Abstract
Extensive research documents the relevance of families and socioeconomic resources to health. This article extends that research to sexual minorities, using 12 years of the National Health Interview Survey (N = 460,459) to examine self-evaluations of health among male and female adults living in same-sex and different-sex relationships. Adjusting for socioeconomic status eliminates differences between same- and different-sex cohabitors so that they have similarly higher odds of poor health relative to married persons. Results by gender reveal that the cohabitation disadvantage for health is more pronounced for different-sex cohabiting women than for men, but little difference exists between same-sex cohabiting men and women. Finally, the presence of children in the home is more protective for women’s than men’s health, but those protections are specific to married women. In all, the results elucidate the importance of relationship type, gender, and the presence of children when evaluating health.
Regardless of the researcher’s ideological or empirical position, families are becoming ever more fluid (Rosenfeld 2007). In a recent national study of American views and attitudes toward family life, Powell et al. (2010) found that the largest perception change regarding family types involves views of same-sex families. For example, in 2006, roughly 33 percent of individuals surveyed believed that two persons of the same sex living together as partners represented a family, a significant jump up from only 25 percent three years earlier. This figure reaches 60 percent, notably a majority opinion, if same-sex couples live together with children (Powell et al. 2010). Same-sex couples still lag behind other family types, including different-sex cohabiting couples, in public approval. For example, in 2006, nearly 40 percent of individuals in the same survey considered an different-sex unmarried couple with no children a family, and over 80 percent believed that an different-sex unmarried couple with children represented a family (Powell et al. 2010). But compared with overwhelming public disapproval of same-sex relationships a half a century ago—resulting, for example, in executive orders banning sexual minorities from federal employment—the changing views of sexual minorities represent a substantial shift in public opinion.
Although public debates on family composition are ongoing, scholars recognize the importance of family life for individual well-being, and researchers have examined the health-related consequences of living in marital and nonmarital relationships for some time now (see review by Wood, Goesling, and Avellar 2007). Although the rise in cohabitation and nonmarital childbearing led to a flurry of research on the economic and health consequences of living in single-parent and cohabiting-parent families, researchers, policy makers, and health care providers know very little about how family relationships affect sexual minority health, despite the growing presence and recognition of sexual minority families in the United States (Institute of Medicine 2011). This is an important limitation, given that families generally serve as an integral source of support (financial and otherwise) and are crucial for individual health and well-being (Carr and Springer 2010). At the same time, as we detail below, researchers have documented health disadvantages among sexual minorities (Institute of Medicine 2011). Yet these literatures have arisen almost entirely separately from each other.
In this article, we ask where sexual minorities fit in the discussion of families and health. Investigation of this question is needed, because research shows that intimate relationships may be especially important for individuals who find themselves outside the norm and often in compromised social positions (Mays and Cochran 2001). An impressive body of work examining health status shows important distinctions relative to the types of relationships people maintain. For example, although both cohabitation and marriage are generally beneficial for health, these benefits may operate more strongly for married adults (Musick and Bumpass 2006, 2012). However, this dichotomous representation is based almost entirely on cohabitors of different sex (see Reczek and Umberson 2012 and Wienke and Hill 2009 for recent and laudable exceptions), leaving it an open question where sexual minorities fall along the health continuum for persons who reside in family relationships, and whether the factors that contribute to health differences across family types, including socioeconomic resources, operate similarly for same- and different-sex cohabitors when contrasted against married adults.
Furthermore, studies of social relationships and health also suggest that gender may complicate the health profile of those in same-sex relationships. The reduced mortality risk associated with marriage is stronger among men, although research examining gender differences in the physical health benefits of marriage is relatively sparse (see review by Wood et al. 2007), and we know almost nothing about the health consequences of cohabitation for men and women in same-sex relationships. Furthermore, although some (but not all) research suggests that the presence of children improves the health of parents, those health benefits are not uniform (Umberson and Montez 2010), and virtually no work on sexual minority health considers whether the health impacts of family life differ by the presence (or absence) of children in the home.
We address these gaps by centering on the role of families in health, for men and women in different- and same-sex relationships. Drawing on data from the 1997 to 2008 waves of the National Health Interview Survey (NHIS), we investigate differences in health evaluations among men and women in same- and different-sex cohabiting relationships, relative to their different-sex married counterparts, both before and after adjustment for socioeconomic status (SES). In doing so, this study addresses a void in the scientific literature on family relationships and health for an ever more visible minority population (Institute of Medicine 2011; U.S. Department of Health and Human Services 2010).
