What have twin studies taught us about educational attainment?

The development in the field of the behavior genetics of educational outcomes has been nothing less than astonishing in the recent decade. Numerous twin studies have demonstrated that roughly half of the individual differences in academic achievement could be attributed to genetic factors. The same studies have also shown important contributions of both family-wide and individual-specific environmental factors. Molecular genetic investigations have thus far identified over a thousand genetic variants that are associated with educational attainment. Researchers are now using these discoveries to predict a wide range of different outcomes in external datasets.

In a series of blog posts, I intend on summarizing some of the important studies in the field. Given the nature of blog posts, this will not be a systematic review of the literature but my aim is to make it as representative as possible.

Why twin models?

The rationale for the classical twin model is the estimate the relative contributions of genetic and environmental influences to a trait of interest. The classical twin model is based on the assumptions that identical twins share all of their co-segregating genes whilst fraternal twins share half of their co-segregating genes. The model further assumes that the early environmental influences are roughly equal across both types of twins. There are a number of additional assumptions as well but the validity of the model remains strong (1, 2). Given these model assumptions, the presence of genetic influences is determined by the extent to which the correlation of a given trait is higher in identical twins as compared to fraternal twins. 

By measuring a trait of interest in a sufficiently large group of identical and fraternal twins, researchers are able to examine to what extent the individual differences in the trait could be attributed to three sources of influences; (A) additive genetic influences or the heritability, (C) shared environmental influences, and (E) unique environmental influences. Shared environmental influences encompass an aggregate of all non-genetic factors within the family that make the twins more similar to one another in regards to the trait of interest. These could include factors like parenting practices and socioeconomic status. Unique environmental influences is a somewhat misleading term as it includes all factors that make the twins different from one another. These include individual-specific environmental risk factors but also measurement error and random (stochastic) events. 

The evidence from the UK
The Twins Early Development Study (TEDS) is a representative study of approximately 16,000 twin pairs in the United Kingdom. In a study of academic achievement measured at the age of 16 years (GCSE exam scores), the TEDS researchers found that an excess of 50% of the individual differences between the twins could be attributed to genetic factors (3). Furthermore, the researchers found that environmental factors shared within the family explained between a fifth to a quarter of the individual differences whilst the rest of the variation was attributed to unique environmental influences. In other words, both nature and nurture explained why the students differed in their achievements. 

In a follow-up study (4), the TEDS team were interested in examining the relative contributions of genetic and environmental influences on intelligence and specific school subjects:

Similar to the previous study, the researchers found that an excess of half of the individual differences across all of the examined outcomes could be attributed to genetic differences in the population. Among the school subjects, shared environmental influences explained a larger proportion of the individual differences than did unique environmental influences. In contrast, they found that shared environmental influences explained a very small proportion of the variation in intelligence.

The second aim of this study was to examine how intelligence and the specific school subjects were related to one another by adopting multivariate twin models. This approach enabled the researchers to study how genetic and environmental influences explained the associations between the outcomes. To put this in more simpler terms, one could say that if the association between two outcomes was stronger among identical twins as compared to fraternal twins then it implied that genetic factors explained a proportion of the observed association in the sample. 

The researchers found, rather expectedly, strong associations between the measures of achievement in the specific subjects and intelligence. In other words, this meant that students who achieved well in a particular subject were on average more likely than their peers to achieve well in other subjects. Interestingly, the researchers also found that the genetic influences overlapped to a large extent between the outcomes. Expressed more technically, the observed genetic correlations between each combination of the specific school subjects and intelligence varied between 44% and 88%. These findings support the existence of both generalist genes that explain differences across all of the examined outcomes and specialist genes that only explain differences in the specific outcomes. The shared environmental correlations were even stronger, ranging between 66% and 95%. This implies that shared environmental factors might be influencing academic performance more generally across all of the subjects rather than being subject-specific. In contrast, the unique environmental correlations were rather low, ranging between 9% to 35%, which could either suggest that the individual-level risk factors were subject-specific or that the estimates were attenuated by a combination of measurement error and random noise.

What’s included in the high heritability of academic achievement?
To many readers, it may seem somewhat strange that the heritability estimates of the academic achievement measures are so large. One of the many advantages of the TEDS study is its comprehensive data collection measuring the participant’s abilities, personality, problem behaviors and living conditions. This allowed the researchers to study how such factors overlapped genetically with school performance, as measured by an index of the GCSE core subjects (5):

The heritability of school performance was estimated to 62% and genetic factors that explained individual differences in intelligence accounted for about half of it. Together, all of the measures explained roughly three quarters of the heritability. 

The stability of school performance
The TEDS team have also examined the stability of school performance and its relative determinants among the twins from primary school up to secondary school (6). The study found that the contributions of genetic and environmental influences were relatively stable across time and subject: 

Similar findings in the Netherlands
Dutch behaviour geneticists have extensively studied measures of educational attainment for a number of decades and their findings have been similar to their British colleagues. In a 2015 study of 12 414 twin pairs, for instance, they observed that the heritability estimates for different indicators of educational attainment measured during primary school varied between 44% and 73% (7):

Even academic choices are heritable
A fascinating line of research has demonstrated that heritable factors do not only contribute to individual differences in academic performance but also to our choices to pursue higher education and the subjects that we wish to specialize in. Importantly, these studies have also shown why twin studies are important to get gain greater insights into how environmental factors also explain such differences.

In a 2016 paper of 6584 twin pairs (8), the TEDS researchers examined how genetic and environmental factors influenced the choice to pursue university preparatory courses (“A-levels”). They additionally examined the subject-specific courses that the students elected.

The results indicated that heritable factors explained an excess of 40 percent of the individual differences across all of the outcomes. The researchers nevertheless observed that whilst shared environmental factors explained roughly the same amount of variation in the choice to pursue any of the A-level courses, their contributions to the subject-specific courses were substantially lower. One interpretation of these findings could be that family members encourage the students to pursue higher education but that they may be less concerned with the specific subjects that the students choose.

You’re on your own once you’ve reached the university
In one of their most recent studies, the TEDS team examined university outcomes among 3000 of the twin pairs (9). Similar to the previous study, they found that genetic and shared environmental influences contributed roughly equally to the choice of enrolling at university. Again, this is consistent with family members encouraging the students to pursue higher education. Interestingly, the authors failed to find that such factors influenced how the students performed in their university courses. The latter was instead explained by genetic (46%) and unique environmental (54%) influences.

Highest achieved education
A 2013 meta-analysis estimated the heritability of educational attainment as defined by the highest achieved educational level to be approximately 40% (10), which is somewhat smaller in magnitude to those presented above. This illustrates the importance of outcome definitions. A crude measure of highest achieved educational level will rank individuals with very heterogenous abilities equally. For instance, two individuals with master’s degrees in astrophysics and history, respectively, will, on average, have very different abilities. The best outcome definition will therefore depend on what the specific research question of interest is.

Where are the genes?
In my next blog post, I will cover the recent advances in the molecular genetics of educational attainment, from genome-wide association studies to studies that adopt genetic risk scores to predict a wide range of outcomes. Please subscribe to the blog to stay updated.