Does the Transition to Grandparenthood Deter Gray Divorce? A Test of the Braking Hypothesis

Abstract

The gray divorce rate, which describes divorce among individuals aged 50 and older, has doubled since 1990. Extending prior research that showed the transition to parenthood has a “braking effect” on divorce, we examined whether the transition to grandparenthood, an emotionally meaningful midlife event that typically renews midlife marriages, exerts an analogous “braking effect” on gray divorce. Using panel data from the 1998–2014 Health and Retirement Study, we found that becoming biological grandparents has a large deterrent effect on gray divorce that persists even after accounting for a host of other factors known to be associated with divorce. However, the transition to step grandparenthood has no protective effect on gray divorce. Our study demonstrates the importance of the larger family system and in particular the life webs connecting the generations for promoting marital stability among midlife couples.

During an era marked by stability in the US divorce rate, divorce among adults aged 50 and older, termed “gray divorce,” has climbed dramatically since 1990 (Brown and Lin 2012; Kennedy and Ruggles 2014). The gray divorce rate doubled from roughly 5 divorces per 1,000 marriages in 1990 to 10 per 1,000 in 2010. This doubling of the gray divorce rate coupled with the aging of the population translates into sizeable growth in the proportion of the divorcing population that is aged 50 or older. In 1990, a mere 8 percent of those getting divorced were in this age category compared with more than one-quarter in 2010 (Brown and Lin 2012).

The rapid acceleration in gray divorce raises new questions about the unique predictors of divorce for this age group. Scholars have theorized that midlife transitions, such as an empty nest, retirement, or declining health, could be key to deciphering the gray divorce phenomenon (Wu and Schimmele 2007; Lin et al. 2018b). These turning points often coincide with spouses coming to the realization that they have grown apart, and this in turn spurs gray divorce, according to recent qualitative work (Bair 2007). However, longitudinal research has found no appreciable linkages between either empty nest or retirement and gray divorce (Lin et al. 2018b) and only mixed evidence of any link between declining health and gray divorce (Karraker and Latham 2015; Lin et al. 2018b). Here, we consider a key milestone of midlife that may be protective against gray divorce: the transition to grandparenthood.

Drawing on the well-established finding that the transition to parenthood exerts a “braking effect” on divorce for married couples of childbearing age (Cherlin 1977; White and Booth 1985b), we extend this logic by examining whether the transition to grandparenthood during midlife has an analogous deterrent effect on gray divorce. Becoming a grandparent is a significant life transition typically marked by feelings of generativity and providing new meaning to life (Cunningham-Burley 1986), which in turn may strengthen and renew midlife marriages. As a living embodiment of the continuity of the family lineage into the future, the grandchild represents a shared project for the grandparents (cf. Berger and Kellner 1964) and thus may promote marital stability by inhibiting couples from growing apart (Bair 2007), ultimately having a braking effect on gray divorce.

We use panel data from the 1998–2014 Health and Retirement Study (HRS) to investigate whether the transition to grandparenthood among married couples with at least one adult child is a deterrent to gray divorce. The HRS is ideally suited for our study because it contains a large sample of married couples followed prospectively, allowing us to employ a couple-level modeling approach that includes time-varying covariates. Importantly, the HRS tracks the number of offspring from each of the couple’s children at every wave, and thus we can identify when the transition to grandparenthood occurs. It also permits us to address whether the transition to grandparenthood operates differently for couples with shared biological versus step adult children. It is possible that becoming a step grandparent does not have as large of a braking effect on divorce as biological grandparenthood given that the adult child is the offspring of only one of the spouses and thus the grandchild is less likely to be a shared project for the couple. The findings from this study contribute to our understanding of how intergenerational relationships, namely, grandparenthood, can shape couple relationship dynamics and the risk of divorce in later life.

Background

The recent rise in gray divorce reflects a confluence of factors first articulated more than 30 years ago, when several scholars asserted that divorce among older adults would be increasingly common in the coming decades (Berardo 1982; Uhlenberg and Myers 1981; Hammond and Muller 1992). A leading contributor is the growth in higher-order marriages, which coincides with the divorce revolution of the 1970s. Many of the Baby Boomers who got divorced as young adults eventually remarried, and remarriages are much more likely to end in divorce than first marriages (Amato 2010; Brown and Lin 2012). Remarriages are at greater risk of divorce because they tend to be of poorer spousal match quality (i.e., spouses are less homogamous) and they are often destabilized by the relationship challenges that accompany the formation of a stepfamily. Whereas some fraction of first married couples are in empty shell marriages that are unsatisfying but which they are unwilling to leave, individuals who have been divorced previously are more willing to divorce again in the event of an unhappy marriage (Uhlenberg and Myers 1981; Wu and Schimmele 2007). Indeed, the gray divorce rate is 2.5 times higher for those in remarriages versus first marriages (Brown and Lin 2012). Another factor is the gains in women’s employment, which have bolstered wives’ economic independence and thus made divorce a viable option. Rising life expectancy has decreased the likelihood of spousal death and increased the length of exposure to divorce. Additionally, longer life expectancies have changed the calculus about marriage in later life. Individuals are more reluctant now to remain in unsatisfactory marriages given a longer time horizon (Wu and Schimmele 2007).

More broadly, the changing meaning of marriage is altering expectations for what constitutes marital success. This paradigm shift towards the deinstitutionalization of marriage applies not only to young adults but to older adults, too (Wu and Schimmele 2007; Brown and Wright 2019). We are in an era of individualized marriages, in which spouses emphasize self-fulfillment, flexible gender roles, and open communication (Cherlin 2004). The stigma of divorce has receded, and nowadays most older adults agree with the notion that divorce is the best option when couples do not get along (Brown and Wright 2019). Wu and Schimmele (2007:42) argued “the ideal of lifelong marriage is becoming obsolete because the demand for receiving personal gratification through marriage appears incompatible with single, lifelong marriages…[,] straining numerous long-term marriages to their breaking points.”

