Marital separation, reconciliation, and repartnering in later life

Abstract

Objective:

The growth in gray divorce raises new questions about the marital dissolution process experienced by older adults. Our goal was to assess patterns of reconciliation among couples following marital separation, treating forming a union with a new partner as a competing risk.

Background:

Repartnering after a gray divorce is common, particularly among men. However, the extent to which older adults reconcile with their spouses is unknown. In line with the few prior studies on marital reconciliation among younger people, we anticipated that spouses with fewer resources and more marital-specific capital would be more likely to reconcile.

Method:

Using the 1998–2018 Health and Retirement Study, we tracked women and men who experienced a marital separation after age 50 to evaluate their propensities to reconcile with their spouse versus form a coresidential union (i.e., cohabitation or remarriage) with a new partner relative to remaining separated.

Results:

Roughly 7% of women and 11% of men reconciled with their spouses, whereas 12% of women and 26% of men instead formed unions with a new partner within 10 years of marital separation. We expected that having fewer resources and greater relationship-specific investments would encourage reconciliation, but results were mixed for women and men alike. Resources did tend to be positively associated with repartnering, particularly for men.

Conclusion:

Our study contributes to the emerging research on repartnering after late-life divorce as well as the limited literature on marital reconciliation by underscoring the utility of examining both reconciliation and repartnering as potential outcomes following marital separation.

INTRODUCTION

The gray divorce rate, which describes divorces that occur among adults aged 50 and older, has doubled since 1990 in the United States (Brown & Lin, 2022). This recent increase coupled with the aging of the married population has contributed to a dramatic shift in the age composition of US adults experiencing divorce. Whereas fewer than 10% of persons divorcing in 1990 were aged 50 or older, by 2019 the share stood at 36% (Brown & Lin, 2022).

Although recent work indicates that many gray divorced individuals go on to repartner through either cohabitation or remarriage (Brown et al., 2018, 2019), patterns of reconciliation among couples who separate (or divorce) during the second half of life are unknown. In fact, marital reconciliation is poorly understood across all stages of the adult life course with only limited attention among young adults (Binstock & Thornton, 2003) and women of childbearing age (Bumpass & Raley, 2007; Tumin et al., 2015). The recent acceleration in late-life divorce points to the importance of examining patterns of marital reconciliation during the second half of life.

Using data from the 1998–2018 Health and Retirement Study (HRS), we explore reconciliation patterns following a late-life marital separation (or divorce) to estimate the prevalence of reunification with one’s spouse. We contribute to the literature in at least three notable ways. First, we focus on marital separations that occur during the second half of life among adults aged 50 and older, many of whom are exiting long-term marriages marked by significant marital-specific capital that could encourage reconciliation. Second, we conceptualize reconciliation as just one of two possible options: separated individuals can reconcile with their spouse, but they also can form a union with a new partner. A key shortcoming of the limited prior research on the predictors of marital reconciliation is that repartnering with someone new has been ignored. Our study treats forming a union with a new partner as a competing pathway. In a sense, marital reconciliation can be conceptualized as a specific form of repartnering. Our approach is to perform event-history analyses that treat reconciliation (with one’s spouse) and repartnering (with a new partner) as competing risks relative to remaining single. Third, unlike prior work on the antecedents of reconciliation which has been restricted to women, we explicitly include both women and men to assess whether the patterns and predictors of reconciliation and repartnering differ by gender. Our framework draws on the limited literature on marital reconciliation (Binstock & Thornton, 2003; Tumin et al., 2015; Wineberg, 1994, 1996) and the emerging work on repartnering following late-life divorce (Brown et al., 2019; Schimmele & Wu, 2016). The findings from our study offer new insights on the marital dissolution process among older adults.

BACKGROUND

Divorce is widely recognized as a process that unfolds over time (Wang & Amato, 2000). Uncoupling between spouses involves many more facets than merely the legal termination of the marriage (Kitson, 1992). Marital separation ruptures the emotional, economic, residential, and parenting ties between spouses. The marital bond is comparable to the parent–child bond in terms of its intensity and centrality to one’s identity, which is why the separation distress entailed in divorce tends to adversely affect both spouses regardless of who initiated the divorce (Weiss, 1976). Adjustment to marital disruption appears to be especially protracted among older adults relative to younger adults (Wang & Amato, 2000).

A recent theoretical treatise on adjustment following gray divorce likened the experience to an extended convalescence with individuals taking upwards of 4 years to recover (Lin & Brown, 2020). This pattern is distinctive from the entrenched conceptual models that frame adjustment as either a short-term crisis period or an interminable chronic strain (Amato, 2000). Unlike the crisis period model, the prevailing pattern of divorce adjustment for older adults is more protracted. But, contrary to the chronic strain model, older adults eventually bounce back. Empirical support for the convalescence model was evident in the trajectories of depressive symptoms experienced by older adults following divorce (Lin et al., 2019). Likewise, in-depth interviews with older adults, who recently got divorced aligned with the convalescence model, revealing that women and men struggled with loneliness, isolation, and financial difficulties after divorce but had reached a new equilibrium roughly 3 years postdivorce (Crowley, 2019).

During this convalescence period, some spouses may decide to reconcile with one another rather than forge ahead on separate paths. Others may emerge from the convalescence period relishing their independence and newfound freedom. Many could be eager to explore new relationships (Harris, 2023) that eventuate in cohabitation or remarriage (McWilliams & Barrett, 2014). Cohabitation has skyrocketed among older adults, with the numbers living with an unmarried partner quadrupling since 2000 (Stepler, 2017). Forming a remarriage after a late-life divorce is nearly as common as cohabitation (Brown et al., 2019). Nonetheless, partnership formation patterns in later life remain understudied (Carr & Utz, 2020; Sassler, 2010) and, to the best of our knowledge, attention to marital reconciliation among older adults is nonexistent.

