Does Sexual Satisfaction Change With Relationship Duration?

Sexuality is an essential part of romantic relationships, and sexual satisfaction is closely associated with relationship stability and satisfaction (Byers, 2005; Laumann, Gagnon, Michael, & Michaels, 1994; Sprecher, 2002). Therefore, it is not surprising that sexual satisfaction has constantly attracted researchers’ attention over the years, although there is still a need both for theoretical models and for sound empirical analyses (Christopher & Sprecher, 2000; Sánchez-Fuentes, Santos-Iglesias, & Sierra, 2014). In particular, little is known about how and why sexual satisfaction changes with relationship duration. Sexual satisfaction is a complex construct, involving both physical pleasure and emotional satisfaction (Laumann et al., 1994), as well as a global subjective evaluation of an individual’s sex life (Lawrance & Byers, 1995).

One established theoretical model used to explain sexual satisfaction is the Interpersonal Exchange Model of Sexual Satisfaction (IEMSS) developed by Lawrance and Byers (1992, 1995), which takes into account the levels of costs and rewards exchanged in a sexual relationship. According to the IEMSS, sexual satisfaction will be high if the accumulated levels of rewards exceed those of costs, if the level of rewards and the reward-to-cost ratio are higher than expected, and if equality is perceived regarding one’s own and the partner’s levels of rewards and costs. A change in sexual satisfaction over time could then be the consequence of sexual exchanges becoming more or less favorable with relationship duration (Byers & Macneil, 2006; Lawrance & Byers, 1995).

Rather than providing arguments for exchanges becoming systematically more or less favorable with relationship duration, this model serves as a framework for the analysis, as it defines the factors determining sexual satisfaction. For instance, it could be that costs accumulate at a constant rate with relationship duration, whereas rewards accumulate at a diminishing rate. Then, according to the IEMSS, a decline in sexual satisfaction would be the consequence. Similarly, satisfaction would decrease if the perceived equality of both partners’ exchanges declined. In fact, Sprecher (2001) reported that, for a sample of romantic couples, perceived sexual equity (which is a concept closely related to equality) decreased over time for both women and men.

One reason for this phenomenon, which was not mentioned by Sprecher (2001) could be that men’s and women’s sexual needs diverge over time. Smith et al. (2011) found that middle-aged men with relationship durations of 6 or more years were less satisfied than other men, and interpreted this result as a sign that these men’s needs are not fulfilled, be it because of competing demands on couples’ time or because women desire sex more often in the earlier periods of a relationship than they do in later periods. Baumeister, Catanese, and Vohs (2001) mentioned a large number of empirical studies which provide evidence of a constantly higher sex drive of men, but they also pointed out the argumentation of Baumeister and Bratslavsky (1999): in the beginning of a relationship, the difference between men’s and women’s sex drive may diminish. Also, Klusmann (2002) and Murray and Milhausen (2012) reported that women’s sexual desire decreased with relationship duration, whereas men’s sexual desire did not. Research has shown that sexual desire and congruency of partners’ level of desire are both associated with sexual satisfaction (Davies, Katz, & Jackson, 1999). Hence, if a woman’s sexual desire diminishes with relationship duration while her male partner’s desire remains constant, both partners will perceive this mismatch and, as a consequence, sexual satisfaction may decline.

Many empirical studies on changes in sexual satisfaction with relationship duration have considered two aspects: On the one hand, a negative effect is assumed, which Liu (2003) explained as diminishing marginal utility of repeated intercourse with the same partner, implying that the utility of further intercourse occurrences decreases relative to the accumulated experience of intercourse within the couple. This effect has been referred to as the habituation effect (e.g., Call, Sprecher, & Schwartz, 1995) or routine (Klusmann, 2002), but as Klusmann (2002) pointed out, it is not clear why routine should have a negative effect in the case of sexual satisfaction, whereas in many other domains of life it does not play a similar role. One explanation could be a male preference for variety (Little, DeBruine, & Jones, 2014; Liu, 2003). An alternative explanation for this habituation effect could be deduced from Baumeister and Bratslavsky (1999) who argued that passion arises only in the case of increasing intimacy, so that intimacy and passion show typical patterns over the course of a relationship: intimacy in a relationship rises rapidly in the initial periods, but thereafter with a decreasing pace and, therefore, passion is high in the beginning of a relationship, but decreases later on. Given that passion is an essential element for high sexual satisfaction (Rubin & Campbell, 2012), sexual satisfaction will be high in the beginning of a relationship and steadily decline later on with subsiding passion.