In-depth interviews conducted by Bair (2007) revealed that gray divorce often occurs because couples drift apart. Spouses change and evolve during the course of a marriage, which can diminish their compatibility. Couples once consumed by their careers and children may find they have less in common once they retire and start spending large chunks of time alone together. This qualitative evidence has not been supported by population-based research using nationally representative samples though. Neither the transition to an empty nest nor retirement is associated with gray divorce, perhaps because these events have mixed effects (Lin et al. 2018b). For couples that have grown apart, an empty nest or retirement may spur divorce, but for other couples, namely, those still in love and committed to their marriages, these life transitions are actually likely to strengthen the marital bond by providing the couples with more opportunities to spend time alone together without the distractions of children or work.

Becoming a Grandparent

The transition to grandparenthood is a significant life event that most midlife and older adults view quite positively because it creates meaningful ties across the generations, building intergenerational solidarity, and is concrete evidence of familial continuity, providing a sense of immortality (Neugarten and Weinstein 1964; Cunningham-Burley 1986; Hagestad and Lang 1986; Taubman-Ben-Ari, Shlomo, and Findler 2018). By introducing new meaning to life, becoming a grandparent is arguably a transformative experience. It is certainly unique in that it is not self-initiated. Life course scholars have labeled becoming a grandparent a “countertransition” because it depends upon someone else’s transition, that is, the offspring becoming a parent (Hagestad and Lang 1986). It is also emblematic of the life course principle of “linked lives,” which describes familial relationship ties within and across the generations (Elder 1994) that shape life events such as divorce.

Grandparenthood is typically a distinct phase of the life course that does not overlap with parenting (Hagestad and Burton 1986). Rising ages at childbirth and smaller family sizes have delayed entry into grandparenthood, whereas longer life expectancies are lengthening time spent in grandparenthood (Leopold and Skopek 2015; Margolis 2016; Margolis and Verdery 2019). By the time they become grandparents, most people are no longer rearing minor children (Sprey and Matthews 1982; Leopold and Skopek 2015; Marhánková 2015). These various family demographic trends mean that more generations are alive at a given point in time and they are spending greater proportions of their lifetimes with one another, underscoring the growing salience of the transition to grandparenthood.

Unlike parenting, which grandparents recall as a stressful and often overwhelming experience, grandparenting is viewed as a rewarding status for several reasons. Grandparents have less anxiety about children having successfully raised their own. They also have accrued more financial resources and have fewer competing demands on their time since they are less likely to be working. Grandparenting affords more opportunities for fun and enjoyment with fewer responsibilities. The birth of a first grandchild often stirs feelings of youthfulness among new grandparents and ultimately imparts new meaning to life (Cunningham-Burley 1986; Hagestad and Lang 1986).

The transition to grandparenthood is not merely an individual event but a familial one, which links together three generations, reinforcing both horizontal and vertical family ties (Sprey and Matthews 1982). A shared life experience that signals the continuity of the family, becoming grandparents, reorients the culture of a marriage towards the new grandchild (Neugarten and Weinstein 1964; Kahana and Kahana 1971). The demographic metabolism of the family (Ryder 1965) is altered as new life webs (Pruchno, Blow, and Smyer 1984) are created and the roles of each generation shift in turn as grandparents become great-grandparents, aging parents become grandparents, and adult children are now parents.

The Braking Hypothesis

Berger and Kellner (1964) asserted that children and other relatives inform a couple’s marital reality, and, by extension, we surmise that grandchildren, particularly the birth of the first grandchild, are likely to have similar effects. Becoming grandparents reshapes the marital reality, prompting couples to reconstruct both the past and present to integrate the cultural script of grandparenthood (Cunningham-Burley 1986). New grandparents report gains in their familial relationships with some pointing to the transition as “cementing” their spousal relationship (Taubman-Ben-Ari et al. 2018:95).

By reorienting and reinforcing the marital relationship, the transition to grandparenthood may deter divorce, exerting a “braking effect” akin to the stabilizing influence that the birth of a first child has on the marriage of a couple of childbearing age (Cherlin 1977; Glenn and McLanahan 1982; White and Booth 1985b). White and Booth (1985b) uncovered no appreciable relationship between the transition to parenthood and multiple dimensions of marital quality, but couples who became parents were significantly less likely to get divorced in the next few years than their counterparts who remained childless. Here, we extend this logic to assess whether the birth of a first grandchild functions as a divorce deterrent.

Biological Versus Step Grandparenthood

Stepfamilies are common in middle and later life. Nearly two in every five (37 percent) married couples aged 50 and older with children are stepfamilies (Lin, Brown, and Cupka 2018a). Step grandparenthood, defined as an adult stepchild having a child,¹ is apparently on the rise as its prevalence is highest for more recent cohorts of older adults (Yahirun, Park, and Seltzer 2018). About 17 percent of women and men in their 60s are step grandparents (Yahirun et al. 2018). On the one hand, the transition to step grandparenthood may have a weaker effect than the transition to shared biological grandparenthood because the grandchild is a step relation for one of the spouses, potentially diminishing the strength of the life webs established by the birth of the grandchild. On the other hand, the transition to step grandparenthood could create conflict between spouses and thus actually destabilize marriages. Step grandparenthood may alter the marital reality, shifting the focus from the couple’s marital tie to the spouse’s filial relationship with the biological adult child who has produced the grandchild, a reorientation that is likely to be more fulfilling for the spouse who is the biological grandparent than the step grandparent. Indeed, the presence of stepchildren is positively associated with divorce among those in remarriages, and thus by extension the birth of a step grandchild may heighten the risk of divorce for older couples (White and Booth 1985a).

