Personality tests for hiring: what they can and cannot do

Personality testing in hiring attracts strong opinions in both directions. The defensible position is narrower than either camp likes: these tests measure something real, the something is weakly related to job performance, and the way most companies use them is worse than not using them at all.

What they measure

The frameworks worth taking seriously measure stable dispositions. The Big Five is the standard: openness, conscientiousness, extraversion, agreeableness and emotional stability. These are reasonably consistent over time and across situations, which is exactly what makes them attractive for prediction.

The catch is effect size. Conscientiousness is the most consistent performance-related trait across job types, and even it explains a modest share of the variation in how well people do their jobs. Modest is not nothing, and a small real signal used carefully beats a large imagined one. But it will not carry a hiring decision on its own.

Where companies go wrong

Treating the profile as a verdict. A candidate scoring low on extraversion is not disqualified from sales. The trait predicts a tendency, not an outcome, and plenty of effective salespeople are introverts who prepared.

Using a framework built for something else. MBTI is the common case. It was designed for self-understanding and development, its publisher says it should not be used for selection, and its test-retest stability is poor enough that the same person can land in a different type weeks later.

Screening on it first. Running personality as the opening filter throws away candidates before you have looked at whether they can do the work. If you are going to use it at all, use it after a work sample, as context for the interview rather than as a gate.

Skipping the adverse impact check. If your selection rate for any protected group falls below about four fifths of the highest group’s rate, you have a problem that needs either a validity argument or a different process. You cannot find this out without collecting applicant flow data, which means the check has to be designed in rather than run once a year in retrospect.

What the rules actually require

Three things are worth knowing before you buy anything:

1. Job-relatedness under adverse impact. The EEOC Uniform Guidelines put the burden on the employer to show the instrument relates to the job. Content validity, meaning the test visibly samples the work, is the cheapest route and the reason work samples are easier to defend than trait inventories.
2. The medical examination line. Instruments that probe psychopathology can be treated as medical examinations, which the ADA restricts sharply in pre-offer hiring. Karraker v. Rent-A-Center is the case that made this concrete.
3. Local rules on automated tools. New York City Local Law 144 requires an annual independent bias audit, a published summary and ten business days of notice to candidates. Illinois regulates AI-analysed video interviews separately. Colorado and the EU both place employment decisions in a high-risk category under their AI legislation.

None of this makes personality testing off limits. It does mean the compliance work is part of the cost, and a vendor who cannot discuss it is telling you something.

When a work sample is simply better

If the job has checkable output, test the output. A bookkeeping candidate reconciling a messy sheet tells you more than a conscientiousness score, and it is easier to justify to a candidate who asks why they were rejected. Personality testing earns its place where the work is hard to sample directly, or as a structured prompt for the interview rather than a filter before it.

How we handle this

SharpAssessment does not sell a personality test, which is a position rather than an omission.

The practical and job-knowledge tests in our library can be defended on content validity because they sample the work itself. A trait inventory cannot, so the moment it produces adverse impact it needs criterion validity, meaning real outcome data collected on a population like yours. Selling one without that would put the hardest problem in the category in front of the product instead of behind it.

If a psychometric profile has to carry weight in a decision you may have to defend, the alternatives page points at the vendors whose normative data goes back decades. That is a real reason to buy elsewhere.