The Roles of Gray Divorce and Subsequent Repartnering for Parent–Adult Child Relationships

Abstract

Objectives

Divorce is now widespread in later life, yet little is known about how older adults and their adult children respond in the aftermath of gray divorce. Guided by the life course perspective, this study examines the consequences of gray divorce and subsequent repartnering for parent–adult child relationships from the parent’s perspective.

Methods

Using longitudinal data from the 1998–2014 Health and Retirement Study in the United States, we estimated growth curve models to compare fathers’ and mothers’ frequent contact with and financial support to their adult children prior to, during, and following gray divorce.

Results

Gray divorce and repartnering had disparate effects on father– versus mother–adult child relationships. Following a divorce, fathers’ frequent contact with their adult children decreased but financial support to their adult children increased. Fathers’ repartnering had an enduring negative effect on frequent contact with their children. Gray divorce did not alter mothers’ financial support to adult children and it actually increased interaction between mothers and adult children as the odds of frequent contact doubled upon divorce. Repartnering had no appreciable effects on mothers’ relationships with their adult children.

Discussion

The results of our study are consistent with prior research showing that divorce creates a matrifocal tilt in our kinship system. The shifting dynamics of parent–adult child relationships in response to gray divorce and repartnering raise questions about whether gray-divorced parents will be able to rely on their adult children for care as they age.

Divorce has become an increasingly common experience for middle-aged and older adults (“older adults” hereafter) in the United States. The divorce rate among adults aged 50 and older, which is termed gray divorce, has doubled since 1990, from five divorcing persons per 1,000 married persons in 1990 to 10 divorcing persons in 2010 (Brown & Lin, 2012). Since 2010, the gray divorce rate has reached a plateau (Cohen, 2019). However, more than one quarter of today’s divorces are gray divorces. Even if the gray divorce rate remains constant, the number of people affected will climb 33%, from about 600,000 in 2010 to more than 800,000 in 2030, simply because of the aging of the population (Brown & Lin, 2012). The growing prominence of gray divorce portends a corresponding increase in subsequent repartnering through remarriage or cohabitation. A sizable share of individuals—about 22% of women and 37% of men—repartner within 10 years after gray divorce (Brown et al., 2019).

Although these demographic trends have piqued researchers’ interest in studying the antecedents of gray divorce and subsequent repartnering (Brown et al., 2019; Karraker & Latham, 2015; Lin et al., 2018), little is known about how older adults and their adult children respond in the aftermath of gray divorce (Shapiro & Cooney, 2007). Marital dissolution typically weakens parent–child relationships, particularly for fathers (Amato, 2010). Nonetheless, most studies on the consequences of divorce for parent–child relationships either examine divorces that occurred earlier in the life course or rely on the adult child’s perspective. Whether these findings are applicable to parents who experience gray divorce remains unknown.

In this article, we draw on key tenets of the life course perspective to assess the roles of gray divorce and subsequent repartnering for parents’ appraisals of their relationships with each of their adult children. Using the 1998–2014 Health and Retirement Study (HRS) to estimate growth curve models, we follow parents prior to, during, and after gray divorce to examine how later-life marital dissolution and subsequent union formation are related to frequent contact with and downward financial transfers to adult children, paying particular attention to the role of parent’s gender. Building on prior work, which only has gauged the timing of parental divorce from the standpoint of the adult child, we shift the focus to the parent to address how gray divorce and repartnering shape fathers’ and mothers’ appraisals of intergenerational ties. We anticipate that gray divorce and subsequent repartnering have negative consequences for father–adult child relationships, but the effects for mother–adult child relationships are less clear given the disparate findings in prior research. Our study contributes to the literature on the consequences of divorce for parent–child relationships by capturing the parent’s position in the life course and relationships to all (not just one) adult children in the family to perform a novel examination of the roles of gray divorce and subsequent repartnering for parent–adult child relationships in later life.

A Life Course Perspective on Parent–Child Relationships

The life course perspective views human development as a lifelong process involving the interplay of individual, social, and historical forces (Elder et al., 2003). Two of the principles—the interdependency of family members (i.e., linked lives) and the timing of life transitions—are particularly relevant for understanding the unique impact of gray divorce on parent–adult child relationships (Shapiro & Cooney, 2007).

Family members’ lives are intertwined. Transitions in one person’s life are inherently linked to the experience and development of other family members (Elder et al., 2003). Divorce often pits parents against each other and these changes in the marital dyad are likely to spill over into parent–child relationships (Booth & Amato, 1994; Orbuch et al., 2000). Children may resent the news of their parents’ divorce and blame one or both parents for the breakup (Cooney et al., 1986). Parental conflict also may force children to take sides in parental disputes during and after the divorce, subsequently strengthening or weakening parent–child ties (Amato & Afifi, 2006; Cooney et al., 1986). However, parents’ appraisals of their relationship dynamics with their adult children following divorce may not align with the perspectives of their adult children.

The developmental consequences of life transitions and events, such as divorce and repartnering, vary according to their timing in one’s life (Elder et al., 2003). Later-life divorce, an off-time event, is likely to present different challenges than early-life divorce. For instance, whereas early-life divorce involves legally enforceable child support and visitation arrangements, later-life divorce typically does not. Nonetheless, most parents continue to maintain frequent contact with and provide financial transfers to their adult children throughout the life course (Kalmijn, 2019; Kalmijn & de Vries, 2009). It is largely unknown how gray divorce affects parent–adult child relationship dynamics and whether gender differences in intergenerational ties after gray divorce are distinct from those after early-life divorce.