The Present Study

Gray divorce accounts for one in four divorces today. Although the gray divorce rate has doubled since 1990, the risk and protective factors for gray divorce have received limited attention in the literature thus far (Brown and Lin 2012; Karraker and Latham 2015; Lin et al. 2018b). Moreover, the few studies to date have been largely unsuccessful in identifying the unique correlates of gray divorce. Despite qualitative work on the centrality of midlife transitions that serve as turning points prompting couples to assess the viability of their marriages (Bair 2007), quantitative research using nationally representative data has not uncovered any appreciable relationships between transitions either to an empty nest or to retirement and gray divorce (Lin et al. 2018b).

We assert that prior studies have overlooked a significant transition that marks midlife and may be protective against gray divorce: the transition to grandparenthood. Becoming grandparents is a life-affirming transition that strengthens familial ties and the marital bond (Taubman-Ben-Ari et al. 2018). Grandparenthood often spurs a new outlook on life, giving grandparents a renewed sense of purpose, youthfulness, and even immortality (Neugarten and Weinsteain 1964; Cunningham-Burley 1986). In short, the birth of a first grandchild can breathe new life into long-term midlife marriages that have become routinized or boring (cf. Wu and Schimmele 2007). If many midlife marriages succumb to growing apart (Bair 2007), then the transition to grandparenthood should act as a countervailing force that alters the marital reality and reinforces the marital tie for couples who were drifting apart, slowing the march towards divorce by exerting a braking effect (White and Booth 1985b).

Our goal is to examine whether the transition to grandparenthood is negatively associated with gray divorce among married couples (in which one spouse is at least aged 50 or older) who are eligible to become grandparents, that is, who have at least one child aged 16 or older. We distinguish between the transitions to shared biological versus step grandparenthood since stepchildren are linked to poorer family functioning and marital instability (Ganong and Coleman 2017). Using a discrete-time event history modeling approach, we account for numerous time-invariant and time-varying factors that have been considered in prior research on the antecedents of gray divorce, including marital biography, marital quality, spousal homogamy, economic resources, and life transitions (Lin et al. 2018b). Marital biography factors include marital duration and marriage order. The risk of gray divorce diminishes as marital duration lengthens, and this association is confounded with the positive relationship between remarriage and gray divorce (Brown and Lin 2012). Husbands and wives who report higher-quality marriages are less likely to experience gray divorce (Lin et al. 2018b). Spousal homogamy promotes marital stability. When husbands are a lot older than their wives, gray divorce is more likely (Karraker and Latham 2015). Likewise, interracial couples are at higher risk of gray divorce than same-race couples (Lin et al. 2018b). Economic resources, such as wealth, are negatively related to gray divorce (Lin et al. 2018b). Finally, midlife transitions including an empty nest, retirement, and declining health appear to be largely unrelated to gray divorce (Lin et al. 2018b), although qualitative studies identify these transitions as pivotal to the decision to call it quits (Bair 2007), and thus we control for them in our study.

Method

Data came from the 1998–2014 HRS, a nationally representative survey of a continuous cohort of US adults born before 1960 and their spouses. We focused on individuals in the original HRS (born 1931–1941) and War Babies (WB; 1942–1947) cohorts, excluding older cohorts (AHEAD and CODA) that did not enter the study until age 68 or older, well beyond the midlife age range (50–64) at which gray divorce most often occurs. The two more recent cohorts added to the HRS in 2004 and 2010, respectively, were excluded because they were not asked the same marital quality questions as were the HRS and WB cohorts. Initial response rates for the HRS and WB cohorts were 82 and 70 percent, respectively, with reinterview response rates over 90 percent. Designed to examine the economics, health, and family ties of older adults, the HRS is well-suited for our purposes because it includes detailed marital histories for both respondents and their spouses. Additionally, at every wave, each couple reported on their children and their children’s children, which means we can capture the transition to grandparenthood. Sampling weights were used to adjust for the unequal probability of selection (for Blacks, Hispanics, and Floridians), nonresponse, and sample attrition (Ofstedal et al. 2011).

In total, there were 37,495 respondents on the RAND HRS 1992–2014 file. We restricted our sample to the 16,390 respondents who belonged to the HRS and WB cohorts. Among them, there were 5,832 marriages observed in 1998 or later. For respondents who had multiple marriages in 1998 or later, we kept only their first marriages, yielding 5,584 married couples. We also excluded same-sex couples (n = 14), couples interviewed only once (n = 193), and couples with a baseline weight of 0 (n = 36). Finally, we removed couples in which spousal and/or child information was missing (n = 149), couples with no children age 16 or older during our observation (n = 29), and couples who already had a grandchild in 1998 (n = 4,118), resulting in 1,045 couples (7,587 couple-interview-year observations) in our analytic sample. Of the 1,045 couples, 719 experienced a transition to any grandparenthood, whereas 326 did not. Of the 719 grandparent couples, 521 transitioned to biological grandparenthood, and 198 transitioned to step grandparenthood.

Measures

Divorce, the dependent variable, was coded 1 if the couple reported having gotten divorce at the interview and 0 otherwise.

The focal independent variable was the transition to grandparenthood. Two variables measured the two possible pathways: biological grandparenthood was coded 1 when a first grandchild to a child who was biologically related to both spouses was reported (and 0 otherwise), and step grandparenthood was coded 1 when a first grandchild to a child who was biologically related to only one of the two spouses was reported (and 0 otherwise).² The reference category for these two variables is no transition to grandparenthood.

Marital biography consisted of marital duration and marriage order. Marital duration was a time-varying covariate indicating the current length in years of the marriage. Remarriage (time-invariant) was coded 1 if for at least one spouse it was a remarriage and 0 when both spouses were in their first marriage.

