The ideal qualities that people wanted in a partner remained fairly stable over the long term, but the importance of status and resources increased over time

Stability and Change of Individual Differences in Ideal Partner Preferences Over 13 Years. Julie C. Driebe et al. Personality and Social Psychology Bulletin, April 8, 2023. https://doi.org/10.1177/01461672231164757

Abstract: Ideal partner preferences for traits in a partner are said to be stable cognitive constructs. However, longitudinal studies investigating the same participants’ ideals repeatedly have so far been limited to relatively short retest intervals of a maximum of 3 years. Here, we investigate the stability and change of ideals across 13 years and participants’ insight into how ideals have changed. A total of 204 participants (M = 46.2 years, SD = 7.4, 104 women) reported their ideals at two time points. We found a mean rank-order stability of r = .42 and an overall profile stability of r = .73 (distinctive r = .53). Some ideals changed over time, for example, increased for status-resources in relation to age and parenthood. We found some but varying insight into how ideals had changed (mean r = .20). Results support the idea of ideals being stable cognitive constructs but suggest some variability related to the demands of different life stages.

Discussion

In this study, employing unique longitudinal data across 13 years, we investigated stability (i.e., retest and profile correlations) and change (i.e., mean-level changes) of ideal partner preferences, and whether individuals possess insight into how their preferences have changed (i.e., correlations of perceived changes with actual changes).

Stability and Change in Ideal Partner Preferences

Regarding our first hypothesis (H1), our results suggested considerable overall stability of participants’ ideals of r = .42, corresponding to a medium-sized to large effect (Cohen, 2013; Gignac & Szodorai, 2016). This stability is smaller than coefficients obtained after 5 months (Gerlach et al., 2019) yet roughly comparable to coefficients found after 3 years (Bleske-Rechek & Ryan, 2015). These results are in a range that has previously been reported for the rank-order stability of personality traits (around r = .60 for a retest interval of 6.7 years, Roberts & DelVecchio, 2000; r = .33 for an interval of 11 years in a more diverse sample, Atherton et al., 2022). However, when compared with a meta-analysis by Anusic and Schimmack (2016), our results are comparable with the retest correlation of affect and self-esteem found after 13 years in a group of 30-year-olds but smaller compared to the retest correlation of broad personality dimensions after 13 years in a group of 30-year-olds. Our results, together with the stability coefficients reported for partner preferences across shorter intervals (e.g., (Fletcher et al., 1999, 2000; Gerlach et al., 2019; Shackelford et al., 2005), are in line with the finding that the strongest declines in stability coefficients are found in the first years after assessment (Anusic & Schimmack, 2016; Costa et al., 2019). The fact that our retest correlations do not further decrease even over such a long timespan suggests that individual differences in ideal partner preferences contain a sizable trait component (Anusic & Schimmack, 2016). However, this stability seems to be more comparable to constructs such as self-esteem (as opposed to broader personality domains), which has been shown to be more susceptible to external influences.

Nonetheless, investigating participants’ profiles revealed that patterns of which traits individuals preferred more or less were surprisingly stable, with overall profile correlations exceeding r = .70. These profile stabilities were only slightly reduced when accounting for normative components (e.g., most people value warmth-trustworthiness more than status-resources) by employing distinct profile correlations. We take this high temporal consistency to suggest that—while individual dimensions may well be affected by external influences, resulting in only moderate stability—idiosyncratic patterns in what people value in a romantic partner may be a very stable individual difference characteristic, even when effects of what is normatively more versus less preferred are taken into account.

We then examined the relationship between parenthood and the stability of preferences. As put forward in hypothesis four (H4), we found that the stability of preferences for status-resources was lower in participants who became parents over the 13-year study period or who had intentions to become a parent at the time of the re-assessment, compared with participants without (the intention to have) children. We assumed that these shifts in partner preferences could be related to shifting priorities and efforts according to different life stages (cf. Del Giudice et al., 2016; Heckhausen et al., 2010, 2019), with parenthood potentially being of particular importance. As having a partner who is able to provide resources facilitates founding a family and raising children, (the decision to) becoming a parent may alter one’s preference for status-resources, explaining the lower stability. Yet, parenthood was also related to the stability of other preference dimensions, suggesting that the decision to become a parent has the potential to shake up how we picture our ideal partner more generally.

