Gene-by-father’s income interactions shape educational attainment

Abstract

This paper improves upon existing measures of childhood financial socioeconomic status (SES) in the Health and Retirement Study (HRS). As a new measure of childhood financial SES, it predicts income of the fathers of HRS respondents when these respondents were 16 years old, using a model calibrated on US census data. These income predictions are up to four times as predictive of educational attainment as are other HRS measures of childhood financial SES. The income predictions are used to estimate how genetic propensity towards education, father's income during childhood, and their interaction shape educational attainment. For the mean respondent, I find that the effect of a standard deviation decrease in the polygenic score for educational attainment is offset by an increase in father's income of 42% to 47%. Arbitrary gene-by-environment correlation may confound the OLS estimates of G×father's income interactions. To address that issue, I construct a Bartik instrument as a plausibly exogenous source of variation in income. It is positively associated with predicted father's income, and is unassociated with predetermined individual characteristics, including genetic propensity for education. The Bartik IV results are in line with the OLS results, but are imprecisely estimated. Although weak in this setting, the Bartik instrument is a promising strategy to obtain causal G×E estimates in the presence of substantial gene-by-environment correlation.