What your genes mean for your education

Zip codes have an impact on educational outcomes. Having a teacher that looks like you has an impact on educational outcomes. Having students that are well fed in a comfortable environment has an impact on educational outcomes. Keeping students away from pollution has an impact on educational outcomes. Class can have an impact on educational outcomes, especially given the relative long-term stability of intergenerational transfer of wealth, as Thomas Piketty flagged in Capital.

But a study recently published in Nature seeks to continue recent research by arguing that genes themselves have an impact on educational outcomes in the U.K. and that ‘success’ in education is something that can be inherited on from one generation to the next.

Currently existing research seems to suggest that this method of prediction can be difficult to do well: an international team of scientists published a study of over one million individuals of European descent that sought to determine how genes influence how long someone stays in school this summer. Yes, the study was able to statistically predict the achievement of 11% of those surveyed as a collective group — which is an achievement in and of itself — but “when the team tried to use these genetic variants to explain differences in schooling time among African-Americans,” Carl Zimmer writes in the Times, “the predictions failed.” The predictions also failed — as Ed Young notes in The Atlantic, with handy charts to back all this up — when it came to “predicting the outcome for any specific individual.”

Which brings us back to the study of twins by Kaili Rimfeld at King’s College London along with others. To study twins is to study genetic difference between two individuals and the impact the environment might have on the twins. (Given how twins share so many traits — they come from the same womb, et. al. — it is easier to track the impact of environmental factors than not.) To examine heritability is to examine the statistical variation of certain traits — i.e., a twenty-percent likelihood that your eyes might be a certain color as opposed to something else. It’s also worth noting that there has been long-standing criticism of using twin studies to determine heritability. As was pointed out in 1996 —

” …the ‘heritability’ estimate – often taken to mean the influence of trait-specific genes – is merely a statistical abstraction derived from a matrix of correlations; this estimate encompasses many buried environmental effects so that ‘heritability’ does not correspond to any underlying DNA structure.”

In the study from Rimfeld and others, they posit that “twin studies have shown that around 60% of individual differences in school achievement are explained by inherited differences in children’s DNA sequence.”

In explaining how they went about getting these results, the study opts for a path that pre-presumes knowledge. It does not go out of its way to earn any awards for rhetorical lucidity.

So let’s go through some of what the study goes through together.

To begin with: the study is probably at its most effective if you think of it as being similar to the study that looked at the linkage between genetic makeup and time spent in school mentioned just a moment ago — that there is a wide-angle lens of statistical truth here that can be examined and mulled over by folks who have a role in education policy.

But, the specifics: when the study speaks of a student’s ‘school achievement,’ the study means to say that they’re talking about “teacher ratings, exam performance, and achievement scores.”

The study does not make its own case in terms of linking a particular genetic trait to eventual potential intergenerational transmission, which would then give lay readers a better potential understanding of the overall argument being made. It could mention particular studies where a single variation in a single nucleotide — that is, the things that become your DNA and RNA — show certain statistical trend lines of being linked to certain educational outcomes when it comes to reading. The study could then talk about how children born to dyslexic parents can be loosely divided into being at a low-risk of developing dyslexia or at a high-risk and that “dyslexia is influenced by a wide range of environmental and genetic risk factors.” The study could then talk about ways in which teachers could engage a student with a learning disability in the hopes of boosting their achievement. The study could then show us how this all plays out at the appropriate level of abstraction.

But this study doesn’t do that. Indeed, that isn’t this study’s purpose or its prerogative. The study’s prerogative is to explain how they determined the math that they determined — the specifics of which you can read for yourself here, but which I will summarize as such: they looked at the average of and of variation in traits in five different kinds of twins. The educational estimates they ended up with were similar to or comparable to already existing scores related to “teacher ratings, exam performance, and achievement scores that combined teacher ratings and exam performance.” The study also notes that there is a low statistical variation in the kinds of traits one would assume are associated with learning (the study — in this section — doesn’t quite make this explicit) amongst twins.

All well and good, more or less, but trouble arises when the study has a go at defining what they mean by ‘heritability’ —

“The proportion of heritability at each age that is accounted for by genetic effects different from those operating at the previous age can be calculated by dividing the sum of the innovation path squared (Ai) and the age-specific genetic path squared (As) by the overall heritability.”

— a definition that only leaves us tempted to leap into the arms of the definition from 1996 once again. “Innovation path” and “age-specific genetic path” each seem to be coined phrases that seek to describe already existing work within the text and don’t have a bearing on the nature of the equation the study is using to determine heritability.

Based on the description, one would think that the equation would look like the result of dividing individual performance by each student and the genetic mean associated with each year (perhaps through a compilation of a wide range of various polygenic scores.) But the study also works to incorporate what was learned the year before into the equation as well.

And while working through the math here is — again — all well and good, it’s important to remember that — as you go forward reading this; even with certain controls put in place — twins aren’t a random sample of the population. Twins are often “born premature” — up to sixty percent! — “have lower birth weights, and are at increased risk for perinatal complications and perinatal death compared with singletons.” And though premature births don’t necessarily have a negative effect on educational outcomes, that doesn’t mean that that’s something that should automatically be factored out of eventual analysis determining the role of your genes in assisting your educational outcome, let alone the open question as to how policy can go about seeking to assist this in a positive, proactive way while thoroughly avoiding pop culture Social Darwin implications.

Or, to put it all another way: “Despite the facts that the classical twin studies are still being guided by assumptions made back in the 1920s and that the inherent limitation lies in the study design itself, the results suggested by earlier twin studies have often been confirmed by molecular genetic studies later.”

In other words: it might make sense for folks to start thinking about what these results might mean now just in case they might be confirmed by a second similar study in the future.