Same-Sex Sexuality and the Duration of First Different-Sex Marriages

Recent research has focused on the once-married and associations between various aspects of same-sex sexuality (i.e., desire/attraction, behavior and identity) and divorce from a different-sex spouse. In this paper, we theorize that same-sex sexuality could be associated with either shorter or longer marital duration, and we use data from the 2011-2013 National Survey of Family Growth (NSFG) to examine the associations between three aspects of same-sex sexuality and marital duration among those who married and divorced once (N=617). Among the once-married/divorced, same-sex sexuality substantially reduces marital duration by approximately 18-24 months, on average, net of other variables. Supplemental analyses indicate that these associations do not differ by sex/gender but do differ in nuanced ways for individuals who are bisexually attracted (identified) compared to those who report, respectively, same-sex only or different-sex only attraction (heterosexual, bisexual and lesbian/gay identity). Between-group differences in age at marriage exert a substantial influence on the associations between same-sex sexuality and marital duration documented in the supplemental analyses of bisexuality. We discuss our findings in relation to the extant literature, the limitations of available data, and directions for future research.


Introduction
Demographers, sociologists, and other family scholars have identified several factors that increase the risk of divorce among different-sex couples: having parents who divorced (Amato 1996;Amato and Cheadle 2005;Wolfinger 2005), premarital cohabitation (Booth and Johnson 1988;Bumpass, Sweet, and Cherlin 1991;Schoen 1992), early entry into marriage (Bumpass, Martin, and Sweet 1991;Raley and Bumpass 2003), extramarital sex (London, Allen, and Wilmoth 2013), lower socioeconomic status (Cherlin 2009;Martin 2006;Raley and Bumpass 2003) and being African American (Cherlin 2009;Kposowa 1998;Raley and Bumpass 2003), among others. Our research provides evidence that, among the once-married, same-sex sexuality is also associated with an increased risk of different-sex divorce (London and Hoy 2017). In well-controlled models estimated on nationally representative data collected in 1992 and during the 2011-2013 period once-married individuals who report experiencing any of the three components of same-sex sexuality -desire/attraction, behavior and/or identity (Laumann et al. 1994) -are more likely to also report that their different-sex marriage ended in divorce.
To date, no research of which we are aware examines whether same-sex sexuality affects different-sex marital duration. It is possible that same-sex sexuality could be associated with either shorter or longer marital duration. On the one hand, individuals who report same-sex sexuality may be invested in maintaining a different-sex marriage, especially if the marriage was initially contracted to avoid the negative social consequences that lesbian, gay and bisexual people often encounter. In addition, many individuals who experience same-sex sexuality married prior to the recent liberalization of attitudes toward homosexuality (see Loftus 2001). By the time social conditions became more favorable for lesbian, gay and bisexual people, their different-sex marriages had become well-established, and they likely faced substantial barriers to ending them, or at least ended them at longer durations. On the other hand, same-sex sexuality within the context of a different-sex marriage remains "an ideological impossibility" (Wolkomir 2009:504). Different-sex couples who confront one or both spouses' same-sex sexuality may struggle briefly to maintain the marriage, but ultimately divorce more quickly. Of course, it is also possible that countervailing influences or mechanisms could result in no (or different) net associations between the components of same-sex sexuality and different-sex marital duration.
In this paper, we use data from the 2011-2013 National Survey of Family Growth (NSFG) to extend our prior research by examining the associations between three aspects of same-sex sexuality and first different-sex marital duration. Specifically, we focus on the once-married, once-divorced and ask: Is same-sex sexuality associated with shorter or longer marital duration, on average, net of demographic and background variables? In our primary analyses, we focus on all three dimensions of same-sex sexuality using dichotomous indicators; however, in supplemental analyses that focused on two of these three dimensions, we specifically address the question of whether differences exist between those reporting exclusively same-sex versus bisexual attraction and identity. Finally, although our primary analyses control for the main effect of sex/gender, in supplemental analyses we also consider whether sex/gender moderates observed associations between same-sex sexuality and marital duration.

Same-sex sexuality, different-sex marriage, and the risk of divorce
Our recent research using high-quality, nationally representative data collected in 1992 and during the 2011-2013 period shows that same-sex sexuality significantly reduces the likelihood of entering into a different-sex marriage (London and Hoy 2017). For example, analyses of the 2011-2013 NSFG show that those who reported same-sex attraction, behavior and identity were significantly more likely than those who did not to have never married (and therefore to not be at risk of divorce), with adjusted odds ratios (AORs) of 1.66 (p < 0.001) for the comparison of those who reported attraction relative to those who did not; 1.67 (p < 0.001) for those who reported same-sex sexual behavior versus those who did not, and 2.33 (p < 0.001) for those who reported lesbian, gay or bisexual identity versus those who did not. Still, some individuals who report same-sex sexuality have at some point contracted a different-sex marriage. Although it is possible that both spouses experience same-sex sexuality in some of these marriages (i.e., both spouses may be bisexual), it is more likely that most are what scholars generally term "mixedorientation marriages" (e.g., Buxton 1994;Hernandez, Schwenke, and Wilson 2011). That is, most are marriages in which one spouse is exclusively sexually oriented toward different-sex individuals, while the other spouse is at least partially, if not exclusively, oriented toward individuals of the same sex. In our analysis below, which uses individual-level rather than couple-level data, we are unable to specify whether a respondent's spouse is also partially or exclusively sexually oriented toward those of the same sex. Nevertheless, for the purposes of theorizing, we assume that most respondents were in mixedorientation marriages.
It is difficult to know how many mixed-orientation marriages exist, but Buxton (1994Buxton ( , 2001 estimates that in the United States approximately 2 million currently exist. According to recent analyses of NSFG data, among the once-married, some of whom had divorced, 9.1% report experiencing same-sex attractions, 9% report having engaged in same-sex behaviors, and 3% identify as lesbian, gay or bisexual (London and Hoy 2017). However, it is important to note that the timing of same-sex sexuality in relation to the different-sex marriage is generally not known; it may be the case that for some individuals, samesex sexuality emerges only after the divorce has taken place.