On the other hand, a positive learning effect is assumed (Laumann et al., 1994), which Liu (2003) described as an investment in partner-specific skills in committed relationships. Partners become familiar with each other’s sexual preferences and thus improve their skills in order to pleasure each other. In contrast to the first mechanism, the second implies an increase in sexual satisfaction with relationship duration. As both effects may be at work, the net effect on sexual satisfaction will depend on effect sizes. Liu (2003) identified four potential patterns of change in sexual satisfaction: a constant decline in case of the habituation effect only, a constant increase in case of the learning effect only, as well as a decline with decreasing rates if the habituation effect dominates, and an increase with decreasing rates if the learning effect dominates, in case of both effects. Even more patterns could be possible (e.g., if the habituation effect sets in later than the learning effect or the learning effect dominates only until a certain point in the relationship) which would lead to a combination of two of the patterns mentioned by Liu (2003).

Empirical evidence regarding the effect of relationship duration on sexual satisfaction is mixed. A negative association was found by Klusmann (2002) in a cross-sectional analysis of a sample of German university students, but due to the particular and presumably rather homogenous sample used these results cannot be generalized. In addition, as mean relationship duration was only approximately 3 years, conclusions about the development over longer time periods cannot be drawn. Using representative cross-sectional data from the U.S. National Health and Social Life Survey (NHSLS), Liu (2003) tested for constant and changing growth rates over time and found a constant decline in sexual satisfaction with marital duration. Also, Edwards and Booth (1994) reported a significant decline in happiness regarding sex with marital duration from their longitudinal analysis based on a national sample of married individuals in the U.S. Based on a sample of 120 newlyweds, McNulty and Widman (2013) studied sexual satisfaction 5 years into marriage and reported a negative slope of the trajectory of sexual satisfaction as well, which did not differ across men and women, but was steeper for spouses with higher levels of sexual narcissism.

A second group of studies has shown differences between men and women regarding patterns of change in sexual satisfaction. Richters, Grulich, Visser, Smith, and Rissel (2003) found such differences using cross-sectional data from the Australian Study of Health and Relationships. For women, they reported a steady decline of both physical and emotional aspects of sexual satisfaction over the course of a relationship, whereas for men they found an inversely U-shaped pattern. This finding stands in contrast to the results reported by Heiman et al. (2011), who analyzed cross-sectional data of middle-aged and older couples in Brazil, Germany, Japan, Spain, and the U.S., finding a significantly positive linear effect of relationship duration on the women’s (but not men’s) probability of being satisfied with their sexual relationship. But, in contrast to the afore mentioned studies, their sample included mostly long-term relationships (only 5 % with 7 years in duration or less) and, even more importantly, due to limited data availability, the analysis used an inaccurate duration variable so that measurement errors, in particular for shorter durations, were likely.

In addition to studies finding a negative effect and those finding differing effects for men and women, a third group of studies did not find significant effects of relationship duration on sexual satisfaction at all. Carpenter, Nathanson, and Kim (2009) did not find any significant effect of relationship duration on physical or emotional sexual satisfaction, using NHSLS data as well. The differing results might be due to the fact that the study was based on a different subsample of the NHSLS than Liu’s (2003) study. Cheung et al. (2008) investigated a large (but also only cross-sectional) sample of Chinese couples without finding a significant effect of marriage duration on sexual satisfaction. Pedersen and Blekesaune (2003) compared the sexual satisfaction of married, cohabiting, and committedly dating young couples in Norway, and did not find a significant effect of relationship duration. It seems, however, that their analysis did not fully exploit the longitudinal structure of the dataset: apparently, observations of the different waves were pooled and standard cross-sectional regression models were applied. With this between-person approach, unobserved heterogeneity between participants is likely to bias estimations as is the case when cross-sectional data are analyzed. In contrast, panel data models, especially fixed effects models, utilize the longitudinal data structure to its full capacity. In fixed effects models, estimations are based on within-person changes over time, instead of comparison between persons, so that time-constant unobserved heterogeneity between participants does not bias the estimations. Regarding the duration effect, Pedersen and Blekesaune differentiated only between relationships of more or less than 2 years, which seems reasonable for their sample of young adults (aged 20–26), but does not allow for generalized conclusions. A similar classification was applied by Neto and Pinto (2013), who distinguished between participants who had been more or less than five years in the same relationship without finding significant effects. Byers and Macneil (2006) did not find a significant change in sexual satisfaction over time either although their findings indicated that consistent with the IEMSS, participants whose sexual exchanges became less favorable reported lower sexual satisfaction over time.

