Comparing same- and different-sex relationship dynamics: Experiences of young adults in Taiwan

Abstract

BACKGROUND

Few studies of same-sex relationships are able to capture the dynamics of these relationships from formation to dissolution, and even fewer provide evidence on these dynamics in a non-Western context.

OBJECTIVE

Using retrospective relationship history data collected from a nationally representative sample of young adults, this study compares the processes of forming and terminating relationships between same- and different-sex couples in Taiwan, an Asian society featuring both strong parental influences on children’s mate selection and an ongoing legislative effort to legalize same-sex marriage.

RESULTS

Results from event-history models show that factors associated with relationship formation and dissolution are largely similar for same- and different-sex unions and that same-sex relationships do not have higher dissolution rates. Nevertheless, premarital coresidence with parents, which is likely to amplify parental influences on children’s mate selection, deters the entry into and accelerates the dissolution of same-sex relationships more than it does different-sex relationships. Moreover, same-sex relationships are more heterogamous in family economic background, but more homogamous in age and education level, than different-sex ones.

CONTRIBUTION

This study is among the first to provide evidence on the dynamics of same- and different-sex relationships in a non-Western context. Aside from a few differences between same- and different-sex relationships related to parental influences, our study provides strong evidence that same- and different-sex couples experience intimacies in similar ways—even in a relatively conservative cultural context like Taiwan.

1. Introduction

Demographers have long been interested in the diversity of family forms (Biblarz and Savci 2010; Cherlin 2004). With the increasing visibility of same-sex unions, an emerging literature has been focusing on the similarities and differences between same- and different-sex unions (Kurdek 2006; Moore and Stambolis-Ruhstorfer 2013; Peplau and Fingerhut 2007). Some researchers show that same-sex couples experience a level of stability similar to that of different-sex couples in the United States (Manning, Brown, and Stykes 2016), leading to the argument that close dyadic relationships work in similar ways, irrespective of sexual orientation (Gates 2015; Kurdek 2006). Others, by contrast, reveal significant differences between same- and different-sex couples. In some European countries, for example, same-sex couples have higher dissolution rates than different-sex ones (Andersson et al. 2006; Lau 2012; Wiik, Seierstad, and Noack 2014). In the United States, the partners of same-sex unions resemble each other less than those of different-sex unions do (Schwartz and Graf 2009). Adding to the debate about similarities between same- and different-sex unions, research further shows that the extent of similarity may depend on social and cultural contexts. Living in U.S. states more supportive of same-sex relationships, for example, is associated with higher levels of formation and lower levels of dissolution of such relationships (Manning, Brown, and Stykes 2016; Prince, Joyner, and Manning 2017).

Despite the potential importance of social and cultural contexts, previous research on same-sex relationships focuses almost entirely on Western countries. The lack of studies of such relationships elsewhere, especially in societies where relatively traditional family practices are still prevalent, constitutes a major void in this otherwise vibrant and growing literature. In this study, we examine the extent to which the relationship dynamics of same-sex unions are similar to those of different-sex ones in Taiwan. Using retrospective relationship history data collected from a nationally representative sample of young adults, we compare the paces of forming and terminating relationships between same- and different-sex couples. We specifically focus on relationship experiences from mid-adolescence to early adulthood, a period during which individuals may begin to express their sexual orientation through relationship formation. Relationships during this period are important because they exert significant influences on future individual development and union formation (Collins, Welsh, and Furman 2009; Sassler 2010). To provide a comprehensive understanding of how the dynamics of same- and different-sex relationships are similar or different, we further ask whether same- and different-sex unions differ in how much the two partners resemble each other in sociodemographic characteristics, and whether this resemblance contributes to relationship stability to the same extent in the two types of unions.

Not only is our study the first to provide evidence about same-sex relationships in a non-Western context, but our use of detailed relationship history data to examine relationship dynamics also improves upon existing research in several ways. First, despite a few studies comparing the levels of stability between same-sex and different-sex unions (Lau 2012; Manning, Brown, and Stykes 2016), a comprehensive analysis of same-sex relationship dynamics, from their formation to dissolution, remains rare (Bennett 2017). By examining factors associated with the formation and dissolution of same-sex relationships, as well as comparing partner similarities between same- and different-sex unions, our study considerably enhances our understanding of same-sex relationship dynamics. Second, prior research on the stability of same-sex relationships commonly observes same-sex couples who have already formed their relationships at the time of interview, failing to assess the relationship stability from the beginning of the union (i.e., left-censoring) (e.g., Manning et al., 2016). Our study, in contrast, can overcome this shortcoming by analyzing individuals’ complete relationship histories through early adulthood. Third, most existing studies use data from same-sex cohabitating unions (Jepsen and Jepsen 2002; Lau 2012; Manning, Brown, and Stykes 2016) or married couples/registered partnerships (Andersson et al. 2006; Noack, Seierstad, and Weedon-fekĵr 2005; Wiik, Seierstad, and Noack 2014). Such research faces the potential problem of differential selectivity of same- and different-sex couples into marriage or cohabitation. The inclusion of all intimate relationships in this study helps avoid this problem.

Beyond enabling our uniquely advantageous research design, the Taiwanese case is worthy of attention in its own right. Like others in East Asia, Taiwan exemplifies societies with both singles’ prolonged coresidence with parents and rapid declines in marriage and fertility rates (Raymo et al. 2015; Yu and Kuo 2016). With marriage and fertility rates on the decline, the proportion of young adults living with parents has increased, and the transition into adulthood has been further delayed over time (Huang 2013; Raymo et al. 2015). The high level of parent-child coresidence, in conjunction with the Confucian cultural tradition that emphasizes children’s obedience and conformity (Smits and Park 2009), may enable Taiwanese parents to exert especially strong influences on children’s relationship formation and stability. Our analysis of Taiwan can therefore shed light on same-sex relationship dynamics for similar cultural contexts, in which traditional family values and parental control over unmarried children remain prevalent.

