Romantic Unions and Mental Health: The Role of Relationship Churning

Abstract

The stress process perspective suggests that romantic relationship transitions can be stressors that impair mental health. Research on romantic relationships and mental health has ignored one common stressor, on-again/off-again relationships, or churning. Using five waves of data from the Fragile Families and Child Wellbeing Study (N = 3,176), we examine associations between relationship churning and mothers’ mental health. We find that mothers experiencing relationship churning have worse mental health than mothers in stably together relationships, net of characteristics associated with selection into relationship instability; these associations persist over four years. Mothers experiencing relationship churning have similar mental health as their counterparts who experience union dissolution (with or without repartnering). Current relationship status and quality explain some of the differences between churning and stably together mothers. Findings emphasize attending to multiple types of family stressors—even stressors and instability in ongoing relationships—and the micro-level ecological factors that shape mental health.

A wealth of research explores the benefits and costs of relationship formation and dissolution to individuals’ mental health (Barrett 2000; Dupre and Meadows 2007; Goldman, Korenman, and Weinstein 1995; Horwitz and White 1991; Hughes and Waite 2009; Lillard and Waite 1995; Lorenz et al. 2006; Marks and Lambert 1998; Umberson 1992). While initially focused on marriage and divorce, the field has diversified to consider other relationship forms, such as cohabitation (Meadows, McLanahan, and Brooks-Gunn 2008; Musick and Bumpass 2012; Rhoades et al. 2011), which is important given the growing diversity of families in the United States (Sassler and Lichter 2020). Nonetheless, this expansion does not capture the common, but understudied, experience of being in an on-again/off-again relationship, known as relationship churning (Halpern-Meekin et al. 2013). Relationship churning is reported by nearly half of young adults in their present or most recent relationship (Halpern-Meekin et al. 2013) and by one-fifth of parents in urban areas by their child’s ninth birthday (Turney and Halpern-Meekin 2021). The stressor of relationship transitions or instability, as opposed to partnership status, is strongly related to mental health (Osborne and McLanahan 2007), and, accordingly, a focus on churning relationships is essential for fully understanding the association between romantic union experiences and mental health.

In the present study, we use data from the Fragile Families and Child Wellbeing Study, a birth cohort study of nearly 5,000 parents from 20 urban areas, to examine the association between relationship churning and mental health. We examine whether mothers’ relationship churning, compared to other relationship experiences (including those with stably together relationships and those with relationships that dissolve, either with or without repartnering), is associated with mothers’ mental health outcomes over time. We also consider the possibility of reverse causality, that is, whether mental health challenges increase the likelihood of relationship churning. These data are ideal for providing the first examination of the association between relationship churning and mental health. First, these data include a large number of parents who identify as members of minoritized groups, a group that is overrepresented among couples experiencing relationship churning (Halpern-Meekin and Turney 2016). Second, these data provide measures of both relationship churning and mental health, which is rare in larger data sets. We operationalize the latter with a diverse set of measures including self-reported depression, heavy drinking, and receiving mental health treatment (seeking mental health services or taking prescribed medication for mental health problems). Third, these data allow us to focus on mothers, an important population given that their mental health is important to their own well-being and to their children’s short- and long-term development (Cavanagh and Fomby 2019; Hardie and Turney 2017; Turney 2011). Understanding the relationship between relationship churning and mental health provides new information on the health repercussions of relationship instability.

BACKGROUND

The Stress Process Perspective as a Lens for Understanding Family Transitions and Mental Health

The stress process perspective provides a lens for understanding the relationship between romantic union experiences and mental health. This perspective suggests that stressors are disproportionately experienced by vulnerable populations and that these stressors have deleterious repercussions for mental health (Pearlin 1989; Pearlin et al. 1981). Exposure to stressors varies among women and men, as do coping responses (Matud 2004); for example, women are more likely than men to drink following romantic relationship troubles (Levitt and Cooper 2010). Furthermore, relationship instability is a stressor that varies by race-ethnicity and socioeconomic status in its frequency and associated outcomes (Kalmijn 2010; Karney 2021; Kuo and Raley 2016; Wu and Thomson 2001). Therefore, there are reasons to expect that it might be uniquely important to understand the experiences of women from minoritized groups with the particular stressors we examine here—relationship instability and its correlates.

Our focus on relationship instability, rather than family structure or marital status, is in line with previous research finding relationship instability is predictive of mental health (Blekesaune 2008; Meadows et al. 2008; Whitton and Whisman 2010). For example, among unmarried parents, those who dissolved their union before their child’s birth reported higher levels of psychological distress and substance abuse than those who stayed together in cohabiting or nonresidential relationships (DeKlyen et al. 2006). In line with the stress process perspective, we are interested in whether individuals’ experiences of change in family roles and relationships is related to their mental health; we are less interested in whether there is a difference between, for example, married and cohabiting individuals, as has been the focus of much previous research.

The stress process perspective offers two more specific theories about how long the consequences of stressors will last. The crisis perspective suggests that the repercussions of stressors, such as relationship transitions, are short-lived. Relationship transitions are a jolt to well-being, followed by a period of adjustment (Meadows et al. 2008; Simon 2002; Simon and Marcussen 1999; Strohschein et al. 2005; Williams 2003; Wu and Hart 2002); that is, there is a crisis and then recovery, such that the negative consequences ease over time. Alternatively, the chronic strain perspective implies both short- and long-term repercussions of stressors such as relationship transitions. As a relationship ends, the costs of union dissolution and the negative consequences for well-being continue over time, inducing chronic strain. That is, differences emerge following a transition and then persist.

