Gender, Union Formation, and Assortative Mating Among Older Women

Abstract

This paper presents a search-theoretic model of union formation among women, aged 55 older. Specifically, it provides new estimates of gender differentials in cohabitation and marriage at older ages, and documents recent patterns of assortative mating using data from the 2008–2017 American Community Survey. Our analyses reveal that cohabitation represents a much smaller share of all older unmarried women, all partnered women, and all women in comparison to patterns observed among their male counterparts. The results also reveal highly uneven patterns of union formation by age, race and marital history, which reflect demographically uneven constraints and preferences. Our analyses also document, for the first time, patterns of assortative mating at older ages. Shortages of similarly-aged men, especially among older African American women, seemingly heighten the likelihood of demographically mismatched unions. Older women are less likely to form unions with same-race or economically attractive partners, defined as men having a college-degree. This study shows that older single women, in general, are at a comparative disadvantage in the marriage market, both in forming co-residential unions and in finding partners who match their own social, demographic, and economic profiles. This paper highlights considerable heterogeneity in the experiences of America’s older women. It calls for new theoretical approaches that acknowledge the unequal resources and bargaining power among older women in the marriage market.

1. Introduction

Cohabitation has become the modal pathway to first marriage in the United States and other high-income countries (Sassler & Lichter, 2020; Smock & Schwartz, 2020). Not surprisingly, previous studies have focused primarily on young adults—on the incidence, timing, and sequencing of cohabitation and marriage. Increasing rates of cohabitation, however, are not limited to young adults, and are now observed across the entire life course, even at older ages (Brown, Lee, & Bulanda, 2006; Sassler, 2010; Vespa, 2012). For persons aged 50 and above, the number of cohabiters increased more than 4-fold, from roughly 951,000 in 2000 to over 4 million in 2016 (Brown & Wright, 2017). Increases in cohabitation at older ages are a direct consequence of high rates of divorce among America’s aging baby boom generation, including rapidly increasing divorce rates at older ages (Brown & Lin, 2021), and continuing declines in remarriage rates in mid-life and beyond (Raley & Sweeney, 2020). For older persons, romantic relationships and co-residential unions—both marriage and cohabitation —are now being transformed as never before.

The graying of America’s baby boomers makes this an especially propitious time to examine growing complexity in union formation among older persons (Carr, 2019; Sassler & Lichter, 2020). Older women are less likely than their male counterparts to be partnered in coresidential cohabiting unions, a gender gap that increases as the supply of potential male partners declines with age (Brown & Wright, 2017). Older single women, especially historically disadvantaged women, face severe demographic shortages of potential partners in the marriage market. Less well appreciated, however, is that large and growing gender imbalances at older ages also affect the likelihood of forming equitable or demographically-matched cohabiting and marital unions. Few studies have examined whether older women are entering cohabiting unions at the expense of marital unions (Brown et al., 2019) or, more importantly, whether patterns of mate selection are expressed in more demographically mismatched unions (e.g., intermarriage). Our theoretical and empirical approach takes an explicit gender perspective—one that acknowledges large gender disparities in the matching of partners at older ages.

Our paper has two main objectives. One objective is to build on previous demographic studies of the rise in elderly cohabitation (Brown et al., 2019; Brown & Wright, 2017; Lin, Brown, & Hammersmith, 2017). Specifically, we evaluate heterogeneity in union formation—both into cohabitation and marriage—among older women, contrasting their experiences with those of older men. Brown et al. (2016) showed that 76 percent of divorced women, aged 50 and older, were unpartnered in 2010, compared with 57 percent of divorced men. Among widows, only about 1 in 20 were partnered compared with 1 in 4 among widowers. In this paper, we show that national averages often obfuscate substantial heterogeneity in the experiences and pathways to cohabitation among America’s older women. Although the reasons are often unclear, cohabitation has seemingly become an alternative to marriage among many older women, even as others have opted in large numbers to live alone. Our study, using up-to-date nationally representative household data (linking the household records of male and female partners) on recent marriages and cohabitations from the American Community Survey, provides an empirical baseline highlighting extraordinary diversity in the living arrangements of older women.

A second objective is to highlight how demographic opportunities and constraints are revealed in patterns of intermarriage—in older women’s parings with demographically mismatched partners. With whom do older women form cohabiting and marital unions? Specifically, how are older women matched to their male partners on conventional demographic traits, such as age and race and ethnicity, and on their education and earnings, which undoubtedly reflect differences in the material resources often needed to effectively act on their preferences? Our analyses highlight uneven patterns of hypogamy, homogamy, and hypergamy among older women living in coresidential intimate unions. Compared with women in first marriages, we show that newly married and cohabiting older women are far more likely to be in hypogamous marital and cohabiting unions—to experience downward mobility on conventional demographic indicators of social status. One implication is that older women—especially those with fewer personal or financial resources—are disadvantaged in the marriage market.

2. Background

2.1. Cohabitation and mate selection at older ages

Conventional theories of marriage and assortative mating are typically rooted in a cost-benefit economic framework or in utilitarian models of marriage that emphasize a fair or equitable exchange in the marriage market (Becker, 1981; Oppenheimer, 1988). A search-theoretic model of marriage argues that marriage markets provide an arena for social interaction among eligible mates, where women (and their potential partners) seek to maximize utility or equitable returns to their own valued traits, such as physical attractiveness, youthful age, or a good job. The marriage market is in disequilibrium if women face constraints or difficulties finding suitable or well-matched partners.¹ Demographic mismatches in the marriage market—or shortages of “marriageable men”—presumably affect both the likelihood of marriage (vis-à-vis singlehood or cohabitation) and prospects for forming homogamous or demographically matched unions (England, 1986; Lichter, Price, & Swigert, 2020; Schwartz, 2013).

In early adulthood, the heterosexual marriage market is balanced numerically, with roughly similar numbers of single men and women available to form unions.² For example, the sex ratio in 2018 at ages 25–34 was 103.4 men per 100 women (U.S. Census Bureau, 2019); young women’s marital prospects presumably benefit from the surfeit of men. Of course, sex ratios often differ widely across geographic space and different population groups (e.g., among African Americans), which contribute to differentials in marriage and cohabitation, and in patterns of assortative mating (likes marrying likes) (Cohen & Pepin, 2018; Qian, Lichter, & Tumin, 2018). Opportunities to form co-residential unions also decline as women age, a result of higher male than female mortality rates at each successive older age (Beltrán-Sánchez, Finch, & Crimmins, 2015). At ages 50–59, for example, there were nearly 20.9 million men and 21.9 million women in 2018. This small surfeit of women almost doubled by ages 60–69, from nearly 1 million excess women at ages 50–59 to 1.9 million. As women become older they face declining opportunities to form unions. The implications seem clear: Older women are increasingly less likely to form unions, especially marital unions.

Sex ratio imbalances for union formation are likely to be highly differentiated among older women. As an illustrative case, African American women, in general, face severe shortages in the demographic supply of economically attractive African American men (Raley, Sweeney, & Wondra, 2016; Wilson, 1987), who continue to face high rates of unemployment and underemployment as a result of job discrimination and low education (Cohen & Pepin, 2018; Kuo & Raley, 2016). Mass incarceration also has removed a disproportionate share of black men from the marriage market (Charles & Luoh, 2010). In addition, rates of out-marriage with whites are much lower for black women than for black men, which places additional downward pressure on union formation (Qian & Lichter, 2011;Van Bavel, 2021). For older black women, sex ratio imbalances are especially large, a result of lifelong gender and racial disparities in mortality. Cultural adaptations, such as the formation of multigenerational households (e.g., doubling up) or supportive fictive kin, may also reduce economic incentives to marry or encourage marrying unwisely (Park, Wiemers, & Seltzer, 2019; Taylor et al., 2021). Historically, the limited marital opportunities among African American women—throughout the life course—have made them less dependent on male providers and encouraged their own financial independence (e.g., higher employment rates among black women). Large shares of African American women today are unmarried at older ages (Addo & Lichter, 2013). For older black or other economically disadvantaged women who marry or cohabit, studies of whether they enter matched or economically advantageous unions are rare.