Marital quality captured patterns of marital interaction as reported independently by husbands and wives at baseline interview. The measure was constructed using two items. One item gauged how husbands and wives spent their free time: 1 = mostly together; 2 = some together, some apart; and 3 = mostly apart. The other item measured how much they enjoyed the time they spent together: 1 = extremely enjoyable, 2 = very enjoyable, 3 = somewhat enjoyable, and 4 = not too enjoyable. Following prior research (Bulanda 2011; Lin et al. 2018b), we combined not too enjoyable with somewhat enjoyable since very few respondents characterized their time together as not too enjoyable. Responses to these two items were summed and reverse coded such that higher values correspond with more positive marital quality, with values ranging from 2 to 6.

Spousal homogamy encompassed spousal age, race, and education. Husband’s age was measured in years. It was a time-varying covariate. Age homogamy captured the difference between the husband’s and wife’s ages (time-invariant). Racial homogamy distinguished among couples in which both spouses were White (1 = yes, 0 = no), both spouses were non-White (1 = yes, 0 = no), or spouses were of different racial backgrounds (1 = yes, 0 = no). A continuous measure of husband’s education captured years of schooling, ranging from 0 to 17. Educational homogamy was the difference between the husband’s and the wife’s years of schooling. Both of these education measures were time-invariant.

Family composition included the numbers of shared and non-shared children that the couple had. Number of shared children was a continuous variable that indicated the number of children aged 16 and older that were the biological offspring of both spouses. Number of non-shared children was a continuous variable that represented the number of children aged 16 and older that were the stepchildren of at least one (or both) of the spouses.

A time-varying covariate, home ownership, captured whether the couple owned their home (1 = yes, 0 = no). The couple’s wealth was a time-varying measure that classified couples into one of five categories: in debt, $0–$50,000 (reference category), $50,001–$100,000, $100,001–$250,000, and $250,001 or more. These economic resources measures were lagged to ensure temporal ordering prior to divorce.

We followed the coding strategy used by Lin et al. (2018b) to measure life course transitions, lagging these transitions by one wave to ensure their temporal order with divorce. The occurrence of an empty nest was coded 1 when there were no offspring in the household and 0 otherwise. The onset of the husband’s and wife’s retirement was coded 1 when they reported being (partly) retired and 0 otherwise. The onset of husband’s and wife’s chronic conditions was coded 1 when respondents reported a doctor had diagnosed them with heart disease, cancer, lung disease, or a stroke and 0 otherwise. Once a life course transition occurred, it was assumed to persist across all subsequent waves of observation.

Interview year was a series of time-varying binary variables that tracked each interview wave the couple was at risk of gray divorce (2000 was the reference category).

Analytic Strategy

First, we estimated the survival curve for couples by marital duration to examine whether and how the transition to grandparenthood, distinguishing between transitions to biological grandparenthood and step grandparenthood, alters the duration-dependent risk of gray divorce. We anticipated that at a given duration, transitioning to biological grandparenthood would be associated with a greater probability of survival, or not experiencing a gray divorce. The transition to step grandparenthood may have a weaker positive association with survival or possibly even a negative association.

Second, we calculated means and distributions (as appropriate) for all study variables separately by whether the couple experienced the transition to biological grandparenthood and step grandparenthood or did not become grandparents to assess how the three groups differ across the various sets of factors associated with gray divorce at baseline.

Third, we estimated a discrete-time event history model predicting the risk of gray divorce using logistic regression. Discrete-time modeling permits the inclusion of time-varying covariates (Allison 1982), and thus we can explicitly account for dynamic features of the marriage, including key life course transitions and changing economic resources. Additionally, discrete-time modeling is more appropriate than continuous time modeling because marriage start and end dates were measured as time intervals. Log-linear methods are used for model estimation, which is specified as follows:

is the hazard rate, defined by , where T is the discrete random variable giving the uncensored time of event occurrence (Allison 1982). In other words, is the conditional probability that divorce occurs to couple i at time t, given that it has not already occurred. We consider how this hazard rate is a function of time () and a vector of explanatory variables (, with its coefficient vector ). Couples were observed from the earliest time point at which they were (1) married and (2) at least one spouse was aged 50 or older. All couples entered the analysis beginning with the first interview at which they were married (1998 or later). They were censored once they divorced, when one of the spouses died, or at the 2014 interview (or attrition). We relied on a straightforward modeling approach in which the transition to grandparenthood and all covariates were inserted simultaneously to assess the net effect of becoming grandparents on the likelihood of gray divorce.

Finally, we conducted two robustness tests to evaluate the potential biases associated with excluding from the analytic sample those married couples that were already grandparents at first observation. First, we pooled our analytic sample (n = 1,045) with those who had already become grandparents (n = 35,908 couple-interview years) and estimated the risk of divorce for three groups: already grandparents at baseline, those who transition to grandparenthood during the observation period, and those who do not make the transition during the observation period. If the first two groups are not significantly different from each other and both exhibit lower odds of divorce than the third group, then this lends credence to the robustness of our original results. Second, we adjusted our original models for nonrandom selection into grandparenthood using a Heckman (1979) two-step approach. The first step involved predicting membership in the analytic sample versus the already grandparents at baseline group, using key covariates (marital biography, spousal homogamy, and family composition variables) along with an additional variable related to grandparenthood status but unrelated to divorce (and therefore not included in the second equation predicting gray divorce). This additional variable was age of the oldest child. Its correlation with grandparenthood status (1 = not already grandparents; 0 = already grandparents) was large at −0.50. Its correlation with divorce was negligible at −0.07. In the second step predicting the likelihood of divorce, all of the original covariates were included along with the inverse Mills ratio generated from the first equation to correct for selectivity into grandparenthood. A significant inverse Mills ratio would signal some selectivity in becoming grandparents.

