A new study examining the genetics of educational attainment and childhood problem behaviour finds that genetic variants linked to higher educational attainment are also associated with fewer behavioural and emotional difficulties in children and adolescents, with the strongest patterns emerging through the cognitive component and notably in girls.
In behavioural genetics, researchers use a tool called a polygenic index, or PGI, to summarise statistical patterns across the human genome. Rather than measuring a fixed trait, a PGI is a mathematical summary that adds up thousands of tiny DNA associations identified in large population studies. When applied to complex outcomes like educational attainment, these indices reflect a blend of genetic variation, social structures and environmental opportunities.
Researchers classify childhood behavioural challenges into internalising problems, such as anxiety and distress, and externalising problems, including rule-breaking and aggression. Some areas, like attention problems, social problems and unusual thoughts, sit between these two categories.
Untangling how genes relate to these behaviours within families is tricky because parents contribute both their DNA and the home environment. In a 2018 study, researchers showed that parental genetics influence children partly by shaping the rearing environment—a dynamic known as genetic nurture. To isolate direct genetic associations, researchers must account for parental DNA to avoid misleading conclusions.
Educational outcomes reflect more than test scores—persistence, curiosity and emotional stability all play a part. Using a technique called GWAS-by-subtraction, a 2021 study split the genetic signal for educational attainment into two parts: a cognitive component tied to performance on cognitive tests, and a “non-cognitive” remainder capturing other traits such as motivation or personality. This separation lets scientists test whether the genetics of educational attainment operate mainly through cognitive channels or broader behavioural tendencies.
Separately, a 2020 study showed that a child’s educational attainment PGI related to fewer attention and hyperactivity symptoms, even after controlling for parental DNA. Since problem behaviour is often linked to socioemotional skills, this provided a natural test of whether decomposed PGIs map onto behaviour as theory would predict. That work paved the way for the current team’s investigation into specific genetic pathways and whether patterns differ by sex or informant.
Lead author Lyydia Alajääskö, a Ph.D. researcher in economics at Vrije Universiteit Amsterdam in the field of genoeconomics, explained: “The educational attainment polygenic index (EA PGI) has become a very powerful predictor in sociogenomics, and is also being adopted into other fields like economics and sociology. I was motivated to better understand what exactly the EA PGI signal contains.”
She added: “From economic theory, we have two types of skills that predict later-life success, cognitive and non-cognitive skills. I was able to test whether the cognitive/non-cognitive distinction holds up when you look at a behaviour that should sit clearly on one side of it: childhood problem behaviour. This is a novel test of this theory and a novel use of the decomposition approach.”
The Lifelines study, a multi-generation health project based in the northern Netherlands, provided the data for the analysis. The researchers examined 4,008 unique children and adolescents aged 4 to 17, contributing a total of 7,787 observations across multiple survey waves. All participants were of European genetic ancestry, with most observations relying on reconstructed parental genomes rather than direct measurements.
Behavioural difficulties were assessed through standardised surveys from the Achenbach system. Parents completed 4,611 observations, while adolescents aged 13 to 17 provided self-reports across 3,176 observations, covering eight specific behavioural areas from anxiety and physical distress to attention problems and aggression.
The study produced overall educational attainment PGIs for each participant and split them into cognitive (Cog) and non-cognitive (NonCog) components. Regression models controlled for sex, age, birth year, survey wave and ancestry, with a strict threshold to guard against false discoveries. By comparing models with and without parental PGIs, the researchers separated the child’s direct genetic effects from indirect family-wide effects.
The findings indicate that higher overall EA PGIs tended to be associated with lower levels of certain problem behaviours, but robust associations were limited to a subset of areas. These associations primarily traversed the cognitive component rather than the non-cognitive component. Specifically, higher Cog values were robustly linked to fewer attention problems in parent reports and fewer physical complaints related to emotional distress in youth self-reports, with a possible reduction in anxious or depressed symptoms in parent reports as well, though the latter was less consistent across models.
In contrast, the non-cognitive component showed weaker and less consistent associations with youth behaviour, not retaining statistical robustness across the main specifications.
Alajääskö cautioned that the lack of robust findings for the non-cognitive path does not automatically render it irrelevant: “We could lack power to detect effects, and there are statistical limitations to the GWAS-by-subtraction decomposition in the first place. If anything, our results hint that the cognitive/non-cognitive distinction might partly reflect a shared underlying domain rather than fully separate pathways – though that remains speculative.”
Parental PGIs did not emerge as statistically significant predictors of children’s problem behaviour in the primary models, suggesting the associations were not largely driven by the home environment shaped by parents’ genetics. The authors noted power limitations in detecting very small indirect effects and that direct associations remain a blend of biology and how traits influence responses from others.
When the data were split by sex, marked differences appeared. Among girls, the cognitive component robustly predicted lower levels of anxious or depressed symptoms, unusual or obsessive thoughts, attention problems and aggression in parent reports. In girls’ self-reports, the overall EA PGI correlated with fewer physical complaints and internalising symptoms, while the cognitive component remained robust only for physical complaints.
Among boys, several initial associations emerged, but none withstood inclusion of parental PGIs in the models, diminishing their robustness. The paper also notes that the respondent type mattered: the link between the cognitive component and fewer physical complaints was stronger in self-reports than in parent reports in the full sample, while informant differences were less pronounced for other behaviours.
The authors propose that cognitive skills and behavioural regulation may share an underlying neurocognitive foundation, such as executive functioning, though they emphasise that executive function was not directly measured in the current study and this remains speculative.
Commenting on the work, Martin Fieder, associate professor of evolutionary demography at the University of Vienna who was not involved in the study, described the research as methodologically sophisticated. “The main message is that the genetic association between educational attainment and childhood problem behaviour seems to operate more through the cognitive component of the genetics of educational attainment (i.e., performance on cognitive tests) rather than through the ‘non-cognitive’ component (i.e., the remaining component of genetic variation associated with educational attainment),” he said.
“Accordingly, genetic differences associated with better cognitive performance and higher educational attainment also appear to be associated with fewer behavioural problems, providing evidence for genetic overlap between cognitive performance and childhood behavioural regulation,” Fieder observed.
He added: “The study confirms previously reported genetic overlap between educational attainment and childhood problem behaviour, but extends this work by showing that the association appears to be driven predominantly by the cognitive component of educational-attainment-related genetic variation.”
Fieder also urged replication: “As always, I would like to see a replication study using an independent dataset.”
The authors, Lyydia Alajääskö, Marina Aguiar-Palma, Abdel Abdellaoui and Titus Galama, titled their work Genetics of Educational Attainment and Childhood Problem Behaviour: Direct Genetic Effects, Cognitive Pathways, and Sex Differences.
