Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Feb 1.
Published in final edited form as: Dev Psychol. 2021 Feb;57(2):191–199. doi: 10.1037/dev0001131

An Examination of Genetic and Environmental Factors Related to Negative Personality Traits, Educational Attainment and Economic Success

Michael C Stallings 1,2, Tricia Neppl 3
PMCID: PMC8269749  NIHMSID: NIHMS1716717  PMID: 33539127

Abstract

Personality variables are associated with educational attainment and socio-economic outcomes. In this study we incorporated a polygenic score derived from the largest GWAS of educational attainment to date (Lee et al., 2018) into the Interactionist Model of R. Conger and colleagues (2020) that describes the influence of socioeconomic factors on individual development. The inclusion of a polygenic score predictive of educational attainment (PS-Edu) into this model, and the use of the multi-generation, longitudinal Family Transitions Project (FTP) provides a unique opportunity to investigate genetic and environmental influences on the development of negative personality traits, and educational and economic outcomes. The FTP is a three-generation sample. This study utilized data from the first generation (G1; mean age 40 at initiation of the FTP) and second generation (G2; assessed at mean ages 18 and 30). Participants are approximately 50% female, 99% of European ancestry, primarily from lower to middle class SES. PS-Edu was significantly correlated with educational attainment in both generations of the FTP, accounting for 4.1% to 6.7% of the variance. Findings confirm that PS-Edu is a complex genetic index that is correlated with all of the socioeconomic constructs in the model. Results suggest potential gene-environment correlation or common genetic influences underlie associations among parenting investments, negative personality traits and educational attainment. Genetic variance captured by PS-Edu was mediated substantially through G1 parental investments. Although study limitations warrant cautious interpretation, we demonstrate the promise of including polygenic scores in developmental models to better understand genetic and environmental influences on human development.

Keywords: Socioeconomic Development, Interactionist Model, Personality and Educational Attainment, Gene-environment interplay

This report provides the final of four empirical papers in this special section on the development of personality and its life course consequences. This special section grew out of a 2014 special section in Developmental Psychology that focused specifically on conscientiousness and its correlates. As noted by R. Conger, Martin, and Masarik (2020), the earlier special section was rich in theoretical reasoning about possible bases for the development of conscientiousness and in the review of earlier studies demonstrating associations between this personality trait and a range of personal competencies. A limitation of the 2014 report, however, was the absence of empirical data that would allow testing of the development and effects of conscientiousness over time. The availability of data from the Family Transitions Project (FTP), which provides information for all four empirical studies in the present special section, corrects this limitation because of its prospective, longitudinal design. In addition, the earlier special section only focused on positive personality traits. We have extended that focus by considering negative personality traits as well. In the present investigation, we examine social, genetic, and economic predictors of negative personality traits during adolescence and its influences on later educational and economic success.

The Interactionist Model (IM) of Personality Development and the Current Study

The series of investigations in this special section employ a theoretical model (see Figure 1) which proposes that personality is influenced both by genetic factors transmitted from parents to offspring and also by environmental contexts ranging from the family of origin to educational and occupational experiences as an adult. The Interactionist Model (IM) hypothesizes that socioeconomic status (SES), family processes (e.g., parental investment in their offspring), and personality are interrelated over time and across generations. Individual development occurs as part of ongoing transactions between individual characteristics and environmental conditions that are mutually influential over time (for details see R. Conger et al., 2020).

Figure 1.

Figure 1.

Open in a new tab

Interactionist Model Including the G2 Target Polygenic Score (PS-Edu)

The consideration of genetic factors in the IM is a unique contribution of two of the papers in this special section. In particular, the current study extends the findings of Donnellan, Martin and Senia (2020) who included a polygenic score (PS) in their analysis of the developmental IM. Specifically, they used a genetic index that had been identified in earlier studies as a predictor of educational attainment (Lee et al., 2018). Although polygenic scores derived from the Lee et al. summary data are ‘predictive’ of educational attainment (and we use the label PS-Edu to note that), they are complex scores that correlate with many factors associated with educational and economic outcomes. Inclusion of PS-Edu into the IM provides a unique opportunity to investigate the interplay between genes, environment, personality, and educational and economic outcomes in a multi-generation, longitudinal model. Here we extend the findings of Donnellan et al. (2020), who focused on conscientiousness, by examining the role of negative personality traits in the developmental processes proposed by the IM. Although Donnellan and colleagues found evidence that conscientiousness linked genetic differences to educational attainment, it is unknown whether negative personality traits operate similarly. We investigate that possibility in the current study.

Evidence of Genetic Influences on Educational Attainment

Findings from genetic studies of academic outcomes provide an important foundation for the present study. Educational attainment is influenced by multiple genetic and environmental factors, as well as the interplay among these influences (i.e., gene-environment correlation and interaction). Gene-environment interaction occurs when environmental factors moderate genetic influences, or vise-versa, when genetic predisposition moderates the effects of environmental factors. Gene-environment correlation (rGE) refers to the fact that the environment one is exposed to is often influenced by genetic factors. For example, passive rGE occurs when the environment parents provide for their children is influence by parental genetic predispositions. Active rGE occurs when an individual’s genetic propensities create or shapes their environment, and evocative or reactive rGE occurs when others modify environment influences in reaction to the genetic predisposition of an individual (e.g., providing lessons for a musically-inclined child).

