Journal of Research in Personality 47 (2013) 369375

Contents lists available at SciVerse ScienceDirect

Journal of Research in Personality

journal homepage: www.elsevier.com/locate/jrp

Personality similarity and life satisfaction in couples

Katrin Furler , Veronica Gomez, Alexander Grob
Department of Psychology, University of Basel, Switzerland

a r t i c l e

i n f o

a b s t r a c t
The present study examined the association between personality similarity and life satisfaction in a large, nationally representative sample of 1608 romantic couples. Similarity effects were computed for the Big Five personality traits as well as for personality proles with global and differentiated indices of similarity. Results showed substantial actor and partner effects, indicating that both partners personality traits were related to both partners life satisfaction. Personality similarity, however, was not related to either partners life satisfaction. We emphasize the importance of thoroughly controlling for each partners personality and for applying appropriate analytical methods for dyadic data when assessing the effect of personality similarity in couples. 2013 Elsevier Inc. All rights reserved.

Article history: Available online 15 March 2013 Keywords: Personality similarity Life satisfaction Couples Differentiated approach Actor effects Partner effects

1. Introduction Personal relationships in general, and romantic relationships in particular, are essential for peoples well-being. Romantic relationships represent a unique social unit in which partners spend a considerable amount of time together and share closeness and intimacy with each other (Argyle, 1999). A recent longitudinal study on development of life satisfaction in couples revealed that both members of a couple inuence each other and mutually affect the other partners well-being (Hoppmann, Gerstorf, Willis, & Schaie, 2011). In this regard, a dyadic approach that focuses on couples as the unit of analysis is of crucial interest for personality psychology. Correspondingly, within- and between-person as well as similarity effects of personality on well-being have received considerable attention in recent research and both ones own as well as the partners personality characteristics have been found to be important factors in intimate relationships (e.g., Barelds, 2005; Dyrenforth, Kashy, Donnellan, & Lucas, 2010; Gonzaga, Campos, & Bradbury, 2007; Luo & Klohnen, 2005). Evidence so far speaks for consistent effects of ones own personality on well-being (actor effects), such that being extraverted, agreeable, conscientious, emotionally stable and open to experience is positively related to a persons well-being (e.g., Steel, Schmidt, & Shultz, 2008). But it is not only ones own personality that affects wellbeing, it is also the partners personality (partner effects). Being in an intimate relationship with someone who is extraverted, agreeable, conscientious, and emotionally stable is associated with
Corresponding authors. Address: Department of Psychology, University of Basel, Missionsstrasse 62, 4055 Basel, Switzerland. E-mail addresses: katrin.furler@unibas.ch (K. Furler), alexander.grob@unibas.ch (A. Grob).
0092-6566/$ - see front matter 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jrp.2013.03.002

higher well-being as well (Barelds, 2005; Dyrenforth et al., 2010; Headey, Muffels, & Wagner, 2010; Robins, Caspi, & Moftt, 2000). However, as far as personality similarity in romantic couples is concerned, the everlasting question about who is a persons perfect match has not yet been answered satisfactorily. Is having a partner whose personality is similar to ones own associated with higher levels of satisfaction? Or do we prefer someone rather dissimilar? The present study tries to shed light on this issue and examines how personality similarity in couples relates to each partners life satisfaction. Research to date leans more towards like attracts like rather than opposites attract and assumes that similarity, as opposed to complementarity (Antill, 1983), is an important factor in romantic relationships. Studies on mate selection for instance support the notion that similarity between partners is essential in forming intimate relationships. People tend to choose partners who are similar to themselves on a number of variables, including age, education, values, physical attractiveness, and intelligence (Epstein & Guttman, 1985; Vandenberg, 1972). Similarity is theorized to be benecial for intimate relationships by coordinating partners thoughts and behaviors, increasing understanding of each others intentions and motivations, and reinforcing their appraisals, leading to relationship satisfaction and longevity (Anderson, Keltner, & John, 2003). Nevertheless, spousal similarity in personality traits is generally quite low (Watson et al., 2004) although people tend to prefer partners with similar personalities (Botwin, Buss, & Shackelford, 1997). The existing body of research on the association between personality similarity and well-being in couples does not provide a clear picture. Some studies show positive effects for relationship or life satisfaction respectively, such that the more similar couples personalities, the higher each partners satisfaction in the

