Evidence for the Scarr-Rowe effect on genetic expressivity in a large US sample

Abstract

Using the Continuous Parameter Estimation Method (CPEM), a large genotyped sample of the population of Wisconsin, USA (the Wisconsin Longitudinal Study, N=8,509) is examined for evidence of the Scarr-Rowe effect, an adverse gene x environment (GxE) interaction that reduces the heritability of IQ among those with low socioeconomic status. This method allows the differential expressivity of polygenic scores predictive of both educational attainment and IQ (EA3) on the phenotype of IQ to be operationalized throughout the full range of these variables. Utilizing a parental SES factor-weighted composite as a measure of childhood SES, evidence for the Scarr-Rowe effect was found, i.e. the genetic expressivity of EA3 on IQ increased with increasing parental SES (β=.064, p=3.82x10 -9 , df=8508). The effect was found for both the male and female samples separately (β(males)=.051, p=.002, df=4062; β (females)=.076, p=1.76x10 -7 , df=4445) -there were no significant sex differences in the effect magnitudes, however. The effects were furthermore robust to removing outlying values of parental SES.

Social Adversity Reduces Polygenic Score Expressivity for General Cognitive Ability, but Not Height

“…A key prediction therefore is that even if hypothetical PGSs could be constructed using only causal variants, such PGSs would still not be fully portable between populations (or even among subgroups of individuals of the same population exposed to different environments), as heritability-attenuating gene-by-environment interactions associated with racial discrimination, poverty and other forms of social adversity would cause them to fail to predict equal amounts of phenotypic variance when examined in the context of differentially socially advantaged groups. Some evidence consistent with this observation comes from two studies (Woodley of Menie et al, 2018;Woodley of Menie et al, 2021) that were able to recover Scarr-Rowe effects using childhood SES as a moderator of the expressivity of educational attainment PGSs on measures of GCA in two large genetically informed US samples (sourced from the Wisconsin Longitudinal and Health and Retirement Studies, respectively). In both cases, the effects were relatively small by the standards of psychological science (specifically ≤.l0, the average meta-analytic effect magnitude in psychology being around .20; Gignac & Szodorai, 2016), but they were statistically significant and robust to the use of different sets of controls and measurement models.…”

“…Possibly the most compelling evidence of this has been found in genetically informed studies employing cohorts that range from mid to late age in the Wisconsin Longitudinal and Health and Retirement Studies, where in both cases the Scarr-Rowe effect was detected specifically using various measures of participant childhood SES (Woodley of Menie et al, 2018, Woodley of Menie et al, 2021. On this basis, it might be reasonable to expect that the discrimination × PGS EDU interaction might not 'fade out' completely either, unless its apparently greater affinity for GCA makes it more sensitive to fading out than the Scarr-Rowe effectcontinued longitudinal data collection on the ABCD cohort could help to resolve this uncertainty in the future.…”

“…It should be noted that the effect sizes associated with the interaction terms are relatively small in magnitude (≤.10; Gignac & Szodorai, 2016), although this is anticipated given that gene-byenvironment interactions are expected to be much smaller than main effects (McGue & Carey, 2017). Moreover, these estimates are very similar in terms of magnitude to those found in PGS-based studies of the Scarr-Rowe effect in other US samples, based on the use of both two-way interaction and CPEM modeling approaches (Woodley of Menie et al, 2018;Woodley of Menie et al, 2021), suggesting cross-study replicative consistency.…”

“…The major disadvantage to using this method is that the resultant effect size will be confounded with unmodeled effects, thus CPEM effect sizes tend to be larger than those generated using two-way interaction models. This approach has been used (along with a two-way interaction model) to test for the presence of the Scarr-Rowe effect in an analysis employing data from the Wisconsin Longitudinal Study (Woodley of Menie et al, 2018). In this study, a CPE was generated via the dot-product of the participant's standardized PGS EDU and their IQ scores.…”