Parental Divorce and Parent–Adult Child Relationships

Parents and adult offspring provide each other with emotional and material support (Seltzer & Bianchi, 2013). Adult children are an important source of psychological and social support for parents across the life course (Kalmijn & de Vries, 2009). Transfers from older parents to their adult children are much more frequent and intense than those in the reverse direction. The net downward flow from parents to children decreases with rising parental age but does not reverse until very late in life (Albertini et al., 2007; Kalmijn, 2019). Gray divorce is likely to disrupt the patterns of parents’ contact with and downward financial transfers to their adult children but whether that results in reductions or increases in contact with and financial transfers to adult children is unclear.

On the one hand, gray divorce could reduce child contact, reflecting strained parent–child relationships as children often blame their parents for the divorce, wish to avoid being caught between two feuding parents, or aim to balance their attention and time given to each parent (Amato & Afifi, 2006; Cooney et al., 1986). Parents may engage less with adult children following divorce, whether because they remind the parent of the ex-spouse or because they are pursuing new romantic relationships. Moreover, gray divorce reduces parents’ economic resources by cutting nest eggs in half (Lin & Brown, 2020; Sharma, 2015), likely diminishing their financial transfers to adult children. On the other hand, gray divorce may increase contact as children form alliances with one parent or divorced parents lean on adult children for emotional support or advice (Arditti, 1999) or even coresidence (Cooney, 1989). Divorced parents may also harbor feelings of guilt that they could attempt to assuage by augmenting their financial transfers to their children (Troilo & Coleman, 2012).

The few studies on later-life divorce paint an incomplete picture of its effects on parent–adult child relationships. Prior research has consistently revealed that parental divorce in later life has negative consequences for father–adult child ties. Parental divorce that occurs after children reach adulthood is associated with less father–child contact (Aquilino, 1994; Cooney, 1994; Shapiro, 2003). In contrast, the consequences of later-life divorce for mothers and their adult children are decidedly mixed. Some scholars find that adult children have less contact with their divorced mothers than continuously married mothers (Aquilino, 1994). Others suggest no significant negative effects of divorce on mother–child contact (Cooney, 1994). Still, others show that divorced mothers are more likely than continuously married mothers to report an increase in frequent contact with at least one adult child (Shapiro, 2003).

The linkage between later-life parental divorce and financial support to adult children is even less clear, perhaps reflecting a paucity of research. In a cross-sectional study, Aquilino (1994) shows a negative association between parental divorce and financial transfers to adult children. In contrast, Shapiro and Remle (2011), using the 1992–1998 HRS, find that relative to parents who are continuously married, parents who become divorced are slightly more likely to provide financial transfers to their children but parents who are continuously divorced are less likely to provide financial transfers, suggesting that a recent divorce may be associated with a temporary increase in the propensity of downward financial transfers, but over time the likelihood of providing financial support declines. Neither study examines these associations separately by parent’s gender so whether the patterns are applicable to both fathers and mothers is unknown.

These disparate findings align with the overarching conclusion that divorce generally exerts more negative impacts on father–child than mother–child relationships (Bulcroft & Bulcroft, 1991), at least from the adult child’s perspective. Scholars have offered two explanations for this gender differential (Kalmijn, 2007). One is the investment explanation (de Graaf & Fokkema, 2007). Relative to mothers, fathers tend to spend less time with their children (Sayer et al., 2004) and are less likely to live with their children after divorce (Cooney, 1994; Shapiro, 2003). Thus, divorce is more consequential for father–child ties than for mother–child ties. The other is the kin keeper explanation (also called the marriage protection hypothesis, see Kalmijn, 2007). Mothers tend to be closer to their children than are fathers and often act as the gatekeeper between their spouses and children (Townsend, 2002). Once divorced, men lose not only a wife but also the conduit that connects them with their children.

However, gray divorce usually does not involve child support and visitation arrangements, and mothers may be less likely to act as a kin keeper after children enter adulthood. Thus, gender differences in parent–child relationships may be smaller following later-life divorce than early-life divorce, underscoring the importance of the timing of divorce in the life course. Indeed, Furstenberg et al. (1995) show that parental gender differences in the likelihood of financial transfers to children narrow as a child’s age at the time of parental divorce increases (from age 0 to 18+). Still, several studies continue to document a greater reduction in parent–adult child contact associated with later-life divorce for fathers than mothers (Aquilino, 1994; Cooney, 1994; Kalmijn, 2007).

Repartnering After Divorce and Parent–Adult Child Relationships

Only a handful of studies have considered the timing of divorce when examining the consequences of divorced parents’ repartnering for parent–adult child relationships. For instance, Noël-Miller (2013), pooling four waves of data from the HRS, finds that divorced fathers’ new unions are associated with less contact with and a lower likelihood of providing financial support to their adult children, relative to divorced fathers who remain single. Repartnering through remarriage appears to be as detrimental as repartnering through cohabitation. Although her study does not specifically focus on gray divorce, the findings hold regardless of whether divorce occurs before or after children reach age 18.

Another cross-sectional study (Kalmijn, 2015) conducted in the Netherlands echoes Noël-Miller’s (2013) results for fathers, showing that those who repartner after divorce exhibit lower levels of contact and financial support than those who remain single. However, mothers who repartner after divorce maintain as frequent contact with and support to their children as mothers who remain single. Additionally, repartnering diminishes parent–adult child contact and downward financial transfers more so for fathers than mothers. According to this study, the timing of divorce matters, at least for fathers. The detrimental effects of repartnering on parent–child relationships are weaker after children reach age 18 for fathers, but there is no appreciable difference for mothers.

