Hereditarians often argue that education doesn't affect intelligence. Common arguments include:
"it's not on g"
“It’s not additive”
"It's overestimated"
"education only works for smart kids(matthew effect)"
In this 🧵, I'll address all of these points:
"it's not on g"
“It’s not additive”
"It's overestimated"
"education only works for smart kids(matthew effect)"
In this 🧵, I'll address all of these points:
https://x.com/notcomplex_/status/1933296049155723527
Let’s start with the “not on g” argument. First of all, the Ritchie &Bates study was run incorrectly, see
@SashaGusevPosts's piece on it(Ritchie agreed with the criticism!)
Some other studies such as Bergold et al(mentioned in Sasha’s article) have found that it is on g.
@SashaGusevPosts's piece on it(Ritchie agreed with the criticism!)
Some other studies such as Bergold et al(mentioned in Sasha’s article) have found that it is on g.
@SashaGusevPosts theinfinitesimal.substack.com/p/does-educati…
@SashaGusevPosts Nevertheless, in my opinion it is unresolvable with current public data.
But good thing, we don’t need to, since Coyle has shown that the non-g residual part of IQ tests predict outcomes just as well as g itself.
But good thing, we don’t need to, since Coyle has shown that the non-g residual part of IQ tests predict outcomes just as well as g itself.
@SashaGusevPosts This is true for grades, income, and job performance. It has also been replicated with the ASVAB residuals, and even group factors!
Whether the gains are on g or not is a red herring. The only thing that matters is whether education induced IQ gains translate to better outcomes
Whether the gains are on g or not is a red herring. The only thing that matters is whether education induced IQ gains translate to better outcomes
@SashaGusevPosts “smart kids gain more”
Heckman used very fancy latent variable models to control for selection effects, and estimated that the effects are indeed additive with very minor diminishing returns, and LOWER IQ PEOPLE BENEFIT MORE FROM EDUCATION!
Heckman used very fancy latent variable models to control for selection effects, and estimated that the effects are indeed additive with very minor diminishing returns, and LOWER IQ PEOPLE BENEFIT MORE FROM EDUCATION!
https://x.com/notcomplex_/status/1933296056114061573
@SashaGusevPosts sciencedirect.com/science/articl…
@SashaGusevPosts So what explains the results Alex has cited?
Simply put, vertical scaling, which is less of an issue in factor analysis than item level data:
Simply put, vertical scaling, which is less of an issue in factor analysis than item level data:
https://x.com/drewhalbailey/status/1559592939067609088
@SashaGusevPosts “Effects fade out”
They don’t fade out, Cremieux’s chart is implying gains themselves are nonlinear… It uses the most biased design tho, and all other data shows that the effects are additive.
They don’t fade out, Cremieux’s chart is implying gains themselves are nonlinear… It uses the most biased design tho, and all other data shows that the effects are additive.
https://x.com/notcomplex_/status/1933296062367699311
@SashaGusevPosts For example, getting a 4 year degree is associated with 6-22(!!!) points of IQ gains, and the effects are additive in grade school too.
pmc.ncbi.nlm.nih.gov/articles/PMC35…
pmc.ncbi.nlm.nih.gov/articles/PMC35…
@SashaGusevPosts (the reason the avg iq of a college grad is not 122+ is that IQ tests are renormed pretty often and most people go to college nowadays. The gains are still there, an SD of 15 now is not the same as the SD of 15 5 decades ago)
@SashaGusevPosts "effects are overestimated due to measurement error" Well, this argument only applies to the design where one controls for prior intelligence. Most of our evidence are from regression discontinuity designs or policy changes.
@SashaGusevPosts @BrainyMarsupial
• • •
Missing some Tweet in this thread? You can try to
force a refresh