Although the specific genetic mechanisms remain unknown, numerous studies have demonstrated substantial genetic influences on educational achievement, attainment, and related factors (Knopik et al., 2017). For example, twin studies of reading ability (Harlaar, 2006; Olson, 2007) have reported heritabilities of approximately 50% to 60% for different aspects of reading, and similar findings have been reported for academic achievement in other domains (Ayorech et al., 2017; Baretls et al., 2002; Kovas et al., 2007; Wainwright et al., 2005). A recent meta-analysis of data from 61 twin studies of educational achievement in primary school (including 11 cohorts from 7 countries), estimated the heritability of general educational achievement at 66% (de Zeeuw et al., 2015). Using alternative methods, Schwabe et al. (2017) reported a pedigree-based heritability estimate of 85% for educational achievement.

Although it is clear that genetic factors play a role in educational attainment, efforts to identify specific genetic variants may provide insights into the biological and environmental pathways underlying educational attainment. Unlike twin studies that quantify the magnitude of latent genetic influences, genome-wide association studies (GWAS) directly interrogate measured genotypes, usually single nucleotide polymorphisms (SNPs) distributed throughout the genome. Early GWAS made use of approximately a million SNPs typically available on commercial genotyping chips. However, modern imputation methods (Fuchsberger, Abecasis & Hinds, 2015) have now enabled researchers to test trait associations for many millions of SNPs across the genome.

The first GWAS of educational attainment was published in 2013 (Rietveld, et. al). This initial meta-GWAS made use of a discovery sample of 101,069 individuals and a replication sample of 25,490. Only three significant, replicated SNPs were identified. More importantly, however, a linear polygenic score (PS) utilizing information from all of the measured SNPs accounted for approximately 2% of the variance in educational attainment in a replication sample. Since this first report, GWAS of educational attainment using considerably larger sample sizes have been conducted (e.g., Okbay et al., 2016b), with the most recent, and largest GWAS conducted by Lee et al. (2018). Lee et al. conducted a meta-GWAS of educational attainment utilizing data from over 1.1 million individuals and identified 1,271 independent genome-wide significant SNPs, implicating genes involved with neural development and neural communication. A PS derived from the Lee et al. summary data accounted for approximately 11% and 13% of the variance in educational attainment in two independent replication samples, indicating that PSs derived from large discovery samples are able to explain a substantial proportion of the variance in educational attainment.

Several investigators (e.g., Belsky et al., 2016; de Zeeuw et al., 2014; Domingue et al., 2015; Mottus et al., 2017; Selzam et al., 2017) have used summary data from these large GWAS and others (e.g., Hill et al., 2016) to demonstrate that educational attainment is a complex phenotype. These studies have shown that PSs that predict educational attainment are also associated with SES, early speech and reading skills, general cognitive abilities, personality variables, and adult economic outcomes—and suggest evidence for gene-environment correlation. Krapohl et al. (2014) used 13,306 twins from the United Kingdom assessed at age 16 to demonstrate that academic achievement is influenced by many factors. Several psychological domains including intelligence, self-efficacy, personality, well-being, and behavior problems correlated substantially with academic achievement and these correlations were largely mediated genetically.

Although many factors influence socioeconomic success, there has been considerable interest in the extent to which personality plays a role in educational and economic outcomes. Personality traits have been associated with educational and/or economic success in numerous studies (e.g., Corker, Donnellan, & Oswald, 2012; Noftle & Roberts et al., 2007; Robins, 2007; Poropat, 2009; Shiner & Masten, 2012; Sobowale et al., 2018). Personality traits such as conscientiousness and persistence are predictive of academic and economic success, but negative personality traits such as neuroticism, and emotional instability also play a role (Damian et al., 2015; van Eijck & de Graff, 2004). Personality traits are also substantially heritable (Loehlin, 1992; Knopik et al., 2017; Sanchez-roige et al., 2018). Recent large-scale meta-analyses of twin data have estimated the heritability of multiple personality measures at about 40% (Polderman et al., 2015; Vukasovic and Bratko, 2015), and meta-GWAS have begun to identify specific genetic variants associated with personality attributes (de Moor et al., 2012; Luciano et al., 2018; Nagel et al., 2018).

These findings provide an important base upon which to build the present investigation. Unique to this study, however, we go beyond the simple correlational analyses typical of these earlier investigations. We examine the role of genetic variation in a complex theoretical model that can identify both direct and indirect genetic influences on developmental pathways predictive of educational and economic success. We believe this approach provides a better account of how genetic variation interrelates with social and behavioral factors to affect individual development over time.

Method

Study Participants

Data for the present study come from the Family Transitions Project (FTP), a longitudinal study of youth and their three-generation families from rural Iowa (see K. Conger et al., 2020 for additional details). FTP families were first recruited in 1989 through public and private schools in eight rural, north central, Iowa counties. The racial/ethnic minority population in this region was about 1% when the study was initiated; thus, essentially all of the FTP participants are of European ancestry and primarily of lower-middle and middleclass SES. Eligible FTP families included a target adolescent (G2) in seventh grade at the initiation of the study, both of his/her biological (G1) parents (but single-parent G1 mother-headed households were added later in 1991), and a sibling within 4 years of age of the G2 target. Although a third generation of participants was added starting in 1997, they are not included in the present analyses.

The current study uses FTP participants from the first two generations: the G1 parents and the G2 targets. At the initiation of the FTP (1989), the mean age of the G1 fathers and mothers was 40 and 38, respectively. The G2 target adolescents ranged in age from 12 to 14 years (M = 13.17) and 52% were female. The current study uses FTP data obtained between 1991 and 2010, when the G2 targets were approximately 15 through 35 years of age. Because we used informant reports to measure negative personality traits (see below), the current analyses are based on the 346 G2 targets for whom informant data was available in adolescence and adulthood. Data collection for the most recent assessment waves of the FTP was approved by the University of California, Davis IRB (#21946-11 “Family Transitions Project”). The IRB protocol was officially closed in 2013 but the closure allows for the analysis of data. This paper involves secondary analysis of existing data only.