In general, three views have been offered in the literature to explain why the father–child relationship following early-life divorce tends to worsen with remarriage. These explanations may apply to repartnering following gray divorce, but the timing of divorce could impose unique effects. First, repartnered fathers shift their investments to the new spouse and her family, including their shared children (or her children) by “swapping” families (Manning & Smock, 2000). Second, some children may not get along with their father’s new partner (Kalmijn, 2015), subsequently withdrawing contact and exchanges with their father. Third, the spouse is usually the caregiver when one needs help, suggesting fathers who have a new spouse would have less need for support from their adult children than those who remain single (Kalmijn, 2015). These explanations can be applied to repartnered mother–child relationships, too, though arguably to a lesser extent because men usually are less supportive in couple relationships than women, and thus repartnered mothers may rely more on their adult children than do repartnered fathers (Kalmijn, 2007).

The Present Study

Our examination of parent–adult child relationship dynamics prior to, during, and following gray divorce and repartnering fills several notable gaps in the existing literature. First, even though gray divorce is on the rise and a sizable share of individuals repartner through remarriage or cohabitation within 10 years after gray divorce, little is known about how parent–adult child relationship dynamics shift in the aftermath of gray divorce, particularly from the perspective of the parent. Nearly all of the studies on how parental divorce affects parent–adult child relationships take the perspective of the adult child by constraining observations to offspring who were adults when their parents divorced. It is critical to assess the role of divorce timing from the parent’s position in the life course because we know that gray divorce presents unique challenges for older adult adjustment (Lin et al., 2019; Lin & Brown, 2020) and thus by extension may have significant consequences for parent–adult child relationship dynamics. Our focus on divorce timing from the parent’s perspective provides novel insights on intergenerational relationships after gray divorce.

Second, prior research tends to focus on one parent–adult child dyad in the family (Aquilino, 1994; Cooney, 1994; Kalmijn, 2007, 2015), but divorce likely has distinctive effects on a parent’s relationships with each child. One study (Shapiro, 2003) shows that while divorced mothers report an increase in weekly contact with one or more adult children, they are also more likely than married mothers to report being estranged from at least one of their adult children. Parents’ lives are uniquely linked to those of each of their children, and parents and adult children continue to negotiate their respective roles through their linked lives (Greenwood, 2012; Pett et al., 1992). Thus, our study considers how ties with each child shift in response to parental divorce and appraises whether the impact of later-life divorce on intergenerational relationships is short- or long term.

A third limitation is that most prior studies on how parental divorce affects parent–adult child relationships use cross-sectional data, thereby failing to account for selection. Parents who have poorer marital quality also tend to have lower relationship quality with their children (Booth & Amato, 1994; Orbuch et al., 2000). To isolate the effect of divorce on parent–child relationships, one needs to control for parent–child relationships prior to divorce. Although some researchers use two waves of data to account for the predivorce parent–child relationships (Kaufman & Uhlenberg, 1998; Shapiro, 2003), they still compare divorced parents with continuously married parents. People experiencing gray divorce are already disadvantaged relative to those whose marriages remain intact (Lin et al., 2018). By comparing divorced parents with continuously married parents, researchers tend to overestimate the effect of divorce. Our study follows the same individuals prior to, during, and after gray divorce (i.e., within-person comparisons) to account for the nonrandom selection into divorce and repartnership.

Drawing on the life course principles of linked lives and the timing of life transitions (Elder et al., 2003), we use data with a long observation window spanning 16 years, including before, during, and after divorce, to address the substantive and methodological limitations in prior research and provide a more nuanced understanding of how gray divorce and subsequent repartnering are related to parent–child contact and downward financial transfers from the perspectives of fathers and mothers. Based on prior research about the effects of parental divorce and remarriage on parent–adult child relationships, we hypothesize that gray divorce is negatively related to father–child relationships. Specifically, we posit that fathers are less likely to engage in frequent contact with their adult children following divorce. Fathers’ provision of financial support to their adult children also is expected to diminish after divorce. Whether gray divorce has similar negative impacts on mother–child relationships is unclear given the inconsistent results in the literature to date. Regardless, the effect of gray divorce on frequent contact and downward financial transfers is expected to be more negative for fathers than mothers. We conjecture that repartnering after gray divorce further weakens intergenerational ties. Repartnered parents have a lesser need for support from adult children, relying instead on their new partner. Also, they may disinvest in their adult children from the marriage that ended in divorce in favor of providing financial support to children from the new union. The negative association between repartnering and intergenerational ties is expected to be larger for fathers than mothers.

In the analysis, we also account for parents’ and adult children’s characteristics that may confound the association of parental divorce with parent–child contact and downward financial transfers. Previous studies have shown that gray divorce is inversely related to age, education, marital duration, household income, and homeownership but positively related to being Black, higher-order marriage, and poorer health (Brown & Lin, 2012; Karraker & Latham, 2015; Lin et al., 2018). Older parents and parents with more education are more likely to have frequent contact with their adult children (Cooney, 1994; Shapiro, 2003). Older children, sons, and adult children who have received more education tend to have less frequent contact with their parents than their respective counterparts (Cooney, 1994; Kalmijn, 2007). Financial support to children is positively associated with parental income but inversely associated with adult children’s age, marital status, parental status, and income (Furstenberg et al., 1995).