The purpose of the present study was to investigate whether sexual satisfaction changes over the course of a relationship. From the theoretical arguments presented above, a positive learning effect, a negative habituation (or diminishing marginal utility) effect, or combinations of both effects could be expected. Thus, rather than by a directed hypothesis, our analysis was guided by the research question of which patterns of change could be found in the data. The purpose of our study was to overcome the main shortcomings of existing empirical studies, which apply cross-sectional analysis prone to bias due to unobserved heterogeneity. For this, we have used a large, randomly sampled German panel study and applied fixed effects regression, which bases estimations on within-person changes over time.

Participants

Our empirical analysis was based on data of 2,814 individuals (reportedly with a partner) who participated in Waves 2–4 of the German Family Panel study (pairfam) Release 4.0 (Nauck, Brüderl, Huinink, & Walper, 2013), a nationally representative longitudinal study with a focus on intimate relationships and family relations. The panel study began in 2008 with a random sample of 12,000 participants of three birth cohorts: 1971–1973, 1981–1983, and 1991–1993. For the first panel wave, addresses were randomly drawn from the local population registers of 343 randomly selected German municipalities. In each wave, participants received a letter signed by the head of project management and were then contacted by the interviewers. In every wave, each participant received a cash incentive of 10 Euro for participating in the interview. Participants were followed over time as long as they did not move abroad and did not refuse to be interviewed. The annual survey included both a personal interview of about 1 h conducted by a professional interviewer (CAPI) and a self-administered questionnaire for intimate questions (CASI), which had to be completed during the interview time using the official survey laptop. All questions about sexuality were posed in the CASI section of the survey and were asked annually. Some questions, however, were introduced starting with the second wave. For additional information on the panel study, see Huinink et al. (2011).

We limited the sample to heterosexual participants who had had the same partner for at least two panel waves. The youngest cohort was excluded as we expected adolescent sexuality and relationships to differ from sexuality in the adult age groups. In addition, cases with long relationship durations were rare in this cohort, which could lead to a selectivity bias in the estimation of duration effects. This left us with a data set of 3604 participants. After removing cases with missing data or inconsistent information on relationships, we retained our analytical sample with 2814 participants and 7385 observations across all three waves.^{Footnote 1}

Measures

Table 1 shows descriptive information for the main variables used. Note that all information used for this analysis, including information about the relationship and the partner’s demographics, was provided by the participant. The first column of the table shows mean values with standard deviation in brackets for metric variables and percentage of observations for dichotomous variables based on all three waves. The second column applies only to dichotomous variables, giving the share of participants who were in the respective group in at least one of the three waves. The last column indicates the percentage of participants with variation in the variable between waves. For instance, in 66.0 % of all observations across the three waves, participants were married (see first column). The number of participants who were married in at least one of the three waves is 68.3 % (second column). The share of participants who got married between waves is 7.4 (column 3). Note that in the first column the share of total observations is given, whereas in the second and third column, percentages refer to the number of participants.

