What Actually Predicts Job Performance? The Numbers Changed.

For twenty-five years, the most-cited fact in hiring was that general cognitive ability predicts job performance better than anything else, at a correlation of 0.51. It shaped how companies screened, how careers advice was written, and how people were told to prepare. In 2022 it was substantially revised, and almost nobody who repeats the old number knows.

What changed, and why

The original figure came from a 1998 meta-analysis by Schmidt and Hunter. Validity studies always understate the truth, because the people you can measure are a narrower range than the applicant pool, so researchers apply a correction for that range restriction.

In 2022, Sackett and colleagues showed the field had built those corrections wrong and applied them too widely, for decades, inflating almost every estimate. Most validity studies test people already hired, where the predictor played no role in selection and so there is little restriction to correct, yet the same large correction was applied anyway.

The authors were unusually direct about their own part in it, writing that they had "errantly drew upon these approaches," cited them, and taught them.

The corrected ranking

The headline is that cognitive ability is no longer the top predictor. Structured interviews are.

• Structured interviews: 0.42 (was 0.51)
• Job knowledge tests: 0.40
• Work samples: 0.33 (was 0.54)
• Cognitive ability tests: 0.31 (was 0.51)
• Unstructured interviews: 0.19 (was 0.38)
• Years of job experience: 0.07 (was 0.18)

The four strongest predictors, the researchers note, are all job-specific measures: how you answer a defined question, whether you know the domain, your track record, and whether you can do a sample of the actual work. The general trait is not the thing.

What it means for you

If you are being hired, the evidence says a well-run structured interview and a work sample will do more to reveal you than any aptitude test, which is good news if your strength is doing the job rather than sitting exams.

If you are hiring, the practical implication is stark and it connects to a separate finding: the free-form conversation most managers trust is close to the bottom of the list. We cover that in why unstructured interviews make hiring worse.

And "years of experience" at 0.07 is worth sitting with. It barely predicts performance, which is consistent with what Botswana employers report in the skills Botswana says it needs, and it undercuts the assumption behind a lot of career-ladder thinking.

The honest caveats

These are averages with wide credibility intervals; structured interviews range from 0.18 to 0.66 across settings. And a later update, using studies from this century only, put cognitive ability lower still, at 0.23. The authors summarise their own work plainly: "Our predictors are useful; but the predictive relationships are considerably weaker than previously thought."