Relationship Churning and Parenting Stress Among Mothers and Fathers

Abstract

Researchers have documented the consequences of relationship instability for parenting stress but have given little attention to within-partner relationship instability. In this study, the authors used data from the Fragile Families and Child Wellbeing Study (N = 3,544) to estimate the association between within-partner relationship instability (known as churning or on-again/off-again relationships) and parenting stress. First, they found that by the focal child’s 5th birthday about 16% of biological parents experience churning. Second, compared to being stably together with or stably separated from the child’s other parent, churning is associated with greater parenting stress for both mothers and fathers. Because parenting stress is the same or higher among churners compared to their counterparts who stably separate, this suggests that, more than a change in partner, relationship instability—whether within or across relationships—is tied to parenting stress.

In a world of fluid family structures, the resources parents have at their disposal to support them in parenting tasks vary as romantic partners enter and exit family life. The result may be heightened parenting stress—the psychological strain created when one perceives that the demands of the parenting role exceed one’s capacity to fulfill them. Relationship transitions among both married and unmarried parents are associated with increased parenting stress (Beck, Cooper, McLanahan, & Brooks-Gunn, 2010; Cooper, McLanahan, Meadows, & Brooks-Gunn, 2009; Meadows, McLanahan, & Brooks-Gunn, 2008; Osborne, Berger, & Magnuson, 2012; Ryan, Tolani, & Brooks-Gunn, 2009). Relationship transitions—ending a cohabiting union with a child’s biological parent or moving in with a new partner, for example—may cause tensions in the family system and affect residential parents’ economic or emotional resources, increasing their parenting stress (Beck et al., 2010; Cooper et al., 2009; Osborne et al., 2012).

Previous research has treated relationship transitions as one-way, one-time events: A couple is either together or not together. This neglects the small but growing set of studies that have found that relationship churning—breaking up and getting back together—is relatively common among unmarried couples (Dailey, Pfiester, Jin, Beck, & Clark, 2009; Halpern-Meekin, Manning, Giordano, & Longmore, 2012; Vennum, Lindstrom, Monk, & Adams, 2014). In other words, unmarried couples experience both stable and unstable breakups, with some couples remaining separated and others reuniting. Grouping all couples who are in relationships and all who have broken up misses the possibility that relationship churning may distinguish the romantic and parenting experiences within and between these groups, with differences between those who are stably together versus stably broken up versus churning.

In the present study we analyzed the association between relationship churning and mother’s and father’s parenting stress using longitudinal data from the Fragile Families and Child Wellbeing Study (FFCWB; http://www.fragilefamilies.princeton.edu/), an urban cohort of children born to mostly unmarried parents. These data allowed us to estimate parenting stress as a function of relationship churning during the child’s first 5 years, adjusting for baseline characteristics at the focal child’s birth. We focused on relationship churning during children’s early years because family structure transitions during this time are particularly influential for children’s development (Cavanaugh & Huston, 2008), and parenting stress is highest when children are young (Crnic & Booth, 1991; Kuczynski & Kochanska, 1990; Neece, Green, & Baker, 2012). Parenting stress is associated with negative outcomes for both parents and children. For adults, it is linked to lower life satisfaction and more psychological distress (Crnic & Greenberg, 1990; Thompson, Merritt, Keith, Bennett, & Johndrow, 1993). For children, caregivers’ experience of parenting stress is negatively associated with socioemotional well-being and cognitive development (Crnic, Gaze, & Hoffman, 2005; Magill-Evans & Harrison, 2001).

Understanding Relationship Churning

Because churning is more common in nonmarital than marital relationships (Vennum et al., 2014), the growth in child rearing outside marriage means that increasing numbers of children likely experience parental relationship churning. However, most of the previous research on relationship churning outside the marital context has largely not focused on parents, relying instead on convenience samples of college students (e.g., Dailey, Hampel, & Roberts, 2010; Dailey, Jin, Pfiester, & Beck, 2011; Dailey, Middleton, & Green, 2012; Dailey, Pfiester, et al., 2009) or a sample of unmarried young adults (Halpern-Meekin et al., 2012; Halpern-Meekin, Manning, Giordano, & Longmore, 2013). An exception comes in research focused specifically on unmarried, cohabiting parents. Nepomnyaschy and Teitler (2013) found higher levels of material hardship but little reduction in father involvement among churning parents. Ethnographic studies suggested that the bond of shared children might have drawn these churning couples back together repeatedly (Cross-Barnet, Cherlin, & Burton, 2011; Roy, Buckmiller, & McDowell, 2008) [Correction added after online publication on May 23, 2016: Author name “Cross-Barnett” changed to correct spelling, “Cross-Barnet” in the previous sentence.]. Given that approximately half of unmarried parents are not cohabiting at their child’s birth (McLanahan & Beck, 2010), the present study makes a significant contribution to existing knowledge by including parents in and out of coresidential relationships in examining churning.

