# Skills-based hiring: why dropping the degree requirement is not enough

> Many employers dropped degree requirements and kept hiring the same people. What skills-based hiring actually requires, what the evidence says about work samples and job knowledge tests, and a planner for building a short work sample.

Jennifer at SortTheStack, October 12, 2026. Original: https://sortthestack.com/blog/skills-based-hiring

Skills-based hiring means judging applicants on evidence that they can do the
work, rather than on credentials that stand in for it. **Removing a degree
requirement from a posting is the easy part, and on its own it rarely changes
who gets hired. What changes outcomes is replacing the proxy with something
better at every step: requirements written as evidence, a screen that reads
for that evidence, and an assessment that tests the skill directly.**

## The gap between announcing it and doing it

In 2024 the Burning Glass Institute and Harvard Business School
[studied 11,300 roles at large U.S. firms](https://burningglassinstitute.org/research/skills-based-hiring-2024)
where they could observe hiring for at least a year before and after the
employer removed a degree requirement. On average, the share of people hired
without a bachelor's degree rose by about 3.5 percentage points. Because only a
small fraction of roles had dropped the requirement, the authors estimated the
overall effect at fewer than 1 in 700 hires in 2023.

The average hides a split. Nearly all of the change came from about a third of
the firms, which the report calls Skills-Based Hiring Leaders. About 45% of
firms changed their postings with no meaningful change in who they hired. A
further group made early gains that didn't last.

The report's explanation is the useful part. Removing a requirement from the
posting doesn't remove it from the hiring manager's head. Under pressure to
fill an expensive vacancy quickly, managers have "a subtle, but real incentive
to revert to relying on proxies, rather than engage in the hard work of
assessing an individual candidate's strengths and skills".

In other words, a degree was never just a requirement. It was a shortcut for
judging people. Take away the shortcut without providing a better way to judge,
and people find their way back to it.

## What a degree was standing in for

Before replacing a credential, be specific about what it was meant to tell
you. For most roles it's some mix of:

- **Knowledge of a field**, which a job knowledge test or a few targeted
  interview questions can check directly.
- **Ability to do specific tasks**, which a work sample can show.
- **Persistence and reliability**, which past behaviour, explored in a
  [structured interview](https://sortthestack.com/blog/structured-interviews), speaks to more directly
  than a credential.

Our guide to [writing must-haves](https://sortthestack.com/blog/must-have-vs-nice-to-have) has a
four-question test for working out which requirements are real and which are
stand-ins.

## Job titles are a stand-in too

Degrees get most of the attention, but requiring a previous job title does the
same thing. LinkedIn's Economic Graph Research Institute
([Lara, 2025](https://economicgraph.linkedin.com/content/dam/me/economicgraph/en-us/PDF/skills-based-hiring-march-2025.pdf))
compared two ways of deciding who is eligible for a job, using its members'
profiles. The first counted people who had held the same or an equivalent title
in the past five years. The second counted people whose recent roles shared at
least half of the job's top skills, limited to roles that people actually move
from into that job, so a nurse doesn't count as a candidate to be a doctor.
Across the 56 countries in that analysis, the skills-based pool for a typical
job was 6.1 times the size of the title-based one. In the United States it was
15.9 times.

Read that for what it is: a model of who could be considered, built from
LinkedIn profile data, not a measure of who was hired or how they performed.
The report itself says "actual hiring practices remain more complicated and
nuanced". Together with the Burning Glass findings above, it points one way:
far more people can do a job than look like it on paper, and reaching them
depends on how you screen, not only on what the posting says.

## What the evidence supports instead

A [2022 re-analysis of selection research](https://sortthestack.com/blog/what-predicts-job-performance)
estimated how well common methods predict job performance. Three that test
skills directly are among the strongest:

| Method                 | Estimated validity | Works best when                                     |
| ---------------------- | ------------------ | --------------------------------------------------- |
| Structured interview   | .42                | Questions come from the job's requirements          |
| Job knowledge test     | .40                | The job has a defined body of knowledge             |
| Work sample test       | .33                | The person must be able to do the task from day one |
| Job experience (years) | .07                | (For comparison: barely predictive)                 |

None of these is decisive on its own, and the estimates vary from study to
study. But each is far closer to the actual work than a credential or a year
count.

## Build a short work sample

A work sample asks applicants to do a small, realistic piece of the job. The
U.S. Office of Personnel Management's
[guidance on work samples](https://www.opm.gov/policy-data-oversight/assessment-and-selection/other-assessment-methods/work-samples-and-simulations/)
notes that applicants often see them as very fair, and that they generally show
little or no difference in performance between men and women or between racial
groups, though that depends on what is being assessed. It also cautions that
they suit roles where people are expected to have the skill on entry; if you'll
train them in it after hiring, don't test for it before.

