Skip to content

IQ and School Performance: What the Correlation Really Says

You will read in a lot of places that IQ explains about a quarter of how well children do at school. That figure comes from a correlation of roughly 0.5, and it is comfortably wrong, or at least comfortably incomplete.

The largest study of this question in the UK followed more than 70,000 children and found a correlation of 0.81 between intelligence measured at age eleven and educational achievement at sixteen (Deary et al., 2007). That is much higher than the usual figure, and understanding why is the most useful thing on this page.

Why 0.81 and not 0.5

The two numbers are measuring different things, and the difference is technical but it matters.

A correlation of around 0.5 is typically what you get when you correlate one IQ test score against one exam result. Both of those measurements carry error, and error drags a correlation down. You are not measuring the relationship between ability and achievement. You are measuring the relationship between two noisy estimates of them.

Deary and colleagues used latent traits instead. They combined multiple indicators to estimate the underlying ability and the underlying achievement, which strips out a good deal of the measurement error. The relationship between the underlying things is 0.81. The relationship between any two single measurements of them is lower.

Correlations with school achievementLatent g at 11 vs latent GCSE achievement at 160.81Conscientiousness vs academic performance0.22Openness vs academic performance0.12
The first is Deary and colleagues (2007), over 70,000 English children. The lower two are from Poropat’s (2009) meta-analysis of personality and academic performance.

Both numbers are honest. The 0.5 tells you what to expect when you compare one test result against one set of grades, which is the situation a parent is usually in. The 0.81 tells you how tightly the underlying abilities are connected, which is the scientific question.

If your child took one test and got one set of school reports, the weaker figure is the one that applies to you.

What is left over

Even at 0.81, a third of the variation in achievement is something other than measured intelligence. At the level of individual test scores and individual grades, most of it is.

Poropat’s meta-analysis of personality and academic performance found conscientiousness correlating with achievement at around 0.22, which is substantial and, importantly, largely independent of intelligence (Poropat, 2009). Being organised, finishing things and turning up is a separate contribution, not a proxy for ability.

Then there is everything the research struggles to quantify: teaching quality, whether the subject interests them, whether home is settled, whether a friendship group treats effort as embarrassing, and whether anyone has spotted that the child cannot see the board. Several of these are more fixable than ability is.

The pattern behind a strong correlation
g score at age 11 Pupils reaching 5 good GCSEs What that means
Top fifth Large majority High ability usually converts, but not always
Middle fifth Around half The outcome is genuinely open
Bottom fifth A minority, but not none Low measured ability is not a ceiling

Illustrative of the distribution reported in large UK cohort studies. The point is that every band contains pupils at both outcomes.

Read that table carefully, because it is the honest version of a strong correlation. High ability at eleven usually converts into good results at sixteen. Usually. Every band contains children who went the other way, and the middle band is genuinely open.

The direction of causation is not one way

There is a trap in all of this that is easy to fall into: treating IQ as the cause and school results as the effect.

Schooling raises measured intelligence. Ritchie and Tucker-Drob pooled studies across three research designs and found each additional year of education worth roughly one to five IQ points, with the gain persisting (Ritchie and Tucker-Drob, 2018). So a child who engages with school becomes, measurably, someone who scores higher on the test that supposedly predicted their engagement.

The two things feed each other. An IQ score at eleven is partly a measure of how school has gone so far, which is one reason scores move through childhood more than a fixed-trait model would suggest.

As a mother, the correlation I keep in mind instead

My daughter’s school sends home a sheet with a predicted grade on it, generated from some prior-attainment model, and the first time I saw one I treated it as information about her.

What changed my mind was watching a friend’s son, who had been predicted very little, get interested in history at fourteen because of one teacher, and then go on to do far better in it than anything about him at eleven would have suggested. Nothing in the model saw that coming, because the thing that happened was a person.

So the number I actually attend to now is not the correlation. It is the residual. A third of the variation at the population level, and most of it at the level of any individual child, is not explained by ability, which means it is still up for grabs. That is not a sentimental reading of the data. It is what the data says.

What to do with a score

If the score is high, do not relax. High ability converts into results most of the time, not all of the time, and the gap between the two is usually conscientiousness and interest.

If the score is low, do not conclude much. One test against one set of grades is the weak-correlation scenario, and the list of ordinary reasons a score comes in low is long.

And in either case, spend your effort on the part that is movable. What personality contributes alongside ability is the more actionable half of this question, and attention difficulties account for a lot of apparent underachievement that is nothing of the kind.

For a current reading, the free IQ test for kids is age-normed and free, and there is an adult version. Whatever comes back, heritability does not mean fixed, which is the point most often lost in this conversation.

References

  1. Deary, I. J., Strand, S., Smith, P., & Fernandes, C. (2007). Intelligence and educational achievement. Intelligence, 35(1), 13-21. doi:10.1016/j.intell.2006.02.001
  2. Poropat, A. E. (2009). A meta-analysis of the five-factor model of personality and academic performance. Psychological Bulletin, 135(2), 322-338. doi:10.1037/a0014996
  3. Ritchie, S. J., & Tucker-Drob, E. M. (2018). How much does education improve intelligence? A meta-analysis. Psychological Science, 29(8), 1358-1369. doi:10.1177/0956797618774253