Under the circumstances, it is hardly surprising that older women are less likely to be married than their male counterparts (Brown et al., 2016). Sex ratio imbalances—and cultural and economic adaptations to them—provide a narrowly circumscribed demographic view of marital opportunities and constraints. In perhaps the most comprehensive study of elderly union formation to date, Brown et al. (2019), using the Health and Retirement Survey, found no evidence of statistically significant differences in the likelihood of either cohabitation or remarriage vs. being unpartnered following “gray divorce” or, more importantly, in the relative risk of transitioning to cohabitation vis-à-vis remarriage. Perhaps surprisingly, divorced older women’s education, employment status, homeownership, wealth, pension income, or health insurance seemingly did not influence the transition to marriage in their discrete-time multinomial logistic regression models of repartnering. Statistical power was limited by HRS’s small sample size (N=534), which prevented the authors from evaluating assortative mating among cohabiting and married couples. Nevertheless, for education, employment status and wealth—all key economic indicators—the relative risk ratios did not differ significantly from 1.0 at conventional statistical levels (p < .05), although Brown et al. (2019) provided some limited evidence that remarried persons were slightly more likely than those remaining single to have more education and be employed. The large sample size of the ACS, which we use here, provides substantially more observations (and statistical power) to revisit the question of whether personal and financial resources affect marital status at older ages and the kinds of unions (i.e., marriage or cohabitation) for very different segments of the U.S. population, including racial and ethnic minorities.

2.2. Opportunity, third-party constraints and preferences at older ages

Patterns of assortative mating, like union formation more generally, are shaped by demographic constraints, marital preferences, and third-party constraints (Kalmijn, 1998). The implications of demographic constraints on patterns of mate selection, both in marriage and cohabitation, seem self-evident but rarely are studied at older ages.³ Difficulties finding suitable partners presumably reduce the likelihood of marriage, but also affect the marital matching process. Older single women may either drop out of the marriage market or choose instead to cohabit. Or they may “cast a wider net,” forming unions with partners who may be of a different age, race, or socioeconomic status. For older women who enter unions, the likelihood of forming hypogamous unions (indicating “downward” mobility on some trait) is likely to increase as opportunities to marry or cohabit decline with age. They may lower their “reservation-quality partner” (England & Farkas, 1986) and form partnerships with “minimally acceptable” older men. The clear implication is that older women may be disadvantaged in the marriage market, especially if measured by the kinds of matches they form in cohabitation and marriage.

Third-party constraints typically refer to the normative environment—both informal and legal constraints—that influence women’s choice in partners. These operate quite differently at different ages.⁴ For example, older women are usually unconstrained by parental disapproval, but nevertheless may be influenced by their adult children. Adult children may discourage romantic relationships of any kind—mismatched or not—or favor cohabitation over marriage if new partners are viewed as a threat to kin alliances or to their inheritances. Older women seemingly have fewer cultural or normative constraints when it comes to forming intimate relationships. There may be a loosening from family and friends of the usual normative pressures for homogamy (e.g., same-race or same-age partners) or for forming relationships with “economically-attractive” partners. Without minor children involved, longer-term emotional and financial commitments also may matter less among older women, especially those who are highly educated and financially secure. For widows, current provisions in Social Security or pension plans also arguably create economic disincentives to forming marriages, including mismatched marriages. But this may not be the case among older economically-disadvantaged divorced women who cannot count on alimony or their former partner’s pension or Social Security (Lin, Brown, & Hammersmith, 2017; Tamborini, Iams, & Whitman, 2009). Even at the same age, widows are far less likely than divorced women to re-partner (Brown et al., 2016; Schimmele & Wu, 2016). For widows, the economic or financial imperative to marry or cohabit may be weak.⁵ The loosening of third-party constraints provides a new freedom to remain single rather than cohabit with or marry demographically mismatched partners.⁶

Preferences in marital or cohabiting partners also may change in predictable ways as women grow older. Infecundity among older unmarried women reduces both the biological ability and expectation to bear additional children; indeed, by age 55, childbearing is completed for virtually all women. For younger women, marriage is often motivated by fertility and the desire to have a family (Brown et al., 2019; Raley & Sweeney, 2020). As women age, the specific circumstances—economic and otherwise—that previously attracted them to specific kinds of partners (i.e., those likely to be good fathers and providers) are likely to shift.⁷ For example, any preferences against forming unions with younger men decline significantly with age, as do preferences excluding divorced men or fathers as potential partners (Rapp, 2018; South, 1997). Preferences are shaped by normative expectations about age-appropriate or age-inappropriate partners. Whether age-graded marital preferences simply reflect demographic reality (i.e., the knowledge that the supply of “suitable” men available to them declines with age) is unclear.⁸ One argument, albeit untested, is that companionship matters most among older women (Watson & Stelle, 2021).⁹ At older ages, cohabitation may be viewed as viable alternative to marriage rather than as stepping-stone or testing ground for marriage and family life, a pattern especially common among young adults (Lundberg, Pollack, & Stearns, 2016; Sassler, Michelmore, & Qian, 2018). Cohabitation may supplant remarriage for many older women. In a study of on-line dating, McWilliams and Barrett (2014) showed that older women are not anxious to remarry, especially if they had concerns about returning to traditional gendered marital roles, including caretaking (Brown & Shinohara, 2013).

3. Current study

Our goal is to document patterns of marriage and cohabitation among older women, recognizing that patterns of assortative mating are bound by demographic constraints, personal preferences, and third-party constraints. We cannot easily disentangle the causative forces influencing older women’s decision-making, but we can provide an empirical baseline and analytic framework that acknowledges competing factors and substantial heterogeneity in older women’s living arrangements. This is important today for developing age-appropriate search-theoretic models of marriage across the life course. Demographic constraints in the marriage market affect the choices older women make in ways that reduce the likelihood of marriage (vis-a-vis cohabitation) and, as we argue here, fundamentally alter patterns of assortative mating. This singular explanation—one of demographic destinies—is commonplace and intuitively appealing. However, it also can mask substantial heterogeneity among older women, minimizes personal agency, and potentially misrepresent the diverse marriage market circumstances and the personal motivations that shape decision-making.

As a guiding principle, we therefore hypothesize that women with more personal and financial resources—those who are younger, white, more educated or have higher personal incomes—will be more favorably positioned to exercise their options in the marriage market (e.g., Han, 2021).¹⁰ Among older women, for example, we expect those at the younger end of the age distribution will have more choices among potential partners. Moreover, in America’s racially stratified society, the prospect of forming unions is likely to be greater among older whites than it is among African American and Latino women (see Schwartz, 2013). From a market perspective, white women also have more choices than disadvantaged women of color, who face clear age-graded shortages of potential male partners. Finally, education and income, including earnings and pensions, presumably provide greater financial security or independence, which give older women more opportunities to be selective in choosing among potential male partners. As shown in this paper, older women experience very different opportunities and constraints in the marriage market, which are revealed in highly uneven patterns of union formation and assortative mating.

4. Methods

4.1. Data

To address our empirical objectives, we use data from the 2008–2017 American Community Survey (ACS). The ACS, which replaced the long form of the decennial census, has many features that are especially well suited to our empirical goals, including its large, nationally-representative samples of older populations. Unlike most other national surveys (such as the HRS), household records include information on both marital and cohabiting partners. Our primary sample includes 5,243,038 women and 4,174,223 men aged 55 and over.¹¹ We also capture 2,754,825 marital or cohabiting unions in those households where women are aged 55 years and above.¹² This sub-sample provides the basis for our analyses of assortative mating. Much of analysis is disaggregated by age, separating older women and men into three age groups (55–64, 65–74, and 75 and older). This is important because marriage market conditions (i.e., sex ratio imbalances) impose additional constraints on women as they age. We use personal weights in percentage distributions for descriptive analyses based on men and women separately and household weights in percentage distributions for descriptive analyses based on couples.

4.2. Measurement

Our analyses focus on whether women have recently formed cohabiting or marital unions, while also considering their previous marital statuses (never married, divorced/separated, and widowed). ACS data identify women’s recent marital transitions, including year last married and the number of times married. The incidence of remarriages can also be examined for both divorced/separated and widowed women, who presumably face different age-related marriage market constraints and may have different motivations—especially economic motivations—for marriage or cohabitation (Qian & Lichter, 2018).