370

K. Furler et al. / Journal of Research in Personality 47 (2013) 369375

respective domain (Arrindell & Luteijn, 2000; Gaunt, 2006; Gonzaga, Carter, & Buckwalter, 2010; Gonzaga et al., 2007; Luo & Klohnen, 2005; Robins et al., 2000). However, a recent study using nationally representative panel data from Great Britain, Australia, and Germany revealed no or only small associations between personality similarity and relationship or life satisfaction and neither of these small effects was consistent across the three samples (Dyrenforth et al., 2010). Thus, even with big sample sizes having the power to detect very small effects, there was little evidence that personality similarity predicted relationship or life satisfaction in those three large samples. Furthermore, in a representative sample of Dutch couples, personality similarity did not affect marital quality (Barelds, 2005) and in a sample of distressed and treatment-seeking couples, similarity on any of the Big Five traits was not a predictor of marital satisfaction (Gattis, Berns, Simpson, & Christensen, 2004). Given the inconsistent empirical evidence, research is needed in order to deepen our understanding of how personality similarity in romantic relationships relates to each partners well-being. For this purpose, three important factors that have been neglected in several previous studies have to be considered. First, when assessing the effect of personality similarity in couples, it is essential to control for each partners individual level of personality to get the unique similarity effect beyond each partners individual contribution (Grifn, Murray, & Gonzalez, 1999; Kenny & Acitelli, 1994). Not controlling for initial levels of both partners personality leads to an overestimation of the association between similarity and wellbeing. Some of the studies reporting signicant similarity effects did not or did not thoroughly control for individual levels of partners personality (Gonzaga et al., 2007; Robins et al., 2000). When main effects were included in the analysis, however, similarity was no longer a unique predictor of well-being (Barelds, 2005; Dyrenforth et al., 2010; Gattis et al., 2004). Second, to gain generalizable results and a more accurate sense of the association between personality similarity and well-being, it is necessary to examine a large representative sample of couples. Several studies that found similarity effects on well-being analyzed relatively small (Arrindell & Luteijn, 2000; Gonzaga et al., 2007) or very specic and thus possibly biased samples such as newlyweds (Luo & Klohnen, 2005), cohort study members (Robins et al., 2000) or couples who met via online dating platforms (Gonzaga et al., 2010). However, the few studies examining large representative samples failed to nd similarity effects on well-being (Barelds, 2005; Dyrenforth et al., 2010). Third, in light of the dyadic nature of couple data it is necessary to make use of proper analytic techniques and similarity measures. Thus far, many studies have treated dyadic data as if they were individual data, for instance by conducting analyses separately for husbands and wives or using simple correlational methods that fail to capture the interdependent nature of couple data. This shortcoming in previous studies has been pointed out as a major problem by relationship researchers and can be overcome by applying appropriate analytical methods that take the interdependence of dyadic data into account and are able to test for the unique effect of each independent variable on well-being (see Kenny, Kashy, & Cook, 2006). Furthermore, studies on personality similarity in couples have used a wide array of similarity measures. The most common measures reported include difference scores and prole correlations. Difference scores are straightforward and intuitively understandable in the way we usually think about differences in daily life (Grifn et al., 1999). These scores are typically computed at the trait level, taking the absolute value of difference between two partners scores on a given trait. They thus indicate how dissimilar two members of a couple are with respect to a specic trait. Yet researchers criticize this approach mainly due to the lowered

reliability inherent in difference scores. If the two component variables are positively correlated, as is often the case given that the scores are usually measured with the same instrument, the reliability of the difference between those two components becomes less reliable (Edwards, 1994, 2001). Furthermore, difference scores can be confounded with each partners individual score. A simple solution to avoid this problem is to include both partners individual scores in the same analysis (Grifn et al., 1999). Prole correlations, on the other hand, are more difcult to interpret. Generally, personality similarity computed at the prole level represents the degree to which both couple members overall personality proles are similar to each other; that is, how well two partners match on a set of personality traits (see Cronbach & Gleser, 1953). To interpret the effects of prole similarity adequately, it is important to know that a prole consists of three elements (Cronbach & Gleser, 1953; see also Furr, 2010). First, every prole has a shape that represents the pattern of scores in a prole. It reects which traits have relatively high scores and which ones have relatively low scores within the same prole. Elevation represents the overall mean across all traits within a prole. And scatter refers to the variability or variance among the scores of a prole. It thus reects how much the trait scores deviate within the same prole. Because a prole comprises these three elements, similarity between different proles can be measured in various ways. A commonly used prole correlation measure is the Intraclass Correlation Coefcient (ICC). It reects a global index because it captures all three characteristics of a prole at once (e.g., Dyrenforth et al., 2010). However, assuming ICC as a global index of personality similarity might confound ndings because shape, elevation, and scatter are conceptually different from one another and should not be mixed within the same analysis. Instead, Furr (2010) suggests a differentiated analysis and argues for the necessity to separately examine all elements of a prole. Hence, shape similarity is calculated by correlating the scores in one partners prole with the scores of the other partners prole using Pearson correlations. Second, elevation similarity is measured using difference scores (i.e., absolute value of difference between the overall mean across traits: mean similarity). Third, scatter similarity is also computed using difference scores (i.e., absolute value of difference between the variances across all traits within the proles: variance similarity). In light of the range of possible measures of similarity, another goal of the present study is to rule out that the use of different similarity indices results in different ndings. By adopting a differentiated approach we can determine whether results vary depending on how prole similarity is measured. Some researchers who applied difference scores and prole correlations in the same study reported that prole-based similarity was more strongly associated with satisfaction (Gaunt, 2006; Luo & Klohnen, 2005), whereas others did not nd different effects with different similarity measures (Dyrenforth et al., 2010). No study to date has analyzed similarity by means of a differentiated approach; that is, examining each element of a prole separately (as proposed by Furr, 2010). In brief, the aim of the present study is to clarify the association between personality similarity and life satisfaction in couples. Three reasons speak for the relevance of this research question. First, we think that some previous ndings mostly represent an overestimation of the effect of similarity because many studies did not take each partners personality into account (e.g., Robins et al., 2000). Thus, we will control for each partners individual level of personality to examine the effect of personality similarity on life satisfaction beyond the effects of ones own and the partners personality. In line with previous research, we expect actor and partner effects for the association between personality and life satisfaction in couples (positive actor and partner effects for

