Narratives of demographic shifts overlook how societal changes shape the family trajectories of sexual minorities. Using sequence analysis, we describe how partnering and parenthood evolve over the life course of lesbian, gay, and bisexual (LGB) women and men in the United Kingdom (N = 455) and how the types of these family trajectories changed across two birth cohorts (born before 1965 and in 1965–1979). We find five distinct trajectories between ages 18 and 40, wherein two thirds of the sample belonged to a family trajectory that did not involve living with children. Partnership-centered trajectories became more common across cohorts, and this increase came at the expense of trajectories characterized by singlehood among gay men and lesbian women. However, parenthood trajectories became less common among all LGB groups. Furthermore, family trajectories became more complex across cohorts, including more transitions, which coincides with trends in the general population. Yet we also find that family trajectories became less diverse among lesbian women and bisexual men, in contrast to the trend among gay men and the general population. The results demonstrate the dynamic, complex, and diverse nature of LGB individuals' family lives and why existing narratives of family-related demographic change should explicitly consider sexual minorities in demographic narratives.
What are the partnering and parenthood trajectories of sexual minorities? Have these life course trajectories changed across cohorts? Existing theoretical and empirical scholarship provides little guidance in answering these questions. Theories of demographic change primarily focus on explaining declines in relationship duration, marriage, and parenthood in the general population (Lesthaeghe 2010; Mills and Blossfeld 2003; van de Kaa 1987; Zaidi and Morgan 2017). Such perspectives are not designed to describe cohort change among the lesbian, gay, and bisexual (LGB) population, which was historically excluded from marriage and various routes into parenthood (Adamczyk and Liao 2019; Moore and Stambolis-Ruhstorfer 2013; Trandafir 2015). Empirically, most representative data sources available to study sexual minorities are limited to coresident couples consisting of two persons of the same sex or gender (Gates 2012; Kolk and Andersson 2020; Lau 2012; Manning and Payne 2021; Russell et al. 2020). Although these data are informative in understanding parenthood and partnership outcomes of same-sex couples at a given time, they hinder our understanding of union formation and parenthood as dynamic processes that evolve over the life course. Moreover, these data represent only sexual minorities that selected into (same-sex) partnering. Recently, large-scale nationally representative surveys started incorporating questions about sexual identity. Nonetheless, the opportunities these data sources provide to study the life course dynamics of partnering and parenthood among sexual minorities have remained underexplored (Badgett et al. 2021; Perales et al. 2020; Reczek 2020).
In this article, we use information on self-reported sexual identity and retrospective and prospective coresidence histories with partners and children to describe cohort change in the family formation trajectories of sexual minorities. We answer two main questions: (1) What are the family formation trajectories of lesbian, gay, and bisexual individuals (LGBs hereafter)?1 (2) Was there a cohort change in the type of trajectories? Using data from the Understanding Society survey (UKHLS), we construct the life courses of a relatively large representative sample of LGB women and men between ages 18 and 40. To examine cohort change, we compare a cohort that reached mid-adulthood (age 40) before most major legal changes took place in the United Kingdom (UK) (born before 1965) to a cohort that experienced those changes during early adulthood (born between 1965 and 1979).
We employ a sequence analysis approach to answer our questions. Previous research focused primarily on parenthood in same-sex unions (Goldberg 2006; Kolk and Andersson 2020; Manning and Payne 2021; Moore and Stambolis-Ruhstorfer 2013) or on union formation and dissolution (Carpenter 2020; Joyner et al. 2017; Kalmijn et al. 2007; Lau 2012; Manning et al. 2016; Ruiz-Vallejo and Boertien 2021). The sequencing of partnerships and parenthood over the life course provides a rich descriptive account of coresident familial experiences by documenting the occurrence, order, timing, and duration of multiple family-related events. We find descriptive evidence that the likelihood of belonging to family trajectories characterized by singlehood and trajectories that include parenthood has declined. In contrast, the likelihood of belonging to trajectories in which individuals live with a partner but not with children has increased across cohorts.
Family Trajectories in the LGB Population
We start the literature review with an overview of existing knowledge about union formation, dissolution, and childbearing among LGBs that guides several expectations regarding family life course trajectories. Subsequently, we discuss how LGBs' trajectories fit within existing theoretical frameworks and general narratives of demographic change.
Union Formation and Dissolution
Cross-sectional research in the United States suggests that LGBs are less likely to be in a coresidential relationship than heterosexuals (Black et al. 2000; Carpenter and Gates 2008). Lesbian women's partnership rates are similar to those of heterosexual women, while partnership rates are considerably lower among gay men and among bisexual men and women (Badgett et al. 2021). Moreover, studies on attitudes toward partnering have found that lesbian, gay, and bisexual women and men desire to be in relationships and to marry at similar rates as heterosexual women and men across various countries (Barrios and Lundquist 2012; D'Augelli et al. 2007; Riskind and Tornello 2017; Tate et al. 2019). Similarly, Meier et al. (2009) found modest differences across sexual identities in relationship values among young men and women in the United States.
There are several factors hindering LGBs' union formation that could shape different partnering trajectories across cohorts. First, fewer opportunities exist to meet suitable partners because fewer people identify as LGB than as heterosexual (Ellingson et al. 2004). Moreover, in contrast to heterosexuals, it is relatively uncommon for same-sex couples to have met their romantic partners through family or in school (Rosenfeld and Thomas 2012). However, these barriers have lessened over time; same-sex relationships are more prevalent, and nonheterosexual identities are more visible (England et al. 2016; Gilroy and Kashyap 2021). Also, decreased parental control and increased geographic mobility allow people more independence to form unions (Rosenfeld and Kim 2005). Furthermore, the emergence of online dating helped to facilitate connections between LGBs more easily.