In fact, marital reconciliation is largely overlooked in the vast literature on divorce and repartnering. This is a surprising omission given that estimates of reconciliation are sizeable, occurring in somewhere between 12% (Tumin et al., 2015), 25% (Binstock & Thornton, 2003), or even 44% (Wineberg, 1996) of marital separations. These wide-ranging estimates likely reflect disparate samples and measurement approaches. Although marital separation is a broadly understood term referring to the breakup of a marriage that typically involves a residential and romantic decoupling, it is an event that is often poorly captured in survey research, leading to artificially low estimates of separation that in turn obscure reconciliations (Binstock & Thornton, 2003; Tumin et al., 2015). For instance, some longitudinal surveys only ask about separation in the event of a divorce or if the respondent’s current marital status is separated, overlooking the possibility that an individual might have separated and reconciled with their spouse between interviews. Others ask for a date when a couple “stopped living together” rather than allowing respondents to report “separation” as a marital transition type. These measurement issues have ramifications for the substantive conclusions we draw about patterns of separation and divorce (Bumpass & Raley, 2007). For example, the relative propensities of women of various racialized groups to reconcile with their spouses during marital separation potentially account for some of the well-established racial and ethnic disparities in marital disruption (Tumin et al., 2015). In sum, the lack of scholarly attention to marital reconciliations echoes the myriad challenges associated with measuring marital separation.

A key contribution of the current study is to investigate reconciliation and repartnering in concert, recognizing that after a marital separation, older adults have three options: they can either reunite with their spouse, form a coresidential union with a new partner, or remain unpartnered (i.e., outside of a coresidential union). Research to date on marital reconciliation has ignored the possibility of repartnering as an alternative pathway. In the remainder of this section, we draw on the separate literatures surrounding marital reconciliation among childbearing-aged women on the one hand and late-life repartnering on the other hand, to elucidate the prominence of resources in shaping the prospects for each of these types of transitions. Whereas marital reconciliation tends to occur most often among spouses with fewer resources, late-life repartnering is most common among those with more resources (Brown et al., 2019; Vespa, 2012, 2013). We then join these disparate literatures on reconciliation and repartnering in the subsequent section to explicate our study hypotheses.

The roles of resources and relationship-specific capital in marital reconciliations

The prevailing theoretical framework guiding research on marital reconciliation centers on the role of resources. Individuals with fewer resources are presumably more motivated to reconcile with their spouses because they lack the instrumental support and economic security needed to go it alone (Morgan, 1988; Wineberg, 1996). Financial difficulties during separation may prompt individuals to reconsider their decision to split despite feelings of dissatisfaction with the marriage. Several studies of marital reconciliation have shown that women with lower (vs. higher) levels of education are more likely to reconcile with their husbands (Morgan, 1988; Tumin et al., 2015; Wineberg, 1994, 1996; Wineberg & McCarthy, 1994). Whether the same pattern holds for men is unclear because to our knowledge prior work has exclusively examined the correlates of reconciliation from the perspective of women.

Education is a resource that is primarily indicative of one’s earnings potential but early research on reconciliation also characterized education as a proxy for one’s attitudes toward gender and marriage as well as one’s personality characteristics (Morgan, 1988). It is unclear whether other indicators of resources, such as family income, are related to the likelihood of reconciliation. To our knowledge, the only study to examine family income found it had no association with reconciliation net of education, although the study is now rather dated (Morgan, 1988). Thus, the literature on the role of resources in marital reconciliation has been rather circumscribed with its narrow focus on women’s education. For older adults, other indicators of economic resources, such as wealth and home ownership, may encourage reconciliation given the significant costs of dividing such assets (Lin & Brown, 2021).

For older adults, health can be viewed as a key resource in the aftermath of a marital separation. In the absence of good health, one is more vulnerable which may make independent living less attractive or less feasible. A health decline, particularly when experienced by the wife, is a risk factor for gray divorce (Karraker & Latham, 2015). Women in poorer health may be more motivated to reconcile with their husbands whether because they need caregiving assistance and support or because they lack the financial resources (e.g., health insurance) to manage their own healthcare. Similarly, men who are in worse health also may have greater incentive to reconcile with their wives to receive care.

A corollary to the resources argument is that individuals with greater relationship-specific capital may have more incentive to reconcile (Wineberg, 1996). This line of reasoning is an offshoot of the overarching resources argument in that it calls attention to those individuals who have more resources within the marital relationship than outside of it. Certainly, a key marriage-specific investment is children. Children are not a readily transferable investment; an extensive literature documents that offspring are a deterrent to remarriage, especially among women who more often have primary custody (Di Nallo, 2019; Goldscheider & Sassler, 2006). Even as most Americans support divorce when children are involved (Brewster & Padavic, 2000), nonetheless children and the coparenting relationship likely factor into decisions to separate and eventually divorce. After a period of separation, some couples may decide to reconcile for the sake of their children. Despite the compelling conceptual rationale for the role children might play in decisions to reconcile, we are not aware of any study that has empirically tested this assertion.

Children are only one example of relationship-specific investments that couples make over the course of a marriage. For older adults, children are often grown and out of the household. But many have formed their own families, providing their parents with grandchildren. In line with prior work indicating becoming grandparents deters late-life divorce (Brown et al., 2021), the presence of grandchildren, a form of relationship-specific capital, may encourage reconciliation among older couples. In fact, relationship-specific investments are likely to be sizeable in long-term marriages in which couples have had not just years but often decades building their lives together, intertwining the pair financially, socially, and emotionally (Wineberg, 1996). For example, family and friendship networks likely become more enmeshed over time, with spouses especially likely to share these ties in longer term marriages. The role of marital longevity is especially relevant for older adults experiencing a separation in the second half of life as most have been married for decades. Thus, those in long-term first marriages should be most likely to reconcile.