Many individuals who enter mixed-orientation marriages experienced same-sex attractions prior to marrying (Stein 1997), and some continue to experience such attractions while married (Bozett 1982;Higgins 2002;Pearcey 2005;Wyers 1987). This pattern holds in terms of behavior, too. Approximately 27.5% of the men in Higgins' (2002) convenience sample engaged in some form of sexual behavior with another man prior to marrying a different-sex spouse, and a similar percentage (28.9%) continued such behaviors while married. In Humphreys' (1970) classic ethnography of anonymous public sex among men, 54% of the men observed were in different-sex marriages. To be sure, few people identify as lesbian, gay or bisexual before entering into a different-sex marriage; however, some do develop such an identity while married to a different-sex spouse (Bozett 1982;Wyers 1987).
Individuals who experience same-sex sexuality early in the life course sometimes contract a different-sex marriage to conform to the pressures of a heteronormative society. Heteronormativity refers to the interwoven set of assumptions, practices and policies that make heterosexuality and the nuclear family appear both natural and normal (see Kitzinger 2005;Martin 2009). Those who defy heteronormative expectations often face negative social consequences for doing so; most are marginalized or even ostracized, while some also face discrimination and violence (see Meyer 2015;Schilt and Westbrook 2009;Tilcsik 2011). Not surprisingly, some individuals who experience same-sex sexuality contract different-sex marriages to appear "normal" and to avoid such consequences (Higgins 2002;Humphreys 1970;Pearcey 2005;Ross 1983). Individuals who became aware of their same-sex sexuality in the mid-20 th century may have been especially likely to marry a different-sex spouse given the intensity of heteronormative expectations during that time (Humphreys 1970;Seidman 2002). Of course, some individuals contract different-sex marriages simply because they want to marry (Stein 1997). Others do so because they want to have children, which they believe is culturally appropriate only within the context of marriage (Higgins 2002). Until recently, marriage was only legally available to different-sex couples. As Higgins (2002:28) explains, these individuals marry because they believe different-sex marriage is natural and normal rather than as a strategy to avoid penalty or sanction for their sexuality. Also, it is worth noting that individuals who are bisexual in their attractions, behaviors and/or identity are especially likely to cite such reasons for marrying (Buxton 2001(Buxton , 2004a(Buxton , 2004bHernandez et al. 2011).
For those who experience same-sex sexuality, marriage to a different-sex spouse is more likely to end in divorce (London and Hoy 2017). Indeed, all three components of same-sex sexuality -desire/attraction, behavior and identity -are associated with an increased risk of divorce from a different-sex spouse, net of a range of demographic and background characteristics. When an individual learns of his or her spouse's same-sex sexuality, either through disclosure or discovery, a negotiation process usually follows in which couples must decide whether to remain married (Wolkomir 2009). Because individuals in different-sex marriages are expected to be heterosexual and marriages are expected to be unique and lifelong, for some, same-sex sexuality and different-sex marriage are an "ideological impossibility" (Wolkomir 2009: 504). In the end, most such couples divorce.
Importantly, not all mixed-orientation marriages end in divorce (Buxton 2001(Buxton , 2004a(Buxton , 2004bSchwartz 2012;Yarhouse, Gow, and Davis 2009). In particular, those in which the non-heterosexual spouse is bisexual, rather than gay or lesbian, seem better able to endure the stresses of disclosure or discovery. For instance, even after informing their wives of their same-sex sexuality, half of all bisexually-identified husbands surveyed by Edser and Shea (2002) described their marriage as good, with another 30% saying it is "workable." Furthermore, most reported feeling little to no sense of loss being in a monogamous different-sex marriage, and 90% reported investing in and working to improve their marriage (Edser and Shea 2002). Mixed-orientation marriages in which one spouse is bisexually-identified are more likely to last in large part because it is possible for the couple to maintain their mutual attraction and to continue having a satisfying sex life (Buxton 2001(Buxton , 2004a(Buxton , 2004bHernandez et al. 2011). For these reasons, it is important for research on mixed-orientation marriages to distinguish between persons who are sexually oriented toward both sexes and persons who are exclusively oriented toward individuals of the same sex. As discussed in greater detail below, we conduct analyses that address this distinction. Although quantitative studies have been able to document associations between same-sex sexuality and divorce (London and Hoy 2017), it is important to note that the existing research has not been able to empirically specify when same-sex sexuality emerges in relation to divorce. Almost all of the available data with large enough samples and measures of key variables are cross-sectional and include no information on timing and sequencing. It may be the case that some aspects of same-sex sexuality precede marriage, emerge during it, or are realized only after the divorce has occurred. Possibly, for some, samesex sexuality results from opportunities that are engendered by a divorce. Such circumstances could contribute to an association between same-sex sexuality and marital duration, although it is not the primary source of the association we theorize here.

Different-sex marriage and marital duration
How long different-sex marriages last, and when they start and end, have long been of central concern to demographers, sociologists and other family scholars. Notably, the percentage of different-sex marriages that last 10 or 20 years, respectively, has not changed significantly since the 1970s. Using 2006-2010NSFG data, Copen, Daniels, Vespa, and Mosher (2012 report that the probability of a first marriage lasting at least 10 years is 68% for women and 70% for men, while 52% of women and 56% of men have marriages that last 20 years or longer. Each of these estimates is consistent with earlier NSFG and vital statistics data dating back to the 1970s (Copen et al. 2012).
Divorce and spousal death are the two primary ways that marriages end, but divorce is now far more common (Cherlin 1992). Although spousal death was more common for centuries, this changed in the mid-1970s, when divorce overtook spousal death as the primary cause of marital dissolution for the first time in U.S. history (Cherlin 1992). Divorce is most likely to occur either around the seven-year threshold or at mid-life, usually around the time a couple's first child reaches the age of 14, when different-sex marital quality tends to hit a low-point (Amato and Cheadle 2005;Cherlin 1992;Gottman and Levenson 2000;Orbuch, House, Mero and Webster 1996;White and Booth 1991). Recent evidence from the Survey of Income and Program Participation (SIPP) suggests that, for first marriages, couples are mostly likely to separate at approximately seven years before finalizing their divorce around year eight (Kreider and Ellis 2011; see also Brentano 2013).
