Evidence Base

Data Sources
& References

Every figure in this campaign comes from publicly available UK data. This page lists all primary sources used in the economic model, the quiz, and the campaign statistics — with direct links to originals. Nothing is paywalled. Everything is checkable.

Transparency note: The economic model uses simplified parameterisations of Acemoglu–Restrepo (2019, 2022) calibrated to UK-specific data from ONS, IPPR, and DSIT. All model limitations and caveats are documented in the interactive model's Methodology tab. The campaign makes no claim that causality is established between AI adoption and job displacement — the GOV.UK (Jan 2026) assessment describes current evidence as "consistent with AI contributing" but not conclusive.
UK Labour Market (3 sources)
Office for National Statistics · Feb 2026
ONS Labour Force Survey — UK Labour Market Bulletin
Primary
Key figures used: 34.24M employed; 75.0% employment rate (aged 16–64); 2.5M unemployed; average weekly earnings; sectoral employment breakdown. Baseline workforce figure for all displacement calculations.
Office for National Statistics · Sep 2025
ONS Business Insights and Conditions Survey (BICS)
Primary
Key figures used: 23% of UK businesses now using AI; 10% of AI-using firms already replacing roles; 27% expect admin/clerical roles most impacted; 23% expect creative/design roles affected.
Office for National Statistics · 2025
ONS Annual Survey of Hours and Earnings (ASHE)
Primary Model Input
Key figure used: Average annual wage £38,000 — used as base in fiscal gap calculation (ΔTax = D × avg_wage × eff_tax_rate).
AI Occupational Exposure (3 sources)
DSIT / Department for Science, Innovation & Technology · Nov 2023
The Effects of AI on UK Jobs and Training
Primary Model Input
Key figures used: Artificial Intelligence Occupational Exposure (AIOE) scores mapped to UK SOC 2010. Finance & Insurance identified as #1 most exposed sector. Used as baseline for sector-level displacement estimates in the model.
IPPR — Institute for Public Policy Research · March 2024
Transformed by AI: How Generative AI is Changing the UK Economy
Primary Model Input
Key figures used: 22,000-task analysis of UK labour market. Wave-1 automation exposure: 11% of tasks. Wave-2 exposure: 59% of tasks. Three displacement scenarios (optimistic/central/pessimistic) — central scenario used as model default (α = 35%, δ = 45%). The 7.9M "at risk" headline figure.
Parliament POST (Parliamentary Office of Science & Technology) · Feb 2026
POST Briefings on Artificial Intelligence & Employment
Policy
Balanced review of AI employment evidence for MPs and peers. Concludes evidence is mixed but notes "partial automation" as primary near-term risk rather than full job elimination.
Government Policy Assessments (3 sources)
GOV.UK / DSIT · Jan 2026
AI and the Labour Market — Evidence Review
Primary
Key figures used: UK job postings in high-exposure occupations fell 38% faster than low-exposure roles between 2022–2025 (McKinsey data, UK-specific cut). Caution: GOV.UK notes "consistent with AI contributing" but causality not established. Used as supporting evidence, not proof of displacement.
DSIT · 2024
DSIT AI Sector Study 2024
Primary
Key figures used: 3,256 UK AI firms; 58% company growth YoY; £6B venture capital raised in 2025. Token market size estimate partially derived from UK share of AI cloud spend in this dataset.
OBR — Office for Budget Responsibility · Nov 2025
Economic and Fiscal Outlook
Model Input
Key figure used: £440B annual income tax and NI base — used as denominator for fiscal gap calculation. Effective tax rate of 28% derived from OBR projections on labour income.
Financial & Economic Analysis (3 sources)
PwC UK · 2023–2024
The Impact of AI on UK Jobs and GDP — PwC UK Report
Primary
Key figures used: England GDP +10.6% from AI adoption by 2030. Scotland/Wales +8.4–9.8%. Northern Ireland +5.4%. Productivity Effect calibration in the GDP impact model section.
Goldman Sachs Global Investment Research · 2025
Generative AI Could Raise Global GDP by 7%
Model Input
Key figures used: +15% global labour productivity potential; +7% global GDP by 2030; GDP inflection point projected at 2027. Used to calibrate the upper bound of productivity effect in the model.
McKinsey Global Institute · 2022–2025
Job posting decline data cited within GOV.UK (Jan 2026) assessment
UK-specific cut of McKinsey's longitudinal job posting analysis, showing −38% in high-exposure occupations vs −21% in low-exposure roles (2022–2025). Accessed via GOV.UK evidence review rather than directly — no standalone UK publication.
Theoretical Framework (3 sources)
Acemoglu & Restrepo · Journal of Economic Perspectives · 2019
Automation and New Tasks: How Technology Displaces and Reinstates Labor
Model Input
Foundation of the model. Introduces displacement effect, productivity effect, and reinstatement effect decomposition. Establishes that automation always reduces labour income share through displacement, but creates new tasks where labour has comparative advantage. The model's ΔE = R − D equation directly implements this framework.
Acemoglu & Restrepo · Econometrica · 2022
Tasks, Automation, and the Rise in US Wage Inequality
Model Input
Introduces the concept of "so-so automation" — technology that automates tasks without generating large productivity gains — as the primary risk in current AI transition. This is the key rationale for the fiscal gap: displacement cost is front-loaded, productivity gains are deferred.
World Economic Forum · 2025
Future of Jobs Report 2025
Figure used: +170M new AI-adjacent roles globally by 2030. Used as upper-bound input for the reinstatement rate (ρ) default value of 40% — representing WEF's "moderate reinstatement" scenario applied proportionally to UK workforce.
Data Estimates & Assumptions (4 sources)
Token Market Size — Derived Estimate
UK Commercial AI Token Market (2026 est.)
Estimated ±50%
No official UK data exists. The £8–12B 2026 estimate is derived from: UK share of global cloud compute spend (~5%) applied to global AI API cost projections (OpenAI, Anthropic, Google published pricing × estimated enterprise volume). Growing at 35%/yr (DSIT AI sector growth rate). Wide confidence interval — ±50%. The levy yield range in the model reflects this uncertainty.
Effective Tax Rate — Derived
Labour Income Effective Tax Rate (28%)
Model Input
Blended effective rate across income tax (basic rate 20%, NI 8%) on average UK wage of £38,000, net of personal allowance (£12,570). Calibrated to OBR Nov 2025 total income tax + NI receipts divided by ONS employed workforce × average wage.
Labour Share of Income — ONS / OBR
Labour Share of UK National Income (s_L = 0.57)
Model Input
UK labour share of gross value added, approximately 57–59% over the past decade (ONS National Accounts). Used in the Acemoglu–Restrepo GDP decomposition equation.