Using a large population sample of twins and siblings, the current study provides detailed insights into the genetic overlap between personality and a broad range of well-being measures. Given our large sample size, the present study was well-powered. Overall, our results are in line with the previous finding that especially Neuroticism, Extraversion, and Conscientiousness are genetically the most important personality traits for well-being (
Hahn et al., 2013;
Røysamb et al., 2018;
Weiss et al., 2008). Furthermore, the heritability of the personality traits of ∼40–55% (
Vukasović & Bratko, 2015) and well-being traits of ∼30%–40% (
Bartels, 2015;
Nes & Røysamb, 2015) are comparable with previous meta-analyses.
Our results indicate that personality traits and well-being traits share considerable amounts of common genetic and environmental influences, yet that they are also influenced by their own domain-specific and trait-specific effects. Additive (vs. non-additive) genetic effects were more shared between personality traits and well-being traits, as no trait-specific additive effects were found after accounting for common effects. Non-additive genetic effects showed a greater variety in effects due to different sources. Below we discuss the results in relation to each of our three research questions in detail.
Genetic and Environmental Overlap Between Personality and Well-Being (RQ1)
Genetic and environmental effects shared between personality and well-being traits varied considerably across traits. Genetic effects due to the general, common factor ranged from 15% (Ag) to 89% (Ne) (
Mdn: 60%). Genetic effects on the personality traits due to the personality-specific factor ranged from 0% (Ne) to 78% (Op) (
Mdn: 2%). Genetic effects on the well-being traits due to the well-being-specific factor ranged from 4% (DEP) to 35% (SAT) (
Mdn: 12%). Finally, trait-specific genetic effects ranged from 0% (SAT) to 65% (Co) (
Mdn: 18%). Environmental effects were mostly trait-specific (
Mdn: 68%, ranging from 26% for DEP to 91% for Op), and much less common (
Mdn: 20%, ranging from 0% for Op to 72% for DEP) or domain-specific (
Mdn: 9%, ranging from 2% for Ne, DEP, and LON to 43% for QOL). Of all personality traits, Neuroticism was most strongly related to well-being, and particularly strongly genetically related to depression and loneliness, in line with previous research (
Abdellaoui et al., 2019;
Fanous et al., 2002;
Kendler et al., 2006;
Okbay et al., 2016;
Schermer & Martin, 2019). Because of its pivotal role, Neuroticism is sometimes included as a well-being trait (
Baselmans et al., 2019a,
2019b). On the other hand, Openness, Agreeableness, and self-rated health appeared to mostly be genetically and environmentally distinct from the other traits.
Importantly, the percentages from the previous section are based on common genetic effects on personality and well-being once their respective shared variances have been taken into account. For example, Neuroticism showed the strongest bivariate genetic correlations with well-being traits, but also with the other personality traits. In the best-fitting theoretical model in which shared domain-specific variance was taken into account, it still showed the strongest overlap with well-being. Thus, genetic effects on Neuroticism and well-being were not due to the genetic overlap that Neuroticism shares with other personality traits, or the genetic overlap that well-being traits share with each other. The same was true for Conscientiousness and Extraversion. Earlier claims that these personality traits and well-being are influenced by cross-domain pleiotropic effects (
Hahn et al., 2013;
Røysamb et al., 2018;
Weiss et al., 2008) thus seem to be robust.
Based on our results, it can be concluded that the genetic overlap between personality and well-being is quite large (
Mdn: 60%). This is in line with a proposed (genetic) “covitality” factor (
Figueredo et al., 2004;
Weiss & Luciano, 2015) influencing the variation in both personality and well-being ratings: the recovering of such an overarching factor in our best-fitting model supports this claim. Based on the substantial genetic overlap, it has previously been suggested that “happiness is a personality thing” (
Weiss et al., 2008). Yet, without explicit modeling of the direction of causation, personality may be a well-being thing just as well as well-being may be a personality thing (
Keyes et al., 2015). At the phenotypic level, both directions of causality may indeed be simultaneously operating (e.g.,
Soto, 2015;
Specht et al., 2013). However, the current study shows that shared genes will act as a confounder for these effects. Additional research on causality in which genetic confounding is taken into account is thus needed (
Briley et al., 2018).
When these causal mechanisms become more clear, our results are informative for future intervention studies. Although both are relatively stable over the lifespan, well-being is thought to be more malleable than personality (
Anusic & Schimmack, 2016) and several well-being interventions have proven to be successful (
van Agteren et al., 2021). Again, genetic effects need to be taken into account, as they play a role in stability and change of both personality and well-being (
Nes et al., 2006;
Pedersen & Reynolds, 1998). By gaining more insights into what (genetically) separates well-being from personality, it will become easier in the future to target interventions specifically at effects unique to well-being.