Background
Families and Sexual Minority Health
Families have substantial implications for individual health outcomes (Carr and Springer 2010). How long and how well we live is partly determined by the various resources that help or harm us, and families serve as immediate sources and distributors of social and economic security. In general, married persons enjoy substantial benefits over single persons, including better health and longer life (Waite and Gallagher 2000). Therefore, commentaries concerned with sexual minority health have begun to highlight the disjuncture between research establishing the health benefits of marriage and the legal hurdles that restrict sexual minorities from formally marrying (Herdt and Kertzner 2006; Herek 2006; King and Bartlett 2006; Umberson and Montez 2010). This is an issue of growing importance, as an estimated nine million individuals who self-identify as lesbian, gay, bisexual, and transgender live in the United States (Gates 2011), and numerous federal initiatives have identified gaps in knowledge surrounding sexual minority health (Institute of Medicine 2011; U.S. Department of Health and Human Services 2010). Important among these limitations is a relative void of social ecological investigations of sexual minority health (Institute of Medicine 2011), examining how collective entities, such as families and communities, affect individual health patterns.
Nevertheless, research on this subpopulation is increasing, and although it is rooted in specific areas of health (most prominently in sexual and mental health arising from the human immunodeficiency virus crises of the 1980s), researchers have begun to extend their focus to other differences in physical health status. Studies show that sexual minorities experience worse physical health than heterosexuals (Conron, Mimiaga, and Landers 2010), although patterns often differ by gender. Although gay men display poorer mental health and health behaviors (Dilley et al. 2010), they are also more fit and more likely to maintain healthier weights than heterosexual men (Brennan et al. 2010). In addition, Bybee et al. (2009) showed that the mental health disadvantage among gay men wanes with age, while across the life course, lesbians display consistently worse mental and physical health than heterosexual women (Case et al. 2004; Dilley et al. 2010), more often smoke cigarettes and drink alcohol to excess (Gruskin and Gordon 2006; Lee, Griffin, and Melvin 2009), and are more often overweight or obese (Boehmer and Bowen 2009).
Although increasing attention has been paid to the effects of family formation and support systems on adult health prospects, this body of work has yet to extend fully to sexual minorities (Umberson and Montez 2010). As a result, much existing research examines sexual minority health as if individuals were disconnected from one another and devoid of influence from essential social contacts via romantic partners. One of the few studies to examine the health status of partnered gays and lesbians (Wienke and Hill 2009) found that sexual minorities, and especially men, rated their health similarly to that of married persons and different-sex cohabitors. However, the study suffered from a small sample of sexual minorities (n = 282) and, importantly, was unable to determine the cohabitation status of all respondents with same-sex partners (Wienke and Hill 2009).
Family Structure, Resources, and Health
Cohabiting unions are similar to marital unions in that they involve coresidence, an intimate sexual relationship, and at least some economic consolidation, and a growing body of research suggests that the benefits of cohabitation and marriage may be quite similar in terms of psychological well-being, relationship quality and satisfaction, and social ties (Hansen, Moum, and Shapiro 2007; Musick and Bumpass 2006, 2012). However, studies have demonstrated that although both cohabitation and marriage are positively associated with health status, the health benefits of marriage appear stronger than those of cohabitation (Carr and Springer 2010; Musick and Bumpass 2006, 2012). For example, Wu et al. (2003) found that the self-rated health status of cohabiting adults is lower than that of married adults but higher than among never-married and previously married adults.
In explaining the positive relationship between marriage and health, scholars often point toward the health-promoting resources that marriage provides, including economic resources (Ross, Mirowsky, and Goldsteen 1990). SES is considered a fundamental cause of disease (Link and Phelan 1995), and scholars argue that health differences based on status inequalities, including sexual orientation and gender, occur through differential exposure to various demands and hardships (Pearlin et al. 2005). Research has also established important differences in the economic profile and benefits associated with marital versus cohabiting relationships (Blumstein and Schwartz 1983). Marriage is associated with a larger wage boost than cohabitation among men (see discussion by Light 2004), and overall, cohabitation is more common among those with less education and income (Smock and Manning 2004). Although the SES profile of sexual minorities is mostly unknown (Institute of Medicine 2011), limited evidence based largely on convenience samples suggests some SES advantages for sexual minorities (Valanis et al. 2000), such as higher levels of education than for heterosexual adults. However, other work suggests that this is often coupled with lower earnings (Factor and Rothblum 2007), at least for gay men. For women, some research suggests that lesbians have higher incomes than heterosexual women, although other work has found no such difference (Institute of Medicine 2011).
Beyond the uncertainty surrounding the SES profile of sexual minority adults, very little research examines the role of SES in contributing to health disparities between sexual minorities and heterosexual adults (Institute of Medicine 2011), especially in the context of family life; it is not known how much SES contributes to physical health differences across same-sex cohabiting, different-sex cohabiting, and married households. SES advantages provide some potential for better health among same-sex cohabitors, at least compared with different-sex cohabitors. However, if marriage confers particular economic benefits that can be applied toward better health (e.g., pooling money, which work by Blumstein and Schwartz 1983 shows is more common among married couples than either different- or same-sex cohabiting couples), then same-sex cohabitors likely have worse health than married persons.
