Abstract
This study builds on Becker’s and Oppenheimer’s theories of union formation to examine the economic determinants of marriage and cohabitation during older adulthood. Based on the 1998–2006 Health and Retirement Study and a sample of previously married Americans who are at least 50 years old, results show that wealthier older adults, regardless of gender, are more likely to repartner than stay single. Wealth has no discernable effect on the likelihood of remarrying versus cohabiting. Among the oldest men, the positive associations between wealth and repartnering are entirely due to housing assets. Results suggest that Oppenheimer’s theory of marriage timing may be more applicable to later-life union formation than Becker’s independence hypothesis. Further, economic disadvantage does not appear to characterize later-life cohabitation, unlike cohabitation during young adulthood. These findings help illuminate the union formation process during older adulthood and are timely considering demographic changes reshaping the American population.
Introduction
Numerous studies have examined union formation, the process of entering into marital and cohabiting relationships. To understand this process, social scientists frequently study women’s growing economic resources whether via increased education, employment, or earnings. Indeed, a long line of research has documented the economic determinants of marriage (Becker 1991; Goldscheider and Waite 1986; Lichter et al. 1992; Oppenheimer 1988, 1994, 2003; Oppenheimer and Lew 1995; Sassler and Goldscheider 2004; Xie et al. 2003) and, as it has become more common, cohabitation (Clarkberg 1999; Oppenheimer 2003; Sassler and McNally 2003; Smock et al. 2005).
The literature on union formation largely addresses the experiences of young adults, however. We know little about marriage and especially cohabitation in later life (Cooney and Dunne 2001), despite demographic changes in the population, age, and household structures that make it timely to do so. Since 1900, the number of older Americans (aged 65 and older) has increased 12 times (from 3.1 million to 37.9 million), while their percentage of the population has tripled (from 4.1% to 12.6%). Their numbers are projected to increase an additional 36%, to 55 million, by 2020 (U.S. Census Bureau 2007).
Although it is easy to picture cohabitation—living together in an intimate relationship without being married—as a young person’s “experiment” with nontraditional family forms, cohabitation is not limited to young adulthood. In 2000, 14.9% of the 4 million U.S. cohabitors were at least 55 years old, and 5% were at least 65 (U.S. Census Bureau 2001). In just one decade, those percentages have increased by a third: in 2010, persons 55 and older and 65 and older made up 21.5% and 7.7%, respectively, of the 7.5 million U.S. cohabitors (U. S. Census Bureau 2010). Older persons comprise the fastest growing group of U.S. cohabitors because aging baby boomers, the first U.S. generation to cohabit in large numbers, are swelling the ranks of older adulthood (for historical estimates, see Chevan 1996; and Fitch et al. 2005). At the same time, acceptance of cohabitation is growing among all age groups (Haskey 2001; Thornton and Young-DeMarco 2001).
The goal of this study is to explore the economic determinants of remarriage and cohabitation among previously married, heterosexual Americans aged 50 years and older. How do economic resources such as Social Security, wealth, and housing shape union formation among older men and women? Previous work in this field includes qualitative studies using mainly European samples (de Jong Gierveld 2002, 2004; de Jong Gierveld and Peeters 2003; Stevens 2002), cross-sectional data (Brown et al. 2006; King and Scott 2005), and indirect measures of cohabitation from the 1960–1990 U.S. censuses (Chevan 1996; Hatch 1995). This research uses a nationally representative sample of older adults and is the first longitudinal study to examine remarriage and cohabitation among older Americans.
This study tests competing theoretical frameworks and has implications for how scholars study and understand later-life relationships. It questions whether resource-poor older adults are more likely to cohabit than marry, a pattern found repeatedly in young adulthood. Understanding the kinds of intimate relationships older adults experience is important because relationships are vital to well-being. Single older adults lack the protective monitoring of a partner (Franks et al. 2004; Goldman et al. 1995) and tend to be lonelier, receive less informal care, and have higher risks of institutionalization and mortality than older adults who are partnered (Grundy 2001; Hays 2002; Manzoli et al. 2007; Peters and Liefbroer 1997). Not all later-life relationships are similarly protective of health, though. Compared with older married persons, older cohabitors are more likely to experience institutionalization, bereavement, and death (Koskinen et al. 2007; Moustgaard and Martikainen 2009). Understanding later-life union formation is therefore not only demographically opportune but offers insight into the daily experiences and well-being of a growing portion of the American population.
Background
Union Formation in Later Life
Marriage and cohabitation share some similarities (for an overview of the characteristics of households, see Burch and Matthews 1987). Partners in both unions benefit from economies of scale and opportunities for intimacy and companionship. But, marriage’s roles and norms are standardized, and the relationship is legally and religiously defined. Marriage involves greater sacrifice of personal freedom and creates interdependencies between spouses via commingling finances and social networks (Cherlin 2004; Nock 1995). In contrast, cohabitation’s roles and norms are nebulous and the relationship lacks legal and religious signification. Cohabitors do not have a “script” for whether household expenses are shared, pension or health benefits include the partner, or families need to socialize with or accept both partners (Karlsson and Borell 2005). Thus, commitment and caregiving obligations are thought to be greater in marriage, whereas the creation of roles and norms within cohabiting unions is more deliberate and specific to the arrangements of each partner.