Our findings on common, domain-specific, and trait-specific effects have implications for molecular genetic studies. GWASs are designed to identify the genetic variants associated with a trait. Several GWASs on personality (
De Moor et al., 2015;
Lo et al., 2017;
van den Berg et al., 2016;
Weiss et al., 2016) and well-being (
Baselmans et al., 2019a;
Okbay et al., 2016;
Turley et al., 2018) have been published in recent years. Recently, multivariate methods have been developed to investigate the (latent) genetic structure underlying traits at a molecular genetic level and use this structure to find new genetic variants for the identified latent factors (Genomic SEM;
Grotzinger et al., 2019). Our models can be used as input for such investigations. Ultimately, this should make it possible in the future to arrive at a clear picture of the variants that are uniquely associated with well-being and personality, or with both.
Based on our results, one could alternatively argue that, overall, personality and well-being are quite distinct (100%–60% = 40%). With regards to the overlap and distinction, we largely concur with Keyes and colleagues (2015) who noted that personality reflects how one functions in life, while well-being reflects how well one functions. Being both part of the process of functioning in life they have much in common, but they also differ in their role in this process. These differences and similarities are likely to be reflected in their genetic makeup.
The Influence and Interpretation of Domain-Specific Shared Variance
Although we fitted domain-specific factors mostly to control for domain-specific variance, our results can provide insights for the interpretation of these factors. In the CP models, we found that loadings of Neuroticism (∼ −.85), Extraversion (∼ .55), and Conscientiousness (∼ .46) on the common personality factor were sizeable, while loadings of Agreeableness (∼ .23) and Openness (∼ −.08) were low. We thus did not find strong support for a phenotypic common personality factor (referred to as the General Factor of Personality;
van der Linden et al., 2016). At the same time, the domain-specific well-being factor was well-defined by all well-being traits in our CP models, with phenotypic loadings ranging from ∼.40 (self-rated health) to ∼ −.84 (loneliness). In addition, in the IP models, domain-specific effects were more pronounced for well-being compared to personality. These results provide evidence for a broad, general well-being factor underlying different well-being measures (e.g.,
Longo et al., 2016) and makes it plausible that this factor has a solid genetic basis (
Bartels & Boomsma, 2009;
Baselmans & Bartels, 2018).
Nevertheless, the superior fit of IP (vs. CP) models implies that these common factors must be interpreted with caution. This finding indicates that they may not be the causal factors influencing their indicators, as the common and unique effects operate at the indicator level, and not at the common factor level (
Franić et al., 2013). Yet, the existence of a latent construct cannot be proven or disproven based on the relative fit of IP over CP models alone. For example, IP models tend to fit better than CP models when fitting them to the facets underlying each of the Big Five factors (
Franić et al., 2014;
Jang et al., 2002). Rather than dismissing the Big Five as constructs altogether,
Jang et al. (2002) concluded that they “do not exist as veridical psychological entities per se, but rather they exist as useful heuristic devices that describe pleiotropic effects and the common influence of environmental factors on sets of individual facets.” (p. 99). Similarly, the common factors in the current study may be viewed as an organization of traits on which common genetic and environmental are operate, each of them also having their own unique influences. Ultimately, to answer the question what these common factors represent, multi-trait-multi-method (MTMM) studies based on ratings of personality
and well-being (see
Schimmack & Kim, 2020) in a genetically informative design are needed to accurately separate trait from method effects (
Bartels et al., 2007;
Borkenau et al., 2001).
Although not providing clear evidence on its meaning, the current study can parsimoniously explain why controlling for the shared Big Five variance reduces their correlations with well-being (
Kallio Strand et al., 2021;
Kim et al., 2018;
Schimmack & Kim, 2020). In the suboptimal CP models, the genetic and environmental correlations between the latent general well-being and general personality factor were much higher (1.00. .96, and .81, for ADE respectively) than in the IP models (.25, 1.00, and .50, respectively). If then, in the CP models, the common genetic effects on indicators are aggregated to a higher level in an unbalanced way (as is the case for the higher-loading Neuroticism, Extraversion, and Conscientiousness, compared to Openness and Agreeableness), then this will artificially lead to higher genetic correlations between the common factors. These stronger genetic correlations translate to the phenotypic level. Thus, when we control for the shared phenotypic personality variance, then we are haphazardly controlling for the “true” underlying genetic and environmental effects at the indicator level, reducing the correlations between the Big Five and well-being. Again, this hypothesis needs to be tested in the future using genetically informative MTMM studies.
