Intelligence test scores and educational duratino are positively correlated. Either students with greater propensity for intelligence go on to complete more education, or a longer education increases intelligence. We meta-analysed three categories of quasi-experimental studies of educational effects on intelligence:
- those estimating education-intelligence associations after controlling for earlier intelligence,
- those using compulsory schooling policy changes as instrumental variables,
- those using regression-discontinuity designs on school-entry age cutoffs.
Across 142 effect sizes from 42 datasets involving 600k+ participants, we found consistent evidence for beneficial effects of education on cognitive abilities of approximately 1 to 5 IQ points for an additional year of education. Education appears to be the most consistent, robust and durable method yet to be identified for raising intelligence.
note ane: all this based in IQ tests? there are other ways to measure intelligence
We analyzed results from the three most prominent quasi-experimental methods for testing the effects of education on intelligence. Each method implements a different approach to minimize effects stemming from selection processes.
The first research design, 'Control prior intelligence', is a longitudinal study where cognitive testing data is collected before and after variation in the duration of education. Results indicated that completing a university education was linked to higher midlife cognitive ability, above and beyond adolescent intelligence.
The second design, 'Policy change', relies on changes in educational duration that are exogenous (having an external cause or origin) to the characteristics of the individuals.
The third design, 'School age cutoff', use regression discontinuity analysis to leverage the fact that school districts implement a date-of-birth cutoff for school entry. Prior studies found that schooling exerted positive effects on all of twelve tests covering a variety of cognitive domains.
After synthesizing the evidence within and across these three research designs, we addressed two further questions. First, which factors moderate the effect of education on intelligence? Are any educational effects subject to decline or fadeout with increasing age? And second, to what extent is there publication bias in this literature, such that the meta-analytic effects might be biased by a disproportionate number of positive rsults?
- The outcome cognitive measures had to be objective and continuous.
- Variation in education had to be after age 6.
- The population under study had to be generally healthy and neurotypical
- Studies had to fit into one of the three study design types described above. They had to either use earlier cognitive test scores as a control variable in a model predicting cognitive test scores after some variation in educational duratino, or use data from a 'natural experiment' that specifically affected educational duration prior to the outcome cognitive tests, or use a regression discontinuity design to analyse cognitive test scores from individuals born either side of a cutoff date for school entry
We gathered a number of publications (more in paper) for the meta-analysis (Supplementary materials).
Calculating effect sizes: We rescaled each effect size into the number of IQ point units in the standard IQ scale (mean=100, SD=15), associated with one additional year of education.
Meta-analytic structural equation models: to produce the main estimates, we used random-effects meta-analytic structural equation modeling. We estimated these models using Mplus.
Moderators: we used the tau statistic, an estimate of the SD of the true meta-analytic effect, as an index of heterogeneity. To attempt to explain any heterogeneity, we tested a somewhat different set of moderators for each of the three study designs, as appropriate given their methodological differences.
We classified outcome tests in two ways: The first was into the broad intelligence subtype:
- fluid intelligence (gf: tests that assessed skills such as reasoning, memory, processing speed, and tasks that could be completed without outside knowledge from the world)
- crystallized intelligence (gc: tests that assessed skills such as vocabulary and general knowledge)
- general intelligence (g: tests that assessed a mixture of fluid and crystallized skills)
The second classification method was to highlight tests that might be considered 'achievement' measures. We classified every test that would likely have involved content directly taught at school as 'achievement' and the remaining (involving IQ-type measures, processing speed, reasoning, vocabulary) as 'other' tests.
Publication bias tests: We used four separate methods to assess the degree of publication bias in the dataset.
First, we tested whether the effect sizes were larger in peer-reviewed, published studies vs. unpublished studies. If unpublished studies have significantly smaller effect sizes, this might indicate publication bias.
The other more complicated methods can be found in paper page 8.
The selection process resulted in a final meta-analytic dataset including 142 effect sizes from 42 datasets, analysed in 28 studies. The total simple size across all 3 designs was 615k.
In three separate unconditional random-effects meta-analytic models (one for each study design) we estimated the effect of one additional year of education on cognitive outcomes. For all three study designs, there was a significant effect of one additional year of education. For the first, 1.2 IQ points, for the second, 2 IQ points, for the third, 5.3 IQ points.
There was significant heterogeneity in the unconditional meta-analyses from all three designs.
Age at early test and time-lag between test: the age at which the early test is taken has no effect to the effect size, but time-lag is. At the smalles gap (5 years) the effect sizewas 2.4 IQ points, and for the largest (72 years), it was 0.3 IQ. This is the gain for one more year of education, which at age 73 doesn't make much difference.
Age at intervention: the age at which the educational policy produces an increment in compulsory schooling doesn't affect the outcomes.
Age at outcome test: outcome age is a significant moderator. At the youngest age (18 years) the effect size in an additional educational year is 2.1 IQ points, whereas at the oldest age (83 years) the effect size was 0.5 IQ points.
Outcome test category: For the first study there's stronger educational impacts on tests of general intelligence, but there's not much difference with the other two.
There is also no difference between achievement tests and other tests.
Male-only studies: there's no change in the results of a male-only database.
There was no significant difference between peer-reviewed studies and unpublished in one design, a bigger difference in the other designs.
Unpublished studies produce larger estimates for the Policy Change design, in opposite direction to what we would expect under publication bias.
We found highly consistent evidence that longer educational duration is associated with increased intelligence test scores. The results support the hypothesis that education has a causal effect on intelligence test scores. The effect of one additional year of education was estimated at 1-5 standardised IQ points.
The advantages and drawbacks of each research designs can be found in the paper page 15.
The following questions remain: are the effects on intelligence additive across multiple years of education? Are there individual differences in the magnitude of the educational effect? Which cognitive abilities were impacted? And what are the underlying psychological mechanisms of the educational effects on intelligence?