However, scholars also point out that strict interpretations of the health benefits of marriage (and cohabitation) are likely skewed to some extent because of positive selection into relationships on the basis of factors that include health status and economic standing (Murray 2000; Oppenheimer, Kalmijn, and Lim 1997). And although unknown for sexual minorities, we can see no reason to presume that these factors would not contribute to selection into same-sex relationships as well. Although data limitations prohibit us from directly testing arguments related to selection versus causation, we are able to control for factors related to relationship status and health in our models, particularly socioeconomic standing but also gender, age, racial and ethnic identity, and whether children are living in the home. In addition, we present findings from supplemental analyses (described below) that permit a more stringent test of the extent to which socioeconomic standing (and other factors) may confound the accuracy with which we can observe the connection between relationship status and health.
Gender and Children
In examining the relevance of relationship status for health outcomes, gender is a key factor to consider. From a health standpoint, research shows that self-reported health status differs substantially by gender, with women reporting worse self-rated health on average than men (Gorman and Read 2006). Although some research questions whether a gender difference exists in the health benefits associated with marriage (Liu and Umberson 2008), other studies have shown that men experience a stronger health advantage associated with marriage versus singlehood than women (Rendall et al. 2011; Teachman 2010). For example, Williams and Umberson (2004) found that marriage is positively related to self-rated health, but only among men. Studies also show that marriage improves health through gender-specific routes, with increased social support and reduced risky behaviors more important for the health gains of married men and enhanced economic standing more important for the health gains of married women (Lillard and Waite 1995; Ross et al. 1990). Indeed, entering into marriage (or different-sex cohabitation) has been shown to increase total family income more for women than for men (Light 2004), and enhanced financial standing and access to health-related resources (e.g., medical insurance) is an important route through which marriage benefits the health of women (Wood et al. 2007).
There is also evidence that, moving beyond marriage, gender shapes the potential health benefits of other types of intimate relationships. For example, although Wu et al. (2003) found that married adults reported better self-rated health than cohabiting adults, this difference was significant only for women, a finding that likely relates to the higher economic standing of married versus different-sex cohabiting couples and the greater importance of SES for shaping health outcomes among women in relationships. An unanswered question is where same-sex cohabitors, male and female, fit into this pattern and its potential relevance for health standing. Research by Reczek and Umberson (2012) does suggest more gender similarity in the relationship dynamics for same-sex cohabitors than between men and women in marital relationships, which implies that the health benefits (or harms) of living in a same-sex cohabiting relationship may be similar for men and women.
Qualitative research also shows evidence of more gender similarity in economic standing between male and female same-sex cohabitors, as Blumstein and Schwartz (1983) demonstrated that most same-sex couples, male and female, believe that their partners should work for pay outside the home; different-sex cohabitors are less likely to adhere to this belief, but they do so more frequently than married adults. However, because of this, we might expect that although men in same-sex relationships would benefit in terms of health from higher average household wealth than men and women in all other relationship types (because of the presence of two male earners in the home and the persistent gender gap in wages in the United States; see Gottschalk and Danziger 2005), their health may also be harmed by diminished essential supports typically provided by women in different-sex relationships (see Reczek and Umberson 2012). Indeed, although men appear to benefit equally from marital and different-sex cohabiting relationships (Wu et al. 2003), this benefit may not hold in the absence of a female partner. At the same time, the health status of women in same-sex relationships might be harmed by a lesser income advantage than would occur if they were a member of a relationship that included a man.
Beyond gender, it is also likely that the health statuses associated with these relationships differ across families that do and do not include children. Census estimates indicate that among cohabitors, 34 percent of sexual-minority women and 22 percent of sexual-minority men are raising children (Gates and Ost 2004). However, the link between parenthood and physical health is not established (see discussion by Teachman 2010), with some researchers documenting similar self-rated health scores for parents and childless adults (Hughes and Waite 2002), others suggesting that the presence of children may be associated with illness and poor health (see Evenson and Simon 2005; Umberson and Montez 2010), and still others finding that the relationship is gender dependent, with parenthood showing no relationship to physical health among men but either a positive (Teachman 2010) or a negative (Wienke and Hill 2009) relationship to physical health for women. Research on same-sex couples with children is even more limited, but studies do suggest that same- and different-sex couples with children have similar partner relationship quality (Peplau and Fingerhut 2007). Although researchers debate whether children’s own health differs depending on whether they have same- or different-sex parents (Carr and Springer 2010; Marks 2012), the effects of children on health for adults in same-sex families remain underexamined. However, parenthood has been shown to increase fathers’ wages and decrease mothers’ wages (see discussion by Light 2004), suggesting a greater likelihood that the health of women (and especially women in same-sex relationships) may be harmed by the financial strains associated with child rearing.
