Repartnering Following Gray Divorce: The Roles of Resources and Constraints for Women and Men

Abstract

The doubling of the gray divorce rate (i.e., divorce at age 50 or older) over the past few decades portends growth in later-life repartnering, yet little is known about the mechanisms undergirding decisions to repartner after gray divorce. Using data from the 1998–2014 Health and Retirement Study, we examined women’s and men’s likelihoods of forming a remarriage or cohabiting union following gray divorce by estimating competing risk multinomial logistic regression models using discrete-time event history data. About 22 % of women and 37 % of men repartnered within 10 years after gray divorce. Repartnering more often occurred through cohabitation than remarriage, particularly for men. Resources such as economic factors, health, and social ties were linked to repartnering, but constraints captured by the contours of the marital biography were also salient, underscoring the distinctive features of union formation in later life.

Introduction

In the United States, marital dissolution is on the rise in middle and later life, reflecting the rapid growth in gray divorce, a term describing divorce among persons aged 50 and older (Brown and Lin 2012; Kennedy and Ruggles 2014). A quarter-century ago, divorce during middle and later life was uncommon. Fewer than 1 in 10 people who got divorced in 1990 were at least age 50. Since then, the gray divorce rate has doubled. By 2010, one in four people getting divorced were aged 50 or older (Brown and Lin 2012). Moreover, the U.S. population is aging, which compounds the effect of the rate increase.

The rise in the gray divorce rate foretells a growth in repartnering during older adulthood. Divorce results in two single people who might form new unions, whereas widowhood leaves just one person available to repartner. Moreover, repartnership following divorce is more common than repartnership after widowhood, and this difference is larger among women than men (Brown et al. 2018; Vespa 2012).

Yet, the mechanisms underlying decisions to repartner following gray divorce are unknown. In fact, very few studies have addressed the determinants of later-life repartnership more generally, and the results have often been inconsistent (Brown et al. 2012; de Jong Gierveld 2004; Schimmele and Wu 2016; Vespa 2012; Wu et al. 2014). Additionally, research to date is marked by methodological weaknesses that may partially account for the disparate empirical findings. Specifically, some of the prior work has been limited to cross-sectional retrospective analyses (Brown et al. 2018; de Jong Gierveld 2004; Schimmele and Wu 2016; Wu et al. 2014). Other studies used longitudinal data but began observation at an arbitrary time point, resulting in samples that are overrepresentative of long-term singles, whose likelihood of repartnering is extremely low (Brown et al. 2012; Vespa 2012).

Our study offers significant advancements over existing research on repartnering among older adults. First, we rely on prospective, longitudinal data from the 1998–2014 Health and Retirement Study to estimate the likelihood of repartnering for women and men who have experienced gray divorce, treating remarriage and cohabitation as competing risks. Second, we track individuals from the time of their dissolution and follow them for up to a decade, which enables us to identify both the determinants and timing of repartnering following a gray divorce. Guided by recent empirical work on union formation during the second half of life (Brown et al. 2012; de Jong Gierveld 2004; Schimmele and Wu 2016; Vespa 2012), we assess how resources (economic factors, health, and social ties) and constraints (the marital biography and demographic characteristics) are linked to repartnering for women and men as well as whether these patterns differ by gender. The findings from our study help to elucidate the pathways to repartnership following a gray divorce, a notable concern in light of the rising rate of divorce during older adulthood.

Background

According to a recent overview and assessment of partnering research, “current literature … has given short shrift to (re)partnering at older ages” (Sassler 2010:561). As outlined by Sassler (2010) and others (Bulcroft and Bulcroft 1991; King and Scott 2005; Vespa 2012), the mechanisms underlying decisions to form a union in later life are distinctive from the factors that drive repartnering at earlier stages of the life course. In particular, the resources and constraints facing older adults are different from those that younger adults experience. The economic, health, and social support resources available to older adults are unique (Brown et al. 2012; Vespa 2012). Moreover, they confront age-related constraints, such as an unbalanced sex ratio that widens with age and may shape partnering opportunities (de Jong Gierveld 2004; Schimmele and Wu 2016).

Resources

Economic resources are positively associated with marriage formation. For women and men alike, factors such as education and wealth are related to increased odds of getting and staying married (Oppenheimer 1988, 1994, 2003; Schneider 2011). Financial stability is viewed as a prerequisite for marriage, whereas cohabitation is an option for those with inadequate or fragile economic resources (Cherlin 2004; Smock et al. 2005). Vespa (2012) examined whether the economic theories used in the literature on union formation during young adulthood, such as those espoused by Oppenheimer (1988), apply in later life for the previously married. Vespa found that consistent with the literature for younger adults, wealthier individuals were more likely to repartner during the second half of life. Wealth was not associated with favoring marriage over cohabitation, though, and thus he concluded that the role of economic factors in later-life cohabitation formation is distinctive. Unlike the evidence suggesting that cohabitation is a “poor man’s marriage” that is most popular among economically disadvantaged younger adults (Clarkberg 1999), Vespa’s findings for older adults provide empirical support for theorizing that conceptualizes cohabitation as a long-term substitute for remarriage in older adulthood (Brown et al. 2012; King and Scott 2005).