K. Furler et al. / Journal of Research in Personality 47 (2013) 369375

371

Extraversion, Agreeableness, Conscientiousness, Emotional Stability as well as positive actor effects for Openness; e.g., Dyrenforth et al., 2010). Second, with the exception of only a few studies (Barelds, 2005; Dyrenforth et al., 2010), previous research mainly used small or specic samples (e.g., Arrindell & Luteijn, 2000; Luo & Klohnen, 2005). Thus, in order to provide generalizable results we use a nationally representative sample of couples living in Switzerland. We expect to replicate ndings of Dyrenforth et al., 2010 who analyzed comparable data sets of other countries. Third, we apply proper methods for dyadic data and extend prior research on similarity and well-being by adopting a differentiated approach: associations between personality similarity and life satisfaction in couples will be analyzed both at the trait and at the prole level. The latter includes global prole similarity (ICC), as well as similarity of shape, elevation (mean), and scatter (variance). 2. Method 2.1. Participants Participants were drawn from the Swiss Household Panel (SHP), a nationally representative household panel study in Switzerland. The SHP is an annual survey, which started in 1999, of a random sample of Swiss households. All adult members of selected households were asked to participate using computer-assisted phone interviewing. Since its start, the SHP has collected data on close to 8000 households, including a variety of social and demographic measures (for further information see Voorpostel et al., 2010). The sample for the present study consists of all heterosexual romantic couples for which life satisfaction and personality data of both partners were available in 2009, leading to a total sample size of 1608 couples (Nind = 3216). Men were on average 51.88 years old (SD = 13.58), women 49.10 years (SD = 13.35). The mean length of the relationship was 24.21 years (SD = 14.11) and the majority of couples were married (85%). 2.2. Measures Life satisfaction was assessed with a single item asking respondents how satised they felt with their life in general ranging from 0 (not at all satised) to 10 (completely satised) (M = 8.14, SD = 1.24). Using single items is nearly custom and widely accepted within large-scale panel studies to minimize subject burden (see also Lucas & Donnellan, 2012). Men and women did not differ in their life satisfaction scores (t(1606) = .51, p = .61). Personality was assessed using a German 10-item instrument derived from the Big Five Inventory (BFI; see John & Srivastava, 1999 for the original version). For each trait, participants reported the extent to which each of the two statements per personality trait described themselves on an 11-point scale ranging from 0 (completely disagree) to 10 (completely agree). Correlations between the two items for Extraversion, Agreeableness, Conscientiousness, Emotional Stability, and Openness were .39, .11, .24, .40, and .21, respectively, and are consistent with previously published research using personality data from the SHP (Anusic, Lucas, & Donnellan, 2012). Furthermore, the 10-item BFI correlates highly with the original BFI-44 scale (r = .83) and predicts around 70% of the variance of the full scale (Rammstedt & John, 2007). 2.3. Rationale of analysis We applied structural equation modeling to estimate different sets of Actor-Partner Interdependence Models (APIM; Kenny et al., 2006). The APIM is a dyadic relationship model that takes interdependencies between two members of a dyad into account

and models each dyad members effect on his/her own outcome (actor effect) as well as on the partners outcome (partner effect). Two sets of analyses were conducted. The rst set of analyses focused on personality similarity at the trait level and comprised ve models (trait level models; one for each Big Five trait). In these models, life satisfaction of both couple members was predicted by each partners personality (actor and partner effects) as well as the couples personality similarity on a given trait. As mentioned earlier, personality similarity at the trait level is represented as the absolute value of difference between two partners scores on a given trait. This difference score was then multiplied by 1 in order to be indicative for similarity instead of difference. Hence, ve models were estimated to analyze actor, partner, and similarity effects in the association between the Big Five traits and life satisfaction of couples (see Fig. 1 for an illustration of the trait level models). The second set of analyses focused on personality similarity at the prole level and consisted of two models. In both models life satisfaction of both couple members was predicted by each partners personality (actor and partner effects of all Big Five traits) as well as couples personality similarity using different prole similarity indices for the two models. (1) The global prole model included a global index of prole similarity represented by the Intraclass Correlation Coefcient (ICC), which captures all elements of a prole at once. This model estimated the overall or global effect of prole similarity above and beyond each partners individual level of personality. (2) The differentiated approach model included three different prole similarity indices to capture each element of the prole separately (i.e., shape, mean, and variance similarity). Thus, this model provided insight into the unique predictive power of each element as proposed by Furr (2010). Moreover, these analyses allowed a thorough examination of whether the application of a differentiated approach with distinguished prole similarity indices for shape, elevation, and scatter yields valuable additional information that goes beyond a global prole similarity index. The similarity indices used in the differentiated approach model were represented by (a) Pearson correlation coefcient for shape similarity, (b) the absolute value of difference between partners overall mean across all Big Five traits (multiplied by 1) for mean similarity, and (c) the absolute value of difference between partners variance within their Big Five traits (multiplied by 1) for variance similarity. Hence, the differentiated approach model estimated the effect of three different prole similarity indices above and beyond each partners individual level of personality (see Figs. 2 and 3 for an illustration of the global prole and the differentiated approach model, respectively). 3. Results Descriptive statistics and correlations of all variables are presented in Table 1. Intercorrelations between similarity indices

T it Trait y Similarity

Life Satisfaction Man Life Satisfaction Woman

e1

Big Five Trait Man Big Five Trait Woman

e2

Fig. 1. Model specication for the ve trait level models with trait similarity and the Big Five traits of both partners as predictors of each partners life satisfaction. Trait similarity is represented by the absolute value of difference between two partners scores on a given trait (multiplied by 1).