In most of the existing research, marital duration is used as an independent variable, with scholars examining how different aspects of the marital relationship change over time. For instance, marital satisfaction has been shown to change with marital duration, although the shape of this association remains a point of disagreement among scholars (e.g., Glenn 1990Glenn , 1998Johnson, Amoloza, and Booth 1992;Johnson, White, Edwards, and Booth 1986;Vaillant and Vaillant 1993;VanLaningham, Johnson, and Amato 2001;Weishaus and Field 1988). Similarly, researchers have examined patterns of sexual frequency (e.g., Blumstein and Schwartz 1983;Call, Sprecher, and Schwartz 1995;Jasso 1985;Liu 2000) and sexual quality (Liu 2003) by marital duration. However, some demographers, sociologists and other family scholars examine different-sex marital duration as an outcome and estimate the probability that marriages will last particular lengths of time as a function of various characteristics (Copen et al. 2012). Given the relevance of marriage to individual and family well-being (Lillard and Waite 1995;Waite and Gallagher 2000;Williams 2003), important questions about the factors that shape marital duration can be posed, such as whether same-sex sexuality shortens or lengthens it.
In the analyses that follow, we use different-sex marital duration as a dependent variable and ask whether three distinct components of same-sex sexuality -attraction, behavior and identity -shorten or lengthen marital duration, net of other relevant demographic influences. We restrict our analysis to the oncemarried, once-divorced on the assumption, grounded in the theoretical literature cited above, that the dissolution of a first marriage because of same-sex sexuality generally precludes individuals from entering into higher-order different-sex marriages. If the initial motivation for marrying was to meet heteronormative expectations, or if same-sex sexuality emerged during the marriage, then the divorce process likely reduces the motivation and desire for subsequent different-sex marriages.

Same-sex sexuality and different-sex marital duration
In theory, same-sex sexuality may be associated with shorter or longer different-sex marital duration, or have no observable association with marital duration. Below, we take each of these possibilities in turn, starting with the less-intuitive possibility that same-sex sexuality might be associated with longer marital duration.
For individuals who experience same-sex sexuality prior to or during a different-sex marriage, there may be much at stake in prolonging the marriage, perhaps even more than in marriages where both spouses are heterosexual. Aside from the stressors with which all people must deal while divorcing, those who experience same-sex sexuality also face the possibility of losing their protective cover. For individuals who are fearful of the judgment and other negative sanctions that lesbian, gay, and bisexual people often encounter, or who have some degree of internalized homophobia (see Meyer 1995;Frost and Meyer 2009), maintaining the marriage for as long as possible may be key to their own sense of safety and wellbeing. It is important to emphasize that this may apply to bisexuals, as well, who often experience invalidation and stigma as a result of their sexuality (see Eliason 2000;Israel and Mohr 2004). In this sense, same-sex sexuality may act as a barrier to divorce, thus increasing marital duration. This may be especially true among those who identify as lesbian, gay or bisexual, since it is identity, rather than attraction or behavior, that is most likely to engender the negative consequences that many initially sought to avoid. Still, same-sex attraction and behavior both might contribute to longer marital duration. In the case of attraction, some individuals may have invested in maintaining a different-sex marriage in order to stave off any suspicions regarding their same-sex attractions (Humphreys 1970;Seidman 2002). In terms of behavior, it is possible that some who engage in extramarital same-sex behavior find that doing so offers them an outlet for fulfilling their sexual desires, thus minimizing their need to leave the marriage in order to achieve sexual satisfaction (see Walker 2014).
The fact that same-sex marriage did not become legal across the U.S. until 2015 is also important to note, as this may have prolonged some marriages. Those who entered into different-sex marriages because they see marriage as natural or normal, or as the most appropriate relationship form for having and raising children, may have left their marriages sooner had they been legally permitted to marry a partner of the same sex. Finally, different-sex marriages contracted during earlier, more restrictive time periods during which homosexuality was more strongly stigmatized may tend to be longer. For instance, some members of mid-20th century birth cohorts who experienced same-sex sexuality married a different-sex spouse because, at the time, lesbian, gay and bisexual people were openly and legally discriminated against, and most were ostracized from their families and friendship circles (Humphreys 1970;Seidman 2002). Attitudes toward homosexuality have recently liberalized (Loftus 2001), but marriages contracted long ago are likely well-established. Couples may have strong family ties, shared assets and routines that act as barriers to exiting the marriage, even if one or both spouses experiences same-sex sexuality. Grace and Frankie, a recent Netflix original series, provides a popular-cultural dramatization of this possibility (for a discussion, see London and Hoy 2017). These circumstances could delay divorce or make it an impossibility for some individuals, thereby increasing marital duration.