Second, minority stress can lower the “returns” from being in a relationship. Minority stress is caused by discrimination and microaggressions related to a specific minority status (Frost and Gola 2015). LGBs experience many of these stressors when they enter a same-sex union, which could deter people from entering relationships (Frost et al. 2017). Studies in Germany and the United States have found that people in same-sex relationships are more likely to experience parental and social disapproval and express more concerns about their partner's acceptance by family and friends than are people in different-sex relationships (Hank and Wetzel 2018; Holmberg and Blair 2016; Reczek 2014). In an analysis of preferences among online daters across eight countries, Potârcă and colleagues (2015) found that residing in a supportive environment or region with formal recognition of same-sex unions was associated with an increase in long-term dating intentions among individuals looking for same-sex partners. Given the rapid increases in acceptance of same-sex couples over the last decades (Rosenfeld 2017; Trandafir 2015), we would expect that younger cohorts of LGBs are more likely to seek long-term unions than older cohorts of LGBs.
Third, institutional discrimination, such as the legal exclusion from marriage, can further increase the costs and reduce the benefits of being in a coresidential union for older LGBs. Same-sex marriage was not legal for large parts of older LGBs' lives. Marriage, unlike cohabitation, is related to institutional benefits in many contexts. For instance, legalizing same-sex marriage improved health outcomes among individuals in same-sex unions by increasing access to health insurance (Gonzales and Blewett 2014). Moreover, reductions in institutional discrimination are generally related to increased well-being for same-sex couples (Gonzales and Blewett 2014; Hatzenbuehler et al. 2009; Kail et al. 2015).
The developments discussed so far would predict an increase in the prevalence of LGBs' coresidential relationships across cohorts. However, reduced stressors and discrimination might not directly increase partnership rates. For example, the legalization of same-sex marriage did not increase the shares of LGBs in a union in Massachusetts (Carpenter 2020). Furthermore, some people in same-sex marriages express ambivalent feelings toward marriage and view it as a patriarchal and heteronormative institution, whereas others express their concerns that marriage could further assimilate sexual minorities into mainstream culture (Bosley-Smith and Reczek 2018; Stambolis-Ruhstorfer and Descoutures 2020). Singlehood could be more appealing to LGBs than to heterosexuals thanks to strong friendship networks (Dewaele et al. 2011; Lau 2012), and younger LGBs might opt to reject the heteronormative script of family formation that is embedded in coresidential partnerships (and childbearing). Taken together, these trends could theoretically imply cohort stability or even a decline, rather than an increase, in the prevalence of partnering among LGBs.
The foregoing discussion has centered on union formation processes, but the same factors are also relevant to understanding union dissolution among LGBs. Same-sex unions are more likely to end in separation than different-sex unions in some countries (Kalmijn et al. 2007; Kolk and Andersson 2020; Lau 2012; Manning et al. 2016), possibly because of the negative effects of stigma, discrimination, and lack of family support on relationship quality. Obstacles toward parenthood and marriage can prevent LGBs from investing in coresidential romantic relationships (Lau 2012). From this perspective, minority stress and stigma are likely to lead to relatively high separation rates, but these could have declined across cohorts as investments in relationships became more available to LGBs (Lau 2012; Ruiz-Vallejo and Boertien 2021).
There are also reasons to expect separation rates to increase or remain unchanged across cohorts. For instance, higher separation rates can stem from LGBs giving less importance to lifelong commitment in relationships than do heterosexual women (Meier et al. 2009). Similarly, Lau (2012) did not find changes in the stability of same-sex unions across cohorts in the UK and suggested that, among older cohorts, only committed same-sex couples might have decided to start living together. Finally, separation and serial cohabitation have become more common across cohorts in the general population (Bukodi 2012; Eickmeyer and Manning 2018; Hiekel and Fulda 2018), which could have made separation and repartnering among younger cohorts of LGBs prevalent as well.
In summary, multiple social and demographic forces suggest that the prevalence of trajectories involving partnerships should increase across cohorts, but expectations regarding changes in the prevalence of union instability are less clear.
Pathways Into Parenthood
Most evidence on parenthood among LGBs is limited to childbearing and parenting in same-sex couples. Overall, there is consistent evidence that women, and especially men, in same-sex couples are less likely to coreside with children than individuals in different-sex couples (Baumle and Compton 2011; Gates 2012; Kolk and Andersson 2020). However, there is variation within the LGB population. Recent research in the United States showed that bisexual women are as likely as heterosexual women to coreside with children. In contrast, bisexual and gay men are less likely than heterosexual men to coreside with children (Badgett et al. 2021). Trends over time are mixed. Kolk and Andersson (2020) found that childbearing within same-sex marriages increased considerably over time for Swedish women. In contrast, Gates (2015) reported that the share of same-sex couples living with children in the United States has declined.
