Skills-based hiring means screening candidates on demonstrated ability rather than on a credential that stands in for it. Removing a degree requirement is the announcement, not the method. Unless the screen it leaves behind is replaced with evidence of the work itself, the degree returns informally — in recruiter shortlisting, in interview judgement, in who gets a callback.
That gap between pronouncement and practice is now measurable, and the most recent data suggests it is widening rather than closing.
Where the numbers actually stand
Degree requirements are rising again, not falling. Indeed's Hiring Lab found that 19.3% of US job postings required a bachelor's degree or higher in November 2025, up from 16.6% in November 2023. On a job-mix-adjusted basis — holding the occupational composition of postings at 2019 weights — the share bottomed at 18.5% in March 2024 and had drifted back to 18.6% by November 2025. A slim majority of postings, 51%, state no formal education requirement at all.
The announcements moved very few hires. The Burning Glass Institute and Harvard Business School's Project on Managing the Future of Work reported in February 2024 that skills-based hiring accounted for "not even 1 in 700 hires" in the year studied. Of the firms examined, 37% had genuinely removed degree requirements, roughly 45% were "in name only" with no measurable change in who they hired, and about a fifth were "backsliders" whose gains did not hold. Firms the report classed as leaders raised their share of hires without a bachelor's degree by nearly 20%.
The affected population is large. Opportunity@Work counts more than 75 million US workers as STARs — Skilled Through Alternative Routes — roughly half the workforce, including 64% of rural workers and 61% of veterans.
A practitioner FAQ
Is a degree requirement a selection procedure in the legal sense?
In the United States, yes. The Uniform Guidelines on Employee Selection Procedures define a selection procedure to include educational and work-history requirements, not only tests. Griggs v. Duke Power Co., 401 U.S. 424 (1971), concerned a high school diploma requirement and a general ability test, and held that a neutral-looking requirement with disparate impact must be shown to be job-related. The EEOC's guidance on tests and selection procedures treats education the same way.
How well does education actually predict job performance?
Weakly, as a standalone screen. Schmidt and Hunter's 1998 synthesis in Psychological Bulletin put the validity of sheer years of education at .10, and noted it predicts performance in job training better (around .20) than performance on the job. Ng and Feldman's 2009 meta-analysis in Personnel Psychology, "How broadly does education contribute to job performance?", found education positively related to core task performance, creativity and citizenship behaviours and negatively related to absenteeism — real relationships, but direction rather than the effect size a pass/fail hurdle needs.
What predicts better?
From the revised validity estimates in Sackett, Zhang, Berry and Lievens, published in *Industrial and Organizational Psychology* (2023):
- Structured interviews: .42
- Job knowledge tests: .40
- Empirically keyed biodata: .38
- Work sample tests: .33
- Cognitive ability tests: .31
- Integrity tests: .31
- Assessment centres: .29
- Years of job experience: .07
The practical reading is that a job knowledge test plus a work sample plus a structured interview carries far more information about a candidate than any credential filter — and that swapping a degree requirement for a years-of-experience requirement swaps a .10 predictor for a .07 one.
How many replacement measures do we need?
Two or three, each mapped to a distinct part of the role. The gain from adding a predictor is incremental validity, not the sum of the coefficients, and predictors that overlap heavily add little. Three well-chosen stages typically exhaust the practical gain; a fourth mostly adds drop-off.
Won't candidate volume explode once the filter comes off?
It usually does, which is why the economics matter more than they used to. A per-candidate variable cost makes the arithmetic explicit instead of hiding it in recruiter hours — on AssessAll, assessment credits run at ₹30 / US$0.50 each, with 100 free credits on individual signup and 250 for corporate accounts. The screening design question is where to put the cheap wide net and where to put the expensive deep measure.
How do we stop the degree creeping back in informally?
Treat the roughly 45% "in name only" finding as the base rate for an unmanaged change. Three controls help: remove the education field from what first-stage reviewers see; score the structured interview against behavioural anchors rather than background; and audit the realised share of hires without a degree quarterly, adjusted for the mix of roles you were hiring for, because an unadjusted share moves with your role mix alone.
Do hires without degrees actually perform and stay?
The Burning Glass and HBS report found that at leader firms, workers hired without degrees had retention rates 10 percentage points higher than degree-holding colleagues, and that those hired into roles that previously required a degree saw average salary increases of 25%. Read this carefully: these are observational comparisons at firms that chose to change, so selection effects are plausible and the retention gap is not a causal estimate. It is evidence against the fear that the quality floor collapses, not proof of a performance premium.
Does dropping the degree requirement reduce adverse impact by itself?
No. Removing one screen re-opens the question for whatever replaces it, and cognitively loaded measures carry subgroup differences of their own. Run the impact analysis on the new funnel stage by stage, keep the records, and treat the four-fifths rule in 29 CFR 1607.4(D) as a flag for investigation rather than a clean bill of health.
Is a certificate or microcredential a drop-in substitute?
Not on its own. A certificate records that someone completed a programme; it rarely records how well they performed against a standard, on which tasks, or when. If you accept credentials in place of a degree, accept them as evidence that routes a candidate into assessment, not as a result.
When is keeping the degree requirement the right answer?
When law or licensure requires it, when the credential is the regulated competence, and in fields where the Indeed data shows degree requirements remain dominant — tech, finance, STEM, medical and professional services. Dropping a requirement you cannot defend operationally, and then rejecting non-degree applicants at interview anyway, is worse than keeping it: it wastes applicant time and generates a record of disparate treatment.
What should we measure to know whether anything changed?
- Share of hires without a bachelor's degree, adjusted for role mix, quarterly.
- Stage-by-stage pass rates, disaggregated by subgroup.
- Six- and twelve-month retention, by entry route.
- Performance rating distributions by entry route, once you have two review cycles.
- Time-to-fill and cost-per-hire, so the honest trade-off is visible.
- Interview-stage rejection reasons, scanned for background language.
The takeaway
The evidence does not say credentials are worthless; it says they are a weak predictor being asked to do a strong predictor's job. Removing the requirement is a two-line change to a job posting, and on the published record it moves almost nothing on its own — what moves hiring is replacing the filter with a measurement, and then auditing whether the filter actually left.