However, it can also be theorized that same-sex sexuality might be associated with shorter marital duration. Same-sex sexuality may shorten marital duration via a variety of mechanisms. Individuals who experience same-sex sexuality and are married to a different-sex spouse most often conceal their sexuality from their spouse, and they sometimes develop elaborate schemes for doing so (Buxton 1994(Buxton , 2001Humphreys 1970;Ross 1983). However, they also risk being discovered, perhaps especially if they are engaged in extramarital sex with same-sex partners or if they disclose same-sex attractions or a gay, lesbian or bisexual identity to others. In the case of extramarital same-sex behavior, some couples may divorce quickly because one spouse feels betrayed by the other's infidelity. Some may also feel that their spouse's same-sex behavior makes them untrustworthy or a risk for sexually transmitted infections. Furthermore, as attitudes toward gay, lesbian and bisexual people become more tolerant and accepting (see Loftus 2001), some may also feel emboldened and choose to divorce in order to explore their samesex sexuality. In such instances, individuals who are attracted to others of the same-sex may see divorce as a pathway to exploring those attractions. For those who already identify as lesbian, gay or bisexual, divorce may allow them to publicly acknowledge their identities. In either event, some different-sex couples are likely to see same-sex sexuality as a serious threat to their marriage. Because same-sex sexuality contradicts widely held assumptions about what a different-sex marriage should be, some of these couples may initiate a separation or divorce after a process of counseling and negotiation, while others may do so immediately (Wolkomir 2009). Whereas other causes of divorce, including a lack of communication or emotional intimacy, may develop over longer periods of time (see Kayser 1993), the revelation of a spouse's same-sex sexuality is more likely to be a sudden and dramatic turning point in the relationship. Indeed, if a spouse is exclusively attracted to persons of the same sex and/or identifies as gay or lesbian, the couple may realize that they are incompatible and bypass the work of attempting to rescue the marriage that would otherwise lengthen marital duration. Thus, unlike other causes of divorce, samesex sexuality may initiate a quick end to the marriage, thus decreasing marital duration. This may even be the case with bisexuality. Although mixed-orientation marriages involving a bisexual spouse can and do survive, heterosexual spouses may worry about their inability to satisfy all of their bisexual spouse's sexual wants and may not fully trust that their spouse will remain faithful. In such circumstances, individuals in mixed-orientation marriages often try to communicate more frequently and more openly in order to maintain the marriage (see Buxton 2001Buxton , 2004aBuxton , 2004bHernandez et al. 2011), but this causes additional stress that may lead couples to divorce more quickly.
It is important to note that it is likely that same-sex sexuality increases marital duration in some cases, while decreasing it in others. If these countervailing scenarios occur with approximately equal frequency, then there may be no aggregate associations between same-sex sexuality and marital duration observed, particularly in cross-sectional data. In the case of mixed-orientation marriages involving a bisexual spouse, couples may be able to manage any complications that arise because of the bisexual spouse's sexuality and thus have marital durations that are no different from those of couples in which both partners are exclusively heterosexual.
Given the dearth of prior research on this topic, we have no empirical basis to strongly favor one hypothesis over another. That said, theoretical considerations lead us to favor the hypothesis that samesex sexuality will be associated with shorter marital durations.

Analysis plan
In what follows, we use data from the 2011-2013 NSFG to examine the associations between three components of same-sex sexuality -attraction, behavior and identity -and marital duration among those who married and divorced once. Restricting the sample in this way while also controlling for demographic and background variables provides a relatively homogenous sample on which to directly estimate associations between the dimensions of same-sex sexuality and different-sex marital duration. Both observed factors, and unobserved factors that select people into different-sex first marriages and divorce, are controlled. In these analyses, we do not include the separated with the divorced as is often done in other studies because it is unclear whether the separated consider their marriages officially over and stop counting duration at the point of separation; some may consider themselves still legally married and count duration up to the time of the divorce.
Following the presentation of results from these primary analyses, we report the results from two supplemental analyses. In the first, we examine whether sex/gender moderates any observed associations between same-sex sexuality and marital duration. In the second, we distinguish those with bisexual attraction and identity from those with exclusively same-or different-sex attractions and identities, respectively, in order to nuance our main findings. We do not analyze bisexual behavior in these analyses on the assumption that all or almost all individuals who are in different-sex marriages and report ever having engaged in same-sex behavior have had sexual experiences with both women and men (i.e., that they consummated their marriages in addition to having had same-sex encounters).

Data
The 2011-2013 NSFG is a cross-sectional probability sample of 10,416 15-44-year-old women and men in the household (i.e., non-institutionalized) population of the U.S. The NSFG is designed and funded by the Center for Disease Control and Prevention's National Center for Health Statistics and several other programs within the U.S. Department of Health and Human Services (Copen et al. 2016). Data were collected by means of in-person interviews. Separate questionnaires were used for women and men, although there is considerable overlap in core content.
Most data were collected via computer-assisted personal interviewing (CAPI), which involves having the questionnaire loaded onto a laptop, and having an interviewer ask the survey questions and record the respondent's answers. Some data, including the data on same-sex attraction, behavior and identity, were collected via audio computer-assisted self-interviewing (ACASI). With ACASI, respondents listen to the interview questions on headphones, read them on a screen, or both, and enter their responses directly into the computer. By removing the interviewer from the process of collecting sensitive data, social desirability bias is reduced. The overall response rate was 72.8%.
The NSFG collects a complete marital and divorce history for each respondent. Using data on the participant's age and a NSFG-derived variable that measures the number of times the respondent married (FMARNO), which is provided in the public-use dataset, we were able to identify persons who were 18 or older and had married once. Given those constraints, persons who reported their current marital status as divorced had, by definition, ended their first marriage and not remarried (N=674). We excluded those who married more than once on the assumption that the hypothesized influence of same-sex attraction, behavior and identity is best examined in relation to first marriages (i.e., among those who never remarried someone of a different sex).
We also excluded those with missing data on any analytic variable. Overall, we dropped 57 once-married, once-divorced participants (8.5%, un-weighted) due to missing data on one or more analytic variable. Same-sex attraction and behavior, respectively, and marital duration were not significantly associated with being dropped from the analytic sample, but lesbian-, gay-, bisexually identified persons as a group were more likely to be missing than those not so identified. Sex/gender (men more than women), age at marriage (older more than younger), and maternal employment when the participant was 5-15 years old (part-time more than full-time) were the only other variables associated with missingness. Although missingness was not random, none of the statistically significant differences were particularly large. Once we constrained the NSFG sample to those who had married and divorced once, dropped those missing on the three focal independent variables and the dependent variable, and dropped a small number of individuals with missing data on the control variables, the analytic sample included 617 individuals with no missing data.

Measures
The continuous dependent variable is the NSFG-derived variable (MAR1DISS), which measures the duration of the respondent's first marriage, in months.