There are numerous pathways into parenthood for LGBs. A prominent pathway among older LGB cohorts was through childbearing in different-sex relationships (Andersson et al. 2006; Gates 2012). Studies have suggested that gay men continue to experience pressures to form a different-sex union (e.g., Schacher et al. 2005). Sexual minority women also continue to have higher rates of unwanted births (Ela and Budnick 2017; Everett et al. 2021). Nonetheless, parenthood through a previous different-sex relationship is becoming less common among same-sex couples in the United States (Gates 2015). Alternative pathways to parenthood, which have become more accessible for LGBs over time, include adoption, assisted reproduction techniques (ARTs), and surrogacy (Moore and Stambolis-Ruhstorfer 2013; Rozental and Malmquist 2015). However, these options require considerable economic resources and planning (Park et al. 2020). Therefore, expanding access to routes into parenthood could have increased the number of same-sex couples with children, but these increases are limited to those with economic resources.
Relatively low levels of parenthood among LGBs can also be related to intentions and desires to become a parent. Studies on parenthood desires have documented that most gay and lesbian people would like to have children, but this share is lower than for heterosexuals and bisexuals (Hank and Wetzel 2018; Riskind and Patterson 2010; Romeu et al. 2015; Tate et al. 2019). Nonetheless, the gap between desired and observed parenthood is greater for gay men and lesbian women than for heterosexual men and women (Riskind and Tornello 2017). Several studies have documented how minority stress and the experience of discrimination reduce parenthood intentions among gay men and lesbian women, while favorable policy environments and involvement in the LGBT community increase parenthood desires (Costa and Bidell 2017; Scandurra et al. 2019). This would imply that the gap in parenthood desires between gay/lesbian and heterosexual individuals is smaller in more favorable contexts. Furthermore, societal pressures to follow mainstream family trajectories, including the formation of “nuclear” families, could increase for LGB individuals as possibilities to access institutions like marriage and parenthood expand (Duggan 2002).
Because of the focus on same-sex couples in the literature, we know very little about LGBs' experiences with single parenthood. However, studies have consistently found that sexual minority women are more likely to experience unwanted births (Ela and Budnick 2017; Everett et al. 2021). Moreover, given that many same-sex parent families are formed after the dissolution of a different-sex union (Gates 2015), it is likely that single parenthood is prevalent for some part of LGBs' life course, especially among older birth cohorts. However, it is unclear to what extent this was common and whether it changed across cohorts.
In summary, evidence suggests that social, legal, and demographic forces have varying implications for younger LGBs' experience with parenthood over the life course. The increasing acceptance of same-sex partnerships could have reduced the prevalence of parenthood within different-sex relationships among LGBs, whereas alternative routes toward parenthood have become more available. How the relative weight of both trends translates into cross-cohort trends in life course trajectories that include parenthood is unclear.
Demographic Change and Sexual Minorities
Existing narratives of demographic change are not constructed to understand family-related changes among the LGB population. These narratives focus on a retreat from hegemonic family trajectories driven by ideational change in the general population (Lesthaeghe 2010; van de Kaa 1987; Zaidi and Morgan 2017), economic insecurity (Mills and Blossfeld 2003), and (heterosexual) gender relations as drivers of demographic change (Esping-Andersen and Billari 2015; Goldscheider et al. 2015). Among these drivers, only ideational shifts help to understand demographic change among LGBs. The increasing prevalence of same-sex unions and families can be perceived as societal changes toward self-realization (Rosenfeld and Kim 2005). However, this general understanding does not predict what type of family trajectories exist among LGBs and how different family trajectories have changed over time as opportunities to form a family expanded. To illustrate, rather than retreating from institutions like marriage and parenthood, the LGB population has made advancements toward accessing these institutions in recent decades in many countries. As a result, demographic change within the LGB population will likely look very different from that of the heterosexual population.
To empirically document demographic change through life courses, past research has quantified change in the type, complexity, and diversity of family trajectories in the general population (Brückner and Mayer 2005; Elzinga and Liefbroer 2007; Van Winkle 2018). Such empirical approaches could be more easily applicable to LGBs than theories of demographic change. Past studies found that family trajectories in the general population became more diverse across individuals. The hegemonic trajectory of early lifelong marriage and parenthood gave way to alternative trajectories, including cohabitation, separation, repartnering, and single parenthood (Elzinga and Liefbroer 2007; Van Winkle 2018). The evidence of cohort changes in the complexity of individual trajectories (i.e., the number of family events experienced) in the general population is mixed. McMunn et al. (2015) found that partnering trajectories became more eventful with increases in repartnering, while parenthood trajectories became less complex over time in the UK. Van Winkle (2018) found that family trajectories became more complex across cohorts in various countries, but cohort changes were relatively minor compared with cross-national differences.
In the case of the LGB population, the option to marry and have children through pathways such as adoption and ARTs could increase the diversity and complexity of family trajectories across cohorts. This can especially be the case if some of the LGB population do not follow heteronormative family trajectories, whereas others embark on more heteronormative pathways of family formation that were previously unavailable (Bosley-Smith and Reczek 2018; Hank and Wetzel 2018; Riskind and Patterson 2010). However, whether complexity and diversity in family trajectories between individuals increased depends on how diverse and complex they were in older cohorts. It is unclear how typical or dominant family trajectories, such as lifelong singlehood, were among LGBs in the past.
In short, previous theoretical perspectives of demographic change have focused on majority populations and population averages, while empirical accounts summarizing demographic change have not yet studied the LGB population. Therefore, in the remainder of the article we present an analysis that directly acknowledges and focuses on the contextually unique experience of minoritized groups (Johnson-Hanks et al. 2011; Zaidi and Morgan 2017).