372

K. Furler et al. / Journal of Research in Personality 47 (2013) 369375

Global Similarity

Life Satisfaction Man

e1

Big Five Trait Man

Life Satisfaction Woman

e2

Big Five Trait Woman

Fig. 2. Model specication for the global prole model with the global index of personality similarity as predictor of each partners life satisfaction, controlling for all Big Five traits of both partners. Global similarity is represented by the Intraclass Correlation Coefcient (ICC). Covariances between all predictor variables were included but are not displayed for reasons of clarity.

Shape Similarity Mean Similarity Variance Similarity

Life Satisfaction Man Life Satisfaction Woman

e1

e2

Big Five Trait Man

Big Five Trait Woman

Fig. 3. Model specication for the differentiated approach model with shape, mean, and variance similarity as predictors of each partners life satisfaction, controlling for all Big Five traits of both partners. Shape similarity is represented by the Pearson correlation coefcient; mean similarity is represented by the absolute value of difference between partners overall mean across all Big Five traits (multiplied by 1); variance similarity is represented by the absolute value of difference between partners variance within their Big Five traits (multiplied by 1). Covariances between all predictor variables were included but are not displayed for reasons of clarity.

were generally low or moderate, which indicates independence and conceptual differences between these indices. As an exception, the global similarity index (measured by ICC) and the shape similarity index (measured by Pearson correlation) were strongly correlated. This is not surprising given the conceptual overlap between these two indices. As mentioned earlier, the ICC captures all three elements of a prole (shape, elevation, and scatter), whereas the Pearson correlation captures the shape of a prole. Values of the different indices ranged from very low to very high (e.g., .98 to .97 for global similarity) meaning that there are both couples that are highly similar as well as couples whose personalities are fairly dissimilar. The rst set of analyses at the trait level consisted of estimating ve separate models for each Big Five personality trait (see Fig. 1). These were saturated models with zero degrees of freedom. Within each model, estimates of actor, partner, and similarity effects did not differ between men and women because constraining the respective paths to be invariant did not lead to a signicant deterioration in model t in either model (Extraversion: Dv2 (Ddf = 3) = 3.35, p = .34; Agreeableness: Dv2 (Ddf = 3) = 1.00, p = .80; Conscientiousness: Dv2 (Ddf = 3) = 1.37, p = .71; Emotional

Stability: Dv2 (Ddf = 3) = 1.83, p = .61; Openness: Dv2 (Ddf = 3) = 1.57, p = .67). Table 2 presents results of the ve trait level models with invariant actor, partner, and similarity effects across men and women.1 In line with previous research, ones own as well as the partners personality was related to ones life satisfaction. Actor effects were found for each Big Five trait such that people high in Extraversion, Agreeableness, Conscientiousness, Emotional Stability, and Openness reported higher levels of life satisfaction. In a similar vein, having an extraverted, agreeable, conscientious, and emotional stable partner was also positively associated with ones life satisfaction. Effect sizes for actor and partner estimates were generally small (according to Cohen, 1988). No partner effects were identied for Openness. Regarding personality similarity in couples on a trait level, partners tended to be more similar than dissimilar on single Big Five personality traits (see means and ranges of trait similarity indices in Table 1). However, there was little evidence for similarity effects beyond the individual contribution of each partners personality (see Table 2). Coefcients were generally small and only one effect reached signicance, which has to be interpreted with caution considering the large number of analyses conducted. The second set of analyses focused on personality proles predicting life satisfaction. First, we established a global prole model (Fig. 2) with the ICC as global prole similarity index controlling for actor and partner main effects of the Big Five traits. This model provides an overall prole similarity effect on both couple members life satisfaction above and beyond the effects of ones own and the partners personality. Second, we adopted a differentiated approach to prole similarity and estimated a differentiated approach model (Fig. 3) with prole indices of shape, mean, and variance similarity. Again, we controlled for actor and partner main effects of all Big Five traits. This model provides the unique prole similarity effect of each element of a prole on both couple members life satisfaction above and beyond the effects of ones own and the partners personality. Actor, partner, and similarity estimates did not differ between men and women because constraining the respective paths to be invariant did not lead to a signicant deterioration in model t in either model (global prole model: Dv2 (Ddf = 11) = 10.05, p = .53; differentiated approach model: Dv2 (Ddf = 13) = 14.09, p = .37). Even though the similarity indices reect conceptually different estimates, results at the prole level revealed no similarity effects, neither in the global prole model (b = .00, p = .84; R2 = .13), nor in the differentiated approach model (bshape similarity = .02, p = .32; bmean similarity = .00, p = 1.00; bvariance 2 similarity = .00, p = .95; R = .13). In short, after controlling for each partners personality traits, prole similarity was unrelated to both partners life satisfaction, both in the global model as well as in the differentiated approach model.2 Additionally, because our sample comprised a variety of couples with a broad range of relationship length, we examined the potential inuence of how long couples had been together by including relationship length as a moderator in all analyses. We added an interaction term (relationship length similarity) to all our trait as well as prole level models. Again, estimates did not differ between men and women because constraining the respective paths to be invariant did not lead to a signicant deterioration in model t in either trait model (Extraversion: Dv2 (Ddf = 5) = 2.25, p = .66; Agreeableness: Dv2 (Ddf = 5) = 1.99, p = .85; Conscientiousness:
1 The invariant standardized regression coefcients across men and women were averaged using Fishers z-transformation. 2 An important issue concerns the fact that people tend to respond to items in a normative or stereotypical way. Because of this response pattern, prole correlations between two random individuals are likely not zero (Furr, 2008; Kenny et al., 2006). To account for this potential source of confounding, the analyses were also conducted using stereotype adjusted scores instead of raw scores (according to Kenny et al., 2006). However, results remained the same and thus, we only reported the results for the raw scores.