Data on the linked lives of married and cohabiting women and their co-residential partners are an especially useful feature of the ACS, which is illustrated in several recent studies of mate selection and assortative mating (Qian & Lichter, 2011; Choi, 2020). Older women may be either married or cohabiting with male householders or are themselves householders living with husbands or partners. The sociodemographic characteristics of women’s heterosexual partners can be compared and any differences in mate selection patterns can be detected.¹³

Our analyses of linked partner characteristics document patterns of assortative mating by age, race and ethnicity, education and income, including the distribution of homogamous and heterogamous unions, both hypogamy (downward partnering) and hypergamy (upward partnering) among older women. We define age-homogamous unions as those including partners within 3 years of age. Race is defined by OMB’s classification: Hispanic or Latino (of any race), non-Hispanic white, non-Hispanic black (or African American), Asian American, American Indian, and Alaska Native, Native Hawaiian and other Pacific Islander. Couples are considered racially homogamous if partners report the same race or ethnicity (as defined above).

We also consider the sorting of married and cohabiting partners on the basis of educational attainment and personal income. Couples are defined as educationally homogamous if both partners share the same broad educational categories (i.e., less than high school, high school degree, some college, and a college degree or more). Following convention in the demographic literature, women with more education than their marital or cohabiting partners are considered educationally hypogamous unions while men with more education than their female partners are in hypergamous unions (Blackwell & Lichter, 2000; Y. Qian, 2017; Van Bavel, Schwartz, & Esteve, 2018). Although women’s education is a common indicator of personal attractiveness, reflecting both cultural capital and earnings, we also supplement these analyses by considering men’s and women’s personal income. For each annual file of the ACS (2008 to 2017), each older woman’s annual income in the previous year (adjusted by income in 2016) is measured by its location in the income percentile distribution of all women aged 55 or older.¹⁴ Our approach eliminates the effects of year-to-year changes in the economy or wages, instead comparing each woman’s annual income each year to other older women in the income distribution (see Chetty et al., 2020, for additional justification).¹⁵ The same gender-specific measurement approach is used for women’s male partners. Women’s and men’s percentile personal income is separated into quartiles, then cross-tabulated, with entries in the diagonal cell indicating homogamous couples and entries in off-diagonal cells indicating heterogamous married and cohabiting couples.¹⁶

4.3. Empirical approach

Our analyses begin by first illustrating the diversity of living arrangements among older women, then identifying characteristics of the men with whom they co-reside, i.e., in intact first marriages, remarriages, and cohabiting unions. We expect that older women are far more likely to be demographically mismatched in recent remarriages and cohabiting unions, experiencing a much higher likelihood as cohabiters than married persons to live with older men, men of less education, and different racial backgrounds. These baseline analyses are then supplemented with multivariate models (i.e., multinomial or binominal logistic regression models) that separate the sources of observed differences into supply-side (e.g., characteristics of women with different marital histories) and demand-side considerations (e.g., imbalances in the size and composition of women and men at older ages).

5. Results

5.1. Heterogeneity in marriage and cohabitation

Our first objective is to document gender differentials in union status by age group among women and men aged 55 and older. The baseline hypothesis is that older women, especially minorities, may be disadvantaged in the marriage market and that gender disadvantages are exacerbated with age, albeit unevenly.

5.1.1. Gender differences in marriage and cohabitation

How prevalent is cohabitation among older women and men? Table 1 identifies sex differentials in the percentage shares of female (top panel) and male cohabiters (bottom panel) among (1) all unmarried persons (distinguished by unmarried status), (2) all persons in co-residential unions, and (3) all older persons. The older men and women in our sample typically cohabit with each other rather than with younger partners. So, it is unsurprising that the numbers of older women and men (55+) living with unmarried partners are in demographic alignment (93,376 women versus 96,868 men). The small numerical differences result from men in this age range (55+) being more likely than older women to cohabit with younger partners (i.e., younger than age 55).

Table 1.

Percent Cohabiting by Marital Status, Individuals Aged 55 and Older, by Age and Sex

	Number of People				Percent Cohabiting
Age	Cohabiting	Unmarried	Married	Total	Among unmarried	Among divorced or separated	Among widowed	Among never married	Among unions	Among older persons
Women
55–64	60,062	874,896	1,389,518	2,264,414	6.3	7.4	4.4	5.0	4.4	2.7
65–74	25,003	708,598	872,830	1,581,428	3.2	4.6	2.2	2.3	2.9	1.5
75+	8,311	998,095	399,101	1,397,196	0.8	1.9	0.6	0.6	2.1	0.6
Total	93,376	2,581,589	2,661,449	5,243,038	3.4	5.8	1.6	3.6	3.6	1.8
Men
55–64	53,881	628,337	1,277,893	1,906,230	8.0	9.7	5.7	5.3	4.3	2.9
65–74	29,960	364,506	989,424	1,353,930	7.7	10.2	5.0	4.2	3.0	2.2
75+	13,027	316,347	597,716	914,063	3.9	7.7	2.6	2.4	2.2	1.4
Total	96,868	1,309,190	2,865,033	4,174,223	7.0	9.6	3.7	4.7	3.4	2.5

Note. Numbers are unweighted and percentages are weighted.

These gender differentials nevertheless diverge as women age. The data in Table 1 show that the number of female cohabiters exceeds that of male cohabiters at ages 55–64, but is roughly two-fifths lower among the oldest age group (8,311 vs. 13,027). The percentage of cohabiting women at ages 55+ is much lower than their male counterparts, if considered as a share of all unmarried women. Only 3.4 percent of all older unmarried women are cohabiting, a share about one-half the size observed among all older unmarried men (7.0 percent). These sex differences reflect, in large part, the much larger percentages of older women than men who are unmarried (whether cohabiting or not). Our sample of unmarried older persons contains 2.6 million women, but only 1.3 million men.

Older women also “cohabit up” in age (a topic to which we shall return). For example, at the oldest ages, 75 and older, unmarried women outnumber unmarried men by over 3-to-1 (998,095 to 316,347). The sex differential in cohabitation widens by age. The gender differential in shares of cohabiters among unmarried women declines from 6.3 percent at ages 55–64 to only .8 percent at ages 75 and older. For older men the decline is from 8.0 to 3.9 percent. The percent cohabiting was 27 percent higher at ages 55–64 and about 3 times higher at ages 75 and older among unmarried men than among unmarried women.

The pathways to cohabitation from being divorced or separated, widowed, or never married among older women exhibit substantial heterogeneity. A large majority (61%) of all older cohabiting women are divorced or separated rather than widowed (21%) or never-married (18%) (data not shown). Divorced/separated and never-married cohabiters are over-represented among older women aged 55–64 and underrepresented among the oldest-old. For women aged 75 and older, over one-half (53%) are widowed. From a demographic perspective—one emphasizing shortages of potential male partners—younger, divorced women may be at once “attractive” partners to older—perhaps much older—men and also have greater economic incentives than never-married or widowed women to re-partner.¹⁷

These data also reveal widely uneven rates of cohabitation by age and previous marital status (Table 1). Older unmarried women have lower shares of cohabitation than their male counterparts (5.8 versus 9.6 percent for the separated/divorced, 1.6 versus 3.7 percent for the widowed, and 3.6 versus 4.7 percent for the never married). Perhaps reflecting economic need, shares of cohabitation are highest among both separated or divorced men and women, especially women before retirement ages when Social Security or pension income cannot be accessed (7.4 percent). Cohabitation among women declines sharply by age. Comparing women aged 55–64 to those aged 75 and older, cohabitation declines from 7.4 to 1.9 among the separated or divorced, from 4.4 to 0.6 percent among the widowed, and from 5.0 to 0.6 percent among the never married. Overall, much lower percentages of women than men are in cohabiting unions.