2. Research on the dynamics of same-sex unions

Intimate partnering is one of the most important life course events during young adulthood. Romantic relationships experienced in adolescence and early adulthood, in particular, are thought to shape future development of personal identities and peer relations (Collins, Welsh, and Furman 2009; Crosnoe and Johnson 2011; Montgomery 2005), in addition to being highly relevant to the formation and quality of intimate relationships in later adulthood (Meier and Allen 2009; Raley, Crissey, and Muller 2007; Sassler 2010). With the increasing acceptance of same-sex unions in Western countries (Bennett 2017; Sassler 2010), research on relationship experiences has expanded to include same-sex relationships in recent years. Most studies examining same-sex relationship dynamics center on union stability and its determinants, with explicit comparisons to different-sex unions. The argument for comparable dissolution patterns between same- and different-sex unions generally highlights that all intimate unions share similar dynamics (Gates 2015; Kurdek 2006). Sexual orientation therefore should not affect how couples relate to each other and decide whether to continue the union. At the same time, the minority stress perspective, which focuses on stressors that are uniquely associated with being members of socially stigmatized groups (Frost and Gola 2015; Meyer 2003), provides a contradictory account. Specifically, this perspective contends that same-sex couples face more barriers and challenges in forming and maintaining high-quality relationships than do different-sex couples, due to hostile and stressful social environments towards sexual minorities (Cao et al. 2017). More than different-sex couples, same-sex couples face stressors deriving from such sources as intrapersonal appraisals (e.g., self-concealment or self-stigma), interpersonal events (e.g., opposition of family and friends), and structural conditions (e.g., norms and policies) (Hatzenbuehler et al. 2014; Leblanc, Frost, and Wight 2015; Mohr and Daly 2008). These stressors may be interactive or additive in how they influence same-sex people’s relationship functioning and well-being.

To date, about 20 studies have compared same- and different-sex relationship stability, but their results are mixed and vary by context (Joyner, Manning, and Bogle 2017). In most European studies, same-sex couples are shown to be less stable than their different-sex counterparts. In Norway and Sweden, for example, registered same-sex partners have a higher divorce risk than different-sex married couples, even after taking relevant socioeconomic characteristics into account (Andersson et al. 2006; Noack, Seierstad, and Weedon-fekær 2005; Wiik, Seierstad, and Noack 2014). Cohabiting same-sex couples in Britain also experience a higher dissolution rate than different-sex couples who are cohabiting or married (Lau 2012). Conversely, using cohabitation samples in the United States, Manning and colleagues (2016) demonstrate similar levels of stability between same- and different-sex cohabiting couples. As Joyner and colleagues (2017) argue, the mixed results may be related to different samples and reference groups selected across different studies. Most prior research compares cohabiting or registered same-sex couples with married/cohabiting different-sex couples, with only one or two exceptions. Considering the different cohabitation rates across countries, as well as between same-sex and different-sex couples (Carpenter and Gates 2008), it is possible that cohabitating couples are selected differently based on sexual orientation.

Although much research on same-sex relationship dynamics centers on dissolution rates, the formation of an intimate relationship, a well-studied topic in the literature concerning heterosexual unions, constitutes another key part of relationship dynamics. Because studies of same-sex unions rarely focus on their formation, we know relatively little about whether the factors that promote or impede the formation of different-sex relationships are the same as those for same-sex ones (Prince, Joyner, and Manning 2017). As involvement in intimate relationships is important to the health and well-being of both same- and different-sex oriented people (Liu, Reczek, and Brown 2013; Sassler 2010), it is critical to pay attention to the formation of same-sex relationships as well.

Forming a same-sex relationship is not just an individual mate-searching behavior; rather, it is tightly linked to individuals’ family contexts. Previous research shows that the family of origin continues to influence same-sex union formation even in the West. Evidence from Britain, for example, shows that moving away from parents for education and employment leads young adults to begin their search for same-sex partners (Strohm 2010). A study in Denmark finds that individuals who grew up in non-intact families, in which parents are less available to monitor children’s mate selection, are more likely to enter same-sex marriages (Frisch and Hviid 2006). Among sexual-minority individuals in the United States, those who have come out to their parents are more likely to form a same-sex union than those who have not (Prince, Joyner, and Manning 2017). Despite the argument for the importance of family influences, to our knowledge, no existing research has explicitly compared parental influences on the formation rates of same- and different-sex relationships.

Beyond relationship formation, prior research also draws attention to how likely same- and different-sex unions are to be formed by people with similar traits. Researchers have proposed two mechanisms for the argument that homogamy is less likely among same-sex unions, compared to different-sex ones. The first concerns meeting opportunities. Because of the much smaller pool of potential partners for same-sex oriented people, they may have greater difficulty finding someone like themselves (Schwartz and Graf 2009). Second, as a nontraditional union, individuals with a same-sex orientation are thought to have less conventional preferences for partners, leading to greater tolerance of partner heterogeneity (Potârcă, Mills, and Neberich 2015). At the same time, however, other researchers argue that same-sex couples might be more homogeneous than different-sex unions, because how they relate to each other is not affected by the gender hierarchy in society, resulting in their more egalitarian preferences (Verbakel and Kalmijn 2014). Empirically, same-sex unions are found to be more heterogamous with respect to age in a few European countries and the United States, but this pattern does not extend to their educational levels (Andersson et al. 2006; Schwartz and Graf 2009; Verbakel and Kalmijn 2014).

Other than understanding how same-sex oriented people may have different partner preferences, another reason to study differences in couple homogeneity by sexual orientation is that such differences may also account for any discrepancy in stability between same- and different-sex relationships. Because more heterogamous unions likely feature greater differences in attitudes, lifestyles, and beliefs between the partners, they tend to have poorer relationship quality and thus higher dissolution rates (Schwartz 2013). Most research on the associations between matching patterns and relationship outcomes, however, focuses on different-sex couples. One exception is Manning and colleagues’ (2016) study of same-sex cohabiting couples in the United States. These researchers find that educationally heterogamous unions face modestly higher dissolution risks, whereas age heterogamy and interracial unions are not tied to higher dissolution risks. Two other studies examining socioeconomic differences within couples indicate that equal earnings within unions reduce the likelihood of breakups for same-sex couples, but increase this likelihood for different-sex couples (Kalmijin, Loeve, and Manting 2007; Weisshaar 2014). All of these studies, however, are based on cohabiting or married/registered couples, and thus the findings might not be generalizable to contexts with relatively low cohabitation rates and where same-sex partnership is not legalized.

3. The Taiwanese context

Taiwan exemplifies industrialized societies that have been experiencing less and later marriage (Raymo et al. 2015). In 2016, the average age of first marriage was 32 years old among men and 30 years old among women in Taiwan (DGBAS 2016), making the society among those with the greatest delays in marriage in the world. Unlike many Western countries where unmarried couples frequently cohabit (Manning, Brown, and Payne 2014), Taiwan, akin to other East Asian countries, has a relatively low level of premarital cohabitation, despite increasing marriage postponement (Raymo et al. 2015). While demographers have proposed various explanations for the rapid marriage declines in a number of East Asian countries (Boling 2008; Jones 2007; Raymo and Iwasawa 2005), Yu and Kuo’s (2016) recent study of Japan, which shares Taiwan’s patterns of cohabitation and marriage, specifically indicates a close connection between relationship formation and marriage entry. As these researchers argue, young people’s lack of romantic involvement to a great extent accounts for their delayed entry into marriage. Understanding relationship dynamics is therefore potentially important to explaining demographic trends in Taiwan.