K. Furler et al. / Journal of Research in Personality 47 (2013) 369375 Table 1 Intercorrelations and descriptives for study variables (N = 1608). 1. Men 1. 2. 3. 4. 5. 6. Women 7. 8. 9. 10. 11. 12. Life satisfaction Extraversion Agreeableness Conscientiousness Emotional stability Openness Life satisfaction Extraversion Agreeableness Conscientiousness Emotional stability Openness .18 .23 .22 .26 .08 .29 .09 .12 .10 .11 .03 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21.

373

.21 .24 .25 .16 .09 .14 .07 .07 .09 .00

.27 .31 .04 .10 .03 .10 .07 .06 .29 .10 .12 .01 .07 .05 .05 .12 .19 .20 .22 .29 .12 .22 .21 .20 .18 .32 .31 .11 .29 .13

.13 .01 .08 .06 .08 .03 .09 .03 .04 .01 .01 .13

.09 .08 .09 .12 .07 .11 .06 .03 .13 .06

.04 .02

Trait similarity indices 13. Extraversion .01 .14 .02 .00 .03 .01 .05 .00 .02 .07 .02 .01 14. Agreeableness .02 .03 .11 .07 .04 .03 .04 .02 .04 .01 .01 .03 15. Conscientiousness .00 .04 .04 .20 .03 .01 .00 .01 .03 .05 .02 .03 16. Emotional .01 .01 .05 .05 .09 .04 .07 .04 .06 .04 .32 .01 stability 17. Openness .02 .02 .09 .03 .07 .06 .02 .03 .06 .06 .03 .00 Prole similarity indices 18. Global 19. Shape 20. Mean 21. Variance M SD Min Max .01 .02 .03 .18 .02 .00 .06 .17 .03 .01 .04 .01 .00 .07 .01 .11 8.13 6.44 6.81 7.49 1.21 1.86 1.36 1.40 0.00 0.00 0.00 0.00 10.00 10.00 10.00 10.00 .11 .11 .03 .06 6.81 1.62 0.00 10.00 .15 .01 .09 .03 .14 .01 .12 .02 .09 .05 .13 .08 .04 .00 .02 .04 .01 .04 .08 .03 .06 .06 .09 .14 6.24 8.14 6.75 6.82 7.67 6.28 1.75 1.26 1.92 1.42 1.44 1.75 0.00 3.00 0.00 1.00 1.00 0.00 10.00 10.00 10.00 10.00 10.00 10.00

.06

.19 .35 .17 .31 .33 .30 .17 .27 .07 .19 .24 .23 .85 .02 .36 .29 .35 .32 .28 .43 .01 .14 .27 .18 .18 .23 .21 .11 .05 .06 6.34 1.97 1.45 1.56 1.90 1.91 0.03 0.14 1.02 2.15 1.86 1.54 1.17 1.21 1.45 1.44 0.44 0.50 0.79 2.28 0.00 9.00 8.00 8.00 8.50 8.00 0.98 0.97 4.60 17.78 10.00 0.00 0.00 0.00 0.00 0.00 0.97 1.00 0.00 0.00

Note: Correlation coefcients printed in bold are signicant at the p < .05 or p < .01 level.

Table 2 Standardized estimates of the ve trait level models predicting life satisfaction of both partners. Trait Standardized estimates Actor Extraversion Agreeableness Conscientiousness Emotional stability Openness .18*** .21*** .22*** .27*** .10*** Partner .07*** .10*** .11*** .11*** .01 Similarity .03 .05** .04 .01 .02 .04 .06 .06 .09 .01 R2

Note: Models were estimated separately for each trait. For each of the ve models, both partners personality on the given trait and trait similarity were included as predictors of life satisfaction. Estimates and explained variance (R2) refer to models with invariant actor, partner, and similarity effects across men and women. ** p < .01. *** p < .001.

