IQ and Test Scores in UK Schools

The Economist’s Science and Technology section published this piece summarizing research by Robert Plomin et al. published in Nature’s Science of Learning (the original research can be found here). 

Plomin is one of the most influential psychologists of the day and one of the most active researchers around twin studies and heritability.

It was very interesting to see The Economist acknowledge that heritability of intelligence exists and plays a role in academic outcomes. However, it notes that the paper finds only 8% of variance in GCSE performance (national exams taken end of sophomore year in highschool) is the result of genetics, while the remainder is due to education and socio-economic background.

From this they conclude that “progressive eugenics” and, by extension, any other form of eugenics is not compelling. Progressive eugenics is the proposal that “poor people should be able to screen embryos for free on the basis of intelligence, if the technology to do so becomes available.” I assume that after this screening is allowed there is the opportunity for embryonic selection via IVF.

I thought I would take this opportunity to share some of my own thoughts on the study and in particular its commentary on the genetics of intelligence:

The highest correlations between IQ and desirable individual outcomes are around 0.5 (societally desirable outcomes have much higher correlations).

r is the correlation between intelligence and the measure of success, k is the number of studies included in the meta-analysis, N is the number of individuals included in the meta-analysis. (Correlations are taken from this.)

From this analysis it seems that the values of 8% should be quite a bit higher. There are a few likely culprits for this discrepancy:

  1. The above chart has correlations for intelligence, not genetics. During childhood heritability is around 30% while by the age of 21 it goes up to 60-80%. This means that the correlation between genetics contributing to intelligence will have a smaller correlation.
    • Moreover, due to childhood IQ malleability, the correlation of 0.58 may be exacerbated because having a good school will likely boost both IQ (temporarily!) and academic achievement.
  2. However, while here the value for genetics is only 8%, as mentioned above, by the time one is 21 the value for heritability has over doubled meaning that genetics should play a much larger contribution to test scores.
  3. The Plomin study used polygenic scoring for academic achievement rather than IQ. Polygenic scoring still has a long ways to go and draws a floor on heritability estimates than a ceiling.
  4. The GCSE exams being analysed have a low ceiling on performance with an A* (star) being the best possible result and one that is not all that difficult to get. This means that any natural variations in performance from intelligence can be compensated for by educational activity.
  5. I want to do a much deeper analysis on literature around correlation between IQ and income plus social mobility more broadly.  However, it seems that even if socio-economic status plays a role in one’s test achievements, one of the best ways to boost future income is through having higher intelligence rather than academic success.
  6. Finally, there is deep and compelling evidence not only for education fade out effects but also weak influence of education on one’s overall IQ. While this is not the end all be all way to success and happiness, it is very significant (here is an extensive literature review on the topic that we have produced, read all of it or head to the “Significance” section) and has a multiplier effect on every other aspect of one’s life, through the entire duration of their life. Educational interventions must be sustained and of high quality, they are also costly, and lead to domain specific knowledge rather than the universal reasoning improvements of IQ.


Leave a Reply

Your email address will not be published. Required fields are marked *

© 2018 Placebos & Panaceas · by MinimaDesign