Using DNA From Mothers and Children to Study Parental Investment in Children’s Educational Attainment. Jasmin Wertz et al. Child Development, October 27 2019.
https://doi.org/10.1111/cdev.13329Abstract: This study tested implications of new genetic discoveries for understanding the association between parental investment and children’s educational attainment. A novel design matched genetic data from 860 British mothers and their children with home‐visit measures of parenting: the E‐Risk Study. Three findings emerged. First, both mothers’ and children’s education‐associated genetics, summarized in a genome‐wide polygenic score, were associated with parenting—a gene–environment correlation. Second, accounting for genetic influences slightly reduced associations between parenting and children’s attainment—indicating some genetic confounding. Third, mothers’ genetics were associated with children’s attainment over and above children's own genetics, via cognitively stimulating parenting—an environmentally mediated effect. Findings imply that, when interpreting parents’ effects on children, environmentalists must consider genetic transmission, but geneticists must also consider environmental transmission.
Discussion
The investments parents make to raise their offspring are thought to be a major contributor to children’s educational success, making parental investment a cornerstone of psychological, sociological, and economic models that seek to explain how educational inequalities are created and perpetuated (Cheng, Johnson, & Goodman, 2016; Feinstein, Duckworth, & Sabates, 2004; Kalil, 2015). However, findings from behavioral‐genetic studies have challenged causal interpretations of parental influence by showing genetic influences on parenting; a gene–environment correlation. Here we tested implications of gene–environment correlations for parental investment in children’s educational attainment using a novel design—in a prospective‐longitudinal study, we collected genotype data from both mothers and children and matched these genetic data with home‐visit measures of parenting behavior. We report three main findings.
First, we found evidence for gene–environment correlations. Both mothers’ and children’s education‐associated genetics, summarized in genome‐wide polygenic scores, were associated with the kind of parenting that is known to be linked with children’s later educational success. By collecting genetic data from both mothers and their offspring, we were able to show that different forms of gene–environment correlations operate in the same family, at the same time. Both active and evocative gene–environment correlations were implicated in the cognitive stimulation and the warm, sensitive parenting that children experienced and in the kinds of households (chaotic; safe and tidy) in which children grew up. Second, we found evidence for slight genetic confounding. The estimated effects of mothers’ parenting on children’s educational attainment were significantly reduced after accounting for education‐associated genetics, consistent with a view of genes as confounding part of the link between parenting and child attainment. However, the magnitude of confounding as measured using the polygenic score was small. Third, we found evidence for genetic nurture. Parenting behavior—particularly mothers’ cognitive stimulation of their children—explained why mothers’ genetics were associated with their children’s educational attainment (independently of children's own genetics). This finding extends recent reports of associations between parental genetics and children’s educational attainment (Bates et al., 2018; Belsky et al., 2018; Kong et al., 2018; Liu, 2018) by showing, for the first time, that parents’ education‐associated genetics shape the features of the family environment that predict the next generation’s educational success.
Our findings need to be interpreted in light of several limitations. First, our approach to estimating genetic nurture relies on the assumption that mothers’ and children’s polygenic scores are measured with identical error. To the extent that this assumption is violated, our estimates of genetic nurture could be upwardly or downwardly biased, depending on whether error is greater in mothers’ versus children’s polygenic scores. However, the assumption is probably defensible, because mothers’ and children’s polygenic scores are identical measurements, that is, sums of the same genotypes transformed using the same weights. Second, although the education polygenic score that we used is based on the largest‐ever social science GWAS, a limitation of this GWAS is that it still reflects only a portion of all genetic influences on educational attainment (approximately one third; Lee et al., 2018). To the extent that the polygenic score is an underestimate of the total genetic influence on educational attainment, our estimates of gene–environment correlations, genetic confounding, and possibly genetic nurture are likely to be underestimates of the true effects. At this point, our findings provide “proof‐of‐principle” of these processes, and the implications they raise can continue to be tested as refined polygenic scores become available. Third, we tested genetic confounding and genetic nurture only for children’s educational attainment, not for other child outcomes. We focused on educational attainment because it is a central determinant of future health, wealth, and well‐being (Cutler & Lleras‐Muney, 2010; Hout, 2012; Oreopoulos & Salvanes, 2011), and because the polygenic score for educational attainment is based