Although relationship instability may lead to consequences for mental health via a crisis or chronic strain, an alternative possibility is that the characteristics that affect the likelihood of relationship instability also affect the likelihood of experiencing mental health challenges, thereby inducing a spurious correlation between relationship instability and mental health. Indeed, the stress process perspective highlights the embeddedness and interconnectedness of stressors (Pearlin 1989; Wheaton 1994). Although selection into relationship instability certainly plays some role in the association between relationship instability and mental health, research indicates that selection processes cannot entirely explain away these associations (Blekesaune 2008; Hill, Reid, and Reczek 2013 Johnson and Wu 2002; Kim and McKenry 2002; Lamb, Lee, and DeMaris 2003). Therefore, in our analyses, we adjust for characteristics associated with selection into relationship instability.

Relatedly, reverse causality may play a role in the association between relationship transitions and mental health, namely, that the stressor of mental health challenges influences relationship transitions. For example, one study finds that selection into marriage among healthier people explains some of the observed positive correlation between depression and union status (Tumin and Zheng 2018). Other research finds no indication of positive selection into relationships but some indications of negative selection out of relationships, with less mentally healthy mothers more likely to experience a breakup (Meadows et al. 2008). Although mental health may be predictive of relationship outcomes, researchers generally conclude that the bidirectional relationship is stronger the other way, running from romantic union experiences to partners’ mental health (Braithwaite and Holt-Lunstad 2017; Whisman, Sbarra, and Beach 2021). We conduct supplemental analyses that capitalize on our longitudinal data to examine whether mental health indicators are predictive of relationship status.

Relationship Churning in the Stress Process Perspective

In alignment with the stress process perspective, there are theoretical reasons to expect that relationship churning, compared to other types of relationship transitions, may be especially detrimental for mental health. Indeed, relationship churning is a stressor. Prior research finds that it is associated with psychological distress net of prior mental health in a sample distinctive from the present study’s—a majority White, highly educated group of 545 parents and nonparents (Monk, Ogolsky, and Maniotes 2022; Monk, Ogolsky, and Oswald 2018). Relationship churning introduces a liminality or ambiguity into the relationship, with the status of the relationship and the accompanying relationship norms and roles potentially unclear to partners. Ambiguity can be uniquely unsettling for individuals and families (Boss 1980; Boss et al. 1990). Unlike being stably together or stably broken up, partners in a churning relationship may be neither trustingly settled into the union nor able to move on. In line with this perspective, Turney and Carlson (2011) find that multiple-partner fertility—which itself can produce ambiguity in relationships because the performance of family roles is spread across households—is associated with an increased risk of maternal depression.

A pattern of relationship churning that involves couples coming back together repeatedly may be stressful because the ability to trust or rely on the resources of the relationship may be undermined by its ongoing instability. In a sample of college students, three-quarters of those who had been in a churning relationship described breaking up and getting back together two or more times, indicating that churning may be an ongoing process, as opposed to a one-time relationship disruption (Dailey et al. 2009). In essence, for churners, the distinction the stress process perspective makes between crisis and chronic strain may be artificial because a state of repeated crisis may create chronic strain for those in churning relationships. As research in the neurodevelopmental field has found, a lack of predictability in one’s environment can induce negative physiological stress responses (Soltani and Izquierdo 2019), underlining the importance of stability to well-being.

Existing Research on Family Transitions and Mental Health

Previous research finds that relationship troubles and transitions are associated with subsequent alcohol use and depression (Levitt and Cooper 2010; Verhallen et al. 2019); therefore, we focus on these outcomes in our study. Although existing research has not examined the associations between relationship churning and mental health, a large body of research examines how the association between relationship transitions and mental health varies by union type. Although some find that marriage is more strongly predictive of positive mental health than is cohabitation (e.g., Kim and McKenry 2002), others find few differences between relationship structures in predicting psychological well-being (Musick and Bumpass 2012; Wu and Hart 2002). Among unmarried couples, those who are cohabiting or have plans to marry experience more psychological distress following a breakup than do those who were less committed (Rhoades et al. 2011; see also Brown 2000). More recent research suggests that the benefits to marriage may have been previously overstated, with limited mental health gains from marriage but relatively larger negative consequences to relationship dissolution (Kalmijn 2017; see also Simon and Barrett 2010; but see Chen and van Ours 2018). In this previous research, it is relationship disruption, rather than relationship formation, that is a stronger predictor of mental health.

The association between relationship instability and mental health may not apply to all demographic groups. Hill et al. (2013) question whether lower-income women in urban areas see benefits to marriage. They find that neither marriage entry nor exit is related to mental health; stably married women, however, have lower rates of psychological distress, which may be due to them being less likely to experience financial struggles. By contrast, Meadows et al. (2008) examine relationship status transitions and mental health among urban mothers in the Fragile Families and Child Wellbeing Study, who are disproportionately low-income and from minoritized groups. Their results show that compared to being stably together, mothers who experience a breakup report poorer mental health. In line with the stress process crisis perspective, these differences in well-being decline over time (see also Hetherington 1999; Leopold and Kalmijn 2016; Lorenz et al. 2006; Soons, Liefbroer, and Kalmijn 2009; Strohschein et al. 2005). Other research using the same data finds that breaking up increased the risk of depression, while repartnering decreased the risk of depression (Osborne, Berger, and Magnuson 2012). These previous studies have not recognized the existence of churning relationships. The present study adds to knowledge in the field by separately considering couples who are stably together versus churning, thereby allowing us to understand more about the role of instability in relationships (as opposed to between relationships) in predicting mental health outcomes.

The Present Study

Research indicates that churning may be a distinctive experience for mothers versus fathers and thus should be theorized and analyzed separately. We focus on mothers because previous research finds that children are more likely to remain with their mothers after their parents separate, that father involvement declines over time when parents are separated, and that non-coresidential fathers have distinctive patterns of child involvement (Carlson, VanOrman, and Turner 2017; Edin, Tach, and Mincy 2009). Among mothers, the negative consequences of relationship transitions may be amplified for those with younger children (Leopold and Kalmijn 2016), making these parents of particular interest.