Differences between marriage and cohabitation matter because older adults face unique circumstances concerning repartnering. The legal ties of remarriage may complicate children’s inheritance and threaten older adults’ receipt of entitlement incomes, such as Social Security (Chevan 1996; Hatch 1995). Single older women especially fear that by remarrying they will become trapped in traditional marital roles, have to assume caregiving burdens, and give up control of their finances to a new husband (Bulcroft and O’Conner 1986; Karlsson and Borell 2005; Talbott 1998). For these reasons, older women express interest in nonmarital unions that preserve their autonomy (Davidson 2002; de Jong Gierveld 2002; van den Hoonaard 2002). It is not surprising then that 8 in 10 single older women oppose remarrying, and half very strongly oppose it (Talbott 1998), while a quarter express interest in cohabitation (Bulcroft and Bulcroft 1991).
Single older men, in contrast, feel deprived of their former wife’s caregiving and have come to expect and depend on traditional marital roles (Lee et al. 1998; Mason 1996). Older men specifically report seeking emotional support as a motivation for repartnering (Bulcroft and O’Conner 1986). These motivations may underlie why older men are more likely to repartner than older women (Brown et al. 2006; Chevan 1996; de Jong Gierveld 2004; Wu and Balakrishnan 1994) and do so sooner after becoming single (de Jong Gierveld 2002).
Later-life repartnering is about constraint as well as choice. A sex-ratio imbalance in older adulthood limits the availability of potential male partners. The imbalance grows markedly after retirement from 114 for 65- to 69-year-olds to 210 for persons 85 and older (Kung et al. 2008). The likelihood of repartnering also declines with age (Brown et al. 2006; Hatch 1995; de Jong Gierveld 2004) and across time since becoming single (de Jong Gierveld 2002, 2004; Wu and Balakrishnan 1994). Men’s tendency to repartner with younger women further exacerbates the sex-ratio imbalance (Goldman et al. 1984). So although men comprise one-quarter of the older population, they make up nearly two-thirds of older cohabitors (Chevan 1996).
Past studies show that other demographic characteristics are associated with relationship status in later life. Most single older adults were once married, and divorced individuals are more likely to remarry or cohabit than the widowed (Chevan 1996; de Jong Gierveld 2002). Many widowed persons feel their former marriages were satisfying and express loyalty to their deceased spouses (Davidson 2002; Stevens 2002), which may lead to prolonged singlehood in later life. Cohabitation is more common among nonwhites than whites (Raley 1996), a pattern that may carry into older adulthood via cohort replacement. In later life, health serves as a proxy for morbidity and general quality of life. Although older adults are reluctant to repartner with poor-health individuals (Bulcroft and Bulcroft 1991), older women weigh a potential partner’s health limitations more seriously than men (Karlsson and Borell 2005). Poor-health older adults are also more likely to be in cohabiting relationships than in remarriages (Brown et al. 2005).
Theory
To examine union formation, scholars have drawn on two competing theories: Becker’s independence hypothesis and Oppenheimer’s theory of marriage timing. Both focus on the role of economic resources, but they take opposing views on how resources affect women’s (but not men’s) likelihood of marrying and cohabiting.
According to Becker (1991), the gains to marriage are greatest when men specialize in the market and women specialize in the home and they exchange the outputs associated with these comparative advantages. Men are evaluated by their current and future economic contributions to the household. Greater economic resources allow men to offer greater financial security than men with fewer resources. As a result, men’s resources increase their likelihood of marriage while decreasing marital alternatives, such as staying single or cohabiting.
As women’s education, employment, and earnings have increased relative to men’s, women have forgone home specialization and gained economic resources instead. Becker argues that resources lower women’s dependence on a potential husband, and so women with greater resources have little incentive to marry. Instead, cohabitation may be a more attractive, and feasible, marital alternative in part because its norms for specialization are ambiguous (Cherlin 2004; Nock 1995). The independence hypothesis posits that women’s resources would reduce the likelihood of marriage while increasing that of marital alternatives.
Contrary to the independence hypothesis, Oppenheimer’s theory (1988, 1994, 1997) suggests that women’s resources increase the likelihood of marriage. For Oppenheimer, union formation depends on perceptions of economic security because marrying requires sufficient resources to maintain an independent household. The feasibility of marriage therefore depends on meeting financial prerequisites and sustaining them because marriage is seen as a long-term commitment (Dixon 1971; Oppenheimer 1988; Oppenheimer et al. 1997).
Historically, men’s resources were critical for meeting such prerequisites (Dixon 1971). Changes in labor markets have deteriorated men’s economic standing while improving women’s, which has transformed the marital bargain. Because the economic threshold for marriage is difficult for a sole male breadwinner to reach, specialization is costly (not advantageous, as Becker argues). Indeed, the “temporary or permanent loss of one specialist in a family can mean that functions vital to the well-being of the complementary specialist . . . are not being performed” (Oppenheimer 1997:447). As a result, women are evaluated on their economic standing, and so their resources would increase the likelihood of marriage over alternatives such as cohabitation or singlehood.
Whereas Becker would reason that marital alternatives are a response to economic security among women, Oppenheimer argues that they are a response to economic uncertainty (Oppenheimer 1988, 2003). Because cohabitation is thought to be a less committed relationship than marriage, its economic prerequisites are relatively lower. Individuals who are unwilling to marry because of economic insecurity may instead cohabit (Oppenheimer 1988, 2003; Smock and Manning 1997). From this perspective, low economic resources would facilitate cohabitation relative to marriage for men (as Becker argues) and for women (contrary to Becker).
