Ok, deep breath, I am going to do a thread on this paper on the genetics of income from the Deary lab.
tl;dr income is broadly downstream of intelligence, and common variant genetics has a v. modest role in determining income, but it is complex.
nature.com/articles/s4146…
tl;dr income is broadly downstream of intelligence, and common variant genetics has a v. modest role in determining income, but it is complex.
nature.com/articles/s4146…
For many people the initial reaction on reading the title or abstract is "what on earth are these scientists trying to show?" (sometimes with a "what is their agenda?" undercurrent ).
As well as curiosity, a big motivation is that income levels are used all over the place in behavioural or social studies as covariates/something to regress out, adjust for, or stratify. An important question therefore is what things are we doing/adjusting for with income levels?
(I know some people now just want to put their fingers in their ears and say "I don't even want to know" because this walks into such societal/debated areas. Not knowing though is just.... being comfortable with ignorance. I'm not a fan).
Onto the paper. They used the amazing UK BioBank (500,000 healthy on recruitment, middle aged, middle class Brits - the world's best studied group of people now and for the next decade) and end up doing genetics on 200,000 (survey variable response; need for a well mixed pop)
Their GWAS is good (note the LD score regression about polygenic vs stratification) and in this "vanilla" take on income their find 30 loci. Those 30 loci overlap with previously identified loci for measures of intelligence / educational attainment and some neuronal function
Broadly one can trust this GWAS - the thing to worry about is cryptic population stratification which (a) would have a poor LD score regression and (b) the PRS later on with the generation scotland cohort would not have worked (c) neuronal biology overlap wouldn't have happened
However, its *not* much variance explained - 7% - from this model. Ie - *no surprise* - the vast majority of variation of income in a population is not due to genetics. I think the authors would have been better putting this in straightforward sentence with figure in the abstract
Furthermore this 7% is in the subset of the UK BioBank which itself is a biased sample (more educated, more wealthy, classically "university professors") of UK. This means the "real" contribution of common variant genetics to income levels in the UK has to be lower than 7%
(It's frankly complete guess work to take this number and adjust it but it's definitely going to be below 5% of variance explained. Rare variants - not in visible in this study - might up that again - something I will return to).
So, like many weak genetic effects due to the ability to measure genetics at scale we can find this small level ... but it is definitively showing that genetics has a very MODEST effect on income. Again, I think geneticists do a disservice by not stressing this.
They then use the unbiased and definitely "upstream" aspect of genetics to try to tease apart how variation in income comes about - graphically shown in figure 1. Here one is using genetics almost as a tracer through the system.
Something they show is that measures of intelligence are "upstream" of income - one could state this just from first principles and time based view point (measures of intelligence, eg exams, IQ tests, happen way before income). Unsurprising but nice to see.
The other thing which caught my eye was that there is something about income levels which is not just intelligence related. I'd need to rummage around in the supplement and the data because this sort of regressing/conditioning stuff can be fragile
They also recapitulate long standing associations between measures of intelligence with psychiatric traits, such as ADHD, Schizophrenia and Autism - income with similar components to intelligence, with this strange positive association with Autism
Stepping back - what does this paper show? For me it shows two big things: (a) there is not much genetic contribution to income - ie, it is broadly valid to regress out/use as a covariate most of the time; be careful though wrt to intelligence/education measures
(b) broadly speaking the unsurprising idea that income is downstream of intelligence/education (duh!) holds up, and this is where a fair chunk of the genetics is flowing, but interesting not all - I think that's worth following up. A sort of "savvy" intelligence?
Remember that in this world of common variation, if there are variants that increase the risk of diseases like... schizophrenia, and if these diseases reduce education and income levels (no surprise!) then this paper is ... to be expected.
Could the authors do better in explanation? There is a healthy discussion in the paper but not enough context in my view. I think the paper would have been better with the 7% number in the abstract and more framing on modest (but important to know about) effects.
It should go without saying given that I am confident there is no good evidence of genetics underlying differences in intelligence between ethnic groups that this is even further "downstream" of intelligence and *absolutely nothing in this paper* speaks ethnic group difference
The authors state this (in more science speak) in the discussion but for sure someone is going to just stick the title on some mad part of the internet. This is why geneticists in this area have to up their communication game!