Linking Relationship Transitions to Parenting Stress

Theoretically, there are reasons to expect that within-partner relationship instability is associated with greater parenting stress among mothers and fathers. Previous studies have suggested that relationship status, economic disadvantage, and mental health problems (e.g., depression) are linked to relationship transitions, such as breaking up, and parenting stress (Cooper et al., 2009; Deater-Deckard & Scarr, 1996; Kalil, Ziol-Guest, & Coley, 2005; Meadows et al., 2008; Ryan et al., 2009). Like any breakup, churning likely results in declines in emotional and practical support between partners when the couple is separated, leaving fewer resources to invest in and cope with parenting tasks. Unlike stable separations, however, churning involves multiple transitions with the same partner, potentially weakening trust between partners and disrupting the establishment of family routines and roles. Likewise, the causal pathway linking churning and parenting stress may run in the opposite direction, with increased parenting stress eroding the couple bond or ability to cope with relationship stressors and conflicts, thereby increasing the possibility of churning.

A host of additional factors may also confound the relationship between churning and parenting stress. For example, teen mothers report heightened parenting stress compared to older mothers (Passino et al., 1993). Nepomnyaschy and Teitler (2013) found that, compared to stable cohabitors, churning cohabitors were more likely to be Black, be under age 21 at the child’s birth, and have less than a high school diploma, and they were less likely to have worked in the past week; churners are also more likely to have been raised outside of a two-parent family (Halpern-Meekin et al., 2012). Those in interracial relationships receive less social support (Chito Childs, 2005), and those with depression may have weakened coping skills (Deater-Deckard, 1998), raising the likelihood of experiencing parenting stress. Mothers with lower levels of education are prone to heightened stress with higher order births (Tach, 2012), meaning the focal child’s birth order should be taken into account in predicting parenting stress. Some research suggests that child gender may be predictive of fathers’ involvement (Lamb, 2000), which has implications for parenting stress. Parenting stress is higher among those whose children have more difficult temperaments (Williford, Calkins, & Keane, 2007). Therefore, we controlled for these factors to isolate the association between churning and parenting stress.

In the present study we make several contributions to research on romantic relationships and parenting in three areas. First, we show how a previously neglected form of relationship instability—within-partner relationship churning—is related to mother’s and father’s parenting stress. In doing so, we describe the frequency of relationship churning and the characteristics of churning parents, compared to nonchurning parents, to reveal the distinct nature of this population. Second, we isolate relationship instability from partner transitions, given that churning partners experience instability within one union, and we can begin to address whether instability matters over and above a change in partner. Finally, examining whether and how relationship churning is associated with mother’s and father’s parenting stress provides an understanding of an important relationship status for which future research and parenting interventions should account. To this end, we analyze and discuss alternative approaches to measuring relationship churning as well as discordance between partners in their reports of churning.

Method

Fragile Families and Child Wellbeing Study

We used data from the FFCWB to estimate the association between relationship churning and parenting stress. The FFCWB is a cohort of urban children, sampled from 20 cities with populations greater than 200,000, born to mostly unmarried parents in 1998–1999 and followed longitudinally. Mothers were interviewed in the hospital after giving birth, and fathers were interviewed in person as soon as possible after the birth (mostly, but not always, in the hospital). Both parents were reinterviewed by telephone when their children were approximately 1, 3, 5, and 9 years old. We used data through the 5-year survey. Response rates were relatively high, with about 86% of sampled mothers and 78% of sampled fathers participating in the baseline survey and, of these, 90%, 88%, and 87% of mothers and 69%, 67%, and 64% of fathers completing the 1-, 3-, and 5-year surveys, respectively. For more information about the survey design, please see Reichman, Teitler, Garfinkel, and McLanahan (2001).

The first analytic sample, used for most analyses, comprised 3,544 of the 4,898 individuals in the baseline FFCWB. We first deleted the 1,265 observations (26%) missing information on relationship status at any wave through the 5-year survey. Nearly all of these observations were missing because of nonparticipation and not item nonresponse (1, 542, 672, and 764 were missing information on relationship status at the baseline, 1-, 3-, and 5-year surveys, respectively, with some missing information at multiple waves). We then deleted the additional 28 observations (<1%) missing mother’s parenting stress at the 5-year survey. Also, to ensure clean comparisons between those who experience relationship churning and others, we deleted the 89 observations (2%) that did not fit into one of our four relationship history categories (see next section). The second analytic sample, used for estimates of father’s parenting stress, comprised 2,331 observations because this analytic sample excluded the 1,213 observations (25%) missing father’s parenting stress at the 5-year survey. Estimates of mother’s parenting stress are robust to using this second, smaller sample (results not presented).

Measures

Parenting stress.

The dependent variable, parenting stress, was an average of a parent’s responses to the following four questions at the 5-year survey (1 = “strongly disagree,” 2 = “disagree,” 3 = “agree,” 4 = “strongly agree”): (a) “Being a parent is harder than I thought it would be”; (b) “I feel trapped by my responsibilities as a parent”; (c) “I find that taking care of my children is much more work than pleasure”; and (d) “I often feel tired, worn out, or exhausted from raising a family” (α = .62 for mothers, α = .58 for fathers).

Relationship history.