Widows/widowers have especially low shares of cohabitation. Yet, gender differentials persist. Only 1.6 percent of widows is cohabiting, compared with 3.7 percent of widowers. The demographic paradox is that over one-half—52.7 percent—of unmarried cohabiting women aged 75 and above but only 39.2 percent of the male counterparts are drawn from the widowed population (data not shown). This sex differential is consistent with the hypothesis that widows may be advantaged relative to their divorced and never-married counterparts in financial resources (see Lin, Brown, & Hammersmith, 2017), having access to their late husband’s pension and Social Security income. Without an economic imperative, there may be little incentive to seek a co-residential relationship and if they do, they may be better able to secure “attractive partners” while eschewing “unattractive” ones.¹⁸

5.1.2. Racial and ethnic differences

Racial gaps in marriage are on full display in Table 2. The ratio of unmarried to married white women was .53 at ages 55–64, .70 at ages 65–74, and 2.29 at ages 75 and higher, a pattern reflective of high rates of marriage throughout the life course, followed by high rates of widowhood at older ages. For African American women, these ratios are higher still: 1.84, 2.24, and 5.89, respectively, for these three age groups. The large majority of older African American women, even at ages 55–64, are unmarried, a result of comparatively low marriage rates during the early stage of the life course, followed by high divorce and low remarriage rates (Z. Qian, 2014). Shortages of black men also play a role; unmarried African American women outnumbered their male counterparts by 61 percent at ages 55–64, compared with by only 30 percent among whites of the same ages.

Table 2.

Percent Cohabiting by Marital Status, Individuals Aged 55 and Older, by Age, Sex, and Racial/Ethnic Group

	Number of People				Percent Cohabiting
Age Group	Cohabiting	Unmarried	Married	Total	Among unmarried	Among divorced or separated	Among widowed	Among never married	Among unions	Among older Persons
Non-Hispanic White
Women
55–64	45,544	596,404	1,129,354	1,725,758	7.1	8.5	5.0	4.8	4.1	2.7
65–74	20,506	516,047	740,666	1,256,713	3.7	5.4	2.5	2.2	2.8	1.6
75+	7,146	799,696	349,832	1,149,528	0.8	2.2	0.6	0.6	2.0	0.6
N	73,196	1,912,147	2,219,852	4,131,999	3.7	6.6	1.6	3.3	3.3	1.8
Men
55–64	40,882	457,765	1,043,715	1,501,480	8.3	10.4	6.0	4.6	3.9	2.7
65–74	24,106	277,479	834,618	1,112,097	8.2	11.1	5.5	3.7	2.9	2.2
75+	11,020	255,317	518,840	774,157	4.1	8.6	2.8	2.0	2.1	1.4
N	76,008	990,561	2,397,173	3,387,734	7.3	10.4	4.0	4.1	3.2	2.3
Black
Women
55–64	5,894	147,276	80,085	227,361	3.8	3.8	3.0	4.4	7.3	2.6
65–74	1,837	98,813	44,125	142,938	1.7	2.1	1.2	2.2	4.1	1.2
75+	454	96,644	16,398	113,042	0.4	0.8	0.3	0.5	2.7	0.4
N	8,185	342,733	140,608	483,341	2.4	2.9	1.2	3.5	5.9	1.7
Men
55–64	6,140	91,402	76,746	168,148	6.5	7.3	4.8	5.9	7.8	3.7
65–74	2,855	45,729	55,176	100,905	5.9	7.1	3.8	5.2	5.2	2.8
75+	935	29,450	28,200	57,650	3.0	5.0	1.6	3.3	3.4	1.6
N	9,930	166,581	160,122	326,703	5.8	7.0	3.1	5.6	6.3	3.1
Hispanic
Women
55–64	5,457	80,516	97,280	177,796	6.0	6.2	3.7	7.4	5.8	3.0
65–74	1,627	56,069	47,781	103,850	2.5	3.3	1.5	3.1	3.5	1.5
75+	464	60,557	19,011	79,568	0.6	1.3	0.4	1.1	2.5	0.5
N	7,548	197,142	164,072	361,214	3.5	4.8	1.4	5.3	4.8	2.1
Men
55–64	4,663	49,934	87,767	137,701	8.5	9.0	5.4	8.2	5.3	3.4
65–74	2,013	25,658	54,611	80,269	6.9	8.4	3.5	6.9	3.7	2.5
75+	735	19,365	28,984	48,349	3.3	5.5	1.7	4.7	2.5	1.4
N	7,411	94,957	171,362	266,319	7.1	8.5	3.0	7.6	4.4	2.8

Note. Numbers are unweighted and percentages are weighted.

The large racial gap at older ages is also revealed in cohabitation patterns. The percent cohabiting among unmarried African American women, aged 55–64, is only 3.8 percent, compared with 7.1 percent among unmarried white women. At each age, the percentage cohabiting also is lower among divorced or separated and widowed African Americans than among their white counterparts (2.9 and 1.2 percent versus 6.6 and 1.6 percent, respectively). Yet, the share of cohabiting unions, measured by the percentage of all co-residential unions, is substantially higher among African American women than among white women (5.9 percent and 3.3 percent respectively), indicative of lower marriage rates among African American women.

Hispanic patterns of marriage more closely resembled white than black patterns. Sex ratio imbalances rather than economic hardship seemingly underlie black-Hispanic differences in marriage and cohabitation at older ages. Other studies have also acknowledged familism as an important cultural dimension of Hispanic family life (Landale & Oropesa, 2007), which is expressed in high marriage rates (including early marriage), low divorce rates, and strong intergenerational ties, including child-parent relationships and kin-care at older ages.

5.1.3. Marriage and remarriage at older ages

Table 3 documents large differences between older women and men in cohabitation and marriage, both before and after age 55. The large majority of men, but a minority of women, are partnered (in marriage or cohabitation). For older men, 45.9 percent are still in first marriages plus another 20.2 percent are in remarriages, but for older women, less than half are married (i.e., 48.1 percent). These gender differentials in marriage widen with age. At ages 55–64, the marital status distributions among women and men generally are similar and the large majority of pre-retirement men and women are in unions. The index of dissimilarity indicates that only 2.5 percent of women would need to change marital status to achieve the same marital status distribution of men (data not shown). At ages 75 and above, however, the index of dissimilarity increases 13-fold, to 37.5; 72.5 percent of women are single, compared with only 34.9 percent of their male counterparts. First marriage is the modal category for the oldest men (46.0), plus 10.5 and 7.2 percent remarried before and after age 55, compared with 21.1, 3.9, and 2.0 percent, respectively, among the oldest women. These results, once again, provide evidence of large gender differences in marriage.

Table 3.

Union Status by Age Group and Sex, Individuals Aged 55 and Older, 2008–2017

	Age Group
Union Status	55–64	65–74	75+	All 55+
Women
Single (never married, divorced, or widowed)	39.7	46.5	72.5	50.2
Cohabiting	2.7	1.5	0.6	1.8
First Married	39.0	37.0	21.1	33.8
Remarried before age 55	17.3	12.1	3.9	12.3
Remarried at age 55 or later	1.4	2.8	2.0	2.0
N	2,264,414	1,581,428	1,397,196	5,243,038
Men
Single (never married, divorced, or widowed)	33.1	26.8	34.9	31.5
Cohabiting	2.9	2.2	1.4	2.4
First Married	44.9	47.5	46.0	45.9
Remarried before age 55	17.4	18.2	10.5	16.2
Remarried at age 55 or later	1.8	5.2	7.2	4.0
N	1,906,230	1,353,930	914,063	4,174,223

Note. Numbers are unweighted and percentages are weighted. Separated people are included among the divorced.

Our analyses again reveal large racial disparities in union formation. Data in Appendix Table 1 provide evidence of exceptionally high rates of singlehood among older African American women. About 71 percent are single, a figure almost 25 percentage points higher than their white counterparts. Not surprisingly, the percentage of first-marriages or remarriages is much lower among African American than among white women. Only in the case of cohabitation are comparable figures observed among whites, African Americans, and Hispanics.

5.2. Mate selection and assortative mating at older ages

Our second objective examines, for the first time, patterns of assortative mating among older women. We consider three traits commonly considered in the demographic literature on intermarriage: age, race, and socioeconomic status, including education and personal income (Schwartz, 2013). Our expectation is that older women are less likely to receive an equitable exchange, which will be observed in increasing hypogamy as women age.