The UK's fertility and union dissolution rates are higher than those of other European countries but lower than those of the United States (Lesthaeghe 2010; Musick and Michelmore 2018). The environment for LGB individuals has changed rapidly and positively over time. In 2005, civil partnerships became available to same-sex couples. The same year, adoption became available to same-sex couples and single individuals in England and Wales (2009 in Scotland and 2013 in Northern Ireland). Sexual orientation was incorporated into antidiscrimination laws in 2007 and 2010. In 2009, ARTs and the possibility of having two mothers on a birth certificate became available. In 2014, same-sex marriage was legalized in England, Wales, and Scotland (2020 in Northern Ireland). Attitudes toward sexual diversity are relatively positive in the UK compared with other countries (Adamczyk and Liao 2019).
We compare LGBs born before 1965 to LGBs born between 1965 and 1979. The older cohort reached age 40 before significant legal changes took place. The younger cohort saw these changes unfold during early or mid-adulthood and could, to some extent, have taken advantage of the changing possibilities to form unions and transition into parenthood. Younger cohorts born after 1979 are excluded from our analysis because they have yet to complete their prime family formation years, albeit being the cohorts to experience the most extensive range of possibilities yet. Nonetheless, our analysis provides the first benchmarks of family trajectories for LGB individuals.
Data and Sample
We use the Understanding Society survey data from 2009 to 2019. UKHLS is a representative household panel survey of the UK population (Buck and McFall 2011) and is one of the very few large-scale surveys to collect information about sexuality as well as partnership and parenthood histories. These unique features allow us to identify LGB individuals on the basis of their self-reported sexual identity instead of inferring it from the gender of their partner and to reconstruct their histories of coresidence with partners and children. We include respondents who were present in at least one wave in which complete retrospective histories were collected (Waves 1 and 6)2 as well as one of the waves in which sexual identity was reported on (Waves 3 and 9).3 Subsequently, we exclude all individuals who had not reached age 40 by the date of the last interview, as our analysis relies on analyzing complete partnering and parenthood histories. We also exclude cases with more than six years of missing partnership and parenthood histories or missing information about age and sexuality. The main analytic sample includes 455 LGB individuals (45% women) observed between ages 18 and 40 and born before 1979 (50% born by 1964 and 50% born between 1965 and 1979).
The Understanding Society survey asked respondents, “Which of the following options best describes how you think of yourself?” The answer options were heterosexual or straight, gay or lesbian, bisexual, other, prefer not to say, and don't know. Sexual identity is a time-varying characteristic but is measured only at two points in time in our data. Our analysis, therefore, documents differences in family trajectories based on sexual identity measured in the last wave with nonmissing information.4 We restrict the analysis to two mutually exclusive groups: bisexual (n = 176) and gay/lesbian (n = 279) individuals.5 Gender/sex was collected with the question, “And you are male/female?” thus preventing us from considering other gender identities or from comparing cis- and transgender persons.
Union formation, dissolution, and partnership spells are based on self-reported coresidence with a partner of any gender.6 We do not have information about nonresident partnerships. In Waves 1 and 6, respondents were asked to retrospectively report the starting and ending dates of up to 12 coresidential relationships (marriages, civil partnerships, or cohabiting unions) that lasted at least three months. This information was updated and harmonized with prospective information collected annually in the subsequent waves (Nandi et al. 2020). We use these retrospective histories on coresident partners to construct a person-month file with a dummy variable indicating whether, in that month, the person coresided with a partner.7
Parenthood is recorded from reports about coresidence with children younger than 18. Respondents were asked in Waves 1 and 6 whether they were ever the parent of an adopted, biological, or stepchild, and this information is updated in each wave. For biological children, the survey collects the year of birth and the year the respondent last lived with the child. For adopted and nonbiological children, the survey collects the year the respondent started and stopped living with the child (if applicable). We assume respondents started living with biological children in the year of birth (if they ever lived with their biological child) and create a dummy variable of coresidence with a child in each year of the person-year file accordingly. Focusing on coresidence allows us to capture spells of single parenthood and residence with nonbiological children. However, it does pose some limitations, namely, that it does not account for nonresident children. In an additional analysis, we show that parenthood trends are similar when also considering nonresident biological children (see Table S2 in online Supplementary Materials B).
More generally, we acknowledge that our family states, that is, coresidence with partners and children, do not capture a wider range of family roles and structures in the LGB population. Given sample size limitations, our family sequences cannot differentiate between marriage and cohabitation for those partnered or between biological and nonbiological children. We provide the prevalence of ever having resident or nonresident (biological) children as part of the descriptive statistics to offer a better characterization of family trajectories among sexual minorities. We keep individuals with missing information on partnering or parenthood for part of the observation period and include “missing” as an additional state in their family trajectories.8
We use sequence analysis, in which a person's family trajectory is conceived as a succession of family states during the prime family formation period. In our analysis, each respondent's sequence consists of a series of monthly or yearly states of coresidence with partners and children starting from age 18. Four different family states are observed: (1) single without coresident children, (2) single with coresident children, (3) partnered without coresident children, and (4) partnered with coresident children. Our operational definition of a family sequence enables us to assess multiple partnership episodes (i.e., union formation, dissolution, and repartnering), the transition into parenthood, and their timing and sequencing within a person's family trajectory.