Repartnering in later life

Following marital separation, many individuals form coresidential unions with new partners. Individuals who remain separated can form cohabiting unions but of course remarriage requires that individuals obtain a formal divorce. An impetus for seeking a divorce can be the desire to marry a new partner and in this sense the transition to divorce may be endogenous to the decision to enter a remarriage. Repartnering after gray divorce is fairly common, particularly among men. Roughly 37% of men and 22% of women repartner following a late-life divorce and these unions are more often cohabitations than remarriages, illustrating the growing prominence of cohabitation in the second half of life (Brown et al., 2019). Cohabitation appears to function as a long-term alternative to remarriage for older adults as these unions tend to endure and are unlikely to be formalized through marriage, and older cohabitors and remarried adults exhibit comparable relationship quality (King & Scott, 2005; Wright, 2020).

The antecedents of repartnering are distinctive in later life (Brown et al., 2019; Schimmele & Wu, 2016; Vespa, 2012). Resources, such as economic factors, good health, and social ties, are positively related to forming a new union following late-life divorce although the magnitudes of these effects tend to be modest (Brown et al., 2019). Economic indicators such as employment and home ownership are tied men’s repartnering, whereas wealth is linked to women’s repartnering. Physical health status is predictive of forming a new union after gray divorce. Both the number of chronic conditions and the number of (instrumental) activities of daily living ([IADLs] ADLs) limitations an individual faces are negatively associated with cohabiting or remarrying. Social ties, such as having friends or relatives living nearby, deter repartnering among men but are unrelated to women’s repartnering (Brown et al., 2019). The marital biography is indicative of relationship-specific investments and is related to repartnering after gray divorce (Brown et al., 2019). As time elapses following marital dissolution, the likelihood of repartnering tends to diminish. Marital duration is positively related to repartnering among men. Similarly, the dissolution of a remarriage versus a first marriage is positively associated with repartnering among men but unrelated among women. Finally, age at dissolution is negatively related to repartnering for men and women alike. These disparate patterns of repartnering for women and men underscore the importance of examining reconciliation and repartnering separately by gender.

The role of gender in reconciliation and repartnering

Heterosexual unmarried older adults face a lopsided sex ratio. Women tend to live longer than men, resulting in a surplus of single women at older ages. This imbalanced sex ratio rises with age, meaning that older men enjoy a much larger pool of potential mates than do women. Gendered patterns of mate selection, in which men prefer to partner with younger women, whereas women typically partner with older men, only exacerbates the imbalanced sex ratio (Bischoff, 2024). Men’s pool of potential mates widens as they age yet for women the pool narrows with age. Consequently, repartnering levels are much lower among older women than men (Brown et al., 2018, 2019; Schimmele & Wu, 2016). The ramifications of the imbalanced sex ratio for marital reconciliation are equivocal. On the one hand, women’s relative lack of new partners may heighten their chances of marital reconciliation. On the other hand, men may be less inclined to reconcile, reflecting vaster options in a larger pool of potential new mates.

Beyond the role of the lopsided sex ratio, qualitative work has shown that older women are much less desirous than older men of forming a late-life union (McWilliams & Barrett, 2014; Talbott, 1998). This differential arguably reflects the gendered bargain implied by remarriage and, to a lesser extent, cohabitation. Women exchange household labor and caregiving for men’s economic support (Brown et al., 2018). As they age, men’s financial obligations typically ease with an increasing reliance on retirement and savings as opposed to active employment, whereas for women, caregiving obligations often intensify with age. Older men are reluctant to partner with women who already have caregiving responsibilities (Harris, 2023). Older women also avoid partnering with men who require care, although they view men who are close to their families as more desirable potential partners (Harris, 2023). These disparate patterns exacerbate the gender gap in repartnering with age (Bischoff, 2024).

THE PRESENT STUDY

The recent growth in gray divorce raises new questions about the divorce process in later life. Marital separation does not always eventuate in divorce. Some couples will reconcile and still others will remain separated but not formally divorce. Regardless of whether they remain separated or get divorced, many adults, especially men, form coresidential unions with new partners. Adjustment to late-life marital disruption is protracted (Lin & Brown, 2020; Wang & Amato, 2000) and consequential for individual health and well-being (Lin et al., 2019).

Our goal is to examine how older adults navigate a marital separation (and subsequent divorce, if applicable) by considering two competing outcomes: marital reconciliation versus repartnering. Marital reconciliation has received scattered scholarly attention over the years (Binstock & Thornton, 2003; Morgan, 1988; Tumin et al., 2015; Wineberg, 1994, 1996), but the limited work to date has focused on younger women whose experiences are likely to differ from those who separate later in the life course. On balance, older adults are likely to enjoy more resources but dividing them leads to large declines in standard of living and assets (Lin & Brown, 2021). Unlike younger adults, older adults frequently contend with health difficulties. Moreover, older adults, who often have been married for decades, have sizeable relationship-specific investments, including shared children and grandchildren, family, and friends.

Uniting research on both marital reconciliation and repartnering is the centrality of resources. We conceptualize resources broadly to encompass economic factors, overall health, and social ties with others. In line with prior research, our expectation is that resources are negatively associated with marital reconciliation (Morgan, 1988; Tumin et al., 2015; Wineberg, 1996) and positively associated with repartnering (Brown et al., 2019; Vespa, 2012). Economic resources should confer financial autonomy and security, reducing the necessity for individuals to depend on their former spouse, thereby deterring reconciliation. In contrast, financial hardship may trump marital discord and spur reconciliation. Financial security enhances one’s attractiveness as a potential partner, heightening the chances of repartnering. Ongoing health difficulties such as chronic conditions or limitations in one’s ability to perform ADLs may impede independent living for some separated adults, forcing them to consider reconciliation. Health limitations diminish one’s attractiveness in the repartnering arena, and this appears to be particularly true among women (Brown et al., 2018). Social ties with friends and family as well as having children and grandchildren may encourage marital reconciliation but deter repartnering. Finally, the marital biography is indicative of relationship-specific investments. The longer a couple has been married, the greater the chances of reconciliation and the lower the chances of repartnering. Similarly, those who have separated from a first marriage should be more likely to reconcile than those exiting a remarriage, both because first marriages tend to be of longer duration and because relationship-specific investments are likely to have been larger in first marriages than remarriages. Obtaining a divorce should greatly diminish the chances of reconciling and increase the chances of repartnering. Finally, the quality of the marriage from which one has separated should be positively associated reconciliation (any linkage with repartnering is unclear).

