Does IQ Predict Job Performance? A Shrinking Number

So does IQ predict job performance? Yes, but far less than the headline number long implied. The best 21st-century estimate of the link between cognitive ability and job performance is about r = .22 (Sackett et al., 2024), a fraction of the r = .51 that ruled hiring for a generation (Schmidt & Hunter, 1998).

Martin's experience is the gap this piece unpacks: how much that cognitive edge is really worth, and where it stops. The size of that edge, and its limits, matters to anyone deciding how much weight a test score should carry, whether they are sitting the test or scoring it.

Is IQ a good predictor of job performance?

Cognitive ability is a useful predictor, but a modest one. The strongest 21st-century meta-analysis estimates that general mental ability correlates with job performance at about r = .22 (Sackett et al., 2024), which means it accounts for under 5 percent of the differences in how well people do their jobs.

First, a clarification that clears up a lot of confusion. Hiring research seldom uses a clinical IQ score. It measures general cognitive ability, sometimes called general mental ability, a closely related summary of skills such as reasoning, working memory, and problem-solving speed. These labels overlap so much that the findings carry straight over to what most people mean by IQ at work.

The real story is simpler. It is not whether cognitive ability predicts performance, because it does, in dozens of studies; the 2024 meta-analysis alone pooled 153 of them (Sackett et al., 2024). The question is how much, and the answer keeps sliding under fresh scrutiny.

A correlation runs from -1 to +1, capturing how well a pair of measures move together; the validity coefficients in hiring research sit on the positive end, so the stretch that matters here runs from 0 to 1. An r of 1 would mean a cognitive score predicts performance with no error; an r of 0 would mean it tells you nothing at all. The real value sits low on that scale, closer to nothing than to certainty, yet still high enough to be worth measuring.

Why does a low number still matter? Because hiring happens at scale. A signal that nudges the odds in your favor, applied across hundreds or thousands of decisions, adds up to better average hires. The mistake is reading r = .22 as either useless or decisive. It is neither.

What is the correlation between IQ and job performance?

The number has moved. Schmidt and Hunter's 1998 review reported a corrected correlation of r = .51 (Schmidt & Hunter, 1998). A 2022 reanalysis lowered it to r = .31 (Sackett et al., 2022), and a study of 21st-century data landed on r = .22 (Sackett et al., 2024), the strongest current estimate though still contested.

Read those as a single number migrating, not as separate facts. They all estimate the same thing: how strongly general cognitive ability tracks overall job performance, from the 1998 review to the 2024 update. The trend runs in a single direction, downward, and every endpoint comes from a careful correction rather than a sloppy one.

Squaring each correlation turns it into the share of performance differences the test can explain, and that is where the shrinkage looks dramatic. The chart below tracks how steeply that share has fallen from the 1998 benchmark to the 2024 estimate.

Has the validity of cognitive tests for hiring changed?

Yes, and the revision came in stages. The bulk of the drop, from r = .51 to r = .31, came from redoing a range-restriction correction in the older math (Sackett et al., 2022). Only the smaller final step, to r = .22, reflects new 21st-century data (Sackett et al., 2024). The world changed less than the formula did.

Here is what happened, step by step. The original .51 came from Frank Schmidt and John Hunter, who corrected their 1998 figures for range restriction, the statistical fact that the people who get hired are a narrow, pre-screened slice of all applicants. Range restriction works a little like judging how well height predicts basketball skill by studying only NBA players: once the group is filtered down to a narrow band, the link looks weaker than it would across everyone who tried out, so researchers adjust the raw number to estimate what it would be in the full pool. A team led by Paul Sackett argued that the older correction ran too hot and re-did it on the same 20th-century data, which pulled the estimate from .51 down to .31 (Sackett et al., 2022). That step is arithmetic, not a claim that workers changed. In plain terms, most of the shrinkage is statisticians fixing old math, not employees who somehow got worse at their jobs.

The second move is the empirical one. When Sackett and colleagues restricted the evidence to 21st-century studies, the estimate fell further to .22 (Sackett et al., 2024). Critics pushed back fast. Oh and colleagues challenged the range-restriction assumptions behind the lower figure, but Sackett's team replied without backing down (Oh et al., 2023). The revision is a live scientific argument, not a closed verdict, which is why careful coverage holds the tension rather than declaring that IQ stopped mattering.

Why lean on 21st-century samples at all? Because the work itself shifted. Sackett and colleagues argue that the mix of what jobs reward has broadened since the 20th-century studies, so older data can overstate how much raw reasoning drives modern performance (Sackett et al., 2024).