We also investigated mean-level changes in ideal partner preferences. In line with our second hypothesis (H2), participants placed higher importance on status-resources over time and this increasing preference was stronger for younger participants. Furthermore, although effects were small (Cohen’s d < 0.20), participants placed more importance on warmth-trustworthiness and less on vitality-attractiveness over time. Our third prediction (H3), an increase in family-orientation, was only partly supported: Over time, the preference for family-orientation only increased for younger individuals, yet compared with older participants, younger individuals already reported a higher preference for family-orientation at the initial assessment. Further exploration revealed that participants without children generally placed less importance on family-orientation, whereas the preference for family-orientation increased over time for those with children. While this might be a mere cohort effect, this finding could also be interpreted in light of age-graded opportunity structures and/or developmental deadlines (Wrosch & Heckhausen, 2005). For example, younger participants might picture themselves as likely to begin a family in the future, whereas older participants had already begun to ponder a possible life without children because they already considered themselves to be beyond the ideal age for having children, were pessimistic about finding a suitable partner for such an endeavor, or had already come to cherish a “childfree” lifestyle. Exploring mean-level changes in relation to the number of relationships participants had entered revealed that participants who entered more than one relationship over time reported an increased preference for warmth-trustworthiness and status-resources, whereas participants who entered only one relationship over time showed no significant increase in these preference dimensions. For the dimension status-resources, this may be due to having limited statistical power in these analyses, as participants who entered only one relationship descriptively showed an increased preference for status-resources. It can also be speculated that participants who entered more than one committed relationship after going through one or more break-ups may have realized that having a warm and trustworthy partner may be most vital for a relationship to last. Participants who entered only one committed relationship, however, may not have seen the necessity to update their preferences on this dimension.

In our study, we found considerable stability of preferences over 13 years. As such, our findings cannot explain the mixed findings in previous research on the link between preferences and relationship decisions. An alternative explanation for those mixed findings may be the relationship phase investigated (see Campbell & Stanton, 2014; Gerlach et al., 2019): studies that could not find a link between preferences and relationship decisions, for the most part, investigated the initial stage of getting to know each other (e.g., Eastwick & Finkel, 2008; Joel et al., 2017; Todd et al., 2007), whereas studies finding a relationship between preferences and relationship decisions often investigated already established relationships (e.g., Conroy-Beam & Buss, 2016; Park & MacDonald, 2019) or relationship formation (e.g., Campbell et al., 2016; Gerlach et al., 2019). However, our findings have important implications for future research investigating the association of ideal partner preferences and relationship decisions. First, the relatively high stability of preferences suggests that studies which investigate the association between preferences and relationship decisions do not necessarily need to constantly assess preferences over the investigated timespan. Instead, study designs in which preferences are initially assessed should suffice to investigate the link between these initial preferences and later relationship decisions. Second, a factor to consider when investigating longer timespans or populations more diverse as the typical student sample is that having children may alter preferences. Future studies investigating partner preferences may thus take into account the parenthood status of participants and the presence (vs. absence) of family formation goals more broadly.

Insight Into Preference Change

Over the 13-year study period, preferences for status-resources and warmth-trustworthiness increased and decreased for vitality-attractiveness—but were these changes mirrored in participants’ perceptions? Descriptively, participants perceived increases in their preference for warmth-trustworthiness and perceived decreases in their preference for vitality-attractiveness and status-resources. They also perceived increases in family orientation, intelligence, and humor and decreases in adventurousness-confidence and creativity. One interpretation of these perceived changes may be that participants believe to place more importance on dimensions that become more relevant with increasing age. For example, with increasing age, it may be adaptive to have a partner who is caring and oriented toward the family instead of a partner who is up for adventure and likes taking risks. Although objectively, having a high status and resources might also become more important when one gets older, participants may not perceive this change because they might have already achieved certain resources or status for themselves and may not realize that this increased standard of living has already shaped their preferences for a partner. Another possibility is that participants may answer in a socially desirable way: If participants do not want to admit that having a certain status and monetary resources is relevant to them, they might indicate that this dimension had become less relevant to them over time, while still ascribing considerable importance to it.

Interestingly, around 50% of participants did not report that they changed their ideals, except for family orientation, where only 37% of participants believed that their preferences had not changed. These patterns dovetail with results by Sprecher and colleagues (2018): Around half of their sample perceived not to have changed their ideals, except for “good parenting potential,” a variable close to family orientation. This perception of no change may mirror the previously found stability of ideal partner preferences or changes may have occurred at a younger age (Bleske-Rechek & Ryan, 2015).

When investigating whether perceptions correspond to actual changes, overall, we found a small positive effect. Yet, insight varied considerably between the different dimensions: Participants had the most insight into family orientation and the least for status-resources and intelligence. Contradicting our fifth hypothesis (insight into changes for status-resources, H5.1), participants believed to have decreased in their preference, when in fact they increased over time. One possibility is that participants may perceive themselves in a biased self-enhancing manner via a similar process to what Robins et al. (2005) suggested to be the case for perceived changes in personality. Yet, in line with the second part of this prediction (H5.2), participants showed some insight into changes in their preference for vitality-attractiveness, although the perception of change appears stronger than the actual change. Interestingly, age and sex were not related to participants’ insight.