The key explanatory variable, relationship history, was measured by a series of mutually exclusive dummy variables: relationship churning (reference category), stably together, stably broken up, and repartnered. Relationship churning was measured by a combination of direct and indirect reports of relationship churning. Mothers were asked directly about relationship churning at the baseline, 3-year, and 5-year surveys (but not at the 1-year survey; therefore, the prevalence of relationship churning in our study is underestimated). For example, at baseline mothers were asked to best describe their current relationship with the child’s biological father (response categories included “romantically involved on a steady basis,” “involved in an on-again and off-again relationship,” “just friends,” “hardly ever talk to each other,” and “never talk to each other”), and we considered mothers who reported an on-again/off-again relationship to have experienced relationship churning. At the 3- and 5-year surveys, mothers were asked to characterize their relationships with the biological father as steady or on-again/off-again, and we considered mothers who reported an on-again/off-again relationship to have experienced relationship churning. We also considered an indirect measure of churning, coded affirmatively if a mother reported being in any (marital, cohabiting, or nonresidential romantic) relationship with the biological father at one survey wave, not in a relationship with him at the following survey wave, and in a relationship with him again at a subsequent survey wave (or similar combinations of between-wave churning). The direct and indirect measures of churning were only weakly correlated (r = .106); supplemental analyses, presented below, consider the independent influence of these direct and indirect measures.

In the present analysis we used mothers’ reports of relationship churning to predict both their own and fathers’ parenting stress. We did this for two reasons. First, there were far more missing data from fathers; therefore, the requirement to have no missing data on relationship status at each survey wave resulted in a more restricted and selective sample for fathers than it did for mothers (because mothers had higher response rates). Second, research indicates that male partners are affected by female partners’ perceptions of their relationship, with this association being far stronger than the reverse (female partners being affected by male partners’ relationship perceptions; see, e.g., Faulkner, Davey, & Davey, 2005, and Kurdek, 1995); furthermore, women’s perceptions of the qualities of their relationship are more strongly associated with the risk of relationship dissolution than are men’s (Frisco & Williams, 2003).

We compared parents who experience relationship churning with three groups. Parents were coded as stably together if they reported a marital, cohabiting, or nonresidential romantic relationship with the child’s other biological parent at all four survey waves (baseline, 1-year, 3-year, and 5-year surveys) and reported no churning. Parents were coded as stably broken up if they dissolved their relationship with the other biological parent and reported no repartnering (and no churning). Parents were coded as repartnered if they dissolved their relationship with the other biological parent and reported repartnering (and no churning). Parents in the stably broken-up and repartnered groups could have separated from the other biological parent at any point. Stably together and churning parents may reside together or not, as coresidence status was asked separately from relationship status. As noted above, 89 observations did not neatly fit into one of these four categories, and we dropped these observations from the analyses. Supplemental analyses (not shown) that include these 89 observations in the repartnered category, the relationship history group to which they are most similar, yielded substantively similar findings.

Control variables.

We adjusted for the following characteristics of parents: race (indicated by mother’s race and interracial relationship status), age, childhood family structure (1 = lived with both biological parents at age 15), educational attainment, relationship status at the focal child’s birth, material hardship (a sum of affirmative responses to 12 questions [e.g., evicted from home or apartment for not paying the rent or mortgage in the past year; borrowed money from friends or family to help pay bills in past year]), employment (1 = employed in the past week), and depression (measured by the Composite International Diagnostic Instrument—Short Form). We also adjusted for child temperament, an average of six questions (e.g., child fusses and cries, child gets upset easily; α = .51) measured on a scale that ranged from 1 (not like my child at all) to 5 (very much like my child; Buss & Plomin, 1984), child birth order (1 = first born), and child gender (1 = boy). To ensure appropriate time ordering between the dependent, explanatory, and control variables (and especially to ensure that the control variables were measured at or prior to relationship churning), the control variables were measured at baseline or, when variables were not ascertained at baseline, as close to baseline as possible.

Analytic Strategy

The analyses occurred in two stages. First, we examined the demographic, socioeconomic, and behavioral characteristics of parents who reported relationship churning, compared to parents in the other three groups (stably together, stably broken up, and repartnered). We tested for statistically significant differences, depending on the distribution of the outcome variable, with chi-square or t tests. Second, we used ordinary least squares regression models to estimate mother’s and father’s parenting stress as a function of relationship churning. The first model estimates the unadjusted association. The second model adjusts for a set of basic demographic controls (e.g., race/ethnicity, education) that likely occurred temporally prior to relationship churning. The third model adjusts for additional controls (e.g., material hardship, depression) that may have occurred concurrent to or after relationship churning and therefore may be endogenous. We consider these final estimates as conservative, as we may be controlling for some of the pathways linking relationship churning to parenting stress. Relatively few covariates were missing data, and we preserved data by imputing 20 data sets and averaging results across imputations. All analyses were conducted in Stata 14.0.

Results

Descriptive Statistics of Relationship Churning

In Table 1 we present the percentages of parents who experienced relationship churning. Relationship churning was most common at baseline; 9% directly reported being in an on-again/off-again relationship at the child’s birth. About 2% directly reported being in an on-again/off-again relationship at either the 3- or 5-year surveys, and 5% reported indirect churning (i.e., between-wave churning) through the 5-year survey. Taken together, about 1 in 6 couples (16%) experienced relationship churning by the focal child’s fifth birthday.

Table 1.