Hypotheses and Objectives
The existing empirical evidence suggests conflicting expectations for the influence of same-sex cohabitation on health. On one hand, studies have found fewer health advantages associated with cohabitation than marriage, though this evidence has rarely distinguished between different- and same-sex cohabitors. On the other hand, the only study to do so to date found no health difference between same-sex cohabitors and married persons (Wienke and Hill 2009). Hypothesis 1 serves as our primary objective in this study: to describe how self-evaluations of health differ between persons in same- and different-sex relationships:
Hypothesis 1: All else equal, same-sex cohabitors will be more similar in health to different-sex cohabitors than to married persons and will be more likely than married persons to report fair or poor health.
An alternative hypothesis suggests that same-sex cohabitors will be more similar in health to married persons, whereas different-sex cohabitors will be more likely to report fair or poor health.
Next, we address the role of SES in contributing to group differences in self-rated health among men and women living in different- and same-sex relationships:
Hypothesis 2: Health disparities between adults living in same- and different-sex relationships will be partially explained by adjusting for SES.
In addition, because men are socioeconomically advantaged relative to women, we expect that adjustment for SES will also reduce any observed gender disparities in self-rated health.
We pull from existing evidence in literatures on gender, family, and health to explore two additional, underdeveloped areas in sexual minority health research. First, following research showing more positive health assessments among men, and established male advantages in health benefits from relationships, we posit our third hypothesis:
Hypothesis 3: Men in cohabiting relationships will report better self-rated health than will women in cohabiting relationships.
However, it is uncertain whether this male advantage will be enhanced or diminished for men in same-sex relationships. On one hand, studies have shown less health difference between gay and heterosexual men than between lesbian and heterosexual women, and same-sex cohabiting men may benefit from higher combined family income compared with same-sex cohabiting women. On the other hand, there is some evidence of gender similarity in relationship dynamics across male and female same-sex cohabitors, which suggests that both will receive similar health benefits from cohabitation.
Research is unclear regarding the health consequences of children in the home, although limited research does suggest that the presence of children is less important for shaping the health of men:
Hypothesis 4: Living with children will be less important for the health of men than women.
However, research documenting higher stress and diminished wages among mothers suggests the potential for negative health consequences among women. In addition, little research has been done on the effects of children on parental health across different relationship types, including sexual minorities, but because different-sex cohabitors tend to be of lower economic status than married persons, it may be that children are more of a health burden to cohabitors than a benefit.
Data and Methods
The data for our sample come from combining the 1997 to 2008 years of the National Center for Health Statistics (NCHS) National Health Interview Survey (NHIS), a cross-sectional survey aimed at understanding the correlates of health in the United States. The Integrated Health Interview Series (IHIS), compiled by the Minnesota Population Center (2010), streamlines an otherwise arduous data consistency process and provides the data files used here.
Starting in 1997, an unmarried partner category was included in the NHIS household roster, allowing researchers to examine various married and unmarried relationship types. There are no indicators of sexual attraction or identity in the NHIS, so partnerships are identified by matching the sex of respondent variable with the relationship to householder variable. Married couple households are identified by one man and one woman reporting married status. Same-sex cohabitors are identified by two men or two women reporting as partners and different-sex cohabitors as a man and a woman reporting as partners. Thus, the same-sex cohabiting couples here are consistent with those identified in census data and representative only of the “out” sexual minority population who report living with partners (Rosenfeld 2007). In addition, household roster information is used to identify households that do and do not include children under 18 years of age.
Compiling the 1997 to 2008 data resulted in a sample size of 665,658 adults aged 18 years or older. Nearly 70 percent of the sample over this time period lived in the partnered relationships identified above. Because our research questions center on comparisons between adults in relationships and because the NHIS does not include questions on sexual identity, we drop all adults who do not report partnerships (including those living alone or in other types of households). After we drop an additional 3.7 percent of the sample because of missing values on key variables used in the analysis, an acceptable loss of cases given that less than 5 percent of the sample was removed (Menard 2002), the sample includes 460,459 adults, including 3,219 same-sex cohabiting adults, 419,424 different-sex married adults, and 37,816 different-sex cohabiting adults. Importantly, same-sex cohabitors are likely underreported because they are identified through indirect methods (Gates and Ost 2004; Wienke and Hill 2009).
Measures
The dependent variable, self-rated health, is a subjective measure of well-being that has correlated strongly with objective measures of health, and even mortality, in past research (Idler and Benyamini 1997). An ordinal measure, self-rated health is measured as 0 (poor), 1 (fair), 2 (good), 3 (very good), or 4 (excellent), which we recode to contrast bad health (fair or poor health = 1) with good health (good, very good, or excellent health = 0), as have many previous investigations of self-rated health.
A set of demographic variables captures age, gender (1 = male), and race-ethnicity coded with dummy variables for non-Hispanic black, non-Hispanic other, Hispanic, and non-Hispanic white (reference). Socioeconomic measures include education, measured continuously from 0 (no formal education) to 16 (postbaccalaureate education), employment status (comparing unemployed and out of the labor force to employed persons), and total combined household income. The IHIS (Minnesota Population Center 2010) combines imputed income measures provided by NCHS across all years used here to account for the roughly 18 percent of NHIS respondents missing information on income. The NCHS uses multiple imputation techniques to create the imputed income data for 1997 to 2008 (Minnesota Population Center 2010). The income measure is adjusted for purchasing power over time and for different sized families. Our multivariate models include a logged transformation of the imputed measure to account for its skewed distribution; we estimated models with and without missing income data but found no discernible differences.
