MARRIAGE AND DISSOLUTION AMONG WOMEN’S COHABITATIONS: VARIATIONS BY STEPFAMILY STATUS AND SHARED CHILDBEARING

Abstract

Cohabiting unions increasingly involve children, either born during the union and/or from prior relationships (i.e., stepchildren). Drawing from arguments about the institutionalization of cohabitation and stepfamilies as well as the family systems perspective, this paper examines dissolution and marriage risks among women’s cohabiting unions by stepfamily status, configuration (which partner has children) and shared intended and unintended fertility using the 2006–2013 National Survey of Family Growth. A minority (32%) of 1^st cohabitations, but the majority of 2^nd (65%) and 3^rd (75%) cohabitations, are stepfamilies. Stepfamily cohabitations are less likely to transition to marriage compared to non-stepfamily unions, especially among complex stepfamilies (both partners have children), but neither stepfamily status nor configuration affect dissolution. Shared intended and unintended births are associated with dissolution and marriage risks but largely only for non-stepfamily cohabitations, suggesting that shared childbearing is only indicative of the institutionalization for cohabitations that are not stepfamilies.

Are Shared Children and Stepchildren Associated with the Outcomes of Women’s Cohabitations?

The past few decades have witnessed remarkable changes in cohabitation. One of the most striking has been the growing presence of children. In the 1980s, births to cohabitors were rare, but by 2011–13, a quarter of all births were to cohabiting women (Manning, Brown, & Stykes, 2015). Childbearing within cohabitation indicates a growing acceptance of cohabitation as site for childbearing and childrearing – while marriage remains the ideal context in which to raise children (Pew Research Center, 2010), three-fourths of Americans now agree that it is okay to have and raise children when the parents live together but are not married (Stykes, 2015). From a family systems perspective (Broderick, 1993; Larsen & Olsen, 1990), the ties created by shared childbearing in cohabitation may solidify the roles and relationships in a union that is otherwise viewed as lacking widely accepted guidelines and expectations for behavior and interactions (Brown, Manning, & Payne, forthcoming; Cherlin, 2004; Nock, 1995). This, in turn, may increase overall cohabiting union stability and the likelihood of transitioning to marriage.

However, although more cohabiting unions include children (Manning, 2015), not all children are biologically related to both partners. Union dissolution among parents, combined with childbearing outside of coresidential relationships, increases the number of parents in the relationship market, and cohabitation has emerged as the modal way that all coresidential unions are formed (Manning, Brown, & Payne, 2014; Wu & Schimmele, 2005). As such, a growing proportion of cohabiting unions are stepfamilies (i.e., unions in which either partner has a child from a previous relationship). Because stepfamilies roles and relationships are ambiguous (Cherlin, 1978; Sweeney, 2010), the family systems perspective would suggest that children from prior relationships would negatively impact the stability of a cohabiting union.

To date, work examining children and cohabitation stability has largely focused on shared childbearing, neglecting the potential influence of children from prior relationships. Similarly, although we know that children in cohabiting stepfamilies fare worse than those in marital stepfamilies (Brown, 2006) and that stepfamily instability is negatively associated with well-being (see Sweeney, 2010 for a review), less is known about the stability and transitions of stepfamilies themselves. The demographic and social changes described above have altered the nature of both cohabiting unions and stepfamilies, but the stability of cohabiting stepfamilies has yet to be examined. To fill this gap, I draw from arguments about the institutionalization of cohabitation and stepfamilies as well as the family systems perspective. I examine the outcome (remaining intact, breaking up, or marrying) of women’s cohabiting unions, paying attention to the role of stepchildren and shared children using the 2006–2013 National Survey of Family Growth (NSFG). Specifically, this work examinations stepfamily status, configuration (which partner has children), and shared intended and unintended childbearing.

Theoretical Framework

A family systems approach provides a helpful way to think about cohabitations overall and especially cohabiting stepfamilies (and the difficulties they face). Popularized in the 1960s and 1970s, when married nuclear families were dominant, the family systems perspective views families as a set of individuals who each have a set of rules, roles, and connections between them (Bowen, 1978; Day, 2010; Hill & Hansen, 1960). There are clear boundaries delineating who is in and out of the family system (Boss & Greenberg, 1984; Carroll, Olson, & Buckmiller, 2007), so each member is aware of how they are connected to one another, what is expected of them, and what they can expect from each other. Although individual families vary in these rules and roles (parenthood behavior can vary across families, for example, but within the family, the parent role is clear), there are socially proscribed norms and expectations, evidence of the “institutionalization” of a family form (Cherlin, 1978, 2004; Cherlin & Furstenberg, 1994; Nock, 1995). For instance, there are strong norms about sexual fidelity between spouses (Hemez, 2015), and parents are expected to provide emotional, developmental, and financial support for children (Alstott, 2004). Unlike married nuclear families, however, neither cohabitation (Nock, 1995) nor stepfamilies (Cherlin, 1978; Cherlin & Furstenberg, 1994; Sweeney, 2010) are completely institutionalized. If institutionalization provides the clear expectations associated with functioning and stability, as argued in the family systems perspective, cohabiting stepfamilies would seem to be especially at risk of instability.