on the largest GWAS of a social‐behavior phenotype (Lee et al., 2018). As increasingly larger GWAS are conducted for more developmental outcomes, the same design we present here can be used to test genetic confounding and genetic nurture for other outcomes. Fourth, our study members are still young and most of them have not yet completed their final educational degree. Our measure of educational attainment is therefore only a proxy measure. However, UK students’ qualifications obtained by age 18 are good indicators of their educational pathways beyond age 18 (UK Department for Education, 2018). Furthermore, polygenic‐score associations with educational attainment are very similar between our study and studies of adults (Belsky et al., 2018; Lee et al., 2018), and findings of genetic nurture have been observed in studies of adults who have completed their education (Bates et al., 2018; Belsky et al., 2018; Kong et al., 2018; Lee et al., 2018). Fifth, our labeling of correlations between parents’ genetics and parenting as “active gene–environment correlation” may elicit skepticism among some readers, because correlations between genetics and parenting are typically examined from the perspective of the child and are then referred to as “passive” gene–environment correlations. We would argue that a correlation between parents’ genes and parenting qualifies as active gene–environment correlation, which has been defined as a person contributing to their own environment, and actively seeking an environment related to their genetic propensities (Plomin et al., 1977). Specifically, the aspects of parenting we examine—the cognitively stimulating activities that parents and children engage in together; the warmth and sensitivity of the parent–child relationship; the chaos in the household; and the safety and tidiness of the home—all represent environmental exposures for the parent and the child, that are actively created and shaped by parents to match their genetic dispositions. This interpretation of active gene–environment correlation also follows the concept of niche construction in animal ecology, whereby organisms actively modify their own and each others’ environments; for example by building nests for their offspring (Odling‐Smee, Laland, & Feldman, 2003). Sixth, there is a wide variety of measures available to study parenting and our findings may not generalize to all of these other measures. However, we recently reported very similar associations to the ones observed in our study in an independent sample, using measures of parenting that were derived using what some researchers view as the “gold standard” of parenting assessment—observer ratings of videotaped parent–child interactions (Wertz et al., 2019). The replication across two independent cohorts and different measurements of parenting bolsters the substance of our findings. Seventh, we did not have genetic data from fathers, which means that we were unable to control for fathers’ education polygenic scores when estimating associations between mothers’ and children’s education polygenic scores and parenting. To the extent that fathers’ genes are correlated with parenting, the associations we observed in our study may partly reflect effects of fathers’ genetics, because biological fathers’ and children’s genes are correlated (due to genetic inheritance) and because mothers’ and fathers’ genetics may be correlated (due to assortative mating, i.e., the tendency to select partners with characteristics similar to one’s own). Eigth, even though our research is genetically informative, it is still observational, and hence cannot establish causal relationships between genetics, parenting, and children’s educational attainment. What it can do is (a) point to pathways through which genetic influences may contribute to intergenerational transmission; (b) elucidate processes of gene–environment interplay in parenting and child development; (c) shed light on possible developmental and social mechanisms that link parent and child education‐associated genetics with future attainment; and (d) provide an example of how to integrate new genomic discoveries into developmental psychology to study questions relevant to child development. Against this background, we conclude by discussing the implications of our findings about gene–environment correlations, genetic confounding, and genetic nurture for a more thorough understanding of the developmental processes that shape children’s attainment.
Our findings of gene–environment correlation replicate and extend our prior work on genetic associations with parenting (Wertz et al., 2019). We replicated findings from a previous analysis in a New Zealand cohort, in which we showed that parents’ education polygenic scores were associated with the warm, sensitive, stimulating parenting they provided to their children (Wertz et al., 2019). Here we report the same pattern of results in an independent cohort of British mothers, indicating that genetic correlations with parenting are robust against differences in context and measurements of parenting. We extend this prior work by incorporating children’s polygenic scores in our analyses, finding that children’s genetics are associated with the parenting they receive. Together with other recent studies (Dobewall et al., 2018; Krapohl et al., 2017; Selzam et al., 2018), these findings provide molecular‐genetic evidence for a bidirectional model of parent–child relations, in which parenting is partly a response to children’s characteristics (Bell, 1968; Crouter & Booth, 2003; Pardini, 2008; Sameroff, 2010).
