Skills-based hiring: the evidence and the hype.

“We no longer require a college degree.” Over the past decade that sentence has migrated from contrarian op-eds to the career pages of the world’s largest employers. IBM reframed thousands of openings as “new collar” jobs, and a recognizable roster of technology, finance, and healthcare firms followed with commitments of their own (Fuller, Langer & Sigelman, 2022). From 2022 onward, a wave of U.S. state governments — Maryland, Utah, and Pennsylvania among them — removed the bachelor’s requirement from many civil-service postings.

Behind those announcements sits a genuinely strong research case. In front of them sits an increasingly awkward measurement: when researchers counted actual hires rather than press releases, the behavior change was a small fraction of the announced change. Both halves of that story deserve to be taken seriously, because together they say something precise about what it takes to retire a proxy.

Degree inflation: the original diagnosis

The modern case begins with Dismissed by Degrees, Fuller and Raman’s 2017 Harvard Business School report, which analyzed more than 26 million U.S. job postings (Fuller & Raman, 2017). Its central exhibit is a mismatch the authors called degree inflation: employers demanding bachelor’s degrees for roles whose current occupants mostly do not have one. The signature example is the production supervisor — 67% of postings for the role asked for a college degree, while only 16% of the people actually employed as production supervisors held one.

The report’s employer survey pointed in the same uncomfortable direction. Degree-gated versions of middle-skills roles took longer to fill and cost more to staff, because degree holders commanded a salary premium — yet employers themselves reported little productivity advantage for those hires, alongside higher turnover (Fuller & Raman, 2017). Paying more, waiting longer, and churning faster for similar output is what relying on a bad proxy looks like from the inside.

What the degree was standing in for

None of this is mysterious under classic signaling theory. In Spence’s original model, education can pay off for employers even if it teaches nothing job-relevant, purely because completing it correlates with traits that are otherwise hard to observe (Spence, 1973). A degree is a cheap-to-read signal in a market where the underlying skill is expensive to measure. Proxies of that kind are not abandoned because they are crude; they are abandoned when something cheaper and more accurate arrives.

As a predictor of actual job performance, the credential itself has never ranked well. In Schmidt and Hunter’s synthesis of eighty-five years of selection research, years of education sat near the bottom of the validity table — far below work-sample tests, cognitive-ability measures, and structured interviews (Schmidt & Hunter, 1998). A major re-analysis in 2022 revised many of those validity estimates downward, but preserved the ordering that matters here: direct measures of job-relevant knowledge and behavior dominate distal credentials (Sackett et al., 2022).

The degree proxy also does its damage before any human reads a résumé. Fuller and colleagues’ Hidden Workers report found that the overwhelming majority of large employers use applicant-tracking or recruiting-management software to filter and rank candidates automatically, and that 88% of employers acknowledged these systems screen out qualified high-skills candidates simply because they fail to match the configured criteria (Fuller et al., 2021). The authors estimated a population of more than 27 million “hidden workers” in the United States — people willing and able to work who are systematically excluded by proxy-based filters, with degree requirements prominent among them.

The reset, as announced

Change did show up in the data — first in postings. Analyzing tens of millions of online job advertisements, the Burning Glass Institute documented what it called an emerging degree reset: between 2017 and 2019, employers reduced degree requirements in 46% of middle-skill and 31% of high-skill occupations (Fuller et al., 2022). The report was careful to distinguish structural resets — deliberate, lasting rewrites of role requirements, visible at a set of large named employers — from cyclical resets, the temporary loosening that accompanied pandemic-era labor scarcity.

Writing in Harvard Business Review, Fuller, Langer, and Sigelman read the same data as a genuine and consequential shift, concentrated in middle-skill work and led by firms that had rebuilt job descriptions around skills rather than credentials (Fuller, Langer & Sigelman, 2022). On stated requirements, the degree reset was — and remains — real.

