Personality psychologists have been successful in promoting the Big Five personality factors as a scientific model of personality. Short scales have been developed that make it possible to include Big Five measures in studies with large nationally representative samples. These data have been used to examine the influence of personality on wellbeing in married couples (Dyrenforth et al., 2010).
The inclusion of partners’ personality in studies of well-being has produced two findings. First, being married to somebody with a desirable personality (low neuroticism, high extraversion, openness, agreeableness, and conscientiousness) is associated with higher well-being. Second, similarity in personality is not a predictor of higher well-being.
A recent JPSP article mostly replicated these results (van Scheppingen, Chopik, Bleidorn, & Denissen, 2019). “Similar to previous studies using difference scores and profile correlations, results from response surface analyses indicated that personality similarity explained a small amount of variance in well-being as compared with the amount of variance explained by linear actor and partner effects” (e51)
Unfortunately, personality psychologists have made little progress in the measurement of the Big Five and continue to use observed scale scores as if they are nearly perfect measures of personality traits. This practice is problematic because it has been demonstrated in numerous studies that a large portion of the variance in Big Five scale scores is measurement error. Moreover, systematic rating biases have been shown to contaminate Big Five scale scores.
Anusic et al. (2009) showed how at least some of the systematic measurement errors can be removed from Big Five scale scores by means of structural equation modelling. In a structural equation model, the shared variance due to evaluative biases can be modelled with a halo factor, while the residual variance is treated as a more valid measure of the Big Five traits.
The availability of partner data makes it possible to examine whether the halo biases of husbands and wives are correlated. It is also possible to see whether halo bias of a partner has positive effects on well-being. As halo bias in self-ratings is often considered a measure of self-enhancement, it is possible that partner who enhance their own personality have a negative effect on well-being. Alternative, partners who enhance themselves are also more likely to have a positive perception of their partner (Kim et al., 2012), which could increase well-being. An interesting question is how much partner’s actual personality influences well-being after halo bias is removed from partner’s ratings of personality.
It was easy to test these hypotheses with the correlations reported in Table 1 in van Scheppingen et al.’s article, which is based on N = 4,464 couples in in the Health and Retirement Study. Because information about standard deviations were not provided, all SDs were set to 1. However, the actual SDs of Big Five traits tend to be similar so that this is a reasonable approximation.
I fitted the Halo-Alpha-Beta model to the data, but as with other datasets alpha could not be identified. Instead, a positive correlation between agreeableness and extraversion was present in this Big Five measure, which may reflect some secondary loadings that could be modelled with items as indicators. I allowed for the two halo factors to be correlated and I allowed well-being to be predicted by actor-halo and partner-halo. I also allowed for spousal similarity for each Big Five dimension. Finally, well-being was influenced by self-neuroticism and partner-neuroticism because neuroticism is the strongest predictor of well-being. This model had acceptable fit, CFI = .981, RMSEA = .038.
Figure 1 shows the model and the standardized parameter estimates.
The main finding is that self-halo is the strongest predictor of self-rated well-being. This finding replicates Kim et al.’s (2012) results. True neuroticism (tna; i.e., variance in neuroticism ratings without halo bias) is the second strongest predictor. The third strongest predictor is partner’s true neuroticism, although it explains less than 1% of the variance in well-being. The model also shows a positive correlation between partners’ halo factors, r = .32. This is the first demonstration that spouses’ halos are positively correlated. More research is needed to examine whether this is a robust finding and what factors contribute to spousal similarity in halo. This correlation has implications for spousal similarity in actual personality traits. After removing shared halo variance, spousal similarity is only notable for openness, r = .19, and neuroticism, r = . 13.
The key implications of this model is that actual personality traits, at least those measured with the Big Five, have a relatively small effect on well-being. The only trait with a notable contribution is neuroticism, but partner’s neuroticism explains less than 1% of the variance in well-being. An open question is whether the effect of self-halo should be considered a true effect on well-being or whether it simply reflects shared method variance (Schimmack & Kim, in press).
It is well-known that well-being is relatively stable over extended periods of time (Anusic & Schimmack, 2016; Schimmack & Oishi; 2005) and that spouses have similar levels of well-being (Schimmack & Lucas, 2010). The present results suggest that the Big Five personality traits account only for a small portion of the stable variance that is shared between spouses. This finding should stimulate research that looks beyond the Big Five to study well-being in married couples. This blog post shows the utility of structural equation modelling to do so.