A multiple imputation procedure, multivariate imputation using chained equations, which is Stata’s mi impute chained command, was used to handle minimal missing data for our analytic sample of 1,045 couples. The results presented here were based on five random, multiple-imputed replicates. All estimates were weighted using the baseline weights.

Results

Figure 1 shows the cumulative probability of gray divorce by whether the couple becomes grandparents. Transitioning to biological grandparenthood was protective against gray divorce, as these couples enjoyed a significantly lower risk of divorce regardless of marital duration. In contrast, the risk of gray divorce rose precipitously with marital duration among those couples that experienced a transition to step grandparenthood. For couples married between 15 and 35 years, the transition to step grandparenthood actually was associated with a larger risk of gray divorce than the absence of a transition, with the cumulative risks for the two groups converging at higher marital durations. Ultimately, the cumulative probability of divorce among those who became biological grandparents was negligible at just over 1 percent versus about 12 percent for those who either transitioned to step grandparenthood or did not become grandparents.

Figure 1.

Cumulative probability of divorce by grandparenthood transition type

Bivariate Results

The differentials according to grandparenthood status were evident across several domains, as shown in Table 1. Couples that transitioned to biological grandparenthood had been married about 30 years at baseline interview compared with 11 years for couples who transitioned to step grandparenthood and 25 years for those who did not become grandparents. About 13 percent of couples that became biological grandparents were in a remarriage versus nearly all (97 percent) of those who became step grandparents. Of those that did not transition to grandparenthood, 30 percent were in a remarriage. The three groups did not appreciably differ in terms of marital quality.

Table 1.

Weighted Means (Standard Deviations) or Percentages of Couples’ Characteristics at Baseline by Grandparent Transition Type

	Biological grandparenthood	Step grandparenthood	No transition
Dependent variable
Divorce	1.60 * * *	12.39	11.59
Marital biography
Marital duration	29.94 (16.47)***	11.36 (25.14)***	25.13 (10.29)
Remarriage	12.92 * * *	96.70***	30.16
Marital quality
Wife’s report of marital interaction	4.48 (2.32)	4.63 (3.54)	4.40 (1.16)
Husband’s report of marital interaction	4.56 (2.22)	4.67 (3.35)	4.53 (1.10)
Spousal homogamy
Husband’s age	55.97 (11.31)	55.36 (17.83)	56.56 (6.22)
Difference in husband’s and wife’s ages	3.15 (10.30)*	6.07 (19.28)*	4.25 (6.30)
Husband’s years of education	14.17 (5.57)	14.14 (8.59)	14.32 (2.78)
Difference in husband’s and wife’s years of education	0.36 (5.19)	0.21 (7.96)	0.16 (2.68)
Both spouses are White	86.33	79.77	82.95
Both spouses are non-White	10.62	9.93	10.76
Spouses have different races	3.05 *	10.30	6.28
Family composition
Number of shared children	2.32 (2.01)***	0.29 (2.79)***	1.35 (0.89)
Number of non-shared children	0.09 (1.51)***	2.93 (3.38)***	0.41 (1.03)
Economic resources
Home owner	94.54	87.97	91.49
In debt	2.65	4.15	4.28
0 to $50,000	34.80	28.53	29.48
$50,001– $100,000	11.11	7.54	10.37
$100,001–$250,000	21.09	22.78	27.20
$250,001 or more	30.95	37.00	28.67
Life course transitions
Empty nest	47.72***	46.90**	30.99
Wife retired	12.33	7.93	9.53
Husband retired	16.68	16.64	17.58
Wife’s chronic conditions	12.17	7.24*	14.14
Husband’s chronic conditions	14.72	20.62	18.95
Weighted percentages	49.05	19.11	31.85
Unweighted number of couples	521	198	326

^* P < .05, ^**P < .01 and ^***P < .001 (two-tailed tests) with no transition as the reference category.

Bolded numbers indicate that biological grandparenthood significantly differs from step grandparenthood (reference category) at P < .05.

With respect to spousal homogamy, two significant differences emerged among the three groups. The age difference between husbands and wives was smaller for those who became biological grandparents (3 years) than those who became step grandparents (6 years). Likewise, both groups differed from their counterparts who did not become grandparents, for whom the average age difference was about 4 years. Those who transitioned to biological grandparenthood were disproportionately both White (86 versus 80 percent) and unlikely to be part of an interracial couple (3 versus 10 percent) relative to those who became step grandparents. Family composition varied across the three groups. The average number of shared (i.e., biological) children was 2.3 for couples who transitioned to biological grandparenthood, followed by 1.3 for couples that did not transition to grandparenthood, and finally just 0.29 for couples who transitioned to step grandparenthood. The mean number of non-shared children (i.e., stepchildren) ranged from a high of 2.9 for couples that transitioned to step grandparenthood to 0.41 for those who experienced no transition, to just 0.09 for couples that transitioned to biological grandparenthood.

For the most part, the economic resources of the various groups were comparable. The only difference was couples who transitioned to biological grandparenthood more often owned their home than did couples who transitioned to step grandparenthood (95 versus 88 percent). The groups did not differ on household income. Few differences were evident in life course transitions. About 47 percent of those who became grandparents, whether biological or step, also had experienced the onset of an empty nest by baseline interview compared with just 30 percent of couples that did not become grandparents.

Multivariable Regression Results

As shown in Model 1 of Table 2, the logistic regression estimates from the discrete-time event history model indicate that the transition to biological grandparenthood was negatively associated with gray divorce, whereas the transition to step grandparenthood was positively related to the risk of gray divorce. The odds of divorcing were 73 percent lower for couples who became biological grandparents versus those that did not become grandparents. In contrast, couples who transitioned to step grandparenthood experienced odds of gray divorce that were more than twice as high as those who did not become grandparents. The two types of transitions differed significantly from each other such that couples who became biological grandparents were less likely to divorce than those who became step grandparents (as indicated by the bolded coefficients in Table 2).

