Do Personality Tests Work for Hiring? Weaker Than You'd Think.

The validity numbers, and what they mean in selection

Meta-analyses across thousands of studies find conscientiousness has a mean corrected validity around .20 to .22 for predicting overall job performance, which translates to about 4% of variance accounted for. This is a real relationship, replicable and not zero, and it is also modest enough that many other factors matter more. Emotional stability adds a smaller contribution, and the other three Big Five traits — extraversion, agreeableness, openness — contribute variably or negligibly depending on the job.

That .20 figure is a population average, and it rises in some contexts: conscientiousness predicts better in jobs requiring sustained detail work, and extraversion matters more in roles with a high social component. The problem is that those contexts are where you could observe the behavior directly — the warehouse applicant's prior attendance record or the sales candidate's actual numbers — and both of those outpredict the trait score.

For comparison, cognitive ability tests typically yield validities of .50 or higher for complex roles, and structured behavioral interviews run .40 to .50. A work sample test in a relevant domain is often the single strongest non-GMA predictor. Personality adds something beyond those methods, but it is a small increment over a much stronger foundation, which is why it is defensible as one input among several and a poor choice as the gate.

The faking problem and why it matters differently for selection

People answering a personality test in a hiring context score higher than people answering the same instrument in research, particularly on traits employers obviously prefer. Conscientiousness, emotional stability and agreeableness all shift upward, with effect sizes in some studies exceeding half a standard deviation. This is not subtle and it is consistent enough to meta-analyse.

Two competing interpretations exist. One is that applicants are presenting their ideal selves rather than their typical ones, which means the score is aspirational more than descriptive. The other is that a high-stakes context reveals self-regulatory capacity, and the ability to present well under observation is itself predictive. The evidence leans toward the first: faked scores are generally weaker predictors of later job performance than honest scores, and forced-choice formats designed to resist faking have not eliminated the validity drop.

The practical implication is that rank-ordering candidates by personality score is less meaningful than most selection processes assume. The top scorers are often the most motivated to appear conscientious rather than the most conscientious, and distinguishing the two from a questionnaire is close to impossible.

Legal and fairness concerns

Personality tests carry real legal risk under adverse impact standards, particularly in the United States. If a test disproportionately screens out applicants from a protected class and the employer cannot demonstrate clear business necessity, it can be challenged under Title VII. Demonstrating that necessity requires showing that the test predicts job performance in this specific role, for this employer — and given the modest validities involved, that is a higher bar than many assume.

There is also a professional-ethics question separate from the law. Using a measure with a validity of .20 as a gate throws out a substantial number of people who would have succeeded, and it disproportionately affects those less familiar with the genre — first-generation applicants, older workers returning after a break, anyone who has not learned the test-taking conventions. When the same screen could be replaced with a work sample that predicts three times as strongly, choosing the personality test is a choice, and it is one worth justifying.

When and how they are defensible

As one input in a structured process, behind a resume screen and a cognitive or work-sample test, personality assessment can add value without carrying most of the weight. That is the context where the incremental validity research applies: not personality alone, but personality after the stronger predictors have already narrowed the pool.

For roles where a specific facet genuinely matters — emotional stability in high-stress safety-critical work, for example — and where that trait cannot be observed directly, a validated measure is defensible. But the specificity matters: using a general Big Five instrument when the actual requirement is narrow wastes the signal in noise.

Never use them as the primary gate, never rank-order on personality alone, and always pilot any new instrument on your own employees first to check for adverse impact before deploying it on applicants. Most selection lawsuits happen because a tool was rolled out without local validation and produced a demographic disparity the employer could not justify.