5.2.1. Homogamy and heterogamy

Table 4 provides baseline estimates of age, race, educational, and income homogamy and heterogamy by women’s union status. The results confirm our general hypothesis of substantial heterogamy among older couples, especially cohabiting couples. For example, only a minority share (41.4 percent) of cohabiting women are in age-homogamous unions (within 3 years apart). In fact, cohabiting unions are least likely to be age homogamous. In contrast, among women in first marriages, 67.7 percent are in age-homogamous relationships but, among remarriages formed after age 55, only 45.5 percent are in such relationships. The implication is that later-life unions, especially cohabitations, draw more age-heterogamous unions. Previous research on younger women has shown that age-heterogamous unions are higher in cohabiting unions and remarriages than in first marriages (Blackwell & Lichter, 2000). This also is the case for older women.

Table 4.

Age, Race, and Educational Pairing by Union Status, Women Aged Ages 55 and Older, 2008–2017

		Women’s Union Status
		Cohabiting	First married	Remarried before age 55	Remarried at age 55 or later	All unions
Age pairing	Couples within 3 years apart	41.4	67.7	46.8	45.5	60.7
	Women at least 4 years younger	29.4	28.1	33.7	31.5	29.6
	Women at least 4 years older	29.1	4.3	19.5	23.1	9.7
Racial pairing	Same race	90.9	96.2	93.7	93.8	95.3
	White women nonwhite men	4.2	1.4	2.6	2.5	1.9
	Nonwhite women white men	4.0	2.1	3.4	3.3	2.5
	Nonwhite women nonwhite men	0.8	0.3	0.4	0.4	0.3
Education paring	Both less than high school	7.0	6.4	3.8	4.8	5.7
	Both high school	15.8	15.4	14.6	13.6	15.1
	Both some college	10.9	10.4	13.4	11.4	11.2
	Both college or more	10.7	21.5	14.0	14.4	19.0
	Women less educated than men	26.1	26.8	30.4	30.9	27.8
	Women more educated than men	29.6	19.6	23.8	24.9	21.2
Income paring	Both in bottom quartile	8.8	8.2	6.4	6.6	7.7
	Both in middle two quartiles	12.9	12.3	13.0	16.6	12.7
	Both in top quartile	8.9	10.0	10.4	9.3	10.0
	Women in lower quartile than men	21.1	41.8	38.7	29.8	39.8
	Women in high quartile than men	48.3	27.7	31.6	37.7	29.8
	Gender gap in total income (average)	8,076	35,766	29,847	21,072	32,779
	Gender gap in total income (median)	3,241	19,000	17,074	11,200	17,541
Total		93,376	1,869,462	683,068	108,919	2,754,825

Note. Numbers are unweighted and percentages are weighted.

The second panel (Table 4) reveals the distribution of racially endogamous and exogamous unions. The results are consistent with our hypothesis: Cohabiting women are least likely to be in racially homogamous unions (90.9 percent), while first marriages are the most homogamous (96.2 percent). Remarriages again occupy an intermediate position (about 94 percent). These patterns could reflect increases over time in interracial unions; compared to recently formed cohabitations and remarriages, most first marriages began many years if not decades ago. They imply that older unmarried persons cast a much wider net in search of acceptable partners (Qian and Lichter, 2018).

Indeed, as shown in Appendix Table 2, white women are more likely to form cohabiting unions with nonwhite men than are their counterparts in marriages or remarriages (5.5 percent among cohabiting couples, 1.8 percent among first married couples, and around 3.0 percent among remarried couples). Compared with first marriages, nonwhite women also are generally more likely to be in cohabiting unions and remarriages with white men. This may reflect large marriage market squeezes among older Hispanic and black women.

The data on educational homogamy and heterogamy reveal, overall, that educational homogamy is the statistical (if not cultural) norm, and that deviations from this norm are usually expressed in educational hypergamy (women experience upward mobility). As shown in the third panel (Table 4), more than one-half of all unions are educationally homogamous and about one-quarter of unions involve women with lower levels of education than their partners. (Here we have defined upward and downward education mobility using the 4 educational categories.) Still, recently remarried women are more likely to marry a man with more rather than less education than themselves (30.9 versus 24.9 percent). The opposite patterns are observed among cohabiting women; 29.6 percent have more education than their partners while 26.1 percent have less. This may reflect the overall lower levels of education among older cohabiting women and men (e.g., higher percentages of homogamous couples with a high school education or less, and lower percentages of college educated couples).

Finally, the bottom panel (Table 4) provides evidence or not of income homogamy. We focus on the personal income of each partner. Overall, only 7.7 percent of older women and their male partners are similarly located in the bottom quartile of the income distribution.¹⁹ About 30 percent of all older women co-reside with similarly situated men in the same sex-specific income quartile (i.e., this is the sum of 7.7, 12.7, and 10.0). Instead, these data nevertheless reveal the familiar pattern of upward income mobility: 39.8 percent of these women are located in a lower income quartile than their partners, compared with 29.8 percent in a higher income quartile.

Variation in patterns of income homogamy and heterogamy across categories of union status are large. Indeed, cohabiting women and their male partners are slightly more likely than married or remarried partners to be located in lowest income quartiles. They are instead likely to be slightly underrepresented in the top income quartile. For sample, 8.9 percent of cohabiting women are located in the top quartile as their partners. This contrasts with 10.0 percent among women in first marriages and 10.4 percent among women who remarried before age 55. The adverse marriage market conditions are also revealed in patterns of income hypogamy. Older cohabiting women, especially in comparison to women in first marriage or in remarriages occurring below age 55, are considerably more likely to be living with partners with incomes lower than their own than with partners in higher income quartiles (48.3% vs. 21.1%). This pattern of greater income hypogamy than hypergamy is also revealed among older women who remarried at age 55 or older. These analyses highlight older women’s relative disadvantage in the marriage market (i.e., women’s returns to their relative income). It is important to note, however, that the gender income gap (measured using mean and median differences) is positive (i.e., men earn more), regardless of union status (bottom two rows of Table 4). Still, the gender income gap is much lower among cohabiting women ($8,076) and those women who remarried at older ages ($21,072) than the overall average ($32,779).

5.2.2. Multivariate analyses of union formation and mate selection

We begin in this section by asking whether cohabiting women are more likely than currently married women to have “economically attractive” partners, who we initially define as men with a college degree or more. As an empirical baseline, Figure 1 presents log odds ratios of unions in which both partners have a college education or both have less than college education relative to unions in which one partner has a college education and the other partner has less than college education. By definition, a log odds ratio of 0, when exponentiated, is equal to an odds ratio of 1.0 (i.e., e⁰ = 1); this means that the odds of educational homogamy are equal to the odds of educational heterogamy. The results in Figure 1 clearly indicate that the odds ratio of educational homogamous relative to heterogamous marriages is very high among first married couples. The log odds ratio, when exponentiated, is 10.9 (or e^2.391). That is, highly educated women in first marriages are significantly more likely than their less educated counterparts to have college educated partners. This is clearly consistent with previous studies of young adults which indicate substantial (and increasing) educational homogamy among highly educated women (Schwartz & Mare, 2005).

Figure 1.

Log Odds Ratio of Unions among Couples with Same Versus Different Levels of Education (College or more versus some college or less) by Union Status and Race/Ethncity among Couples in which the Female Partner Is Aged 55 or More

The fact that log odds ratios in Figure 1 are well above 0 for women in other kinds of unions similarly indicates that highly educated women tend to partner with highly educated men, just not to the extent characterizing women in first marriage. Moreover, any observed differences in odd ratios among other kinds of unions (i.e., cohabiting and remarried women) are much smaller. This is especially the case of differences between cohabitation and remarriages before age 55, which are statistically insignificant for all women and for women in each of the ethnoracial groups presented here.