Sexual satisfaction was measured by the question “How satisfied are you with your sex life?” Participants rated their satisfaction on a scale that ranged from 0 (very dissatisfied) to 10 (very satisfied). Due to limited resources, only this single-item measure was included in the questionnaire, but for our research question, the measurement quality of this construct is sufficient (Mark, Herbenick, Fortenberry, Sanders, & Reece, 2014). A drawback of the wording of this question is that it does not refer to a specific time span or to sexuality with the participant’s partner. But the question preceding the satisfaction item, “How often have you had sexual intercourse on average during the past 3 months with your partner?”, might provide context for the question on sexual satisfaction.

Relationship duration was calculated as the number of months since the beginning of the relationship (without taking into account disruptions in the relationship history), irrespective of cohabitation or marriage, which we controlled for separately. Duration ranged from less than 1 month to 25 years, with a median duration of 9 years. We built two 6-month intervals for the first year, 1-year intervals for the following 5 years, and 2-year intervals for the following 10 years. Durations of 16 years and more were not further differentiated. By using this set of binary variables instead of metric variables with quadratic and cubic terms, we were able to capture also small non-linear effects without imposing a specific functional form. Smaller intervals were chosen in the initial phases of the relationship as we expected more rapid changes in the beginning of a relationship than after several years. Additional analyses with different time splits (not shown) did not yield substantially different results.

The couple’s frequency of intercourse in the 3 months preceding the interview was included as a control variable. As the direction of causality between sexual frequency and satisfaction is not clear, and a reciprocal relationship is likely (Yucel & Gassanov, 2010), we analyzed the model both with intercourse frequency and without. The duration effect without controlling for intercourse frequency was the total effect of relationship duration on sexual satisfaction, including changes caused by decreased sexual frequency, whereas the effect in the model controlling for intercourse frequency was the more conservative estimation, measuring changes in satisfaction for constant levels of sexual frequency.

A number of variables capturing characteristics of the relationship were included. First, marital status and cohabitation were controlled for. Second, we tested for potential effects of parenthood, including the age of the youngest child in the household and dummy variables indicating pregnancy, breastfeeding, and if the couple had tried to conceive since the last interview in order to account for the close link between sexuality and fertility. A third set of variables pertained to relationship quality: we measured couples’ conflict style using a 12-item scale containing participants’ ratings of their own and their partner’s conflict styles in the relationship with the same six items (e.g., frequency of yelling at the partner, remaining silent, or insulting the partner when having a disagreement). Further, we measured intimacy in couple communication using two items with ratings on how often participants shared their thinking and secrets/private feelings with their partner (adapted from the Intimacy scale of the Network of Relationships Inventory (Furman & Burmester, 1985). We also controlled for a distance of more than 45 min traveling time between partners’ homes to account for partner’s availability, as limited availability has been found to negatively affect sexual satisfaction, at least in the case of shift workers (White & Keith, 1990).

In addition, we included the month of the interview, panel wave, and both partners’ labor force status and age. We controlled for participants’ health status, expecting a positive effect of good health (Heiman et al., 2011). Information on the partner’s health status was not available in the data. In our sample, participants were rather young, so that health problems may not yet have influenced sexual function to a large extent. In the fixed effects model, it was neither necessary nor possible to include time-constant variables, such as religion or sexual experiences prior to the current relationship, as the estimation was based on the within-person changes of independent and dependent variables over time.^{Footnote 2}

Analytic Strategy

We applied a fixed effects model to analyze changes in participants’ sexual satisfaction throughout the duration of their relationship. Fixed effects estimation only uses intra-individual changes over time, so that we were able to examine within-person associations between sexual satisfaction and relationship duration. That is, instead of comparing individuals with different relationship durations in order to estimate the effect of relationship duration on sexual satisfaction (as would be the cross-sectional approach), estimation was based on participants’ individual changes throughout a relationship and sexual satisfaction over the three panel waves. One of the main advantages of fixed effects regression models is that unobserved heterogeneity caused by time-constant variables, such as gender or childhood experiences, does not bias the estimations (for details of fixed effects estimation, see Brüderl & Ludwig, 2014; Wooldridge, 2010). The regression models were estimated with cluster-robust standard errors (modified Huber-White sandwich estimators). Some of the studies presented above found that the effects of relationship duration (and of some control variables) on sexual satisfaction differ for men and women (see Heiman et al., 2011). For this reason, we first estimated models with all variables interacted with the participant’s gender to allow the coefficients to differ by gender. In a second step, we eliminated all interactions with gender which did not prove to be significant.