The NSFG measured three dimensions of same-sex sexuality: attraction, behavior and identity. We use dichotomous indicators of each as the focal independent variables in our primary analyses. However, as noted above, these indicators invisiblize those who experience both same-and different-sex sexuality. Therefore, in supplemental analyses, we distinguish current bisexual attraction and identity, but not behavior, and use these indicators as independent variables to assess whether individuals who experience some aspects of different-sex sexuality, alongside same-sex sexuality, have shorter or longer marital duration, on average, than individuals who experience only different-or same-sex sexuality, respectively. As noted in the literature review above, qualitative evidence suggests that those with bisexual attractions and/or identities may be better able to maintain their marriages for longer durations (e.g., Buxton 2001Buxton , 2004aBuxton , 2004bEdser and Shea 2002;Hernandez et al. 2011).
Same-sex attraction was measured by a question that asked: "People differ in their sexual attraction to other people. Which describes your feelings? Are you: (1) only attracted to females; (2) mostly attracted to females; (3) equally attracted for females and males; (4) mostly attracted to males; (5) only attracted to males; (6) not sure." For our primary analyses, we recoded this information into a dichotomous variable such that men and women who reported some same-sex attraction were combined into one category (=1), and men who reported only being attracted to women and women who reported only being attracted to men were recoded as not having same-sex attractions (=0). For the supplemental analysis, we derived a three-category variable that distinguishes those with exclusive same-or different-sex attractions, respectively, from those who express at least some attraction to both sexes (categories 2, 3, and 4 above).
Same-sex behavior was measured using a dichotomous NSFG-derived variable that indicates whether the respondent had ever engaged in sexual behavior with a person of the same sex (SAMESEXANY). Sexual identity was measured with a question that asked: "Do you think of yourself as: (1) heterosexual or straight; (2) homosexual or gay; (3) bisexual." For our primary analyses, we recoded this variable as a dichotomy, with answers 2 and 3 combined to signify gay, lesbian or bisexual identity (=1). For the supplemental analysis, we distinguished those reporting bisexual identity from those reporting hetero-and homosexual/gay identities, respectively.
Since duration of marriage partly depends on age at marriage, we include this measure in our analyses. Age at marriage was calculated as the difference between two NSFG-derived variables -the century month of the respondent's first marriage (MARDAT01) and the century month of the respondent's birth (CMBIRTH) -divided by 12.
The NSFG included a range of exogenous demographic and background variables that we include as controls in the multivariate analyses. One demographic control variable that has received considerable attention in the literature on marriage and divorce -different-sex premarital cohabitation -is recoded dichotomously (yes=1). The other demographic variables include: sex/gender (female=1); age, recoded categorically as 18-24, 25-29, 30-34, 35-39, and 40-44 years; race/ethnicity, recoded categorically as non-Hispanic White, non-Hispanic African American, Hispanic, and non-Hispanic other races/ethnicities; foreign-born, recoded dichotomously (yes=1); and maternal education, recoded as less than high school, high school graduate, some college, and college graduate or more. Additionally, participants retrospectively reported on a range of background characteristics related to their childhood and adolescence. Respondents indicated whether their mother worked when they were 5-15 years old, recoded categorically as did not work for pay, full-time, part-time, and equal amounts full-and part-time. Respondents also indicated with whom they lived at age 14, recoded categorically as with both parents, only one parent, a parent and step-parent, and neither parent/other. Finally, respondents indicated the religion in which they were raised, recoded categorically as no religion, Baptist, Catholic, Protestant, and some other religious affiliation. A variable measuring the year in which the survey was conducted is also included as a control variable.

Analytic approach
We begin by describing the characteristics of the total population of once-married, once-divorced persons, as well as the characteristics of the sub-populations of persons who report same-sex attraction, same-sex behavior, and lesbian, gay or bisexual identity, respectively. Then, we describe the bivariate associations between the available measures of same-sex sexuality and the control variables, respectively, and the duration of different-sex marriages. Next, focusing on the once-married once-divorced, we present the results from multivariate ordinary least squares (OLS) regression analyses of the association between each measure of same-sex sexuality and the number of months the different-sex marriage lasted. Finally, we report the results from two supplemental analyses: the first focuses on whether sex/gender moderates associations between same-sex sexuality and duration of marriage; and the second focuses on distinguishing those with bisexual attraction and identity from those with exclusively same-or differentsex attraction and identity, respectively. We conducted all analyses using the SVY commands in STATA 14.0 (StataCorp 2015). All analyses are weighted and the standard errors are corrected to take the NSFG's complex sampling design into account.

Population description
As seen in the first column of Table 1 (all Tables and Figures are in Annex 1 to the present document), 15.3% of once-married, once-divorced individuals report current same-sex attraction and 18.7% report having ever engaged in same-sex sexual behavior. However, only 6.5% identify as gay, lesbian or bisexual. It is noteworthy that the three measures of same-sex sexuality are not perfectly correlated with one another. Specifically, the correlation between current attraction and having ever engaged in same-sex behavior is 0.78, the correlation between current attraction and current identity is 0.61, and the correlation between having ever engaged in same-sex behavior and current identity is 0.54 (not shown in table). These patterns of discordance are also evident in Table 1. Among the once-married, once-divorced reporting same-sex attraction, 90.1% report that they have ever engaged in same-sex sexual behavior; however, only 41.9% currently identify as lesbian, gay or bisexual. Among those who report ever engaging in same-sex sexual behavior, 73.8% report current same-sex attraction, while only 34.3% report lesbian, gay or bisexual identity. Among those who report a lesbian, gay or bisexual identity, almost all report current attraction and ever engaging in same-sex sexual behavior. These patterns are consistent with what has been previously reported in the literature (Laumann et al. 1994). It is partly because these dimensions of same-sex sexuality are not more highly correlated that we analyze each distinctly.