Existing research on the formation or dissolution of relationships in Taiwan, or elsewhere in Asia, is scarce. The potentially close connection between relationship and marriage formation, however, suggests that previous research on the determinants and patterns of marriage formation in Taiwan and other East Asian countries with similar cultural and demographic characteristics is likely to shed light on relationship dynamics, especially for different-sex couples, in Taiwan. More than in Western countries, studies of marriage formation in East Asia have emphasized parental influences (Raymo et al. 2015). Although the tradition of arranging marriage for their children has declined, East Asian parents continue to facilitate their children’s marriage by providing monetary support to their children when they enter marriage (Lin and Pei 2016) or by becoming involved in children’s mate selection (Wang and Chen 2017). In Taiwan, parental expectations still exert some degree of influence on children’s choice of mate, as well as affect children’s perceptions of their relationship quality (Chen and Chen 2014; Tsay and Wu 2006).

Part of the reason why Taiwanese parents remain influential in regards to their children’s marriage formation is Taiwan’s high level of parent-child coresidence before marriage. Similar to their counterparts in some Southern European and other East Asian countries, adult children in Taiwan expect and are expected to remain in the parental home until marriage (Yu and Kuo 2016; Zeng et al. 1994), unless their job or school makes commuting from the parental home infeasible (Huang 2013; Nauck, Gröpler, and Yi 2017). This family practice, combined with the rising age of marriage, leads more than one half of young adults in Taiwan to live in their parental home at age 30, which is drastically different from Western countries, such as the United States and Germany, where near 90% of young adults have left the parental home by the same age (Nauck, Gropler, and Yi 2017).

The cultural tradition of parental intervention in children’s marriage choices, further empowered by children’s prolonged stay in the parental home, has important implications for same-sex relationship dynamics. Although public acceptance of homosexuality has been increasing, with Taiwan about to make legislative changes to enable same-sex marriage,¹ this acceptance is far less common among older adults (Cheng, Wu, and Adamczyk 2016). That is to say, parents of today’s young adults in Taiwan are likely to disapprove of their children’s same-sex relationships. If pressure from parents or other family members serves as a stressor that could accelerate the dissolution of same-sex relationships, as indicated by the minority stress perspective (Cao et al. 2017; Meyer 2003), then we should observe a higher dissolution rate of same-sex relationships in the Taiwanese context, especially among those living with their parents.

Premarital coresidence, as a proxy for parental influences on children, may also differentially affect same- and different-sex relationship formation. Previous research demonstrates that Taiwanese men and women living with parents actually experience a greater delay in their first marriage than those who do not (Yu, Su, and Chiu 2012). A recent study further suggests that this delay may have to do with coresident children’s lower likelihood of forming relationships (Yu and Kuo 2016). The argument for the deterrent effect of coresidence on different-sex relationship formation specifically focuses on coresident children’s lower psychological need for companionship, resulting in their higher standards for romantic partners. In the case of same-sex oriented people, coresidence with parents may even more strongly discourage relationship formation, as they simultaneously may face disapproval from their parents.

In addition to how parents may shape children’s paces of entering marriage, previous research on marriage formation in Taiwan also indicates a strong tendency of assortative mating (Tsay 1996). Although the traditional gender norm favoring the husband having higher status than the wife leads a sizeable proportion of Taiwanese women to marry men who are older and more educated than themselves, the percentages of relationships that feature educational and age homogamy have increased steadily in Taiwan (Cheng 2014; Chu and Yu 2010). Furthermore, previous research indicates a strong tendency for people in Taiwan to marry those from a similar family socioeconomic background (Chu and Yu 2010). Parents’ relatively great influence on children’s marriage choices is thought to partly contribute to this high level of family background homogamy, as parents, following the tradition of arranging marriage with a family of equivalent socioeconomic standing, are likely to prefer such a mating pattern more than their children (Tsay and Wu 2006).

Another dimension of assortative mating in Taiwan has to do with ethnicity, with each main ethnic group demonstrating a strong tendency to marry within the group (Tsay and Wu 2006). To provide some background, the ethnic divides in Taiwan mainly follow the timing of arriving in the island. The aborigines were in Taiwan the earliest, followed by Han Chinese migrating from Coastal China in the 16^th and 17^th centuries, who are further divided as Hokkien and Hakka based on their original region and dialect. The fourth ethnic group, Mainlanders, consist of Chinese migrants who arrived around 1949 with the Kuomintang government, after its loss of the civil war, and these migrants’ descendants. As of the early 2000s, over 80% of Taiwanese were married to a member of their own ethnic group (Tsay and Wu 2006).

Given the likely close link between romantic relationships and marriages in Taiwan, one would expect different-sex relationships to have considerable levels of couple similarities in age, education, ethnicity, and family background, just like marriages. Regarding our question of how same-sex relationships may differ in this aspect, however, the level of couple homogeneity may depend on the specific characteristic in question. Because same-sex couples would not be affected by the gender norm that prefers the male partner in a relationship to have higher status than the female partner, same-sex unions may be more homogamous with respect to age and education. For ethnicity and family economic status, however, not only might same-sex oriented people be more tolerant, as claimed by researchers using Western cases (Potârcă, Mills, and Neberich 2015; Schwartz and Graf 2009), but also, in the context of Taiwan, they may be subject to less parental pressure to find partners from comparable family backgrounds. As same-sex marriage is still illegal in Taiwan, parents’ preference for their children to marry someone from a similar family background, so that they can keep the kinship homogeneous, would not apply to same-sex relationships. Thus, same-sex relationships may be less homogamous in ethnicity and family economic status.

4. Methods

4.1. Data

The data used in study are from the Taiwan Education Panel Survey and Beyond (TEPS-B), conducted by a team of researchers at National Chengchi University, National Taiwan University, and Academia Sinica in Taiwan, with funding from the Ministry of Science and Technology in Taiwan. The TEPS-B is a follow-up survey of the Taiwan Education Panel Survey (TEPS), which collected data from a nationally representative sample of 20,055 seventh-grade students in 2001 and followed a subsample of them through 12^th grade (i.e., TEPS core sample, including 4,261 individuals). The TEPS-B selected all 4,261 individuals from the TEPS core sample and added a random subsample from the rest of the TEPS sample for interviews in 2014. Using this sampling procedure and the face-to-face interview technique, the TEPS-B collected information from 5,172 Taiwanese men and women, who, altogether, can be considered as a nationally representative sample of young adults born in 1988–1989 (Kuan 2017).