Log odds ratios (or odds ratios) provide a useful descriptive tool but may be misleading if age, race, and education are interrelated statistically. For example, older women are likely to be less highly educated than younger women. Racial minorities, especially African American women, may be overrepresented if their marital biographies are marked by lower rates of first marriage and remarriage. The descriptive results in Figure 1 summarize educational assortative mating among all women currently in marital and cohabiting unions, but ignore variation in recently formed unions. To address this question, we first present multinomial logistic regression model predicting whether older women; (1) married in the past year, (2) were cohabiting at the time of the survey, or (3) were unpartnered at the survey (the reference category). We then present the logistic regression models predicting whether older women formed unions with men with college degrees. Our independent variables are women’s age, race, and education among others.

In Table 5, the multinomial model shows, as expected, that transitions to marriage or cohabitation vis-à-vis remaining single decline steadily with age. These data also show that getting married after age 55 is less common than cohabitation. Compared to women aged 55–64, the relative risk ratio of marriage is 61 percent lower for women aged 65–74 and 91 percent lower for women aged 75 or older. For cohabitation, the relative risk ratios are 53 and 89 percent lower, respectively, for the two age groups. White women are more likely than any other group (including Asian women) to marry or cohabit compared to Hispanics (9 percent or 7 percent higher, respectively). Black women are least likely to transition to either marriage or cohabitation (with ratios 35 percent and 43 percent lower than those of Hispanics, respectively). College-educated women are most likely to marry. This is clearly consistent with the idea that a college education—and the financial independence it affords them—incentivizes entry into marriage. Finally, older never-married women are least likely to form marital or cohabiting unions while previously remarried women are most likely to enter into another remarriage or cohabiting union.

Table 5.

Relative Risk Ratios Predicting the Likelihood of Marriage or Cohabitation Relative to Remaining Single and Odds Ratios of Forming Unions with Male Partners with Completed College, Women Aged 55 and Older, 2008–2017

	Model 1		Model 2
	Marriage vs. Single	Cohabitation vs. Single	Male Partner with Completed College
Age Group
55–64
65–74	0.39***	0.47***	1.20***
75+	0.09***	0.11***	1.30***
Race/ethnicity
Non-Hispanic White	1.09*	1.07***	1.78***
Non-Hispanic Black	0.65***	0.57***	0.65***
Asian American	1.10	0.81***	1.82***
Native American	0.70***	1.17***	0.78*
Hispanic
Education
Less than high school
High school	1.39***	1.15***	2.43***
Some college	1.58***	1.06***	4.81***
College or more	1.69***	0.98	17.65***
Year unions formed
2008–2012
2013–2017	1.06**	1.24***	0.96*
Type of residence
City	0.60***	0.83***	2.07***
Suburban	0.86***	0.97***	1.68***
Non-metropolitan area
Nativity
Native-born
Immigrant	1.10**	0.92***	1.51***
Previous marital history
Never married
Married once	1.09***	1.27***	1.13***
Remarried	1.51***	1.53***	1.02
Union type
Cohabitation	-
Marriage	-		1.39***
Constant	0.01***	0.05***	0.19***
LR Chi2	70,139		16,960
BIC	876,289		94,902
Observations	2,564,135		104,207

^***p<0.001

^**p<0.01

^*p<0.05

Older women also are increasingly likely over time to form unions, especially cohabiting unions. The ratios of marriage and cohabitation are, respectively, 6 percent and 24 percent higher in 2013–2017 than in 2008–2012. Older immigrant women also are more likely to marry but less likely to cohabit than their U.S.-born counterparts, reflective of cultural differences in attitudes and behavior toward marriage and cohabitation. Where older women live also matters. Older women living in non-metropolitan areas are more likely to form either marital or cohabiting union than those living in cities or suburban areas.

Our logistic regression model (Model 2, Table 5) shows that married women’s partners are significantly more likely than cohabiting women’s partners to be college graduates (OR = 1.39). The likelihood of living with a college-educated male also increases with women’s age, which is consistent with observed educational gradients in mortality among men (Montez et al., 2009); that is, older partnered men are more likely to be college-educated than their younger counterparts. As expected, historically disadvantaged black and American Indian women are least likely to be partnered with college-educated men, with odds being 35 percent and 22 percent lower, respectively, than those for Hispanics. White and Asian American women are most likely to have college-educated partners. Hispanics (the reference category) occupy an intermediate position.

Educational homogamy is the norm. Older college-educated women are significantly more likely to be partnered with college-educated men, with an odds ratio of 17.65. This finding cannot be explained by age and racial differentials in education, which are controlled in the model. The partners of older women living in cities or suburbs are most likely to have completed college. Older immigrant women, interestingly, are more likely to form unions with college-educated male partners, possibly due to strong educational differences in union formation among immigrants. Finally, older women who married only once have higher odds of union formation with college-educated male partners than never-married women or women who married more than once. The differences are consistent with previous studies reporting that older women often “upgrade” their partners as they move from one marriage to the next (Graefe & Lichter, 2007).

Finally, as a robustness test, we also consider patterns of income hypergamy among older women. The descriptive results, reported in Table 4, indicated that women’s returns to cohabitation, if measured by their partner’s location in income distribution of all older men, is much lower than their counterparts still in first marriages. Whether this is a matter of selection into first marriage or divorce is unclear. The multivariate logistic regression models presented in Table 6 revisits this general conclusion, considering where income “returns” to women’s recent union formation. Specifically, we focus on whether older women are newly partnered with men who are located in the highest quartile of the income distribution.

Table 6.

Odds Ratios Predicting the Likelihood of Forming Unions with Male Partners of Top Quartile Income, Women Aged 55 and Older, 2008–2017

	Model 1	Model2	Model3	Model 4
Age Group
55–64
65–74	0.67***	0.72***	0.71***	0.89***
75+	0.56***	0.59***	0.61***	0.78***
Race/ethnicity
Non-Hispanic White	1.39***	1.66***	1.35***	1.35***
Non-Hispanic Black	0.60***	0.67***	0.61***	0.64***
Asian American	1.54***	1.88***	1.52***	1.51***
Native American	0.66***	0.76*	0.68***	0.68***
Hispanic
Education
Less than high school
High school	1.93***		1.77***	1.77***
Some college	2.72***		2.26***	2.34***
College or more	5.19***		3.72***	3.95***
Year unions formed
2008–2012
2013–2017	0.99	1.02	1.00	1.01
Type of residence
City	2.04***	2.05***	1.89***	1.93***
Suburban	1.88***	1.85***	1.76***	1.81***
Non-metropolitan area
Nativity
Native-born
Immigrant	1.05	1.06	1.08*	1.04
Previous marital history
Never married
Married once	1.27***	1.20***	1.25***	1.29***
Remarried	1.17***	1.09***	1.17***	1.22***
Union type
Cohabitation
Marriage	1.61***	1.69***	1.62***	1.62***
Total income percentile
Bottom quartile
Second quartile		0.84***	0.82***
Third quartile		1.13***	1.01
Top quartile		2.42***	1.75***
Absolute Income
Earning				1.03***
Social security				0.96***
Retirement				1.01***
Investment				1.06***
Constant	0.03***	0.06***	0.04***	0.03***
LR Chi2	6,186	6,186	7,910	7,814
BIC	93,051	93,051	91,361	91,468
Observations	104,207	104,207	104,207	104,207

^***p<0.001

^**p<0.01

^*p<0.05

Model 1 (Table 6) replicates the model 2 in Table 5, but replaces the dependent variable with men’s income (in the highest quartile). Highly educated older women are far more likely than less educated women to form unions with economically-attractive older men. Nevertheless, women’s education gradient is much flatter with respect to attracting an economically attractive partner based on her partner’s personal income (OR = 5.19) than education (OR = 17.65; Table 5). Model 2 replaces women’s education with her income (measured by her location in women’s income distribution). The income gradient indicates that older women in the upper half of the income distribution (but especially in the top quartile) are most likely to form unions with high-income men, which is consistent with the education results in Tables 4 and 5. Observed education and income gradients in forming coresidential unions with affluent men are reduced but not eliminated when both variables are included in model 3.