Other studies on the linkages between economic resources and union formation in later life are not consistent with Vespa (2012). Using the same data as Vespa (the Health and Retirement Study), Brown et al.’s (2012) analysis revealed no effects of wealth or other economic indicators (e.g., homeownership, pension receipt, employment, and health insurance) on the likelihood of forming a remarriage or a cohabiting union in later life, suggesting that economic resources may be less salient to later-life partnering decisions than partnering that occurs during young adulthood. In another study, which examined Canadians aged 45 and older, repartnering following marital disruption was unrelated to economic characteristics for women (except that having some savings was positively associated with remarriage); for men, having a pension and being employed were associated with a greater likelihood of forming a cohabiting union but had no appreciable effects on the risk of remarriage (Schimmele and Wu 2016). Other research conducted in the Netherlands showed that education and employment were positively associated with forming a remarriage but unrelated to forming a cohabiting union (de Jong Gierveld 2004). The inconsistencies marking research to date may stem in part from cross-national variation, but the opposing results from Brown et al. (2012) and Vespa (2012) are puzzling given that they relied on the same data set and took similar analytic approaches.¹

Health is an additional resource that is uniquely salient for older adults. It appears to be more consequential for union formation among women than men, but here again, the evidence is mixed. For example, better self-rated health was positively associated with remarriage among midlife women aged 50–64 but not for similarly aged men (Vespa 2012). Yet, self-rated health was unrelated to entry into cohabitation among midlife and older women. For men, self-rated health was immaterial to union formation (remarriage or cohabitation). This latter finding was echoed by Brown et al.’s (2012) analysis of union formation, but they also found that health was unrelated to women’s transitions into either remarriage or cohabitation. In a retrospective cross-sectional study, self-rated health was positively related to being remarried (but not to cohabiting) among women who experienced a later-life marital dissolution. No appreciable effects of health on repartnership status were found for men (Brown et al. 2018). Attention to other indicators of health, such as the number of chronic conditions or experiencing difficulties with activities of daily living (ADL), has been limited. For older Canadians, having a chronic illness was negatively associated with remarriage but unrelated to forming a cohabiting union among women and men alike (Schimmele and Wu 2016). In a U.S. sample of older adults, Brown et al. (2012) found no association between having at least one ADL difficulty and repartnering.

The social ties of older adults are also linked to union formation. Having friends and family members who live nearby may deter union formation in later life. Widowed men’s interest in remarriage drops to the weaker level of widowed women’s only when they have adequate sources of social support through friends or family (Carr 2004). The presence of nearby friends and family is associated with a lower likelihood of remarriage (but not cohabitation) among women (Brown et al. 2012; Vespa 2012) and cohabitation (but not remarriage) among older men (Vespa 2012).

Resident children are another type of social tie that tends to reduce the likelihood of union formation among older adults. The presence of a child in the household may make an older adult a less-attractive potential partner. But it is also likely that older adults are more hesitant to form a coresidential union when they have a child living with them (de Jong Gierveld and Merz 2013). Among older adults, a resident child was associated with lower odds of remarriage among women and reduced odds of forming a cohabiting union among men (Vespa 2012). Similar results were obtained from a cross-sectional sample of older adults who had experienced a later-life marital dissolution: women with a resident child were less likely to be either remarried or cohabiting, and men were unlikely to be cohabiting relative to those without any children in the household (Brown et al. 2018). In a study conducted in Canada, a resident child reduced the odds of transitioning to either cohabitation or remarriage among women but had no appreciable effects on men’s union formation (Schimmele and Wu 2016).

Constraints

Repartnering in later life is a distinctive process in large part because older adults face a unique set of structural constraints (Schimmele and Wu 2016). Of particular importance is the lopsided sex ratio. On average, women live longer than men, meaning that at older ages, more women than men are in the population. The situation is exacerbated by the fact that women tend to marry older men—or, stated differently, that men often marry younger women. Ultimately, the repartnering pools for women and men shift in opposing directions in later life. For women, the pool of potential partners shrinks rapidly as they age; for men, the pool of eligible partners grows with age. Consequently, it is not surprising that repartnership levels are much lower among women than men (Brown et al. 2018; Schimmele and Wu 2016).

Not only does the imbalanced sex ratio shape opportunities for repartnering but so too do gendered preferences to repartner versus remain single at this life course stage. Qualitative research has shown that older women are much less interested in repartnering than are older men (de Jong Gierveld 2002; McWilliams and Barrett 2014; Talbott 1998; Watson and Stelle 2011). Women often enjoy their newfound autonomy and independence following marital dissolution and view this stage of their lives as time for themselves. They are reluctant to form a new union, particularly a remarriage, because of the gendered expectations for caregiving (Talbott 1998). They already cared for a husband and children during their marriage and are not eager to repeat this experience, preferring instead to go solo. Men are typically much more enthusiastic about repartnering precisely because they are motivated to find a wife who will care for them (McWilliams and Barrett 2014). Older men tend to repartner quickly following dissolution to reestablish the social and emotional support that their wives provided. Both the imbalanced sex ratio and gendered preferences for repartnering combine to yield significantly lower levels of remarriage and cohabitation among older women than men (Schimmele and Wu 2016).

In addition to the demographic constraints of the lopsided sex ratio, which are exacerbated by women’s weaker affinity toward repartnering, the marital biography also shapes the union formation behaviors of older adults. Age at dissolution is negatively associated with the formation of either a remarriage or a cohabitation (Brown et al. 2012; Schimmele and Wu 2016; Vespa 2012). Schimmele and Wu (2016) maintained that age at dissolution is the most important predictor of repartnering in later life. The length of time since marital dissolution, or the time spent single, is negatively related to repartnering, which is most likely to occur soon after marital dissolution (Schimmele and Wu 2016). A key limitation of the research to date on later-life union formation is that it has lumped together the long-term single with the recently divorced or widowed (Brown et al. 2012; de Jong Gierveld 2004; Vespa 2012). This approach arguably leads to faulty conclusions about repartnering dynamics because the long-term single who have been unpartnered for several years or possibly decades are unlikely to ever repartner. Meanwhile, individuals most likely to repartner are underrepresented in the sample because they have already formed a new union. Other aspects of the marital biography, including the duration of the dissolved marriage and whether it was a first marriage or a remarriage, could be consequential for repartnering behaviors but have been overlooked in the scholarship on union formation and repartnering in later life.