Figure 1 shows three fictional sequences of family states over a 10-year period. In this example, S1 depicts a trajectory with no family transitions, wherein a person does not live with children or a partner throughout the observation period; S2 features a transition to single parenthood in the sixth observation; and S3 features the start of a partnership in the second observation, and the presence of a child since the seventh observation.
On the basis of these sequences, we generate a typology of underlying family trajectories to answer the first research question.9 First, we compare each pair of sequences in our sample and calculate their similarity or optimal matching (OM) distance.10 These pairwise OM distances are based on criteria that account for differences in timing, sequencing, and duration of the different states observed across sequences (Aisenbrey and Fasang 2010). Second, we use hierarchical cluster analysis with the Ward link to cluster similar sequences into a typology of trajectories or profiles. We chose a five-cluster solution among the available solutions given by the cluster analysis because of its empirical fit to the data at hand and because it renders a set of theoretically meaningful pathways (see the cutoff criteria for different cluster solutions in Figure S1 in online Supplementary Materials C). Even though sequences that belong to the same group are not identical, they conform to a general trajectory pattern, allowing us to better categorize family trajectories than what would be obtained by using simple indicators of the prevalence of events.
We explore cohort change through multinomial regression analysis predicting cluster membership by birth cohort. We measure the birth cohort as a dummy variable that compares LGBs born before 1965 and between 1965 and 1979. We control for gender, sexual identity, education, race-ethnicity, and family structure at age 16. We also present predicted probabilities of cluster membership by gender and sexuality while controlling for the same covariates. In a supplementary analysis, we calculate composite measures of changes in the complexity and diversity of sequences used by previous empirical research (see online Supplementary Materials D for a more technical explanation and results).
We present results only for LGBs in the main text because they are the focus of our analysis. We refer to heterosexuals when it is useful for comparative context and present results for heterosexuals in the online Supplementary Materials.
Descriptive Family Formation Statistics
Figure 2 shows the distribution of the four family states at every age between 18 and 40 for LGB women and men. The graph shows that singlehood, that is, not living with a partner or children, is a prominent state for LGBs across all ages in our sample. About 40% of LGB sample members were not living with a partner or children at age 40, compared with about 10% of the heterosexual sample members (see Figure S2 for heterosexuals' state distribution in online Supplementary Materials E). Being partnered with children is also a prominent state of LGBs' life courses, but being in a union without children is more common (in contrast to heterosexuals, see Figure S2). Moreover, having children within partnerships occurs late in the life course. Finally, single parenthood is the least common state. When it does occur, it is later in the life course.
Family Formation Trajectory Typology
Figure 3 presents results for the cluster analysis that produced five clusters of family formation trajectories among LGB women and men in the sample: (1) mostly single, no children (n = 172, 38%); (2) mostly partnered, no children (n = 128, 28%); (3) early partnering and parenthood (n = 87, 19%); (4) delayed partnering and parenthood (n = 48, 11%); and (5) single parenthood (n = 20, 4%).
Two thirds of the sample belonged to Clusters 1 and 2, which are characterized by not residing with children but differ in partnering patterns. The largest cluster in the sample is Cluster 1, with 38%; it is characterized by stable singlehood or short partnership spells, mostly between ages 24 and 34. A significant share of the trajectories in this cluster (58%, see Table 1) had a partnership at least once by age 40, but these partnerships were short-lived (28 months on average). Twenty-eight percent of the sample belonged to Cluster 2, which is characterized by long partnership spells after age 30 or earlier partnering and repartnering throughout the observed life course.
A third of the sample was distributed across Clusters 3, 4, and 5, characterized by coresident parenthood but differing partnering patterns. Nineteen percent belonged to Cluster 3, characterized by early partnering and parenthood transitions by the mid-20s. Another 11% belonged to Cluster 4, with delayed partnering and parenthood transitions. Specifically, this cluster includes family formation trajectories that start with early union formation and later parenthood transition or delayed union formation with a transition to parenthood soon after. Finally, a small group of LGB sample members (4%) belonged to Cluster 5, characterized by long single parenthood spells. Because of the small sample size, we combined Clusters 3, 4, and 5 into one group of parenthood clusters for the rest of the analysis.
What sociodemographic characteristics are associated with belonging to each cluster? Table 1 shows bivariate descriptive sociodemographic characteristics for each of the three (groups of) clusters: mostly singlehood (labeled “singlehood” hereafter for brevity), mostly partnership (“partnership” hereafter), and coresident parenthood (“parenthood” hereafter).
Overall, women and bisexual people were underrepresented in the singlehood cluster but overrepresented in the parenthood clusters. White LGBs were relatively less likely to belong to a parenthood cluster. LGBs whose parents were not together when they were 16 years old were more likely to be in the singlehood cluster, whereas LGBs whose parents were together were more likely to be in the partnership cluster. Overall, there are no notable differences in education across the clusters in our sample.
Table 1 also shows that more than half of the people in the singlehood cluster have had a coresident partner, but as Figure 3 showed, most of these partnerships were short-lived. On average, the number of partners was similar among LGBs in the partnership and parenthood clusters (1.5 and 1.4 partners, respectively). More than half (57%) of the people in the parenthood clusters have ever had nonresident minor children, in contrast to 6% and 4% among LGBs in the singlehood and partnership clusters, respectively. This striking contrast implies that a very small proportion of LGBs in these clusters were ever parents.