We examine patterns of reconciliation and repartnering separately for women and men. To date, the literature on reconciliation has focused exclusively on women, ignoring men. In addition to being the first study to examine reconciliation among older adults, another contribution of our work is the inclusion of men. Repartnering in later life varies markedly by gender, reflecting the imbalanced sex ratio and age-graded mate selection preferences that facilitate repartnering for men and impede repartnering among women (Bischoff, 2024). In line with prior research on repartnering in later life, we examine women and men separately (Brown et al., 2019; Vespa, 2012) to elucidate how various economic, health, and social resources as well as the marital biography are linked to marital reconciliation and repartnering.

METHOD

Data came from the 1998–2018 HRS, a nationally representative longitudinal study of adults over the age of 50 conducted by the University of Michigan (https://hrs.isr.umich.edu/about). The HRS interviewed 33,855 older adults in 1998 and continue to interview them biennially, folding in new cohorts of 51–56-year olds every 6 years (2004, 2010, and 2016). The HRS is well suited for our aims because it includes a sizeable number of older adults who experience marital separation, allowing us to prospectively track marital reconciliation versus repartnering. No other US data set has a sufficient number of respondents who experience a gray marital separation. Across the two decades of data, we identified 24,552 respondents aged 50 and older who reported being married. After dropping 128 respondents in same-sex marriages, among the 24,424 respondents aged 50 and older who were in different-sex marriages, a total of 1115 experienced a marital separation at or after their initial wave of entering the HRS, ensuring that we avoided left-censoring bias. For the 47 individuals who experienced more than one marital separation at age 50 or older, we used only their first separation. An additional 40 respondents with a zero sampling weight were excluded from the analytic sample. The final analytic sample included 526 women and 549 men who experienced a marital separation in later life. Among women, 52 reconciled with a former spouse, 78 repartnered with a new partner, and 396 reported that they remained unpartnered. Among men, 60 reconciled with their former spouse, 171 formed a union with a new partner, and 318 remained unpartnered. Note that among women who repartnered, 46% formed a cohabiting union, whereas 56% remarried. Among men who repartnered, 60% entered a cohabitation and only 40% remarried.

Measures

Transition type

The dependent variable for this study captured whether the respondent exited the separated state by either reconciling with their spouse or forming a union with a new partner (i.e., repartnering). We created a marital history file to track the marital transitions experienced by each respondent. Using the marital history file, we identified individuals who transitioned from married to separated. The wave at which a respondent reported being separated marked the beginning of their risk period for reconciliation or repartnering. At each subsequent wave, the respondent’s marital status report was used to determine whether they experienced either type of transition (reconciliation vs. repartnering) or if they continued to be separated (or transitioned into divorce as explained later in this section). The HRS assigned unique spouse/partner identification numbers to spouses/partners, allowing us to distinguish between reconciliations and new unions. Specifically, when a respondent transitioned from separation to marriage to a spouse whose identification number matched that of the spouse from whom the respondent had been separated, we coded this transition as a marital reconciliation. When a respondent transitioned from separated (or divorced) to either cohabiting or married and the spouse/partner identification number did not match that of the spouse from whom they had separated (or divorced), we coded this transition as repartnering. We grouped together new cohabitations and remarriages into an umbrella category of repartnering because cohabitation is more common than remarriage after gray divorce (Brown et al., 2019) and cohabitation appears to function as a long-term alternative to remarriage in later life (Wright, 2020).

Economic resources

Years of education measured the years of schooling the respondent completed. Employed was a time-varying indicator that measured whether the respondent was full- or part-time employed (0 = No, 1 = Yes). Owns home, a time-varying covariate, measured whether the respondent owned a home (0 = No, 1 = Yes). Wealth was constructed as a time-varying measure of the logged value of the respondent’s nonhousing household wealth. Negative and zero wealth values were recoded as $1 and then logged. Pension was a time-varying measure of whether the respondent had received or would receive pension income (0 = No, 1 = Yes). Health insurance was a time-varying measure that indicated whether the respondent had health insurance coverage (0 = No, 1 = Yes).

Health

Number of chronic conditions was a time-varying measure of the sum of chronic diagnoses, ranging from 0 to 8. Chronic conditions included psychiatric disorders, heart disease, hypertension, cancer, lung disease, arthritis, diabetes, and stroke. Number of ADL/IADL difficulties was a time-varying measure of the sum of reported difficulties with activities ADLs and IADLs, ranging from 0 to 12. ADL included difficulties in walking, dressing, getting in and out of bed, bathing, eating, and toileting; IADL included difficulties in preparing hot meals, shopping for groceries, making phone calls, taking medications, managing money, and using a map.

Social ties

Relatives or friends nearby was a time-varying indicator of whether the respondents had relatives or friends living in close geographical proximity to them (0 = No, 1 = Yes). A categorical measure captured whether the respondent had children or grandchildren, distinguishing among those with children and grandchildren, children but no grandchildren, and no children (reference category).