The Decision to Remarry Versus Cohabit

Nowadays, older adults who form unions are increasingly choosing to cohabit instead of remarry. Cohabitation offers older adults many of the benefits of marriage without the constraints (Brown et al. 2012; Chevan 1996; Hatch 1995). Cohabitation offers a more flexible living arrangement in which partners can negotiate their own norms and expectations for interdependence, caregiving, and resource pooling. Cohabitation may be particularly appealing to older women who want to avoid the gendered caregiving obligations of marriage. It is also a way in which partners are able to maintain their financial autonomy, protecting their assets for themselves and their offspring. Couples who remarry become entwined economically and assume legal responsibility for the spouse’s medical and other debts. Remarriage frequently involves relinquishing a pension received from a previous spouse, whereas cohabitation enables individuals to continue to receive those benefits. Cohabiting women receiving a pension are less likely to marry than their counterparts who do not have this resource (Vespa 2013), suggesting that cohabitation is a desirable alternative to marriage for women who are already financially secure. Indeed, cohabiting unions appear to operate similarly to marriage for older adults (King and Scott 2005). Older cohabitors and married individuals report comparable relationship quality (Brown and Kawamura 2010). Also, cohabiting couples are unlikely to either get married or separate, instead staying together until one partner dies (Brown et al. 2012; Vespa 2013).

The Present Study

The incidence of repartnering in later life is rising, reflecting the growth in divorce and the emergence of cohabitation as a viable alternative to remarriage (Schimmele and Wu 2016). Despite the empirical evidence that later-life repartnering is increasingly common and expected to grow for the foreseeable future (Schimmele and Wu 2016), it remains poorly understood. Only a handful of studies have investigated the determinants and timing of later-life repartnering, and the results to date are inconsistent. Divorced individuals are especially likely to repartner, and their numbers are increasing. According to one estimate, the number of adults who experience a gray divorce will climb by upwards of 33 % over the next two decades, assuming a stable gray divorce rate (Brown and Lin 2012). This increase foregrounds the importance of examining repartnership pathways following gray divorce.

Our goal is to examine repartnering following gray divorce. In particular, we estimate the overall levels of repartnering for women and men as well as whether they form remarriages versus cohabiting unions. We anticipate that cohabiting unions are as common as remarriages given the rapid acceleration in later-life cohabitation (Brown and Wright 2017) and the preference of divorced individuals to form cohabiting unions rather than remarriages (Bumpass et al. 1991).

Drawing on the existing literature about union formation in later life, we investigate how resources and constraints are related to transitions into either cohabitation or remarriage versus remaining single. Economic resources, such as education, employment, wealth, homeownership, receipt of a pension, and having health insurance, are expected to be positively associated with women’s and men’s repartnership. Women who are financially autonomous may be especially likely to choose cohabitation over remarriage (Vespa 2013), although the empirical evidence is equivocal (Brown et al. 2012; Vespa 2012).

Health is a resource that is likely to be differentially related to repartnering among women versus men. Poor health appears to be a deterrent to repartnership only among women, but the evidence is mixed (Brown et al. 2012; Vespa 2012). For men, health is unlikely to be associated with either remarriage or cohabitation, although it is possible that better health may be tied to a greater likelihood of forming a remarriage than a cohabiting union as found by Schimmele and Wu (2016) in their examination of union formation after dissolution among Canadians during the second half of life.

Social ties with friends and family may discourage repartnering, particularly among men. Men’s primary source of social support is a spouse or partner, and thus men who experience a gray divorce may be especially likely to repartner when they lack other social ties. Women tend to enjoy greater social ties, which could account for some of their reluctance to repartner (McWilliams and Barrett 2014). Social support should play a larger role in men’s than women’s repartnering, and it is expected to be more closely associated with remarriage than cohabitation. Older cohabitors exhibit fewer social ties than their remarried counterparts (Brown et al. 2006). The presence of a resident child, particularly among women, is likely to be negatively associated with forming a new union (Brown et al. 2018; Schimmele and Wu 2016).

Our examination of women’s and men’s repartnering following a gray divorce includes adjustments for the constraints they face, including marital biography and demographic characteristics associated with remarriage and cohabitation. The marital biography encompasses time since divorce, duration of prior marriage, age at divorce, and marriage order. Repartnering tends to occur quickly after dissolution, if it occurs at all (Brown et al. 2018; Schimmele and Wu 2016). Age at the time of gray divorce is negatively associated with repartnering: younger individuals are more likely to repartner than older individuals, especially among women (Brown et al. 2018; Schimmele and Wu 2016). Those who ended a remarriage are more likely to repartner than those whose first marriage ended, at least among women (Brown et al. 2018). Repartnering is less common among blacks and Hispanics than whites, and this racial difference is larger for remarriage than cohabitation (McNamee and Raley 2011). We include a control for Baby Boom cohort (born 1946–1964) membership: repartnering could be less frequent for Boomers than older cohorts given that Boomers are more often unmarried than were earlier generations (Lin and Brown 2012). Also, the propensity for Baby Boomers to cohabit rather than remarry has likely increased given the rise in supportive attitudes toward cohabitation that has occurred among this group (Brown and Wright 2016).

Data

Data came from the 1998–2014 Health and Retirement Study (HRS). Beginning in 1998, the HRS surveyed a nationally representative sample of 33,855 persons over the age of 50 (i.e., born 1947 or earlier) who have been interviewed biennially. Refresher samples aged 51–56 were added in 2004 (early Baby Boomers) and 2010 (middle Baby Boomers), respectively, to maintain the sample’s representativeness. The HRS is ideally suited to our study because it is sufficiently large to prospectively identify a sizable number of individuals who experience gray divorce. No other U.S. data set yields an adequate sample size to examine repartnering among gray divorced individuals.