Consistent with Becker and Oppenheimer, most studies have found a positive relationship between men’s economic resources and marriage (Cooney and Hogan 1991; Goldscheider and Waite 1986; Oppenheimer et al. 1997; Sassler and Schoen 1999). Regarding the role of women’s resources, most evidence supports Oppenheimer. Whether measured as education, employment, or income, resources typically increase women’s likelihood of marriage, though education may delay the transition to marriage (Goldscheider and Waite 1986; Lichter et al. 1992; McLaughlin and Lichter 1997; Oppenheimer and Lew 1995; Oppenheimer et al. 1997; Sassler and Schoen 1999). Using aggregated data, some studies have found that marriage rates are lower in areas with high female employment (Cready et al. 1997; Lichter et al. 1991; Waite and Spitze 1981). These studies are criticized, however, because aggregate-level variation may not accurately reflect individual-level behaviors (Oppenheimer 1997).
No study has found that economically advantaged individuals are more likely to cohabit than marry, but poorer persons have higher likelihoods of cohabiting than their nonpoor counterparts (Clarkberg 1999; Oppenheimer 2003; Xie et al. 2003). Among cohabitors, those with good economic prospects are the most likely to marry (Brown 2000; Sanchez et al. 1998; Smock and Manning 1997). Thus, contrary to the independence hypothesis, women’s resources do not appear to facilitate marital alternatives—at least among young adults.
Little work has been done on the economic determinants of later-life union formation, and what exists has produced mixed findings. Some have reported that education and income negatively associate with cohabitation compared with marriage among older men (Hatch 1995). In line with Oppenheimer, older female cohabitors tend to have lower household incomes than remarried older women (Brown et al. 2006). Because older cohabitors appear to be disadvantaged compared with older married persons, some scholars have concluded that poverty characterizes later-life cohabitation (Brown et al. 2006; Chevan 1996).
Others have found that employment and education positively associate with later-life cohabitation (Chevan 1996; de Jong Gierveld 2004), consistent with Becker. Older female cohabitors have a greater likelihood of full-time employment (Brown et al. 2006) and higher earnings (Hatch 1995) than older married women. Yet others have found no relationship between education and later-life cohabitation (Chevan 1996; Wu and Balakrishnan 1994), while Brown et al. (2006) found no difference between the incomes of older remarried and cohabiting men. It is also unclear how economic resources correlate with later-life singlehood. Education may positively associate with later-life remarriage versus singlehood (de Jong Gierveld 2004). Yet others have found the opposite relationship for older widows (Smith et al. 1991), suggesting that older women’s resources subsidize singlehood.
Current Study
Economic Resources
In later life, it seems unlikely that economic potential (education) or work-related measures (employment, earnings) fully capture older adults’ resources. Pensions, Social Security, and other assets may better reflect economic standing, but their receipt often varies by gender and marital status (i.e., widows’ collection of Social Security death benefits).
To capture some of this complexity, we should consider wealth as a measure of economic resources. Wealth reflects a lifetime of savings from earnings and investments in stocks, pensions, and real estate. Wealth includes financial (liquid) assets and nonfinancial (illiquid) assets such as housing. Homeownership is a core asset of Americans (McNamee and Miller 1998), especially older adults, for whom it represents financial and personal independence (Chevan 1996; Hatch 1995). Because wealth is a stock of savings, it is less sensitive than income to changes in earnings and employment because of retirement. Incomes are still important resources for older adults. About two-thirds of single older women receive at least 50% of their income from entitlement sources (McGarry and Schoeni 2000). Nonetheless, older adults could have substantial assets that their incomes do not reflect.
Whereas wealth represents a stock of resources, transfers could reflect a resource drain. Transfers are money given in the form of gifts, loans, or aid from one family member to another. They are an enduring aspect of later life, and older adults feel substantial obligation to give financial assistance to family (Cheal 1983; Hatch 1995; Morgan 1982). One-half of older Americans make financial transfers, usually to offspring (Hogan et al. 1993). Coresidence can be a form of assistance that involves entrenched transfers. Compared with older men, older women are more likely to have coresiding family (Spitze and Logan 1990) and are 86% more likely to give financial assistance to family (Hogan et al. 1993).
Wealth and Later-Life Union Formation
Becker and Oppenheimer have predicted that men’s resources positively associate with marriage and negatively associate with marital alternatives, such as cohabitation or singlehood. For older women, the theories offer competing predictions. Based on the independence hypothesis, single older women’s wealth may positively associate with singlehood or cohabitation relative to marriage. Because they face marriage-market constraints that compound with age, even if older women’s resources delayed remarriage, thereby subsidizing the search process (Oppenheimer 1988), delay would be synonymous with nonmarriage. Single older women’s resources would, in effect, negatively associate with remarriage versus singlehood, as some research has found (Smith et al. 1991).
Resource-poor older women have a financial incentive to remarry. Widows who remarry experience gains to wealth as large as if they were never widowed (Zick and Smith 1988). Compared with the continuously married, divorced and widowed cohabitors have one-third the wealth while their remarried counterparts have three-quarters the wealth (Wilmoth and Koso 2002). Thus, cohabitation is a poor economic substitute for remarriage and, like singlehood, may be feasible mainly for wealthier women (who can afford to live independently). Cohabitation may appeal to women who want to avoid caregiving burdens and protect their assets, particularly entitlement incomes whose receipt often depends on marital status. Consistent with Becker, older women’s wealth may therefore positively associate with singlehood or cohabitation relative to marriage.
