Genes and Diplomas: New Study Maps How DNA Signals Shift Across Each Educational Milestone

For more than a decade, geneticists have been chipping away at one of the most studied traits in behavioral science: how far people go in school. The landmark genome-wide association studies of educational attainment, which now aggregate millions of participants, have produced polygenic indices that predict years of schooling with modest but real accuracy. Yet almost all of that work rests on a single, deceptively simple measurement—the total number of years a person spent in education. A new study published in PLOS Genetics argues that this familiar yardstick may be hiding as much as it reveals, and that the hidden structure becomes visible only when researchers stop counting years and start tracking the discrete milestones that actually shape an educational career.

The research, led by Eirik Haugland Kvalvik and colleagues including Yunpeng Wang, Kristine Beate Walhovd, Torkild Hovde Lyngstad, and Ole Røgeberg, drew on two extraordinary Norwegian data resources: the Norwegian Mother, Father and Child Cohort Study, known as MoBa, with 120,527 participants, and the Norwegian Twin Registry, contributing 8,910 individuals. Norway is an ideal setting for this kind of analysis because its education system is standardized and its registries are unusually complete. Rather than collapsing educational histories into a single number, the team treated each credential as its own event: finishing high school, earning a bachelor’s degree, completing a master’s, and reaching a doctorate. That reframing allowed them to ask a question the standard design cannot answer—do the genetic associations with education stay constant as people climb the ladder, or do they change from rung to rung?

The technical machinery behind the study combined three complementary approaches. First, the researchers ran genome-wide association studies, or GWAS, which scan hundreds of thousands of genetic variants for statistical links to each milestone. Second, they built polygenic indices, abbreviated PGIs, which condense the tiny effects of many variants into a single score for each individual and can be tested for how well they discriminate between people who do and do not reach a given milestone. Third, they applied twin models, which compare identical and fraternal twins to estimate heritability from family resemblance rather than measured DNA variants. Running all three on the same cohorts let the investigators cross-check their conclusions in ways no single method allows.

The headline finding is a striking inverse-U pattern. When the team analyzed each transition separately—conditioning on having reached the previous milestone—the heritability attributable to measured common genetic variants, known as SNP heritability, and the predictive power of the polygenic indices both rose and then fell across the educational sequence. The peak came at the transition from high-school completion to attaining at least a bachelor’s degree, where SNP heritability reached roughly 0.14 and the polygenic index achieved a Tjur R-squared of about 0.05, a measure of discrimination for binary outcomes. Beyond that point, at the postgraduate transitions, both measures declined. In other words, the genetic signal is strongest in the middle of the educational distribution and weakest at its very top.

That decline at the doctoral transition comes with an important statistical caveat that the authors are careful to flag. Estimates for the master’s-to-PhD transition were computed only among individuals who had already attained at least a master’s degree, a highly selected group. This creates range restriction: when nearly everyone in the analyzed sample shares the preceding credential, the variation available for genetic associations to explain is compressed, and attenuation of the estimates may partly reflect that selection rather than a genuine weakening of genetic influence. The study does not claim that genes stop mattering at the doctoral level; it claims that the standard interpretation of per-year effects cannot tell you what is happening there at all.

A second set of analyses probed how the genetics of education relate to the genetics of intelligence, a question that has long shadowed the field. The researchers computed genetic correlations between their milestone-specific results and large-scale GWAS of educational attainment and of intelligence. The pattern was clear: genetic overlap was high at earlier transitions but dropped at later ones. With the widely used EA4 educational attainment GWAS, the genetic correlation was approximately 0.92 at the transition from high school to a bachelor’s degree, but only about 0.38 at the transition from master’s to doctoral attainment. Sensitivity analyses using an alternative intelligence GWAS and additional polygenic indices reproduced the same transition-specific trajectories, suggesting the pattern is not an artifact of one particular reference panel or scoring method.

The twin data added a third perspective, and one of the study’s most surprising results. In cumulative analyses of attained status across the full cohort, the gap between twin-based heritability estimates and SNP-based estimates—the so-called missing heritability—shrank as the milestones became more advanced. At high-school completion, the point-estimate gap was roughly 0.38; at the doctoral level, it narrowed to about 0.11. This convergence challenges a common intuition that rare, hard-to-detect genetic variants must account for the missing heritability in complex traits. It suggests instead that part of the discrepancy in education research may arise from how the phenotype itself is constructed, with years-of-schooling scales blending together biological and social processes that behave differently at different points in the sequence.

Genetic correlations between the milestones themselves told a complementary story. Adjacent milestones—say, high-school completion and a bachelor’s degree—showed very high genetic overlap, around 0.92, indicating that largely the same set of variants is associated with both. But distant milestones were less genetically aligned: the correlation between high-school completion and a doctorate was approximately 0.71. The genetic architecture of education, in other words, is not a single unified target. Different stages of the educational career share much of their biology with their immediate neighbors while drifting apart from stages further away, a structure that a single years-based phenotype is mathematically incapable of expressing.

Why does this matter beyond the statistics? Years-of-education GWAS convert distinct credentials and progression steps into one numerical scale, and the resulting estimates are routinely reported as effects per additional year of schooling. The study’s central warning is that such estimates should not be read as though one year carries the same meaning across the entire educational sequence. A year that ends with a high-school diploma, a year in the middle of an undergraduate program, and a year of doctoral research are not interchangeable units of exposure; they sit at different points of a selective, staged process with different social and biological correlates. Polygenic indices built on the years-based phenotype inherit that ambiguity, which may explain why their predictive performance varies so sharply across the milestones examined here.

The practical implications reach into several active debates. For researchers using educational attainment as a proxy in studies of health, fertility, social mobility, or cognitive aging, milestone-preserving analyses offer a way to check whether conclusions depend on how the phenotype was coded. For geneticists, the results suggest that future GWAS of education could gain resolution by modeling transitions explicitly rather than pooling them. And for anyone interpreting polygenic scores in social science, the study is a reminder that a score’s predictive power is not a fixed property but a function of where on the educational ladder the question is being asked. By making the hidden structure of the years-based measure visible, the Norwegian team has sharpened the interpretation of education GWAS and polygenic indices alike—and shown that sometimes the most informative move in genetics is not to gather more data, but to look more carefully at the phenotype already in hand.

Subject of Research: Genetic associations with discrete educational milestones in Norwegian cohorts

Article Title: Beyond years of schooling: Genetic associations across educational milestones in two Norwegian cohorts

Article References: Kvalvik, E. H., Wang, Y., Walhovd, K. B., Lyngstad, T. H., & Røgeberg, O. (2026). Beyond years of schooling: Genetic associations across educational milestones in two Norwegian cohorts. PLOS Genetics, 22(9), e1012310. https://doi.org/10.1371/journal.pgen.1012310

Image Credits: AI Generated

DOI: 10.1371/journal.pgen.1012310

Keywords: educational attainment, genetics, GWAS, polygenic indices, twin studies, heritability, Norway, MoBa, intelligence, doctoral attainment, SNP heritability, genetic correlation