I found this study published in Nature via prof. Wouter Duyck and while I know some people will get itchy feelings because of it, it’s worth sharing.
First of all, although it’s a twin pair study, it’s not only about heritability, as the researchers also looked at the actual DNA in relation to university enrolment. Let’s jump to the conclusion (pun intended):
…Twin analysis revealed substantial heritability for all university success measures, including university entrance exam achievement (57%), the choice to study at university (51%), the quality of university attended (57%) and achievement at university (46%). In addition to twin analysis, we also found evidence for genetic influence using DNA alone. Indeed, a genome-wide polygenic score (GPS) for adult educational attainment explained up to 5% of variance in the university success variables. Taken together, these results highlight that the appetite and aptitude young adults have for higher education is, in part, genetically influenced.
The crucial words in this conclusion is of course in part. Often people get itchy because genetical and heritability research is often associated with determinism, which it shouldn’t be. Our genetics give us the natural possibilities which will interact with the environment one lives in. And do note: some people will find up to 5% of variance explained underwhelming, but do regard this as an ongoing search to discriminate different factors.
For me more surprising was this result:
Indeed, shared environmental influences account for up to 20% of the variance in the compulsory school years, but make a non-significant contribution to variance in achievement at university.
But (self-)selection might be the explanation:
One explanation for this pattern of results is that in the early school years children’s environments are largely the same across multiple life domains, for example siblings go to the same school, have many of the same friends, and spend much of their time at home under substantial parental influence. By contrast, young adults have more freedom of choice in their education, in terms of the subjects they take, the extracurricular activities they engage in and how they spend their time. This increase in choice leads to greater genetic influence and decreased shared environmental influence across development. We see this same developmental decrease in the influence of the shared environment for other educationally-relevant traits such as intelligence. This suggests that as children gain more freedom to choose their environments, they increasingly select environments that correlate with their genotype.
While this is imho a bit more frightening, there is also a kind of consolation:
An exception to this developmental decrease in shared environmental influence pertains to decisions about whether to continue in education. For example, shared environmental influences account for nearly 40% of the variance in the choice of whether or not to take university entrance exams and the choice of whether or not to pursue a university degree, yet shared environmental influence is not evident for university achievement. It is possible that families and schools influence educational choices to a greater extent than educational achievement.
This is why this kind of research can be so relevant, as it can help distint which influence is working on what. Do note, there are a lot of guesses still around. And even a paradox:
Although we found moderate twin heritability estimates for university achievement, the polygenic score prediction of this trait was small in magnitude, only predicting 0.7% of the variance. Furthermore, even when we split university achievement into two groups: STEM-related subjects and humanities subjects, we did not find any differences in EduYears GPS prediction between subjects. This suggests that even within subject field, the GPS is not discriminative of achievement. In contrast, EduYears GPS explains 9% of the variance in achievement at age 16. There are several possible reasons for this paradox. First, a polygenic score based on years of education might be less discriminative for individuals who have all obtained a university degree. Second, examinations at university level are not standardized, which means that results may be less comparable between universities; a first-class degree at an elite university will be weighted the same as one from a lower-level university.
To conclude, for me the worst sentence is actually in the abstract:
…and highlight the potential for DNA-based predictions of real-world outcomes, which will continue to increase in predictive power.
Why? Imho, this can invite people to think more deterministic, more deterministic science will probably ever allow.
Abstract of the study (open access):
University success, which includes enrolment in and achievement at university, as well as quality of the university, have all been linked to later earnings, health and wellbeing. However, little is known about the causes and correlates of differences in university-level outcomes. Capitalizing on both quantitative and molecular genetic data, we perform the first genetically sensitive investigation of university success with a UK-representative sample of 3,000 genotyped individuals and 3,000 twin pairs. Twin analyses indicate substantial additive genetic influence on university entrance exam achievement (57%), university enrolment (51%), university quality (57%) and university achievement (46%). We find that environmental effects tend to be non-shared, although the shared environment is substantial for university enrolment. Furthermore, using multivariate twin analysis, we show moderate to high genetic correlations between university success variables (0.27–0.76). Analyses using DNA alone also support genetic influence on university success. Indeed, a genome-wide polygenic score, derived from a 2016 genome-wide association study of years of education, predicts up to 5% of the variance in each university success variable. These findings suggest young adults select and modify their educational experiences in part based on their genetic propensities and highlight the potential for DNA-based predictions of real-world outcomes, which will continue to increase in predictive power.