Based on Oppenheimer’s theory, on the other hand, older women’s wealth may positively associate with remarriage. Older adults are deeply concerned about financial security and maintaining an independent household (Chevan 1996; Hatch 1995; Karlsson and Borell 2005). If specialization is risky during young adulthood, it could be perilous in later life because of morbidity and the financial distresses of widowhood or divorce. In settings where it is precarious to specialize, women’s resources are valued for their economic contributions to a household (Oppenheimer 1988, 1994). Wealth may be an attractive characteristic in potential older husbands and wives because it increases the gains to marriage. Wealth therefore may positively associate with remarriage among men and women while discouraging singlehood or cohabitation.
The relationships between resources and repartnering may further vary with age. Older adults’ resources begin shifting in their 60s. Income flows constrict because of retirement, when older adults begin relying on entitlement resources and accumulated savings. Because older adults have few chances to recoup lost savings, their perceptions of financial security may change—perceptions that are critical for union formation (Oppenheimer 1988). Therefore, the associations between wealth and repartnering may strengthen with age. For Becker, this would translate into greater negative effects on the likelihood of remarriage versus singlehood or cohabitation for the oldest women; for Oppenheimer, greater positive effects on the likelihood of remarriage versus singlehood or cohabitation for the oldest men and women alike.
Financial Transfers and Later-Life Union Formation
Transfers may be costs to repartnering and therefore decrease the likelihood of union formation. Older adults may be uninterested in assuming the financial burden of a potential spouse’s family. The perceived costs may not be as high for cohabiting partners who have fewer legal and normative obligations to one another than do spouses. Transfer recipients also have vested interests in discouraging their older kin’s remarriage. The remarriage of older adults reduces both the amount and frequency of assistance they give to family (Pezzin and Schone 1999). Indeed, offspring often resent their parents’ remarriage for this reason and may even pressure them to cohabit to safeguard their assets (Chevan 1996; de Jong Gierveld and Peeters 2003; Hatch 1995). Older adults are sensitive to such pressures and sometimes employ a strategy of “passive acquiescence” to their children’s wishes to avoid family conflict (Beel-Bates et al. 2007; Pyke 1999).
As persons rely more on fixed resources and savings, transfers may have greater negative effects on remarriage at the oldest ages because of the relatively greater perceived costs. The relationship between transfers and repartnering may be gendered as well. Offspring, especially residential ones, have negative effects on older women’s chances of remarriage (Smith et al. 1991; Wu 1995). More widowed women than men report that they are not remarried because their children discouraged it (Davidson 2002). Offspring do not associate with older men’s chances of remarrying perhaps because their family relationships deteriorate after marital dissolution (Di Leonardo 1987).
Method
Data and Sample
This study draws on five waves of longitudinal panel data from the 1998–2006 Health and Retirement Study (HRS). The data are collected biennially and constitute a nationally representative sample of over 26,000 noninstitutionalized Americans who are at least 50 years old. Five cohorts comprise the HRS: Asset and Health Dynamics Among the Oldest Old (born 1923 and earlier); Children of the Great Depression (1924–1930); the original HRS (1931–1941); War Babies (1942–1947); and Early Baby Boomers (1948–1953), who were first interviewed in 2004.
When respondents first enter the survey, HRS collects reports on marital and cohabitation experiences. After entering the survey, HRS collects information for both unions biennially. These histories make possible a longitudinal analysis of later-life union formation. Respondents in the sample for this current study must either be single at the baseline interview or become single between 1998 and 2006.1 The sample excludes the continuously married and never-married. To track repartnering, the sample includes only respondents who are interviewed at least twice. The final analytic sample contains 8,348 previously married older Americans (aged 50 and older) who were single between 1998 and 2006.2
Outcome: Repartnering
Using information on respondents’ residential unions, I construct a history of coresidential repartnering between 1998 and 2006. Repartnering is a time-varying dummy variable that measures whether individuals (1) remain single, (2) remarry, or (3) cohabit during any given wave. HRS does not capture intra-wave cohabitations, which means the sample may be selective of more durable cohabiting unions. HRS collects self-reported statuses for marriage and cohabitation, the latter defined as “living with a partner as if married.”3
Predictors: Economic Resources
Education is a time-constant, continuous measure (in years) of respondents’ schooling. It is measured the first time respondents are interviewed. Income is a time-varying, logged variable that includes monies from wages and salaries (up to two jobs), government transfers (unemployment, worker’s compensation, veteran’s benefits), annuitized payments, Social Security disability income, Supplemental Security Income, and Social Security Retirement (including spousal and widowed benefits). The latter four sources are especially important because many older adults receive annuitized payments and are collecting Social Security benefits. Because the variable is time-varying, it accounts for changes in income flows across waves.
Wealth is a time-varying, logged variable measuring net worth (the value of an asset less any debt, lien, or balance due).4 It is a composite measure of two components. (1) Financial assets include the net worth of pensions, retirement accounts, and mutual funds (up to three); checking, savings, and money market accounts; and certificates of deposit, savings bonds, and treasury bills. (2) Nonfinancial wealth includes the net worth of housing and real estate, vehicles, businesses, and other durable goods. Wealth is therefore an aggregate measure, the sum of financial and nonfinancial net worth. Those with no or negative net worth are treated as having zero wealth.