The present results for perceptions are in line with previous research (Bleske-Rechek et al., 2009) that found participants to predict that they would value intrinsic characteristics (i.e., warmth-trustworthiness, family orientation) more and appearance (i.e., vitality-attractiveness) less over time, suggesting that participants may be more oriented toward committed relationships over time. At the same time, perceptions of change were somewhat exaggerated and for the most part only achieved modest accuracy (family orientation being a notable exception), showing that perceptions do not necessarily correspond to actual changes. These results highlight the necessity to conduct longitudinal studies when one is interested in preference change and underscore that intraindividual processes should not be investigated in cross-sectional data: Self-perceptions of change do not reflect actual changes accurately enough to allow them to be used as a substitute.

Strengths

The longitudinal design of this study, covering 13 years, makes it unique among studies on the stability and change of partner preferences, which have so far investigated much shorter time periods. Even over this long timespan, we managed to rerecruit a sizable proportion of the initial sample, and participant retention was better than expected over such a large time interval. For example, while in the current study we found a retention rate of 59% after 13 years, Gerlach et al. (2019) reported a retention rate of more than five months of 64%, whereas a study by Gustavson et al. (2012) covering a time span of 15 years reported a retention rate of 44%. A special feature of our sample is that it is a community sample not restricted to the typical student population. In particular, our sample spanned a wide age range, allowing us to investigate intraindividual stability and change of preferences across a period when participants were still single until much later in life when they may have found a partner with whom they then had to decide whether to have children or not. Investigating this life stage may be of particular interest since it does not only involve the time in which participants start having a family but also a time in which important career decisions take place. Finally, we used comprehensive measures of participants’ ideals at both assessments and complementary indices to investigate their stability and change.

Limitations and Future Directions

Although our community sample was arguably more diverse than the typical student sample, it was still highly educated and came from a Western background. The generalizability of our results may be limited because preferences and their importance could not only vary by education but also across different cultures. For example, in a study involving samples from Taiwan, Lam et al. (2016) uncovered preference attributes referring to the extended family previously overlooked in Western samples. Furthermore, education might be related to how much importance individuals ascribe to attributes conducive to a partner’s career advancement (e.g., successful, ambitious). Future studies should strive to recruit participants with more diverse educational backgrounds, ideally also from non-Western countries (Henrich et al., 2010).

Furthermore, although the large retest period is unique and showed that ideal partner preferences contain a sizable trait component, life events may still be associated with a change in preferences. The fact that we only had two assessments available precludes an in-depth analysis of further factors that might have driven preference change. Future research should include multiple assessments of preferences and important events (e.g., parenthood; entering Gerlach et al., 2019 or ending relationships; experiences of romantic rejection and acceptance Charlot et al., 2020). Additional factors influencing changes in partner preferences may be the increased occurrence of specific life events in a persons’ social environment. For example, the importance of having a partner with a high family orientation may increase when more and more people in one’s environment are trying to or are indeed having a child (Keim et al., 2009). Another possible change in partner preferences may be that people lower their expectations after a long period of time not being able to find a partner (Gerlach et al., 2019). For example, people lower their standards regarding a partner’s physical attractiveness. Finally, a recent study has found divorce to be associated with changes in self-esteem (Bleidorn et al., 2021). Similarly, relationship dissolution may be a life event associated with changes in preferences. For example, after a relationship dissolution fraught with conflict, individuals may increase their preference for having a kind, trustworthy partner because they recently got to know the disagreeable side of their ex-partner. Future research with multiple assessments should also include participants’ perception of change to investigate what drives the accuracy of preference change perceptions and whether the perception of change may be associated with future dating or relationship decisions.

Finally, we deviated from our preregistered analytic plan in three analytic decisions (see S4). Therefore, only our hypotheses and design can be regarded as preregistered. In particular, the diverging assessment of initial ideals between the two samples led to larger problems than anticipated, which led us to the decision to analyze both samples separately and interpret results based on the BSDS only. Unfortunately, this also lowered our sample size, hence the power of our study, which is especially relevant for the analyses comparing participants with and without children. Furthermore, as the instruction for rating partner preferences was not completely identical across T1 and T2, we also checked for measurement invariance across the two time points according to the procedure as suggested by Mackinnon et al. (2022). We found scalar invariance partly supported, suggesting that participants may have interpreted our response scale slightly differently at T1 and T2. We therefore recommend future studies to adopt the exact same wording of their instructions at all assessments.