Dv2 (Ddf = 5) = 0.51, p = .99; Emotional Stability: Dv2 (Ddf = 5) = 5.28, p = .38; Openness: Dv2 (Ddf = 5) = 2.21, p = .82) or prole model (global prole model: Dv2 (Ddf = 13) = 10.51, p = .65; differentiated approach model: Dv2 (Ddf = 17) = 16.70, p = .48). More importantly, no substantial effects point to similarity as a function of relationship length. Standardized estimates for the interaction effects ranged from .02 to .01, and were all not significant (all ps > .35). Only one interaction effect for similarity in Conscientiousness reached signicance (b = .05, p < .01), but the effect was too small to merit attention.

4. Discussion The goal of the present study was to clarify the effects between personality similarity and life satisfaction in couples by examining

a large representative sample of romantic couples and adopting a differentiated methodological approach. Overall, the results of our study reveal little evidence for associations between personality similarity in couples and each partners life satisfaction. This does not mean, however, that two partners personalities are unrelated to their individual life satisfactionrather the opposite is true: in line with previous research, ones own personality as well as the partners personality is substantially related to both partners life satisfaction (Barelds, 2005; Dyrenforth et al., 2010). The fact that not only within-person characteristics but also partner characteristics affect each partners life satisfaction gains remarkable importance, and even more so when bearing in mind the myriad of potential (within-person) correlates of subjective well-being. Nevertheless, we tried to plunge even further into dyadic effects and laid our main emphasis on personality similarity as a predictor of life satisfaction over and above actor and partner effects of the Big Five traits. Neither at the trait nor at the prole level did our results provide evidence for personality similarity effects on life satisfaction. Only one effect for similarity at the trait level reached signicance. However, the effect was too small to merit interpretation. Our results rather afrm previous research with comparable data sets of nationally representative samples that did not nd any evidence for effects of personality similarity on well-being (Barelds, 2005; Dyrenforth et al., 2010). Although an issue in everyday life, empirical evidence does not support the common saying like attracts like, that is, personality similarity is not substantially related to life satisfaction in couples. Three main factors may account for this lack of relevance of personality similarity for each couple members well-being. First, the present study controlled for actor and partner effects and results

374

K. Furler et al. / Journal of Research in Personality 47 (2013) 369375

conrm previous research showing that personality similarity is not a predictor of life satisfaction above and beyond each partners individual level of personality (Barelds, 2005; Dyrenforth et al., 2010; Gattis et al., 2004). Thus, given the magnitude of both actor and partner effects, we again emphasize the importance of thoroughly controlling for each partners individual level of personality when assessing the effect of similarity in couples. Several studies reporting signicant similarity effects have neglected this issue (e.g., Gonzaga et al., 2007; Robins et al., 2000). Second, one of the strengths of the present study is the use of panel data. Most previous research had smaller sample sizes (Arrindell & Luteijn, 2000) or laid the focus on specic samples such as newlyweds (Luo & Klohnen, 2005). In contrast, results presented here are representative of heterosexual couples living in Switzerland. Our results support ndings of Dyrenforth et al. (2010) as well as Barelds (2005), who also made use of nationally representative samples of couples. When examining very large data sets, similarity in personality is not systematically related to satisfaction. Third, we estimated different sets of Actor-Partner Interdependence Models to take account of the dyadic nature of the data. Several previous studies have failed to capture the interdependent data structure in their analyses, which renders interpretation of similarity effects in couples problematic (e.g., Gonzaga et al., 2007). Thus, our results demonstrate why it is necessary to apply adequate analytic techniques to dyadic data (e.g., as proposed by Kenny et al., 2006). Furthermore, a recurring issue in this area of research is the question of which similarity measure to use. Hence, another asset of our study is the adoption of a differentiated approach to similarity. We captured conceptually different estimates of personality similarity and fragmented our analyses through separate examinations of trait and prole (global, shape, mean, and variance) similarity indices as predictors of life satisfaction. Even though the similarity measures used in our study are conceptually different from one another, none of them had an effect on either partners life satisfaction. Thus, adopting a differentiated approach did not provide additional information that goes beyond the global similarity index. This supports the conclusion of Dyrenforth et al. (2010) that as long as main effects are controlled, ndings suggest no relation between personality similarity and life satisfaction regardless of the similarity index used. However, we do not rule out the possible advantage of a differentiated approach in general. It is essential to understand the conceptual difference of the applied measures and we therefore believe that future research could benet from this differentiated approach to analyze similarity. Alternative theoretical explanations for mixed ndings in past research can be ascribed to the assumption of general effects of similarity on well-being. Future research should rather investigate differential effects between couples: whereas being similar might be benecial for some couples, for others it may not be benecial as they are more satised with a partner who is dissimilar and thus complements them. Alternatively, underlying relationship structures or relationship maintenance processes might lead some people to be more satised with a similar partner, whereas others prefer a dissimilar partner. Moreover, future research should more thoroughly investigate the processes and mechanisms through which personality is linked to important life outcomes. Motivational aspects of personality for instance could explain associations between personality similarity on single traits and life satisfaction in couples and could serve as a potential mediator. Such analyses have been fairly neglected in couple research so far. Furthermore, although we did not nd evidence that similarity on the Big Five traits is related to life satisfaction in couples, similarity in other domains (within or outside personality) may very well be important for well-being. In this vein, previous research provides evidence for the relevance of similar attachment characteristics for marital sat-