The results from models 1–3 also clearly indicate that returns to marriage (in the form of marrying high-income partners) is much greater than returns to cohabitation. The odds ratio for getting married in the past year is about 1.60, which suggests a significant marriage premium. That is, getting married (vis-à-vis cohabitation) leads to a 60 percent greater likelihood of partnering with a man in the top income quartile. Cohabitation among older women is much less likely to be pathway to upward socioeconomic mobility. The results also show that non-Hispanic whites and Asians are also far more likely to form unions with men located in the top quartile, suggesting that the economic incentives to marriage or cohabitation are much lower among other older women of color, who face shortages of “economically attractive” potential partners. Economic incentives to forming unions also decline with age, and are lower among nonmetropolitan women than city and suburban women. Indeed, women of working age (i.e., ages 55–64) are more likely to marry or cohabit with men in the top income quartile than are retirement age women. And, not surprisingly, languishing income and earnings in rural America is undoubtedly responsible for the lower likelihood that rural women will form unions with rich men.²⁰

Finally, as a coda to these analyses, model 4 models the effects of specific sources of women’s income: Earnings from employment, Social Security income, retirement income (pensions), and income from personal investments (i.e., income from an estate or trust, and interest, dividends, royalties and rents). These analyses address the question of whether specific kinds of income deter or disincentivize marriage or cohabitation or, instead, provide women with the economic resources required to attract partners. Here, absolute dollar values, for both women and men, are logged, yielding a straightforward interpretation: A percent change in women’s income (for a particular source) is associated with a percent change in the likelihood of forming unions with men at the top income distribution. Interestingly, earnings, retirement income, and investment incomes are each positively associated with forming unions with the most affluent men. To illustrate, the likelihood of union formation with high-income males increases by about 6 percent with each 1 percent increase in women’s annual investment income. Clearly, more affluent (and apparently wealthy) women attract more affluent men. In contrast, Social Security income is negatively associated with forming unions with high income older men.

6.1. Discussion and conclusions

This paper provides an up-to-date demographic portrait of union formation and mate selection at older ages. We use nationally-representative data from the ACS to highlight exceptional heterogeneity in the living arrangement of contemporary older women. We also document their comparative disadvantages in the marriage market, which are expressed in large gender disparities in patterns of assortative mating (including hypogamy). Our basic premise is that the motivations and constraints on cohabitation or marriage are both different from those of younger women (Sassler & Lichter, 2020) and change considerably as women become older. Our analyses also acknowledge that personal and financial options provide older women with more options—more decision-making power—to realize their aspirations for cohabitation or marriage.

Our study provides, first and foremost, updated national estimates of past-decade patterns of union formation at older ages. We showed that cohabitation, perhaps surprisingly, is now less prevalent among older women than men, and that this gender gap increases with age. Our study, like previous studies, show that older women are much less likely to be partnered than older men (Brown et al., 2016). Cohabitation represents a much smaller share of all unmarried women, all partnered women, and all women in comparison to patterns observed among their male counterparts. As such, our results are generally consistent with demographic explanations that emphasize shortages of men at older ages, especially among the oldest old. Less than 5 percent of older single women, aged 55 and older, are cohabiting, a figure one-half as large as their male counterparts. If partnered, they are more likely than men to be cohabiting and less likely to be married. They also are more likely to be in hypogamous cohabiting and marital unions.

A second contribution of our study is its emphasis on heterogeneity—older women’s uneven experiences in the marriage market. Indeed, the likelihood entering a cohabiting union or remarriage varies widely across different sub-populations of older women, including those with different racial backgrounds and marital histories. The marriage gap among African American women is especially large at older ages in comparison to their white counterparts. Most currently unmarried older women were previously married but now divorced or separated, but these patterns change rapidly with age as married women experience the death of a spouse. Widows are far less likely to cohabit than their divorced counterparts. Moreover, widows as compared with widowers are much less likely to form cohabiting unions. At every age, divorced women are less likely than divorced men to cohabit or marry. These results provide a caveat to previous studies: Low and uneven rates of cohabitation and marriage among women are not simply a function of demographic constraints associated with sex ratio imbalances at older ages.

Finally, our analyses build on previous studies by documenting recent patterns of assortative mating, a topic that has received little if any previous empirical attention. For older women, the disadvantages in the market for partners, at least as reflected in gender differentials in singlehood, cohabitation, and marriage, suggest that older women have more difficulty forming homogamous or demographically well-matched unions with men. If older women are poorly matched with their husbands/partners, this raises new questions for future research about any implications for older women’s emotional and financial well-being. The fact that older cohabiting women also are more likely to be living with poorly-educated (i.e., a high school degree or less) or low income men also raises obvious questions about women economic security.²¹ Indeed, we found that “returns” to older women’s valued traits, such as education, were lower among cohabiting than among recently married older women, and especially among older women still in first marriages. Hypogamy at older ages is commonplace.

Our results are largely consistent with theoretical explanations that emphasize demographic constraints on partnering at older ages. But demography is not destiny. Indeed, aggregate or “on average” estimates mask substantial heterogeneity among older women. Personal agency and diverse motivations also shape decision-making in the marriage market. This is plainly indicated by widely disparate patterns of marriage and cohabitation at older ages (Carr & Utz, 2020; Halpern-Manner, et al., 2015; King & Scott, 2005). Some older women are also much less able than others to form homogamous unions. This matters—at least potentially—because previous studies at younger ages often indicate that homogamy is associated with union stability and life satisfaction (Schwartz, 2013). Whether this is true at older ages is unclear. We do not know whether the counterfactual—if older cohabiting women had married instead or remained single—would be better off in some objective (e.g., poverty) or subjective (e.g., happiness) sense. If preferences are revealed in behavior (i.e., revealed preferences), then older women who entered and stayed in cohabiting relationships presumably did so because they expected a better or more satisfying life. Whether cohabitation represents a salutary alternative to marriage depends, however, on whether the decision is hers alone or reflects a joint decision-making process with her partner. Additional theory-driven research is needed, especially qualitative studies that emphasize interpersonal processes (with families, friends, and partners), marital aspirations, and idiosyncratic conditions that select older women into different kinds of co-residential unions.

Our study of union formation and mate selection at older ages is not without some limitations. Our analyses are based on a repeated cross-sectional study design (i.e., the ACS), which means that we are unable to fully monitor the timing, duration, and sequencing of cohabiting and marital unions. What we have gained in sample size and sampling design (i.e., household record) are offset in part by our inability to track the actual processes of union formation. We know from samples of young adults that declining shares of cohabiting couples subsequently marry and that increasing shares dissolve their relationships (Lichter et al., 2016). Serial cohabitation is on the rise, but this seems less common among older cohabiting women; in fact, committed cohabiting relationships often end in the death of partners rather than in marriage or dissolution (Brown et al., 2016). Our prevalence measures also undoubtedly underestimate the incidence of cohabitation at older ages or other kinds of intimate relationship (“living apart together”). Despite its many advantages, the ACS is ill-suited for analyses of instability among cohabiting relationships or determining whether cohabiting relationships represent an alternative to marriage. Studies based on younger couples reveal substantial union homogamy, especially in marriages, and that couples typically experience an upgrade in the “quality” of partners with each successive co-residential relationship (Graefe & Lichter, 2007; Wright, 2017). We do not know whether these conclusions apply to older women, who face very different demographic opportunities, third-party constraints, and preferences. For older persons, marriage markets are not likely to be national in scope. Opportunities to form well-matched unions are instead likely to vary across local areas or institutional settings (e.g., retirement village versus long-term care).

The aging of the baby boom generation makes this a propitious time to study mate selection patterns among older persons. The baby boomers were in the vanguard of the divorce revolution, of feminism and growing gender equality (e.g., the gender revolution), and the transformation of traditional marriage (i.e., husband as provider, wife as homemaker) during the 1960s and 1970s. The baby boom peaked in 1957, meaning that this age cohort will reach age 65 in 2022. How older women (and men) live out their later years provides potentially important lessons for healthy aging and wellbeing and for evolving patterns of intimacy and the quality of co-residential unions (Lin & Brown, 2012). America’s baby boomers are, once again, forerunners of change in American marriage and family life, but this time at older ages.

Appendix Table 1.