Although the mothers in the present study vary in the number and ages of their children, all gave birth to the study’s focal child between 1998 and 2000 and are surveyed longitudinally when their focal children are similar ages, allowing us to observe their experiences throughout the focal child’s early years. As the existing research elucidates, examining the association between relationship instability and mental health for a population of lower-income, urban mothers, predominantly from minoritized groups, as the present study does, is particularly important. The present study asks whether these trends hold when we separate out temporary breakups (churning) from stable breakups and new partnerships (repartnering), and we compare the associations of these forms of instability to couples remaining stably together or stably separated without repartnering or churning.

In the present study, we draw on the stress process perspective to examine the role of relationship churning in mental health for a diverse sample of mothers. Hypothesis 1 predicts that those in churning relationships, compared to those who are stably together, will report more negative mental health outcomes over the shorter term and longer term. Hypothesis 2 predicts that those in churning or repartnered relationships, compared to those who are stably together or stably separated without repartnering, will report the most negative mental health outcomes at both time points. We treat as an open empirical question whether those in churning versus repartnered relationships report more positive mental health outcomes. We can observe whether a breakup followed by a relationship with a new versus the same partner matters for mental health. Reunifying with the same partner potentially creates less of a transition (because partners may know and understand each other already), but the resources the churning relationship offers could be fewer due to its ongoing instability. That is, we could observe similar patterns of outcomes in both groups, but this could be driven by distinct relationship dynamics because the stressors each face are different.

The prediction in Hypothesis 2 that those who experience relationship churning compared to a stable breakup without repartnering will report poorer mental health outcomes is based on the idea that churning partners, as discussed previously, may remain in a liminal or ambiguous state. The ways that a relationship can be beneficial to mental health—through the protective effects of trust, care, shared support, and social control (Kamp Dush and Amato 2005; Umberson 1992)—may be weakened by the relationship’s ambiguous state. Repeated family transitions can be particularly challenging (Osborne and McLanahan 2007), which could suggest that churning is uniquely consequential for mental health because it maintains liminality and prevents adjustment.

Furthermore, to the extent that churning may draw partners back into a lower-quality relationship, it will increase risks to their well-being that are associated with lower-quality unions (Holt-Lunstad, Birmingham, and Jones 2008; Simon and Barrett 2010; Umberson et al. 2006), particularly because instability in both relationship status and quality is linked to a higher risk of health problems (Blekesaune 2008; Whitton and Whisman 2010). Therefore, we additionally test two competing hypotheses. Hypothesis 3a predicts that churning is associated with mental health outcomes over and above number of relationships and relationship quality because of the liminality that churning induces. Conversely, Hypothesis 3b predicts that number of relationships and relationship quality will account for differences in mental health by relationship status (churning, stably together, stably separated with repartnering, or stably separated without repartnering).

DATA AND METHODS

Data

We considered the role of relationship churning in the association between romantic unions and mothers’ mental health using data uniquely positioned to examine this topic. The Fragile Families and Child Wellbeing Study, a cohort of mostly unmarried urban parents with a child born around the turn of the twenty-first century who were followed over time, was designed to study the correlates and consequences of parents’ relationships (Reichman et al. 2001). Baseline interviews with parents occurred immediately after their child’s birth, between 1998 and 2000, and parents were interviewed an additional five times over a 15-year period (corresponding to when their children were about 1, 3, 5, 9, and 15 years old). We used data through the 9-year survey given the relatively closer spacing between the first five survey waves (compared to the 6-year gap between the 9- and 15-year surveys), which was necessary to observe quickly unfolding relationship processes and their proximal outcomes.

The analytic sample included 3,176 of the 4,898 mothers in the baseline sample. We first excluded the 1,606 mothers who did not participate in the five- and nine-year surveys (with 223 not participating in the five-year survey, 847 not participating in the nine-year survey, and 536 not participating in both the five- and nine-year surveys), which was when our outcome variables are measured, and the additional 40 observations missing data on any mental health indicators at the five- or nine-year surveys (described in the following). We also excluded an additional 76 mothers who did not fit into one of the four relationship history categories (also described in the following).

There were several small observed demographic differences between the baseline and analytic samples. Mothers in the analytic sample, compared to mothers in the baseline sample, were more likely to identify as non-Hispanic Black (50.5% compared to 47.6%, prior to imputation), less likely to identify as Hispanic (24.4% compared to 27.3%), less likely to be born outside the United States (13.3% compared to 17.0%), and less likely to have less than a high school diploma (31.8% compared to 34.7%). Mothers in the analytic sample were also more likely than those in the baseline sample to have more than one child with the focal child’s father (59.3% compared to 55.2%).

Measures

Mental health problems.

We measured mental health problems, at both the five- and nine-year surveys, with three binary variables: (1) depression, (2) heavy drinking, and (3) receiving mental health treatment. First, depression was measured by mother’s responses to the Composite International Diagnostic Instrument-Short Form (CIDI-SF), a commonly used measure of depression in large samples (Kessler et al. 1998). Mothers were considered depressed in the past year if they answered affirmatively to at least one of two stem questions (feeling sad, blue, or depressed or losing interest in normally pleasurable things) most of the day every day for a period of at least two weeks over the past year and three additional questions (about losing interest in things, feeling tired, experiencing a weight change of at least 10 pounds, having trouble sleeping, having trouble concentrating, feeling worthless, and thinking about death). Second, mothers were asked about alcohol consumption, and heavy drinking indicated the mother reports having four or more drinks in one sitting at least once a month over the past year. Third, a binary variable indicated that the mother reported at least one of the following two types of mental health treatment: (1) receiving counseling, therapy, or other treatment for personal problems (including feelings of depression, worry, alcohol, or drug use problems) in the past year and (2) regularly taking prescribed medication for depression, anxiety, or other mental health conditions. These three measures allowed us to capture an array of expressions of mental health problems, following previous scholars (Aneshensel, Rutter, and Lachenbruch 1991; Simon and Barrett 2010; Uecker 2012). It also allowed us to capture individuals whose treatment (via receiving mental health services or taking prescribed medication) for mental health problems may mean they did not report symptoms. Because each survey asked mothers to recall experiences over the past year and because the surveys are administered four years apart, measures at the five- and nine-year surveys asked about distinct periods and sets of experiences.