Despite being from one specific cohort, the TEPS-B data are exceptionally rich. In addition to full educational and residential histories, the TEPS-B asked all respondents to report detailed information on their current and previous romantic relationships that had lasted one month or more since age 15, including the month and year they began and ended each relationship. Such information enables us to analyze the paces and patterns of entering and exiting relationships for all respondents from mid-adolescence to early adulthood. Although retrospective life history data face the potential problem of recall errors, we believe that such errors are likely few in our data on romantic relationships, because the TEPS-B respondents were relatively young, and most of them had just one or two romantic relationships to report.² The TEPS-B did not include questions about respondents’ sexual orientation, but it asked respondents to report the sex and other characteristics (e.g., age, education, ethnicity, and family economic status) of each romantic partner. Based on these responses, we can determine whether a relationship is a same- or different-sex one. In an exploratory analysis, we found that among those reporting having had any same-sex relationship, their romantic partners tended to all be same-sex ones, with few reporting both same- and different-sex relationships over time. This consistency provides us with some confidence that the same-sex relationships recorded in our sample are unlikely to result from coding or reporting errors. The detailed information about each romantic partner also enables us to compare the matching patterns between same- and different-sex unions.

To conduct the statistical analysis that utilizes respondents’ life history information, we converted the retrospective data into person-month observations, with time-varying information for each respondent. Because our focus is romantic relationships since age 15, we constructed the person-month sample from the month when respondents turned 15 years of age through the month of the interview. The 5,172 respondents in the sample reported 8,482 relationships in total. After excluding observations with missing values on key variables, the analytical sample contains 8,306 relationships (with 332 same-sex ones) and 568,209 person-months.

Unlike previous studies, which mostly focus on cohabiting/married same-sex couples, our data enable us to capture all romantic relationships respondents had experienced since age 15. This study can therefore provide a comprehensive view of relationship dynamics, from formation to dissolution. Our focus on romantic relationships, instead of cohabiting unions, is also more appropriate for non-Western contexts where cohabiting is relatively rare.³ Like many other studies of same-sex unions (Umberson et al. 2015), however, we are limited by the modest number of same-sex relationships in our sample. Also akin to previous research, we face potential social desirability bias; that is, some same-sex relationships might not have been reported, because respondents felt less comfortable reporting such relationships (Lau 2012). These potential limitations notwithstanding, the percentage of unions that are same-sex in the TEPS-B is quite similar to those in other studies of Western countries. In the TEPS-B, same-sex relationships account for 4% (332/8,306) of all romantic unions, compared to 3% (263/8,437) in a British study (Lau 2012), 5% (126/2,283) in one U.S. study (Manning, Brown, and Stykes 2016), and 2% (335/14,005) in another study conducted in the United States (Joyner, Manning, and Bogle 2017).

Consistent with the U.S. finding that men report fewer same-sex relationships than women (England 2016), our sample also includes more female same-sex unions than male ones. Specifically, out of the 332 same-sex relationships, 236 involve women and 96 involve men. Given the potentially different social desirability bias between men and women engaged in same-sex relationships, in an exploratory analysis, we fitted models separately for men and women. Although some modest gender differences in same-sex relationship dynamics were observed, regarding our primary research concern, whether same- and different-sex relationship dynamics significantly differ, the results were consistent regardless whether we separated or combined men and women in the statistical models. Because of the modest number of same-sex relationships in our sample, combining men and women in the analysis, as do most prior studies of same-sex unions (Lau 2012; Manning, Brown, and Stykes 2016), has the benefit of providing greater statistical power. We therefore decided not to separate the sample by gender in the analysis.

4.2. Variables and analytic strategies

To investigate how same- and different-sex relationships differ in the patterns of formation, dissolution, and couple homogeneity, the statistical analysis in this study contains three parts, with different dependent variables for each part. First, we conduct an event history analysis of Taiwanese young adults’ paces of transitioning into a same-sex or different-sex romantic relationship (Yamaguchi 1991). Specifically, we use competing-risk discrete-time hazard models, which employ the multinomial logit regression technique, to estimate the odds of individuals entering into a same-sex or different-sex relationship at a given month, conditional on the event not occurring before that time. We code the dependent variable for this part of the analysis as 1 if respondents started a different-sex relationship at month t and 2 if they began a same-sex relationship; otherwise we code it as 0. To ensure that the predictors used to study the entry into a relationship precede the transition, we use independent variables at month t-1 to predict the relationship entry at t. Because only the person-months subject to the risk of entering a relationship can be included in event history models, this part of the analysis is restricted to the person-months when respondents were not in any relationship (364,576 person-months).

Second, to show whether same- and different-sex relationships differ in the levels of couple homogeneity, we fit a series of binary logit regression models predicting whether respondents’ relationships are homogamous in age, education, ethnicity, and family economic status, conditional on their entering a relationship. That is, only the person-months when respondents entered a relationship are selected for this part of the analysis (n = 8,306). For the dependent variables for this set of models, we create binary indicators of age, education, ethnicity and family homogamy, defined as the couple being the same age, having attended equivalent educational levels, reporting the same ethnic origin, and having similar family economic conditions, respectively, at the time of entering the relationship (coded as 1). We code these outcome variables as 0 if the relationship is heterogamous in age, education, ethnicity, and family economic status, respectively, or if respondents did not know the attribute of concern for their partner.⁴ Previous research on assortative mating uses various age differences as cutoffs when defining age homogamy (Qian and Qian 2014; Tsay and Wu 2006). In an exploratory analysis not presented here, we tried the different cutoffs (from 1–4 years of age differences), and the results were consistent regardless of the cutoffs. We measure education homogamy based on the highest levels of education that respondents and their partners had attended at the time. Specifically, we consider a relationship as homogamous if the couple’s highest educational levels were both: (1) high school or less, (2) technical or general university, or (3) postgraduate school. Regarding ethnicity, the TEPS-B asked respondents to report their father’s ethnic origin, a typical way for young adults to identify their ethnicity (and to be identified as such) in Taiwan.⁵ For the partner’s ethnicity, we rely on respondents’ reports. Following previous research on ethnic homogamy (Tsay and Wu 2006), we divide respondents and their partners into four major ethnic groups: (1) Hokkien, (2) Hakka, (3) Mainlanders, and (4) Aborigines. As to family background homogamy, we measure it based on respondents’ reports of how their romantic partners’ family economic status compared to their own. We consider cases where respondents reported to have similar or same family economic conditions as their partner’s as homogamous.