Findings of gene–environment correlations with parenting imply that the family environments children experience while growing up are partly a function of their own and their parents’ genetics. For example, we found that children of parents who carried a higher number of education‐associated variants were exposed to greater cognitive stimulation in the home compared to children of parents who carried fewer of these variants. Because biological parents and children share genes, family environments shaped by parents’ genes will tend to match and reinforce children’s genetic dispositions (Knafo & Jaffee, 2013; Scarr & McCartney, 1983; Tucker‐Drob & Harden, 2012). Such a match can positively influence children’s development; for example, when a child with a high education polygenic score is born into a family that provides cognitive stimulation. However, the same match also implies that a children with lower education polygenic scores will tend not to experience exactly the kind of stimulating and supportive parenting that could make a difference for their attainment. Thus, for better and for worse, correlations between genes and environments can reduce the availability of experiences that alter individuals’ developmental trajectories. This also applies to the reproduction of educational success across generations. To the extent that educational outcomes are influenced by genetics, genes will tend to be a force for intergenerational stability in educational attainment, both via direct genetic transmission and via indirect effects of genes on caregiving environments that shape future generations’ behaviors. This tendency means that is it important to improve children’s access to interventions that may be able to break reinforcing links between genes and environments, such as high‐quality early skill‐building programs (Heckman, 2006).
Given how much attention critics of parenting effects devote to the possibility of genetic confounding (Harris, 1998; Rowe, 1993; Sherlock & Zietsch, 2018), it may seem surprising that our estimates of genetic confounding were so small. There are two possible explanations for this finding: either genetics do little to confound associations between parenting and children’s educational attainment, or we have underestimated the true magnitude of genetic confounding. The observation that polygenic‐score associations with educational attainment are substantially lower than heritability estimates of educational attainment (Branigan, McCallum, & Freese, 2013) suggests that our findings underestimate genetic confounding. Currently, even the best and biggest efforts to capture the genetic variants associated with educational attainment are still missing a substantial part of its heritability (Manolio et al., 2009; National Human Genome Research Institute, 2018). Until more of this “missing heritability” can be accounted for at the molecular‐genetic level, the safest way to rule out genetic confounding is to continue to use family‐based designs, such as discordant‐twin designs (McGue, Osler, & Christensen, 2010; Vitaro, Brendgen, & Arseneault, 2009), parent–child adoption designs (Leve et al., 2013) or children‐of‐twin designs (D’Onofrio et al., 2003), that can estimate associations between parenting and children’s educational attainment free from genetic influences shared between parents and children (Turkheimer & Harden, 2014).
Debates about parental influences on children’s development tend to contrast the effects of parents’ genes—assumed to influence children via genetic transmission—with the effects of parenting—assumed to influence children via environmental ways. Our finding of genetic nurture draws a more nuanced picture, by showing that mothers’ genetics were associated with children’s attainment over and above genetic transmission, via parenting. This finding has three implications. First, over and above a persons’ own genetics, their development will be shaped by the genetics of significant others. We demonstrate this here for effects of parents’ genetics on children’s outcomes, but this observation likely extends beyond parents to everyone who creates environments inhabited by people: family members; individuals residing outside the family context, such as peers and partners (Conley et al., 2016; Domingue et al., 2018); even people to whom a child may be exposed to only indirectly, such as the grandparents who raised a child’s parents (Hällsten & Pfeffer, 2017; Kong et al., 2018; Liu, 2018). The existence of a “social genome” broadens the scope of the study of genetics, from an individual’s genes and their effects on an individual’s phenotype, to the genomes of the individuals making up an individual’s social context (Domingue & Belsky, 2017). Second, much has been written about the need to integrate genetics into parenting research and socialization theory, but there is also a need to integrate environments into how we think about and collect genetic data. Correlations between genes and environments are a challenge not only for socialization research, but also for genetics research: Although DNA sequence cannot be modified by the environment, our findings show that environments still pose a threat to causal inference, because associations between a person’s DNA and developmental outcomes may partly reflect effects of environments created through genes of other individuals (Bates et al., 2018; Kong et al., 2018). As much as genetic confounding needs to be considered when estimating environmental effects, “environmental confounding” needs to be taken into account when estimating genetic effects (Krapohl et al., 2017; Young et al., 2018). Third, environments are part of the pathway from genotype to phenotype (Kandler & Zapko‐Willmes, 2017; Scarr & McCartney, 1983). Specifically, we found that genetic influences on children’s educational attainment partly manifested through parenting; an environmentally mediated genetic effect. The finding shows that new GWAS discoveries are not inimical to socialization theories, because these genetics partly work through factors that socialization researchers have studied for decades, such as the home environment. Combining genetic data with measures of individuals’ social environments is key to tracing how genetics affect life outcomes. By joining forces in this way, genetics and socialization researchers will be able to strengthen causal estimates and obtain a more complete understanding of the processes shaping children’s attainments.
Posts
