Ability tests in recruitment: what the research shows

For decades, cognitive ability tests were held up as the single best predictor of job performance. Recent research has revised that claim in important ways — without dislodging ability testing from the core of a well-designed selection process. Here is where the evidence now stands.

The classic case

Schmidt and Hunter's 1998 review of 85 years of research concluded that general mental ability was the most valid single predictor of job performance available to employers, with an estimated validity of about .51, rising with job complexity, and the strongest known predictor of success in training.^[1] That conclusion anchored selection practice for over twenty years: ability tests are cheap, quick, and predict learning — which is why they generalise across roles in a way few other measures do.

The 2022 correction

In 2022, Sackett, Zhang, Berry and Lievens re-examined the statistical corrections underlying those estimates and showed that the standard treatment of range restriction had systematically inflated them. Under more defensible corrections, the mean validity of cognitive ability tests fell to about .31, and structured interviews emerged as the predictor with the highest mean validity, at about .42.^[2] Most procedures that ranked high before still rank high — but the gaps narrowed, and ability lost its stand-alone crown.^[2,3]

The correction also sharpened a long-standing concern. Cognitive ability tests show larger subgroup score differences than most alternatives, creating a validity–diversity trade-off. Follow-up work using the updated estimates found that removing ability tests from a well-constructed battery costs little validity while substantially reducing adverse impact^[4] — which argues not for abandoning ability testing, but for using it in proportion, alongside predictors with smaller group differences.

What survived the correction is worth stating plainly: ability tests remain among the most valid, most economical and most role-general instruments available, and their validity for predicting how quickly people learn new work is not in serious dispute.^[1,2]

What ability tests are for

The practical distinction is between what a candidate can do today and what they will learn to do. Skills tests measure the former; ability tests are the best available measure of the latter. That makes ability testing most valuable where the job will keep changing, where training investment is large, or where candidates cannot yet be expected to have the skills — graduate and entry-level hiring being the obvious cases. Sackett and colleagues make a related point: predictors like knowledge and work-sample tests can be improved by study and practice, while ability is comparatively stable — so the right mix depends on whether you are selecting for current capability or future growth.^[3]

Verbal, numerical and abstract reasoning sub-tests also carry role-specific signal: numerical reasoning for analytical and financial roles, verbal for roles built on written communication, abstract for novel problem solving. A well-designed process matches the sub-test mix to the job's demands rather than administering a generic battery.

Using ability tests defensibly

The research supports four rules. Use ability as one component of a battery, never a sole gate — its incremental combination with structured interviews, skills tests and personality is where selection accuracy comes from.^[1,3] Set cut-offs from job analysis, not percentile ambition. Monitor group differences and weigh the validity–diversity trade-off explicitly.^[4] And time the test generously enough that it measures reasoning rather than test-taking nerve — speed matters in some jobs, but rarely as much as accuracy.

Read together with the skills-testing evidence, the division of labour is clean: skills tests tell you what a candidate can do on day one; ability tests tell you how fast they will pick up what they can't do yet. Most good hiring decisions need both numbers.