isfaction in a sample of newlywed couples (Luo & Klohnen, 2005) or for the relevance of gender-related attitudes on dyadic adjustment (Aube & Koestner, 1995). Finally, one could argue that personality similarity may operate differently as a function of relationship stage. For instance, being similar to ones partner might be of greater relevance for ones well-being during acquaintanceship and mating, whereas complementing each other might be more benecial in a more advanced phase of the relationship. Because our sample comprised a broad range of couples in different relationship stages, we included relationship length as a potential moderator into our models. However, these additional analyses barely yielded evidence for alternative associations between personality similarity and satisfaction as a function of relationship stage. Besides these strengths, we should also consider the limitations of this study. First, similar to other studies using panel data, we were conned to using short scales for the analyses. To ensure a broad assessment of the construct, personality traits were assessed with very few items per trait with rather low intercorrelations. As a consequence, internal consistencies of the Big Five traits were rather low in this sample. However, short scales are known to have lower reliabilities and are widely used in large-scale panel studies due to time constraints. Another limitation of this study concerns life satisfaction as the primary outcome variable within a dyadic research question of romantic couples. Life satisfaction represents a cognitive evaluation of ones life in general (Diener, Suh, Lucas, & Smith, 1999; Schimmack, 2008) and is thus a more or less highly individual appraisal of ones life. At least in the context of romantic relationships, personality similarity may be more relevant for a dyadic well-being outcome such as relationship quality or relationship satisfaction. Unfortunately, however, the SHP does not assess relationship satisfaction. In sum, personality is highly relevant in explaining life satisfaction. Of remarkable interest, not only ones own personality but also the partners personality matters for each partners well-being. However, after controlling for these actor and partner effects, being similar to ones partner is not related to life satisfaction. Adopting a differentiated approach revealed similar results for all similarity indices. Acknowledgements This study has been realized using the data collected by the Swiss Household Panel (SHP), which is based at the Swiss Centre of Expertise in the Social Sciences FORS. The SHP is nanced by the Swiss National Science Foundation. This research was supported by the Swiss National Science Foundation Sinergia program Co-Development in Personality: Longitudinal Approaches to Personality Development in Dyads across the Life Span (CRSI11_130432/1). We thank Laura Wiles for proofreading. References
Anderson, C., Keltner, D., & John, O. P. (2003). Emotional convergence between people over time. Journal of Personality and Social Psychology, 84, 10541068. http://dx.doi.org/10.1037/0022-3514.84.5.1054. Antill, J. K. (1983). Sex role complementarity versus similarity in married couples. Journal of Personality and Social Psychology, 45, 145155. http://dx.doi.org/ 10.1037/0022-3514.45.1.145. Anusic, I., Lucas, R. E., & Donnellan, M. B. (2012). Cross-sectional age differences in personality: Evidence from nationally representative samples from Switzerland and the United States. Journal of Research in Personality, 46, 116120. http:// dx.doi.org/10.1016/j.jrp.2011.11.002. Argyle, M. (1999). Causes and correlates of happiness. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology (pp. 353373). New York: Russell Sage Foundation.

K. Furler et al. / Journal of Research in Personality 47 (2013) 369375 Arrindell, W. A., & Luteijn, F. (2000). Similarity between intimate partners for personality traits as related to individual levels of satisfaction with life. Personality and Individual Differences, 28, 629637. http://dx.doi.org/10.1016/ S0191-8869(99)00125-7. Aube, J., & Koestner, R. (1995). Gender characteristics and relationship adjustment: Another look at similaritycomplementarity hypotheses. Journal of Personality, 63, 879904. http://dx.doi.org/10.1111/j.1467-6494.1995.tb00319.x. Barelds, D. P. H. (2005). Self and partner personality in intimate relationships. European Journal of Personality, 19, 501518. http://dx.doi.org/10.1002/per.549. Botwin, M. D., Buss, D. M., & Shackelford, T. K. (1997). Personality and mate preferences: Five factors in mate selection and marital satisfaction. Journal of Personality, 65, 107136. http://dx.doi.org/10.1111/j.1467-6494.1997. tb00531.x. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Cronbach, L. J., & Gleser, G. C. (1953). Assessing similarity between proles. Psychological Bulletin, 50, 456473. http://dx.doi.org/10.1037/h0057173. Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. L. (1999). Subjective well-being: Three decades of progress. Psychological Bulletin, 125, 276302. http://dx.doi.org/ 10.1037/0033-2909.125.2.276. Dyrenforth, P. S., Kashy, D. A., Donnellan, M. B., & Lucas, R. E. (2010). Predicting relationship and life satisfaction from personality in nationally representative samples from three countries: The relative importance of actor, partner, and similarity effects. Journal of Personality and Social Psychology, 99, 690702. http://dx.doi.org/10.1037/A0020385. Edwards, J. R. (1994). Regression analysis as an alternative to difference scores. Journal of Management, 20, 683689. http://dx.doi.org/10.1016/01492063(94)90011-6. Edwards, J. R. (2001). Ten difference score myths. Organizational Research Methods, 4, 265287. http://dx.doi.org/10.1177/109442810143005. Epstein, E., & Guttman, R. (1985). Mate selection in man: Evidence, theory, and outcome. Social Biology, 31, 243278. http://dx.doi.org/10.1080/ 19485565.1984.9988579. Furr, R. M. (2008). A framework for prole similarity: Integrating similarity, normativeness, and distinctiveness. Journal of Personality, 76, 12671316. http://dx.doi.org/10.1111/J.1467-6494.2008.00521.x. Furr, R. M. (2010). The double-entry intraclass correlation as an index of prole similarity: Meaning, limitations, and alternatives. Journal of Personality Assessment, 92, 115. http://dx.doi.org/10.1080/00223890903379134. Gattis, K. S., Berns, S., Simpson, L. E., & Christensen, A. (2004). Birds of a feather or strange birds? Ties among personality dimensions, similarity, and marital quality. Journal of Family Psychology, 18, 564574. http://dx.doi.org/10.1037/ 0893-3200.18.4.564. Gaunt, R. (2006). Couple similarity and marital satisfaction: Are similar spouses happier? Journal of Personality, 74, 14011420. http://dx.doi.org/10.1111/ J.1467-6494.2006.00414.x. Gonzaga, G. C., Campos, B., & Bradbury, T. (2007). Similarity, convergence, and relationship satisfaction in dating and married couples. Journal of Personality and Social Psychology, 93, 3448. http://dx.doi.org/10.1037/0022-3514.93.1.34. Gonzaga, G. C., Carter, S., & Buckwalter, J. G. (2010). Assortative mating, convergence, and satisfaction in married couples. Personal Relationships, 17, 634644. http://dx.doi.org/10.1111/j.1475-6811.2010.01309.x.