Method

Data came from the 1998–2014 HRS in the United States, harmonized by the RAND Corporation. The HRS is a prospective, nationally representative survey of American adults aged 51 and older and their spouses (or partners). The biennial study began in 1992 with the original HRS cohort (born 1931–1941) and merged with the Study of Asset and Health Dynamics of the Oldest Old (born before 1924) in 1998 with the addition of two new birth cohorts, the Children of the Depression Age (born 1924–1930) and the War Baby (born 1942–1947) cohorts. Since 1998, the HRS has added a refresher cohort every 6 years to ensure generalizability to the U.S. population aged 51 and older. In 2004 and 2010, the Early Baby Boomer (born 1948–1953) and the Mid Baby Boomer (born 1954–1959) cohorts were added, respectively. Baseline response rates for new cohorts averaged around 73%, while re-interview response rates ranged from 85% to 90% between 1998 and 2014 (HRS Sample Sizes and Response Rates, 2017). The HRS collects information on marital status and children, making the data ideal for tracking respondents’ relationships with each child before and after gray divorce (repartnering). Information on respondents’ children was obtained from the RAND Family Data File 1992–2014 (V1).

We began by constructing a marital history file that captured the marriages and cohabitations formed and dissolved by all HRS respondents. Overall, 22,448 respondents aged 50 or older reported being married in 1998 or later. We restricted the sample to the 920 respondents who reported a divorce at age 50 or older. If respondents divorced more than once during the observation period, we used information from their first gray divorce. We removed respondents in a same-gender relationship (n = 6), those who had a sample weight equal to zero or missing throughout the observation period (n = 31), and those who reported a race–ethnicity other than White, Black, or Hispanic (n = 19). Next, we used the RAND Family Data File to link respondents to their biological children, dropping 64 respondents who reported having no biological children. We further limited our sample to observations in which biological children were aged 25 and older at the time of their parents’ divorce, resulting in the removal of 146 respondents. Defining adult children as aged 25 and older rather than using traditionally defined ages of adult status (e.g., 18 or 21 years old) is preferable given that many young adults remain dependent on their parents through their early 20s. Using age 25 to demarcate adult child status is consistent with other recent work using HRS data (Seltzer et al., 2013). Supplemental analyses indicated our results were robust to various specifications of adult child status, including defining adult children as those offspring aged 18 or older. Our final analytic sample consisted of 654 biological parents, 334 women and 320 men, all of whom eventually experienced a gray divorce. Of these, 86 women (25.75%) and 114 men (35.63%) repartnered following divorce.

Most respondents had multiple adult children and thus changes in parent–child ties surrounding gray divorce were not only contingent on the respondent’s characteristics but also conditioned by each child’s characteristics, individually as well as collectively. We employed a hierarchical, three-level approach that nested time within children within parents (see the Analytic Strategy section for detailed information). In total, the 334 women in our sample had 1,032 adult children, and the 320 men had 1,097 adult children. Respondents in re-interviewed households (as opposed to respondents in households newly entering the survey) in 1998, 2006, 2010, and 2014 were not asked about the frequency of contact with children. Thus, we did not have data for all respondents at all waves for our analysis of parent–child contact, resulting in a total of 3,126 adult child-year observations for men and 3,243 for women. However, in each wave of the HRS, respondents were asked about financial support to children. This complete information enabled us to use all nine waves of data between 1998 and 2014 for our analyses of changes in financial support both before and after gray divorce, resulting in a total of 5,156 adult child-year observations for men and 5,268 for women.

Measures

Frequent contact, a time-varying measure, captured whether each child had at least once a week contact (either in person or by phone, mail, or email) with the respondent during the past 12 months (1 = yes, 0 = no). Contact was gauged only for children who did not reside with the respondent. For adult children living with the respondent, we also included them in the analysis and coded them as having frequent (i.e., weekly) contact (Noël-Miller, 2013).

Downward financial transfers were obtained from respondents by asking them whether they gave financial help totaling $500 or more in the past 2 years to any of their children and if so, which child or children. For each adult child, we constructed a time-varying measure gauging whether they received financial support from the parent (1 = yes, 0 = no). Financial support was measured in the HRS for all children, regardless of residence status (lived with the respondent vs. not).

Time was measured in years, coded 0 when the respondent entered the sample, adding 2 years for each subsequent wave of observation.

Four time-varying variables captured the immediate and lasting effects of divorce and repartnering. Divorce was coded 0 until the respondent experienced a gray divorce, after which it was coded 1. Repartnering after divorce was coded 1 when the divorced respondent got remarried or began cohabiting with a new partner and in all subsequent waves, and 0 otherwise. These two variables gauged the immediate effects of divorce and repartnering, respectively. Years since divorce (repartnering) was included to measure the effect of divorce (repartnering) over time and was coded 0 at the time of divorce (repartnering), adding 2 years for each subsequent wave. For respondents who never repartnered during the observation period, years since repartnering was coded 0 throughout their observation.

We included several parental characteristics. Five time-invariant variables were measured at baseline. Marital duration was measured in years and indicated the length of the marriage that ended in gray divorce. Higher-order marriage was coded 1 if the marriage that ended in gray divorce was a remarriage and 0 if it was the first marriage. To avoid collinearity with the counting of time in the model, age was measured in years at baseline. Race–ethnicity was gauged as White (reference category), Black, or Hispanic. Educational attainment was a four-category variable: less than high school (reference category), high school, some college, or college or higher.

