--- title: "Bridging Skill Gaps for the Future: New Jobs Creation in the AI Age" aliases: - "IMF Skill Gaps SDN" - "Jaumotte et al 2026" - "IMF SDN/2026/001" tags: - cross-country - multilateral - skills - polarisation - wages - lightcast - working-paper source: "/research/2026-01-imf-skill-gaps-new-jobs.pdf" extract: "/research/extracts/2026-01-imf-skill-gaps-new-jobs.md" authors: "Florence Jaumotte; Jaden Kim; David Koll; Elmer Z. Li; Longji Li; Giovanni Melina; Alina Song; Marina M. Tavares" publisher: "International Monetary Fund" date: "2026-01" --- # Bridging Skill Gaps for the Future: New Jobs Creation in the AI Age > [!info] Source > PDF (human): [2026-01-imf-skill-gaps-new-jobs.pdf](/research/2026-01-imf-skill-gaps-new-jobs.pdf) · Raw extract (machine): [2026-01-imf-skill-gaps-new-jobs.md](/research/extracts/2026-01-imf-skill-gaps-new-jobs.md) > IMF Staff Discussion Note SDN/2026/001 — January 2026 — Cross-country (advanced + emerging economies), 49 pages > **JEL:** J24, J31, O33 · **ISBN:** 979-8-22902-819-6 ## TL;DR Cross-country analysis of "new skills" demand using Lightcast vacancy data across Brazil, Denmark, Germany, South Africa, the UK, and the US. **About 1 in 10 job postings in advanced economies demand at least one new skill** (about half that in emerging markets). New skills appear first in the US, then spread to other countries. New skills carry a **3–3.4% wage premium** in the US and UK. Critically, **AI-related new skills do not boost overall employment** — and in occupations highly exposed to AI with low complementarity, employment levels are **3.6% lower** in regions with higher AI-skills demand 5 years on. Polarising: benefits high- and (via services consumption) low-skilled workers, **shrinking the middle class**. ## Key findings ### Demand for new skills - ~1 in 10 advanced-economy job postings demand at least one new skill (half that in emerging markets). - New skills appear first in **US labor demand** (especially California) and spread to other countries with a measurable lag. - IT skills account for **>50% of new skills**; AI a growing share. Sector-specific (health-care) skills also growing. - Concentrated in **professional, technical, managerial occupations**. - Driving firms: young, innovative, less financially constrained. ### Wage and employment effects - **All new skills:** posting-level wage premium 3–3.4% in US and UK. In US local labour markets, +1pp share of new-skill postings → +2.3% wage gain, +1.3% employment gain. Germany: +0.9% wage, no significant employment effect. - **Polarisation:** benefits high-skilled workers (direct) and low-skilled workers (via service-sector consumption spillovers). **No significant benefits for middle-skilled workers** — reinforces job polarisation, **potentially shrinking the middle class**. ### AI-specific findings - **AI-related skills post higher wages** at the vacancy level. - **But greater AI-skills demand is linked to no overall employment gains**, and to lower employment for some groups. - **For occupations highly exposed to AI but with low complementarity:** employment levels **3.6% lower** in regions with greater AI-related-skills demand **five years after** the appearance of these skills. - Particularly challenging for: white-collar middle-skilled jobs, young workers, and some IT specialist categories. ### Skill supply - Heavily reliant on tertiary-educated workers, especially STEM and IT graduates. - IT skills found across all fields of study, not just IT graduates. - **Skill Readiness Index** combines: share of recent graduates supplying new IT/non-IT skills, retraining frequency, workforce literacy/numeracy. ### Skill Imbalance Index - Distinguishes economies by relative demand vs supply of new skills. - High-demand / low-supply economies: priority is **expand worker training**, integrate IT across all fields of study, favour labour mobility, strengthen STEM. - Low-demand / high-supply economies: priority is **stimulate innovation, improve access to finance** so firms absorb skills. - IMF AI Preparedness Index can guide policy focus. ### Mergers & Acquisitions - Securing scarce expertise can spur talent-driven M&A ("acquire-hire") — raises concerns about market concentration and skill diffusion. ### Cross-cutting policy needs - Active labour market policies + affordable housing to facilitate worker mobility across occupations and regions. - Firm-union collaboration for adjustment. - Limit non-compete agreements that slow talent diffusion. ## Methodology in brief - **Core:** Lightcast vacancy data (millions of online job postings) across 6 countries (Brazil, Denmark, Germany, South Africa, UK, US). - US vacancy data merged with **American Community Survey** at local labour market level. - German data: administrative records (Sample of Integrated Labour Market Biographies). - US firm characteristics from **Compustat**. - Lightcast worker profiles for skill supply. - Skills Readiness Index combines vacancy data with ILO/OECD employment & education statistics. ## Implications for AdaptAI This is the **strongest cross-country empirical case for caution about AI-related upskilling alone**. The paper finds AI skills boost individual wages but not aggregate employment — which is exactly the warning workers need to hear when planning a career pivot. ### Calculator (`/calculator`) - **High-AI-exposure + low-complementarity** is the paper's risk band — it maps to AdaptAI's High exposure tier (>61%). The 3.6% employment loss after 5 years in those regions is a useful calibration data point. - **Polarisation finding** suggests the result-band copy could distinguish more sharply between "augmented" and "displaced" within the High band. Not currently the case. - **Time horizon:** 5-year delta (+30) directly corresponds to the paper's measurement window — supports keeping the 5-year cap conservative. ### Big Picture (`/big-picture`) - `bigPictureData.skillsDemand.growing` — AI/ML Engineering, Prompt Engineering, AI Ethics & Governance entries are exactly the new-skill categories the paper identifies. Refresh growth percentages with paper's wage-premium framing. - `bigPictureData.skillsDemand.declining` — middle-skill polarisation supports the listed declining skills. - `bigPictureData.industryExposure` — supports the high-exposure ranking. - `bigPictureData.regionalData` — paper's local-labour-market analysis is methodologically similar; the regional impact figures (UK regions) are consistent with cross-country evidence of polarisation. - `bigPictureData.unemploymentByAge.youth16to24` — paper specifically calls out young workers; provides cross-country support. - **Add:** a new key stat about "polarisation / shrinking middle class" framing on `/big-picture` would be supported by this paper. ### Industry Insights - `/industries/professional-services` — paper concentrates new skills in professional/technical/managerial occupations. - `/industries/tech` — paper specifically calls out IT specialists as a category facing employment challenges from AI-skills diffusion. - `/industries/finance` — high adoption + finance among the priority verticals. - `/industries/healthcare` — paper notes sector-specific skills (health-care) are gaining ground; supports the existing "augmentation, not displacement" framing. ### How to Adapt - `/adapt/upskilling` — paper supports upskilling for individual wage gains BUT cautions that AI-skills training alone doesn't generate aggregate employment. The Skill Readiness Index policy framing is useful for the section on national/government action. - `/adapt/career-transitions` — middle-skilled workers face the "no significant benefits" finding. Career transitions guide should be explicit that the polarisation pattern means "moving up" or "moving sideways into low-skill services" are more viable than staying in mid-skill. - `/adapt/rights-protections` — paper recommends limiting non-compete agreements; useful UK-relevant policy point. - `/adapt/human-advantage` — middle-skill cognitive work is the squeeze zone; emphasising hard-to-replicate human skills is exactly what the paper suggests is needed to escape polarisation. ## Related notes - [[BIS European Firm-Level Study]] — firm-side adoption complement; provides the productivity gain (4%) consistent with this paper's wage-premium findings. - [[AI Layoff Trap]] — theoretical counter-argument that upskilling alone cannot fix the demand externality. - [[DSIT UK Labour Market Assessment]] — UK exposure and complementarity figures using the same IMF framework. ## Caveats & limitations - Lightcast vacancy data over-represents formal-sector, online-posted jobs — may miss informal/SME hiring. - Six-country sample; UK is included but findings driven by US data. - "Highly exposed + low complementarity" definition relies on Felten et al. and Pizzinelli et al. exposure indices — not validated against real outcomes. - 5-year measurement window is finite — paper does not address long-run new-task creation that may eventually offset. - Wage premium for new skills (3–3.4%) is at the posting level — does not measure realised wages. - Polarisation conclusion is about the marginal effect of new skills, not absolute middle-class trajectories (which depend on many other factors).