Announcements versus hires

Stated requirements are not hires. In 2024, the Burning Glass Institute and Harvard Business School’s Managing the Future of Work project published the follow-up study, tracking what actually happened at firms after they removed degree requirements from specific roles (Burning Glass Institute & Harvard Business School, 2024). The headline finding: for all the pronouncements, the resulting shift in who gets hired amounted to fewer than one in 700 U.S. hires in 2023.

The texture of the report is as telling as its headline. A minority of firms — the authors call them leaders — changed their postings and then measurably changed whom they hired. But for a large share of roles where the requirement was deleted, the mix of new hires barely moved: the degree line left the job advertisement and kept operating inside the screen. Much of the announced change, as the authors characterize it, was change in name only (Burning Glass Institute & Harvard Business School, 2024).

Why behavior lags the press release

The simplest explanation is that deleting a requirement removes information without adding any. A recruiter facing two hundred résumés and no degree filter still needs some basis for ranking them; absent a better signal, the familiar proxy re-enters — through habit, through hiring-manager preference, or through applicant-tracking configurations that were never updated to match the new posting language (Fuller et al., 2021).

A second explanation is that some of the reset was never conviction in the first place. Modestino, Shoag, and Ballance showed that stated requirements breathe with the business cycle: in the slack labor market after the Great Recession, employers raised education and experience requirements for the same occupations — opportunistic “upskilling” of postings when applicants were plentiful (Modestino, Shoag & Ballance, 2020) — and as the market tightened after 2010, those requirements fell again (Modestino, Shoag & Ballance, 2016). Some fraction of the degree reset is therefore a tight-labor-market phenomenon that could quietly reverse in the next downturn, with no announcement at all.

What the evidence doesn’t show

This literature is often summarized more confidently than it should be, in both directions. Four limits are worth stating plainly:

• Outcome data on non-degree hires is thin. The evidence that workers hired without degrees perform comparably rests heavily on employer self-report and on middle-skills roles (Fuller & Raman, 2017). Large-scale, causal comparisons of post-reset hires — performance, promotion, retention over years — mostly do not exist yet.
• Postings measure stated preferences, not selection. Requirements listed in advertisements move with the business cycle and with posting conventions (Modestino, Shoag & Ballance, 2016), and firms sometimes hired non-degree candidates even while postings demanded degrees. Both the inflation and the reset are partly artifacts of what gets written down.
• The one-in-700 figure is a short-window estimate. It captures near-term behavior at firms that changed posting language (Burning Glass Institute & Harvard Business School, 2024). It cannot rule out slower cultural change, spillovers to firms that never made announcements, or effects that compound over a decade.
• “Skills-based hiring” is not one intervention. The label spans everything from deleting a sentence in a job ad to redesigning selection around validated assessment. Studies of the former say little about the latter, and much public argument mixes the two.

What would make the reset real

Read together, the two halves of this literature are not in tension. The diagnosis — degrees are an expensive, low-validity proxy that excludes millions of capable workers — has held up. The prescription implied by the follow-up data is that proxies persist until they are replaced, not merely renounced. And the selection literature has been unusually consistent about what a replacement looks like: measure the skill directly. Work samples, job-knowledge tests, and structured evaluations sit at or near the top of every validity ranking, while credentials sit near the bottom (Schmidt & Hunter, 1998) (Sackett et al., 2022).

That reframes skills-based hiring as a measurement program rather than a communications program. The firms that moved their hiring numbers paired the deleted requirement with the harder work of defining role skills and assessing them (Burning Glass Institute & Harvard Business School, 2024). The firms that didn’t, deleted a sentence.

The bottom line

Degree inflation is one of the better-documented distortions in modern labor markets, and the case against the degree-as-proxy is strong on both economic and psychometric grounds. The hype is not in the diagnosis. The hype is in treating an edited job posting as a changed hiring process — a conflation the follow-up evidence now measures precisely, and unflatteringly.

The proxy-shaped hole in the hiring funnel has to be filled with actual measurement. Where it is, the reset shows up in hires. Where it isn’t, it shows up only in the press release.