Cohort Change in Family Formation Trajectories
Cluster Membership Across Cohorts
We first examine how the two LGB cohorts distribute across clusters to explore cohort change in family trajectories. Table 1 shows that LGBs born between 1965 and 1979 were more likely to belong to the partnership cluster than LGBs born before 1965. Table 2 shows results for a multinomial regression analysis predicting cluster membership. These results confirm that the younger LGB cohort was more likely than the older cohort to experience partnership trajectories than parenthood or singlehood trajectories.
Figure 4 gives better insight into absolute changes in the prevalence of family trajectories and how these changes differed by gender and sexual identity (using a three-way interaction of gender, sexual identity, and cohort to predict the probability of belonging to each of the three groups of clusters). We observe that both gay men and lesbian women have become considerably more likely to follow partnership trajectories across cohorts. This has come at the expense of trajectories characterized by parenthood and singlehood, although few gay men followed parenthood trajectories in the older cohort. Among gay men, singlehood was clearly the most common family trajectory type among the older cohort (60%), but partnership trajectories were most common (46%) among the younger cohort. For lesbian women, both parenthood and singlehood trajectories were prevalent among the older cohort (39% and 40%, respectively), yet partnership trajectories became the most common trajectory among the younger cohort (51%).
These results align with the expectation that partnership trajectories will become more prevalent across cohorts as the opportunities to meet same-sex partners increase and the stigma and discrimination decline over time. However, our results regarding parenthood suggest that only relatively small shares of the cohorts studied have taken advantage of increasing access to alternative routes into parenthood, such as ARTs and adoption.
Results for bisexual women and men show different patterns of change than those observed among gay/lesbian individuals. Although the sample size calls for caution in interpretation, the descriptive patterns among bisexuals align with general narratives of demographic change. Bisexuals from the younger birth cohort in our sample have experienced a decline in parenthood and partnership trajectories but an increase in singlehood trajectories. This change was more substantial among bisexual men than bisexual women.
Additional Analysis: Measures of Diversity and Complexity
Empirical studies on demographic change often provide summary measures of how eventful family trajectories are (complexity) and how diverse trajectories are across individuals (diversity) to quantify other forms of change in life trajectories. Figures 5 and 6 illustrate such measures for our sample, and we briefly summarize the main takeaways here (see Supplementary Material D for more methodological information and elaborated results).
Overall, we find that family trajectories of lesbian women and bisexual men became less diverse over time, meaning their family pathways are becoming more similar to each group's representative trajectory (Figure 5). This result does not imply that all lesbian women and bisexual men follow the same family formation pathways, nor that the subjective experience of similar pathways is alike. However, regarding coresident partners and children, the sequences of younger lesbian women and bisexual men in our sample were more likely to be similar to one another than the sequences of their peers born before 1965.
Results for complexity—that is, the number of events (transition from one state to another) people experience—show a relatively uniform increase across cohorts for all groups (Figure 6). This probably reflects the replacement of low-complexity trajectories, such as those characterized by singlehood, with trajectories characterized by partnership.
Research on demographic change has mostly overlooked the family trajectories of sexual minorities (Badgett et al. 2021; Perales 2016; Russell et al. 2020). Existing empirical research has yet to document how partnerships, parenthood, and singlehood evolve over the life course of sexual minorities. Similarly, existing theoretical narratives of demographic change focus on the general population and can, at best, explain why same-sex unions have become more visible in recent decades. These narratives provide little understanding of demographic change among sexual minorities in a context where possibilities to pursue partnerships, marry, and have children within same-sex relationships have expanded. In this article, we provide what is, to our knowledge, the first quantitative description of family life courses among LGB persons and the first assessment of how family trajectories changed across two cohorts of LGBs in the UK. Our results provided several novel empirical observations.
First, we identified five distinct profiles of family trajectories between ages 18 and 40. Trajectories characterized by singlehood or partnership without children were the most prevalent. Previous cross-sectional research in the United States demonstrated that LGBs were less likely to have a partner at a given time than heterosexuals, except for lesbian women (Badgett et al. 2021). We add to this body of work that a considerable share of LGBs born before 1979 did not coreside with a partner before age 40 (or experienced very short partnership spells) in the UK. For some groups, these shares were substantial. For instance, 60% of gay men born before 1965 followed a trajectory that mainly consisted of singlehood. This pattern means that empirical studies based on same-sex couples omit a large part of the LGB population, especially among older cohorts of gay men.
We also found some differences compared with studies from the United States (Badgett et al. 2021). For example, trajectories characterized by singlehood were more prevalent among lesbian women than among bisexual women in the UK. We also found that gay men were more likely to follow singlehood trajectories than were bisexual men. Both results differ from patterns observed in the United States, where bisexual individuals were more likely to be single than were gay/lesbian individuals (Badgett et al. 2021). This divergence could be because we take a life course rather than a cross-sectional approach. This divergence highlights the importance of a life course perspective in understanding family formation processes among LGBs.11 However, we also observed that singlehood trajectories became more common among bisexuals across cohorts, an issue we elaborate on later.