Five time-varying parental characteristics were also included. Employment status was a dichotomous measure capturing working versus not working (reference category). Household income was a logged measure of total household income using 1998 dollars, including the respondent’s earnings, employer pensions and annuities, social security income (i.e., supplementary, disability, and retirement), unemployment and worker’s compensation, veterans’ benefits, welfare, and food stamps, as well as capital income and other forms of income obtained in the previous year. Homeownership was a dichotomous measure, indicating whether the respondent owned a home. Self-reported health was a five-category ordinal measure that asked respondents to rate their general health from poor (coded 1) to excellent (coded 5). Nonadult children captured whether the respondent had any biological children younger than age 25 (1 = yes, 0 = no).

Several demographic characteristics of adult children also were included. Three time-invariant variables, child’s age (in years), gender (0 = daughter, 1 = son), and whether the child was born in the marriage that ended in divorce (i.e., shared child, = 1) or a prior marriage (= 0), were measured at baseline, while the remaining variables were time-varying. Partnership status was coded 1 if the child was married or cohabiting and 0 otherwise. Educational attainment was a continuous measure in years. Student status was a dichotomous measure indicating whether the child was attending school at any level. Employment status was captured by working versus not working (reference category). Homeownership was binary and measured whether the child owned a home. Parental status gauged whether they had any offspring.

Missing data were modest at the respondent level, ranging from 0.08% for self-reported health to 1.72% for marital duration, with the exception for homeownership (12.50%). Missing data were greater at the child level, ranging from 0.05% for children’s partnership status to 33.01% for children’s homeownership. Most imputation methods do not consider the clustered nature of multilevel data. Because our analyses were estimated using three-level nested data, we used a two-step imputation strategy to account for the hierarchical data structure. First, we imputed missing values at the respondent level using mean and mode. Second, we performed multiple imputation using chained equations for multilevel data, also known as multilevel fully conditional specification, at the child level. Auxiliary predictors in the chained equations included all respondent characteristics, all child characteristics, and within-respondent averages of children’s characteristics (i.e., group means). The within-family averages were included to account for the variance shared between children nested within the same respondent to preclude biased estimates in our multilevel models (Grund et al., 2018, 2019). We could not employ multiple imputation simultaneously at the respondent and child levels, as such an approach would have generated unbalanced respondent characteristics across children in the same family. In summary, the two-step strategy generated 10 random, multiply imputed data sets from which our analyses were drawn.

Analytic Strategy

Our analytic strategy was twofold. First, we compared baseline characteristics (prior to gray divorce) of fathers and mothers who reported experiencing a gray divorce between 1998 and 2014 using means or percentages as appropriate. This analysis also included a comparison of adult children’s characteristics for fathers versus mothers. Next, to investigate whether and how parent–adult child relationships change over time in response to gray divorce and repartnering, we estimated changes in the likelihood of (a) frequent contact with an adult child and (b) financial support to an adult child using three-level growth curve models (Rabe-Hesketh & Skrondal, 2012). The model specification is as follows:

$${\displaystyle \begin{array}{l}{\displaystyle {\mathrm{\eta }}_{ijk}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1^{\prime }{x}_{ijk}+{\mathrm{\beta }}_2^{\prime }{x}_{jk}+{\mathrm{\beta }}_3^{\prime }{x}_k+{\mathrm{\zeta }}_{jk}^{(2)}+{\mathrm{\zeta }}_k^{(3)}}\end{array}}$$

where ${\mathrm{\eta }}_{ijk}=\ln \left[\frac{{\mathrm{\pi }}_{ijk}}{1-{\mathrm{\pi }}_{ijk}}\right]$, ${\mathrm{\pi }}_{ijk}$ denotes the probability of respondent k reporting frequent contact or downward financial transfers at time i with adult child j. ${\mathrm{\beta }}_0$ is the intercept. $x_{ijk}$ is a vector of time-varying covariates, $x_{jk}$ is a vector of child-level characteristics, and $x_k$ is a vector of respondent-level characteristics. ${\mathrm{\beta }}_1^{\prime }$ includes the vector of regression parameters for the focal time-varying covariates at the child-year level (Level 1): time, divorce, repartnering after divorce, years since divorce, and years since repartnering. This vector also captures the parameters for all time-varying control variables, including parents’ employment status, household income, homeownership, self-reported health, and the presence of nonadult children as well as adult children’s partnership status, educational attainment, student status, employment status, homeownership, and parental status. ${\mathrm{\beta }}_2^{\prime }$ and ${\mathrm{\beta }}_3^{\prime }$ denote the remaining time-invariant regression parameters at the child level and respondent level, respectively. ${\mathrm{\zeta }}_{jk}^{\left(2\right)}\sim N(0,{\mathrm{\psi }}^{\left(2\right)})$ is a random intercept varying across adult children (Level 2) and ${\mathrm{\zeta }}_k^{(3)}\sim N(0,{\mathrm{\psi }}^{\left(3\right)})$ is a random intercept varying across respondents (Level 3). The random intercepts ${\mathrm{\zeta }}_{jk}^{(2)}$ and ${\mathrm{\zeta }}_k^{(3)}$ are assumed to be independent of each other and independent across respondents, and ${\mathrm{\zeta }}_{jk}^{(2)}$ is assumed to be independent across adult children as well. Both random intercepts are assumed to be independent of the covariates $x_{ijk}$, $x_{jk}$, and $x_k$. The responses are independently Bernoulli-distributed given the random effects and covariates. We conducted all analyses separately for men and women. We used a two-tailed test for the descriptive analysis and gender-specific models, but because we expected that the effect of gray divorce on parent–adult child relationships to be more negative for fathers than mothers, a one-tailed test was used for examining gender differences.