Union Status by Age Group, Sex, and Race/Ethnicity, Individuals Aged 55 and Older, 2008–2017

	Age Group
Union Status	55–64	65–74	75+	All 55+
Non-Hispanic White
Women
Single (never married, divorced, or widowed)	34.9	42.0	70.3	46.7
Cohabiting	2.7	1.6	0.6	1.8
First Married	41.2	39.6	22.7	35.7
Remarried before age 55	19.8	13.7	4.2	13.7
Remarried at age 55 or later	1.5	3.1	2.2	2.2
N	1,725,758	1,256,713	1,149,528	4,131,999
Men
Single (never married, divorced, or widowed)	30.3	24.4	32.9	29.0
Cohabiting	2.7	2.2	1.4	2.3
First Married	46.0	48.5	47.4	47.1
Remarried before age 55	19.0	19.5	10.9	17.4
Remarried at age 55 or later	1.9	5.5	7.4	4.3
N	1,501,480	1,112,097	774,157	3,387,734
Black
Women
Single (never married, divorced, or widowed)	64.7	70.2	86.3	70.9
Cohabiting	2.6	1.2	0.4	1.7
First Married	22.6	20.0	9.8	19.1
Remarried before age 55	9.1	6.7	2.5	7.1
Remarried at age 55 or later	1.0	1.8	1.1	1.3
N	227,361	142,938	113,042	483,341
Men
Single (never married, divorced, or widowed)	52.9	45.4	52.2	50.6
Cohabiting	3.7	2.8	1.6	3.1
First Married	28.6	31.9	28.6	29.6
Remarried before age 55	13.1	14.7	10.4	13.1
Remarried at age 55 or later	1.8	5.2	7.3	3.7
N	168,148	100,905	57,650	326,703
Hispanic
Women
Single (never married, divorced, or widowed)	47.4	58.3	79.1	57.1
Cohabiting	3.0	1.5	0.5	2.1
First Married	37.4	31.3	16.7	31.3
Remarried before age 55	11.2	7.2	2.7	8.3
Remarried at age 55 or later	1.1	1.8	1.1	1.3
N	177,796	103,850	79,568	361,214
Men
Single (never married, divorced, or widowed)	36.5	33.1	42.4	36.5
Cohabiting	3.4	2.5	1.4	2.8
First Married	46.3	47.5	42.3	46.0
Remarried before age 55	12.3	13.1	8.6	11.9
Remarried at age 55 or later	1.5	3.9	5.3	2.8
N	137,701	80,269	48,349	266,319

Note. Numbers are unweighted and percentages are weighted. Separated people are included among the divorced.

Appendix Table 2.

Age, Race, and Educational Pairing by Union Status and Race/Ethnicity, Women Aged Ages 55 and Older, 2008–2017

		Women’s Union Status
		Cohabiting	First married	Remarried before age 55	Remarried at age 55 or later	All unions 55+
Non-Hispanic White
Age	Couples within 3 years apart	29.9	26.8	33.9	31.7	28.9
	Women at least 4 years younger	42.5	69.9	47.3	46.6	62.2
	Women at least 4 years older	27.7	3.3	18.8	21.7	8.9
Race	Same race	94.5	98.2	97.0	97.1	97.7
	Nonwhite men	5.5	1.8	3.0	2.9	2.3
Education	Both less than high school	3.7	3.5	2.8	3.5	3.3
	Both high school	16.4	16.3	15.2	14.0	15.9
	Both some college	11.5	10.8	13.5	11.5	11.5
	Both college or more	12.3	22.8	14.7	15.2	20.1
	Women less educated than men	26.3	27.4	30.7	31.5	28.4
	Women more educated than men	29.7	19.2	23.2	24.3	20.8
Income pairing	Both in bottom quartile	6.6	6.4	5.7	5.5	6.2
	Both in middle two quartiles	13.5	12.7	13.1	17.1	13.0
	Both in top quartile	10.1	10.6	10.8	9.9	10.6
	Women in lower quartile than men	20.8	43.7	39.4	30.0	41.3
	Women in high quartile than men	49.0	26.6	31.0	37.5	28.9
	N	73,196	1,537,414	588,870	93,568	2,293,048
Black
Age	Couples within 3 years apart	30.3	31.2	34.0	30.0	31.7
	Women at least 4 years younger	39.5	61.6	46.0	42.1	55.7
	Women at least 4 years older	30.2	7.3	20.0	27.9	12.6
Race	Same race	94.0	97.7	95.4	93.8	96.8
	White men	4.4	1.5	3.5	5.0	2.3
	Men of other races	1.7	0.8	1.1	1.2	1.0
Education	Both less than high school	11.3	8.9	6.8	9.9	8.6
	Both high school	18.1	15.6	13.6	13.1	15.2
	Both some college	9.9	11.6	14.3	12.1	12.2
	Both college or more	3.3	12.6	8.3	8.3	10.8
	Women less educated than men	23.7	20.5	22.8	21.8	21.3
	Women more educated than men	33.8	30.8	34.3	34.9	32.0
Income pairing	Both in bottom quartile	17.1	11.6	9.3	12.3	11.4
	Both in middle two quartiles	10.5	12.6	13.9	14.3	12.9
	Both in top quartile	3.5	7.6	7.5	5.6	7.3
	Women in lower quartile than men	18.4	28.1	27.0	21.6	27.0
	Women in high quartile than men	50.6	40.1	42.3	46.3	41.5
	N	8,185	97,488	36,530	6,590	148,793
Hispanic
Age	Couples within 3 years apart	24.7	32.1	30.4	26.7	31.3
	Women at least 4 years younger	36.2	58.0	43.1	37.2	53.5
	Women at least 4 years older	39.1	9.9	26.4	36.1	15.2
Race	Same race	76.0	84.7	63.1	65.4	79.6
	White men	20.4	14.0	34.0	31.9	18.6
	Men of other races	3.6	1.3	2.9	2.7	1.8
Education	Both less than high school	29.0	31.8	16.7	20.3	28.5
	Both high school	10.3	10.7	10.8	9.9	10.7
	Both some college	7.1	7.6	11.1	9.8	8.3
	Both college or more	4.1	7.3	8.0	7.0	7.3
	Women less educated than men	25.7	24.0	30.9	30.1	25.6
	Women more educated than men	23.8	18.6	22.5	23.0	19.7
Income pairing	Both in bottom quartile	16.8	20.5	12.4	15.4	18.6
	Both in middle two quartiles	11.7	10.5	11.3	13.4	10.8
	Both in top quartile	5.0	4.5	7.7	5.1	5.2
	Women in lower quartile than men	25.3	37.9	38.3	31.9	37.2
	Women in high quartile than men	41.2	26.7	30.2	34.2	28.3
	N	7,548	124,984	33,924	5,164	171,620

Note. Numbers are unweighted and percentages are weighted.

Appendix Table 3.

Descriptive Statistics (Percentages/Means), by Single, Marriage in the Past Year, or Current Cohabitation, Women Aged 55 and Older, 2008–2017.

	All	Single	Marriage	Cohabitation
Age Group
55–64	36.5	35.3	69.7	64.2
65–74	27.1	27.2	23.2	26.8
75+	36.4	37.5	7.1	8.9
Race/ethnicity
Non-Hispanic White	72.0	71.8	79.3	79.4
Non-Hispanic Black	14.5	14.6	9.1	8.9
Asian American	3.6	3.6	3.3	2.3
Native American	9.4	9.4	7.6	8.2
Hispanic	0.6	0.6	0.7	1.3
Education
Less than high school	20.6	20.9	10.3	14.0
High school	33.8	33.8	29.8	34.1
Some college	26.1	25.9	33.4	30.7
College or more	19.5	19.4	26.6	21.2
Year unions formed
2008–2012	47.0	47.2	45.6	41.9
2013–2017	53.0	52.8	54.4	58.1
Type of residence
City	11.8	11.8	16.4	15.2
Suburban	15.1	15.2	9.3	10.7
Non-metropolitan area	73.1	73.0	74.3	74.1
Nativity
Native-born	85.2	85.1	88.3	89.5
Immigrant	14.8	14.9	11.7	10.5
Previous marital history
Never married	13.7	13.6	15.0	13.3
Married once	60.8	61.1	48.5	52.5
Remarried	25.6	25.3	36.5	34.3
N	2,564,135	2,488,213	12,130	93,376

Note. Numbers are unweighted and percentages are weighted.