Our second novel observation is the increasing prevalence of trajectories characterized by coresidential partnerships among the younger gay and lesbian individuals in our sample (born between 1965 and 1979). This trend was expected amid the increasing availability and visibility of potential LGB partners, the decreasing stigma and discrimination toward same-sex couples (Rosenfeld 2017; Trandafir 2015), and LGBs' expressed interest in having a partner (Barrios and Lundquist 2012; D'Augelli et al. 2007; Riskind and Tornello 2017; Tate et al. 2019). Which mechanism specifically drives this trend is a question for future research. However, if the underlying mechanism is the decreasing stigma and institutional discrimination, we expect that singlehood trajectories will continue to be highly prevalent among gay/lesbian individuals in contexts where stigma and discrimination are higher than in the UK. Nonetheless, our results for gay and lesbian individuals contradict predictions derived from theories of demographic change. The quest for independence is central to the second demographic transition framework, which should manifest in more singlehood trajectories across cohorts rather than a decrease. Therefore, to understand demographic change among LGBs, demographers should also consider the diversity of family formation experiences within societies.
The results for bisexuals, however, were consistent with the trend in the general population. Bisexuals in the younger birth cohort, especially bisexual men, were more likely to experience a singlehood trajectory than the older birth cohort. Barriers to same-sex partnering restricted options for both bisexual and gay/lesbian people. However, the availability of different-sex partnering could explain why bisexuals in the older cohort were more likely to follow a partnering or parenthood trajectory than gay and lesbian people. As some barriers to same-sex partnering have been reduced, younger bisexual men could opt out of different-sex relationships as a form of independence.
However, the decline in bisexuals' partnership trajectories could also reflect variation in stigmas that different sexual minority groups experience amid the second demographic transition. For instance, bisexuals experience double erasure by both heterosexuals and other sexual minorities, which reduces their partnering opportunities (Morgenroth et al. 2022). This biphobia impacts bisexual men more than bisexual women, which could explain the more pronounced increase in singlehood trajectories among bisexual men. In other words, although bisexuals' trends align with trends in the general population, they could be driven by different societal forces. Hence, our results underscore the importance of understanding the diversity of family trajectories across different sexual minority groups in future research.
A third novel empirical observation was that a nonnegligible but relatively small share of UK LGBs experienced life courses that included prolonged spells of coresidence with children. This share was smallest among gay men and largest among bisexual men and women. Parenthood trajectories became slightly less common across cohorts for gay men, bisexual men, and bisexual women, but a pronounced decline was observed among lesbian women. This trend aligns with evidence of a decline in same-sex couples' likelihood to coreside with children born in previous different-sex relationships (Gates 2015). This explanation is plausible because the most notable decline in our data was among lesbian women, who are more likely to live with children from past different-sex relationships than are gay men.12
Nonetheless, the decline in parenthood trajectories is paradoxical amid relative expansion in access to planned pathways to parenthood for same-sex couples, including adoption and ARTs. What could explain this paradox? First, younger LGBs might be more likely to reject heteronormative family trajectories (Hank and Wetzel 2018; Riskind and Patterson 2010). Second, planned pathways to parenthood might be out of reach for younger LGBs. Despite significant progress in ARTs, the procedures are costly (Cozzani et al. 2021; Park et al. 2020). Moreover, subsidy eligibility is determined by heteronormative criteria, such as not conceiving after having had unprotected intercourse for at least a year. Consequently, women in same-sex couples are often denied subsidies unless they try to conceive through privately funded insemination donors or have undergone an infertility test (Warnes 2019). Similarly, although adoption became available to same-sex couples in the 2000s, rates are still low and same-sex couples are overrepresented among the couples who adopt hard-to-place children in the UK (Tasker and Bellamy 2019). Future research should investigate cohort change in LGBs' pathways to parenthood across more and less favorable contexts than the UK.
The decline in parenthood trajectories also raises questions about LGBs' future care needs as they age. Sexual and gender minorities have historically relied on chosen families and their communities for care support owing to individual and institutional discrimination (Allen and Lavender-Stott 2020; Boyd and Wei forthcoming; Jackson Levin et al. 2020; Reczek 2020). However, amid population aging and increasing demand for family-provided care, an additional unmet need for care could emerge among LGBs of the younger cohorts if they continue to be less likely to experience parenthood trajectories. Future research should investigate how family-related changes affect different sexual minority groups' access to care support over time.
Besides describing family trajectory types, we also explored the complexity and diversity of LGBs' family trajectories, which is another pillar of demographic change research. Previous research argued that family trajectories have become more eventful (i.e., more complex) and diverse (i.e., more heterogeneous) among the general population (Van Winkle 2018). However, our results were mixed. On the one hand, LGBs' family trajectories became more complex across cohorts owing to a decrease in singlehood trajectories. On the other hand, the trend in diversity, that is, the extent to which people's trajectories are different from a single dominant trajectory, was different across groups. Lesbian women's trajectories became slightly less diverse, while gay men's trajectories became slightly more diverse across cohorts. Both groups experienced increases in partnership trajectories that increased diversity, and both groups experienced declines in parenthood trajectories that decreased diversity. However, for lesbian women, the decline in parenthood trajectories was much more substantial, leading to an overall decline in diversity.
Our study has several limitations that offer directions for future research. First, our small sample limited our ability to break down trends by education, race, ethnicity, or more detailed cohorts. In addition, our LGB sample could only identify as women or men, and sexual identities beyond LGB were not available as answer options. Nonetheless, using sexual identity measures is one step forward beyond existing practices in family scholarship that focus on same-sex couples. The sexual identity measure allowed us, for instance, to uncover differences between bisexual and gay/lesbian individuals and to look at LGB individuals who are single or single parents.