The descriptive analysis used baseline sample weights to adjust for the unequal probability of selection into the HRS (for Blacks, Hispanics, and Floridians) and nonresponse (Ofstedal et al., 2011). For the growth curve models, we applied normalized wave-specific weights to Level 1 and baseline sample weights for Level 3 of the model (Heeringa et al., 2017). The HRS does not provide child-level weights. However, child-level information is reported by respondents at each wave, meaning that children’s probability of selection into the sample and patterns of nonresponse mirror that of their parents. Thus, the wave-specific Level 1 weights effectively adjust for these issues at the child and respondent levels. We also estimated robust standard errors to account for the intraclustering correlations related to the HRS sampling strata. To control for potential biases arising from differential attrition from our sample, we included eight wave-specific dropout indicators (1 = dropout at a given follow-up, 0 = remains in the study) at Level 1 (Muthén et al., 2011). Analyses were conducted using the most recent version of Stata (16.1).

To pool parameter estimates across multiply imputed data, Stata users typically rely on the -mi estimate- command, which applies Rubin’s combination rules to adjust model parameters for the variability within and between imputations (Rubin, 1987). Although this command can be used for many multilevel regression commands (e.g., -mixed-), it does not yet support the regression command necessary for a weighted, three-level logistic regression (i.e., -melogit-). Therefore, we effectively performed the -mi estimate- command in Excel by pooling regression parameters and robust standard errors using Rubin’s combination rules (Rubin, 1987). To check our estimation technique against Stata’s -mi estimate- command, we regressed a random continuous variable at Level 2, adult children’s educational attainment, on respondents’ age at baseline (at Level 3) using both strategies. When using a continuous dependent variable, our pooled parameters and standard errors in Excel were identical to the output produced using the -mi estimate- command to the 10-millionths place.

Results

Parental and Adult Child Characteristics

Table 1 presents the means (or percentage distributions as appropriate) of all study variables measured at baseline prior to gray divorce. A lower percentage of fathers than mothers reported frequent contact with their adult children (63% vs. 78%, p < .001). However, approximately one in five parents gave financial support to their children irrespective of gender.

Table 1.

Weighted Means (standard deviations) or Percentages of Baseline (predivorce) Characteristics for Respondents Who Eventually Get Gray Divorce by Gender

	Men	Women
Dimensions of parent–adult child relationships
Frequent contact	63.28	78.45***
Downward financial transfer	20.98	21.71
Parental characteristics
Marital duration	17.16 (13.22)	17.11 (13.29)
Higher-order marriage	65.05	65.61
Age	59.16 (7.37)	56.20 (6.41)
Race–ethnicity
White	76.84	78.75
Black	13.89	12.34
Hispanic	9.27	8.91
Educational attainment
High school or lower	47.89	49.63
Some college	27.28	28.42
College or higher	24.83	21.95
Median household income (in 1998 dollars)	$49,360.00	$50,747.90
Employment status	58.82	54.14
Homeownership	71.77	74.66
Self-reported health	3.18 (1.20)	3.13 (1.22)
Nonadult children (<age 25)	42.21	37.37
Adult child characteristics
Age	32.15 (7.12)	32.06 (7.31)
Son	52.50	50.32
Shared biological child	42.44	44.11
Married or cohabiting	55.21	50.57
Educational attainment (years)	13.39 (2.31)	13.34 (2.36)
Student status	9.69	10.11
Employment status	78.98	81.14
Homeownership	34.10	34.38
Parental status	49.07	50.65
Unweighted number of respondents	320	334
Unweighted number of children	1,097	1,032

Note: Column totals may exceed 100% because of rounding. No means or percentages were significant at the p < .05 or p < .01 level.

Source: Health and Retirement Study, 1998–2014.

***p < .001.

Across all parental and adult child characteristics, no significant parental gender differences emerged. On average, the duration of the marriage ended in divorce was about 17 years at baseline and two thirds of these marriages were higher-order marriages. Parents were in their mid- to late 50s at baseline. More than three quarters of parents were White, 13% were Black, and 9% were Hispanic. Nearly one quarter of parents had a college degree or higher, with a median household income of $50,000 (in 1998 dollars). Around half of parents were currently employed and over 70% owned a home. Parents were generally in good health, and more than one third had a child younger than age 25 at baseline.

Adult children in the sample averaged 32 years of age at baseline and were equally split between sons and daughters, partnered and unpartnered, and having at least one child of their own versus no children. Just less than half of adult children were born in a marriage that ended in divorce. In general, adult children had achieved 13 years of education and only 10% were in school at baseline. The vast majority of adult children were employed and one third were homeowners.

Relationship Dynamics for Fathers and Their Adult Children

Table 2 depicts the odds ratios (ORs) from the growth curve models predicting the likelihoods of having frequent contact with and giving financial support to adult children for fathers. Gray divorce reduced the odds of frequent contact by 46%, but the odds of frequent contact gradually increased, about 10% per year, following divorce. Nonetheless, repartnering after gray divorce further reduced the odds of fathers’ frequent contact with their adult children, by about 15% a year. Notably, the transition to gray divorce almost doubled the odds of fathers’ providing financial support to their adult children (OR = 1.89).

Table 2.