The institutionalization of cohabitation and stepfamilies

Compared to marriage, cohabitation has fewer clear rules and expectations – cohabitors lack a terminology to describe their relationships (Manning & Smock, 2005); have lower levels of income pooling and joint financial decisions (Kenney, 2004; Hamplová, LeBourdais, & Lapierre-Adamcyk, 2014); and report fewer intergenerational exchanges (Eggebeen, 2005). Cohabiting unions are more unstable than marriages and have become less likely to transition to marriage over time (Guzzo, 2014). Others, though, argue that the diffusion of cohabitation seems to have led to partial institutionalization (Brown, Manning, & Payne, forthcoming; Liefbroer & Dourleijn, 2006). Cohabitation is now part of the standard pathway of union formation, with young adults reporting that cohabitation helps ensure compatibility and prevent later divorce (Manning & Smock, 2009). As such, the majority of marriages are now preceded by cohabitation (Manning & Stykes, 2015), and the negative association between cohabitation and subsequent marital stability observed in the 1980s and 1990s has since dissipated (Manning & Cohen, 2012).

Compared to nuclear families, there are also fewer rules, obligations, and boundaries in stepfamilies (Cherlin, 1978; Stewart, 2005). Stepfamilies lack clear terminology (Thorsen & King, forthcoming), and stepparent and stepchild relationships vary widely both within and across families (Ganong, Coleman, & Jamison, 2011). Not surprisingly, boundary ambiguity is common in stepfamilies (Brown & Manning, 2009). The incomplete institutionalization seen more generally among cohabitations means, for cohabiting stepfamilies, that there are fewer guiding norms and rules upon which to model behavior as individuals simultaneously negotiate both informal partnership roles and informal stepparenting roles. Given that stepchildren increase the risk of dissolution among remarriages (Coleman, Ganong, & Fine, 2000), cohabitations in which either partner has children from a prior union have higher chances of dissolution and lower chances of marriage than cohabitations involving two childless individuals.

Stepfamily configuration

However, stepfamilies are not simple, monolithic families but instead vary widely based on who has children. Differences in configuration may be linked to the stability of cohabiting unions due to variation in the intensity of the stepparent roles and in extra- vs. intra-household strains and stressors (Sweeney, 2010). Because mothers more often retain physical custody of children (Grall, 2013), cohabiting stepfamilies formed when a mother is living with a childless man are negotiating their various roles on a continual basis. Further, the mother must divide her time and attention between her biological child and her partner; she cannot focus solely on building the romantic relationship with the partner nor can she entirely focus on her child’s needs. Conversely, in a cohabiting stepfamily in which a childless woman is partnered with a father, the children less often live in the household on a regular basis (Stewart, 2007). The childless partner may only be a part-time stepparent and the lack of clear roles and guidelines may arise infrequently, reducing the possibility for conflict, and the father may have fewer competing obligations between his partner and his children and thus able to spend more time focusing on the romantic relationship. Alternatively, fewer opportunities for stepmothers to form relationships with stepchildren and to develop routines may heighten opportunities for conflict, though it is less likely this would increase stability relative to other stepfamily configurations.

The scenario is further complicated if both partners already have children; this occurs for nearly half of currently cohabiting stepfamilies (Guzzo, forthcoming). Although the male partner’s children are usually nonresident, both partners have biological parent roles and stepparent roles, adding complications because each parent may differ in how they treat their own children relative to each other. Further, residential parents also often treat their children differently than non-residential parents (Furstenberg & Nord, 1985). Putting this all together, it seems that stability and the likelihood of transitioning to marriage would decline across a rough continuum, with childless women partnering with a father least likely to experience dissolution and most likely to marry, followed by mothers cohabiting with a childless man, and then by cohabitations in which both partners had children from a prior union.

The role of shared childbearing

Childbearing in cohabiting unions, which has become more common (Kennedy & Fitch, 2012; Lichter, 2012) may indicate increased institutionalization for cohabitation (Kiernan, 2001). Generally, shared childbearing is reflective of union commitment and stability (Lillard & Waite, 1993); that is, the most stable couples are those most likely to have a child, and the ones most likely to have a child are those that expect their union to last. Given that marrying between conception and birth is an option, those who choose to have a child while cohabiting may be demonstrating a view that cohabitation is an appropriate union in which to raise children; indeed, prior work finds that having a child while cohabiting is not significantly associated with marriage or dissolution relative to staying cohabiting (Guzzo & Hayford, 2014).