In Table 2, we have listed the results of our fixed effects estimations. The interaction with gender did not prove to be significant for any of the variables, so that the resulting models were estimated without any interactions. In the first model, we controlled for age and wave in addition to relationship duration; in the second model, all variables, except sex frequency, were controlled for and, in the third model, the whole set of variables were used. The coefficients have been interpreted as changes in sexual satisfaction due to changes in the respective independent variable over time, holding all other variables constant.

Regarding the development of sexual satisfaction with relationship duration, Models 1 and 2 showed two effects. We found a significant and positive development in the first year of the relationship. That is, participants reported higher satisfaction with their sex life in the second half of the first year of their relationship than in the first 6 months. However, after the first year, sexual satisfaction declined steadily over time. In Model 1, which controls only for age and wave, the negative development of sexual satisfaction with relationship duration was even more pronounced than in Model 2, indicating that the effect was, in part, mediated by the variables controlled for in Model 2. Here, where all variables except sex frequency have been controlled for, in the last duration split (relationships with 16 years duration or more), sexual satisfaction was approximately 2 points lower than at the beginning of the relationship, and almost 3 points lower than in the second half of the first year (on an 11-point scale). The same pattern was visible in Model 3, which measured the development when also controlling for the frequency of intercourse. As in the other models, the positive effect in the second half of the first year was significant. The decrease in satisfaction at later stages of the relationship, however, was significant only when defining the second half-year of the relationship as the reference category (the corresponding model is not shown). For frequency of intercourse, we found the expected significant and positive effect.

We did not observe any significant effects of cohabitation or marriage on sexual satisfaction.^{Footnote 3} We found significant and positive effects for expecting a child, as well as if participants had tried to conceive in the period prior to the interview. On the other hand, breastfeeding mothers and their partners did not report significantly higher or lower levels of sexual satisfaction.

Regarding parenthood, this model did not reveal a consistent significant influence. The coefficients showed a U-shaped pattern, but with only the largest coefficient (having a youngest child aged 3–6) being significant. As the significance tests of the child variables refer to the comparison of each dummy variable to the reference group, we additionally tested the whole set of dummies together for significance using a Wald test, which did not reveal significant differences between the child variables, test statistics for Model 2: F(7, 2813) = 1.54, p = .15. As children might have an indirect effect on sexual satisfaction mediated by relationship quality, partners’ living distance, labor force status, and health, we also estimated a model without these variables (not shown). In this model, the U-shaped pattern persisted, and the coefficient for having a youngest child between 1 and 3 years proved to be significant as well, as did the test for the whole set of dummy variables. In sum, these results indicate that sexual satisfaction decreased in the first years after the birth of a child and increased again once children grew older, but that part of this effect was mediated by relationship quality, partners’ living distance, labor force status, and health.

Relationship quality, measured by negative communication styles and intimacy in couple communication, was found to play a significant role in sexual satisfaction. In line with our expectations, a rise of intimacy in communication had a significant positive influence, and a worsening of the conflict style a significant negative influence in both models. Effect sizes were not reduced largely when controlling for frequency of intercourse, implying that relationship quality directly influences the quality of intercourse rather than affecting sexual frequency.

Age was not found to affect sexual satisfaction to a significant extent.^{Footnote 4} Probably this was due to the high correlation between within-changes of the age variables and within-changes of the wave variables. In addition, our sample included only participants who were 41 and younger, so that physical limitations associated with age might not have played a significant role yet. Better subjective health was associated with higher sexual satisfaction, which was in line with our expectations.

The remaining control variables did not exhibit significant effects. Neither a distance of more than 45 min driving time between partners’ homes, nor both partners’ labor force status, interview month, or year affected participants’ ratings of sexual satisfaction.