Table 2.

Odds Ratios from Discrete-Time Logistic Regressions Predicting Divorce (n = 7,587 Couple-Interview Years)

	Model 1	Model 2
Transition to grandparenthood
Transition to biological grandparenthood	0.27 * * *	0.41 *
Transition to step grandparenthood	2.16 *	1.58
No transition (reference)
Marital biography
Marital duration		0.97
Remarriage		0.69
Marital quality
Wife’s report of marital interaction		0.71*
Husband’s report of marital interaction		0.74*
Spousal homogamy
Husband’s age		0.95
Difference in husband’s and wife’s ages		1.00
Husband’s years of education		1.06
Difference in husband’s and wife’s years of education		0.94
Both spouses are White (reference)
Both spouses are non-White		1.26
Spouses have different races		1.76
Family composition
Number of shared children		1.05
Number of non-shared children		1.21*
Economic resources
Home owner		0.30**
In debt		1.18
0–$50,000 (reference)
$50,001–$100,000		1.01
$100,001–$250,000		0.96
$250,001 or more		0.93
Life course transitions
Empty nest		0.71
Wife retired		1.06
Husband retired		1.27
Wife’s chronic conditions		1.19
Husband’s chronic conditions		1.24
Interview year
2000 (reference)
2002	2.47	2.85
2004	0.94	1.20
2006	1.19	1.80
2008	1.08	1.77
2010	0.79	1.48
2012	1.40	3.03
2014	0.34	0.84
Constant	0.01***	6.13
Model fit statistic	F (9, 49.1) = 5.49	F (31,47.4) = 37.92

^* P < .05, ^**P < .01 and ^***P < .001 (two-tailed tests).

Bolded coefficients within the model are significantly different, P < .05.

In the full model (Model 2), the transition to biological grandparenthood remained negatively associated with gray divorce (odds ratio = 0.41), but the transition to step grandparenthood had no appreciable effect on the likelihood of divorce (odds ratio = 1.58). However, the two types of transitions continued to differ significantly from each other with the transition to biological grandparenthood more protective against divorce than the transition to step grandparenthood.³ In short, the transition to biological grandparenthood exhibited a braking effect on gray divorce that is much like the effect established for the transition to parenthood among couples of childbearing age (White and Booth 1985b). This strong association was evident net of controls for marital biography, marital quality, spousal homogamy, family composition, economic resources, and other life transitions. Some of these factors also were associated with the risk of gray divorce. Specifically, both the wife’s and the husband’s reports of marital interaction were negatively associated with divorce. The number of stepchildren was positively related to the risk of divorce. The odds of gray divorce were especially low among homeowner couples at just 0.30 times that of couples that were not homeowners. Although neither marital duration nor remarriage were significantly associated with the likelihood of gray divorce in the full model, these two variables accounted for the positive effect of transitioning to step grandparenthood on gray divorce (results not shown). The transition to biological grandparenthood significantly interacted with marital duration such that the protective effect of the transition to biological grandparenthood diminished as marital duration increased (results not shown, available upon request), which is a pattern consistent with the cumulative probabilities depicted in figure 1.

Both of the robustness checks confirmed our study’s findings. First, we estimated models that compared the risks of gray divorce for three groups: couples that were already grandparents at baseline, couples that became grandparents during the observation period, and couples that did not become grandparents during the observation period. As expected, those who became grandparents did not differ from those who were already grandparents in terms of their risk of divorce. Moreover, those who never became grandparents were at higher risk of divorce than either those who were already grandparents or those who transitioned to grandparenthood during the observation period. This pattern of findings held without and with the inclusion of the control variables as shown in Models 1 and 2 of Table 3. Again, the control variables operated in the expected fashion.

Table 3.

Odds Ratios from Discrete-Time Logistic Regressions of Divorce on Grandparenthood Status (n = 35,908 Couple-Interview Years)

	Model 1	Model 2
Grandparenthood status
Grandparents at baseline (reference)
Grandparents during observation	0.71	0.68
Never grandparents	2.54 * * *	2.58 * * *
Marital biography
Marital duration		0.96
Remarriage		1.08
Marital quality
Wife’s report of marital interaction		0.84*
Husband’s report of marital interaction		0.71***
Spousal homogamy
Husband’s age		0.97
Difference in husband’s and wife’s ages		1.01
Husband’s years of education		1.03
Difference in husband’s and wife’s years of education		0.97
Both spouses are White (reference)
Both spouses are non-White		1.16
Spouses have different races		1.62
Family composition
Number of shared children		1.00
Number of non-shared children		1.12***
Economic resources
Home owner		0.66*
In debt		1.01
0 to $50,000 (reference)
$50,001–$100,000		0.78
$100,001–$250,000		0.72
$250,001 or more		0.60*
Life course transitions
Empty nest		1.07
Wife retired		0.79
Husband retired		1.06
Wife’s chronic conditions		1.25
Husband’s chronic conditions		1.21
Interview year
2000 (reference)
2002	1.79*	1.96*
2004	0.94	1.14
2006	1.03	1.44
2008	1.20	1.91*
2010	0.88	1.60
2012	0.96	1.93
2014	0.74	1.69
Constant	0.01***	0.90
Model fit statistic	F (9,50.1) = 5.49	F (31,49.4) = 37.92

^* P < .05, ^**P < .01 and ^***P < .001 (two-tailed tests).

Bolded coefficients within model are significantly different, P < .05.

Second, we estimated our main models using a Heckman two-step approach (see Table 4). The results mirrored those shown in Table 2, and the Mills ratio was not significant in the bivariate (Model 1) or full (Model 2) models as shown in Table 4, an additional evidence that any selection into our analytic sample versus already being grandparents (and thus excluded from our analytic sample) was unrelated to the risk of divorce. The consistent pattern of results across these two robustness checks reinforces the validity of the original study findings shown in Table 2.

