Iceland's Employment Policy 2035: A Systematic Blind Spot

The Core Finding

The silence is structural before it is lexical. Iceland's Employment Policy 2035 comes in three documents — the Vaxtarplan til 2035, its Aðgerðaáætlun 2026–2027, and the Stöðugreining that diagnoses the starting position. Between them they commit to 35 numbered actions and nine headline metrics. Not one action funds displacement monitoring, a retraining entitlement for workers whose jobs are automated, or transition support — the nearest thing, ISK 200 million a year for flexible university study pathways, is voluntary upskilling. Not one of the nine metrics tracks employment stability, worker transitions, or welfare inflows, and not one carries a target, a baseline, a stated method, or a date. The plan puts ISK 1,100 million into gervigreind infrastructure and nothing into measuring what that infrastructure does to the workforce.

The vocabulary follows the architecture. Across all three documents, automated text extraction — reproducible in minutes — finds zero occurrences of sjálfvirkni (automation), röskun (disruption), atvinnumissir (job loss), endurþjálfun (reskilling) and umskipti (transition). Gervigreind appears thirteen times, every time as opportunity, capability or infrastructure — never once as a risk to employment. Atvinnuleysi (unemployment) appears four times: three of them looking backwards, at the fall from 8.3% to 3.4% over the 2010s, and one flagging a divergence between Icelandic-born and foreign-born workers. Not once does the word stand in the same breath as the technology.

Reproduce this yourself:

pdftotext -layout data/*Vaxtarplan*.pdf strategy.txt
pdftotext -layout data/*2026-2027.pdf actions.txt
pdftotext -layout data/St*greining*.pdf analysis.txt

The ICT Pattern

Iceland's ICT sector (NACE J, the European industry classification for information and communications) moves on four measures at once, and they point the same way. Annual figures, conservative 2022 baseline:

• Revenue (J62 computer services): +28.1% (2022→2024)
• Employment: flat — 9,029 (2024) to 8,848 (2025) on Hagstofa's full-population tax register, down 2.0% from peak and up 10% on 2019. (The Labour Force Survey shows a steeper fall, 11,400 to 10,100, but publishes ±1,200 and ±1,100 confidence intervals on those two estimates; that difference is not statistically significant. The register is the harder source.) Employment held while revenue grew 28% — which is the productivity story, stated correctly
• Vacancy rate: 3.98% → 0.73% (means of the four quarterly rates, 2022→2025, in the archived Hagstofa extract) — a fall of 82%, the steepest among the sectors in that extract, against 45% for the economy as a whole
• Labour productivity: +10.5% (highest of any sector in Iceland, 2025)

Revenue up, headcount flat, vacancies gone, productivity the highest in the country. That is not what a post-COVID correction looks like — revenue would have fallen too, and it grew 28%. But consistent is not the same as proven, and this analysis does not claim otherwise. A demand correction after an exceptional hiring run, changed subcontracting, mergers, a shift in what the classification counts, and above all the global contraction in technology funding across 2022–2024 all fit the same timing. Manufacturing vacancies fell almost as steeply over the same period (−77%), in an industry where the displacement story does not apply. Something cooled hiring across the tradeable economy, with information and communication at the front of it.

Which of those it was, nobody in Iceland can say — and that is the finding. The signals sit in four separate public tables: JVS00001 (vacancies), VIN10022 (register employment), FYR04101 (turnover), THJ11011 (productivity). Assembling them took a laptop and an afternoon. No published Icelandic work assembles them: of 174 theses on gervigreind in the Skemman repository, one BS thesis touches labour displacement, and none of the eight Rannís grants awarded in the field funds labour-market research. That search covers universities and student work — it says nothing about unpublished analysis inside Hagstofa, Vinnumálastofnun, Seðlabankinn or the ministries, and no claim is made about those. The point is not that the pattern was hidden. It is that no Icelandic institution has a standing duty to look for it.

The Knowledge-Occupation Collapse

"This is not a forecast. This is a record of what has already occurred."

Between 2024 and 2025, the estimated number of employed people in managers, professionals, associate professionals and clerical occupations fell by 9,300 (ISCO groups 1–4, −6.9%), while employment in ISCO groups 5–9 rose by 6,900 (+7.4%). This is a net change in occupational employment; it does not identify abolished jobs or establish an AI effect. The policy prescribes more STEM investment for a knowledge sector that is already contracting. The prescription directly contradicts the data.

Iceland Is Flying Blind

Iceland has gaps in European AI and labour comparisons, although domestic enterprise AI data are now available:

• Eurostat isoc_eb_ai: Denmark 42.0%, Finland 37.8%, Sweden 35.0%, Norway 28.9% — Iceland: surveyed in 2025, with results published on 10 March 2026; 25% reported AI use in specified tasks. Direct comparability with these Eurostat percentages has not been established here
• EWCS 2024 (Eurofound worker conditions survey): 35 countries, Iceland not among them. An Icelandic equivalent would cost, on this project's own costing, ISK 8–12 million. One percent of the plan's gervigreind infrastructure line is ISK 11 million, within that estimated range. It is not funded
• Linked administrative data: Denmark has DREAM+IDA, Sweden LISA, Norway FD-Trygd and Finland Findata. Iceland also has a legal framework: Articles 9 and 13 of Act 163/2007 provide for Statistics Iceland to link registers and grant controlled research access. Whether the specific linked data required by this project are available and can be approved remains to be established.
• Union AI agreements: Denmark (Hilfr2), Norway (LO-NHO), Sweden (Teknikföretagen-IF Metall) — Iceland: none

The parliamentary review of Ríkisendurskoðun’s report identifies differences between Statistics Iceland and Directorate of Labour measures arising from their different definitions and methods, and calls for better coordination.

The Institutional Silence

The AI Action Plan consultation ran from 7 to 21 November 2024 and received twelve submissions. ASÍ, VR, Efling, BSRB, SFR and KÍ do not appear in the government’s respondent list. BHM did submit a response, so this was not complete silence from organised labour. The claim that ASÍ’s 46th congress ignored AI also requires correction: a participating union’s report describes a future-committee proposal addressing automation and AI.

A Varða report published in February 2026, based on nearly 25,000 responses from ASÍ and BSRB members, found that 23% reported actual or expected changes to their jobs from technology or climate change. Among that group, 36% worried about job security, 44% reported increased workload (56% among immigrants), and 55% agreed they had received appropriate training.

Theoretical Framework

This analysis applies two concepts from Beyond Fragmentation (MA thesis, University of Akureyri, 2026) to a live policy document:

• Responsibility Fog: Systematic diffusion of accountability — several major labour organisations did not respond to the AI consultation, although BHM did; the analysis asks who is accountable for gaps in participation.
• Cognitive Debt: A strategy built on international estimates never validated domestically, governing through borrowed statistics rather than observed data. The government's own "130,000 affected workers" projection uses a framework it has not validated for Icelandic conditions.

---

Verification

Findings were run through two independent systems (Manus 1.6 and Claude Code) and three precision corrections were applied. A later audit found what both had missed: the text analysis had been run against the November 2025 consultation deck rather than the strategy published in April 2026. It has been re-run against all three published documents, and the figures on this page are the corrected ones. Two systems agreeing is not verification if both are reading the wrong file — document provenance is checked first now. The correction is in the repository with everything else.

Eight Formal Data Requests

The project’s request register records eight formal information requests sent on 14 April 2026. It records replies from Statistics Iceland on 15 April, the Directorate of Labour on 20 April and the Social Insurance Administration on 27 April. The correspondence and resulting datasets are documented in the research repository.