The present study investigated the development of sexual satisfaction over the course of the relationship based on data from a large, randomly sampled German panel survey (pairfam). We analyzed persons aged 25–41 in heterosexual relationships, measuring how relationship duration as well as changes in relationship characteristics, such as marital status, relationship quality, parenthood, and personal characteristics such as age and health, affect sexual satisfaction. We found a peak in sexual satisfaction in the second half of the first year of the relationship, followed by a significant decline from the second year on. This result was in contrast to prior research, which has found only linear (negative or positive) trends over relationship duration and has not investigated potential non-linear effects in more detail. When controlling for frequency of intercourse, the positive effect persisted, but the negative effect was significant only in contrast to the second half of the first year. This implies that a large part of the negative effect was mediated by intercourse frequency. Differences between men and women could not be detected in our model.

The increase in sexual satisfaction in the beginning of a relationship could be due to a learning process, or to use Liu’s (2003) wording, the partners’ investment in partner-specific sexual skills that increase the physical pleasure of sexual activities. For the subsequent decline, we see two possible explanations: on the one hand, changes in sexual desire could lead to a mismatch between the man and the woman, as the diverging patterns of male and female sexual desire over the course of the relationship are well documented in the literature. On the other hand, as sexual satisfaction is linked to passion, its development may follow the pattern delineated for passion by Baumeister and Bratslavsky (1999), i.e., high levels in the beginning of a relationship followed by a decline for both sexes. As we did not have data about sexual desire or desired frequency of sex, however, we were not able to test for these mechanisms. The second argumentation, referring to intimacy and passion (Baumeister & Bratslavsky, 1999), was further supported by the positive coefficient of the intimacy variable in couple communication, which shows that intimacy matters for sexual satisfaction. This variable captures only one aspect of intimacy (communication), but not intimacy as a whole as Baumeister and Bratslavsky (1999) have conceptualized it. Hence, although the relationship duration effect persists, also when controlling for intimacy in couple communication, the effect could be caused by further (unobserved) aspects of intimacy.

Several of the control variables yielded notable results as well. We did not find that having children played a major role in a couple’s sexual satisfaction, which is remarkable as research has shown that sexual frequency is heavily influenced by the existence and age of children (Call et al., 1995; Schröder & Schmiedeberg, 2015). Our analyses revealed a small U-shaped effect on sexual satisfaction which was, in part, mediated by relationship characteristics, indicating that sexual satisfaction was lowest when the youngest child was between 3 and 6 years of age. Similarly, in contrast to past research (Laumann et al., 1994), we did not find significant differences between LAT, cohabiting, and married couples.

We see a number of limitations to our study. First, to analyze the development of sexual satisfaction over time in more detail, more information about the participant’s sex life or aspects of sexual satisfaction would be required. The German Family Panel study has the advantage of a broad range of questions over various domains, of which we made use when controlling the regression models. But on the other hand, the number of sex-related questions was limited. A more detailed analysis would consider the development of both partners’ sexual skills and sex drive, as well as the level of passion in the sexual relationship over time as potential drivers of sexual satisfaction.

Second, due to the age groups available in our data, we were not able to make any statements concerning sexual satisfaction at higher ages, although it seems reasonable that sexuality is affected by phenomena such as menopause or erectile dysfunction. Similarly, sexual satisfaction of adolescents and young adults in their first sexual experiences was beyond the scope of our study. It would be interesting, nevertheless, to replicate our analysis with samples of older or younger individuals. Further investigations should consider data from other countries as well as cultural norms may differ between countries to an extent that our results might not be generalizable to different cultural contexts. Furthermore, a topic we did not analyze exhaustively was the link between fertility and sexual satisfaction. The significant and positive effect of concrete fertility plans in our estimations implies that sexuality might be more satisfying when it is perceived in its biological function as opposed to simply a means for physical pleasure.

In conclusion, applying sound empirical methods and using a large, nationwide randomly drawn sample of young and middle-aged adults in Germany, our study showed that in the long run, sexual satisfaction diminishes over the course of a relationship, whereas in the initial stages positive effects of relationship duration prevail. Although we presented theoretical explanations for these effects, our investigation was not targeted at testing for the underlying mechanisms. Hence, further research is necessary in order to shed light on the different aspects of sexuality that drive sexual satisfaction.