We created a marital history file to track the marriages formed and dissolved by all respondents. The HRS also includes information on cohabitation status at each interview, which allowed us to track cohabitation experiences among HRS respondents. Of the 26,361 respondents aged 50 and older who reported being married, a total of 1,222 experienced a divorce at age 50 or older at the time of their first interview or after the respondent entered the 1998 HRS, allowing us to avoid left-censoring bias. For the 51 respondents who divorced more than once after age 50, we used the information from their first gray divorce. We removed the 6 respondents who were in a same-sex union and 85 respondents with a sampling weight of 0 from the analytic sample. The final analytic sample included 1,131 gray divorced individuals, of whom 352 repartnered by 2014. Of those who repartnered, 165 (67 women and 98 men) formed a remarriage, and 187 (58 women and 129 men) formed a cohabiting union.

Measures

Economic Resources

Education was measured as years of schooling completed. Employment was a time-varying covariate that measured whether the respondent engaged in full or part-time employment (0 = No, 1 = Yes). Homeownership was a time-varying covariate that measured whether the respondent owned a home (0 = No, 1 = Yes). Wealth was captured by a time-varying measure of the logged value of the respondent’s nonhousing household wealth. All wealth values were adjusted to real 2017 dollars to account for inflation. Those with wealth beyond the 99th percentile were top-coded to the 99th percentile. Respondents reporting negative wealth were assigned a wealth value of $1 prior to taking the natural log. Pension was a time-varying covariate that captured whether the respondent currently received or planned to receive pension income in the future (0 = No, 1 = Yes). Health insurance was a time-varying covariate that indicated whether the respondent had health insurance coverage at any given interview wave (0 = No, 1 = Yes). Given the requirement of the Affordable Care Act (ACA) that all Americans have health insurance, it is likely that in the earlier years, this measure of health insurance was an indicator of resources; but most recently, it also could be viewed as signaling compliance.

Health

Number of chronic conditions was a time-varying measure that summed whether a doctor had diagnosed the respondent with a psychiatric disorder, heart disease, hypertension, cancer, lung disease, arthritis, diabetes, or a stroke. Values could range from 0 to 8. Number of ADL/IADL difficulties was a time-varying covariate that summed the number of reported difficulties with ADLs and instrumental activities of daily living (IADLs). ADLs gauged difficulty walking, dressing, getting in and out of bed, bathing, eating, and toileting. IADL items tapped into difficulty preparing hot meals, shopping for groceries, making phone calls, taking medications, managing money, and using a map. ADL/IADL difficulties could range from 0 to 12.

Social Ties

Relatives or friends nearby was a time-varying measure that captured whether respondents had either relatives or friends living in close proximity (0 = No, 1 = Yes). Resident child was a time-varying dummy variable indicating whether any child of the respondent lived in the household (0 = No, 1 = Yes).

Marital Biography

Years since divorce was a time-varying covariate that accounted for the number of years since the respondent’s gray divorce occurred. Duration of prior marriage was a time-invariant covariate that measured the number of years spent in the prior marriage. Age at dissolution was a time-invariant measure that captured the respondent’s age at the time of gray divorce. Dissolved a remarriage was a time-invariant covariate that represented whether the respondent was in a higher-order marriage (0 = No, 1 = Yes) at the time of gray divorce.

Demographic Characteristics.

Race/ethnicity was a binary variable that indicated whether the respondent reported being white (coded 0) or nonwhite (coded 1). Baby Boomer cohort distinguished respondents born 1946–1964 (coded 1) from older cohorts (coded 0).

Analytic Strategy

Our first step was to estimate survival probabilities for women and men by years since divorce to examine the levels and timing of forming either a remarriage or a cohabiting union following a gray divorce. Because repartnering typically occurs soon after divorce, individuals were censored after 10 years at risk of repartnering. Next, we calculated baseline bivariate statistics according to transition type separately for women and men to evaluate how individuals compared on the study variables depending on whether they ultimately formed a remarriage, formed a cohabiting union, or remained unpartnered. Last, we performed multinomial logistic regression analyses using discrete-time event history models to predict the roles of resources and constraints on the relative likelihoods of remarriage or cohabitation versus remaining unpartnered following gray divorce.

Discrete-time event history models are advantageous because they allow us to take into account two possible ways of exiting from the divorced state: through cohabitation or remarriage. The hazard rate for competing risks can be defined as P_tj = Pr(T = t, J = j|T ≥ t). The hazard of exiting via means J (= 1, …, m) at time T is the probability of doing so given survival up to at least time T. The model is specified as follows (Allison 1982):

$$P_{tj}=\text{exp}[{\alpha }_{jt}+{\beta }_j^{\prime }{x}_t]/(1+\sum _{j=1}^m\text{exp}[{\alpha }_{jt}+{\beta }_j^{\prime }{x}_t])$$

We consider how this hazard rate is a function of time (α_jt) and a vector of explanatory variables (x_t, with its coefficient vector β_j). This model allows explanatory variables to have different associations with the transitions into cohabitation or remarriage. Individual observations were reshaped into person-wave format. We coded the dependent variable 0 if the individual remained single, 1 if remarried, and 2 if cohabited during that person-wave. Multinomial logistic regression was used to estimate the risks of transitioning from the divorced state through either cohabitation or remarriage relative to remaining divorced. Individuals were censored after they cohabited or remarried, after 10 years at risk of repartnering, or at the 2014 interview. Censoring respondents after 10 years at risk of repartnering aligns with the approach taken in prior research. For example, in their study of later-life repartnering, Schimmele and Wu (2016) censored respondents after 10 years at risk.

Our full model estimated the associations between resources (economic factors, health, and social ties) and constraints (the marital biography and demographic characteristics) with the likelihood of forming a remarriage versus a cohabiting union (the reference category was remained unpartnered). Models were estimated separately for women and men because the two groups face a lopsided sex ratio in the repartnering market and exhibit different preferences for repartnering. Research to date on repartnering in later life has consistently relied on gender-specific models (Brown et al. 2012, 2018; Schimmele and Wu 2016; Vespa 2012; Wu et al. 2014).