Finally, we include a region-of-residence variable to capture potential confounding in both regional differences in health and residential choice on the basis of relationship type and an indicator of the year in which the survey was conducted. More detailed geographic indicators are not available in the public-use version of the NHIS.
Information on health behaviors is available only in NHIS sample adult files, and because information is obtained only from a randomly sampled household adult, including these measures substantially reduces the sample of same-sex cohabiting adults. This represents an important omission in the present study, as it limits our ability to identify factors that contribute to differences in health status by household type and gender. A dichotomous measure of health status, the presence of activity limitations, was available for all years and was included in supplementary models, with no limitations serving as the referent. As we expected, activity limitations exerted powerful direct effects on health evaluations, but in general, the measure did not alter the relationships between relationship type and self-rated health and thus was excluded in the final models for parsimony (results available on request).
Estimation
We begin by presenting sample characteristics for all respondents and stratified by gender. Next, we estimate weighted logistic regression models for the full sample and include interactions between relationship type and gender as well as relationship type and the presence of children in the household. Next, we estimate gender-stratified models to examine differences in the association between relationship type and self-rated health both within and between gender groups. We use Stata 12 (StataCorp LP 2010) for all analyses and use “svy” commands to adjust for clustering within households.
Results
Table 1 provides weighted means and proportions for our analytic sample and includes tests of significance between same-sex and other relationship types for the full sample and by gender. The table reveals some descriptive differences in self-rated poor to fair health between the same-sex cohabiting respondents and different-sex married and cohabiting respondents, particularly among men. Significantly fewer same-sex cohabitors (9.2 percent) reported poor health than different-sex married (11.0 percent) and cohabiting (10.6 percent) persons. This statistically significant pattern in reported poor health was present for men but not women.
Weighted Means and Proportions by Relationship Type and Gender (N = 460,459).
Source: Integrated Health Interview Series 1997 to 2008 (Minnesota Population Center 2010).
Note: The table presents tests of significant differences between same-sex cohabitors and different-sex married persons and between same-sex cohabitors and different-sex cohabitors for full samples and by gender.
p ≤ .05. **p ≤ .01. ***p ≤ .001.
Table 1 shows many other significant differences across same- and different-sex relationships. About one in five same-sex cohabitors live with children, whereas half of married persons and over 40 percent of different-sex cohabitors do. There are important gender differences within and between relationship types as well. Only 14 percent of same-sex cohabiting men live with children, compared with nearly 30 percent of same-sex cohabiting women. Same-sex cohabitors are also younger, on average, than married persons but older than different-sex cohabitors. In racial composition, same-sex cohabiting men are more similar to married men, while same-sex cohabiting women are more similar to different-sex cohabiting women. In general, more same-sex cohabitors are non-Hispanic white than their different-sex counterparts, and different-sex cohabitors are more often non-Hispanic black or Hispanic.
Finally, Table 1 shows important differences in SES across relationship types. Same-sex cohabitors hold significant advantages over married persons and different-sex cohabitors in education, and the gaps are similar for men and women. We also see significantly higher household income for same-sex cohabitors, especially when compared with different-sex cohabitors. Although in different-sex married and cohabiting households, there is almost no gender difference in household income, same-sex cohabiting men live in households that earn almost $9,000 more than households of same-sex cohabiting women. In terms of employment, same-sex cohabiting women work at higher rates than different-sex cohabiting and especially married women (we see a nearly 30 percent gap in the employment rate between same-sex cohabiting and married women), while same-sex cohabiting men work at a somewhat higher rate than married men but slightly less than men in different-sex cohabiting relationships.
To examine the association of relationship type and the other covariates on reporting poor or fair health, we turn to the logistic regression results in Table 2. Model 1 includes all covariates except for SES and shows that same- and different-sex cohabitors are 19 percent and 72 percent more likely than married persons to report poor or fair health, respectively. Although both same- and different-sex cohabitors have increased odds of poor health compared with married persons, additional tests show that the difference between the two is significant (t = −4.28, p ≤ .001), meaning that different-sex cohabitors have significantly higher odds of poor health than same-sex cohabitors. Model 1 also shows that the presence of children in the household decreases the odds of reporting poor health by 6 percent. Furthermore, Model 1 shows that men are less likely to report poor health, but once we adjust in Model 2 for the lower socioeconomic standing of women, men emerge as slightly more likely to report poor health than women. The likelihood of reporting poor health increases with age and is nearly two times higher among blacks and Hispanics and 18 percent higher among non-Hispanic others compared with whites. However, these increased odds by race are suppressed considerably with the inclusion of SES in Model 2. Finally, Models 1 and 2 show that there are important regional differences in self-reported health and that the odds of reporting poor health have increased over time. 1
Odds Ratios and 95% Confidence Intervals, Logistic Regression Predicting Poor or Fair Self-Rated Health (N = 460,459).
