Cited by 130Labor market dynamics and wage inequalityHigher Education and EmployabilityEmployer Branding and e-HRM

The DOI matches its Crossref record; 51 of the 51 references it cites were found in a registry or an index; 3 of the 36 sources its claims cite were checked for support.

Skills or degree? The rise of skill-based hiring for AI and green jobs

Open access
Bone, Matthew, González Ehlinger, Eugenia, Stephany, Fabian·2025·Technological Forecasting and Social Change

Abstract

Emerging professions in fields like Artificial Intelligence (AI) and sustainability (green jobs) are experiencing labour shortages as industry demand outpaces labour supply. In this context, our study aims to understand whether employers have begun focusing more on individual skills rather than formal qualifications in their recruitment processes. We analysed a large time-series dataset of approximately eleven million online job vacancies in the UK from 2018 to mid-2024, drawing on diverse literature on technological change and labour market signalling. Our findings provide evidence that employers have initiated “skill-based hiring” for AI roles, adopting more flexible hiring practices to expand the available talent pool. From 2018 to 2023, demand for AI roles grew by 21 % as a proportion of all postings (and accelerated into 2024). Simultaneously, mentions of university education requirements for AI roles declined by 15 %. Our regression analysis shows that university degrees have a significantly lower wage premium for both AI and green roles. In contrast, AI skills command a wage premium of 23 %, exceeding the value of degrees up until the PhD-level (33 %). In occupations with high demand for AI skills, the premium for skills is high, and the reward for degrees is relatively low. We recommend leveraging alternative skill-building formats such as apprenticeships , on-the-job training, MOOCs , vocational education and training, micro-certificates, and online bootcamps to fully utilise human capital and address talent shortages.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

Using ~11 million UK online job vacancies (2018–mid-2024), the paper finds employers increasingly use skill-based hiring for AI roles, with AI skills carrying a 23% wage premium and degree requirements declining 15%, while green roles show rising demand but weaker skill premia. It concludes AI hiring is shifting from degrees to skills, but green hiring has not yet shown the same shift.

Key findings

  • AI postings grew by 21% as a proportion of all postings from 2018–2023 and accelerated into 2024; green postings grew by 67%. Mentions of university education requirements for AI roles declined by 15% from 2018–2023.
  • AI skills command a wage premium of 23%, second only to a PhD (33%) and higher than a master’s degree (13%); green skills do not exhibit a significant wage premium. For AI roles, the education premium nearly disappeared—only a slight master’s premium and no significant PhD premium—while for green roles the educational gradient flattened.
  • AI roles are more skill-intensive: average 15–16 skills per posting, versus about 10 for green roles and about 5 for the broader labour market. AI roles average around £50k, green roles trend toward £40k, and average postings remain slightly below £30k.
  • For AI skills, wage premiums increase with demand and degree premiums decrease with demand; for green skills, the relationship between demand and premium is flat.

Methods

  • Dataset: Lightcast/Burning Glass online job vacancy database covering ~11 million UK vacancies from January 2018 to June 2024, with around 80 million associated skills; advertised salary information is available for about 39% of the sample.
  • Classification: a bottom-up, skill-based approach labels a vacancy as AI-related if it contains at least one AI skill from Lightcast’s AI/ML subcategory (157 skills) and green if it contains at least one green skill from a 2019 list (259 still in the taxonomy).
  • Analysis: descriptive time series plus OLS regressions of log advertised wage on AI/green skill dummies, education dummies, experience, and fixed effects for years, 12 regions, 16 industries, and 9 occupational groups; additional occupation-by-year regressions relate skill and degree premiums to demand.

Limitations

  • Stated in the paper: online job vacancy data cannot fully reveal candidate selection; employers may consider criteria beyond postings; online vacancies miss other talent channels such as contractors or in-house development; OJVs are biased toward certain occupations, sectors, and industries; only advertised salaries are observed, not actual negotiated pay; London and business administration/support services are overrepresented, while motor trades, wholesale/retail, and elementary occupations are underrepresented.
  • Stated in the paper: salary information is available for only about 39% of the sample, though AI (31%) and green (37%) postings are as likely as other vacancies to contain wage information.
  • Evident from the text: the AI/green classification uses a threshold of at least one skill, which is broad; the green skill list was compiled in 2019 and may miss newer green skills; the regression results are associational and do not establish causality.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1016/j.techfore.2025.124042