Relationship churning.

We measured relationship churning between the baseline and five-year surveys with both direct and indirect reports of churning. We used mothers’ reports of churning (rather than fathers’ reports or a combination of the two partners’) because their reports of churning are theoretically most likely to be associated with their mental health and because, as described earlier, they are more likely to participate in all survey waves (and therefore less likely to have missing data). This approach is consistent with prior research (e.g., Halpern-Meekin and Turney 2016).

Direct churning was measured by mothers’ reports that they were in an on-again/off-again relationship with the focal child’s father at the baseline, three-year, or five-year surveys (with direct reports of churning unavailable at the one-year survey). We supplemented these direct reports of churning with indirect reports of churning, which considered relationship transitions between survey waves. Indirect churning included one of the following two types of transitions: (1) mothers who reported being in a (marital, cohabiting, or nonresidential romantic) relationship with the father at one survey wave, not in a relationship with him at the next survey wave, and in a relationship with him at the following survey wave and (2) mothers who reported not being in a relationship with the father at one survey wave, in a relationship with him at the next survey wave, and not in a relationship at the following survey wave. About 12.3% and 5.5% of mothers in the analytic sample experienced direct and indirect churning, respectively, between the baseline and five-year surveys (with 1.4 % experiencing both direct and indirect churning).

We compared mothers who experience relationship churning between the baseline and five-year surveys to three other groups of mothers, with all groups based on relationship status between the baseline and five-year surveys (Halpern-Meekin and Turney 2016). The first group included mothers in stably together relationships—those who report any (marital, cohabiting, or nonresidential romantic) relationship with the child’s father at all time points (and no relationship churning with the child’s father). The second group included mothers who report dissolving their relationship with the child’s father and do not report repartnering (and no relationship churning with the child’s father). The third group included mothers who report dissolving their relationship with the child’s father and do report repartnering (and no relationship churning with the child’s father).

Control variables.

We adjusted for characteristics that might have rendered the association between relationship churning and mental health spurious, most measured at the baseline or one-year surveys, because the stress process perspective highlights the contextual embeddedness of stressors (Pearlin 1989). Demographic characteristics included parents’ race-ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic other race), age, childhood family structure (1 = lived with both biological parents at age 15), and education (less than high school, high school diploma or GED, some college, college). Family characteristics included parents’ baseline relationship status, relationship duration at baseline, having a shared child (besides the focal child), and multipartnered fertility. Socioeconomic characteristics included parents’ material hardship and employment, and behavioral characteristics included parents’ impulsivity (Dickman 1990) and cognitive ability (Wechsler 2001).

Additionally, we adjusted for measures that capture mothers’ predisposition to mental health problems. We adjusted for three indicators of mothers’ mental health prior to baseline, taken from the mothers’ medical records from her hospitalization at the focal child’s birth. These included binary variables indicating drug use during pregnancy, alcohol use during pregnancy, and pre-pregnancy mental illness diagnosis (all ascertained from all possible sources in the medical record). We also adjusted for mothers’ parents’ mental health, measured with a binary variable indicating that the mother reported, at the three- or five-year surveys, her biological mother or father ever had periods lasting two weeks or more when they were depressed, down in the dumps, or blue most of the time.

Current relationship status.

In some analyses, we adjusted for relationship characteristics measured contemporaneously with the outcome variables (i.e., at the five- or nine-year surveys). First, a series of mutually exclusive variables indicated mother’s relationship with the father: no relationship (reference category), lower-quality relationship (mother reports the overall relationship quality as being poor, fair, or good), and higher-quality relationship (mother reports the overall relationship quality as being very good or excellent). Second, a continuous variable indicated the number of romantic relationships since the last interview that lasted at least one month (with mothers in a continuous relationship with the father receiving a value of 1) because some in churning relationships with the child’s father or in repartnered relationships also reported relationships with other partners.

Analytic Strategy

The analytic strategy occurred in three primary stages. All analyses paid careful attention to the time ordering of the variables, with the outcomes measured after the key independent variables and the control variables measured at baseline (or when unavailable at baseline, the first survey wave they are available). We used multiple imputation to impute observations missing values, pooling results across 20 imputed data sets (Allison 2001).

First, we estimated the short-term association between relationship churning—measured through the five-year survey—and the three indicators of mothers’ mental health problems (Hypotheses 1 and 2). We first examined the frequency of mental health problems (depression, heavy drinking, and receiving mental health treatment), measured at the five-year survey, across the four categories of relationship history, comparing mothers in churning relationships to mothers in the other three groups (those who experienced stably together relationships, those who experienced dissolution without repartnership, and those who experienced dissolution with repartnership). We used chi-square tests to identify statistically significant differences across groups. We next used linear probability models to estimate each indicator of mental health problems as a function of relationship history (similarly comparing mothers in churning relationships to other mothers). The first model adjusted for a limited set of control variables that include demographic characteristics and other characteristics that precede the measure of churning (e.g., pre-pregnancy mental health). The second model further adjusted for an extended set of control variables, including those variables that may have occurred after relationship churning begins (e.g., material hardship, which is first available at the one-year survey); this model may have overcontrolled for factors that could render the association between relationship churning and mental health null. Therefore, we took a conservative approach in this model and risk underestimating this association. We also estimated the possibility of reverse causality, with earlier mental health challenges predicting later relationship transitions.

Second, we estimated the long-term association between relationship churning and the three indicators of mothers’ mental health problems, with churning measured between the baseline and five-year surveys and mental health problems measured at the nine-year survey (also Hypotheses 1 and 2). We again first examined descriptive differences across groups, using chi-square tests to identify statistically significant differences across groups, and then used linear probability models to estimate mental health problems as a function of relationship history (with the limited and extended set of control variables).