The distributions for all of the background and demographic control variables for the total population of once-married, once-divorced persons, as well as the subpopulations of persons who reported same-sex attraction, same-sex sexual behavior, and lesbian, gay or bisexual identity are also provided in Table 1. A few patterns are noteworthy. First, the percentage of the once-married, once-divorced who married prior to the age of 20 is substantially higher among individuals in each of the subpopulations who report some aspect of same-sex sexuality than among individuals in the total population (32.4% to 49.2% versus 19.5%). They are also somewhat more likely to cohabit with someone of a different sex prior to marriage (69.4% to 75.4% versus 63.9%). These patterns are consistent with selection processes that may channel some people who experience same-sex sexuality into different-sex relationships and marriages at younger ages because they are seeking normative cover for covert same-sex sexuality or because they believe that heterosexual marriage and childrearing are both natural and normal (Higgins 2002;Humphreys 1970;Pearcey 2005;Ross 1983). Second, women, younger persons, whites, the native-born, and those who did not live with both parents at age 14 are all over-represented among the once-married, once-divorced who report some aspect of same-sex sexuality.

Bivariate analysis
Overall, the mean duration of marriage among the once-married, once-divorced is 61.6 months (5.1 years). As seen in Table 1, two of the three measures of same-sex sexuality, age at marriage, and three of the 10 control variables have statistically significant bivariate associations with duration of marriage.
Same-sex attraction and behavior are associated, respectively, with a significantly shorter duration of marriage. The difference for each is approximately 16 months shorter. Lesbian, gay, and bisexual identity is not associated significantly with marital duration in the bivariate analysis; the difference is only 5 months shorter for those who identify as such relative to those who do not. In addition to these measures of same-sex sexuality, age at marriage is inversely associated with duration of marriage, with the longest mean duration observed among those who married at ages under 20 and the shortest mean duration observed among those who married at age 30 or more. Among the other demographic and background variables, sex/gender, age, and living arrangements at age 14 are each associated significantly with duration of marriage. Table 2 presents the results of an OLS analysis of marital duration among the once-married who divorced (N=617). Constraining the sample to the divorced increases the homogeneity of the sample substantially and allows us to obtain direct estimates of the association between each dimension of same-sex sexuality and marital duration among those who volitionally ended their first different-sex marriage. We estimate separate models for each dichotomous indicator of same-sex sexuality and include all of the controls, including age at marriage. As seen in Table 2, net of controls, each dimension of same-sex sexuality is associated with a substantial and statistically significant reduction in martial duration among the oncemarried, once-divorced. Estimates range from approximately 18 months shorter for lesbian, gay and bisexual identity to 24 months shorter for having ever engaged in same-sex sexual behavior. Marital duration is also shorter among those who married at older ages, women, younger persons, and those who lived with only one parent at age 14. Marital duration is significantly longer among those with some college education and those who completed college, although the latter estimate is only marginally significant. Marital duration is also marginally longer among those interviewed in 2013.

Supplemental analyses
In addition to the primary analyses reported above, we conducted two supplemental analyses. First, we examined whether our results regarding same-sex sexuality varied by sex/gender by re-estimating all of the models with an interaction between sex/gender and the included dimension of same-sex sexuality. In no instance was the coefficient on the interaction term statistically significant. Thus, among the oncemarried, once-divorced, the association between each dimension of same-sex sexuality and marital duration appears to be the same for men and women.
Second, recognizing the limitations of the dichotomous measures of same-sex sexuality used in our primary analyses, we recoded the attraction and identity measures to distinguish respondents who are bisexual from those who report only same-sex attraction and gay/lesbian identity, respectively, and reestimated all of the models. We did not do this with behavior because we assumed individuals who had entered different-sex marriages had consummated those marriages; thus, we assumed all once-married persons are behaviorally bisexual from a life-course perspective.
As seen in Table 3, among the once-married, once-divorced, individuals who report bisexual attraction have significantly shorter marital duration than individuals who report different-sex only attraction in Models 1 through 3, although the difference is only marginally significant in Model 1. Additionally, persons who report same-sex only attraction have shorter marital duration than persons who report different-sex only attraction in Models 1 and 2 (marginally significant), but not in Model 3. It is noteworthy that the marital duration of persons who report bisexual and same-sex only attraction, respectively, differ significantly from one another only in Model 3. In Model 3, persons who report bisexual attraction have shorter marital duration than persons who report same-sex only attraction, as well as persons with different-sex only attraction. Given that the only difference between Models 2 and 3 is the inclusion of age at marriage, it is clear that controlling for age at marriage in Model 3 produces two substantive changes: (1) it reduces to non-significance the difference between persons who report samesex only and different-sex only attraction that is evident in Model 2; and (2) it results in the emergence of a significant difference between those who respectively report bisexual and same-sex only attraction in Model 3.
A supplemental analysis helps explain these results. Specifically, we find that there are large differences in age at marriage between the once-married, once-divorced who report different-sex only, bisexual and same-sex only sexual attraction. As seen in Figure 1, 76.8% of persons reporting bisexual attraction were 24 or younger at the time they married, compared to 61.4% of persons reporting different-sex only attraction, but only 31.6% of persons reporting same-sex only attraction. Stated otherwise, relative to individuals who report different-sex only attraction, those who report bisexual attraction marry at younger ages, while those who report same-sex attraction only marry at older ages. Since older age at marriage is generally associated with shorter marital duration among the once-married, once-divorced (see Table 2), these results indicate that the difference in marital duration between persons reporting same-sex only attraction and persons reporting different-sex only attraction is accounted for by the substantially older ages at marriage among same-sex only-attracted individuals. Once age at marriage is controlled statistically, there is no difference between those who report same-sex only and different-sex only attraction. However, these results also indicate that age at marriage does not explain the difference between persons reporting bisexual and different-sex only attraction, respectively. Those reporting bisexual attraction both marry at younger ages ( Figure 1) and have significantly shorter marriages than persons who respectively report different-sex only and same-sex only attraction. As a result, individuals who report bisexual attraction also have shorter marital duration than persons who report same-sex only attraction.