Missing data were modest, ranging from 0.01 % for gender to 6 % for resident child, with the exception of homeownership (15 %), relatives or friends nearby (22 %), and pension (34 %). We performed multiple imputation using chained equations (MICE) using the mi impute chained command in Stata, which imputed missing values for a given variable as a function of other covariates and the dependent variable in the model (Raghunathan et al. 2001; van Buuren et al. 1999). The results were based on 20 random, multiple-imputed replicates. The bivariate and multivariate analyses were adjusted using the baseline weights (Ofstedal et al. 2011).

Results

Figures 1 and 2 show the cumulative probabilities of forming either a remarriage or a cohabitation for women and men, respectively, by years since gray divorce. The probabilities of repartnering were considerably lower for women than men, as expected. Repartnering also proceeded at a slower pace for women than men. Relative to women, the transition to either remarriage or cohabitation among men tended to occur more quickly, and the rise was sustained for a longer period, especially for cohabitation. Women’s probabilities of repartnering through remarriage and cohabitation were largely similar to each other, although during the first five or so years following divorce, women were more likely to form a cohabiting union than to remarry. Thereafter, women’s probability of remarrying slightly exceeded that of cohabiting. In contrast, men’s repartnering probabilities were higher for cohabitation than remarriage, and this difference grew with time since gray divorce. Men’s probability of repartnering rose during the first eight years following gray divorce and then began to plateau.

Fig. 1.

Cumulative probability of repartnering for women. N = 534. Source: Health and Retirement Study, 1998–2014.

Fig. 2.

Cumulative probability of repartnering for men. N = 597. Source: Health and Retirement Study, 1998–2014.

Tables 1 and 2 depict the weighted means or percentages (as appropriate) of all variables at baseline according to the type of repartnering transition, for women and men, respectively. Women’s economic resources were largely comparable regardless of whether they remarried, cohabited, or remained unpartnered. Compared with those who did not repartner, women who remarried were marginally (p < .10) more often employed (61.0 % vs. 47.0 %). The logged mean values of wealth were significantly higher among women who formed either a remarriage (10.23) or a cohabiting union (10.09) versus remained unpartnered (7.91). No other economic differences emerged across the three transition types, contrary to our expectations.

Table 1.

Weighted means and percentages by repartnering transition type for women

	Remarried		Cohabiting		Unpartnered
Economic Resources
Years of education	13.1		13.4		13.0
Employed	61.0	†	54.2		47.0
Owns home	68.5		52.1		54.4
Wealth (logged)	10.23	**	10.09	*	7.91
Pension	64.7		61.9		58.0
Has health insurance	86.0		90.6		81.8
Health
Number of chronic conditions	2.00		1.87	*	2.18
Number of ADL/IADL difficulties	0.63	**	0.63	***	1.43
Social Ties
Relatives or friends nearby	63.8		50.4		60.6
Resident child	35.7		31.9		38.4
Marital Biography
Years since divorce	4.3	**	2.9	***	5.5
Duration of prior marriage	17.6	*	23.8		21.5
Age at dissolution	59.2		57.7	†	58.9
Dissolved a remarriage	70.2		52.5		63.2
Demographic Characteristics
Nonwhite	24.1		16.6		25.0
Baby Boomer	37.7		36.1		51.2
Unweighted N	67		58		409
Weighted %	11.7		9.9		78.3

Notes: Data are from the Health and Retirement Study, 1998–2014. Covariates are from baseline waves. Significance levels indicate statistical differences between remarried and unpartnered respondents or between cohabitors and those who are unpartnered. Bold numbers indicate that remarried respondents significantly differ from cohabitors at p < .10.

^†p < .10;

^*p < .05;

^**p < .01;

^***p < .001

Table 2.

Weighted means and percentages by repartnering transition type for men

	Remarried		Cohabiting		Unpartnered
Economic Resources
Years of education	13.5	†	13.3		12.6
Employed	62.5	†	69.1	**	48.5
Owns home	60.8		49.1		50.9
Wealth (logged)	9.66	†	9.75	†	8.18
Pension	69.6		72.9		60.5
Has health insurance	90.1		82.9		85.0
Health
Number of chronic conditions	1.46	*	1.84	†	2.30
Number of ADL/IADL difficulties	0.31	***	0.52	**	1.15
Social Ties
Relatives or friends nearby	66.1		49.3	**	66.2
Resident child	28.8		18.7	†	26.9
Marital biography
Years since divorce	4.2		2.8	***	4.4
Duration of prior marriage	21.7		21.8		20.0
Age at dissolution	59.7		58.0	**	60.5
Dissolved a remarriage	58.4		61.3		61.2
Demographic Characteristics
Nonwhite	24.5		18.4	†	26.8
Baby Boomer	32.9	**	45.6		54.3
Unweighted N	98		129		370
Weighted %	14.8		22.4		62.8

Notes: Data are from the Health and Retirement Study, 1998–2014. Covariates are from baseline waves. Significance levels indicate statistical differences between remarried and unpartnered respondents or between cohabitors and those who are unpartnered. Bold numbers indicate that remarried respondents significantly differ from cohabitors at p < .10.

^†p < .10;

^*p < .05;

^**p < .01;

^***p < .001

Women who formed a cohabiting union reported significantly fewer chronic conditions (1.87) than women who remained unpartnered (2.18), but these two groups did not significantly differ from women who remarried (2.00). The number of ADL/IADL difficulties was significantly lower for women who formed a remarriage (0.63) or a cohabitation (0.63) relative to those who did not repartner (1.43). No significant variation in social ties among women was found across relationship transition type, inconsistent with our expectations.

