The Good News Bible of Educational Attainment

In an effort to keep the "Educational Attainment" study relevant, we get yet more mileage from this polygenic risk score in another study.  It's a Plomin study, so you know where it's going, of course.  What strikes me most about this study is the absolute optimism and failure to consider than any of the findings can be anything other than proof of the genetics of "educational attainment," an absurd notion if you really turn off your filters for a second.  So let's go through it, a bit, shall we? 
Let's start with the opening sentence, which I think sets the scene for all of the rest that follows:

Ever increasing sample sizes and methodological advances in polygenic methods have made it possible to powerfully predict complex traits such as cognitive abilities without knowing anything about the causal chain between genes and behaviour. 

The question, of course, here, is what you are actually predicting? 
 The assumption off the bat is that is that there is an actual "trait," variably defined as intelligence, cognitive ability, or most speciously, "educational attainment," and this trait is a function of hundreds or thousands of genes and that slight variations in these genes (SNP's) are what makes the difference between an educational attainer and a non-educational attainer.  This is not based on anything other than the failure to find a smaller number of specific genes that convey such a phenotype.  If we can't find a few, there must be very many, the logic goes.  Thus, it is assumed, there is a "causal chain," between genes and behavior.  How do we map out or describe that causal chain?  Well, we don't bother.  Since we can match some genetic variants to a slightly higher percentage of educational attainers than non-educational attainers, then that somehow proves that there is a causal chain to be discovered later.
So we have no actual explanation as to how these genes or genetic variants would be able to convey educational attainment.  Do they make educational attaining enzymes, for example?  We don't even have a theory as to what the human mind is.  All we have is an easy to quantify trait (educational attainment) and a slight preponderance of certain genes for those who have a higher educational attainment.  Might there be other reasons than a "causal chain" for this preponderance?  One would certainly think so, if one were not one of the indoctrinated authors of this study, but no alternative is proffered (I will proffer one below).
The study also seems quite focused on getting the best possible results.  This is an interesting aspect of these polygenic score studies.  The idea here is that the most positive results are the most valid results.  I cannot think of another area of science where someone would claim such a thing, but it is certainly the implication of the authors of this study - examples below:

...The aim of the present study is to estimate how much variance in intelligence and educational achievement can be predicted by applying several state-of-the-art multivariate genomic approaches and leveraging highly powered GWA summary statistics.       ...In order to boost power of IQ3 (N = 266,450) and EA3 (N = 766,345) GWA results and thus precision of beta weights to construct more predictive IQ3 and EA3 polygenic scores, we jointly analysed these summary statistics with three cognitive and educationally relevant traits:      ...We used three recently developed multi-trait methods, one of which is specifically designed to boost polygenic score prediction: SMTpred       ...MTAG is a generalization of inversevariance weighted meta-analysis, which jointly analyses univariate GWA summary statistics. It boosts power for discovery for each trait conditional on the other traits and outputs traitspecific summary statistics 

So the intent is, obviously, to get as high a predictive score as possible, using whatever means are available.  The bigger the number, the better, apparently.  To what extent are these predictions a function of population stratification?  Well, the authors don't appear to consider this possibility at all.  They go from, "powerful" predictive ability to high (but unspecified) causality. 
Thus, I will consider it for them.  There are many ways to obtain a certain amount of population stratification, let's consider this one:  Suppose, for the moment, that there are no causal genes for higher educational attainment.  I will invent a term for this possibility that I will call "the null hypothesis."
I think that we can agree that people with high educational attainment, tend to marry other people with high educational attainment and that there is, shall we say, a "clustering" of individuals who marry amongst each other in a particular social group in which high educational attainment is bit of a prerequisite.  Sure, you can argue that these individuals have better educational attainment genes, but this assertion is unproven, as the authors appear to concede in the opening sentence of this study ("... without knowing anything about the causal chain between genes and behaviour. ") Is it not possible that what we are really discovering, is a somewhat closed gene pool.  In that sense, one would expect such individuals in that group to have many genetic variants in common.  Much like Ancestry.com and 23andMe use such clustering to determine racial/geographic heritage, we might be discovering genetic variations that have nothing to with educational attainment, but are merely markers for a particular group.  This might explain some, or all, of the genetic clustering we are seeing, so that the polygenic score predictions are really only showing that people are in this group.  Then we are dealing with basic social stratification.  These predictive scores then, would really only be highlighting disparities in our socioeconomic system and a certain lack of social mobility.
We can look at this study and some of the studies it is feeding from and get some clues that this might be the case. 
For example, the Educational Attainment study  (EA3), notes that their predictive value was not particularly effective for African Americans or other non-white/European populations.  If my point above is correct, this would be expected, since there is, frankly, very little mixing of racial populations, so we are talking about different gene pools.  If such is the case, then you would expect that, to the extent that a particular racial population intermarries based on educational attainment, we would find high polygenic scores based on entirely different genetic variants (SNP's).  It would seem odd that different races would have entirely different genes involved in educational attainment.  In fact, the most logical conclusion, in my view, would be that these polygenic predictions are mostly measuring population stratifications amongst different, closed populations.
A couple of other points from this study, I think, speak to population stratification.  The first is the fact that the predictive value for educational attainment significantly exceeds that for intelligence.  Even for the die-hard genetic view of these traits, this should seem to be counter-intuitive.  Educational attainment is a relatively recent phenotype in the history of mankind; far too recent to consider significant genetic development for the trait.  Conversely, intelligence has presumably been a quantifiable and useful phenotype dating back to prehistoric times.  Again, this suggests that we have a population stratification boosting the predictive results. 
The second point relates to the first.  That is the fact that educational attainment was more predictive with age, with 16 year olds being more predictive than 12 year olds.  To look at this simply, let's say you have 2 children who are C students at age 12.  Which of the two do you think would be likely to improve their grade performance and educational attainment by age 16?  The child with little support or the child with parents of a high educational attainment (presumed wealth) and the motivation to help their child improve performance (and perhaps pull some strings to help them advance)? 
If one really reads through these studies, what is most obvious is that the authors have an ideological bent that pervades their approach to the study.  Few quotes can be more illustrative of that than this in the closing paragraph of the study:

Second, unlike other correlations, associations between DNA variants and behaviour are causal from DNA to behaviour in the sense that there can be no backward causation from behaviour to DNA. These unique features will put genomic prediction of cognitive traits in the front line of the DNA revolution. 

The circular nature of this speaks for itself.