Research

Does a high IQ predict success?

It predicts a real slice of academic and job outcomes, and a much smaller slice than either the enthusiasts or the dismissers claim. Here are the actual correlations.

The correlations, stated plainly

The relationship between IQ and outcomes is one of the better-studied questions in social science, and the results are consistent enough to summarise.

Academic achievement: correlations typically around 0.5, higher in earlier schooling and lower at postgraduate level where the range of ability has already been narrowed by selection.

Job performance: the classic meta-analytic figures sat around 0.5 for complex jobs and 0.2 to 0.3 for simple ones. Recent re-analyses correcting the range-restriction assumptions in the older work have pushed the estimates down, and the size of that correction is currently a live dispute.

Income: lower, commonly around 0.2 to 0.3, and heavily mediated by education and occupation.

Job training success: among the stronger relationships, since training is essentially an academic task in a workplace.

A correlation of 0.3 accounts for about nine percent of the variance. A correlation of 0.5 accounts for about twenty-five percent. Those are the numbers, and everything else in this article is about what they mean.

What “accounts for 25 percent of the variance” feels like

Percentages of variance are hard to feel, so here is a more concrete version.

Take two people at random. If one has a higher IQ than the other, and the correlation with the outcome you care about is 0.5, the higher-IQ person does better roughly 67 percent of the time. At a correlation of 0.3, roughly 60 percent of the time.

So it is a real edge and nothing like a determination. A correlation of 0.5 means one person in three does better than someone with a higher score on the trait. That is a lot of people.

This is the number to hold onto when reading either kind of overclaim. It is far too large to dismiss and far too small to predict an individual life.

The threshold hypothesis

A popular idea holds that IQ matters up to a point — often placed around 120 — after which additional points stop buying anything and other factors take over.

It is intuitively appealing and the evidence for it is weak. The Study of Mathematically Precocious Youth followed people identified in adolescence by extremely high test scores, and found that differences within the top one percent still predicted later outcomes: the top quarter of that top percentile produced substantially more doctorates, patents and publications than the bottom quarter of it.

So the relationship does not appear to switch off at a threshold. What does happen is that the practical importance of ability differences shrinks relative to everything else, because at the high end the population is already selected on ability and differences in opportunity, persistence and circumstance carry proportionally more weight.

What else predicts the same outcomes

IQ does not compete alone, and several other measured traits predict overlapping outcomes with respectable effect sizes.

Conscientiousness predicts job performance about as well as cognitive ability in many roles, and predicts academic grades better than ability does at university level — largely because grades reward turning work in on time.

Family background predicts income and educational attainment strongly, and is partly confounded with measured IQ, which makes clean attribution difficult.

Education itself carries predictive weight beyond the ability it selects for, through credentials, networks and specific skills.

Health and circumstance. A chronic illness, a caring responsibility or a recession at graduation can dominate any of the above.

Of that list, the one most responsive to deliberate effort is conscientiousness, which is worth noticing if you came to this article wondering what to do with the information.

Three ways this research gets misused

Reading a group statistic onto an individual. Population correlations say nothing determinate about any particular person. A 0.3 correlation is entirely compatible with the highest earner in a group having a below-average score.

Treating correlation as mechanism. IQ correlates with income partly because it correlates with education, which correlates with occupation. Each link is doing work, and none of it is a direct causal arrow from reasoning to salary.

Circular definitions of success. The outcomes in this literature are the ones that are easy to measure — grades, salary, occupational prestige. They are not a definition of a life going well, and the research does not claim to be one.

What to do with all this

If you took a test and got a number, here is the useful framing.

A high score is an advantage in tasks that reward fast pattern-learning, and it is one input among many. It does not guarantee anything, and the research is unambiguous that plenty of people with higher scores end up doing less with them.

A middling score is not a ceiling. Half of everyone is in the average band, and that band contains the overwhelming majority of competent professionals in every field, because there are so many more people in it.

And a score from a ten-minute web test carries an error band wide enough that a good chunk of the difference between your result and someone else's is noise. Whatever conclusion you were about to draw, draw a smaller one.