The number of years since gray divorce was negatively associated with repartnering. The mean number of years was 4.3 for those who remarried and 2.9 for those who cohabited, versus 5.5 years for women who did not repartner. The time to remarriage was significantly longer, on average, than the time to cohabitation. Women who remarried had been in marriages of shorter average duration prior to gray divorce (17.6 years) than either those who formed a cohabiting union (23.8 years) or those who formed no union at all (21.5 years). Women who eventually formed a cohabiting union (57.7 years) were marginally (p < .10) younger, on average, than those who remained unpartnered (58.9 years) but did not significantly differ from those who remarried (59.2 years). Women who remarried had more often dissolved a remarriage (70.2 %) than women who formed a cohabiting union (52.5 %), but both groups were comparable with women who remained unpartnered (63.2 %). Demographic characteristics were similar for women across transition types. The racial/ethnic composition and cohort membership of women did not differ according to whether they remarried, cohabited, or remained unpartnered.

Turning now to men, Table 2 shows that those who remarried or cohabited differed on some of the economic resource indicators from those who remained unpartnered. The average level of education among men who remarried (13.5 years) was marginally (p < .10) higher than for men who remained unpartnered (12.6 years). The two groups did not differ from men who formed a cohabiting union (13.3 years). Employment levels were marginally higher among those who formed either a remarriage (62.5 %; p < .10) or a cohabiting union (69.1 %) than among those who stayed unpartnered (48.5 %). Men’s logged wealth was marginally significantly higher for those who either remarried (9.66, p < .10) or formed a cohabiting union (9.75, p < .10) than those who did not repartner (8.18). The pension and health insurance status of men was comparable across the three transition types.

The average number of chronic conditions was lowest for men who entered a remarriage (1.46), followed by men who formed a cohabitation (1.84) and then by men who did not repartner (2.30). Men who remarried (0.31) or cohabited (0.52) reported fewer ADL/IADLs than men who remained unpartnered (1.15). Social ties with relatives or friends nearby was related to men’s repartnering through cohabitation but not to remarriage compared with remaining unpartnered. Just 49.3 % of men who formed a cohabiting union reported having relatives or friends nearby compared with 66.1 % of men who remarried and 66.2 % of men who stayed single. The proportion of men with a resident child was marginally lower among men who formed a cohabiting union (18.7 %, p < .10) than among men who formed no union at all (26.9 %), but neither group statistically differed from men who remarried (28.8 %).

For men, the transition to either remarriage or cohabitation occurred roughly 4.2 and 2.8 years, respectively, following gray divorce. Transitions to cohabitation occurred more quickly, on average, than transitions to remarriage. Men who transitioned to cohabitation (58.0 years of age) were approximately two years younger, on average, than men who remained unpartnered (60.5 years), but both were similar to men who remarried (59.7 years). Men did not differ by transition type on either duration of the marriage that ended through gray divorce or by whether that marriage was of higher order. Men who formed a cohabiting union (18.4 %) were marginally (p < .10) less likely to be nonwhite than were men who remained unpartnered (26.8 %). Roughly 24.5 % of men who remarried were nonwhite. Baby Boomer men were underrepresented among those who remarried (32.9 %) compared with those who stayed single (54.3 %). Roughly 45.6 % of men who cohabited were Baby Boomers.

Tables 3 and 4 depict the relative risk ratios from the multinomial logistic regressions predicting the likelihood of forming either a remarriage or cohabitation following gray divorce, treating these two partnership types as competing risks. As shown in Table 3, for women, few indicators of economic resources were related to repartnering. Wealth was linked to a marginally (p < .10) higher risk of forming a remarriage compared with staying single. This positive association between wealth and remarriage echoed Vespa’s (2012) results for later-life union formation among previously married older adults. Also in line with Vespa (2012), wealth was not related to the type of union formed among those women who repartnered. However, contrary to Vespa’s (2012) results, wealth was not related to women’s formation of a cohabiting union. The absence of a significant association between wealth and cohabitation aligned with the findings from Brown et al. (2012).

Table 3.

Relative risk ratios from discrete-time multinomial logistic regression of repartnering through remarriage versus cohabitation for women (unweighted N =534)

	Remarried vs. Unpartnered		Cohabiting vs. Unpartnered		Cohabiting vs. Remarried
Economic Resources
Years of education	0.93		0.98		1.05
Employed	0.94		0.99		1.05
Owns home	2.04		0.90		0.44
Wealth (logged)	1.14	†	1.10		0.97
Pension	1.57		0.95		0.60
Has health insurance	1.34		1.02		0.77
Health
Number of chronic conditions	1.13		1.13		1.00
Number of ADL/IADL difficulties	0.82	†	0.89		1.09
Social Ties
Relatives or friends nearby	0.83		0.63		0.76
Resident child	0.68		0.64		0.94
Marital Biography
Years since divorce	1.06		0.83	***	0.78	**
Duration of prior marriage	0.98		1.01		1.03
Age at dissolution	1.02		0.95	*	0.93	†
Dissolved a remarriage	0.89		0.98		1.10
Demographic Characteristics
Nonwhite	1.39		0.93		0.66
Baby Boomer	1.39		0.62		0.45
Constant	0.00	**	1.07		186.97
Model Fit Statistics			F(32,49.5) = 5.95

Note: Data are from the Health and Retirement Study, 1998–2014.

^†p < .10;

^*p < .05;

^**p < .01;

^***p < .001

Table 4.