Marital biography

Years since separation was a time-varying indicator of the number of years since the respondent’s marital separation occurred, effectively functioning as a counter of time elapsed since the respondent entered the risk period (this counter continues even for respondents who experienced divorce during the observation period). Duration of prior marriage was a time-invariant measure that indicated the number of years that the respondent’s prior marriage lasted. Age at separation was a time-invariant measure of the respondent’s age at the time of their separation. Separation from remarriage was a time-invariant covariate that accounted for whether the respondent was in a higher order marriage at the time of separation (0 = No, 1 = Yes). Divorced from prior marriage was a time-varying covariate that indicated the respondent had transitioned from separated to legally divorced. Once a respondent had reported being divorced, the measure was continuously coded as 1 (it was coded 0 while the respondent was only separated). Marital quality of prior marriage was a time-invariant covariate because these questions were only asked of partnered respondents when they first entered the HRS (or were first observed as married). The binary measure distinguished between individuals who reported that they had spent their free time with their (now former) spouse mostly together and had rated their time together as extremely enjoyable (=1) and 0 otherwise.

Demographic characteristics

Race/ethnicity was a binary measure that indicated whether the respondent self-identified as white (coded 0) or nonwhite (coded 1). Baby boomer cohort was another binary measure that differentiated between respondents who were born in 1946–1964 (coded 1) and those who were born prior to 1946 (coded 0).

Analytic strategy

We began by estimating survival probabilities for women and men separately to chart the timing and levels of marital reconciliation and repartnering after marital separation. Since reconciliation and repartnering generally occur shortly after marital separation, respondents were censored after 10 years at risk. Next, we calculated baseline bivariate descriptive statistics for women and men to examine how the covariates differed across each pathway: reconciliation, repartnering, or continuously separated. Lastly, we performed multinomial logistic regression analyses using discrete-time event-history models to examine the role of resources and relationship-specific investments on the relative risks of reconciling with one’s spouse, repartnering (with a new partner), or remaining separated. Multinomial logistic regressions were conducted separately for women and men. Our models estimated the relative risk ratios (RRRs), or the associations of economic, health, and social resources as well as the marital biography and demographic characteristics with the relative risks of either reconciling with one’s spouse or repartnering with a new partner versus remaining continuously separated.

Discrete-time event-history models are well suited to our study because they allow us to incorporate time-varying covariates to examine their distinct linkages with multiple outcomes, namely reconciliation versus repartnering. Data were reshaped into person-wave format (1571 person-waves for women and 1419 person-waves for men) and individuals were followed beginning at the time marital separation was first reported (minimizing any left-censoring bias) until they either reconciled with their spouse, formed a new coresidential union with another partner, or remained separated (or divorced) after 10 years had elapsed since the start of separation. Censoring respondents after 10 years of exposure to risk reflects the approach employed in prior studies of later life repartnering (Brown et al., 2019; Schimmele & Wu, 2016). We also censored respondents who transitioned from separated to widowed because they were no longer able to reconcile with their previous spouse. These respondents were censored at the wave in which they reported their marital status as widowed.

Missing data were minimal (<3%) except for homeownership (13%) and whether relatives or friends lived nearby (34%). These latter indicators were not obtained at every wave. We performed multiple imputation using chained equations (MICE) in Stata to impute missing values on each independent variable as a function of the other covariates and the dependent variable, union transition type. The results came from 10 random, multiple-imputed replicates. All results were calculated using the baseline weights (Ofstedal et al., 2011).

RESULTS

Figures 1 and 2 show the cumulative probabilities of reconciliation and repartnering over time for women and men, respectively. Reconciliation levels climbed monotonically beginning at the wave (i.e., 2 years) after marital separation and both the rate of increase and the overall levels appeared similar regardless of gender. Reconciliation was much less common than repartnering, however, and this was especially true among men as the probability of repartnering accelerated more quickly than that of reconciliation. In contrast, the rate of increase in repartnering largely paralleled that of reconciliation among women. Consistent with prior research, levels of repartnering were considerably lower among women than men and this gap widened over time.

FIGURE 1.

Cumulative probabilities of reconciliation versus repartnering for women.

FIGURE 2.

Cumulative probabilities of reconciliation versus repartnering for men.

Table 1 depicts the baseline characteristics of women according to whether they reconciled, repartnered, or remain separated. Economic resources were largely unassociated with either reconciliation or repartnering, but women who repartnered had much higher median wealth than women who stayed separated. Two significant differences emerged between women who reconciled with their spouses and women who repartnered. The former group was less educated and had much less wealth than the latter group, aligning with our resources hypothesis.

TABLE 1.

Weighted means Standard Deviations (SD) or percentages of baseline characteristics by separation transition type for women.

	Reconciled	Repartnered	Continuously separated
Economic resources
Years of education	12.43 (3.21)	13.78 (2.72)	13.14 (2.97)
Employed	35.35	46.40	45.40
Owns home	59.85†	56.89†	46.56
Median wealth ($)	9379	65,882 ***	9157
Has pension	23.20	20.16	12.72
Has health insurance	79.17	91.96	85.13
Health
Number of chronic conditions	2.57 (1.89) †	1.64 (1.25) †	1.98 (1.55)
Number of ADL/IADL difficulties	2.18 (2.94)	0.49 (1.28) *	0.96 (1.85)
Social ties
Relatives or friends nearby	65.42	52.16	60.64
Has children and grandchildren	80.75	72.70	65.88
Has children, no grandchildren	7.12*	21.04	26.26
Has no children, no grandchildren	12.13	6.13	7.79
Marital biography
Years since separation	2.02 (3.26)	1.57 (1.52)	1.71 (2.77)
Duration of prior marriage	25.39 (13.93)	21.45 (14.15)	22.38 (13.79)
Age at separation	62.16 (5.06)	60.30 (6.98)	60.88 (7.25)
Separated from remarriage	71.42	77.92 **	60.94
Divorced from prior marriage	27.63	74.54 ***	44.73
Marital quality of prior marriage	81.53	86.63†	81.53
Demographic characteristics
Nonwhite	32.48	19.07 *	21.71
Baby boomer	58.38	67.40	73.92
Weighted %	7.25	11.64	81.10
Unweighted N	52	78	396

Note: Bolded numbers indicate reconciled and repartnered respondents differ at p < .10. ^†p < .10, *p < .05, **p < .01, and ***p < .001 for differences between respondents who eventually reconciled or repartnered compared with those who remained continuously separated.