Measures

Brief descriptions of the model constructs are provided below. For measurement details see Martin and Donnellan (2020).

G1 and G2 educational attainment.

Educational attainment was coded on a 6-point scale (1 = less than high school; 2 = high school degree/GED; 3 = degree from junior, vocational, or community college, or attended college but did not earn a 4-year degree; 4 = degree from 4-year college; 5 = Master’s degree; and 6 = PhD or professional degree). G2 educational attainment was assessed via self-reports of schooling completed in 2003 (mean age 27.7). G1 education (assessed via G1 report) was assessed in 1991. The average of G1 father and G1 mother educational attainment was used as an index of G1 education in the model.

G1 and G2 income.

G1 and G2 income were measured as an income-to-needs ratio by dividing the total family income for the past year by the U.S. Department of Health and Human Services poverty guidelines by family size. As with education, G1 income was also assessed in 1991 and G2 income in 2003 via self-report. Family income included all wages, salaries, and other sources of income.

G1 parenting investments.

Parenting investments were assessed via coded observer reports. Trained observers were used to captured the degree to which the G1 parents provided emotionally supportive and appropriate parenting behavior toward their G2 child. This was assessed as a latent variable with three indicators: warm parenting, low (reversed) hostile parenting, and low (reversed) harsh and inconsistent discipline. G1 parenting investments were assessed in 1992 during family discussion and problem-solving tasks (see Martin et al., 2020 for details). Higher scores represent greater parental investment.

G1 material investments.

A cumulative index of material investments (assessed via interview and self-report) was created by dichotomizing seven variables related to parental investments (extracurricular activities, parental aid during the transition to adulthood, parental aid with talents and skills, and four interviewer reported items assessing the family’s living environment); higher scores represent greater investments.

G2 Negative Personality Traits.

Negative personality traits were assessed in adolescence (1994 assessment; mean age 18.1) and adulthood (2005; mean age 29.6). We defined a negative personality construct (NEG) using the alienation and well-being (reverse coded) scales of the Multidimensional Personality Questionnaire (MPQ; Harkness, Tellegen, & Waller, 1995). Individuals with high scores on the MPQ alienation scale tend to feel they are treated poorly and unfairly by others, feel unlucky, alienated and exploited by others. Individuals with high scores on the low well-being scale are not interested in life, unenthusiastic, disillusioned and unoptimistic. Informant reports were used to assess personality traits. The G1 parents of the G2 targets served as informants in adolescence and the romantic partners of the G2 targets served as informants for the assessment in adulthood. Informant reports by individuals who are well acquainted with the subject have been shown to generate valid information (e.g., Connelly & Ones, 2010).

Genotyping.

Details regarding genotyping can be found in Donnellan et al. (2020). Briefly, because the Illumina HumanOmni1-Quad genotyping platform was discontinued midway through our genotyping effort, approximately half the FTP participants were genotyped using the Illumina HumanOmni1-Quad v1 platform (Omni-Quad) and half were genotyped using the Illumina HumanOmniExpressExome v1 platform (Express-Exome). Quality control (QC) analyses were conducted within platform, and across platforms using a common set of markers. Marker map locations and annotations were based on the Human Genome Reference Consortium GenomeBuild 37.1. SNP markers that did not have a reliable chromosome or base-pair location using Build 37.1 were not used, and markers with minor allele frequencies (MAF) < .01 and call rates less than 95% were also removed. Because markers included in the HapMap Project and 1000 Genomes Project have been extensively validated in independent samples, we further limited our SNPs to those validated in these databases.

The final marker set included 582,624 markers common to both genotyping platforms that met all QC filters. Final genotypes were imputed, separately by platform, from this common set of SNPs using the 1000 Genomes phase 3 (www.ncbi.nlm.nih.gov/bioproject/28889) reference panel via the Michigan Imputation Server (https://imputationserver.sph.umich.edu). Imputation yielded approximately 48 million SNPs. Only SNPs with MAF > 0.01 were used in the current analyses (~9 million SNPs).

Polygenic Scores.

We used summary results from a recent meta-GWAS of educational attainment made available through the Social Science Genetic Association Consortium (SSGAC; https://www.thessgac.org), to generate polygenic scores (Durbridge, 2013) for the FTP participants. The source of the meta-GWAS data was Lee et al. (2018) but due to IRB restrictions the 23andMe cohort was not available through the SSGAC. Thus, summary data (file: GWAS_EA_excl23andMe.txt) were based on a meta-GWAS of 766,345 individuals. All individuals were of European Ancestry (EA), 54% were female, and all were at least 30 years of age at assessment of educational attainment (mean birth year 1955). Polygenic scores (PS-Educ) were generated using PLINK (www.cog-genomics.org/plink/1.9/; Chang et al., 2015). PSs were constructed from 7,187,581 SNPs available in both the FTP sample and the SSGAC summary data (for details see Donnellan et al., 2020). Using various p-value thresholds to select subsets of SNPs (e.g., selecting only SNPs with p-values less than 0.50, 0.25, 0.1, 0.05) did not yield markedly different correlations. Thus, polygenic scores for the FTP participants were scored using all 7,187,581 common SNPs.