Third, we examined how much of the association between relationship churning and mothers’ mental health problems stems from contemporaneously measured relationship characteristics (Hypotheses 3a and 3b). Here we extended the linear probability models to further adjust for two indicators of current relationship status—relationship quality between the parents (no relationship, lower-quality relationship, and higher-quality relationship) and mothers’ number of relationships since the last interview—to examine how these contemporaneous characteristics changed the association between relationship churning and mothers’ mental health problems.

Sample Description

Characteristics of the analytic sample are presented in Table 1. Relationship churning was reported by 16.2% of mothers. As for the other types of relationship histories, stably together relationships were most common, reported by more than two-fifths (41.2%) of mothers, followed by dissolution with repartnering (27.9%) and dissolution without repartnering (14.7%).

Table 1.

Descriptive Statistics of Sample.

	Mean, or % (SD)
Relationship history (b, y1, y3, y5)
Churning	16.2%
Stably together	41.2%
Dissolution without repartnering	14.7%
Dissolution with repartnering	27.9%
Mother race-ethnicity (b)
White, non-Hispanic	22.0%
Black, non-Hispanic	50.5%
Hispanic	24.1%
Other race, non-Hispanic	3.3%
Mother and father mixed-race couple (b)	14.4%
Mother age (b)	25.231 (6.058)
Father age (b)	27.810 (7.271)
Mother lived with both parents at age 15 (b)	42.0%
Father lived with both parents at age 15 (b)	43.4%
Mother and father relationship status, (b)
Married	24.9%
Cohabiting	35.7%
Nonresidential romantic	27.5%
Separated	11.9%
Mother and father relationship duration (b)	4.801 (4.575)
Mother education (b)
Less than high school	31.6%
High school diploma or GED	31.7%
Some college	25.3%
College	11.4%
Father education (b)
Less than high school	31.2%
High school diploma or GED	37.3%
Some college	21.2%
College	10.3%
Mother used drugs during pregnancy (b)	8.7%
Mother used alcohol during pregnancy (b)	6.1%
Mother had pre-pregnancy mental illness diagnosis, (b)	11.8%
Mother had parent with depression (y3, y5)	34.4%
Mother and father share additional child (y1)	59.1%
Mother multipartnered fertility (y1)	35.7%
Father multipartnered fertility (y1)	32.8%
Mother material hardship (y1)	1.154 (1.594)
Father material hardship (y1)	.416 (1.079)
Mother employment (y1)	55.1%
Father employment (y1)	76.9%
Mother impulsivity (y3)	2.032 (.610)
Father impulsivity (y1)	2.011 (.671)
Mother cognitive ability (y3)	6.816 (2.670)
Father cognitive ability (y3)	6.541 (2.714)
Mother and father relationship status (y5)
No relationship	53.4%
Low-quality relationship	13.0%
High-quality relationship	33.6%
Mother number of relationships since last interview (y5)	.957 (.556)
Mother and father relationship status (y9)
No relationship	57.7%
Low-quality relationship	14.4%
High-quality relationship	27.9%
Mother number of relationships since last interview (y9)	1.052 (.581)
N	3,176

Note: Data source: Fragile Families, and Child Wellbeing, Study. b = measured at baseline survey; y1 = measured at one-year survey; y3 = measured at three-year survey; y5 = measured at five-year survey; y9 = measured at nine-year survey.

Mothers were predominantly members of minoritized groups, with about half (50.5%) of mothers identifying as non-Hispanic Black and one-quarter (24.1%) identifying as Hispanic. More than two-fifths of parents (42.0% of mothers and 43.4% of fathers) reported living with both biological parents at age 15. At baseline, most parents were in a romantic relationship with one another, with 24.9% married, 35.7% cohabiting, and 27.5% in anonresidential romantic relationship. Slightly more than one-third (34.4%) of mothers reported having a parent who experienced depression.

RESULTS

Frequency of Mothers’ Mental Health Problems by Relationship History

Table 2 presents frequencies of mental health problems first for the full sample and then across the four categories of relationship history. At the five-year survey, depression was the most commonly reported indicator of mental health problems, reported by nearly one-fifth (17.4%) of mothers, followed by receiving mental health treatment (10.7%) and heavy drinking (6.7%).

Table 2.

Frequency of Outcome Variables for Full Sample and by Relationship History.

	Full Sample		Relationship History
		Churning	Stably Together	Dissolution without Repartnering	Dissolution with Repartnering
	N = 3,176	n = 502–525	n = 1,294–1,322	n = 461–482	n = 876–896
Short-term outcomes (5-year)
Depression	17.4 %	24.5%	12.3%***	17.8%**	20.8%^
Heavy drinking	6.7%	8.3%	4.8%**	6.4%	8.6%
Receiving mental health treatment	10.7%	13.7%	8.5%**	11.6%	11.8%
Long-term outcomes (9-year)
Depression	17.2%	21.9%	13.6%***	15.9%**	20.5%
Heavy drinking	8.7%	11.3%	6.5%***	9.6%	9.9%
Receiving mental health treatment	13.8%	16.6%	10.9%**	11.9%*	17.4%

Note: Data source: Fragile Families and Child Wellbeing Study. Subgroup ns vary across imputed data sets. Asterisks compare, stably together, dissolution without repartnering, and dissolution with repartnering to churning.

^{^}p < .10

^*p < .05

^**p < .01

^***p < .001.

The frequency of these mental health problems varied across categories of relationship history, especially between mothers experiencing churning and mothers experiencing stably together relationships. Mothers reporting churning, compared to stably together mothers, were about 2 times as likely to report depression (24.5% compared to 12.3%, p < .001). They were also nearly twice as likely to report heavy drinking (8.3% compared to 4.8%, p < .01) and receiving mental health treatment (13.7% compared to 8.5%, p < .01). Mothers in churning relationships are also more likely to report depression than those who experience dissolution without repartnering (17.8%, p < .01) and those who experience dissolution with repartnering (20.8%, p < .10). There are no differences in heavy drinking or receiving mental health treatment between mothers in churning relationships and mothers who experience dissolution with or without repartnering.