Abbreviations: ADL, activity of daily living; IADL, instrumental activity of daily living.

Additional evidence for this hypothesis was evident from the health indicators. Women who repartnered reported the best health (as signaled by the lowest average numbers of chronic conditions and ADL/IADL difficulties), followed by women who remained separated, and then lastly women who reconciled with their spouses. In contrast, women’s levels of social ties largely did not vary by their transition type (although women who reconciled less often had children only compared with women who remained unpartnered). Likewise, few differences emerged on indicators of the marital biography, which can be viewed as proxies for relationship-specific investments. Women who repartnered were more likely to have dissolved a remarriage than either women who reconciled with their spouse or remained separated. Those women who transitioned from separation to divorce during the observation period exhibited much higher levels of repartnering than reconciling (women who divorced were also less likely to remain single than to repartner). Women who reconciled with their spouses were disproportionately nonwhite compared with those who repartnered.

Turning now to men’s baseline characteristics as shown in Table 2, we again see that economic resources differed only modestly by separation transition type. Men who repartnered were more often employed than those who remained separated or reconciled. Additionally, the median wealth of men who repartnered was more than 10 times that of men who reconciled with their spouses (men who remained single also had more wealth than did men who reconciled). These associations are consonant with the resources hypothesis.

TABLE 2.

Weighted means (Standard Deviations) or percentages of baseline characteristics by separation transition type for men.

	Reconciled	Repartnered	Continuously separated
Economic resources
Years of education	13.27 (3.80)	13.27 (3.06)	12.83 (3.40)
Employed	39.62	57.38 **	45.45
Owns home	51.23	49.98	48.24
Median wealth ($)	2283 †	29,980	10,649
Has pension	15.63†	19.93	18.29
Has health insurance	80.29	84.42	86.21
Health
Number of chronic conditions	1.72 (1.71)	1.72 (1.51)†	2.05 (1.69)
Number of ADL/IADL difficulties	0.86 (1.54)	0.28 (0.81) **	0.93 (2.00)
Social ties
Relatives or friends nearby	61.34	58.90	63.55
Has children and grandchildren	76.33	64.35	67.36
Has children, no grandchildren	8.01*	28.75*	23.87
Has no children, no grandchildren	16.01	6.86	8.70
Marital biography
Years since separation	2.49 (3.93)†	1.92 (3.42)	1.66 (2.76)
Duration of prior marriage	26.05 (14.83)*	22.27 (12.85)	20.33 (12.88)
Age at separation	63.07 (5.74)	61.19 (7.00)	62.71 (8.37)
Separated from remarriage	78.39 †	83.91 ***	62.46
Divorced from prior marriage	19.45 ***	70.20 ***	54.03
Marital quality of prior marriage	80.24	81.46	83.03
Demographic characteristics
Nonwhite	42.42	25.18 †	27.13
Baby boomer	58.26	65.84	62.75
Weighted %	10.52	26.42	63.05
Unweighted N	60	171	318

Note: Bolded numbers indicate reconciled and repartnered respondents differ at p < .10. ^†p < .10, *p < .05, **p < .01, and ***p < .001 for differences between respondents who eventually reconciled or repartnered compared with those who remained continuously separated.

Abbreviations: ADL, activity of daily living; IADL, instrumental activity of daily living.

Health also operated in the expected fashion, although variation primarily emerged only for the number of ADL/IADL difficulties, which were fewest among men who repartnered versus men who reconciled or remained separated (these latter two groups did not differ). Men who repartnered also reported fewer chronic conditions than men who remained separated. Similar to women, social ties largely did not differ across the three groups, although men with children but no grandchildren were less likely to reconcile and more likely to repartner than remain single. The marital biography figured prominently for men. Men who reconciled had longer marriages than those who remained separated. Men who reconciled with their spouses had less often been in a remarriage (vs. a first marriage) than men who repartnered. And men who transitioned from separation to divorce more often repartnered than either remained single or reconciled. Men who reconciled were more often nonwhite compared with men who repartnered.

Table 3 shows the RRRs from the multinomial logistic regressions predicting the likelihoods of women either reconciling with their spouse or forming a union with a new partner relative to remaining separated. In the full model, few economic factors were associated with separation transition type. Notably, women’s education was unrelated to either reconciliation or repartnering. Employment was marginally negatively associated with repartnering versus remaining separated. Wealth was linked to an increased risk of repartnering relative to either remaining separated or experiencing reconciliation. Number of chronic conditions was positively associated with reconciliation. Consistent with the bivariate findings, social ties were unrelated to women’s patterns of reconciliation and repartnering. Marital biography indicators were related to women’s transitions. The more time that had elapsed since marital separation, the lower the risk of repartnering, signaling that if women repartner, they tend to do so relatively quickly. The longer the duration of the dissolved marriage, the higher the risk of repartnering (p < .10). And separation from a remarriage was positively associated with repartnering relative to remaining separated. Getting divorced was negatively related to reconciliation (p < .10) and positively related to repartnering. The chances of reconciliation were significantly lower than the chances of repartnering for women who got divorced. Nonwhite women were more likely to repartner than were their white counterparts (p < .10).

TABLE 3.

Relative risk ratios from discrete-time multinomial logistic regression of repartnering or reconciling for women (N = 1571 person-waves).