Analysis

The parameters of the structural equation model (SEM) depicted in Figure 1 were estimated using Mplus 8.0 (Muthén & Muthén, 1998-2018) using full information maximum likelihood (FIML) estimation. FIML provides less biased estimates than ad hoc procedures for dealing with missing data. Path coefficients shown in Figure 1 and Table 2 are standardized regression coefficients, partialed for all other coefficients in the model. The significance of individual path coefficients was determined via estimated standard errors. Overall model fit was determined by conventional criteria: a comparative fit index (CFI) and Tucker-Lewis index (TLI) of .950 or more and a root mean squared error of approximation (RMSEA) value below .06 (Hu & Bentler, 1999). For evaluating the significance of the 23 path coefficients in the model (see Figure 1) we used a conservative Bonferroni correction (.05/23 = .0022).

Table 2.

Standardized Path Coefficients: Negative Personality Traits and Socioeconomic Success

Model Path (or correlation) Beta SE p
G1 Income on G1 Education .256 .052 <.001
G1 Parenting Investments on G1 Education .456 .067 <.001
G1 Parenting Investments on G1 Income .183 .067 .006
G1 Material Investments on G1 Education .229 .054 <.001
G1 Material Investments on G1 Income .157 .054 .003
G2 Education on G1 Education .208 .063 .001
G2 Education on G1 Parenting .280 .082 .001
G2 Income on G1 Material Investments .137 .055 .012
G2 Income on G2 Education .354 053 <.001
G2 NEG [adolescence] on G1 Parenting −.176 .087 .044
G2 NEG [adolescence] on G1 Material Investments −.279 .069 <.001
G2 Education on G2 NEG [adolescence] −.273 .057 <.001
G2 Income on G2 NEG [adolescence] −.152 .065 .019
G2 NEG [adult] on G2 Education −.120 .072 .093
G2 NEG [adult] on G2 Income −.101 .069 .143
G2 NEG [adult] on G2 NEG [adolescence] .344 .077 <.001
Correlation: G1 Education with G2 PS-Edu .203 .051 <.001
G1 Parenting Investments on PS-Edu .114 .077 .138
G1 Material Investments on PS-Edu .098 .059 .093
G2 Education on PS-Edu .098 .051 .054
G2 Income on PS-Edu .047 .056 .400
G2 NEG [adolescence] on PS-Edu −.127 .064 .046
G2 NEG [adult] on PS-Edu −.006 .069 .931
Open in a new tab

Fit: Chi-Square: 60.479, df = 45; CFI = .982, TLI = .969, RMSEA = .032, SRMR = .038; NEG = index of negative personality traits; G2 PS-Edu = G2 Target polygenic score, computed using summary data from the meta-GWAS of educational attainment conducted by Lee et al. (2018); given the number of path coefficients tested a conservative Bonferroni correction (.05/25 = .002) was used to determine significance.

Results

Fit indices suggest the model (Figure 1) fits the data well (CFI = .982, TLI = .969, RMSEA = .032). Table 1 shows the correlations among the latent constructs in the model. The last line of the table highlights the correlations between the G2 PS-Edu (the genetic index derived from the Lee et al. 2018 summary data) and each of the model constructs. The PS-Edu is substantially correlated with both G2 (r = .258) and G1 educational attainment (r = .203). Unadjusted squared correlations indicate that the PS-Edu accounts for 6.7% of the variance in educational attainment among the FTP G2 targets and 4.1% of the variance among their G1 parents. However, PS-Edu correlates with G1 parental income, parenting investments and material investments as well. The PS-Edu also correlates with G2 NEG in adolescence (r = −.208), but less so with NEG assessed in adulthood (r = −.128).

Table 1.

Correlations Among Model Latent Variables (N = 346)

1 2 3 4 5 6 7 8
1 G1 Income 1.000
2 G1 Education .256 1.000
3 G1 Material Investments .221 .290 1.000
4 G1 Parenting Investments .306 .526 .445 1.000
5 G2 Income .114 .232 .309 .298 1.000
6 G2 Education .177 .430 .303 .500 .473 1.000
7 G2 Adolescent NEG −.122 −.199 −.377 −.328 −.365 −.426 1.000
8 G2 Adult NEG −.075 −.145 −.198 −.204 −.284 −.316 0.433 1.000
9 G2 Polygenic Score (PS-Edu) .052 .203 .153 .216 .191 .258 −.208 −.128
Open in a new tab

Table 1 shows the zero-order correlations among the latent constructs in the model.

The PS-Edu appears to correlate more strongly with education than with income for both G2 targets and their G1 parents. As might be expected, the magnitude of the correlations between PS-Edu and G2 target’s income and education appear greater than with G1 parental income and education. It is also notable that the correlation between PS-Edu appears higher with G1 parenting investments than with G1 material investments. These correlations suggest possible gene-environment correlation (rGE) effects. These effects might be expected to be stronger for specific actions by the parents toward the youth, such as warmth and support (parenting investments), than for material provisions such as income, housing and family location in the community (material investments). In general, substantial correlations between the G2 PS-Edu and most of the model constructs suggest common genetic influences and/or gene-environment correlation (rGE) processes (i.e., individuals with higher scores on the educational attainment PS-Edu tended to have G1 parents with higher education and greater income who tended to provide greater parenting and material investments in their G2 children). This is referred to as passive rGE because the trait-relevant environment that parents provide for their children is often correlated with genetic influences also transmitted to their children.

Compared with G1 income, G1 education appears to correlate more highly with G2 education and income. G1 material and parenting investments correlate similarly with G2 income, but G1 parenting investments correlate more highly with G2 education (r = 0.50) than with G2 income (r = 0.298). For NEG, stronger correlations were observed between model constructs and NEG assessed in adolescence (mean age ~18) than for NEG assessed in adulthood (mean age ~30). As expected, NEG, assessed at both ages, correlates negatively with all other constructs in the model.