With respect to identity, the pattern of results is somewhat different. In Model 1, which only includes year of data collection, there is a significant difference between bisexual and lesbian/gay-identified individuals; however, it is only once-married, once-divorced individuals who identify as lesbian or gay who have shorter marital duration relative to individuals who report heterosexual identity. Including the demographic and background variables in Model 2 reduces the difference between the bisexual and lesbian/gay-identified groups to non-significance, although the coefficient on lesbian/gay identity remains negative and statistically significant. Adding age at marriage to Model 3 reduces the difference in marital duration between lesbian/gay-and heterosexually-identified persons to non-significance, while the association between bisexuality and marital duration becomes marginally significant. For bisexually-and lesbian/gay-identified persons, respectively, the reduction in marital duration relative to heterosexuallyidentified persons is around a year and a half in the fully specified model. Similar to what was discussed above for attraction, among the once-married, once-divorced, differences in age at marriage between the sexual identity groups -relatively younger age at marriage among bisexuals and older age at marriage among lesbians/gays (see Figure 2)-accounts for the different pattern of associations observed in Model 3 relative to those observed in Model 2.

Discussion
Recent research has documented that same-sex sexuality -desire/attraction, behavior, identity -is associated with a reduced likelihood of entering into a different-sex marriage and an increased likelihood of a different-sex divorce among the once-married (London and Hoy 2017). However, existing research has not addressed the association between same-sex sexuality and the duration of first marriages. To address this gap in the literature, we used the 2011-2013 NSFG to examine whether the three dimensions of same-sex sexuality are associated with shorter or longer marital duration among the once-married, once-divorced. By using a sample comprised only of the once-married, once-divorced, we are able to get a direct estimate of each of these associations in a relatively homogenous sample, where both observed and unobserved factors that select people into marriage and divorce are controlled. The results indicate that persons who report any of the three aspects of same-sex sexuality divorced at shorter marital durations, on average, than persons who did not report such experiences. Specifically, same-sex attraction is associated with marital duration that is approximately 20 months shorter, ever engaging in same-sex sexual behavior is associated with duration that is approximately 24 months shorter, and identifying as gay, lesbian or bisexual is associated with duration that is approximately 18 months shorter. Together, these findings contribute to the literature on same-sex sexuality and different-sex marriage, often called "mixed-orientation marriages," by empirically documenting that the components of same-sex sexuality are associated with reduced marital duration. Although not definitive due to data limitations, which we discuss below, this suggests that same-sex sexuality may indeed have been present during, and shortened some, first marriages via divorce.
In addition, we conducted two supplemental analyses. First, we sought to determine whether the associations between same-sex sexuality and shorter marital duration might be different for women and men. However, we did not observe any moderation by sex/gender, although we do observe a main effect of sex/gender in all of our analyses (see Table 2; present but not shown in Table 3). Taken together, among the once-married, once-divorced, sex/gender and same-sex sexuality are independently associated with marital duration in countervailing ways; women have significantly longer duration than men, while persons who report same-sex sexuality have shorter duration than persons who do not. Given differences in various aspects of same-sex sexuality among women and men (Diamond 2009;Humphreys 1970;Stein 1997), we think further theorizing about and empirical testing of sex/gender and same-sex sexuality interactions are warranted in future research.
Second, because our primary analyses used dichotomous measures of same-sex sexuality, we conducted supplemental analyses to determine whether those who report bisexual patterns of attraction and/or a bisexual identity are different from those who report same-sex only, or different-sex only, sexuality. Our results suggest that differences in marital duration between groups with same-sex only, bisexual and different-sex only attraction and identity, respectively, are influenced substantially by differences in age at first marriage. Specifically, as we document in Figures 1 and 2, among the once-married oncedivorced, persons who report bisexual patterns of attraction tend to marry at younger ages than persons who respectively report different-or same-sex only attraction. The same holds for bisexually-versus heterosexually-and gay/lesbian-identified persons. At the same time, many of those who report same-sex only attraction or gay/lesbian identity marry at relatively older ages. Thus, controlling for age at marriage: explains the difference in marital duration between persons with same-sex only and different-sex only attraction; results in the emergence of a significant difference in marital duration between persons with bisexual and same-sex only attraction; explains the difference in marital duration among lesbian/gay-and heterosexually-identified persons; and results in the emergence of a marginally significant difference in marital duration between bisexually-and heterosexually-identified persons.
In addition, among the once-married, once-divorced, we find that there is a subset of persons who report same-sex sexuality who marry at young ages, a sizeable percentage under the age of 20 (see Table 1). The earlier age at marriage for this subpopulation seems consistent with the literature we cite above, which suggests that individuals who experience same-sex sexuality often marry under the pressures of a heteronormative society, either because they want to pass as heterosexual or because they believe that heterosexual marriages are natural and normal, especially for having and raising children (e.g., Higgins 2002;Humphreys 1970;Pearcey 2005;Ross 1983). The differences in age at marriage among those who report bisexual versus same-sex only attractions and identities, respectively, is harder to explain. The age at marriage patterns we document among the once-married, once-divorced indicate that many of those who report bisexual attraction and/or identity marry early in life and divorce at shorter durations of marriage. Additionally, they indicate that many who report same-sex only attraction and/or a gay/lesbian identity delay marriage until their mid-to late-20s and divorce at durations similar to those of the heterosexually identified once age at marriage is controlled. For the gay/lesbian-identified, remaining single may not arouse suspicions at an earlier age since many of their peers are likely single, as well. However, once their peers begin to marry, lesbian or gay individuals in some contexts may find that their sexuality appears suspect and may elect to marry to curtail such suspicions. Although we believe these kinds of social forces may help explain the patterns we observe, we caution against drawing definitive conclusions based on this interpretation given that we focus on the once-married once-divorced population, our measures of attraction and identity are taken post-divorce, and there may be un-measured factors that simultaneously influence sexual attraction/identity and decisions about the timing of marriage and/or divorce. We believe that these findings related to same-sex sexuality and age at marriage raise questions that warrant additional theorizing and further investigation.
