Skills-Based Hiring: What It Actually Takes

Skills-based hiring is the most repeated promise in the sector in recent years. Drop the degree requirement, evaluate what the person can do, widen the pool.

The Burning Glass Institute and Harvard Business School measured what happened after the announcements. They studied 11,300 roles at large firms, comparing at least one year before and one year after degree requirements were removed.

The increase in hiring people without a degree was 3.5 percentage points. Translated into volume, roughly 97,000 people out of 77 million annual hires. Fewer than one in seven hundred.

Where it breaks

The same work splits companies into two groups, and that is where the explanation sits.

37% changed for real, and in that group the share of hires without a degree rose by close to 20%. 45% dropped the requirement in name only, with no measurable change in who they ended up hiring.

The difference between the two groups is not intent. It is that deleting a requirement from a posting does not change the criterion sitting in the recruiter’s and the manager’s heads.

The degree was doing a job

A degree requirement is a shortcut. A bad one, noisy and with well-documented exclusion effects, but a shortcut that was solving something. It compressed into one résumé line a bet about baseline knowledge, persistence, and familiarity with a certain kind of work.

Remove it without putting anything in its place and the decision does not become fairer. It becomes less structured. And a less structured decision is more exposed to the impression left by the interview, which is where most of the noise enters. How that noise gets measured is in hiring bias.

Removing a proxy is half the work. The other half is directly measuring what the proxy was approximating.

The two layers hidden inside “skills”

“Skills” gets used as one thing and it is two, with different instruments.

What the person knows and can do today. Measured with job knowledge tests and work samples. In Sackett and colleagues’ updated validity estimates, those two families sit among the strongest on the board, with mean operational validities of 0.40 and 0.33.

How the person works. How much structure they need, how they move through ambiguity, how they respond to pressure. None of that shows up in a technical test, and it explains a good share of what we later call fit. It is the trait and competency layer, developed in the OCEAN model explained.

A program that measures only the first layer replaces a degree filter with a technical-résumé filter. It widens the pool very little and does not touch the decision.

The keyword-match trap

The most common version of skills-based hiring today is a system matching words on the résumé against words in the posting.

It is the same old filter under a new name. It rewards whoever writes their résumé in the right vocabulary, not whoever has the skill. And it penalizes exactly the non-traditional profiles the program claimed to want.

A skill is declared on a résumé and demonstrated in an assessment. Those are different things.

What it takes to work

1. A written role profile, not a job description. The three problems of the first year, competencies with levels, and behavioral anchors. Without that, “skills” is a word. The guide is in job profile with data.

2. One instrument per layer. A test or work sample for the technical side. An assessment with published validity for traits. What to demand of each is in pre-employment assessments.

3. A structured interview with a written scale. Same questions for everyone, individual scoring before the group meeting. It is what makes two candidates with different trajectories comparable, which is the central case for this approach. The detail is in competency-based interviews.

4. Demonstrable job-relatedness. Everything you measure has to be defensible against the actual demands of the work. Good psychometric practice, and it reduces legal exposure too.

5. Adverse impact measured, before and after. The change is verified by looking at who you hired, not at the text of the posting. A program without that control does not know whether it worked.

What this approach does not solve

Skills-based hiring does not eliminate bias. It moves the criterion somewhere it can be written down, argued with, and measured, which is a concrete improvement and a different one from what usually gets promised.

It does not widen the pool on its own either. What widens it is removing requirements nobody was going to evaluate and adding a way to evaluate what actually matters. The first step without the second produces exactly the 3.5 points from the study.

Where to start

Take one open search and run a thirty-minute exercise. List every requirement in the posting and write beside it how it will be evaluated. The ones with no answer come out of the posting, because nobody was going to look at them. The ones that stay need an instrument or a question with a scale.

That exercise does more for your pool than any policy announcement. And it leaves the criterion written down, which is the part 45% of the companies in the study never did.