A widely repeated claim says Nobel laureate James Heckman found that IQ explains only 1% to 2% of differences in income. The number has a real basis: in one 2016 analysis co-authored by Heckman, IQ explained around 1% of wage variation, while a set of personality measures explained more.

It is not a universal estimate. Other analyses by Heckman and colleagues produced larger figures, and researchers who revisited the comparison in 2022 concluded that intelligence predicted pay better than Big Five personality on average. The careful version of the story is more interesting than the slogan: IQ is neither irrelevant nor destiny, and the answer changes with what researchers measure.

This is a contested research literature, not settled consensus.

The “around 1%” result came from one wage model

The most direct source is a 2016 paper in Proceedings of the National Academy of Sciences by Lex Borghans, Bart Golsteyn, James Heckman and John Eric Humphries. The researchers used four datasets to ask what IQ, personality, grades and achievement tests capture, and how well those measures predict later outcomes.

One part of the paper used the Midlife in the United States survey, known as MIDUS. Its Figure 4 compared regressions predicting wages, physical health limitations, depression, smoking and incarceration. In the wage model, IQ alone produced an adjusted R-squared of roughly .01. Put plainly, IQ accounted for about 1% of the observed wage variation in that particular sample and statistical specification.

That is the strongest basis for the headline’s 1% to 2% figure. It does not mean that a one-point rise in IQ increases income by 1%, that intelligence causes only 1% of financial success, or that 99% has been assigned to personality. Explained variance is a feature of a model fitted to a population, not a set of ingredients inside an individual life.

Heckman’s other results did not produce one fixed number

Four years earlier, Heckman and economist Tim Kautz reviewed evidence on cognitive and personality skills. A table based on the National Longitudinal Survey of Youth 1979 reported that IQ explained 7% of earnings variation among men at age 35 and 1% among women. For hourly wages, the figures were 7% and 5%.

An achievement-oriented cognitive measure, the Armed Forces Qualification Test, explained still more: 17% of earnings variation among men and 9% among women in those models. The authors did not treat the AFQT as a pure IQ test because achievement scores also reflect schooling, knowledge, motivation and personality.

The shifting values are not a nuisance to be averaged away. They show that predictive power depends on the population, age, income measure, cognitive test, personality measures and controls. “IQ explains 1% of wages in one MIDUS model” is precise. “IQ explains only 1% to 2% of income” everywhere is not.

Why conscientiousness entered the argument

Conscientiousness is one of the Big Five personality dimensions. In research questionnaires, it covers tendencies such as organization, responsibility, persistence and self-control. These patterns can plausibly help with completing education, meeting obligations and sustaining effort through long projects.

The 2016 PNAS paper argued that personality was generally more predictive than IQ across the adult outcomes it examined. It also found that grades and achievement tests often predicted outcomes better than tests intended to isolate fluid intelligence. The proposed reason was not that grades are better IQ tests. Grades combine cognitive skill with learned knowledge, attendance, sustained effort and behavior across time.

There is an important limit. The authors explicitly wrote that they were making no causal claims. A correlation between conscientiousness and income cannot show that changing a personality score would produce a matching change in pay. Family background, education, occupation and health can influence both the traits researchers record and the outcomes that follow.

A 2022 reanalysis challenged the personality-first conclusion

Chen Zisman and Yoav Ganzach reanalyzed two datasets used in the 2016 paper and added four more. Their 2022 study in the journal Intelligence reached the opposite ranking. Across six datasets, intelligence explained an average of 8% of the variation in pay, compared with 4% for Big Five personality. For educational attainment, the averages were 23.2% for intelligence and 5.3% for personality; for grades, 22.9% and 2.4%.

Golsteyn, Heckman and Humphries replied that the disagreement turned partly on measurement. Some tests used by the critics, particularly the AFQT, combine acquired knowledge and other influences rather than measuring only fluid intelligence. Zisman and Ganzach answered that intelligence remained the stronger predictor even under narrower definitions.

A 2023 review of the dispute by psychologist Lazar Stankov traced much of the conflict to differences in the datasets, populations and measures chosen. Stankov concluded that both cognitive and noncognitive variables predict important outcomes, while warning that the Big Five’s advantage was poorly supported when compared with stronger cognitive measures.

So the later literature does not erase Heckman’s result. It narrows it.

“IQ” and “success” are both less tidy than they sound

An IQ score is performance on a particular set of tasks under particular conditions. Achievement tests add learned material. Grades compress months or years of work, classroom conduct and teacher assessment. Personality scores usually rely on answers to questionnaires or observations of behavior.

Income is equally sensitive to definition. Hourly wages, annual earnings, household income and wealth are different outcomes. A model may include or exclude people with no earnings, control for education, split results by sex, transform pay using logarithms or remove extreme values. Each decision can change the amount of variance assigned to a predictor.

Then there are influences outside either IQ or personality: family resources, school quality, discrimination, health, local labor markets, chosen occupation, economic shocks and luck. A low R-squared for IQ does not tell us which of these filled the remaining space. It only says the model left most differences unexplained.

Intelligence is one part, but there is no simple replacement

The defensible conclusion is not that conscientiousness matters and intelligence does not. Nor is it that the later replication settled the contest in intelligence’s favor. Different traits predict different outcomes, and their apparent rankings depend heavily on how researchers define and measure them.

In my reading, the lasting contribution of Heckman’s work is the refusal to treat one score as a complete account of human potential. Cognitive ability can predict education and pay, sometimes substantially. Patterns of persistence and self-control can add information of their own. Grades and achievement tests often predict well precisely because they contain several kinds of information at once.

Being smart is one part of where people end up. The evidence does not support shrinking that part to a universal 1% to 2%, and it does not support replacing IQ with a single personality trait. Human outcomes are produced by abilities, behavior, institutions and circumstances interacting over time.