Two time-varying dummy variables measure transfers. Financial transfer captures whether respondents gave at least $500 to a child, sibling, or parent within the two years prior to interview (reference is no). Income provision captures more substantial transfers, whether respondents provided at least half of the income for a child, parent, or sibling within the two years prior to interview (reference is no).5
Controls
The likelihood of repartnering declines with age (de Jong Gierveld 2004; Wu and Balakrishnan 1994), so I measure age (in years) at the time of risk with a time-varying variable. I also measure time single (since respondents’ most recent widowhood or divorce) with a time-varying variable. To capture curvilinear effects of time, I include a quadratic for time single. Repartnering varies by gender, marital history, and race/ethnicity (Brown 2000; Chevan 1996; de Jong Gierveld 2002; Raley 1996). Time-constant dummy variables capture whether respondents are female, divorced (reference is widowed), and white (reference is nonwhite).6 Time-varying, self-rated variables measure respondents’ health as excellent or good, fair, or poor (reference category). Two time-varying dummy variables measure family contact and social networks. Residential offspring captures whether respondents share a household with at least one child (reference is no), while high network contact captures whether respondents have close friends and family who live nearby them in the same neighborhood (reference is no).
Analytic Strategy
The study uses discrete-time multinomial logistic regression to model the relative risks of repartnering. The method is appropriate because it avoids proportionality assumptions (Allison 1982), permits the inclusion of fixed and time-varying covariates (Yamaguchi 1991), and models predictors separately by transition type. It also minimizes bias that arises from left-truncated data (Guo 1993). Data are left-truncated when some subjects come under observation after having been exposed to the risk of an event. Left-truncated data overrepresent low-risk cases because high-risk subjects tend to experience the event and drop out of the sample prior to observation. As a result, parameter estimation can be biased because risk tends to be underestimated. Using discrete-time models that include duration of exposure at the start of and over the course of observation effectively minimizes the bias (Guo 1993).7
Results are performed separately by gender because of theoretical differences in union formation between men and women, and many variables related to repartnering differ between older men and women (e.g., marital status, health, coresidence). Results are also performed separately by age category at the time of risk (ages 50–64 and 65 and older) because repartnering experiences may differ between the two age groups. For example, the older group faces distinct marriage markets because of sex-ratio imbalances (Goldman et al. 1984). Moreover the effects of economic resources on repartnering may differ by age, as discussed earlier.8 To determine statistical significance across models, I interact the chief predictors (economic resources) with the variable of interest (female or age category) in the full model. Significant differences (at least p < .10) are reported in the tables.
Results are presented in two models. Model 1 includes education and income as well as controls. Model 2 retains these measures and introduces wealth, transfers, and income provision. The two-step process allows assessment of the effects of wealth and transfers net of education and income. Wealth is an aggregate measure, but results could be sensitive to asset type. For example, financial wealth is relatively accessible compared with nonfinancial wealth. But the latter, although more difficult to convert to cash, is more valuable. Being asset-rich in terms of housing may shape union formation differently than being asset-poor in terms of liquid wealth, especially if housing is viewed as a core component of financial security or if older adults need liquid wealth to make transfers. To assess sensitivity by wealth type, I perform supplemental analyses that separate financial and nonfinancial assets (see Table 4).
Results
Summary Statistics
Most of the sample remains single between 1998 and 2006. Overall, about 1 in 8 single older adults repartners (Table 1). Among younger men (aged 50–64), the incidence of repartnering is 2 in 5, while among the oldest women (aged 65 and older) it is 1 in 25 during the period of observation. In general, at least twice as many men remarry or cohabit as women. Among those 65 and older, the gender difference is starker: five times as many men remarry and nearly four times as many cohabit.
Economic resources vary considerably by gender and age. Regardless of gender, 50- to 64-year-olds are better educated and have higher incomes than their older counterparts. Although older adults’ average incomes are small, their wealth is substantial. The typical woman aged 65 or older has a yearly income of $15,822, about three-quarters the size of her younger counterpart’s.9 Her average net worth, however, totals $200,445—significantly more than that of her younger counterpart. Typically, the oldest adults have more wealth than 50- to 64-year-olds, while women have significantly less wealth than men. Compared with the oldest age group, a higher proportion of 50- to 64-year-olds makes transfers to family.
Figures 1 and 2 show wealth and income at the baseline interview by repartnering status. Regardless of age, older women who remained single during the period of observation were economically disadvantaged at their baseline interview compared with those who repartnered. Women who cohabited typically had about two-thirds more wealth, and women who remarried about twice as much wealth, as those who stayed single (Fig. 1). The economic disadvantage of single men is most evident among those 65 and older (wealth differences by repartnering status are small among men aged 50–64) (Fig. 2). In sum, later-life singlehood is typified by economic disadvantage, which appears more widespread among women who remained single than among women who entered any kind of union.
On average, those who remarried had more wealth than those who cohabited, regardless of gender and age. The story is more nuanced than the figures reveal, however. Compared with older adults who remarried, individuals who cohabited had statistically similar incomes and financial wealth. Only the nonfinancial wealth of the remarried was greater than that of cohabitors. When housing wealth is excluded from nonfinancial assets, the significant difference disappears. Thus, older adults who remarried had more wealth than those who cohabited, on average, because they owned more valuable real estate.
