Sponsored by:
We conducted the first molecular genetic analyses of individual income, using high-density scans of common genetic variants among individuals of European ancestries in three large population cohorts in the UK and the US. We find that ≈10% of the variance in occupational wages in the UK sample can be attributed to genetic similarities between individuals who are not or only distantly related to each other. A genome-wide association study (GWAS) in the UK sample (N = 282,963) identifies 45 approximately independent genetic loci for occupational wages, each with a tiny effect size (R2<0.04%). An aggregated genetic risk score constructed from these GWAS results accounts for ≈1% of the variance in self-reported income in the two independent US samples (N = 29,440) and improves upon the variance captured by a genetic risk score obtained from previous GWAS results for educational attainment. Furthermore, we exploit the random genetic differences between 38,698 biological siblings in a hold-out sample to show that roughly half of the covariance between genes and income is due to causal effects of genes. In particular, the sibling who was by chance endowed with the higher polygenic score for socio-economic status is more likely to obtain a college degree, to live in a better neighborhood and to have a well-paid job. Mediation analyses show that these random genetic advantages for socio-economic status are also linked with better health outcomes later in life, in particular lower BMI and lower waist-hip ratio. The genetic effects we identified partly work via malleable behavioral and environmental pathways. For example, roughly half of the genetic influences on income operate via educational attainment. Furthermore, the estimated returns to schooling remain positive and substantial even after controlling for genetic confounds, with on average 8% - 11% higher hourly wages for each additional year of education. Thus, while genetic factors contribute towards inequalities in education, income, and health, their effects are not deterministic.