Cites

51 of 51 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1. Panel A1 shows an increasing share of job postings requiring AI and green skills. Panels A2 and A3show the trends of the most common underlying skills which reveals that techniques like generative adversarial networks are increasingly important for AI/ML roles and skills related to waste management and renewable energies for green roles. For AI roles, the generic skills "Artificial Intelligence" and "Machine Learning" are excluded. Panel B1 highlights that AI roles are three times more skill-intensive, and green roles twice as skill-intensive, compared to the labour market average. Panel B2 illustrates that AI roles command wages around 75% higher than the average, while green roles see slightly higher wages, with growth accelerating from 2022. The strongest growth in skills is observed in renewable energy, waste removal, nuclear power, and machine learning technologies like generative adversarial networks. Panel C1 shows the fastest growing skills in each category between 2018-2023 with at least 100 related posts in 2018.
Figure 1. Panel A1 shows an increasing share of job postings requiring AI and green skills. Panels A2 and A3show the trends of the most common underlying skills which reveals that techniques like generative adversarial networks are increasingly important for AI/ML roles and skills related to waste management and renewable energies for green roles. For AI roles, the generic skills "Artificial Intelligence" and "Machine Learning" are excluded. Panel B1 highlights that AI roles are three times more skill-intensive, and green roles twice as skill-intensive, compared to the labour market average. Panel B2 illustrates that AI roles command wages around 75% higher than the average, while green roles see slightly higher wages, with growth accelerating from 2022. The strongest growth in skills is observed in renewable energy, waste removal, nuclear power, and machine learning technologies like generative adversarial networks. Panel C1 shows the fastest growing skills in each category between 2018-2023 with at least 100 related posts in 2018.CC-BY · From the PDF
Figure 2 : Figure 2: Panel A1 illustrates the wage premium associated with higher education across all roles, with PhD holders earning the highest salaries, reaching around £50k by the end of the time series, while job postings without degree requirements offer significantly lower wages around £25k. Panel A2 shows a different trend for AI roles, where the highest-paying jobs are associated with master's degrees, but wages for roles with no formal education requirements
Figure 2 : Figure 2: Panel A1 illustrates the wage premium associated with higher education across all roles, with PhD holders earning the highest salaries, reaching around £50k by the end of the time series, while job postings without degree requirements offer significantly lower wages around £25k. Panel A2 shows a different trend for AI roles, where the highest-paying jobs are associated with master's degrees, but wages for roles with no formal education requirementsCC-BY · From the PDF
Figure 3 . Figure 3. (A1) Shows the AI, Green, and Degree premiums at the SOC1 level controlling for experience, year, SIC1, and NUTS1. For Professionals and Management jobs, AI skills have a significantly higher premium than degrees.
Figure 3 . Figure 3. (A1) Shows the AI, Green, and Degree premiums at the SOC1 level controlling for experience, year, SIC1, and NUTS1. For Professionals and Management jobs, AI skills have a significantly higher premium than degrees.CC-BY · From the PDF
Figure A1 . FigureA1. When do we define OJVs as related to AI or green? For our analysis, we label a posting as AI or green once it asks for at least one skill from the respective domain. The reason for this decision is that the number of labelled postings would decline quickly if one were to set a higher threshold of two or more skills, see panel A. On the other hand, we notice that for both AI and green postings, the educational requirements (B) and the offered wages (C) increases with the number of respective skills required in the OJVs.
Figure A1 . FigureA1. When do we define OJVs as related to AI or green? For our analysis, we label a posting as AI or green once it asks for at least one skill from the respective domain. The reason for this decision is that the number of labelled postings would decline quickly if one were to set a higher threshold of two or more skills, see panel A. On the other hand, we notice that for both AI and green postings, the educational requirements (B) and the offered wages (C) increases with the number of respective skills required in the OJVs.CC-BY · From the PDF
Figure A2 . FigureA2. When we increase the number of required skills to define an AI or Green OJV, we see decreasing premiums for education (that is, requiring an HNC/HND or higher) and an increasing skill premium for AI roles.The premium for green roles stays relatively stable as the definition becomes more stringent.
Figure A2 . FigureA2. When we increase the number of required skills to define an AI or Green OJV, we see decreasing premiums for education (that is, requiring an HNC/HND or higher) and an increasing skill premium for AI roles.The premium for green roles stays relatively stable as the definition becomes more stringent.CC-BY · From the PDF
Figure A3 . Figure A3. Roughly 38.6 percent (4,187,603) of all job postings in our analysis contain wage information. Wages are log-normal distributed and slightly right-skewed. Therefore, we consider the log of wages for the subsequent analysis.
Figure A3 . Figure A3. Roughly 38.6 percent (4,187,603) of all job postings in our analysis contain wage information. Wages are log-normal distributed and slightly right-skewed. Therefore, we consider the log of wages for the subsequent analysis.CC-BY · From the PDF
Figure A4 . FigureA4. This shows the AI/ML and Green premiums across degree requirements. It shows that AI premiums are consistently positive with the potential exception of HNC/HND-level postings. The premium for Green skills is less clear for Bachelor's and Master's degrees and becomes negative for PhDs. This could be due to many environmental researchers holding lower-paying academic positions.
Figure A4 . FigureA4. This shows the AI/ML and Green premiums across degree requirements. It shows that AI premiums are consistently positive with the potential exception of HNC/HND-level postings. The premium for Green skills is less clear for Bachelor's and Master's degrees and becomes negative for PhDs. This could be due to many environmental researchers holding lower-paying academic positions.CC-BY · From the PDF
Figure A5 . FigureA5. This is a job advertisement from the Greater London area. Information about the salary (highlighted in red), required education (blue), or skills (orange) had been extracted from all parts of the text of the posting, including headlines.
Figure A5 . FigureA5. This is a job advertisement from the Greater London area. Information about the salary (highlighted in red), required education (blue), or skills (orange) had been extracted from all parts of the text of the posting, including headlines.CC-BY · From the PDF
Image 9
CC-BY · From the PDF
Image 10
CC-BY · From the PDF
Image 11
CC-BY · From the PDF
Image 12
CC-BY · From the PDF
Image 13
CC-BY · From the PDF
Image 14
CC-BY · From the PDF