Table 4.

Odds Ratios from Discrete-Time Logistic Regressions Predicting Divorce, with Inverse Mill’s Ratio (n = 7,587 Couple-Interview Years)

	Model 1	Model 2
Transition to grandparenthood
Transition to biological grandparenthood	0.31 * * *	0.44 *
Transition to step grandparenthood	2.64 * *	1.65
No transition (reference)
Marital biography
Marital duration		0.97
Remarriage		0.79
Marital quality
Wife’s report of marital interaction		0.71*
Husband’s report of marital interaction		0.73*
Spousal homogamy
Husband’s age		0.98
Difference in husband’s and wife’s ages		0.97
Husband’s years of education		0.98
Difference in husband’s and wife’s years of education		0.97
Both spouses are White (reference)
Both spouses are non-White		1.38
Spouses have different races		1.67
Family composition
Number of shared children		1.20
Number of non-shared children		1.39**
Economic resources
Home owner		0.30**
In debt		1.22
0–$50,000 (reference)
$50,001–$100,000		1.12
$100,001–$250,000		1.05
$250,001 or more		0.97
Life course transitions
Empty nest		0.87
Wife retired		1.05
Husband retired		1.33
Wife’s chronic conditions		1.19
Husband’s chronic conditions		1.28
Interview year
2000 (reference)
2002	2.39	2.55
2004	0.90	0.99
2006	1.11	1.41
2008	1.00	1.31
2010	0.72	1.06
2012	1.27	2.00
2014	0.30	0.50
Constant	0.01***	3.83
Inverse Mill’s ratio	0.62	0.37
Model fit statistic	F (10,49.1) = 4.26	F (32,47.3) = 7.15

^* P < .05, ^**P < .01 and ^***P < .001 (two-tailed tests).

Bolded coefficients within model are significantly different, P < .05.

Discussion

Gray divorce is an increasingly common event with one in four divorces in the United States involving an individual aged 50 or older (Brown and Lin 2012). Yet, the unique factors shaping the risk of gray divorce remain poorly understood as few studies have investigated the antecedents of divorce in later life (Bair 2007; Karraker and Latham 2015; Lin et al. 2018b). Moreover, theoretical suppositions about the salience of life course transitions, such as an empty nest, retirement, and failing health, for the risk of gray divorce, remain unsupported by quantitative research (Lin et al. 2018b).

Drawing on longstanding literature showing that the transition to parenthood has a “braking effect” on divorce (Cherlin 1977; White and Booth 1985b), we extended this framework to posit that the transition to grandparenthood is likely to have an analogous deterrent effect on the risk of gray divorce. Becoming a grandparent is a generative experience that provides the couple with a shared project (Berger and Kellner 1964) and adds new meaning to life (Cunningham-Burley 1986), reinforcing familial solidarity and strengthening the couple relationship tie. Ultimately, becoming grandparents arguably reduces the chances of gray divorce by inhibiting couples from growing apart, a primary reason couples have pointed to as the triggering factor for their split (Bair 2007).

Using prospective data from 16 years of the HRS, we uncovered dramatic differentials in the risk of gray divorce for couples who transition to biological grandparenthood versus who do not become grandparents or transition to step grandparenthood. After becoming biological grandparents, the risk of gray divorce plummets to only about 1 percent. In contrast, couples who either transition to step grandparenthood or do not become grandparents have expected likelihoods of divorce that hover around 12 percent. Prior to controlling for factors associated with either the transition to grandparenthood or gray divorce, couples who transitioned to step grandparenthood actually were more likely to get divorced than their counterparts who did not transition to grandparenthood, aligning with research showing that stepchildren are positively associated with divorce among remarried couples (White and Booth 1985a). Likewise, the braking effect established for couples of childbearing age who experience the transition to parenthood (White and Booth 1985b) parallels the negative association we uncovered between the transition to biological grandparenthood and gray divorce. Becoming biological grandparents is protective against gray divorce much like becoming parents shields couples of childbearing age from divorce.

Transitioning to step grandparenthood does not offer comparable benefits. The positive bivariate relationship between this transition type and gray divorce reduced to nonsignificance after accounting for confounding factors. Becoming grandparents in a stepfamily is not protective against divorce, perhaps reflecting the weaker (or even conflictual) familial ties characterizing most stepfamilies (Ganong and Coleman 2017). A step grandchild is unlikely to create the same type of shared project as a grandchild who is biologically related to both spouses. Rather, the birth of a step grandchild may effectively exclude the non-biological grandparent from the transformative experience, reshaping the marital reality and creating a wedge between spouses.

By bridging prior research on parenthood and divorce with literature on the meaning and significance of grandparenthood, we developed a novel approach to conceptualizing the role of grandparenthood in divorce during the second half of life. The rapid acceleration in divorce among older adults coupled with the aging of the US population underscores the urgency of deciphering the unique risk factors associated with gray divorce (Brown and Lin 2012). We posited that the transition to grandparenthood could exert a braking effect on gray divorce akin to the effect documented for the transition to parenthood among married couples of childbearing age. Our study shows the transition to biological grandparenthood has a sizeable deterrent effect on gray divorce, signaling the powerful ways that becoming grandparents transforms the marital reality, providing couples with a shared project that rejuvenates midlife marriages (Berger and Kellner 1964). By fortifying the couple relationship (Taubman-Ben-Ari et al. 2018), the transition to biological grandparenthood exerts a braking effect that mitigates against the drifting apart which frequently precipitates gray divorce among midlife couples who have tired of their stale, routinized marriages (Bair 2007).