Multivariate Analyses: Time
Tables 2 and 3 present results from multivariate analyses. Among 50- to 64-year-olds, time single has no statistically significant relationship with the risk of repartnering for either men or women (Table 2). Among persons 65 and older (Table 3), time single has a strong negative association with remarrying or cohabiting versus staying single. Further, distinct gendered patterns in the effects of time emerge among those aged 65 and older. A one-year increase in time single among men associates with .95 times the risk of remarrying versus staying single (Panel B, Table 3). Among women, the effect size is significantly greater (.91), while time single negatively associates with the risk of cohabiting versus staying single (Panel A, Table 3). Thus, time single decreases the risk of any kind of repartnering for the oldest women but decreases the risk of remarriage only for men—and even then to a smaller extent than it does for women. These results are consistent with findings from previous research (Brown et al. 2006; de Jong Gierveld 2002, 2004; Wu and Balakrishnan 1994) showing that the odds of repartnering decline across time, more so for women than men.
Multivariate Analyses: Education and Income
Model 1 in Table 2 (50- to 64-year-olds) and Table 3 (65 and older) presents results using education and income to predict the relative risk of remarrying, cohabiting, or staying single. Based on these models, education and income have little relationship with older men’s repartnering and contradictory ones with women’s repartnering.
Among men aged 50–64, education and income never attain statistical significance for any kind of repartnering versus staying single, though income negatively associates with the risk of cohabiting versus remarrying (Panel B, Table 2). Neither does income attain significance among men 65 and older, though education does (Panel B, Table 3). Each one-year increase in education associates with 1.07 times greater risk of remarrying than staying single among the oldest men (a significant difference from the effect of education for men aged 50–64).
Results for women are less clear. Among women aged 50–64, education decreases the risks of any kind of repartnering. Income decreases the risk of remarrying versus staying single and increases the risk of cohabiting versus remarrying: each logged unit increase in income associates with 1.12 times greater risk of cohabiting versus remarrying (Panel A, Table 2). In other words, women’s education and income positively associate with marital alternatives. Among women 65 and older, education increases the risks of remarrying versus staying single (Panel A, Table 3), an effect that is significantly different—indeed opposite—from that for women aged 50–64 (Panel A, Table 2). For women, therefore, financial resources sometimes positively associate with remarrying relative to staying single or cohabiting; other times they negatively associate with remarrying.
Multivariate Analyses: Wealth and Transfers
Model 2 introduces wealth and transfer measures. Results from this model suggest that older adults with greater wealth are more likely to remarry or cohabit than they are to stay single, regardless of gender or age. Among women aged 50–64, each logged unit increase in wealth associates with 1.10 and 1.06 times greater risk of remarrying or cohabiting, respectively, than staying single (Panel A, Table 2). These effects are significantly more pronounced among women who are at least 65, for whom the risks of remarrying or cohabiting versus staying single are 1.34 and 1.22 times greater, respectively, per logged unit increase in wealth (Panel A, Table 3). In Model 1, education and income are significant for both age groups of women. With the addition of wealth and transfers in Model 2, however, education and income no longer attain significance. Interestingly, for neither age group of women does wealth associate with the risk of cohabiting versus remarrying. Thus, if older women repartner, their odds of cohabiting versus remarrying does not depend on their level of wealth.
Having made a transfer reduces the likelihood of remarrying versus staying single by about one-third for women aged 50–64 and one-fifth for women aged 65 and older. Transfers have no significant effect on cohabiting versus staying single, though they have pronounced effects on whether older women remarry versus cohabit. Single women who are 50–64 and who made transfers to family are 1.67 times more likely to cohabit than remarry, while their older counterparts are 2.06 times more likely (a significant difference). Income provision does not attain significance for older women regardless of age.
In Model 1, education and income had almost no association with men’s repartnering. The same is not true for wealth. Each logged unit increase in wealth for 50- to 64-year-old single men associates with 1.12 and 1.06 times greater risk of remarrying or cohabiting, respectively, than staying single (Panel B, Table 2). Among men 65 and older, the results are nearly identical (they are statistically indistinguishable from the effects of wealth for men aged 50–64). Wealth is the only economic measure that significantly associates with men’s repartnering. Neither transfers nor income provision attain significance for older men regardless of age. Similar to results for women, wealth does not associate with the risk of cohabiting versus remarrying for single older men.
Sensitivity Test: Wealth Type
Table 4 compares the effects of financial and nonfinancial assets on the relative risks of remarrying, cohabiting, or remaining single. Total wealth is provided as a reference and is taken from Model 2 (in Tables 2 and 3). Results for financial and nonfinancial assets are from separate analyses with the full model.
For women, regardless of age, the risk of repartnering versus staying single is the same whether using the disaggregated wealth measures or the aggregated ones presented in Model 2 (see Panel A, Table 4). Moreover, it makes no difference whether nonfinancial assets include or exclude housing wealth. In short, increases in any kind of wealth (liquid or illiquid) for single older women raise their risk of remarrying or cohabiting relative to staying single.
For single older men, results are sensitive to how wealth is measured. Among men who are at least 65, financial assets have no significant relationship with repartnering, while only nonfinancial wealth retains a significant and positive relationship. The positive effect is reduced to nonsignificance, however, when housing wealth is excluded from the measure of nonfinancial holdings. Therefore, not only do financial resources have no significant effect on the repartnering experiences of the oldest men, but the positive effects of wealth are attributable to the net worth of housing. Effects of nonfinancial wealth are not sensitive to housing assets among men aged 50–64.
