Published under EU open data policy — free reuse with attribution. L4.
Technical T3 Tabular data available via Eurostat; requires aggregation across citizenship groups. T3.
-
IND (Netherlands Immigration and Naturalisation Service) — Public Register of Recognised Sponsors (2026)
Legal L2 Publicly accessible list; no download API; terms of reuse not explicitly stated. L2.
Technical T2 Browsable web register only, updated monthly; no bulk export available. Cross-referenced via DutchReview summary (May 2026). T2.
-
OECD — International Migration Outlook 2024
Legal L3 Report freely accessible; underlying microdata proprietary to OECD member states. L3.
Technical T2 Published as PDF and web summary; headline figures used as cited in report. T2.
-
CaixaBank Research — The Changing European Labour Market (June 2025)
Legal L2 Summary report publicly accessible; full model and dataset not open. L2.
Technical T2 Published as PDF/web only; no structured download available. T2.
-
ResumeBuilder — AI in Hiring: 2025 Employer Survey (n=948 business leaders)
Legal L2 Survey summary publicly accessible; microdata proprietary. L2.
Technical T2 Published as web article only; no structured download. T2.
-
Gartner — Candidate Trust in AI Hiring (July 2025, n=2,918)
Legal L2 Press release and summary publicly accessible; full report gated. L2.
Technical T2 Headline figures used as cited in official press release. T2.
-
Brookings Institution / University of Washington — Gender, Race and Intersectional Bias in AI Resume Screening (Kyra Wilson & Aylin Caliskan, 2025)
Legal L3 Research paper freely accessible via Brookings; underlying experimental data not publicly released. L3.
Technical T3 Findings reported in academic paper format; 3M+ resume comparisons described in methodology. T3.
-
Stanford HAI — GPT Detectors Are Biased Against Non-Native English Writers (Liang et al., 2023; replicated through 2025)
Legal L3 Research paper freely accessible; experiment data not publicly released in raw form. L3.
Technical T3 Results reported across 7 AI detectors with described methodology; widely cited and independently replicated 2024–2025. T3.
-
EU AI Act (Regulation 2024/1689) — High-Risk AI in Employment, Articles 5 & 6 (2024–2026)
Legal L4 EU legislation — freely accessible and reusable with attribution. L4.
Technical T3 Structured legal text; compliance dates and penalty figures drawn from official regulation and DLA Piper GENIE briefing (April 2026). T3.
-
Mobley v. Workday Inc. — Class Certification Order, NDCA (May 2025)
Legal L2 Court filing publicly accessible via PACER and CourtListener. L2.
Technical T2 Case documents available in PDF; 1.1B figure drawn from Workday's own filings as cited in class certification order. T2.
-
KCL / The Conversation — 123-Study Meta-Analysis of Resume Discrimination (2021–2024)
Legal L2 Research summary publicly accessible; underlying 123 studies vary in openness. L2.
Technical T2 Meta-analytic findings reported in summary form; 12,000-application KCL study figures used as representative data point. T2.
Tools Used
- HTML
- JavaScript
- SVG
- Chart.js
- Python