The large association we uncovered between the transition to grandparenthood and gray divorce also informs our understanding of the gray divorce phenomenon by offering new insights into why the rapid acceleration in gray divorce has occurred. Family population changes such as delayed fertility and increased childlessness have ripple effects across the life course, contributing to shifts in the experience of biological grandparenthood in later life, which is more often either delayed or foregone (Margolis 2016; Margolis and Verdery 2019). If older married couples increasingly either take longer to become biological grandparents or never transition to biological grandparenthood, then we would expect the level of gray divorce in the population to rise given the sizeable protective effects the transition to biological grandparenthood has on gray divorce. Likewise, the growth in remarriage and stepfamilies across the life course presage a shift in the composition of older adult married couples, and about 40 percent are stepfamilies (Lin et al. 2018a). This means that the transition to grandparenthood increasingly occurs through an adult stepchild rather than a biological child, and we found that step grandparenthood offers no protection against gray divorce, revealing a new mechanism through which remarriage is likely contributing to the climbing level of gray divorce in the older adult population. In short, it is possible that some of the recent rise in gray divorce reflects corresponding population changes in family dynamics, including fertility timing and stepfamily formation, which in turn have diminished the chances of experiencing a transition to biological grandparenthood. This is an important topic for future research to explore empirically.

Our study also makes notable contributions to the larger literatures on both divorce and grandparenthood. Prior work has shown that children’s behaviors while the parental marriage is intact are associated with their parents’ risk of divorce (Cherlin et al. 1991; Sun 2001). Here, we demonstrate that the fertility behavior of an adult child (i.e., producing a grandchild) is protective against divorce for parents, providing new evidence that the actions of adult children can shape the likelihood of parental divorce into mid and later life. Additionally, our work underscores “the long reach of divorce” across three generations, a term invoked by Amato and Cheadle (2005) to describe the intergenerational transmission of divorce from the grandparent to the grandchild generation. Here, we show the long reach of the birth of a third (grandchild) generation for the marital stability of the first (grandparent) generation. Finally, prior research has documented that time spent with grandchildren is linked to less frailty among older adults (Chen et al. 2015). Our work adds to this literature by showing that a benefit of becoming a grandparent is its negative relationship with marital instability.

We recognize that in addition to the many strengths of our study, it also has some limitations. Although we capture the transition to grandparenthood prospectively, we miss transitions that occur prior to age 50. Thus, couples who become grandparents at relatively young ages are largely excluded from our study. However, given that we found the negative effect of the transition to biological grandparenthood declined as marital duration increased, we anticipate that our exclusion of young grandparents actually biases downward our estimate of the magnitude of the benefit of transitioning to biological grandparenthood. More generally, our study design excluded couples who were already grandparents, a group that represented a sizeable share of married couples. To gauge the potential biases associated with this exclusion, we performed two robustness checks. Both of these robustness checks yielded results that mirrored our study findings.

Another limitation of our study is that we were not able to capture different pathways to step grandparenthood (Ganong and Coleman 2017) in the HRS. We can only measure the relationship between the grandchild and grandparents according to the relationship between the grandparents and their adult children. For example, we define the transition to biological grandparenthood as having occurred when couples report that their shared biological child now has a child. However, the relationship between the adult child and grandchild is unknown; it is possible that the grandchild is not the adult child’s biological offspring but instead is a stepchild. These cases are likely to be rare because we would expect older adults to be less likely to report a grandchild that is not the biological offspring of their adult child. Family boundary ambiguity research indicates that people often do not report stepkin (White 1998; Stewart 2005). Regardless, if we have misclassified some step grandchildren as biological grandchildren, then our results are conservative estimates of the beneficial effects of transitioning to biological grandparenthood.

There are a few additional limitations to note. The fertility behaviors of adult children should be largely immune to parental marital dynamics, but some adult children might delay their own fertility if they believe their parents’ marriage is unstable, which would suggest the positive association between not becoming grandparents and the risk of gray divorce is overestimated. Since the HRS does not include interviews with the children of respondents, it is not possible to test this potential confounder. Also, we would have liked to have had longitudinal measures of marital quality, which we could only gauge at baseline. Presumably, the transition to grandparenthood is associated with corresponding gains in marital quality. Additionally, we only considered the transition to grandparenthood and did not explore the effects of subsequent grandchildren on the risk of gray divorce. In supplemental analyses (results not shown, available upon request), we uncovered no evidence that the birth of a second biological grandchild offered an additional protective effect beyond the first biological grandchild. Nor was a second step grandchild associated with gray divorce. Although it seems there is no dose-response relationship between grandchildren and divorce, our findings are equivocal due to diminished statistical power and await future study. Finally, the HRS does not include measures of parent–child relationship dynamics in the core survey. It is possible that becoming biological grandparents has a larger protective effect for couples who are closer to their children, but we are unable to test this possibility.

Our study offers notable new insights on gray divorce and the significance of grandparenthood for older adults. Building on an extensive literature that documents the braking effect of transitioning to parenthood on the risk of divorce for couples of childbearing age, we uncover new evidence that a comparable effect operates for the transition to biological grandparenthood. Becoming biological grandparents is associated with a dramatic reduction in the risk of gray divorce. This large braking effect of grandparenthood on divorce should encourage future researchers to investigate whether the benefits of transitioning to grandparenthood extend beyond marital stability to other domains of well-being such as mental and physical health.

About the Authors

Susan L. Brown is distinguished research professor and chair of sociology at Bowling Green State University. A family demographer, her research examines intimate partnership dynamics and their implications for well-being. She is the author of Families in America (University of California Press).

I-Fen Lin is professor of sociology at Bowling Green State University and fellow of the Gerontological Society of America. Her research focuses on family change, dynamics, and functioning in later life, with attention to intergenerational transfer, caregiving, gray divorce, and reporting bias. She has published extensively in family and gerontology journals.

Kagan A. Mellencamp is a doctoral student and research assistant in the Department of Sociology at Bowling Green State University. His research interests center on aging and life course, bereavement, and health and well-being in later life.