Figure 1 shows the model and standardized path coefficients. Before considering the main effects in the theoretical model, all possible interactions between the PS-Edu and predictor variables in the model were estimated. Because none of these GxE interactions were statistically significant, they were dropped from further consideration. Nominally significant paths (p < .05) are depicted as solid lines; non-significant paths are shown as dotted lines. Table 2 shows the standardized parameter estimates, standard errors, and p-values. Given 23 path coefficients were tested in the model a conservative Bonferroni correction (.05/23 = .0022) was used to evaluate significance. This model is an extension of the model described by R. Conger et al. (2020), with the addition of the G2 polygenic score (PS-Edu).

Consistent with R. Conger et al., G1 education significantly (p < .002) predicts G1 parenting and material investments. G1 income is less predictive of these variables and does not meet our Bonferroni threshold for significance. G2 education is significantly predicted by G1 parenting while the influence of material investments does not meet the multiple testing threshold. However, there remains a substantial direct effect of G1 education on G2 education suggesting that other factors not captured by parenting and material investments or the genetic index (PS-Edu) also contribute. As expected, NEG shows negative relationships with G1 parenting and material investments and G2 education and income; with G1 material investments and G2 education demonstrating significant relationships. However, NEG in adulthood is not significantly predicted by G2 education and income. These unexpected results may stem from the fact that negative personality traits are relatively stable from adolescence to adulthood (r = .443, b = .344), potentially mitigating genetic and environmental effects over time. The pattern may also suggest that NEG mediates genetic and environmental influences on educational attainment rather than the reverse—educational success/failure influences personality.

The bottom, shaded portion of Table 2 shows the path coefficients relating model constructs to the G2 genetic index: PS-Edu. The path from PS-Edu to NEG during adolescence (b = −.127, p = .046), suggests a modest genetic influence on NEG even after economic and social conditions in the family are taken into account. However, this finding should be interpreted with caution given multiple testing issues. Other potential trends in the findings warrant some mention. First, PS-Edu was positively related to parenting (b = .114, p = .138) and material investments (b = .098, p = .093), consistent with the correlations in Table 1 suggesting possible passive or evocative rGE effects. Second, PS-Edu was positively related to G2 education after all other earlier predictors in the model were taken into account (b = .098, p = .054). Even though this finding does not meet our threshold for statistical significance given multiple testing, it is consistent with the zero-order correlation suggesting a genetic influence on G2 education in the overall model.

Note the reduction in the magnitude of the direct relationships between PS-Edu and other constructs in the model when the path coefficients are compared to the correlations in Table 1. This suggests a number of indirect genetic effects. For example, from Table 1 the correlation between the G2 PS-Edu and G2 educational attainment is fairly substantial (r = 0.258). However, after including the G2 PS-Edu in the model the direct effect of the PS-Edu on G2 education reduces to 0.098. This indicates that a substantial proportion of the relationship between the G2 PS-Edu and G2 education is mediated through upstream constructs in the model. There is also no significant direct effect of the G2 PS-Edu on G2 income. Again, there is a substantial zero order correlation, but that relationship is mediated through other constructs in the model. Mediation analyses (conducted in Mplus; Muthén & Muthén, 1998-2018) indicated that the total effect of the G2 PS-Edu on G2 education is significant (p = .001), but the total indirect effect is not significant (p = .012) given the number of tests conducted. Further, none of the specific indirect paths linking the PS-Edu to G2 education were individually significant (all p-values > .054). Similarly, correlations between the G2 PS-Edu and NEG in adolescence (r = −.208) and adulthood (r = −.128) were modest. There is a modest direct effect of the G2 PS-Edu on NEG in adolescence but a negligible direct effect on NEG in adulthood. Total indirect effects of PS-Edu on NEG were not significant in adolescence (p = .054). Total indirect effects on NEG in adulthood trended towards significance (p = .006), but none of the individual direct paths linking PS-Edu and NEG in adulthood were significant (all p-values > .071). These findings have important implications for future research and theory regarding genetic effects within the context of larger developmental models.

Discussion

This report provides the final empirical paper in this special section on the development of personality and its life course consequences. The study makes use of the prospective, multi-generation data of the Family Transitions Project (FTP) to examine social, genetic, and economic predictors of negative personality traits (NEG) during adolescence, as well as influences on NEG, educational and economic success in adulthood. The Interactionist Model (IM) of R. Conger et al. (2020) proposes that personality is influenced both by genetic factors and family environmental factors as well as educational and occupational experiences throughout development. In this study we extended the IM by including a genetic index predictive of educational attainment (PS-Edu). The PS-Edu was derived from summary data obtained from Lee et al. (2018)—the largest meta-GWAS of educational attainment to date. The inclusion of a measured genetic index into the IM provides a unique opportunity to investigate direct and indirect genetic influences on developmental pathways predictive of educational and economic success. Here we extend the findings of Donnellan et al. (2020), who focused on conscientiousness, by examining the role of negative personality traits (NEG; characterized by disillusionment, lack of persistence, alienation and non-conventionality) in the developmental processes proposed by the IM.

This study yielded a number of notable findings. First, the G2 target genetic index (PS-Edu) was significantly correlated with both G2 target educational attainment and the educational attainment of their G1 parents. Uncorrected for other variables in the model, PS-Edu accounted for 6.7% of the variance in G2 educational attainment. This represents a substantial improvement in prediction over polygenic scores derived from earlier smaller GWAS (Rietveld et al., 2013; Okbay et al., 2016). Although much of this correlation is explained by indirect effects in the model, the effect of polygenic indices have a unique predictive status (Selzam et al., 2017). Although gene expression may change over time, the influence of DNA variation on G2 education (assessed via PS-Edu) cannot be attributed to reverse causation (Selzam et al., 2017), while the direction of influence of other factors in the model is more difficult to determine, and may even change over time in a transactional way (e.g., the magnitude and/or direction of influence may change over development).