Discussion
Drawing on Becker’s (1991) independence hypothesis and Oppenheimer’s (1988) theory of marriage timing, this study examines union formation among older Americans (aged 50 and older)—in particular, the role of economic resources in remarriage and cohabitation. Overall, about 1 in 8 single older adults remarries or cohabits during the sample’s eight-year period, although the incidence varies by age and gender. Whereas 2 in 5 younger men (aged 50–64) repartner, only 1 in 25 older women (aged 65 and older) do so. Wealthier older adults are more likely to remarry or cohabit than to stay single compared to their less wealthy counterparts, regardless of gender and age. For the oldest men (65 and older), housing wealth alone drives the positive relationship between wealth and repartnering. In other words, single older men with more valuable property are more likely to enter any kind of union than to remain single. Notably, wealth is not associated with older adults’ risk of cohabiting versus remarrying. Here is an instance when a nonsignificant result is just as intriguing as a significant one. The result is noteworthy because it is at odds with past research, which has repeatedly found that, relative to marriage, cohabitation is selective of the economically disadvantaged (Clarkberg 1999; Oppenheimer 2003; Xie et al. 2003). At older ages, this does not appear to be true.
Alongside presenting new findings to a literature that has largely relied on qualitative work mainly from Europe (de Jong Gierveld 2002, 2004; de Jong Gierveld and Peeters 2003; Stevens 2002) and cross-sectional data (Brown et al. 2006; Chevan 1996; Hatch 1995; King and Scott 2005), this study has theoretical implications. Its findings do not support Becker’s independence hypothesis, which is somewhat surprising considering its face validity at older ages. Older women may be wary of remarriage, fearful of losing their assets and reluctant to assume gendered roles typical in marriage (Bulcroft and Bulcroft 1991; Talbott 1998). Some scholars have even suggested that financially secure older women may avoid remarriage (Davidson 2002; de Jong Gierveld 2002; de Jong Gierveld and Peeters 2003; Smith et al. 1991). Single older women may indeed feel reluctant to remarry and prefer marital alternatives, but their behaviors tell a different story: even small increases in wealth greatly raise single older women’s risk of remarrying. If behaviors reflect latent values, then wealthier older women may prefer remarriage to singlehood. Ultimately, this study examines behaviors. Older adults’ preferences and motivations are beyond its scope.
This study’s findings offer implications for how we conceptualize later-life union formation. First, it is important to distinguish resource types. Only housing wealth matters for the oldest men’s repartnering. Housing is a sign of financial security, stability, and independence among older adults (Chevan 1996; Hatch 1995). Yet, if economic security is so important, why do other assets, such as pensions and retirement accounts, not matter just as much? One interpretation is that valuable—or debt-free—housing may be so important in older marriage markets that other kinds of wealth pale in comparison. Perhaps autonomy and noninstitutionalized living, reflected in homeownership, are the most salient issues to the oldest of adults.
If this conclusion were accurate, then a “good” male partner in later life would be a provider of valuable housing. Despite being past the traditional breadwinning age, the oldest men would still be evaluated as partners based on their status as providers of housing. What is more intriguing is that housing wealth plays analogous roles in the likelihood of remarrying or cohabiting among the oldest men. One could imagine that marriage would provide more secure access to a husband’s housing than would cohabitation. Nonetheless, housing wealth appears to be as attractive a trait in potential husbands as it is in cohabiting partners. Having valuable or secure housing may be an attractive trait independent of the greater commitment thought to characterize marriage.
On the other hand, the effect of housing wealth may not reflect marriage market attractiveness or signal a partner’s willingness to remarry or cohabit. Rather, it may be an indicator of the respondent’s own inclination to repartner. For example, older men with their own homes may want and seek a wife to help with household labor and domestic care. This interpretation could explain the analogous role that housing wealth plays in the oldest men’s likelihood of repartnering. Without couple-level data on marriage markets or data on older adults’ repartnering motivations, this study cannot draw definitive conclusions about either interpretation.
Second, the economic gulf that separates marriage and cohabitation during young adulthood is considerably narrower in later life. Some studies have suggested that poverty is a backdrop to later-life cohabitation (Brown et al. 2006; Chevan 1996). Results here do not support this conclusion. For one, neither the incomes nor financial wealth of older adults who cohabit statistically differ from the resources of older adults who remarry. The remarried do have more valuable real estate than cohabitors (about $82,500 more valuable, on average). Nevertheless, it is difficult to conclude that “poverty” characterizes later-life cohabitation even relative to marriage when the average net worth of older cohabitors exceeds a quarter-million dollars. We might, on the other hand, draw such a conclusion if income alone were used to measure older adults’ resources. Older adults have substantial assets that their income does not reflect.
It is telling that wealth plays no significant role in the likelihood of remarrying versus cohabiting. During young adulthood, economic advantage facilitates marriage relative to cohabitation. Scholars have theorized that cohabitation during young adulthood may be a response to economic uncertainty (Oppenheimer 1988) and that young cohabitors are delaying marriage until they meet its economic prerequisites (Smock et al. 2005). Yet at older ages, cohabitation may have unique functions (Brown et al. 2006; Chevan 1996; Hatch 1995; King and Scott 2005) that are not tied to economic uncertainty. Perhaps some older adults are not interested in meeting the economic prerequisites of marriage because they desire a long-term marital alternative (King and Scott 2005). Future research on the stability of later-life cohabiting unions could help answer these questions.