Second, our conceptualization of family formation is limited to coresidence with partners and children, representing a heteronormative definition of family life (Reczek 2020; van Eeden-Moorefield et al. 2011). Our approach does not include nonresident partners and other forms of partnerships and relationships that could be more prominent in LGBs' perception of a family (Allen and Lavender-Stott 2020; Reczek 2020; van Eeden-Moorefield et al. 2011). However, we demonstrated a cohort change within the scope of coresidential familial relationships. These coresidential trends are essential to document amid existing social and institutional barriers. Moreover, our sequence analysis allowed us to focus on the diversity and complexity of family trajectories amid a small sample. Future studies should investigate further the family dynamics of sexual minorities over the life course, namely, the life courses of younger LGB individuals born after 1980, who are in the process of family formation within a rapidly changing context.
Despite these limitations, we expanded existing research about family formation among sexual minorities by taking a novel life course approach that centers on the dynamic, complex, and diverse ways that partnership and parenthood evolve over the life course of LGB women and men. This study contributed by documenting cohort change in family formation among LGBs and how these changes are prolonged and unequal despite legal expansions and social changes. Our results have consequences for general narratives of demographic change over time. Whereas changing social norms have made it easier for significant parts of the general population to “deviate” from the normative trajectory of early family formation, the same social changes might have facilitated coresidential partnering trajectories for the LGB population. We also expected parenthood trajectories to become more common across cohorts for the LGB population but found the opposite for the cohorts studied here. This seems to imply that sexual minorities continue to face obstacles to access alternative routes into parenthood, such as adoption and ARTs. These results illustrate that demographic change manifests differently for different groups within a society. Hence, existing narratives of family-related demographic change should explicitly consider how different groups within societies, such as sexual minorities, experience and react to social and demographic changes.
Earlier versions of this manuscript were presented at the 2022 annual meeting of the Population Association of America (Atlanta, GA) and the 2022 European Population Conference (Groningen, the Netherlands). The research has benefited from funding from the European Research Council (ERC-2020-STG-948557-MINEQ; PI: Diederik Boertien). Ariane Ophir (FJC2021-048042-I) and Diederik Boertien (PID2021-124267OB-I00; RYC2019-026510-I) also acknowledge funding from the Spanish Ministry of Education.
Unfortunately, we do not have data on other sexual identities, such as asexual, pansexual, or queer, nor on gender minorities, such as transgender and nonbinary people, thus preventing us from studying other groups within the LGBTQ+ population.
Even though new respondents in other waves were asked about their relationship and parenthood histories, this information is not sufficient to reconstruct coresidence histories across the life course.
We also include new respondents who were asked about sexual identity in Waves 5 and 7.
Note that people’s sexual identity might have changed during the observation period, but we are unable to incorporate such changes in the main analysis. Table S1 (in the online Supplementary Materials A; all tables and figures designated with an “S” appear in the online supplements) shows that results are robust when excluding 13 respondents who reported different sexual identities across waves.
The “other” category is complicated to interpret as previous research suggests that this category includes both nonheterosexual individuals and heterosexual individuals who do not understand the question or response set (Elliott et al. 2019). In our data, 75% of individuals identifying as “other” in Wave 3 identified as heterosexual in Wave 9. Eighty-seven percent of persons identifying as gay/lesbian in Wave 3 also identified as such in Wave 9, whereas the corresponding number for bisexual individuals is 45%.
Note that we cannot distinguish different-sex from same-sex partnerships because this information is not collected retrospectively. Therefore, we cannot estimate changes in the prevalence of different- versus same-sex partnerships across cohorts (Gates 2015).
For part of our analysis, we use person-year files based on the status at the start of each year. In particular, our analyses based on optimal matching (i.e., typology of trajectory pathways and diversity index) deploy sequences of yearly states to obtain meaningful results (i.e., distances between sequences) since there are few changes in family states in our sequence data, and multiple changes occur rarely within the same year.
This applies to missing dates at the start or end of the observation period (i.e., starting date of the first union/child or the end date of the last union/child). For other missing dates, we interpolated dates by taking the middle point between the previous and subsequent nonmissing dates. If two subsequent dates were missing, we interpolated by dividing the period between the previous and subsequent nonmissing dates into three. We exclude cases with more extended periods of missing data.
We use the software R and the TraMineR package for optimal matching and cluster analyses (Gabadinho et al. 2011).
The OM distance (or dissimilarity) between two sequences is the minimum cost of transforming one sequence into another sequence by means of substitutions between elements of the sequence, or their insertions and deletions (Abbott and Tsay 2000). We use the OM-Future algorithm to emphasize that relevant patterns in family trajectories substantively differ in the timing of key transitions across family states. The OM-Future variant considers that states in a given position in the sequence are more similar if they share a common future state (Studer and Ritschard 2016). This renders an OM distance more sensitive to the timing when transitions occur. Since operation costs are derived from state distributions, our distance of choice is also less affected by the criticism that cost assignment is arbitrary and has no sociological meaning.
To illustrate, if we cross-tabulate partnership status with sexual identity using the 2017/2018 wave, differences between heterosexual and lesbian women disappear: 54% of lesbian women are in a coresidential relationship compared with 59% of bisexual and 57% of heterosexual women. However, for bisexual men, the relatively high rates of partnering persist, with 37% of gay men in a coresidential relationship compared with 49% of bisexual men and 64% of heterosexual men.
In online Supplementary Materials, we show the percentages of ever having a biological child regardless of coresidence. These percentages are higher, especially among gay men in the older cohort. Nonetheless, the supplementary analysis also shows a decline in the prevalence of parenthood across cohorts when using this measure.