Parent and offspring polygenic scores as predictors of offspring years of education

A polygenic score (PGS) derived from a 2013 genome-wide association study (N=127,000) explains ~2% of the variance in years of education (EduYears). Here, using a PGS derived from a 2016 study (N=329,000), we tested the... [ view full abstract ]

A polygenic score (PGS) derived from a 2013 genome-wide association study (N=127,000) explains ~2% of the variance in years of education (EduYears). Here, using a PGS derived from a 2016 study (N=329,000), we tested the association between the PGS and EduYears within families in an independent sample of 2,517 twins and their parents from the Minnesota Center for Twin and Family Research.

Our first important result is that within the same DZ twinship, the twin with the higher EduYears PGS is significantly more likely to have a higher IQ (p < 0.01) but not a higher EduYears. This is perhaps consistent with the ability of parents to provide educational opportunities to their less able children. We also found that while offspring PGS predicts up to ~9% of the variance in their own EduYears, adding the midparent PGS significantly increases the R² by about a percentage point. Furthermore, the addition of parental characteristics such as SES, years of education, and IQ to the model diminishes the value of the midparent PGS in predicting offspring EduYears, suggesting that these variables lie on an environmental causal path between parental genotype and offspring EduYears. Interestingly, we observe no such pattern of results when the outcome variable is offspring IQ. That is, while the offspring PGS significantly predicts offspring IQ, the midparent PGS adds no significant increment to the R². 

These findings represent the first use of polygenic scores for years of education in clarifying the effects of passive gene-environment correlation on trans-generational outcomes, which may be crucial for disentangling the transactions between genetic and environmental influences on educational attainment.