Relative risk ratios from discrete-time multinomial logistic regression of repartnering through remarriage versus cohabitation for men (unweighted N = 597)

	Remarried vs. Unpartnered		Cohabiting vs. Unpartnered		Cohabiting vs. Remarried
Economic Resources
Years of education	1.07		1.02		0.95
Employed	1.45		1.90	*	1.31
Owns home	1.96	*	1.36		0.69
Wealth (logged)	0.98		1.05		1.07
Pension	1.42		1.21		0.85
Has health insurance	1.24		0.83		0.67
Health
Number of chronic conditions	0.85	†	1.12		1.32	*
Number of ADL/IADL difficulties	0.86		1.03		1.20	†
Social Ties
Relatives or friends nearby	0.78		0.36	***	0.46	*
Resident child	0.90		0.47	*	0.52
Marital Biography
Years since divorce	1.15	*	0.94		0.81	**
Duration of prior marriage	1.02		1.04	*	1.02
Age at dissolution	1.01		0.95	*	0.94	†
Dissolved a remarriage	1.75		2.69	**	1.54
Demographic Characteristics
Nonwhite	1.23		0.84		0.68
Baby Boomer	0.76		0.85		1.13
Constant	0.00	**	0.33		93.72
Model Fit Statistics			F(32,48.5) = 6.40

Note: Data are from the Health and Retirement Study, 1998–2014.

^†p < .10;

^*p < .05;

^**p < .01;

^***p < .001

Health was only modestly associated with women’s repartnering. The number of ADL/IADLs was marginally (p < .10) negatively associated with women’s risk of remarriage but did not deter cohabitation relative to remaining unpartnered. Chronic conditions were unrelated to women’s repartnering. Contrary to our expectations, social ties—whether having friends and relatives nearby or having a resident child—were not associated with women’s repartnering following gray divorce.

The longer the time since gray divorce, the less likely it was that women would cohabit compared with either remaining unpartnered or remarrying. Younger women were more likely to repartner through cohabitation. And given that they formed a union, younger women were marginally (p < .10) more likely to cohabit than to remarry. These findings underscore the salience of the marital biography in constraining women’s repartnering behaviors. Further evidence is gleaned through significant interactions between age at gray divorce and time since gray divorce (results not shown). As time elapsed since gray divorce, older women were marginally less likely to remarry (RRR = 0.99, p = .06) and more likely to cohabit (RRR = 1.01, p = .04) relative to younger women. Immediately following gray divorce, women were more likely to cohabit than to remarry regardless of age. As time passed, the crossover in the likelihood of cohabiting versus remarrying (as illustrated for women regardless of age in Fig. 1), occurred more quickly for older than younger women. Additionally, the risk of transitioning into remarriage remained flat across time for older women, whereas it increased for younger women. Demographic characteristics were not associated with the risk of repartnering among women, net of other factors.

As shown in Table 4, relatively few indicators of economic resources were related to men’s repartnering following a gray divorce. Men’s employment was positively associated with the risk of cohabitation versus remaining unpartnered. Homeownership among men was related to a higher risk of remarriage than remaining single. Contrary to our expectations, wealth, a pension, and health insurance were not associated with men’s repartnering. These findings are at odds with those of Vespa (2012) and instead are consistent with the results of Brown et al.’s (2012) analyses of later-life union formation.

The number of chronic conditions reported by men was negatively associated (p < .10) with the risk of remarriage relative to remaining single. Also, among men who formed a union, those with more chronic conditions were more likely to cohabit than to remarry. Likewise, the number of ADL/IADLs was positively related to entering cohabitation versus remarriage among men who repartnered. The bivariate pattern between social ties and repartnering persisted in the multivariate model. Men with relatives or friends nearby were less likely to form a cohabiting union than either remain unpartnered or remarry. Social ties were unrelated to men’s risk of remarriage compared with remaining single. The presence of a resident child was associated with a reduced risk of cohabiting compared with remaining single but was not related to the likelihood of remarriage, in line with Vespa’s (2012) findings.

The relative risk of forming a union following a gray divorce did not diminish over time for men. In fact, men’s likelihood of forming a remarriage relative to either a cohabiting union or remaining single increased over time. Men’s marital duration prior to gray divorce was positively associated with forming a cohabiting union versus staying single. Younger men were more likely to form a cohabiting union than to remain single or form a remarriage. Dissolving a remarriage (vs. a first marriage) was associated with greater risk of forming a cohabiting union than not repartnering. Men’s demographic characteristics were unrelated to repartnering.

Discussion

The doubling of the gray divorce rate in the United States over the past few decades portends growth in later-life repartnering, yet little is known about the mechanisms undergirding decisions to form either a remarriage or a cohabiting union after divorce during the second half of life. The existing research on union formation among older adults has yielded mixed findings, probably in part because these studies have potentially significant methodological weaknesses. Some of the prior research relied on retrospective cross-sectional data, limiting the set of predictors available (de Jong Gierveld 2004; Schimmele and Wu 2016). Other studies used longitudinal data but faced extreme left-censoring bias because by initiating the risk period at a given point in time (e.g., 1998 when the HRS began), the analytic samples were primarily populated by long-term singles who already had been at risk of forming a union for more than 15 years, on average (Brown et al. 2012; Vespa 2012). Using data from the 1998–2014 HRS, our event history analyses involved prospectively following individuals for up to a decade beginning at the time of gray divorce, which allowed us to avoid the biases characterizing prior research, including left-censoring. At the same time, our study sheds new light on repartnering after a later-life divorce, a rapidly growing phenomenon that cross-sectional research has linked to relatively high levels of repartnering (Brown et al. 2018).

Our study is the first to document that older adults more often repartner through cohabitation than remarriage. We note several possible explanations for this difference. First, unlike prior work on union formation in later life, we were able to observe unions formed immediately following gray divorce when cohabitation might be most likely. Second, our study encompassed more recent cohorts of older adults, including Baby Boomers, who tend to favor cohabitation but were excluded in earlier studies. Third, divorced individuals often prefer cohabitation over remarriage because they are disillusioned by the institution of marriage but still want an intimate, coresidential union (Bumpass et al. 1991). Finally, cohabitation offers numerous benefits for older adults compared with remarriage (Brown et al. 2012; Chevan 1996) and is now more accepted as a living arrangement among older adults (Brown and Wright 2016), making it a viable or even preferable alternative to remarriage.