The patterns are similar when examining outcomes at the nine-year survey. There are again statistically significant differences in mental health between mothers in churning relationships and mothers in stably together relationships. These differences exist for depression (21.9% compared to 13.6%, p < .001), heavy drinking (11.3% compared to 6.5%, p < .001), and receiving mental health treatment (16.6% compared to 10.9%, p < .001). At the nine-year survey, mothers in churning relationships also have worse mental health than those who experience dissolution without repartnering; this exists for two of the measures (21.9% compared to 15.9% [p < .01] for depression and 16.6% compared to 10.9% [p < .01] for receiving mental health treatment).

Short-Term Association between Relationship Churning and Mothers’ Mental Health

The frequencies previously described show that mothers in churning relationships experience more mental health problems than their counterparts in stably together relationships. These group differences may stem from characteristics associated with both relationship churning and mental health, and we consider this possibility in the next set of analyses. Table 3 presents results that estimate the association between relationship churning and mothers’ mental health, focusing on outcomes measured at the five-year survey.

Table 3.

Linear Probability Models Estimating the Short-Term Association between Relationship Churning and Mental Health (N = 3,176).

	Model 1	Model 2
	+ Limited Controls	+ Extented Controls
Panel A. Depression
Relationship history (reference = churning)  Stably together	−.130*** (.023)	−.107*** (.023)
Dissolution without repartnering	−.066* (.026)	−.048^ (.026)
Dissolution with repartnering	−.043^ (.024)	−.035 (.024)
Panel B. Heavy drinking
Relationship history (reference = churning)  Stably together	−.039* (.015)	−.034* (.016)
Dissolution without repartnering	−.017 (.017)	−.015 (.017)
Dissolution with repartnering	−.008 (.015)	−.007 (.015)
Panel C. Receiving mental, health treatment
Relationship history (reference = churning)  Stably together	−.090*** (.019)	−.075*** (.019)
Dissolution without repartnering	−.023 (.020)	−.011 (.020)
Dissolution with repartnering	−.023 (.019)	−.015 (.019)

Note: Data source: Fragile Families and Child Wellbeing Study. Mental, health measured at the five-year survey.

^{^}p < .10

^*p < .05

^***p < .00l.

We turn first to estimates of depression. The results from Model 1, which adjusts for a limited set of control variables, are consistent with the descriptive statistics presented earlier. Mothers in stably together relationships, compared to mothers in churning relationships, report less depression (b = −.130, p < .001). Mothers who experience dissolution without repartnering report less depression than mothers in churning relationships (b = −.066, p < .05). Mothers who experience dissolution with repartnering also report less depression than mothers in churning relationships, although this association is only marginally statistically significant (b = −.043, p < .10). These associations are reduced in magnitude and, in some cases, statistical significance in Model 2, which adjusts for an extended set of control variables. This model shows that compared to mothers in churning relationships, mothers in stably together relationships (b = −.107, p < .001) and mothers who experience dissolution without repartnering (b = −.048, p < .10) report less depression.

We turn next to the additional outcomes, heavy drinking and receiving mental health treatment. Model 1 shows that mothers in stably together relationships, compared to mothers in churning relationships, report less heavy drinking (b = −.039, p < .05) and less mental health treatment (b = −.090, p < .001). There are no statistically significant differences between churning mothers and mothers who experience dissolution with or without repartnering. Model 2 shows that the statistically differences between mothers in stably together relationships and mothers in churning relationships persist for both heavy drinking (b = .034, p < .05) and receiving mental health treatment (b = −.075, p < .001).

Considering reverse causality.

As discussed previously, although our primary analyses focus on the direction of the association from relationship churning to mental health (and we have paid careful attention to the time ordering of the variables), mental health problems may lead to relationship churning. We considered this possibility in supplemental analyses, for which we used linear probability models to estimate the association between pre-pregnancy mental health problems and the direct measure of churning at the three- and five-year surveys. There are no substantively or statistically significant associations between pre-pregnancy mental health and churning. For example, pre-pregnancy mental illness (b = .004, n.s.) or alcohol use during pregnancy (b = .014, n.s.) is not associated with churning at the five-year survey, providing further support for the stress process perspective.

Long-Term Association between Relationship Churning and Mothers’ Mental Health

Table 4 presents results that estimate the association between relationship churning and mothers’ mental health, focusing on outcomes measured at the nine-year survey. These regression results remain consistent with the descriptive statistics presented earlier. In Model 1, compared to mothers in churning relationships, mothers in stably together relationships report less depression (b = −.050, p < .05), heavy drinking (b = −.040, p < .05), and receipt of mental health treatment (b = −.076, p < .01). Additionally, compared to mothers in churning relationships, mothers who experience relationship dissolution without repartnering report marginally less depression (b = −.045, p < .10) and receipt of mental health treatment (b = −.042, p < .10). In Model 2, which adjusts for additional control variables, these associations are reduced in magnitude and statistical significance, with only one statistically significant difference remaining. Receipt of mental health treatment is less common among mothers in stably together relationships than among mothers in churning relationships (b = −.063, p < .01).

Table 4.

Linear Probability Models Estimating the Long-Term Association between Relationship Churning and Mental Health (N = 3,176).