Odds Ratios From Growth Curve Models Predicting Changes in Parent–Adult Child Relationship Over Time for Men

	Frequent contact	Downward financial transfer
Time	0.96	0.96
Union transitions
Divorce	0.54*	1.89*
Years since divorce	1.10*	0.95
Repartnering after divorce	0.98	0.48
Years since repartnering	0.85**	1.08
Parental characteristics
Marital duration	1.02	1.01
Higher-order marriage	1.33	1.06
Age	0.99	1.02
Race–ethnicity (ref. = White)
Black	0.84	0.55
Hispanic	2.51	0.43
Educational attainment (ref. = High school or lower)
Some college	1.70	2.28**
College or higher	3.95***	4.30***
Household income (in 1998 dollars, logged)	1.10	1.20
Employment status	0.81	1.38
Homeownership	0.91	1.64*
Self-reported health	0.91	1.29**
Nonadult children (<age 25)	0.84	0.94
Adult child characteristics
Age	0.97	0.92***
Son	0.52**	0.57**
Shared biological child	4.11***	1.98
Married or cohabiting	1.01	0.86
Educational attainment (years)	0.98	1.01
Student status	1.95*	1.71*
Employment status	1.18	1.07
Homeownership	1.04	0.61**
Parental status	1.56*	1.48
Intercept	1.53	0.01*
Random effects
Adult child level	2.15***	0.97***
Respondent level	2.98***	2.82***
Log likelihood	−6,301,850.39	−7,145,249.77

Source: Health and Retirement Study, 1998–2014.

*p < .05, **p < .01, ***p < .001.

Some of the fathers’ characteristics also are significantly associated with their relationships with adult children. Fathers with a college degree or higher were more likely than their less-educated counterparts to engage in frequent contact with their adult children. Fathers who had at least some college education, who owned a home, and who had better health were more likely to give financial support to their adult children compared with their respective counterparts.

Fathers were less likely to engage in frequent contact with and give financial support to sons than daughters. Fathers were also less likely to give financial support to older adult children but more likely to have frequent contact with and give financial support to adult children who were enrolled in school. In addition, fathers were more likely to be in frequent contact with adult children who were born in the marriage leading to divorce as well as adult children who had at least one child of their own compared with their respective counterparts. Adult children’s homeownership was inversely associated with fathers’ financial support.

Relationship Dynamics for Mothers and Their Adult Children

The multivariable results for mothers are given in Table 3. We uncover some evidence of intergenerational ties strengthening after divorce. Specifically, the transition to gray divorce increased the odds of frequent contact by 211%. However, downward transfers are impervious to gray divorce and repartnering. The odds of downward financial support were unrelated to divorce, years since divorce, repartnering after divorce, or years since repartnering.

Table 3.

Odds Ratios from Growth Curve Models Predicting Changes in Parent–Adult Child Relationship Over Time for Women

	Frequent contact	Downward financial transfer
Time	0.93	0.99
Union transitions
Divorce	2.11*	0.83
Years since divorce	1.03	1.01
Repartnering after divorce	0.42	0.83
Years since repartnering	1.04	0.91
Parental characteristics
Marital duration	0.98	0.99
Higher-order marriage	0.53	0.60
Age	1.09**	1.03
Race–ethnicity (ref. = White)
Black	0.60	0.40*
Hispanic	0.56	0.32**
Educational attainment (ref. = High school or lower)
Some college	0.35**	1.90
College or higher	0.32	7.72***
Household income (in 1998 dollars, logged)	1.04	1.11
Employment status	1.43	1.14
Homeownership	1.40	1.06
Self-reported health	1.12	1.06
Nonadult children (<age 25)	1.17	0.96
Adult child characteristics
Age	0.91**	0.96
Son	0.22***	0.48***
Shared biological child	1.21	0.74
Married or cohabiting	0.89	0.88
Educational attainment (years)	1.03	0.87**
Student status	2.30**	2.29**
Employment status	0.56**	0.82
Homeownership	0.93	0.49***
Parental status	0.94	1.40
Intercept	5.19	0.58
Random effects
Adult child level	3.99***	0.56**
Respondent level	2.08***	2.87***
Log likelihood	−5,285,444.290	−7,016,248.81

Source: Health and Retirement Study, 1998–2014.

*p < .05, **p < .01, ***p < .001.

Consistent with prior research, some of the mothers’ characteristics are significantly associated with their relationships with adult children. Older mothers were more likely to have frequent contact with their adult children than younger mothers, whereas mothers with some college were less likely to have frequent contact with their adult children compared with mothers who had a high school education or lower. Mothers with more education had a greater propensity of giving financial support to their adult children compared with those with less education. Black and Hispanic mothers were less likely to provide financial support to their children than were White mothers.

Adult children’s characteristics also relate to intergenerational ties. Mothers were less likely to engage in frequent contact with and give financial support to sons (vs. daughters) and less likely to engage in frequent contact with older children. As expected, being in school was positively correlated with mothers’ frequent contact and financial transfers. The odds of frequent contact with the mother were roughly 50% lower for adult children who were working versus not working. Mothers’ financial support was inversely associated with adult children’s educational attainment and homeownership.

We also tested gender differences in the effects of divorce, repartnering after divorce, years since divorce, and years since repartnering on frequent contact and downward financial transfers (results not shown). The likelihood of frequent contact differed by gender (p < .001), dropping precipitously after divorce for fathers but increasing for mothers. The odds of frequent contact after repartnering also varied by gender (p < .05) with sustained declines over time for fathers but not mothers. In contrast, fathers were more likely than mothers to provide financial support to their adult children upon gray divorce (p < .05).