The cost is time: yours to design and score it, and the applicant's to do it.
That's why a small team's work sample should be short and focused. Our
recommendation:

1. **Pick one task that matters most.** The one where a weak hire would show up
   first. Not a full day's work.
2. **Keep it under an hour,** and tell applicants how long it should take. If
   it would take longer, consider paying for it.
3. **Use realistic material, not your real work.** A made-up dataset, a
   fictional customer email, a sample file with a few planted problems.
4. **Decide what good looks like before anyone does it.** Write what a weak,
   acceptable and strong result look like, as you would for interview rating
   anchors.
5. **Give everyone the same task, the same instructions and the same time.**
   Offer adjustments to anyone who needs them for a disability.
6. **Ask them to walk you through it.** Ten minutes of "why did you do it this
   way?" tells you more than the output alone, and makes it far harder to
   submit someone else's work, or a language model's.

### Example: the Senior Accountant

For the fictional Larkspur Partners role used throughout our guides:

> **Task (40 minutes, done live on a shared screen):** Here is a bank statement
> and the general ledger cash account for one month. They don't reconcile.
> Find the differences, propose the adjusting entries, and write two sentences
> to the partner explaining what happened.
>
> **Strong:** finds all three planted items, including the timing difference;
> entries are correct; the note says what the partner needs to decide.
> **Acceptable:** finds two, entries mostly correct, note is accurate but
> technical. **Weak:** finds one or none, or forces the reconciliation with a
> plug figure.

Twenty minutes of setup, forty minutes per finalist, and you know more about
their close skills than any resume could tell you.

### Examples for other roles

| Role                 | A short, realistic work sample                                                       |
| -------------------- | ------------------------------------------------------------------------------------ |
| Office administrator | Prioritise a fictional inbox of 12 messages and draft replies to the two most urgent |
| Customer service     | A 10-minute role-play with a frustrated customer, using a one-page policy sheet      |
| Bookkeeper           | Categorise 20 transactions and flag the three that need a question to the owner      |
| Junior developer     | Fix a small, documented bug in a short code sample, then explain the fix             |
| Sales representative | Prepare and deliver a five-minute call opening to a fictional prospect               |

Our [work sample planner (PDF)](https://sortthestack.com/downloads/work-sample-planner.pdf) walks
through the six steps on one page, with space to write the task and the
anchors. Free, no sign-up.

## Change the screen, not just the posting

The report's "in name only" firms show where skills-based hiring fails: in the
pile of applications, not in the posting. If the job no longer requires a
degree, the screen has to stop rewarding one. Practical changes:

- **Screen on the evidence you wrote down, and nothing else.** A degree, a
  well-known employer or a prestigious school should not move anyone up the
  pile unless it's on the list.
- **Read the whole work history.** Skills built in a different industry, in a
  volunteer role, or in a job with a modest title count if they show the
  evidence.
- **Mark "not shown" as unclear, not as no.** Applicants without a degree often
  describe their experience differently, and a phone call settles more than a
  rejection does. Our
  [screening guide](https://sortthestack.com/blog/how-to-screen-a-pile-of-resumes) shows how.
- **Check the result.** After a few hires, look at who reached the interview
  stage. If the share without degrees hasn't changed, your process hasn't
  either.

If the problem is that too few qualified people apply at all, see
[not finding qualified candidates?](https://sortthestack.com/blog/finding-qualified-candidates)

## Limits

Work samples and knowledge tests measure what they test, not everything that
matters. A strong reconciliation doesn't tell you how someone handles a
difficult client. They cost applicants time, which not everyone can give, so
keep them short and use them late, for a handful of finalists. And they are only as fair as their design: a task that
depends on knowing your industry's jargon tests familiarity, not skill.

The report also found that at the firms that did change their hiring, workers
hired without degrees into roles that used to require them stayed longer than
their colleagues with degrees. That's an association within those firms, not
proof that the policy caused it, but it's a sign that the people screened out
by a degree requirement are not the weaker candidates the requirement assumes.

## Where SortTheStack fits

SortTheStack screens on the requirements you set, and nothing else. If a degree
isn't on your list, it doesn't count for or against anyone, and each
requirement is marked met, not met or unclear with the evidence quoted from the
application. It doesn't run work samples or skills tests; those happen after
the shortlist, with you. [See how the screening works](https://sortthestack.com/product).

## Sources

- Sigelman, M., Fuller, J., & Martin, A. (2024).
  [_Skills-based hiring: The long road from pronouncements to practice_](https://burningglassinstitute.org/research/skills-based-hiring-2024).
  Burning Glass Institute and Harvard Business School Project on Managing the
  Future of Work.
- Lara, S. (2025).
  [_Skills-based hiring: Increasing access to opportunity_](https://economicgraph.linkedin.com/content/dam/me/economicgraph/en-us/PDF/skills-based-hiring-march-2025.pdf).
  LinkedIn Economic Graph Research Institute.
- Sackett, P. R., Zhang, C., Berry, C. M., & Lievens, F. (2022). Revisiting
  meta-analytic estimates of validity in personnel selection. _Journal of
  Applied Psychology, 107_(11), 2040–2068.
  [doi.org/10.1037/apl0000994](https://doi.org/10.1037/apl0000994)
- U.S. Office of Personnel Management,
  [Work samples and simulations](https://www.opm.gov/policy-data-oversight/assessment-and-selection/other-assessment-methods/work-samples-and-simulations/),
  accessed 12 October 2026.