Guided by the existing literature on later-life union formation emphasizing the twin roles of resources and constraints (Brown et al. 2012; Schimmele and Wu 2016; Vespa 2012), our study investigated whether economic indicators, health, and social ties were related to the likelihood of remarrying or cohabiting, accounting for marital biography and demographic characteristics. We anticipated that for women and men alike, economic resources would encourage repartnering. Wealth was marginally positively associated with the formation of a remarriage among women but had no appreciable effects on repartnering for men. The absence of an association between wealth and repartnering for men was in line with prior work on repartnering in both the United States (Brown et al. 2012) and Canada (Schimmele and Wu 2016) but is at odds with Vespa’s (2012) work. However, we found that homeownership was positively associated with remarriage among men, which echoed Vespa’s (2012) finding that housing wealth was the sole wealth predictor of remarriage among older men. Ultimately, for men and women alike, economic factors play only a modest role in the repartnering process after a gray divorce.

Our study also revealed few associations between health and repartnering for either women or men. Among women, the number of ADL/IADLs was marginally negatively associated with remarriage. Among men, the number of chronic conditions was marginally negatively related to entry into remarriage, mirroring results established using a Canadian sample (Schimmele and Wu 2016). Also, men in poorer health were more likely to form cohabiting than marital unions. Prior research has tended to show that health matters for women’s repartnering (Brown et al. 2018; Schimmele and Wu 2016; Vespa 2012), although at least one prior study found no association (Brown et al. 2012).

The linkages between social ties and repartnering were also weak. We anticipated that having friends and relatives nearby or living with a resident child would diminish the odds of repartnering. Among women, these indicators of social ties were unrelated to repartnering. Among men, lacking relatives or friends nearby was positively related to forming a cohabiting union but was unrelated to remarriage, partially supporting prior research showing that social support diminished widowers’ interest in dating (Carr 2004). Men with a resident child were unlikely to cohabit but no less likely to remarry than those without a child living with them.

Constraints captured by the marital biography were linked to repartnering among women and men alike. Among women, time since divorce was negatively related to forming a cohabiting union; among men, time since divorce was positively associated with forming a remarriage. Younger men and men who had been in longer marriages or remarriages were especially likely to form a cohabiting union. The age differences in the effects of time since gray divorce on repartnering for women signaled that younger women were decreasingly likely to cohabit and increasingly likely to remarry as time since divorce elapsed. For older women, entry into cohabitation declined with time, whereas the risk of remarrying remained relatively stable. Thus, women tended to favor cohabitation in the short term, but those who waited longer to form a union were especially likely to remarry.

Our analytic approach was consistent with that of prior studies and treated remarriage and cohabitation as competing risks. Yet, resources and constraints were sparsely associated with the decision to remarry versus cohabit among those who formed a union (as shown in the third column in Tables 3 and 4), particularly among women. In other words, conditional on forming a union, resources and constraints did not appreciably discriminate according to union type, which lends additional empirical support to the notion that cohabitation and remarriage operate similarly during the second half of life (Brown et al. 2012; King and Scott 2005).

The apparent similarity between cohabitation and remarriage is just one example of how the partnering process is distinctive in later life. For younger adults, the bar for marriage is substantially higher than for cohabitation. Young people will often cohabit until they can afford to marry (Smock et al. 2005). For older adults, marriage and cohabitation are largely interchangeable (Brown and Wright 2017; King and Scott 2005). Our study also demonstrates that economic resources, which are pivotal to union formation for young adults (Clarkberg 1999; Oppenheimer 1994, 2003), are not appreciably related to repartnering among older adults who experience divorce. Perhaps the availability of Social Security for many older adults diminishes the centrality of traditional indicators of economic resources. But other factors are tied to repartnering in later life. Social ties are salient for men, underscoring the unique role of intimate partnerships for men’s health and well-being in older adulthood (Wright and Brown 2017). In contrast, health is not closely tied to women’s repartnering net of other factors (although in the bivariate, healthier women are more likely to repartner). Finally, constraints matter for older adults. The features of the marital biography—such as time since divorce, duration of marriage prior to divorce, marriage order, and age at divorce—have enduring influences on decisions about repartnering. The centrality of the marital biography for repartnering affirms the value of our focus on individuals who experience gray divorce because dissolution pathway itself is a notable feature of the marital biography (Hughes and Waite 2009). The repartnering behaviors of gray divorced women and men are distinctive from those of other single older adults.

By using prospective, longitudinal data to begin to decipher the mechanisms behind women’s and men’s decisions to repartner following a gray divorce, our research fills a key gap in the literature on later-life repartnering. But it also has some weaknesses. Attrition is always a concern in longitudinal analyses, and it is possible that repartnering was related to dropping out of the study, which could bias our results by leading us to undercount union formation. The HRS does pick up respondents who initially dropped out of the sample, and they were included in our analyses. Unmeasured factors, such as the desire to repartner, are likely driving decisions to cohabit or remarry in later life, but the HRS does not include indicators of interest in or preferences for forming a union. Finally, the HRS permitted us to examine only coresidential repartnerships, but living apart together (LAT) and dating relationships are also on the rise among older adults and may be attractive alternatives to cohabitation and remarriage (Brown and Shinohara 2013; Connidis et al. 2017).

Some women and even more men repartner after gray divorce through either remarriage or cohabitation. Cohabitation seems to be the preferred union type, particularly among men. Economic resources, health, and social ties are related to repartnering, but their effects unfold against a backdrop of structural constraints, including the lopsided sex ratio and gender differences in desires to repartner. Features of the marital biography figure into women’s and men’s repartnering behaviors, signaling the centrality of the circumstances of the marriage ending in gray divorce for subsequent repartnering. The rising gray divorce rate underscores the importance of sustained research attention on older women and men’s repartnering patterns as well as the ramifications of repartnering versus remaining single for health and well-being.