The third part of the statistical analysis focuses on union dissolution, once again using an event history approach (Yamaguchi 1991). We select all the person-months during which respondents were subject to the risk of union dissolution – that is, when respondents were in a relationship – and use discrete-time hazard rate models to estimate the odds for their relationship to end at a given month, conditional on it has not occurred earlier. We measure the dependent variable for this part of the analysis based on respondents’ reports of when each of their relationships ended. Among the 8,306 relationships in our analytical sample, 5,930 were terminated by survey time. A total of 203,633 person-months, during which respondents reported to be in a relationship, are included in the event history models predicting relationship dissolution.

With respect to the independent variables for our analysis, a key predictor included in most of our models is whether a relationship is a same-sex one. We determine this relationship characteristic based on respondents’ report of the sex of their partners. For about 22% of respondents, who never had any relationship, we cannot tell their sexual orientation. As a result, we can only include the indicator of same-sex status in the models restricted to the person-months during which respondents were in a relationship (i.e., the parts of analysis that focus on whether a relationship is homogamous and the pace at which a relationship dissolves).

The rest of the independent variables are similar across different parts of the analysis. To begin, we use coresidence with parents to gauge parental influences on children’s relationship dynamics. We measure coresidence with parents based on respondents’ self-reported residential histories, with a dummy variable indicating whether respondents lived with at least one parent in a given month.

We also include several other time-variant variables that may shape relationship dynamics, such as age, educational level, school enrollment, personal income, and level of urbanization of the area in which respondents lived. Age is measured in three categories: age 17 and younger, 18–22 years of age, and age 23 and older. Educational level is measured as the highest level that respondents had attended at the time, in six categories: (1) general high school, (2) vocational high school, (3) junior college, (4) technical college, (5) university, and (6) postgraduate school. We also construct a binary indicator of whether respondents were enrolled in school during the observed month, for the reason that being in school may facilitate more opportunities to meet romantic partners. We measure personal income based on respondents’ reports of the initial wage of each job they had held since age 15. To prevent the results from being overly influenced by outliers, we take the natural logarithm of the reported wage (in New Taiwanese dollars).⁶ Because the level of urbanization of the area where respondents live may affect the chances to meet romantic partners, as well as their beliefs and preferences about romantic relationships and views on same-sex relationships, we further control for the level of urbanization. The TEPS-B asked respondents to report the ZIP code of their residence since age 15. After matching the ZIP code information with the categorization of urbanization level commonly used in major social surveys in Taiwan (Chang, Tu, and Liao 2012), we further divide the residential area into three categories: (1) large metropolitan areas, (2) other urban localities, and (3) rural areas.⁷ The models also control for respondents’ gender and whether any of their parents have high school or more education. Given that previous research finds that number of siblings is related to the timing of marriage (Yu, Su, and Chiu 2012), we further introduce respondents’ number of siblings into the models. In addition, we control for the number of romantic relationships experienced up to the observed month in the models, as those with more prior relationships may enter or exit relationships at a faster pace and be choosier about their partners’ characteristics.

For the event history models, we also include the duration of being exposed to the risk of the event of interest. Specifically, for the models predicting the entry into a same- or different-sex relationship, duration is measured as the number of months since respondents turned age 15, or since they exited the last relationship if they had any other relationship prior to that month. Because our preliminary analysis indicated a quadratic effect of the duration of singlehood on the odds of transitioning to a romantic relationship, we use both duration and duration squared in the event history models predicting relationship formation. For the models predicting relationship dissolution, the duration of exposure is measured as the number of months since respondents entered the relationship. Different from the models about relationship formation, we use only duration in the models predicting dissolution, as an exploratory analysis showed a linear – not curvilinear – association between the union’s duration and its dissolution. Finally, because our event-history data may include multiple relationship episodes from the same respondents, we further cluster the standard errors by individuals to allow for arbitrary within-person correlation.

5. Results

To provide an overview of how same- and different-sex relationships differ, Table 1 show descriptive statistics for the respondents engaged in same- and different-sex relationships separately, using their information in the month they entered the relationship. The table includes only those who reported ever having been in a romantic relationship. About one half of the relationships were formed during the ages of 18-22, regardless of sexual orientation. Nearly three quarters of the relationships were formed when respondents were in school. On average, respondents had only about one prior relationship when they entered a relationship, indicating that the Taiwanese form relatively few relationships from mid-adolescence to young adulthood. Most of the characteristics of respondents who formed same- and different-sex relationships are similar. One exception is the percentage of respondents living with parents at the time they entered a relationship: Whereas 49% of young adults were living with their parents when they began a same-sex relationship, it was 57% for different-sex couples. Moreover, respondents in same-sex relationships were more likely to live in large metropolitan areas than those in different-sex relationships, suggesting that it is easier to find same-sex partners in highly urban areas. The table also shows comparable percentages of same- and different-sex relationships that are homogamous in education and ethnicity, but a larger proportion of same-sex relationships are characterized by age homogamy. In terms of family economic background, about one half of heterosexual couples were formed by partners from the families of similar economic status, but only 39% of same-sex unions were formed by partners with this similarity.

Table 1:

Descriptive statistics of same-sex and different-sex couples

	Same-sex	Different-sex
Living with parent(s) (%)	49.4	56.5*
Education (%)
General high school	20.2	17.8
Vocational high school	20.2	18.8
Junior college	5.1	5.0
Technical college	18.1	23.6
University	31.0	31.2
Postgraduate school	5.4	3.6
Enrolled in school (%)	72.3	72.2
Ethnicity (%)
Hokkien	77.4	74.4
Hakka	6.6	12.5*
Mainlanders	9.3	8.7
Aborigines	6.7	4.4*
Age at entering union (%)
15-17	29.2	24.3
18-22	49.4	55.7
23-26	21.4	20.0
Income (in NT dollars)	7274.7 (12695.9)	7518.6 (12629.4)
Number of siblings	1.6 (0.8)	1.6 (0.8)
Parent more than high-school education (%)	33.7	34.2
Previous number of relationships	1.2 (1.3)	1.0* (1.1)
Residential area (%)
Large metropolitan areas	61.5	54.4*
Other urban localities	30.1	36.7
Rural areas	8.4	8.9
Couple homogeneity (%)
Age homogamy	45.2	41.5*
Education homogamy	70.8	68.1
Ethnicity homogamy	58.1	57.4
Family homogamy	38.6	47.2*
Union dissolution (%)	67.5	71.5
N of unions	332	7,974

Notes: Values for categorical variables are in percent. The mean values, followed by standard deviations in parentheses, are presented for all other variables.