Descriptive Statistics of Relationship Churning (N = 3,544)

Relationship churning	%
Direct churning
Churning at baseline surveya	8.7
Churning at 3-year surveya	2.3
Churning at 5-year surveya	1.6
Any direct churning	11.6
Indirect churning
Between-wave churning through 5-year surveyb	5.4
Total churning
Any churning through 5-year surveyc	15.7

^aIndicates that the mother reported that she and the child’s biological father were in an “on-again/off-again” relationship.

^bIndicates that the mother reported a relationship with the child’s biological father at one wave, no relationship with him at the next wave, and again a relationship with him at the next wave (or similar combinations of between-wave churning).

^cIndicates that the mother reported any churning at the baseline, 3-year, or 5-year surveys or that she reports between-wave churning.

Comparing Parents Who Report Relationship Churning to Other Groups of Parents

In Table 2, we examine characteristics of parents who experienced relationship churning (measured either directly or indirectly) and compare these characteristics to three other groups of parents with different relationship histories (stably together, stably broken up, and repartnered). Overall, parents who experienced churning were a relatively disadvantaged group. Compared to stably together parents (those in a stable relationship throughout the survey waves), relationship churners were less likely to live with both biological parents at age 15 (34% of mothers and 37% of fathers, compared to 56% of mothers and 57% of fathers). Relationship churners had lower educational attainment (42% of mothers and 39% of fathers had less than a high school diploma at baseline, compared to 24% of stably together mothers and 25% of stably together fathers). At the birth of their child, these parents were less likely than those who stayed stably together to be in marital (2% compared to 51%) or cohabiting (30% compared to 38%) relationships and more likely to have nonresidential romantic relationships (61% compared to 11%) or no relationships (8% compared to 0%). They were less likely to be employed (47% of churners compared to 55% of stably together mothers; 65% compared to 87% among fathers), reported more material hardship (1.643 for churners compared to 0.786 among stably together mothers; 0.708 compared to 0.052 among fathers), and were more likely to report depression (21% of churners compared to 11% of stably together mothers; 15% compared to 6% among fathers).

Table 2.

Descriptive Statistics of All Variables Included in Analyses, by Relationship History (N = 3,544)

		Relationship history
		Churning	Stably together (n = 1,488)			Stably broken up (n = 508)			Repartnered (n = 993)
Variable	Entire sample	(n = 555; % or M)	% or M	Cohen’s d	p	% or M	Cohen’s d	p	% or M	Cohen’s d	p
Mother race (b)
Non-Hispanic White	22.3%	9.4%	33.9%	−0.567	***	13.8%	−0.139	*	16.4%	−0.205	***
Non-Hispanic Black	48.9%	64.3%	31.7%	0.696	***	58.7%	0.117	†	61.1%	0.066
Hispanic	25.3%	23.8%	29.4%	−0.126	*	25.2%	−0.033		20.0%	0.091	†
Non-Hispanic other race	3.5%	2.5%	5.0%	−0.121	*	2.4%	0.010		2.4%	0.007
Mother and father are mixed-race couple (b)	14.2%	14.6%	12.2%	0.073		12.0%	0.076		18.0%	−0.092	†
Mother age (b)	25.215	24.211	27.182	−0.488	***	25.181	−0.161	**	22.847	0.256	***
Father age (b)	27.693	26.563	29.562	−0.426	***	27.290	−0.102	†	25.730	0.125	*
Mother lived with both biological parents at age 15 (b)	42.9%	34.4%	55.8%	−0.435	***	37.0%	−0.054		31.2%	0.068
Father lived with both biological parents at age 15 (b)	46.2%	36.9%	56.9%	−0.406	***	42.7%	−0.118	†	37.0%	0.000
Mother educational attainment (b)
Less than high school	32.1%	42.2%	23.5%	0.421	***	35.4%	0.138	*	37.5%	0.096	†
High school diploma or GED	31.0%	34.1%	26.9%	0.157	**	33.1%	0.021		34.4%	−0.008
Some college	25.4%	21.8%	26.9%	−0.118	*	25.4%	−0.085		25.1%	−0.077
College degree	11.5%	2.0%	22.6%	−0.567	***	6.1%	−0.213	**	3.0%	−0.065
Father educational attainment (b)
Less than high school	31.4%	39.1%	25.4%	0.304	***	32.3%	0.142	*	35.5%	0.074
High school diploma or GED	36.3%	41.4%	27.9%	0.294	***	42.5%	−0.022		42.9%	−0.030
Some college	21.8%	17.7%	26.1%	−0.198	***	20.3%	−0.067		18.3%	−0.017
College degree	10.6%	1.8%	20.6%	−0.534	***	4.9%	−0.175	**	3.2%	−0.087	†
Child is firstborn	38.6%	35.0%	36.8%	−0.039		40.0%	−0.104	†	42.7%	−0.158	**
Child is a boy	52.4%	50.1%	52.5%	−0.048		48.6%	0.029		55.5%	−0.108	*
Mother and father relationship status at birth (b)
Married	25.8%	2.2%	50.5%	−1.116	***	12.8%	−0.419	***	8.7%	−0.269	***
Cohabiting	36.0%	29.5%	38.4%	−0.186	***	41.5%	−0.253	***	33.1%	−0.077
Nonresidential romantic	26.7%	60.7%	11.0%	1.346	***	29.1%	0.668	***	30.1%	0.651	***
Separated	11.4%	7.6%	0.0%	0.549	***	16.5%	−0.280	***	28.1%	−0.522	***
Mother material hardship (y1)	1.120	1.643	0.786	0.570	***	1.154	0.286	***	1.313	0.185	***
Father material hardship (y1)	0.394	0.708	0.052	0.767	***	0.477	0.179	**	0.687	0.016
Mother employed (y1)	53.9%	47.4%	54.8%	−0.148	**	56.3%	−0.179	**	55.0%	−0.152	**
Father employed (y1)	77.0%	64.5%	86.7%	−0.580	***	75.6%	−0.248	***	70.1%	−0.120	*
Mother depression (y1)	15.3%	21.4%	11.0%	0.304	***	14.6%	0.179	**	18.7%	0.068
Father depression (y1)	10.7%	15.3%	6.2%	0.328	***	11.8%	0.102	†	14.3%	0.029
Child temperament (y1)	3.406	3.300	3.499	−0.265	***	3.375	−0.097		3.342	−0.055