The argument that childbearing within cohabitation is evidence of institutionalization seems to assume that such births are intended – that is, cohabitors have deliberately decided to get pregnant and have a child while cohabiting. Only half of births to cohabiting women are intended (Mosher, Jones, & Abma, 2012). In fact, many births to cohabiting women result from pregnancies conceived prior to the start of cohabitation (Lichter, 2012), with so-called “shotgun marriages” declining over time (England, Wu, & Shafer, 2013). Compared to intended births, unintended births to cohabitors increase the risk of dissolution (Guzzo & Hayford, 2014), though birth intentions are unrelated to the likelihood of transitioning to marriage. Shared births among stepfamilies are common (Stewart, 2002; Holland & Thomson, 2011), but births in stepfamily unions are more likely to be unintended than births in other unions (Guzzo, 2016). While one might expect that unintended births would be detrimental for relationship functioning and thus stability, changes in relationship functioning across the transition to parenthood do not seem to vary by intendedness (for a review of this literature, see Doss & Rhoades, 2017).

Whether intended and unintended births are associated with outcomes among stepfamilies, however, remains to be seen. Since stepfamily and non-stepfamily cohabitations can marry prior to a birth, there may be little difference in the link between shared childbearing and outcomes across stepfamily status, regardless of birth intendedness. Conversely, shared childbearing may function differently for stepfamily cohabitations than non-stepfamily cohabitations. On the one hand, shared children may have a unique purpose for a stepfamily by serving as a way to “cement” the bonds (Ganong & Coleman, 1988). Through the lens of family systems theory, a shared child becomes the connection between members, reducing boundary ambiguity and creating at least some clearly defined roles and obligations. If this is the case, shared childbearing could be more strongly linked to the outcome for a stepfamily cohabitation than a non-stepfamily cohabitation. On the other, having a shared child together when one or both partners already have children – and doing so while cohabiting – is a complicated phenomenon because it represents a transition to a higher couple-level parity overall (Holland & Thomson, 2011) and a potential mismatch in parity and obligations across partners. It is possible that a shared birth could be more disruptive to a stepfamily because they face more challenges already, especially if the birth was unintended. Taken together, it is not clear whether shared childbearing affects stability differently across stepfamily status or varies by configuration.

In sum, cohabiting unions have become increasingly diverse – while many cohabitations involve two childless individuals, others include shared children, stepchildren, or both, and it is unclear how the presence of these children relate to the likelihood of dissolution and marriage. Based on the theories and literature discussed above, I posit the following hypotheses:

• Hypothesis 1

  Cohabiting unions in which one or both partners have children from a prior relationship are less likely to marry and more likely to dissolve than remain cohabiting relative to unions in which neither partner has children from a prior union.

• Hypothesis 2

  Cohabiting stepfamilies in which both partners have children from previous relationships are the least likely to transition to marriage and the most likely to dissolve relative to remaining cohabiting, followed by unions in which only the female partner has children and then unions in which only the male partner has children.

• Hypothesis 3

  Shared intended children reduce the likelihood of both marriage and dissolution, while shared unintended children reduce the likelihood of marriage and increase the likelihood of dissolution relative to remaining cohabiting compared to those with no children, but it is unclear whether this process works equally stepfamily status.

Other factors associated with outcomes and stability

There are a host of other factors related to cohabiting union stability and transitions, such as prior marriage and cohabitation experience, plans to marry, and socioeconomic and demographic characteristics. Although cohabitation is a common experience, the stability and outcomes vary across socioeconomic and demographic characteristics, with minorities and less-educated individuals at a higher risk of dissolution whereas whites and better-educated individuals are more likely to marry (Rose-Greenland & Smock, 2013). Further, childbearing within cohabitation is more often concentrated among race-ethnic minorities and the less advantaged (Lichter, 2012), as is having a child from a prior relationship (Guzzo, 2016). Older ages at the start of cohabitation increase the chances of marriage and reduce the chances of breaking up (Guzzo, 2014). Being engaged or having definite plans to marry sharply increases the chances of marriage and reduces the chances of dissolution (Guzzo, 2009). Cohabitation order is also associated with outcomes; first and third or higher cohabitations are more likely to dissolve than remain intact relative to second cohabitations (Guzzo, 2014). Prior union experiences also matter, with post-marital cohabitations more likely to transition to marriage and less likely to dissolve than cohabitations among the never-married (Guzzo & Hayford, 2014).