Source: Integrated Health Interview Series 1997 to 2008 (Minnesota Population Center 2010).
p ≤ .05. **p ≤ .01. ***p ≤ .001.
Model 2 of Table 2 shows that more education and income decrease the odds of reporting poor health, and not working increases them. Supplementary analyses interacting relationship status with the measures of SES found that SES effects do not vary across relationship types (available on request). Importantly, accounting for SES adjusts for advantages for same-sex cohabitors and disadvantages for different-sex cohabitors, placing them at similarly increased odds of reporting poor health (odds ratio [OR] = 1.29 vs. 1.25, t = .36, p = .72), relative to their married counterparts.
Model 3 of Table 2 interacts gender with relationship type and indicates that men in cohabiting relationships fare better than women in self-ratings of health. Although the interaction ORs are similar in magnitude and both less than 1.0, the interaction for same-sex cohabitors does not reach significance, but the interaction for different-sex cohabitors does. Furthermore, the interaction for children by relationship type in Model 4 provides potential evidence that children are more of a burden than a benefit for same-sex cohabitor health, though the interaction does not reach significance. To further explore these differences, Table 3 stratifies by gender.
Odds Ratios and 95% Confidence Intervals, Logistic Regression Predicting Poor or Fair Self-Rated Health, by Gender.
Source: Integrated Health Interview Series 1997 to 2008 (Minnesota Population Center 2010).
p ≤ .05. **p ≤ .01. ***p ≤ .001.
Indeed, Model 1 for men and women in Table 3 shows that, relative to their married counterparts, different-sex cohabiting men and women face similar increased odds of poor health before adjusting for SES (OR = 1.68 vs. 1.76, t = −1.35, p = .18). After adjustments for SES, Model 2 shows that different-sex cohabiting men face increased odds relative to married men (1.14), but different-sex cohabiting women (1.34) face even higher odds relative to married women (t = −4.33, p ≤ .001). For same-sex cohabiting men and women, the evidence suggests both groups face increased odds compared with married persons. The ORs for same-sex cohabiting men are above 1.0 but do not reach significance at the .05 level in either Model 1 or 2. However, women in same-sex cohabiting relationships do have statistically higher odds (1.34) of poor health compared with married women only after controlling for SES, but the difference across gender for same-sex cohabitors does not reach significance (OR = 1.18 vs. 1.34, t = −.76, p = .45). Furthermore, after adjusting for SES in Model 2, the odds of poor health for same-sex cohabiting men do not differ significantly from the odds for different-sex cohabiting men (t = .30, p = .76), and the same comparison is true of women (t = .02, p = .98). In all, the evidence is suggestive that both men and women in same- and different-sex cohabitations have higher odds of poor health than married persons, though different-sex cohabiting men fare slightly better than different-sex cohabiting women after adjusting for SES.
The sociodemographic, socioeconomic, geographic, and year-of-survey covariates affect the health ratings of men and women in a similar fashion. But the presence of children in the household protects against poor health ratings for women but not for men in the baseline model. After adjusting for SES (Model 2), children are protective for both men and women, but additional tests reveal that children are significantly more protective for women than for men (OR = .86 vs. .72, t = 6.62, p ≤ .001). Model 3 of Table 3 examines whether the effect of children differs by relationship type. For men, the OR for same-sex cohabiting men by children is far below 1.0 but has a wide confidence interval (CI), in part because of the relatively few same-sex cohabiting men who live with children. Similarly, different-sex cohabiting men do not appear to differ significantly from married men on the protective effect of children. For women, both interaction ORs exceed 1.0, indicating that children are more of a burden than a benefit for health for cohabiting women relative to married women. This effect does not reach significance for different-sex cohabiting women, but it does so for same-sex cohabiting women. Again, the CI for the interaction for same sex women is wide, so the magnitude of this effect should be viewed with caution. In all, this suggests that the protective effect of children is larger for women than for men but that, among women, the protective effect of children is clearest among married women. Indeed, the protective effect of children on health is specific to both gender and relationship type.