Note. Timing of measurement of all variables in parentheses (b = baseline interview, y1 = 1-year interview). Asterisks compare mothers who reported churning with other groups of mothers.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Parents who experienced relationship churning are also disadvantaged compared to the remaining two groups of parents: (a) the stably broken up (parents who broke up with one another and did not repartner) and (b) the repartnered (parents who broke up with one another and started relationships with new partners). For example, parents who experienced relationship churning, compared to the stably broken up, were less likely to be married (2% compared to 13%) or cohabiting (30% compared to 42%) at the child’s birth. They were also less likely to have a college degree (2% compared to 6% among mothers, 2% compared to 5% among fathers), reported more material hardship (1.643 compared to 1.154 among mothers, 0.708 compared to 0.477 among fathers), and were less likely to be employed (47% compared to 56% among mothers, 65% compared to 76% among fathers).

Parents who repartnered had demographic characteristics most similar to parents who reported churning, but differences also exist between these two groups. Parents who experienced relationship churning, compared to the repartnered, were less likely to be married (2% compared to 9%) or cohabiting (29% compared to 33%). They reported more material hardship (1.643 compared to 1.313 among mothers) and were less likely to be employed (47% compared to 55% among mothers).

Estimating Mother’s Parenting Stress as a Function of Relationship History

Table 3 presents estimates of mother’s parenting stress as a function of relationship history (comparing the stably together, the stably broken up, and the repartnered to those who experience relationship churning, measured both directly and indirectly). Model 1, the unadjusted association, shows that mothers in stably together relationships reported significantly less parenting stress than mothers who experienced relationship churning (b = −0.184, p < .001). This was also true of stably broken up mothers (b = −0.140, p < .01). The difference in parenting stress between repartnered mothers and mothers who reported relationship churning was marginally significant (b = −0.068, p < .10). Model 2, which adjusts for basic demographic controls, shows that mothers in stably together relationships (b = −0.167, p < .001) and mothers in stably broken up relationships (b = −0.128, p < .01) reported less parenting stress than their counterparts who experienced relationship churning. These associations remain statistically significant, albeit reduced in magnitude, in Model 3, which adjusts for several potentially endogenous characteristics. We tested for statistically significant differences among the three comparison groups in the final model and found that stably together mothers reported significantly less parenting stress than their repartnered counterparts (p < .05).

Table 3.

Ordinary Least Squares Regression Models Estimating Mother’s Parenting Stress as a Function of Relationship History (N = 3,544)

	Model 1 Unadjusted relationship			Model 2 + Basic controls			Model 3 + Endogenous controls
Predictor	b	SE	p	b	SE	p	b	SE	p
Relationship history (ref.: churning)
Stably together	−0.184	0.034	***	−0.167	0.036	***	−0.107	0.039	**
Stably broken up	−0.140	0.042	**	−0.128	0.041	**	−0.084	0.042	*
Repartnered	−0.068	0.036	†	−0.055	0.036		−0.024	0.037
Mother race (ref.: non-Hispanic White)
Non-Hispanic Black				−0.022	0.033		−0.030	0.034
Hispanic				−0.053	0.036		−0.045	0.035
Non-Hispanic other race				0.078	0.066		0.063	0.065
Mother and father are mixed-race couple				−0.005	0.035		−0.013	0.035
Mother age				0.002	0.003		0.003	0.003
Father age				0.004	0.003		0.003	0.003
Mother lived with both biological parents at age 15				−0.005	0.025		0.007	0.025
Father lived with both biological parents at age 15				−0.061	0.027	*	−0.060	0.027	*
Mother educational attainment (ref.: less than high school)
High school diploma or GED				−0.132	0.030	***	−0.104	0.030	**
Some college				−0.177	0.034	***	−0.142	0.035	***
College degree				−0.136	0.054	*	−0.077	0.055
Father educational attainment (ref.: less than high school)
High school diploma or GED				−0.050	0.030	†	−0.047	0.030	†
Some college				−0.041	0.360		−0.032	0.036
College degree				−0.012	0.055		0.004	0.054
Child is firstborn				−0.026	0.026		0.007	0.026
Child is a boy				0.027	0.023		0.030	0.022
Mother and father relationship status at birth (ref.: married)
Cohabiting							−0.011	0.035
Nonresidential romantic							0.013	0.041
Separated							−0.022	0.049
Mother material hardship							0.031	0.008	***
Father material hardship							0.001	0.013
Mother employed							−0.063	0.024	**
Father employed							−0.026	0.034
Mother depression							0.206	0.032	***
Father depression							0.020	0.041
Child temperament							−0.088	0.015	***
Intercept	2.300			2.316			2.508
df	3,540			3,524			3,514
R2	.010			.028			.062