Non-additive Genetic Effects (RQ2)
In line with previous work, significant amounts of non-additive variance were found to influence both personality and well-being, and their overlap (
Bartels & Boomsma, 2009;
Hahn et al., 2013;
Keller et al., 2005). Non-additive genetic effects accounted for between 14% (depressive symptoms) to 95% (Agreeableness) of the total genetic variance in the traits (
Table 4). In the Cholesky model, absolute non-additive genetic correlations ranged from .13 to .93 (
Mdn: .47). This is important, for example, for future molecular genetic studies trying to identify the genes associated with personality and well-being, since the methods used in such studies often assume additive genetic effects (
Visscher et al., 2017). The amount of non-additive variance present in traits is also important for theoretical reasons, as it is assumed to be indicative of the evolutionary pressures that have caused these traits to emerge (
Penke et al., 2007;
Verweij et al., 2012).
With our current sample size, we had sufficient power to detect non-additive genetic effects (D), but this does not apply to all previous studies on this topic. We found that especially for D, traits differed in the amount of effects due to common, domain-specific, and trait-specific effects. This will obscure results when effects are aggregated to higher trait levels. For example, when one creates a general well-being scale from multiple scales that differ in their common and unique additive and non-additive effects, then the resulting general measure will be a cloudy mix of these different genetic effects. These findings stress the importance of modeling higher order factors (e.g., “general well-being”) as latent variables in twin designs, to uncover the nuances in their underlying genetic effects.
Sex Differences in Genetic and Environmental Effects (RQ3)
In our large sample, we found moderate to small mean sex differences on the Big Five. In line with previous studies (
Costa et al., 2001;
Schmitt et al., 2008;
Weisberg et al., 2011), females scored higher on Neuroticism and Agreeableness, and somewhat higher on Conscientiousness. In contrast to other studies, we found no sex differences in Extraversion, which may be due to our focus on the Big Five factors rather than facets residing below the Big Five. Females tend to score higher on the facet Enthusiasm and males on Assertiveness (
Costa et al., 2001;
Feingold, 1994;
Weisberg et al., 2011). At the aggregate factor level, these differences may have canceled each other out. Sex differences on well-being traits were generally small, with the largest effect found for depression, also replicating previous work (
Batz & Tay, 2018;
Batz-Barbarich et al., 2018;
Eaton et al., 2012).
Given our large sample and similar results from previous studies (
Bartels, 2015;
Keyes et al., 2010;
Røysamb et al., 2018;
South et al., 2018;
Vukasović & Bratko, 2015), it seems safe to assume that, at the aggregate level, the same genes influence personality and well-being for males and females, and to the same extent. This is important information for theoretical and practical reasons as it suggests that mean differences are probably due to non-shared environmental circumstances. These non-shared environmental exposures reflect idiosyncratic experiences that only a single twin within the same family experiences, making them more different from their siblings. This may include life events, differences in socialization, different opportunities, or specific gender roles (
South et al., 2018). Our results further imply that in future gene finding studies, male-specific and female-specific genes for personality and well-being are unlikely to be found.
It is tempting to conclude that the mean sex differences on personality and well-being are completely unrelated to genetic differences. However, genes may still play a role through more subtle processes such as gene-environment interplay. For example, we investigated genetic and environmental influences independent of age effects by regressing them out from the traits. It may be that a sex by age interaction is present, implying that quantitative or qualitative sex differences are only apparent at specific ages (e.g., during adolescence). For instance, puberty seems to coincide with increases in mean levels of internalizing symptoms and with increases in its heritability, particularly in girls (
Bergen et al., 2007;
Patterson et al., 2018). Future studies investigating genetic and environmental effects as a function of both age and sex are needed to confirm such processes for personality and well-being.
It is also possible that genetic differences exist between males and females, but that these are masked by unmodeled gene by environment interaction (GxE) effects. Traditional twin models assume that GxE is not present, that is, that genetic effects are similar across different environments and/or subgroups. This may not be the case;
Nes et al. (2010b), for example, showed that the environmental exposure marriage influenced the heritability estimates of SWB. Importantly, these marriage effects differed across males and females. GxE effects may also explain why gender differences tend to be larger in more prosperous societies: possible genetic differences between males and females may be more easily expressed in developed countries (
Schmitt et al., 2008). In our study, we investigated a sample from the Netherlands, a highly developed country with relatively equal opportunities for males and females. Within our egalitarian sample, the smaller amount of variance in opportunities and gender roles between males and females may have attenuated the expression of genetic sex differences. Future studies that explicitly model GxE effects for males and females, preferably across countries with different developmental standards, are thus needed.
Posts