375

Grifn, D., Murray, S., & Gonzalez, R. (1999). Difference score correlations in relationship research: A conceptual primer. Personal Relationships, 6, 505518. http://dx.doi.org/10.1111/j.1475-6811.1999.tb00206.x. Headey, B., Muffels, R., & Wagner, G. G. (2010). Long-running German panel survey shows that personal and economic choices, not just genes, matter for happiness. Proceedings of the National Academy of Sciences of the United States of America, 107, 1792217926. http://dx.doi.org/10.1073/Pnas.1008612107. Hoppmann, C. A., Gerstorf, D., Willis, S. L., & Schaie, K. W. (2011). Spousal interrelations in happiness in the Seattle Longitudinal Study: Considerable similarities in levels and change over time. Developmental Psychology, 47, 18. http://dx.doi.org/10.1037/a0020788. John, O. P., & Srivastava, S. (1999). The Big Five trait taxonomy: History, measurement, and theoretical perspectives. In L. A. Pervin & O. P. John (Eds.), Handbook of personality: Theory and research (pp. 102138). New York: Guilford Press. Kenny, D. A., & Acitelli, L. K. (1994). Measuring similarity in couples. Journal of Family Psychology, 8, 417431. http://dx.doi.org/10.1037/0893-3200.8.4.417. Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic data analysis. New York: Guilford Press. Lucas, R. E., & Donnellan, M. B. (2012). Estimating the reliability of single-item life satisfaction measures: Results from four national panel studies. Social Indicators Research, 105, 323331. http://dx.doi.org/10.1007/s11205-011-9783-z. Luo, S. H., & Klohnen, E. C. (2005). Assortative mating and marital quality in newlyweds: A couple-centered approach. Journal of Personality and Social Psychology, 88, 304326. http://dx.doi.org/10.1037/0022-3514.88.2.304. Rammstedt, B., & John, O. P. (2007). Measuring personality in one minute or less: A 10-item short version of the Big Five Inventory in English and German. Journal of Research in Personality, 41, 203212. http://dx.doi.org/10.1016/ J.Jrp.2006.02.001. Robins, R. W., Caspi, A., & Moftt, T. E. (2000). Two personalities, one relationship: Both partners personality traits shape the quality of their relationship. Journal of Personality and Social Psychology, 79, 251259. http://dx.doi.org/10.1037/ 0022-3514.79.2.251. Schimmack, U. (2008). The structure of subjective well-being. In M. Eid & R. J. Larsen (Eds.), The science of subjective well-being (pp. 97123). New York: Guilford Press. Steel, P., Schmidt, J., & Shultz, J. (2008). Rening the relationship between personality and subjective well-being. Psychological Bulletin, 134, 138161. http://dx.doi.org/10.1037/0033-2909.134.1.138. Vandenberg, S. G. (1972). Assortative mating, or who marries whom? Behavior Genetics, 2, 127158. http://dx.doi.org/10.1007/BF01065686. Voorpostel, M., Tillmann, R., Lebert, F., Weaver, B., Kuhn, U., Lipps, O., et al. (2010). Swiss Household Panel Userguide (19992009), Wave 11, October 2010. Lausanne: FORS. Watson, D., Klohnen, E. C., Casillas, A., Simms, E. N., Haig, J., & Berry, D. S. (2004). Match makers and deal breakers: Analyses of assortative mating in newlywed couples. Journal of Personality, 72, 10291068. http://dx.doi.org/10.1111/j.00223506.2004.00289.x.