Note. Coefficients for stably together and repartnered are statistically different from each other (p < .05). ref. = reference category.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

The other covariates, though not the central focus of these analyses, worked as expected. Compared to their counterparts with less than a high school diploma, mothers with a high school diploma or GED (b = −0.104, p < .01) and mothers with some college (b = −0.142, p < .001) reported less parenting stress in the final model. Mother’s material hardship (b = 0.031, p < .001) and depression (b = 0.206, p < .001) were positively associated with parenting stress, and mother’s employment was negatively associated with parenting stress (b = −0.063, p < .01). Child’s temperament was negatively associated with parenting stress (b = −0.088, p < .001).

Estimating Father’s Parenting Stress as a Function of Relationship History

In Table 4 we consider father’s parenting stress as a function of relationship history. Model 1, the unadjusted association, shows that, compared to fathers who experienced churning (measured directly and indirectly), stably together (b = −0.200, p < .001), stably broken up (b = −0.189, p < .01), and repartnered (b = −0.138, p < .01) fathers all reported less parenting stress. These associations persist (though are reduced in magnitude) in Model 2, which adjusts for basic demographic controls, as well as in Model 3, which adjusts for potential endogenous controls. Because it is more likely for fathers than mothers to engage in nonresidential parenting (see http://www.fragilefamilies.princeton.edu/documents/FragileFamiliesandChildWellbeingStudyFactSheet.pdf), we reran these analyses including a measure of father’s coresidence with children at the 5-year survey (when parenting stress was measured); the findings remained robust (results not presented). Similar to our estimates of mother’s parenting stress, we again tested for statistically significant differences among the three comparison groups in the final model. There were no statistically significant differences in father’s parenting stress among the stably together, the stably broken up, and the repartnered.

Table 4.

Ordinary Least Squares Regression Models Estimating Father’s Parenting Stress as a Function of Relationship History (N = 2,331)

	Model 1 Unadjusted relationship			Model 2 + Basic controls			Model 3 + Endogenous controls
Predictor	b	SE	p	b	SE	p	b	SE	p
Relationship history (ref.: churning)
Stably together	−0.200	0.043	***	−0.184	0.045	***	−0.120	0.049	*
Stably broken up	−0.189	0.058	**	−0.173	0.058	**	−0.134	0.059	*
Repartnered	−0.138	0.051	**	−0.127	0.052	*	−0.111	0.052	*
Mother race (ref.: non-Hispanic White)
Non-Hispanic Black				−0.005	0.041		−0.016	0.042
Hispanic				−0.057	0.044		−0.037	0.044
Non-Hispanic other race				0.125	0.078		0.110	0.077
Mother and father are mixed-race couple				−0.063	0.046		−0.074	0.046
Mother age				−0.008	0.004	*	−0.009	0.004	*
Father age				0.002	0.003		0.001	0.003
Mother lived with both biological parents at age 15				0.043	0.032		0.048	0.032
Father lived with both biological parents at age 15				0.022	0.033		0.022	0.033
Mother educational attainment (ref.: less than high school)
High school diploma or GED				−0.098	0.040	*	−0.078	0.041	†
Some college				−0.060	0.045		−0.035	0.046
College degree				0.001	0.065		0.031	0.068
Father educational attainment (ref.: less than high school)
High school diploma or GED				−0.034	0.039		−0.023	0.039
Some college				−0.111	0.046	*	−0.098	0.046	*
College degree				−0.030	0.066		−0.020	0.067
Child is firstborn				−0.082	0.033	*	−0.081	0.033	*
Child is a boy				−0.014	0.029		−0.013	0.029
Mother and father relationship status at birth (ref.: married)
Cohabiting							−0.041	0.041
Nonresidential romantic							0.006	0.051
Separated							0.116	0.081
Mother material hardship							0.006	0.010
Father material hardship							0.029	0.018	†
Mother employed							−0.034	0.030
Father employed							−0.111	0.041	**
Mother depression							0.028	0.042
Father depression							0.163	0.055	**
Child temperament							−0.001	0.020
Intercept	2.182			2.454			2.469
df (residual)	2,327			2,311			2,301
R2	.010			.026			.039

Note. ref. = reference category.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

The other covariates worked as expected. For example, father’s employment was negatively associated with parenting stress (b = −0.111, p < .01), and father’s depression was positively associated with parenting stress (b = 0.163, p < .01).