^*Indicates a significant difference (p < 0.05) from same-sex couples based on chi-square test or t-test.

Turning to the multivariate analysis, Table 2 presents discrete-time multinomial logit models predicting young adults’ entry into either a same-sex or a different-sex relationship. The first two columns present the odds ratios of individuals entering into same-sex and different-sex relationships, respectively, as opposed to not entering any relationship, in a given month. We also include a third column indicating whether the odds ratios for forming same- and different-sex unions are statistically different. Consistent with previous research showing that coresidence with parents delays marriage formation (e.g., Yu and Kuo 2016), the results indicate that young Taiwanese adults living in the parental home enter a romantic relationship at a slower pace. The odds ratios of parent-child coresidence, however, differ for entering different types of relationships. Whereas the odds for those coresiding with their parents to enter a different-sex relationship in any given month are about 14% less than those who do not live with parents, the odds for the former to enter a same-sex relationship are near 47% less than the odds for those who have left the parental home. As the third column in Table 2 shows, the odds of entering a different-sex relationship are significantly different from those of entering a same-sex relationship (p < 0.001). Thus, living with parents deters entry into same-sex relationships much more than it does different-sex relationships.

Table 2:

Competing-risk event history models predicting entry into romantic relationships

	Same-sex/No relationship	Different-sex/No relationship	Differencea
Living with parent(s)	0.531*** (0.076)	0.862*** (0.023)	***
Female	3.343*** (0.646)	1.050* (0.025)	***
Education (ref. general high)
Vocational high school	1.252 (0.317)	1.250*** (0.053)
Junior college	1.048 (0.346)	1.426*** (0.089)
Technical college	1.043 (0.317)	1.554*** (0.075)
University	1.304 (0.366)	1.592*** (0.074)
Postgraduate school	2.072 (0.938)	1.726*** (0.143)
Enrolled in school	1.104 (0.231)	1.167*** (0.046)
Ethnicity (ref. Hokkien)
Hakka	0.542* (0.154)	1.061+ (0.037)
Mainlanders	1.069 (0.396)	1.087+ (0.048)
Aborigines	2.289* (0.812)	1.335*** (0.082)
Age (ref. 15-17)
18-22	1.141 (0.252)	1.820*** (0.080)	*
23-26	2.040* (0.713)	3.413*** (0.230)
Log income	1.033+ (0.018)	1.047*** (0.003)
Number of siblings	0.921 (0.100)	0.987 (0.015)
Parent more than high-school education	0.896 (0.191)	1.019 (0.028)
Previous number of relationships	1.223*** (0.076)	1.029* (0.014)	**
Residential area (ref. large metro)
Other urban localities	0.799 (0.121)	1.021 (0.026)
Rural areas	0.888 (0.226)	1.096* (0.048)
Duration of exposure	0.961*** (0.007)	0.970*** (0.001)
Duration squared	1.000* (0.000)	1.000*** (0.000)
Constant	0.001*** (0.001)	0.023*** (0.001)	***
N of person-month observations		364,576

Note: Results are presented in odds ratios, with values in parentheses representing cluster-corrected robust standard errors.

^aIndicates whether the odds ratios for forming same- and different-sex relationships are statistically different.

^***p < 0.001,

^**p < 0.01,

^*p < 0.05,

⁺p < 0.10 (two-tailed tests)

The first two columns of Table 2 also indicate that women and those with more prior relationship experiences have greater odds of entering both same- and different-sex relationships during the observed time period. Being enrolled in school, having a higher income, and being older also increase the odds of forming a relationship, regardless of sexual orientation. The third column in Table 2, however, shows that the odds ratios of many factors shaping the formation of either type of relationship, are not statistically different according to the relationship’s sexual orientation. One exception is that the gender gap in the odds of entering a same-sex relationship is greater than that in the odds of entering a different-sex relationship (p < 0.001). As discussed earlier, this result may have to do with men’s greater tendency to underreport same-sex relationships (England 2016). The tendency for young Taiwanese to enter a relationship when they are between 18 and 22 years of age, rather than at a younger age, is also stronger for those entering same-sex than different-sex relationships (p < 0.05), suggesting that Taiwanese youth are especially unlikely to form same-sex relationships during mid-adolescence. Overall, the generally few differences in factors associated with entries into same- and different-sex unions provide support for the argument that romantic relationships are formed in similar ways regardless of sexual orientation.

Next, we present results from a series of binary logit regression models predicting whether respondents’ relationships are homogamous in age, education, ethnicity, and family economic status, conditional on respondents’ entering a relationship (Table 3). To conserve space, we show only odds ratios for sexual orientation, even though the models control for all the other variables predicting relationship formation included in Table 2. The results indicate that same-sex couples are more homogamous in age and education, while less homogamous in family economic background than heterosexual couples. There is no difference in the tendency of ethnic homogamy between the two types of relationships. To be specific, for age homogamy, the odds of the partners being similar in age is 28% higher among same-sex couples than among their heterosexual counterparts. Similarly, same-sex couples’ odds of educational homogamy are 32% higher than those of different-sex couples. By contrast, the odds that the two partners in the union come from families with comparable economic conditions are about 25% less for same-sex couples than for heterosexual ones.

Table 3:

Partial results from logit models predicting relationship homogamy on various characteristics

	Age Homogamy	Education Homogamy	Ethnicity Homogamy	Family Homogamy
Same-sex	1.282* (0.151)	1.324* (0.188)	0.936 (0.138)	0.746* (0.109)
Log-likelihood	−5,142	−4,346	−4,569	−5,689
N of observations	8,306	8,306	8,306	8,306

Notes: Results are based on logistic models predicting relationship homogamy, which also include coresidence with parents, gender, educational level, school enrollment, ethnicity, age, personal income, number of siblings, previous number of relationships, parents’ education, and residential area as predictors. Results are presented in odds ratios, with values in parentheses representing cluster-corrected robust standard errors.