Second, our findings confirm that the PS-Edu is a complex genetic index that also correlates positively with G1 parental income, as well as G1 material and parenting investments. Several researchers have reported similar correlations between polygenic indices predictive of educational attainment and measures related to socioeconomic status (e.g., Belsky et al., 2016; de Zeeuw et al., 2014; Domingue et al., 2015; Mottus et al., 2017; Selzam et al., 2017). These correlations may suggest possible gene-environment correlation (rGE) effects. G2 targets with higher scores on the PS-Edu genetic index tend to have G1 parents with higher education and greater income who also tend to make greater parenting and material investments in their G2 children. These findings are consistent with active/evocative rGE, where the genetic predispositions of children evoke or create environments (including parenting behavior) consistent with their genotype. However, passive rGE, whereby parents provide environments that are correlated with their own genetic predispositions could also explain the results. Further, given that parents and children share half of their additive genetic effects, these correlations could also be explained by common genetic factors transmitted from parents to children that influence both parent and child phenotypes (vertical pleiotropy).

Third, the G2 genetic index (PS-Edu) correlates negatively with G2 NEG in adolescence and adulthood. Correlations between genetic indices of educational attainment and negative personality traits have been reported by others (e.g., Okbay et al., 2016; Mottus et al., 2017; Smith-Wooley et al., 2019). Although these findings may also be explained by common genetic influences underlying personality and other traits associated with educational success, an alternative explanation is that personality traits may mediate, in part, the genetic variance in educational attainment (Rimfeld et al., 2016). Figure 1 provides some evidence to support this mediating role. The correlation between the G2 genetic index (PS-Edu) and G2 educational attainment is fairly substantial (r = 0.258). However, after including PS-Edu in the model, the direct effect of the genetic index on G2 education is reduced to 0.098. This suggests that a substantial proportion of the relationship between PS-Edu and G2 education is mediated through upstream constructs in the model—including NEG. Further, there is a significant direct effect of PS-Edu on G2 NEG in adolescence, which in turn significantly predicts G2 education, income and NEG in adulthood. On the other hand, G2 education and income assessed in 2003 do not significantly predict NEG assessed later in 2005. Though this provides evidence that some of the genetic variance assessed by the PS-Edu may be mediated through NEG, as well as other constructs in the IM. However, the presence of the direct effect of PS-Edu on G2 education (a trend, p = .054) suggests that educational attainment is explained by genetic factors. Although this small effect is limited by the predictability of current polygenic scores, it suggests that individuals’ socio-economic success, whether they are born into advantaged or disadvantaged families, is due in part to their genetic predisposition (Ayorech et al., 2017; Conley, 2015). Further, in contrast to the findings of Mottus et al., (2017), and more in line with the interpretations of Rim et al., (2016), our longitudinal findings find some support that genetic variance predictive of educational attainment may be mediated, in part, through negative personality traits. Unfortunately, although we had sufficient power to detect significant total and direct effects in our mediation analyses, tests of specific indirect paths lacked sufficient power.

Limitations

A number of limitations should be considered when interpreting these findings. Although the prospective Family Transitions Project (FTP) provides important advantages for examining genetic and environmental influences on emotional and socio-economic development, the number of G2 targets with available data was not large (N = 346). Consequently, although the Lee et al. (2018) meta-GWAS data available to use was derived from 766,345 individuals, the power to detect associations with polygenic scores derived from these summary data would be expected to be limited in analyses of the FTP data (Durbridge, 2013).

Although the unadjusted polygenic score (PS-Edu) accounted for 6.7% of the variance in G2 educational attainment in the FTP sample—this represents only a portion of the heritability estimated from twin studies. However, unlike twin study heritability estimates, a genetic index derived from GWAS summary data is limited to the additive effects of common SNPs. SNP-based heritability estimates have been much lower than twin-based estimates (Krapohl & Plomin, 2015). Nevertheless, despite improved prediction compared to polygenic scores based on smaller GWAS (Rietveld et al., 2013, Okbay et al., 2016b) our genetic index accounted for only a relatively small proportion of the total genetic variance associated with educational attainment. Although many relationships described by the Interactionist Model (IM) are maintained after including PS-Edu in the model, the genetic index is only explaining a small portion of the genetic variance. Unexplained genetic variance likely influences constructs in the IM in ways that may become clearer as better polygenic scores become available.

In the current study we had two assessments of our negative personality index (NEG): first assessed in adolescence (mean age ~18) with a second assessment in adulthood (mean age ~30). Our findings are consistent with the influence of evocative or reactive effects on parenting, but they may also arise from passive gene-environment correlation processes and/or vertical pleiotropy. Limited power of mediational analyses made it difficult to distinguish among these processes. With additional longitudinal assessment, particularly in childhood, we would have been in a better situation for disentangling the sources of the gene-environment correlations observed. Given evidence from twin and adoption studies that environmental influences tend to decrease while genetic influences increase across development for many traits (Knopik et al., 2017), a longitudinal pattern of decreasing gene-environment correlations may be consistent with passive rGE. Active and evocative gene-environment correlation refers to processes whereby individuals construct or evoke environments that are correlated with their genetic predispositions. One might expect these influences to increase gene-environment correlations from childhood into adulthood. Simply put, even with our longitudinal data we have stepped into the developmental stream at a relatively late date.

