Reading focus
Syllabus: What Is Intelligence?
Compare the assumptions of all three quasi-experimental approaches and separate measured outcomes from latent g, long-term persistence, and later life consequences.
Why you’re reading it. Use evidence of change to test the meaning of intelligence claims rather than assuming either total immutability or unlimited malleability.
A question to consider. The programme reports higher scores after an extra year of education. Which design would most improve its causal argument, and what would still be needed to identify the ability that changed?
Why this reading
People with more education tend to score higher on cognitive tests, but the direction of influence is not established by that association. This meta-analysis compares three approaches intended to estimate education's effect. It makes improvement a question about both causal design and the meaning of the outcomes measured.
Before you read
The designs adjust for earlier intelligence, use changes in schooling policy, or compare children around school-entry cutoffs. Each needs different assumptions. An instrumental-variable estimate uses a source of variation in schooling intended to isolate its effect; a regression-discontinuity design compares observations near a threshold. Standardizing outcomes into IQ points does not make the underlying tests identical.
Read for
- Identify what each design attempts to rule out, including selection into education, and what assumptions remain vulnerable.
- Compare estimates across designs and consider which people, schooling changes, and testing ages they describe.
- Distinguish gains on measured tasks from changes in latent general ability, persistence over time, and benefits in outcomes not directly tested by the synthesis.
Consider afterwards
- A school-leaving reform is associated with higher later cognitive scores among affected students. What supports a causal interpretation, and what further evidence would justify saying the reform changed general intelligence rather than particular tested skills?
Editorial context
The converging evidence supports an effect of education on cognitive test performance while leaving its precise composition and mechanisms open. The reported average effects are not guaranteed gains for each child or quantities that can be multiplied indefinitely by years of schooling. Much of the evidence cannot isolate latent g from particular abilities. Effects associated with a policy change need not generalize to every form of instruction, and the paper does not establish that score changes mediate all later benefits of education.