Whereas wealth facilitates repartnering among older adults, transfers discourage older women from repartnering. Transfer recipients (family members) may disapprove of older women’s repartnering because the recipients could lose control of assets (Chevan 1996; de Jong Gierveld and Peeters 2003; Hatch 1995). But transfers discourage only remarriage, not cohabitation. Because cohabitation entails fewer legal and economic ties than marriage, cohabitation may be a “safer” alternative in terms of safeguarding inheritances (Cherlin 2004; de Jong Gierveld 2002; Karlsson and Borell 2005). Other explanations are possible. Transfers may limit the marital search process for older women; or older adults may be uninterested in remarriage, but not cohabitation, when their family needs financial assistance. Whatever the reason, the transfer relationship with family members is meaningful only for older women’s repartnering experiences.
This study only broaches the topic of later-life union formation. Future research should use couple-level data to examine partner exchanges in later-life marriage markets. Without it, we cannot conclude whether variable effects indicate individuals’ motivation to repartner or their attractiveness to potential partners. Future research should consider nonfinancial transfers, such as instrumental support and caregiving, as well as the role of debt in union formation. Measuring later-life resources is complex. Take, for example, potential income variations by gender and marital status from Social Security spousal and survivor benefits. This study only counts these monies and cannot assess whether the desire to receive them, or fear of losing them, affects older adults’ repartnering. In a similar vein, it has relied on rational-choice assumptions in Becker’s and Oppenheimer’s theories. Other factors affect older adults’ interest in repartnering, including loneliness and the desire for intimacy (Carr 2004). Because of data limitations, older adults’ motivations and preferences remain unmeasured.
Future research would benefit from additional longitudinal data. This study examines a narrow window of older adults’ lives. Because the data are left-truncated, results are generalizable to a select sample of low-risk older adults. Other union types should be studied as well. For example, in living-apart-together relationships (LATs), couples keep separate residences but maintain intimate relationships (de Jong Gierveld 2004; Levin 2004; Stevens 2002). Such unions may appeal to older women who prefer to avoid the gendered housework that typifies residential unions. Estimates from the United States suggest that about one-third of all persons who are not married or cohabiting are in LATs (Strohm et al. 2009). The HRS does not identify these relationships, so this study may miss a potentially important form of later-life repartnering while miscoding some individuals as “single” even though they are in a nonresidential union.
Nevertheless, this study makes substantive contributions to understanding later-life union formation. It adds to the growing literature on marriage and cohabitation during older adulthood by refuting Becker’s independence hypothesis, showing that wealth does not affect the likelihood of remarrying relative to cohabiting, and revealing that housing wealth is a driving economic resource underlying older men’s repartnering. Scholars face a paucity of research on later-life relationships at a time when changes in the population, age, and household structures are reshaping the demographic landscape of the United States. Moreover, intimate relationships have a profound influence on individual well-being, social interaction, and caregiving and support in later life. As aging baby boomers swell the ranks of the older population, it is essential to understand the factors shaping older adults’ experiences with union formation.
Acknowledgments
For their advice and time spent reviewing this article, I thank the former and current editors of Demography, Kenneth Land and Stewart Tolnay, the anonymous reviewers, Elizabeth Cooksey, Adrianne Frech, Jamie Lynch, Elizabeth Menaghan, Kimberley Murphy, Matthew Painter, and Zhenchao Qian. Earlier versions of this study were presented at the annual meetings of the Population Association of America in Dallas, TX, April 15–17, 2010.
Notes
For example, the sample would include persons who were married in 1998 and divorced in 1999, but exclude those who divorced in 1989 and remarried in 1996. Consequently, the sample tends to exclude high-risk cases (younger, male, employed, and divorced) while overrepresenting low-risk cases (older, female, widowed, and retired).
Respondents remain in the sample until they remarry, cohabit, or are censored because of attrition or death. Most variables have no more than 4%–5% missing data in any given wave. About 3% and 11% of cases have indeterminate marital statuses and dates of most recent marriage, respectively. Multiple imputation is used to replace missing data. Respondents do not contribute any data to the analysis for rounds in which they are not interviewed.
Because cohabitation is less normative among older adults than younger ones, cohabitation may be underreported.
Persons with higher incomes likely have more wealth. Wealth and income are only weakly correlated, though (.13).
Financial transfer and income provision may be correlated, but modeling them separately did not change the results. These measures may be endogenous as well. One potential solution is to lag the transfer and provision variables. Doing so changed results slightly but introduced problems with interpretation and precision. I show results using nonlagged variables.
Cell sizes for disaggregated racial/ethnic categories are too small to yield reliable estimates. The aggregated racial/ethnic measure preserves cell sizes but also masks variation in union formation.
Another solution is discarding left-truncated data (Allison 1982). In supplemental analyses, I tried excluding persons who were single for more than 30 years, 20 years, or 10 years. More restrictive exclusions (i.e., 10 years) reduced sample sizes and thus changed parameter estimation somewhat.
Resources could also change because of retirement. Rather than separate models by age, I tried including a series of dummy variables for labor force participation (e.g., full-time employment, fully or partially retired, any employment). I also tried redefining the age cutoff at 62 (to capture early retirees). None of the dummy variables reached statistical significance, and results from the “early retiree” model did not differ from those presented here.
The percentage of older women in the labor force differs by age. About 11% of women 65 and older are employed (whether part or full time), whereas 58% of their younger counterparts are employed.