Robustness of Results
Given significant differences in the distribution of many of the covariates across relationship types (see Table 1), we used a matching methodology and reestimated the regressions presented in Tables 2 and 3 on a sample matched by relationship type, with belonging to a same-sex couple serving as our treatment group and belonging to an different-sex couple as our control group. We use a matching strategy as a robustness check on our results, as it serves as a method of adjustment that minimizes differences across relationship types for important covariates such as education and income before estimating the regressions. Given the likelihood that selection effects operate along sociodemographic and socioeconomic lines to shape who does and does not enter into a coresidential union, this preprocessing step may better account for some of those factors than traditional methods used on cross-sectional data. We used coarsened exact matching (CEM) techniques because CEM directly reduces imbalance in covariates and deals with model dependence and bias before estimation, unlike other matching techniques that deal only indirectly, and after the fact, with these issues (see Blackwell et al. 2009 and Iacus, King, and Porro 2008 for a detailed discussion of CEM and how to implement it using Stata). 2
Providing increased confidence in our reported results, the regressions estimated on this matched sample are consistent with results presented above (matched results available upon request). For example, after matching, same-sex (OR = 1.44; 95 percent CI = 1.24–1.68) and different-sex (OR = 1.39, 95% CI = 1.29–1.49) cohabitors had a statistically similar increased odds of reporting poor health compared with their married counterparts. That is, after adjusting away the influence of SES (and the other covariates) a priori through matching, same- and different-sex cohabitors are, once again, similarly at higher odds of reporting poor health. Furthermore, in gender-stratified models using the matched sample, the odds of reporting poor health for men and women in same-sex relationships did not differ significantly from each other or from the odds of men and women in different-sex cohabitations.
In addition, the matched results show that the presence of children is more protective of women’s than of men’s health. Finally, the matched results for women suggest the protective effects of the presence of children on health prospects are specific to married women, as found in the unmatched sample.
Discussion
With over 600,000 same-sex households in the United States (Gates and Ost 2004), sexual-minority families are an important, yet poorly understood group. Although previous research has been hampered by inadequate sample size, reliance on convenience samples, and no record of relationship status (Institute of Medicine 2011), existing work does document a general health disadvantage among sexual minorities. Drawing on household roster information from the NHIS, our study allows us to produce a rigorous comparison of the health status of adults living in same- versus different-sex relationship families. Although data limitations restrict what we can say about sexual orientation and health, we do contribute to a growing body of work specifying the effects of relationship orientation on health (Reczek and Umberson 2012). We posited four hypotheses involving the associations between family type, gender, resources, and self-rated health status and find varying support for these in our analyses.
First, after adjustment for demographic and regional confounders across groups, Hypothesis 1 finds some support in that same-sex cohabitors, like different-sex cohabitors, are significantly more likely to report poor or fair self-rated health than married adults. But additional tests reveal that the increased odds of poor health for same-sex cohabitors are less than the increased odds for different-sex cohabitors. These findings add to those of Wienke and Hill (2009) and are in line with research on different-sex cohabitation that documents health disadvantages relative to married adults (e.g., Wu et al. 2003). This suggests that same-sex cohabitation, much like different-sex cohabitation, is not equivalent to marriage in terms of the health protections it affords. Indeed, providing legal avenues for same sex couples to marry offers at least the possibility of improved health.
Research on different-sex cohabitors suggests that these relationships can involve fewer resources and less relationship satisfaction, both of which could impair health. Research also indicates, however, that most same-sex couples are in dual-earner relationships and appear to value power equality as an important part of their relationships (Peplau and Fingerhut 2007). Past research also suggests that same-sex cohabitors and married persons report similar relationship quality (Kurdek 2004), and that may translate into greater instrumental and emotional support that can be used to improve health. In sum, the mechanisms producing similarly poor health for same- and different-sex cohabitors may be quite different. Whereas we are unable to investigate these plausible explanations, it is important that population-level research on sexual minority health do so as soon as data collection allows (Institute of Medicine 2011).
Our data do let us test explanations relating to SES, and as expected, we find a significant, positive association between SES and self-rated health. More education, more income, and better employment prospects correspond with better health across relationship types. Importantly, and in support of our second hypothesis, we find that accounting for SES adjusts away observed advantages for same-sex cohabitors and disadvantages for different-sex cohabitors so that they have similar increased odds of reporting poor health, compared with their married counterparts. In addition, SES has a larger explanatory effect among men than women. Adjusting for SES reduced the disparity between married and different-sex cohabitation more so for men than for women; as a result, after adjustment for SES, different-sex cohabitation is more strongly associated with poor health for women than for men.
Regarding Hypothesis 3, we find limited support for our prediction that men in cohabiting relationships would report better health than women in cohabiting relationships. After accounting for SES and other important confounders, both men and women in same-sex cohabiting relationships show higher odds of poor health than married persons, and the difference between same-sex cohabiting men and women is negligible. Different-sex cohabiting men and women also display higher odds of poor health, but we found further evidence that men in these relationships have lower odds than different-sex cohabiting women, relative to their married counterparts. Historical (Blumstein and Schwartz 1983) and recent (Reczek and Umberson 2012) evidence documenting more similar gender role expectations within same-sex relationships versus those in different-sex relationships provides a framework in which to understand these findings. Men partnered with men and women partnered with women may assign gendered roles and expectations within relationships in a more collaborative fashion, and that may result in more similar health prospects for men and women in same-sex partnerships.