Verification

Claims the paper makes

105 Own result · 28 Own interpretation · 21 Sources not found · 12 Own method · 9 No basis given · 2 Partly supported · 2 Not checked yet · 1 Couldn’t confirm · accuracy not yet measured

Show all 104 claims

  • Probably not a substantial claim. 21
  • It may state two separate claims as one. 62
76 of 76
  • The authors state this as an assumption.

    AI and green skills contribute positively to firms' profitability and are progressively being demanded by employers while remaining relatively scarce in supply.

    “assuming that AI and green skills contribute positively to firms' profitability and are progressively being demanded by employers while remaining relatively scarce in supply, we anticipate an increase in their market value, leading to wage premium offerings and a limited focus on candidates' qualifications for AI and green jobs as well as an adoption of skills-based hiring for jobs in these”
  • The paper relies on this without support.

    Posted salaries serve as an indicator of companies' willingness to compensate for specific skill sets demanded in job postings.

    “Although these posted salaries may not reflect the accurate remuneration received by hired employees, they serve as an indicator of the companies' willingness to compensate for specific skill sets demanded in the job postings.”
  • The paper relies on this without support.

    A wide range of countries are facing similar skills mismatches.

    “Furthermore, given that this is not an isolated problem and a wide range of countries are facing similar skills mismatches, we can expect the increasing competitiveness to trigger a global competition for human capital, making the issue even more urgent.”
  • The paper relies on this without support.

    Job vacancy postings accurately reflect employers' demand for skills and their hiring practices.

    “By analysing job vacancies with the lists of skills, we found that employers are looking for a wide variety of skills related to the development and implementation of AI.”
  • The paper relies on this without support.

    The UK job vacancy data is representative of the UK labour market.

    “Our analysis of eleven million OJVs reveals a rising demand for AI and green skills in the UK, reflecting the continuous efforts of firms to adopt AI -and the AI-human complementarity-and meet environmental sustainability targets.”
  • The paper relies on this without support.

    The demand for green and AI skills far exceeds supply, making it hard for employers to find candidates with the needed skills.

    “As mentioned above, the literature suggests that the demand for green and AI skills is far exceeding supply, making it hard for employers to find candidates with the skills needed to navigate societal and technological change.”
  • The paper relies on this without support.

    Employers place weight on university degrees.

    “Given the weight placed by employers on university degrees, a good avenue for doing so might be through higher education.”
  • The paper relies on this without support.

    Education and industry will keep evolving at uneven paces.

    “Given that education and industry will keep evolving at uneven paces, employers should progressively focus on candidates' skills rather than on their formal qualifications.”
  • The paper relies on this without support.

    Employers will not immediately respond to a tight labour market by increasing offered wages.

    “We used a two-year lag variable as employers will not immediately respond to a tight labour market by increasing offered wages”
9 of 9

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 47 references

Research works printed but not linked: 4

  1. Cambridge Industrial Innovation Policy
  2. Fuller, 2022, Skills-based hiring is on the rise, Harv. Bus. Rev., 11, 1
  3. Illanes, 2018, 8(1), 1
  4. Lancaster, V. A., Mahoney-Nair, D., and Ratcliff, N. (2021) Review of burning glass job-ad data. Proceedings of the biocomplexity institute 2021-013, biocomplexity institute.

Cited by

Sorted by citations

Items 101–128 of 128

Metadata

Authors
Bone, M., González Ehlinger, E., & Stephany, F.
Affiliations
  • University of Oxford (GB)
  • Bruegel (BE)
Corresponding author
Stephany, Fabian
Venue
Technological Forecasting and Social Change
Publisher
Elsevier BV
Published
2025-05
Volume
214
Pages
124042
ISSN
0040-1625 (ISSN-L)
OpenAlex
W4407964516
Cited by
130
References
47
Citations by year
2026: 96 · 2025: 32
FWCI
92.06
Type
journal article
Language
en
Open access
Locations
  • Technological Forecasting and Social Change (publishedVersion, cc-by)
  • RePEc: Research Papers in Economics (submittedVersion)
  • arXiv (Cornell University) (submittedVersion, cc-by)
  • arXiv (Cornell University) (cc-by)
Topics
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Labor market dynamics and wage inequality
  • Social Sciences › Education › Higher Education and Employability
  • Social Sciences › Business, Management and Accounting › Organizational Behavior and Human Resource Management › Employer Branding and e-HRM
Keywords
degree requirements, green jobs, labor shortage, job vacancies, wage premium, emerging professions, human capital, vocational education and training, apprenticeship, MOOCs, UK labour market

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-06, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-06 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review