	Model 1	Model 2
	+ Limited Controls	+ Extented Controls
Panel A. Depression
Relationship history (reference = churning)  Stably together	−.050* (.023)	−.033 (.024)
Dissolution without repartnering	−.045^ (.026)	−.031 (.026)
Dissolution with repartnering	−.009 (.023)	−.001 (.023)
Panel B. Heavy drinking
Relationship history (reference = churning)  Stably together	−.040* (.017)	−.026 (.018)
Dissolution without repartnering	−.011 (.020)	−.004 (.020)
Dissolution with repartnering	−.014 (.017)	−.010 (.017)
Panel C. Receiving mental health treatment
Relationship history (reference = churning)  Stably together	−.076** (.022)	−.063** (.022)
Dissolution without repartnering	−.042^ (.024)	−.032 (.024)
Dissolution with repartnering	.012 (.021)	.017 (.021)

Note: Data source: Fragile Families and Child Wellbeing Study. Mental health measured at the nine-year survey.

^{^}p < .10

^*p < .05

^**p < .01.

The Role of Current Relationships

The prior models primarily provide evidence that mothers who experience relationship churning, compared to mothers who experience stably together relationships, have worse mental health in both the short term (at the five-year survey) and the long term (at the nine-year survey). We next consider whether contemporaneously measured relationship characteristics explain the association between relationship churning and mothers’ mental health problems. We present all outcomes at both the five- and nine-year surveys in Table 5 but focus our discussion on the statistically significant associations.

Table 5.

Linear Probability Models Estimating the Association between Relationship Churning and Mental Health Considering Relationship Status (N = 3,176).

	Short Term		Long Term
	Model 1	Model 2	Model 1	Model 2
	+ Extended Controls	+ Contemporaneous Relationship Status	+ Extended Controls	+ Contemporaneous Relationship Status
Panel A. Depression
Relationship history (reference = churning)
Stably together	−.107*** (.023)	−.103*** (.029)	−.033 (.024)	−.008 (.025)
Dissolution without repartnering	−.048^ (.026)	−.032 (.028)	−.031 (.026)	−.034 (.026)
Dissolution with repartnering	−.035 (.024)	−.028 (.026)	−.001 (.023)	−.007 (.023)
Mother and father relationship status (reference = no relationship)
Low-quality relationship		−.076* (.032)		−.026 (.018)
High-quality relationship		−.024 (.032)		−.004 (.020)
Mother number of relationships since last interview		.027* (.012)		−.010 (.017)
Panel B. Heavy drinking
Relationship history (reference = churning)  Stably together	−.034* (.016)	−.027 (.020)	−.026 (.018)	−.018 (.019)
Dissolution without repartnering	−.015 (.017)	−.009 (.018)	−.004 (.020)	.002 (.020)
Dissolution with repartnering	−.007 (.015)	−.010 (.017)	−.010 (.017)	−.010 (.017)
Mother and father relationship status (reference = no relationship)
Low-quality relationship		.017 (.022)		.009 (.016)
High-quality relationship		−.020 (.022)		−.011 (.016)
Mother number of relationships since last interview		.026** (.008)		.035*** (.009)
Panel C. Receiving mental health treatment
Relationship history (reference = churning)  Stably together	−.075*** (.019)	−.047* (.024)	−.063** (.022)	−.022 (.024)
Dissolution without repartnering	−.011 (.020)	−.017 (.022)	−.032 (.024)	−.038 (.024)
Dissolution. with. repartnering	−.015 (.019)	−.023 (.021)	.017 (.021)	.003 (.021)
Mother and father relationship status (reference = no relationship)
Low-quality relationship		.009 (.026)		−.048* (.020)
High-quality relationship		−.057* (.026)		−.099*** (.020)
Mother number of relationships since last interview		.007 (.010)		.024* (.01 1)

Note: Data source: Fragile Families and Child Wellbeing Study.

^{^}p < .10

^*p < .05

^**p < .01

^***p < .001.

The first two columns of Table 5 present results estimating mental health problems at the five-year survey. Model 1 adjusts for all control variables (and is identical to Model 2 of Table 3). In Model 2, which adjusts for current relationship characteristics, the stably together coefficients estimating depression, heavy drinking, and receiving mental health treatment decline in magnitude (by 4%, 21%, and 31%, respectively). The differences in depression (b = −.103, p < .001) and receiving mental health treatment (b = −.047, p < .05) between those in churning relationships and those in stably together relationships persist. Model 2 also shows that these measures of current relationships are associated with mothers’ mental health. Being in a low-quality relationship with the father, compared to having no relationship with the father, is negatively associated with depression (b = −.076, p < .05). Being in a high-quality relationship, compared to having no relationship with the father, is negatively associated with receiving mental health treatment (b = −.057, p < .05). Additionally, the number of relationships is positively associated with depression (b = .027, p < .05) and heavy drinking (b = .026, p <.01).

The last two columns of Table 5 present results estimating mental health problems at the nine-year survey. Model 1 adjusts for all control variables (and is identical to Model 2 of Table 4). Here we focus on the statistically significant relationship between relationship churning and receiving mental health services. Adjusting for current relationships in Model 2 shows that the stably together coefficient declines in magnitude (by 67%) and statistical significance. After accounting for current relationships, there are no statistically significant differences in mental health treatment between mothers in churning relationships and mothers in stably together relationships (b = −.022, n.s.). Being in a low-quality or high-quality relationship with the father, compared to not being in a relationship with the father, is negatively associated with receiving mental health treatment (b = −.114, p < .001). Number of relationships is positively associated with receiving mental health treatment (b = .024, p < .05).

In sum, we find support for Hypothesis 1 of variation in mental health outcomes between those in churning versus stably together relationships, although this holds more strongly at the short-term than the long-term follow-up. We find limited support for Hypothesis 2, with the stably separated generally similar to those in churning and repartnered relationships once we include our extended set of controls; as predicted in Hypothesis 2, those in churning and repartnered relationships have similar mental health outcomes. We see support for Hypothesis 3a when we examine short-term outcomes, with the stably together less likely to report mental health issues than mothers in churning relationships, even accounting for number of relationships and relationship quality. However, Hypothesis 3b appears to be supported in the long term, with relationship history no longer associated with mental health indicators once we control for relationship status and quality.