Supplemental Analyses

The measure of relationship churning, as described above, is limited because (a) it combines both direct and indirect report of churning, (b) does not take into account the timing of churning, and (c) does not consider the potential discordance between mother’s and father’s reports of churning. We considered each of these three points in supplemental analyses. First, we found that indirect reports of churning, compared to direct reports of churning, were associated with less parenting stress among mothers and fathers (see Appendix Table A1). Second, we found proximal churning (measured as relationship churning at the 3- or 5-year surveys) was more strongly associated with mother’s and father’s parenting stress at the 5-year survey than distal churning (measured as relationship churning at the baseline survey; see Appendix Table A2). Third, we found that discordance between mother’s and father’s reports of churning (present in 14% of observations), net of mother-reported churning, was associated with greater parenting stress among mothers, but not fathers (see Appendix Table A3).

Discussion

In this study we used data from the FFCWB, a survey that sampled mostly unmarried parents, to provide the first examination of the association between relationship churning and parenting stress among cohabiting and noncohabiting mothers and fathers of young children. The analyses suggest two main conclusions.

First, we found that relationship churning was fairly common among parents: More than 1 in 6 reported breaking up and getting back together by the time the child was 5 years old. This prevalence of relationship churning is lower than that found in previous studies (Halpern-Meekin et al., 2012; Vennum et al., 2014), likely because we exclusively focused on parents (who may be less likely than their nonparent counterparts to experience relationship churning) and because we measured churning only from the child’s birth (so we did not capture all churning a couple experienced prior to the pregnancy). On a related note, we found that relationship churners are a distinct population of parents. Compared to their stably together or broken-up counterparts, churners had lower educational attainment and, shortly after the child’s birth, reported greater material hardship and lower employment rates. These findings are in line with those of previous studies of young adult, national, and cohabiting parent samples of churners (Halpern-Meekin et al., 2012; Nepomnyaschy & Teitler, 2013; Vennum et al., 2014).

Second, after adjusting for an array of covariates, we found that relationship churners experienced more parenting stress than those in stable relationships as well as those who separated from their child’s other parent. Previous studies have shown that relationship transitions, such as union dissolution, are associated with parenting stress (Beck et al., 2010; Cooper et al., 2009) and that transitions out of relationships mattered more for parenting stress than transitions into new relationships (Osborne et al., 2012). Our results indicate that partner change is less important than relationship instability in predicting parenting stress. Understanding the effects of relationship instability on parents and children therefore requires capturing the on–off cycles partners can experience within one relationship; that is, there can be relationship instability, with its attendant consequences, without partner change. The results also suggest there might be some small differences in the association between churning and parenting stress for mothers versus fathers (e.g., there is a difference in parenting stress for fathers who churn vs. repartner, but not for mothers); future research should explore and explain such gender differences.

The results of the present study also offer insights into measurements of churning. First, we found that the direct and indirect churning measures—the former captured through direct reports and the latter by comparing changes in couple relationship status over survey waves—largely captured different sets of respondents and that churning measured directly was more strongly associated with parenting stress. Although fully understanding this result likely requires qualitative research to better know how partners perceive and define their own relationship transitions, we speculate that viewing one’s own relationship as volatile (e.g., describing it as “on-again/off-again” on a survey) may be more likely to indicate a relationship experience that is disrupting emotional, financial, and coparenting support, thereby leading to more parenting stress. This suggests that studies without direct measures of churning, in which the researcher relies on changes between waves in the reported status of the relationship, may be missing information necessary to distinguish between churning experiences. Second, our results indicate that more proximate measures of churning are more strongly associated with parenting stress. This underlines the importance of longitudinal data that comprise information on within-partner instability at each wave; knowing whether or not partners have ever experienced churning may not be adequate. Third, we noted some discordance between mothers and fathers in reports of relationship churning. What this discordance means about partners’ relationship experiences is an empirical question, likely best answered with qualitative research; it also raises issues for measurement, serving as a reminder that surveying only one partner about relationship status provides incomplete data.

It is possible that we underestimated the association between relationship churning and parenting stress, for two reasons. First, if a substantial amount of time passed between a relationship disruption and the survey, mothers may not report being in an on-again/off-again relationship, or if a couple are surveyed immediately postbreakup, a reconciliation may be coming in the future but is not yet observed. Second, we did not account for any association between churning and parenting stress in the latest relationships of parents who have repartnered (who are in a relationship with someone other than the child’s biological parent), masking the net effect of churning overall. Taken together, these factors suggest that we underestimated the frequency of churning as well as the association between relationship churning and parenting stress; the results presented here therefore likely are conservative estimates.

Future research should build on the present study and explicitly compare cohabiting and noncohabiting churners, given that these living arrangements may be fluid and ill defined over the 5-year period observed (Manning & Smock, 2005; Teitler, Reichman, & Koball, 2006). In our sample, 59% of churners reported cohabiting at some point during the 5-year period of observation. Supplemental analyses (available on request) indicated no statistically significant interactions between relationship transitions and coresidential status when predicting parenting stress. In addition, the results of the present study may be limited by the relatively low reliability of the measures of child temperament and parenting stress; although these two scales are common in studies relying on FFCWB data (e.g., Bronte-Tinkew, Horowitz, & Carrano, 2010; Meadows, McLanahan, & Brooks-Gunn, 2007; Turney & Wildeman 2013), they nonetheless should be verified in future research.

