Conventional twin studies overestimate the environmental differences between families relevant to educational attainment

Abstract

Heritability estimates for educational attainment (EA) average 41-43% across international studies using the Classical Twin Design (CTD) while estimates of shared environmental influence are among the highest for any behavioural trait, averaging 31-36%. However, high parental correlations for EA in this literature and high correlations for dizygotic twins relative to studies of non-twin siblings suggest these CTD-based estimates may be biased by unmodelled assortative mating and/or twin-specific shared environments. We investigated this question in the German TwinLife sample by comparing results from a CTD model fit to twin-only data with results from Nuclear Twin and Family Design (NTFD) models fit to twin, sibling, and parent data from the same 982 families. Models assuming phenotypic assortment and social homogamy were compared. Our CTD model estimated heritability at 34% and shared environmental influence at 43%. By contrast, our best-fitting NTFD model was a phenotypic assortment model which estimated heritability at 51% and sibling shared environments at just 10% — attributing 16% to twin-specific shared environments which don’t generalise to non-twin siblings or singletons. These results suggest the far-reaching conclusions sometimes drawn from high CTD-based estimates of shared environmental influence for EA may be misplaced.

Social origins and socioeconomic outcomes: A combined twin and adoption study

“…Coefficients for adoptees were slightly higher in these analyses. The presence of own-birth siblings in the family could contribute to between-sibling interaction effects, leading to similarity of adoptees and siblings and in turn more similarity with the parents as has been alluded to in the literature (Nielsen 2006;Wolfram and Morris 2023). The coefficients for Norwegian-born siblings were nevertheless larger than those for adoptees.…”

Assessing the impact of 20th century internal migrations on the genetic structure of Estonia

“…In fact, the level of education has been shown to directly influence migration behaviour in Europe and the US [20][21][22] . EA is a heritable trait with heritability estimates ranging from 4% to more than 50%, depending on the definition and study design [23][24][25][26] . Thus, it is natural to expect that recent migrations can be associated with EA-associated genetic variants and so affect the geographical pattern of allele distribution in a non-random fashion.…”

Challenging the utility of polygenic scores for social science: Environmental confounding, downward causation, and unknown biology

“…Although Burt is correct that using current PGSs to control for genetic influences is partial at best, a well-established literature going back to the 1970s has used genetically sensitive study designs to investigate social science outcomes. These not only include conventional twin studies (which consistently show genetic differences account for a substantial share of the observed individual differences in social science outcomes, e.g., Frisell, Pawitan, Långström, & Lichtenstein, 2012;Hyytinen, Ilmakunnas, Johansson, & Toivanen, 2019;Silventoinen et al, 2020) but also a panoply of other genetically sensitive designs, such as adoption designs, extended twin designs, sibling difference designs, and more (Baier, Eilertsen, Ystrom, Zambrana, & Lyngstad, 2022;Björklund & Salvanes, 2011;Holmlund, Lindahl, & Plug, 2011;Sariaslan et al, 2021;Wolfram & Morris, 2022). These various designs show substantially attenuated statistical associations between predictor and outcome after controlling for genetic confounds and sometimes remove the original association altogether.…”