^***p < 0.001,

^**p < 0.01,

^*p < 0.05,

⁺p < 0.10 (two-tailed tests)

Turning to the analysis of relationship stability, Table 4 shows results from a series of discrete-time logit models predicting union dissolution. Model 1 presents the baseline model, without considering the stress respondents were potentially under because of living in the parental home. We add coresidence with parents in Model 2 and test whether this factor is associated with same- and different-sex relationships differently in Model 3. Interestingly, although Taiwan and other Confusion countries have a relatively lower social acceptance of same-sex relationships, especially among the older generation, and stronger parental influences on children’s mating choice than many Western countries (Adamczyk and Cheng 2015; Cheng, Wu, and Adamczyk 2016; Wang and Chen 2017), the odds of exiting a same-sex relationship is not significantly different from those of a different-sex union, regardless of the children’s living arrangements. This finding, once again, suggests that same-sex relationship dynamics are similar to different-sex ones. Model 3 further shows that living with parents accelerates union dissolution considerably more for same-than different-sex relationships. Among young Taiwanese in different-sex relationships, the odds for those living with parents to exit their relationships at a given month are 16% higher than the odds for those living independently. Conversely, among those in same-sex relationships, the odds of terminating the relationship are 63% higher if they live with parents (1.156*1.412=1.632).

Table 4:

Discrete-time event history models predicting union dissolution

	Model 1	Model 2	Model 3
Same-sex	1.030 (0.073)	1.050 (0.073)	0.880 (0.091)
Living with parent(s)		1 172*** (0.036)	1.156*** (0.036)
Same-sex × Living with parent(s)			1.412* (0.203)
Female	0 755*** (0.021)	0.748*** (0.021)	0.748*** (0.021)
Education (ref. general high)
Vocational high school	1.012 (0.048)	1.019 (0.048)	1.019 (0.048)
Junior college	0.757*** (0.056)	0.777*** (0.058)	0.780*** (0.058)
Technical college	0.663*** (0.037)	0.686*** (0.038)	0.686*** (0.038)
University	0.637*** (0.034)	0.672*** (0.036)	0.671*** (0.036)
Postgraduate school	0.425*** (0.039)	0.458*** (0.042)	0.457*** (0.042)
Enrolled in school	1.256*** (0.054)	1.266*** (0.054)	1.267*** (0.054)
Ethnicity (ref. Hokkien)
Hakka	1.014 (0.041)	1.017 (0.041)	1.017 (0.041)
Mainlanders	1.008 (0.048)	1.007 (0.048)	1.006 (0.048)
Aborigines	0.855* (0.060)	0.890 (0.063)	0.895 (0.063)
Age (ref. 15-17)
18-22	0199*** (0.064)	1.248*** (0.068)	1.248*** (0.068)
23-26	0.112 (0.085)	1.143+ (0.087)	1.142+ (0.087)
Log income	0.995 (0.003)	0.994+ (0.003)	0.994+ (0.003)
Number of siblings	0.970+ (0.016)	0.973+ (0.016)	0.972+ (0.016)
Parent more than high-school education	1.056+ (0.032)	1.054+ (0.032)	1.054+ (0.032)
Previous number of relationships	1.020 (0.015)	1.023 (0.015)	1.022 (0.015)
Residential area (ref. large metro)
Other urban localities	0.978 (0.029)	0.985 (0.029)	0.986 (0.029)
Rural areas	0.954 (0.046)	0.967 (0.047)	0.968 (0.047)
Duration of union	0.987*** (0.001)	0.987*** (0.001)	0.987*** (0.001)
Constant	0.064*** (0.005)	0.055*** (0.004)	0.056*** (0.004)
Log-likelihood	−26,314	−26,300	−26,297
N of person-month observations	203,633	203,633	203,633

Note: Results are presented in odds ratios, with values in in parentheses representing cluster-corrected robust standard errors.

^***p < 0.001,

^**p < 0.01,

^*p < 0.05,

⁺p < 0.10 (two-tailed tests)

Also about the patterns of union dissolution, the final part of our analysis examines whether couple homogeneity contributes to relationship stability differently for same- and different-sex couples. Table 5 shows partial results from discrete-time logit models predicting same-sex and different-sex couples’ union dissolution, with the addition of variables measuring union homogamy (in age, education, ethnicity, and family economic status) and interaction terms between homogamy variables and the same-sex indicator. The models also include all the other variables predicting union dissolution included in Table 4. Models 1 and 2 indicate that age homogamy is hardly associated with the pace of relationship dissolution, and this result does not differ between same- and different-sex couples. Perhaps romantic relationships formed from middle adolescence to young adulthood tend not to feature a great age gap between the two partners, leading to less difference between couples who are homogamous in age and those who are not during this stage.

Table 5:

Partial results from event history models predicting dissolution of unions with different matching patterns

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8
Same-sex	1.050 (0.073)	1.027 (0.104)	1.054 (0.074)	1.050 (0.124)	1.039 (0.073)	1.052 (0.098)	1.025 (0.072)	0.984 (0.089)
Age homogamy	0.994 (0.028)	0.992 (0.028)
Same-sex × Age homogamy		1.044 (0.142)
Education homogamy			0.835*** (0.026)	0.835*** (0.027)
Same-sex × Education homogamy				1.006 (0.145)
Ethnicity homogamy					0.735*** (0.023)	0.736*** (0.024)
Same sex × Ethnicity homogamy						0.977 (0.129)
Family homogamy							0.732*** (0.019)	0.729*** (0.020)
Same-sex × Family homogamy								1.120 (0.149)
Log-likelihood	−26,298	−26,298	−26,279	−26,279	−26,248	−26,248	−26,230	−26,230
N of person-month observations	203,633	203,633	203,633	203,633	203,633	203,633	203,633	203,633

Notes: Results are based on discrete-time hazard rate models predicting union dissolution, which also include duration of union, coresidence with parents, gender, educational level, school enrollment, ethnicity, age, personal income, number of siblings, previous number of relationships, parents’ education, and residential area as predictors. The estimates are presented in odds ratios, with values in parentheses indicating cluster-corrected robust standard errors.

^***p < 0.001,

^**p < 0.01,

^*p < 0.05,

⁺p < 0.10 (two-tailed tests)

For education, ethnicity, and family economic conditions, the results are consistent with the argument that couples that share similar traits enjoy better relationship quality and last longer (Models 3-8). Interestingly, the associations between different types of homogamy and relationship dissolution hazards are not statistically different between same- and different-sex couples. Thus, despite somewhat different odds of entering relationships that are homogamous in age, education, and family economic background, once same- and different-sex oriented people form relationships, whether their relationship is homogamous does not affect their relationship stability differently, suggesting that both groups value having a romantic partner resembling themselves to the same extent. In this sense, results from this part of analysis provide further support for the argument that close dyadic relationships work in similar ways irrespective of sexual orientation.