The present study documents the sort of relationship instability among parents that is generally ignored. Relationship churning is not uncommon and is associated with experiencing higher levels of parenting stress. This is consequential because parenting stress is predictive of more negative outcomes for both parents and children (Crnic et al., 2005; Crnic & Greenberg, 1990; Magill-Evans & Harrison, 2001; Thompson et al., 1993). Although respondents who broke up only once and then entered a relationship with a new partner experienced several relationship transitions, churners also experience several relationship transitions without a change in partner. The similarity in parenting stress between churners and those who repartner suggests the possibility that it is relationship instability, rather than partner change, that may exacerbate parenting stress. Both future research and interventions for parents must attend to the possibility of relationship churning and the potential consequences that coping with such uncertainty and change may bring for parenting behaviors and, ultimately, children’s well-being.

Appendix

Table A1.

Estimating Mother’s and Father’s Parenting Stress as a Function of Relationship History, Distinguishing Between Direct and Indirect Churning

	Model 1 Unadjusted relationship			Model 2 + Basic controls			Model 3 + Endogenous controls
Variable	b	SE	p	b	SE	p	b	SE	p
Panel A: Mother’s parenting stress (N = 3,544)
Relationship history (ref. = direct report of churning)
Indirect report of churning	−0.174	0.066	**	−0.164	0.065	*	−0.156	0.065	*
Stably together	−0.229	0.038	***	−0.210	0.039	***	−0.151	0.043	***
Stably broken up	−0.185	0.045	***	−0.170	0.045	***	−0.127	0.046	**
Repartnered	−0.113	0.040	**	−0.098	0.040	*	−0.069	0.041	†
Intercept	2.345			2.361			2.554
R2	.012			.030			.064
Panel B: Father’s parenting stress (N = 2,331)
Relationship history (ref. = direct report of churning)
Indirect report of churning	−0.070	0.084		−0.068	0.084		−0.075	0.085
Stably together	−0.221	0.049	***	−0.204	0.051	***	−0.143	0.056	*
Stably broken up	−0.210	0.063	**	−0.193	0.063	**	−0.158	0.065	*
Reparterned	−0.159	0.057	**	−0.147	0.057	*	−0.135	0.059	*
Intercept	2.203			2.475			2.493
R2	.010			.027			.039

Note. Some observations (N = 50) report both direct and indirect reports of churning; those observations are considered to have direct reports of churning. Models adjust for all control variables from corresponding models in Tables 3 and 4. ref. = reference category.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Table A2.

Estimating Mother’s and Father’s Parenting Stress as a Function of Timing of Churning

	Model 1 Unadjusted relationship			Model 2 + Basic controls			Model 3 + Endogenous controls
Variable	b	SE	p	b	SE	p	b	SE	p
Panel A: Mother’s parenting stress (N = 3,544)
Distal churning	0.122	0.041	**	0.100	0.041	*	0.052	0.043
Proximal churning	0.261	0.061	***	0.231	0.061	***	0.197	0.061	**
Intercept	2.164			2.229			2.434
R2	.008			.027			.063
Panel B: Father’s parenting stress (N = 2,331)
Distal churning	0.109	0.058	†	0.087	0.059		0.048	0.062
Proximal churning	0.265	0.073	***	0.253	0.073	**	0.222	0.073	**
Intercept	2.004			2.315			2.360
R2	.008			.026			.040

Note. Distal churning is a dummy variable indicating relationship churning at baseline survey. Proximal churning is a dummy variable indicating relationship churning at 3- or 5-year surveys. Models adjust for all control variables from corresponding models in Tables 3 and 4. In Model 3, coefficients for distal churning and proximal churning are statistically different from one another (p = .002 for estimates of mother’s parenting stress, p = .006 for estimates of father’s parenting stress).

^†p < .10.

^*p < ..05.

^**p < .01.

^***p < .001.

Table A3.

Estimating Mother’s and Father’s Parenting Stress as a Function of Discordance in Mother’s and Father’s Reports of Relationship Churning

	Model 1 Unadjusted relationship			Model 2 + Basic controls			Model 3 + Endogenous controls
Variable	b	SE	p	b	SE	p	b	SE	p
Panel A: Mother’s parenting stress (N = 3,544)
Discordance in churning	0.117	−0.039	**	0.104	0.039	**	0.094	0.039	*
Churning	0.087	−0.036	*	0.071	0.036	*	0.028	0.036
Intercept	2.153			2.219			2.435
R2	.008			.027			.062
Panel B: Father’s parenting stress (N = 2,331)
Discordance in churning	0.072	0.049		0.059	0.050		0.039	0.050
Churning	0.151	0.048	**	0.139	0.048	**	0.102	0.049	*
Intercept	1.992			2.291			2.353
R2	.010			.026			.039

Note. Discordance is coded affirmatively if the mother and father report differently on relationship churning at the baseline, 1-year, or 5-year surveys. Models adjust for all control variables from corresponding Models in Tables 3 and 4.

^*p < .05.

^**p < .01.