Discussion

The doubling of the gray divorce rate, compounded with the aging of the population, foretells that an increasing number of adults will experience divorce in later life (Brown & Lin, 2012). Yet, little is known about the consequences of gray divorce and subsequent repartnering for parent–adult child relationships (Shapiro & Cooney, 2007). According to the life course perspective (Elder et al., 2003), life transitions in one generation are likely to influence the relationship dynamics with other generations. In this study, we considered the parent’s position in the life course and relationships with all adult children in the family to examine how gray divorce and subsequent repartnering are related to frequent contact and downward financial transfers, paying particular attention to gender differentials.

Consistent with prior studies showing that divorce creates a matrifocal tilt in our kinship system (Furstenberg et al., 1995), we found that gray divorce coincided with an increase in interaction between mothers and adult children, with the odds of frequent contact doubling upon divorce. Among fathers, gray divorce cut the odds of frequent contact by nearly half, although father–adult child frequent contact gradually increased thereafter. It is important to acknowledge that an increase in frequent contact does not necessarily mean that the mothers and their adult children become closer following gray divorce or experience improved relationship quality. Likewise, a drop in frequent contact may not reflect any change in the quality or closeness of the father–adult child relationship. In some cases, frequent contact can characterize ambivalent or even negative parent–child relationships (Birditt et al., 2010).

Regardless, the gendered patterns of frequent contact following gray divorce that we uncovered are consonant with prior work showing that adult children often blame fathers for the breakup (Cooney et al., 1986), weakening father–adult child contact immediately after gray divorce. However, children’s feelings of being caught between two parents tend to fade over time (Amato & Afifi, 2006), which may contribute to the gradual restoration of frequent father–adult child contact we observed. Repartnering further weakened father–adult child ties as the odds of frequent contact decreased by 15% per year afterward. This finding suggests fathers who experience gray divorce and then repartner are effectively swapping families. Upon repartnering, divorced fathers often shift their investments to the new spouse and their shared children or her children (Manning & Smock, 2000), reducing their reliance on adult children from the previous marriage for emotional support (Kalmijn, 2015) which in turn has an enduring negative effect on contact with adult children.

The likelihood of fathers’ provision of financial support to their adult children increased upon divorce, a finding that is contrary to previous research by Aquilino (1994) but consistent with Shapiro and Remle (2011). Divorced fathers often increase their financial support of adult children in the face of losing frequent contact with them, perhaps reflecting either feelings of guilt or attempts to maintain their provider role (Troilo & Coleman, 2012). However, the increase in the likelihood of downward financial transfers that occurred upon gray divorce did not persist over time. And, downward financial transfers by mothers remained unchanged by either gray divorce or repartnering.

Our study design improves on prior research by following respondents prior to, during, and after gray divorce for a longer period of time; conducting within-person comparisons; and considering all adult children of respondents in the analysis. Nevertheless, our study also has limitations that merit future research. For instance, consistent with a life course perspective that emphasizes the salience of the timing of events, we expected the effects of gray divorce on parent–child relationships to be weaker than those of early-life divorce on minor children, given that adult children are likely to be independent, emotionally mature, and in families of their own that offer a support system, but we are not able to compare gray divorce with early-life divorce using the HRS. Fathers’ decreased contact with adult children echoes the pattern of father–child relationships observed in early-life divorce (Seltzer, 1994). In contrast, fathers’ increased odds of making downward transfers to their adult children following gray divorce contrast with the high levels of noncompliance with child support orders among many nonresident fathers following early-life divorce (Seltzer, 1994). Still, the growth in contact between mothers and their adult children following gray divorce aligns with prior work showing particularly close relationships between single mothers and their children (McLanahan & Booth, 1989). Moreover, individuals are situated within a multigenerational structure (Shapiro & Cooney, 2007) of linked lives and thus the effect of divorce may manifest itself across generations. We examined relationship dynamics between gray-divorced adults (G2) and their adult children (G3) but did not consider how gray divorce affects G2’s relationships with their own aging parents (G1). For instance, gray divorce may reduce G2’s resources to help their aging parents (G1) or it may increase downward transfers from aging parents (G1) to gray-divorced adults (G2) to buffer economic and emotional losses. These questions are beyond the scope of the current study but are pivotal to further unpack the consequences of gray divorce for intergenerational ties.

Despite these limitations, this study sheds new insight on the roles of gray divorce and subsequent repartnering for parent–adult child relationships from the parent’s perspective. Congruent with the few studies examining intergenerational relationships from the adult child’s perspective, we also find that gray divorce and subsequent repartnering have unique effects for fathers and mothers. Whereas father–adult child frequent contact diminishes upon gray divorce, their downward financial transfers to adult children increase, at least in the short term. Repartnering, experienced by about one third of men following gray divorce (Brown et al., 2019), is linked to reduced contact with adult children. In contrast, mother–adult child frequent contact rises upon gray divorce even as their financial transfers to their adult children remain unaltered. Uncommon among women, repartnering has no appreciable effects on their relationships with their adult children. These patterns raise new questions about whether and to what extent gray-divorced fathers and mothers can rely on their adult children for care and support as they age.

Funding

This research was supported by a grant from the National Institute on Aging (R15AG047588 to S. L. Brown and I.-F. Lin). Additional support was provided by the Center for Family and Demographic Research, Bowling Green State University, which has core funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (P2CHD050959). Any opinions expressed here are solely those of the authors and not of the funding agency or center.

Conflict of Interest

None declared.

Author Contributions

I.-F. Lin and S. L. Brown conceived and planned the study. K. A. Mellencamp prepared the data file and performed the analyses with input from I.-F. Lin and S. L. Brown. All three authors wrote and revised the manuscript.