Conclusion

In the current study we incorporated a polygenic score (PS-Edu), derived from the largest GWAS of educational attainment to date (Lee et al., 2018), into the Interactionist Model (IM) developed by Conger and colleagues (see R. Conger et al., 2020). Inclusion of the genetic index into the IM allowed us to investigate the interplay between genes, environment, personality, and socio-economic outcomes in a multi-generation longitudinal model. Genetic variance assessed by the PS-Edu correlated with all constructs in the model, suggesting that common genetic factors and/or gene-environment correlation processes are important influences of socioeconomic outcomes. Although limitations of the study warrant cautious interpretation of the findings, we demonstrate the promise of including polygenic scores in developmental models. As better polygenic scores are developed, the value of these tools for better understanding genetic and environmental influences on human development will be realized.

References

  1. Ayorech Z, Kraphol E, Plomin R, von Strumm S (2017). Genetic influences on intergenerational educational attainment. Psych Science, 28(9): 1302–1319. doi: 10.1177/0956797617707270. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bartels M, Rietveld MJ, van Baal GG, Boomsma DI (2002). Heritability of educational achievement in 12-year-olds and the overlap with cognitive ability. Twin Res., 5(6): 544–553. doi: 10.1375/136905202762342017. [DOI] [PubMed] [Google Scholar]
  3. Borriello GA, Ramos AM, Natsuaki MN, et al. (2020). The intergenerational transmission of mathematics achievement in middle childhood: A prospective adoption design [published online ahead of print, 2020 Apr 23]. Dev Sci. 2020; e12974. doi: 10.1111/desc.12974. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Chang CC, Chow CC, Tellier LCAM, Vattikuti S, Purcell SM, Lee JJ (2015). Second-generation PLINK: rising to the challenge of larger and richer datasets. GigaScience, 4. doi: 10.1186/s13742-015-0047-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Conger KJ, Neppl TK, Scaramella L (2020). Special section on personality development: Testing environmental and genetic associations across generations. Developmental Psychology. [DOI] [PubMed] [Google Scholar]
  6. Conger R,D, Martin MJ, & Masarik AS (2020). Dynamic associations among socioeconomic status (SES), parenting investments and Conscientiousness across time and generations. Developmental Psychology. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Damian RI, Su R, Shanahan M, Trautwein U, & Roberts BW (2015). Can personality traits and intelligence compensate for background disadvantage? Predicting status attainment in adulthood. J Pers Soc Psych., 109: 473–489. doi: 10.1037/pspp0000024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. de Moor MH, et al. (2012). Meta-analysis of genome-wide association studies for personality. Mol Psychiatry, 17(3): 337–349. doi: 10.1038/mp.2010.128. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. de Zeeuw E,L, van Beijsterveldt CE, Glasner TJ, Bartels M, Ehli EA, Davies GE, Boomsma DI, et al. (2014). Polygenic scores associated with educational attainment in adults predict educational achievement and ADHD symptoms in children. Am J of Med Genet Part B: Neuropsych Genet., 165(6): 510–520. doi: 10.1002/ajmg.b.32254. [DOI] [PubMed] [Google Scholar]
  10. Domingue BW, Belsky DW, Conley D, Harris KM, Boardman JD (2015). Polygenic influence on educational attainment: New evidence from The National Longitudinal Study of Adolescent to Adult Health. AERA Open. Jul-Sep; 1(3): 1–13. doi: 10.1177/2332858415599972. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Donnellan MB, Martin MJ, & Senia JM (2020). Genetic influences on the Interactionist Model of socioeconomic development: Incorporating polygenic scores for educational attainment into developmental research. Developmental Psychology. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Dudbridge F (2013). Power and predictive accuracy of polygenic risk scores. PLoS genetics, 9(3), e1003348. doi: 10.1371/journal.pgen.1003348. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Fuchsberger C, Abecasis GR, Hinds DA (2015). Minimac2: faster genotype imputation. Bioinformatics, 31(5): 782–784. doi: 10.1093/bioinformatics/btu704. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Harlaar N (2006). Individual differences in early reading achievement: Developmental insights from a twin study. Universisty of London. [Google Scholar]
  15. Hill et al. (2016). Molecular Genetic Contributions to Social Deprivation and Household Income in UK Biobank, Current Biology, 26: 1–7. doi: 10.1016/j.cub.2016.09.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Hu L, Bentler PM (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6: 1–55. [Google Scholar]
  17. Knopik VS, Neiderhiser JM, DeFries JC, Plomin R (2017). Behavioral Genetics (7th Ed). Worth Publishers, New York, NY. [Google Scholar]
  18. Kovas Y, Haworth CMA, Dale PS, Plomin R (2007). The genetic and environmental origins of learning abilities and disabilities in the early school years. Monogr of the Soc Res Child Develop., 72: 1–144. doi: 10.1111/j.1540-5834.2007.00439.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Krapohl E, Rimfeld K, Shakeshaft NG, Trzaskowski M, McMillan A, Pingault JB, Asbury K, Harlaar N, Kovas Y, Dale PS, Plomin R (2014). The high heritability of educational achievement reflects many genetically influenced traits, not just intelligence. Proc Natl Acad Sci., 111(42): 15273–8. doi: 10.1073/pnas.1408777111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Lee et al. (2018). Gene discovery and polygenic prediction from a 1.1-million-person GWAS of educational attainment. Nature Genet., 50 (8), 1112–1121. doi: 10.1038/s41588-018-0147-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Loehlin JC (1992). Genes and Environment in Personality Development. Sage Publications, Inc., Newbury Park, CA. [Google Scholar]