Finally, we find support for Hypothesis 4 in that living with children is more strongly related to the health of women than men. However, we identify an additional caveat in that this applies mostly to women in different-sex married relationships. In fact, for women in cohabiting relationships, the presence of children may be more of a burden than a benefit to health. This is consistent with identified social and economic consequences of child rearing for women (Light 2004). Previous research suggests that social and economic disadvantages come with health consequences for different-sex cohabiting women, and we have identified a health deficit for same-sex cohabiting women, though we are limited in what we can say causes this disadvantage. Nevertheless, these health disadvantages suggest the presence of additional stress and strains for women in cohabiting relationships, compared with their married counterparts, and this may be exacerbated by the presence of children.
The NHIS provides a unique opportunity to study the health experience of adults in same-sex relationships, but it has some important limitations, most notably the lack of direct measurement of sexual identity, which forced us to determine sexual minority status indirectly via the household roster. This is potentially problematic, because existing work on sexual minority health and health behaviors has revealed the multidimensionality of sexual orientation (McCabe et al. 2009). Even so, examining the health of sexual minorities in self-identified relationships with partners in which they share a home, and in some instances live with children, provides important insights into the health of sexual minorities within the realm of social relationships and health (Carr and Springer 2010; Umberson and Montez 2010). In addition, the NHIS lacks information on relationship duration or quality, despite research showing that these factors are quite relevant for a relationship’s impact on health and health behavior (Wu and Hart 2002). These concerns highlight the need for more consistent population-based initiatives to collect detailed information on sexual orientation and relationship duration and quality (Dilley et al. 2010).
The uncertainty surrounding selection into relationships and how selection factors may vary across same- and different-sex relationships (Carpenter and Gates 2008), together with our inability to adequately account for these issues because of the cross-sectional nature of the NHIS data, means that some caution should be exercised in interpreting the association observed between relationship type and health from our findings. Although some scholars have found evidence that theories relating to social causation offer a better explanation for differences in well-being than selection effects (e.g., Brown 2000), these processes which select persons into relationships call for more research and dedicated data collection efforts aimed at gathering information on relationship types and transitions over time among sexual minority and majority adults.
Importantly, although we cannot elaborate empirically on some of our findings, emerging work provides a base from which to move forward. For example, the minority stress framework links discrimination to poorer health for sexual minorities because of hostile and stressful social environments created through fear, stigma, and prejudice (Meyer 2003). Sexual minorities report more lifetime and day-to-day discriminatory strains, which are associated with poor self-rated health, chronic health conditions, high blood pressure, and psychological distress and disorders (Thoits 2010). Without NHIS measures on discrimination, it remains unknown how those effects are either enhanced or mediated through a same-sex relationship. However, it is plausible that the residual negative health effects of same-sex cohabitation lie in these discrimination and stigma pathways.
Furthermore, understanding gender roles in relationship context likely plays a large part in understanding health across same- and different-sex relationships. Men have historically enjoyed larger benefits from intimate social relationships, in part because of the heavy influence of wives in promoting their health. Recent work by Reczek and Umberson (2012) identified a process they termed “cooperative health behavior work,” in which persons in same-sex relationships more often mutually influence each other’s health behaviors, compared with those in different-sex relationships. This process, in all likelihood, applies to both positive and negative health behaviors and operates to some unknown extent in the health patterns observed in our sample as well.
Recent years have witnessed a growing public debate in the United States as to what constitutes a family and who deserves legal recognition as part of one. The importance of marriage as the standard for meaningful, productive, and thus supportive relationships has divided scholars into competing camps (Wienke and Hill 2009). On one side, arguments for alternative family configurations suggest that marriage as the standard is becoming outdated, though institutionally supported. In the modern world, intimacy, not formality, may be the key to productive and healthy relationships (Giddens 1992), and the importance of marriage as an institution may be waning (Cherlin 2004). On the other side, researchers contend that the health benefits of marriage still outweigh the benefits of other relationship types (Waite and Gallagher 2000), and empirical evidence on health and social relationships generally finds continued support for those notions.
If marriage can improve health more than other family formations, what are the ramifications for groups with restricted marriage rights? Although legalizing same-sex marriage will not complete the movement toward marriage equality or improve health right away (Lau and Strohm 2011), it provides the potential for better health for a growing and increasingly visible minority population. It remains to be seen whether sexual minority health will improve along with growing public acceptance of alternative unions (Powell et al. 2010) and movements toward equality and whether policies will be put in place to nurture such progress.
Footnotes
Acknowledgements
We thank Rachel Kimbro, Rose Meideros, Hilary Dowdy, and the anonymous reviewers and editors for comments on earlier drafts; the Minnesota Population Center for producing the Integrated Health Interview Series data; and the Kinder Institute Urban Health Program at Rice University for supporting the research.
Authors’ Note
This article benefited from comments generated during presentations at the University of Texas at Austin’s Population Research Center, Rice University’s Department of Sociology Speaker Series, and at the 2012 American Sociological Association annual meetings in Denver, Colorado.