6. Conclusions

Despite rising attention to the similarities and differences between same- and different-sex relationships, existing studies rarely compare these relationships from their formation to dissolution, or shed light on same-sex unions in non-Western contexts. Using retrospective data on young adults’ relationship histories in Taiwan, this study is the first to provide evidence on the dynamics of same- and different-sex relationships in a relatively conservative cultural context. Results from our analysis demonstrate that young adults’ experiences of homosexual and heterosexual relationships, for the most part, share similar dynamics in Taiwan. Although Taiwan society has only recently become relatively tolerant of homosexuality, many of the factors associated with the formation and dissolution of same- and different-sex relationships are alike among young people. Moreover, contrary to the expectation that the higher prevalence of parental intervention in children’s mate selection may lead to greater relationship instability for same-sex couples in Taiwan, we find no significant differences between young people’s paces of exiting same- and different-sex relationships.

Our result that the level of relationship stability does not differ between same- and different-sex couples in Taiwan is parallel to recent U.S. findings (e.g., Manning et al., 2016), while differing from evidence based on European countries (e.g., Lau, 2012). It is noteworthy, however, that many of the previous studies showing different levels of relationship stability by sexual orientation either compare cohabiting couples with married/registered couples or focus only on cohabiting unions. Unlike such studies, our comparison of same- and different-sex relationships uses all romantic relationships formed from mid-adolescence to early adulthood, without requiring individuals to select themselves into cohabiting unions or marriages. Given the potential problem of differential selectivity of same- and different-sex couples into marriage or cohabitation, our relationship-based analysis is far less likely subject to bias. Therefore, our finding that same- and different-sex relationships have similar levels of stability is especially important and useful in addressing the relevant debate.

Even though we do not find differences in relationship stability between same- and different-sex couples in Taiwan, some of our results suggest that parents, when able to influence children’s mate selection, may still discourage same-sex relationships. Specifically, our models indicate that coresidence with parents deters entry into same-sex romantic relationships more than different-sex relationships. Coresidence also accelerates the dissolution of same-sex relationships more than different-sex ones. Same-sex oriented young adults who live with their parents appear to be more sensitive to their parents’ disapproval of same-sex unions, hence encountering greater difficulties forming and maintaining a romantic relationship. The results about coresidence are somewhat consistent with the minority stress perspective, if we view having parents in close proximity as a stressor. Rather than sexual minority groups being generally subject to greater stress, however, we suggest that their relationships are more at risk when they are in close contact with members of an older generation, who tend to find same-sex relationships unacceptable. As premarital coresidence is common in many parts of the world (Yu and Kuo 2016), our results about coresidence and same-sex unions have general implications for understanding how social norms regarding living arrangements may slow the rises in same-sex unions in those parts of the world.

Another important finding from this study has to do with the different matching preferences among same-sex couples as compared with their different-sex counterparts. We show that same-sex relationships are more likely to be homogamous in age and education, but are less likely to be homogamous in family economic status than different-sex ones. These findings are consistent with the argument that same-sex couples are less bounded by societal norms that prescribe unequal age and educational statuses within the union, but similar economic backgrounds of the two families connected through the couple. Whereas previous research comparing the patterns of assortative mating among same- and different-sex couples in the West often focuses on how same-sex couples are limited by having fewer available mates (e.g., Schwartz and Graf 2009), our finding of their lower tendency of homogamy in family economic background in Taiwan suggests that they may be less restricted in other ways. Like many Asian societies, marriage is viewed in Taiwan as a union between two families, not just individuals in love. Societies with such a view tend to highlight the importance of finding romantic partners from similar family economic backgrounds. Same-sex oriented people, however, are less likely to be under the same pressure, because marriage is still illegal and hence cannot be the destination for a same-sex relationship. Overall, this study paints a mixed picture for how parental influence may determine same-sex relationships in Taiwan. On the one hand, young adults living in the same household with their parents appear to encounter greater difficulties forming and maintaining same-sex relationships than those living independently. On the other hand, same-sex oriented people seem to benefit from being less bound by parental preferences for their partners’ characteristics; they can potentially be more autonomous in mate selection.

Certain limitations of this study are worth noting. First, same-sex couples included in this study are likely younger than those in Western countries, given that our data include only relationship experiences through respondents’ mid-20s. Hence, results from this study may not be entirely comparable to those from Western countries. However, by focusing on relationships from mid-adolescence to early adulthood, our study helps enhance our knowledge of early relationship dynamics for same-sex oriented people. Second, like many other studies (Lau 2012; Manning, Brown, and Stykes 2016), we also face the challenge of not having a direct measure of sexual orientation. Respondents in our study, however, were generally consistent in the sex of their intimate partners, with few reporting both same- and different-sex relationships over time. The lack of data on sexual orientation further limits our ability to examine whether same-sex oriented people may have stronger preferences for living apart from parents than their heterosexual counterparts, especially if they are highly interested in romantic relationships. This differential selectivity into living independently, if exists, could explain some of association between coresidence and relationship dynamics presented in this study. Nevertheless, we should note that the strong cultural norm for single adult children to remain in the parental home and the high housing cost in East Asia make the decision to live away from parents largely dependent on the feasibility of commuting from the parental home to schools or jobs , rather than concerns about romantic relationships (Raymo et al. 2015). Thus, same- and different-sex oriented people’s differing motivations to live apart from parents alone are unlikely to fully account for the stronger association between living independently and forming a same-sex relationship. Besides, this difference cannot easily explain why coresidence accelerates the termination of same-sex relationships more than it does different-sex ones, given that all those already in relationships should be similarly interested in romantic relationships.

A third limitation is that our sample of same-sex couples includes more lesbian than gay couples, perhaps as a result of different social desirability bias between men and women engaged in same-sex relationships. Although the main results were consistent regardless whether we separated or combined gay and lesbian couples in the statistical models, future research will benefit from larger samples that enable scholars to further distinguish and systematically compare relationship dynamics between gay and lesbian couples (e.g., Joyner, Manning, and Bogle 2017).

Despite these limitations, findings from our study have important policy implications for societies in which homosexuality is still highly stigmatized and same-sex marriage remains illegal. Aside from a few differences between same- and different-sex relationships related to parental influences, our study provides strong evidence that same- and different-sex couples experience intimacies in similar ways—even in a relatively conservative cultural context like Taiwan. Thus, arguments that would deny equal rights for legal marriage based on the assumption that same- and different-sex unions differ in their meanings of intimacy have no legitimate foundation. To the extent that marriage is considered as a reasonable destination for heterosexual intimate relationships, marriage should also be an option for same-sex relationships.