DISCUSSION

The present study is aligned with the stress process perspective (Pearlin 1989; Pearlin et al. 1981), providing the first examination of how the stressor of relationship churning, a stressor unequally distributed across the population (Halpern-Meekin and Turney 2016), is associated with mental health problems even after taking into account the embeddedness of stressors (Wheaton 1994). That is, observed selection factors alone do not account for differences in mental health across those with varying romantic relationship histories (although unobserved characteristics may still render these relationships spurious). We also provide the first accounting of the fact that instability in a relationship with one partner (i.e., relationship churning) is similarly associated with mental health problems as instability through repartnering. Notably, we find that a history of relationship churning is predictive of mental health outcomes even after controlling for marital and cohabitation status, underlining the importance of attending to relationship instability, rather than just relationship status, in considering the mental health repercussions of this stressor. We also find no indications of reverse causality, with pre-pregnancy mental health not predictive of later churning.

Specifically, in the short term, mothers experiencing relationship churning are at higher risk for reporting a variety of mental health problems (including depression, heavy drinking, and receiving mental health treatment) compared to those in stable romantic unions. In the long term, mothers experiencing relationship churning are at higher risk for receiving mental health treatment compared to those in stable romantic unions. This could suggest that relationship instability stressors create challenges with depression and alcohol in the short term, which then have lingering consequences that mothers address by seeking treatment over the long term. Therefore, we observe differences in mental health symptoms in the short term and mental health treatment in the long term. Additionally, we find few significant differences in mental health outcomes between mothers in churning relationships and those who stably dissolved their unions and then repartnered. This tells us about the type of relationship stressor that may give rise to later mental health challenges. Instability, rather than a change in partner, may be key in the processes connecting relationship experiences with mental health outcomes. Additional research further elucidating the churning concept will continue to shed light on whether and how it is distinct from repartnering, both as an experience and in its consequences for individual and family well-being.

These results are consistent with those of previous studies, such as findings that those experiencing multiple partner fertility are more likely to report depressive symptoms (Turney and Carlson 2011), churners report more distress (Monk et al. 2022), and those who experience a breakup (vs. staying together) report poorer mental health (Meadows et al. 2008). Our study represents an important contribution to our understanding of these issues, though, because it is the first to examine churning and mental health outcomes longitudinally in a large sample of mothers who predominantly identify as members of minoritized groups, groups most likely to experience stressors such as relationship churning (Halpern-Meekin and Turney 2016; Pearlin 1989).

We are cautious in arguing whether this is indicative of support for the crisis model, which suggests that the repercussions of stressors are short-lived, or the chronic strain model, which suggests that the repercussions of stressors endure over time (Meadows et al. 2008; Simon 2002; Simon and Marcussen 1999; Strohschein et al. 2005; Williams 2003; Wu and Hart 2002). This is because those in churning relationships may experience multiple points of crisis (because the relationship is disrupted multiple times), and therefore, we may not similarly observe the period in which chronic strain could develop for those in churning versus stably ended relationships. However, that relationship characteristics explain the association between relationship history and mental health outcomes in the long term but not in the short term could be seen as indicating some support for the crisis perspective. Future research with more fine-grained temporal data on relationships and mental health should consider adjudicating between these theoretical perspectives.

Limitations

This study is, of course, not without limitations. First, our measures of relationship churning likely miss some instances of relationship instability between survey waves, potentially leading to underestimates of the association between churning and mental health. Second, our assessment of short-term versus long-term implications of relationship churning are structured by the timing of the survey waves, and accordingly, we were unable to use data from the 15-year survey (given the 6-year period between the 9-year and 15-year surveys—a period we were concerned would be too long to have lapse without measurement of relationship events for the purposes of the present study). Third, some models adjusted for covariates that capture mothers’ mental health prior to the focal child’s birth; this may have post-dated the beginning of the romantic partnership and, therefore, be endogenous, meaning we could have overcontrolled our models, biasing them against finding an association between churning and mental health outcomes. Finally, given our sample size, we are unable to explore a number of potential sources of variation among subpopulations (e.g., by race-ethnicity, by coresidential status at the focal child’s birth), and our data also limit us to examining mothers who have had a child with an opposite-sex partner (for a comparison of churning experiences in same-sex and opposite-sex couples, see Monk et al. 2018, 2022). Similarly, because of attrition among fathers and the differential experience by gender parents may have after separation, we focus only on mothers in the present analysis. These limitations present ripe areas for future research on relationship churning.

CONCLUSION

Using the stress process perspective, which highlights the unequal distribution of stressors and the repercussions of stressors for mental health, the present study’s findings bolster the importance of attending to multiple types of family instability—even instability in ongoing relationships—and contribute to our understanding of the micro-level ecological factors that shape mental health outcomes. It demonstrates the importance of these factors in a key population of interest because the well-being of mothers has cascading effects in family systems, in general, and for children, specifically (Cavanagh and Fomby 2019; Hardie and Turney 2017; Turney 2011). Future research should explore ecological factors that are protective against relationship instability and against negative mental health outcomes to inform intervention and treatment efforts.

Biographies

Sarah Halpern-Meekin is a professor in human development and family studies in the School of Human Ecology and La Follette School of Public Affairs at the University of Wisconsin-Madison. She studies family relationships and family finances and the role of government policy in intervening in those areas. Her current scholarship focuses on churning relationships and their associated outcomes for parents and children, the perspectives of prime-age men who are out of the formal labor force, and the experiences of mothers with limited incomes who are receiving an unconditional cash transfer during the early years of their children’s lives.

Kristin Turney is a professor in the Department of Sociology at the University of California, Irvine. Her research examines how stressors—such as criminal justice contact, physical and mental health problems, and relationship instability—affect well-being among individuals and families. Specifically, this research considers how individuals and families have differential risks of exposure to stressors based on their positions in the social structure, how stressors influence well-being via stress contagion and stress proliferation, and how stressors have unequal consequences for families.