	Reconciled vs. continuously separated	Repartnered vs. continuously separated	Reconciled vs. Repartnered
Economic resources
Years of education	1.01	1.00	1.00
Employed	1.17	0.51†	2.36
Owns home	1.72	1.06	1.62
Wealth	1.00	1.25**	0.80*
Has pension	0.91	1.15	0.79
Has health insurance	0.49	1.61	0.30
Health
Number of chronic conditions	1.36*	1.15	1.18
Number of ADL/IADL difficulties	0.92	0.83	1.10
Social ties
Relatives or friends nearby	1.81	0.57	3.20
Has children and grandchildren	0.49	2.17	0.23
Has children, no grandchildren	0.16	0.91	0.18
Marital biography
Years since separation	1.00	0.86**	1.17*
Duration of prior marriage	1.02	1.03†	0.99
Age at separation	1.03	0.94	1.08
Separated from remarriage	1.86	4.60*	0.40
Divorced from prior marriage	0.36†	5.27**	0.07***
Marital quality of prior marriage	1.79	1.27	1.41
Demographic characteristics
Nonwhite	1.49	2.34†	0.64
Baby boomer	0.91	1.10	0.83

Abbreviations: ADL, activity of daily living; IADL, instrumental activity of daily living.

^†p < .10;

^*p < .05;

^**p < .01;

^***p < .001.

The multivariable model for men is shown in Table 4. Contrary to the resources hypothesis, education was positively related to reconciliation versus remaining separated or repartnering. Men who were employed were marginally more likely to repartner than their counterparts who were not working (p < .10). Homeownership was positively related to both reconciliation and repartnering among men. The more ADLs/IADLs men reported, the less likely they were to repartner, whether compared with remaining single (p < .10) or reconciling. In other words, poorer health was linked with an increased chance of reconciling versus repartnering. Having relatives or friends nearby was negatively associated with repartnering among men but was unrelated to reconciliation. Men with children (or children and grandchildren) were less likely to reconcile than either remain separated or repartner relative to their childless counterparts, which was contrary to our expectations. The amount of time that had elapsed since marital separation was positively related to reconciliation relative to either remaining separated or repartnering. Men who experienced separation from a marriage of longer duration were more likely to transition to either reconciliation or repartnering. Similarly, having separated from a remarriage was positively associated with both reconciliation and repartnering among men. Getting divorced was negatively associated with reconciliation and positively related to repartnering. As expected, divorced men were much less likely to experience reconciliation than repartnering. Demographic characteristics were largely unrelated to men’s transitions after separation although nonwhite men were more likely to reconcile than their white counterparts.

TABLE 4.

Relative risk ratios from discrete-time multinomial logistic regression of repartnering or reconciling for men (N = 1419 person-waves).

	Reconciled vs. continuously separated	Repartnered vs. continuously separated	Reconciled vs. Repartnered
Economic resources
Years of education	1.12*	0.97	1.15†
Employed	1.35	1.81†	0.74
Owns home	2.58*	1.98*	1.30
Wealth	0.96	1.01	0.95
Has pension	0.80	0.62	1.29
Has health insurance	0.68	1.09	0.62
Health
Number of chronic conditions	0.97	0.98	0.99
Number of ADL/IADL difficulties	1.17	0.87†	1.36*
Social ties
Relatives or friends nearby	0.75	0.44*	1.70
Has children and grandchildren	0.22*	1.13	0.19†
Has children, no grandchildren	0.10**	1.22	0.08*
Marital biography
Years since separation	1.12***	1.02	1.10*
Duration of prior marriage	1.06***	1.05***	1.02
Age at separation	1.00	1.01	1.00
Separated from remarriage	5.26**	7.78***	0.68
Divorced from prior marriage	0.26**	7.62***	0.03***
Marital quality of prior marriage	1.33	1.08	1.23
Demographic characteristics
Nonwhite	2.06*	1.19	1.74
Baby Boomer	1.13	1.57	0.72

Abbreviations: ADL, activity of daily living; IADL, instrumental activity of daily living.

^†p < .10;

^*p < .05;

^**p < .01;

^***p < .001.

DISCUSSION

Marital disruption increasingly occurs among older adults, yet researchers know relatively little about how older adults navigate the dissolution process. Men and, to a lesser extent, women often move on to form new partnerships. Older adults with greater resources tend to be more likely to form a new cohabiting or marital union, presumably because their economic security and robust health make them more attractive partners (Brown et al., 2018, 2019; Vespa, 2012). In contrast, individuals facing marital separation with fewer resources may be more inclined to reconcile with their spouse because it is more difficult to go it alone (Morgan, 1988; Wineberg, 1996). Moreover, relationship-specific investments are sizeable for older adults exiting long-term marriages, which may encourage reconciliation and deter repartnering.

Unlike the limited prior research on marital reconciliation earlier in the adult life course, we explicitly modeled repartnering as a competing risk with reuniting with one’s spouse. Using prospective data, we tracked women and men who experienced a marital separation after age 50 and followed them until they either reconciled with their spouse or repartnered with someone else, censoring them after 10 years of exposure to risk. Our analyses revealed that marital reconciliation is not uncommon; roughly 7% of women and 11% of men reconciled with their former spouse. By comparison, about 12% of women and 26% of men instead repartnered with someone else. Together, these levels are consistent with those found in prior studies that effectively lumped together reconciliations with a former spouse and repartnering with someone new (Brown et al., 2019).

It is important to note that our estimates of marital reconciliation are conservative. The HRS data do not permit us to capture cases in which couples separate and then reconcile between interview waves. We can only identify reconciliations that occur at an interview wave after the wave at which marital separation was reported. Given that each interview wave occurs roughly every 2 years, we are arguably missing a nontrivial share of all reconciliations because short-lived separations that culminate in reconciliation are probably fairly common. Our inability to measure this scenario is a limitation of our study, but one that is shared with prior research on marital reconciliation (Tumin et al., 2015).