The older, larger estimate has not vanished, though. It still describes complex jobs well. Salgado and Moscoso's reanalysis of 630 job samples found that cognitive ability remains a strong predictor, and a stronger one as job complexity rises (Salgado & Moscoso, 2019). The deflated .22 is an average across many roles, and the average hides a real spread: more cognitively demanding work leans harder on cognitive ability.

What predicts job performance better than IQ?

No single method dominates anymore. Structured interviews, job-knowledge tests, work samples, and conscientiousness predict performance about as well as cognitive ability, and blending several methods beats any used alone (Sackett et al., 2022). Personality and interpersonal skill carry more of the load than the old hierarchy assumed.

The clearest signal in the revised numbers is that performance is multi-dimensional. Sackett's team describes the criterion space, the full set of things that count as doing the job well, as broader than the older research assumed, with interpersonal skills carrying a larger share (Sackett et al., 2024). A separate 2023 review of the selection literature reaches the same verdict from a wider angle, placing cognitive ability among several procedures rather than ahead of them once its corrected validity is revised down and the criterion is read more broadly (Van Iddekinge, Lievens, & Sackett, 2023). That is why a personality assessment can add real predictive power on traits like conscientiousness that a cognitive score never captures.

Set the major selection methods, such as cognitive tests, structured interviews, and work samples, side by side, and the old hierarchy flattens. Each tool measures something different, carries a different fairness profile, and plays a different role in a modern hiring process.

For a hiring team, the practical move is not to drop cognitive testing but to demote it. Build the process around a structured interview or a work sample, both of which rival cognitive tests as top predictors in the revised hierarchy (Sackett et al., 2022), treat a cognitive score as one input weighed against several, and validate the whole battery against the actual job rather than leaning on any single hurdle. That is what the flattened hierarchy looks like in a real selection pipeline.

Think back to Martin. The cognitive test caught his reasoning speed, the trait it was built to measure, and missed the rest, including the patience, follow-through, and people-reading that carried his work in the end. He was not an exception. He was the rule that a broader criterion space predicts.

A separate worry runs alongside the validity debate. A test can predict performance and still raise fairness and legal questions, because cognitive tests often show large differences between demographic groups. The work psychologist Stephen Woods, of the University of Surrey, argues that heavy reliance on cognitive testing in high-volume graduate hiring rests on shaky assumptions, including consistent adverse impact on some groups (Woods & Patterson, 2023). Validity and fairness are separate questions, and a high score on the first does not settle the second.

What does a small correlation mean for a real hire?

More than the tiny percentage suggests, less than the old number promised. A binomial effect-size display of r = .22 (Sackett et al., 2024) implies a 61/39 split: 61% of higher-scoring candidates, but only 39% of lower scorers, reach the top half of performers. The edge shifts the averages, not any single call.

Picture a pair of applicants. Knowing their cognitive scores tells you little about which will be the better hire in any single head-to-head, yet across 500 hires the higher scorers come out ahead a bit more often. That is what a small but real correlation buys you: a gentle, repeatable tilt, not a crystal ball.

This is also why few researchers want to scrap cognitive tests. A small edge, multiplied across an organization's hiring, still raises average performance enough to matter. Precisely because no single number settles much, the useful signal lives in your separate domain strengths. A cognitive assessment that breaks the result into separate domain scores reports working memory, verbal reasoning, and the rest individually rather than as one figure that explains under 5% of performance on its own (Sackett et al., 2024). When a single number settles so little, a profile of distinct domain strengths gives you more to plan a career around than a lone composite score.

Frequently asked questions

What the shrinking number means for you

The honest headline is not that IQ stopped predicting job performance. It is that the famous figure was part math all along, and careful re-checking shrank it from r = .51 to r = .22, with most of that fall coming from corrected statistics rather than a changed world (Sackett et al., 2024). Cognitive ability still matters at work, but it matters less, and alongside more, than a generation of hiring lore assumed.

If you are an experienced candidate worried that a single test score could gate a career you have already proven, the practical takeaway is freeing. That score is a modest signal, not a verdict, and it says little about what an IQ test can and cannot predict about your career. The richer question is which of your cognitive strengths, such as working memory or verbal reasoning, fit which kind of work, and whether the role leans on raw reasoning or on judgment you have yet to build. The same strengths shape how cognitive ability links to income over a long career and whether scores track day-to-day job satisfaction. Matching your cognitive strengths to the right role is where the useful work begins.