  22. Luciano M, Hagenaars SP, Davies G, et al. (2018). Association analysis in over 329,000 individuals identifies 116 independent variants influencing neuroticism. Nat Genet., 50(1): 6–11. doi: 10.1038/s41588-017-0013-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Martin MJ, & Donnellan MB (2020). Socioeconomic status, parenting investments, and negative personality traits over time and across generations. Developmental Psychology. [DOI] [PubMed] [Google Scholar]
  24. Mottus R, Realo A, Vainik U, Allik J, Esko T (2017). Educational attainment and personality are genetically intertwined. Psychol Sci., 28(11): 1631–1639. doi: 10.1177/0956797617719083. [DOI] [PubMed] [Google Scholar]
  25. Nagel M, Jansen PR, Stringer S, Watanabe K, de Leeuw CA, et al. (2018). Meta-analysis of genome-wide association studies for neuroticism in 449,484 individuals identifies novel genetic loci and pathways. Nat Genet., 50(7): 920–927. doi: 10.1038/s41588-018-0151-7. [DOI] [PubMed] [Google Scholar]
  26. Noftle EE, & Robins RW (2007). Personality predictors of academic outcomes: Big Five correlates of GPA and SAT scores. Journal of Personality and Social Psychology, 93: 116–130. doi: 10.1037/0022-3514.93.1.116. [DOI] [PubMed] [Google Scholar]
  27. Okbay A Baselmans BML, et al. (2016a). Genetic variants associated with subjective well-being, depressive symptoms, and neuroticism identified through genome-wide analyses. Nature Genetics, 48: 624–633. doi: 10.1038/ng.3552. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Okbay A, Beauchamp JP, et al. (2016b). Genome-wide association study identifies 74 loci associated with educational attainment. Nature, 553: 539–542. doi: 10.1038/nature17671. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Olson RK (2007). Introduction to the special issue on genes, environment and reading. Reading and Writing, 20(1-2): 1–11. [Google Scholar]
  30. Polderman TJ, Benyamin B, De Leeuw CA, Sullivan PF, Van Bochoven A, Visscher PM, & Posthuma D (2015). Meta-analysis of the heritability of human traits based on fifty years of twin studies. Nature genetics, 47(7), 702–709. doi: 10.1038/ng.3285. [DOI] [PubMed] [Google Scholar]
  31. Poropat AE (2009). A meta-analysis of the Five-Factor Model of personality and academic performance. Psychological Bulletin, 135: 322–338. doi: 10.1037/a0014996. [DOI] [PubMed] [Google Scholar]
  32. Reitveld CA, et al. (2013). GWAS of 126,559 individuals identifies genetic variants associated with educational attainment. Science, 340: 1467–1471. doi: 10.1126/science.1235488. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Roberts BW, Kuncel NR, Shiner R, Caspi A, & Goldberg LR (2007). The power of personality: The comparative validity of personality traits, socioeconomic status, and cognitive ability for predicting important life outcomes. Perspect on Psychol Science, 2: 313–345. doi: 10.1111/j.1745-6916.2007.00047.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Rimfeld K, Kovas Y, Dale PS, Plomin R (2016). True grit and genetics: Predicting academic achiemement from personality. J Pers Soc Psych., 111: 780–789. doi: 10.1037/pspp0000089. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Selzam S, Kraphohl E, von Stumm S, O’Reilly PF, Rimfeld K, Kovas Y, Dale PS, Lee JJ, Plomin R (2017). Predicting educational achievement from DNA. Mol Psychiat., 22(2): 267–272. doi: 10.1038/mp.2016.107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Schwabe I, Janss L, van den Berg SM (2017). Can We Validate the Results of Twin Studies? A Census-Based Study on the Heritability of Educational Achievement. Front Genet., 26;8:160. doi: 10.3389/fgene.2017.00160. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Shiner RL, Masten AS (2012). Childhood personality as a harbinger of competence and resilience in adulthood. Dev Psychopathol., 24(2): 507–28. doi: 10.1017/S0954579412000120. [DOI] [PubMed] [Google Scholar]
  38. Smith-Woolley E, Selzam S, Plomin R (2019). Polygenic score for educational attainment captures DNA variants shared between personality traits and educational achievement. J Pers Soc Psychol., 117(6): 1145–1163. doi: 10.1037/pspp0000241. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Sobowale K,. Ham SA, Curlin FA, Yoon JD (2018). Personality traits are associated with academic achievement in medical school: A nationally representative study. Acad Psychiatry, 42(3): 338–345. doi: 10.1007/s40596-017-0766-5. [DOI] [PubMed] [Google Scholar]
  40. van den Berg SM, et al. (2016). Meta-analysis of genome-wide association studies for extraversion: findings from the Genetics of Personality Consortium. Behav Genet., 46(2): 170–82. doi: 10.1007/s10519-015-9735-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. van Eijck K, & de Graff PM (2004). The Big Five at school: The impact of personality on educational attainment. The Netherlands’ J Soc Sciences, 40: 24–40. [Google Scholar]
  42. Vukasovic T, Bratko D (2015). Heritability of personality: A meta-analysis of behavior genetic studies. Psych Bull., 14(4): 769–785. doi: 10.1037/bul0000017. [DOI] [PubMed] [Google Scholar]
  43. Wainwright MA, Wright MJ, Luciano M, Geffen GM, Martin NG (2005). Multivariate genetic analysis of academic skills of the Queensland core skills test and IQ highlight the importance of genetic g. Twin Res Hum Genet., 8(6): 602–608. doi: 10.1375/183242705774860259. [DOI] [PubMed] [Google Scholar]

RESOURCES