Dropping a Degree Requirement Does Nothing Until Something Replaces It

What happened when employers actually dropped degree requirements?

Almost nothing, at the level of the labor market. The Burning Glass Institute and Harvard Business School matched 11,332 roles at large firms that dropped a degree requirement against the career histories of roughly 65 million US workers, and found the net change in hiring of candidates without a bachelor's degree was 0.14 percentage points: about 97,000 workers a year out of 77 million hires ¹.

Two numbers inside that study point in opposite directions and both are real. Within the roles that changed, the share of hires without a degree rose about 3.5 percentage points, which is a genuine effect. But only 3.6% of roles dropped a requirement at all, so the aggregate barely moved ¹. Quoting either number alone misleads: the first oversells reach, the second undersells what happened inside the roles that did change.

The split between firms is the finding worth acting on. Nearly all the real hiring change came from the 37% of firms the authors class as leaders, which raised their share of non-degree hires in the analyzed roles by nearly 20% in relative terms. About 45% of firms changed the posting with no meaningful difference in who they hired, and roughly a fifth made gains that did not stick ¹. Removing the requirement is not the intervention. What the leaders did to the rest of the process is the intervention, and the study cannot see it directly.

Two limits keep this honest. The population is large firms only, meaning those posting more than 500 unique job ads a year, and education was inferred from resumes and career-history profiles rather than verified ¹. And the pattern is older than the study: measured across posting text, IBM, the employer that had gone furthest publicly, raised the share of its IT postings specifying a degree from 21% in 2017 to 29% in 2021, while Accenture's fell from 54% to 43% ². A stated policy does not hold itself in place. Something has to keep checking the postings against it.

Decide what the degree was standing in for

A degree line was doing one of two jobs, and they need different replacements. Sometimes it stood in for trainability, meaning somebody who can absorb a body of material and be useful in it within a quarter. Sometimes it stood in for written reasoning under a deadline. Pick the one this role actually needs, because the instrument that buys the evidence differs.

The gap between what postings asked for and who was doing the work was documented long before the reset. Analyzing more than 26 million postings, Harvard Business School's Dismissed by Degrees found that in 2015, 67% of production supervisor postings asked for a college degree while only 16% of people employed as production supervisors held one ³. That is one illustrative occupation from 2015 rather than an average, and it predates every change described here. It is still the cleanest statement of what the line often was: a habit, not a specification.

What makes this harder now is the fallback. Take the degree off and the screen leans on whatever is left, which is usually the written application, and a written application stopped separating people the moment a model started drafting all of them. That is the same problem as screening a stack of new-grad resumes with no work history behind them, and it is why the honest replacement has to be a stage that produces work the reviewer can read.

A job-shaped exercise is the usual answer, and it is worth knowing what it buys. The current meta-analytic estimate for work sample validity is .33, revised down from the .54 that still circulates in ranked-methods posts, and 53 of the 54 studies behind it tested people already doing the job ⁴. Read that alongside the .42 estimated for structured interviews. The credibility intervals overlap heavily, and the paper treats the revised table as coherent. What it means in practice is that a work sample earns a place in the round without carrying it alone.

Expect the description to get longer. Across 33 million postings from 2017 to 2019, employers who cut credential requirements went on to ask for a wider and deeper set of skills, with the biggest increases in explicitly stated soft skills such as communication, detail-oriented and problem solving ². Removing one vague requirement and adding five leaves the posting exactly as vague, with the vagueness moved somewhere the reader trusts less.

Expect screening to stop being free

A degree field costs nothing per candidate and a replacement costs reviewer minutes, so count the minutes while the change is still a plan. A twenty-minute graded exercise across sixty first-round candidates is twenty hours of somebody's month, and that somebody has to be a person who can read the output and tell a strong answer from a plausible one.

The expensive artifact is not the exercise. It is the key: a written description of what a strong answer contains, what a weak-but-plausible answer contains, and the decision rule that separates them. Without it, two reviewers grade the same submission differently and the process you installed to replace a credential filter is a taste filter with extra steps.

Three costs to put in the plan before anything is announced:

• Authoring. One practitioner, a paid afternoon, to write the task and the key out of work the team did last quarter.
• Reading. Minutes per candidate, times expected volume, assigned to a named person with room in their week.
• Calibration. Two reviewers grade the first ten submissions independently and compare, before the exercise gates anybody.

If none of that fits, the honest move is to say so and keep the current screen until it does. Going straight to a work sample is a real option with real tradeoffs, and it deserves its own decision: whether to drop the resume screen entirely is a different question from whether to drop the degree line inside it.

Keep the requirement where a rule sets it, and say which rule

Licensure, accreditation and customer contracts set real minimums, and a posting carrying one should say which, so a reader can tell a legal fact from a hiring habit. Everywhere else the line is a choice somebody made once, and a choice nobody can attribute is the kind that survives ten years of copying the last requisition.

That choice is a selection procedure under US federal law, open to challenge on adverse impact and defended on job-relatedness. Griggs v. Duke Power, decided in 1971 on a high school diploma requirement, is where adverse-impact analysis in US hiring starts, and it put the touchstone at business necessity ⁵. The Civil Rights Act of 1991 later wrote the burdens into the statute, which asks the employer to show that a challenged practice is job related for the position in question ⁶. That is public legal fact, and whether a particular line clears it for a particular role is a question for counsel.

The test that settles it takes an afternoon. Pull the last ten hires into this role, and mark which of them the degree line would have blocked. If the answer is zero, the requirement screens nobody who reaches your process, and its only remaining effect is on who decides to enter it. If the answer is more than zero, look at how those people performed, because that is the closest thing to local evidence you will get.

Then decide what re-checks it. A stated policy on requirements is not self-enforcing, which is what the posting-text measurement showed at named employers: the share of postings asking for a degree can climb again after the announcement, one copied requisition at a time ². Put a date on it, name an owner, and re-read the live postings for the role twice a year.

The wider version of this decision reaches past degrees. It is whether the process has an instrument for the thing the credential was standing in for, which is the whole argument in why skills-based hiring needs a test behind it. A pledge without one moves the wording. Only the instrument moves the hiring.

See a sample report >