For the former group, comprised of individuals who report bisexual patterns of attraction and/or a bisexual identity, our finding that they tend to marry early in life and divorce with shorter marital duration complicates the existing literature on bisexual people in mixed-orientation marriages. In general, the literature suggests that mixed-orientation marriages in which a spouse is bisexual are less likely to divorce because couples can maintain a mutual attraction and a satisfying sex life (Buxton 2001(Buxton , 2004a(Buxton , 2004bEdser and Shea 2002;Hernandez et al. 2011). Being bisexual, rather than gay or lesbian, enables one to continue having a romantic relationship with a different-sex spouse, and therefore, the couple can better endure the stresses that same-sex sexuality can pose within the context of a different-sex marriage (see Wolkomir 2009). As such, it would seem likely that even among those who do eventually divorce, doing so at a longer marital duration would be most likely, given how bisexuality, rather than exclusive samesex sexuality, can enable couples to endure. However, focusing specifically on the once-married, oncedivorced, our analysis suggests that those who report bisexual attraction and/or identity are more likely to divorce at shorter durations. Furthermore, we find that this is explained by age at marriage. It is possible that only those who marry at a young age are inclined to divorce at a shorter duration. Yet again, we believe that these findings raise questions that should be addressed in future research.
The present research uses a high-quality, nationally representative dataset to estimate well-specified models of different-sex marital duration, but it is not without limitations. A primary limitation is that the data are cross-sectional and do not allow us to know when the components of same-sex sexuality emerged in relation to marriage and divorce; it is possible that in some cases, same-sex attractions, behaviors, and/or identities developed only after the marriage had ended. Longitudinal data that track sexual development in relation to marriage and divorce are needed to more clearly establish the extent to which same-sex sexuality that develops before or during a different-sex marriage contributes to its dissolution. A second limitation of the current study is that we do not know what individuals themselves think caused the end of their marriage; subjective appraisals and attributions of cause would help clarify the role, if any, that same-sex sexuality plays in divorce. A third limitation of the current study pertains to the fact that the NSFG does not include individuals older than 44. As such, older adults who were born and entered prime marital ages earlier in the 20 th century when attitudes toward homosexuality were less progressive are not represented. Given that it is these individuals who might experience same-sex sexuality from within very long-term and potentially enduring marriages, our capacity to find evidence that same-sex sexuality is associated with longer marital duration is somewhat constrained. Fourth, the measure of premarital cohabitation that we have available pertains only to different-sex premarital cohabitation. It would be useful if data on same-and different-sex cohabitation experiences were collected. Finally, it is possible that a small number of the marriages captured in the NSFG were same-sex marriages since same-sex marriage had been legalized in some jurisdictions during the period of data collection. In this case, it would be incorrect to analyze them as mixed-orientations marriages, as we have done. Unfortunately, to the best of our knowledge, it is not possible to specify whether any of the marriages in the NSFG were same-sex marriages. These limitations could be addressed in future research if relevant data are available or collected.
This research raises other questions that might be addressed in future research. First, with the recent nationwide legalization of same-sex marriage, do individuals who experience same-sex sexuality now leave different-sex marriages more quickly? And, if so, how do individuals make such decisions? The NSFG data used in this study were collected between 2011 and 2013, when same-sex marriage was legal in several states. However, same-sex marriage only became legal nationwide in 2015. Thus, many individuals currently in a different-sex marriage have only recently been given access to same-sex marriage. Because many people who experience same-sex sexuality enter into different-sex marriages because they value marriage or see it as "normal" (Higgins 2002;Pearcey 2005;Ross 1983), the possibility of marrying someone of the same sex may decrease the appeal of a different-sex marriage and incentivize a more rapid divorce than was the case in earlier periods. Second, the role of marital quality has been a major focus of the literature on different-sex divorce, marital duration, and their consequences, but the influence of same-sex sexuality on such dynamics and outcomes has been under-studied. For instance, might some of the negative effects of divorce -perhaps especially in terms of emotional and psychological well-being -be more or less severe for those whose marriages ended due to the presence of same-sex sexuality? Perhaps these effects are mediated by one's attitudes toward homosexuality. Third, although separation is sometimes the focus of analysis, in this paper, we focus exclusively on the once-divorced and exclude the once-separated from the analysis. Future research might focus on the association between same-sex sexuality and separation, the time to separation, and/or the time from separation to divorce.
Overall, future research on marriage and divorce should take same-sex sexuality into account whenever doing so is possible. To date, few researchers have considered same-sex sexuality, perhaps because of the under-theorized assumption that it is irrelevant to different-sex marriages. Data limitations may also have limited the ability of researchers to address such questions. However, as previous research has shown (e.g., Laumann et al. 1994), same-sex attractions, behavior, and identity are distinct and can be present among those who identify as heterosexual, and among those in different-sex relationships and marriages. Recent evidence elaborates this notion by showing that sexually active adolescents in New York City who identify as lesbian, gay, or bisexual are more likely than heterosexually identified sexually-active adolescents to report at least one pregnancy (Lindley and Walsemann 2015). Thus, same-sex sexuality is significant for greater numbers of people than most assume, and as our analyses here and elsewhere (London and Hoy 2017) have shown, same-sex sexuality can shape a person's experiences with marriage and divorce in important ways.    0.33 0.34 0.32 Significance Levels: # = p < 0.10; * = p < 0.05; ** = p < 0.01; *** = p < 0.001.  -17.17 (11.74) (9.38) (16.07) Significance Levels: # = p <0.10; * = p < 0.05; ** = p < 0.01; *** = p < 0.001; Ŧ = Bisexual Differs from Same-Sex Only/Gay/Lesbian, p<0.05. Notes:

Annex
1. Model includes control of survey year only. 2. Model includes controls for demographic and background variables. The demographic variables are: premarital different-sex cohabitation; gender; age; race/ethnicity; place of birth; and mother's education. The background variables are: maternal employment status when the respondent was between 5 and 15 years old; living arrangements at age 14; and religion in which the respondent was raised. Survey year is also controlled. 3. Model adds age at marriage to Model 2.