Our study provides mixed support for our hypotheses that resources are negatively associated with reconciliation but positively related to repartnering. Although some indicators (e.g., education and wealth) of economic resources operated in the expected directions for women at the bivariate level, none was related to reconciliation in the multivariable model, an unexpected finding given the emphasis in the literature on women’s greater proclivity to reconcile when facing economic disadvantage. These results challenge conventional framing of marital reconciliation that has emphasized economic dependence as a key motivating factor. Of course, much of this theorizing is now dated (Wineberg, 1996), eliding the fact that nowadays most wives have at least some economic autonomy. In addition, it is possible that even for women with relatively few economic resources, a desire to exit the marriage may supersede the potential economic benefits of marital reconciliation. Women initiate most marital separations in the second half of life (Montenegro, 2004). Unfortunately, the HRS does not collect information on who initiated the breakup. Women’s economic resources were largely unrelated to forming a new union, although wealth was positively associated with repartnering. And, among women who did not remain single, wealthier women more often repartnered than reconciled. The modest role of economic resources echoes the findings from Brown et al. (2019) on repartnering after gray divorce. Health operated as expected at the bivariate level for women. In the multivariable model, women’s number of chronic conditions was positively associated with reconciliation, which aligned with our expectations. Yet, women’s health was not appreciably tied to repartnering, which is in contrast to at least one prior study showing that women’s vitality was positively related to union formation (Brown et al., 2018). Likewise, women’s social ties were associated with neither reconciliation nor repartnering. Having children and even grandchildren, for example, did not alter women’s propensities to reconcile with their spouse or form a new union. The absence of significant associations may reflect women’s relatively broad social networks that typically include many friends (Carr, 2004).

At the same time, resources are largely unrelated to marital reconciliation among men, which is contrary to our expectations. In fact, the only significant associations between resources and reconciliation operated in the opposite direction: both education and home ownership were positively associated with men’s transitions to reconciliation, and the presence of children or grandchildren reduced the risk of reconciliation. Perhaps childless men have fewer social ties to others (e.g., children and grandchildren) and are thus motivated to reconcile with their spouses. Still, our expectation was that children and grandchildren are forms of marital-specific capital and thus they would spur reconciliation. Regardless, this pattern of findings calls into question the existing frameworks for understanding reconciliation that have been formulated for men. Our results signal that such frameworks do not necessarily apply to older separated men. Resources largely operated as expected for men’s repartnering, which was more common among those who were employed, owned a home, had fewer ADL/IADLs, and no relatives or friends nearby. Comparing decisions to reconcile versus repartner among men who did not remain single aligns more closely with our expectation in that reporting more ADL/IADLs was more often linked with reconciliation than repartnering. However, more educated men more often reconciled than repartnered.

Although few indicators of marital-specific capital were linked to marital reconciliation (or repartnering), the marital biography was closely associated with reconciliation and repartnering for women and men alike. Marital duration was positively related to reconciliation among men, aligning with the notion that the accumulation of marital-specific capital encourages reconciliation. And exiting a remarriage was tied to increased chances of both reconciliation and repartnering for men. For women, it was positively related to repartnering. Upon divorce, women and men were much less likely to reconcile and much more likely to repartner, reinforcing our assertion that the transition to divorce is endogenous to forming a new union. The salience of the marital biography attests to the life course principle that earlier events and experiences have enduring consequences for subsequent transitions (Elder et al., 2003). Arguably, marital biography factors were more prominently tied to reconciliation and repartnering than were economic resources, health, and social ties, which challenges the existing theoretical frameworks that emphasize resources.

The patterns we uncovered appeared distinctive for women and men, suggesting that the mechanisms underlying decisions to reconcile with one’s spouse versus repartner with someone new or just remain single could differ by gender. Researchers recognize the centrality of gender in terms of structuring the likelihood of repartnering in later life (Bischoff, 2024) but lack theory to decipher whether and how the mechanisms may differ for women and men. Our work reveals that prevailing explanations for reconciliation, for example, do not align with older men’s experiences. Rather, resources were positively related to men’s marital reconciliations. We also uncovered some unexpected linkages between resources and repartnering for women and men. An important next step in this line of inquiry is theory development to generate testable hypotheses about the unique factors contributing to women’s and men’s late-life union formation.

Even as this study contributes to both the literature on marital reconciliation and scholarship on repartnering, it also has some limitations. First, as noted earlier, the design of the HRS precludes us from capturing those short-term separations that eventuate in reconciliation entirely between interview waves, meaning our study underestimates levels of marital reconciliation. It also means our study is biased toward lengthier separations that culminate in reconciliation. It is possible that the antecedents of shorter separation to reconciliation spells are distinctive from those identified in our findings. Research on partnered young adults reveals that relationship churning or cycling is fairly common (Dailey et al., 2009; Halpern-Meekin et al., 2013; Monk et al., 2018). Whether and to what extent married couples, especially those in late-life, experience such cycling is unknown but is worthy of future investigation. Second, although we were able to include a snapshot measure of marital quality, it was typically obtained well before marital separation occurred. Ideally, ongoing measurement of marital quality could tap into churning by identifying peaks and valleys in couple marital dynamics over time. However, even contemporaneous measures of multiple dimensions of marital quality at the time of separation would likely be more indicative of the chances of reconciliation as positive and negative dimensions of quality could be uniquely associated with reuniting with one’s spouse. Finally, we recognize that many older adults may form non-coresidential dating or living apart together relationships after a late-life marital separation (de Jong Gierveld & Merz, 2013; Harris, 2023) and these relationships likely reduce the chances of reconciliation and augment the likelihood of repartnering through cohabitation or remarriage. Unfortunately, the HRS does not collect information on respondents’ non-coresidential romantic relationships.

Marital disruption is increasingly common among older adults as evidenced by the rapid growth in gray divorce over the past half century. Our study demonstrates that at least 7%–11% of older couples reconcile and get back together, and this estimate is conservative. Unlike the patterns uncovered for younger women, we did not find consistent evidence that lacking resources or having more relationship-specific investments necessarily encouraged reconciliation. This pattern held across some dimensions but not others among women, and for men we uncovered contrary evidence (i.e., resources were sometimes positively related to reconciliation), suggesting that the mechanisms underlying the marital reconciliation process may be distinctive for older adults in unique ways for women and men. Our work signals not only the value of conducting future research on the determinants of marital reconciliation among older adults but also the consequences of reconciliation for subsequent health and well-being.