Data and methods

This research uses the 2006–2013 cycles of the National Survey of Family Growth (NSFG). The NSFG is a nationally representative cross-sectional survey of individuals aged 15–44 and contains a detailed retrospective history of unions and childbearing. Analyses are restricted to women because the NSFG did not collect birth intendedness in a similar manner for both men and women and because the quality of the data on men’s births, particularly for births outside of coresidential unions and for children for which men have little involvement, is problematic in most datasets (Joyner, Peters, Hynes, Sikora, Taber, and Rendall, 2012).

The analytical sample includes first and higher-order (i.e., second, third, etc.) cohabiting unions to women 15 and older at the start of cohabitation with valid start and end dates. As such, the unit of analysis is cohabitations rather than women (n = 13,417 cohabitations, of which n = 6,374 are stepfamily cohabitations, to 9,367 women). Individuals can contribute more than one spell of cohabitation if they had multiple cohabitations, and the analyses control for clustering. The analyses use event history techniques, where the data is converted into person-months, and multinomial logistic regression, with a time-varying three-category dependent variable indicating whether the cohabiting union was still intact, had transitioned to marriage, or dissolved prior to marriage. Respondents enter the risk set the month the cohabitation begins and leave when the cohabiting union dissolves or transitions to marriage or are censored at the time of interview.

To compare the outcomes of cohabiting unions by stepfamily status and configuration, I first had to identify stepfamily unions. Stepfamilies are identified by the presence of the respondent’s children born prior to the union and by responses to questions about partner’s children. Specifically, each of the respondent’s births are matched by date of birth to union start date. If a birth occurred more than six months prior to the start of cohabiting union, it is assumed to be with a different partner and thus the cohabiting union is considered a stepfamily; children born within six months of the start of cohabitation are assumed to be with the cohabiting partner (alternate specifications yielded virtually identical results). Respondents are also asked whether their partner had any children from prior relationships (though information about the age, number, and coresidence of partner’s children were not collected). Thus, stepfamily status is a dichotomous indicator of whether the respondent, her partner, or both had a child prior to the start of cohabitation, regardless of whether the children live with the cohabiting couple.

There are two variables aimed at capturing configuration. The first is a time-invariant measure: no stepchildren, only the respondent had children, only the partner had children, or both had children at the start of the cohabitation. The second measure incorporates shared intended and unintended childbearing, which varies over the course of the union. Intendedness is based on responses to a series of questions asked for every birth. Respondents are first asked “Right before you became pregnant, did you yourself want to have a(nother) baby at any time in the future?” Negative answers are characterized as unwanted births. If a woman responds affirmatively, she is asked about the timing of the pregnancy: “So would you say you became pregnant too soon, at about the right time, or later than you wanted?” Births that are identified as too late or at about the right time are also considered wanted. Births that are identified as occurring too soon are asked a follow-up question: “How much sooner than you wanted did you become pregnant?” I used two years as a cutoff point to define the extent of mistiming, with less than two years considered only moderately mistimed and births two or more years too soon as seriously mistimed and further dichotomize intendedness into wanted or moderately mistimed (intended) vs. seriously mistimed or unwanted (unintended). I then created a variable to indicate parity by intendedness: no shared births, only intended births, and any unintended births. This is a time-varying measure – the month of a first birth, women are coded as having only intended births if the birth was intended and as having any unintended births if the birth was unintended. They retain this coding for all subsequent months, with women whose first birth is intended but who experience a subsequent unintended birth being recoded as having any unintended births in the month of the first unintended birth and for all subsequent births. This operationalization glosses over potential differences by the number of births and combinations of intendedness (i.e., having both an intended and an unintended birth and having only one unintended birth are both in the latter category) but permits the interaction of stepfamily type with shared childbearing and intendedness. Rather than an actual interaction term, I created a twelve-category measure as the second configuration variable: 4 categories of stepfamily type, each with 3 categories for shared fertility. By using different omitted categories, I test whether shared intended and unintended childbearing is differently associated with cohabiting outcomes within stepfamily types.

Several covariates are also included in the model. Demographic characteristics include the race-ethnicity and the respondent’s age at union start. To proxy socioeconomic status, family background characteristics include maternal education and family structure at age 14 as well as a time-varying indicator of whether the respondent had a high school degree. Information about the respondent’s family formation experiences include cohabitation number (i.e., first, second, or third or higher cohabitation) and an indicator of prior marriage, along with information on whether the partner has ever been married (there is no information on partner’s prior cohabitation status). A dichotomous indicator of engagement status/definite plans to marry at the start of the union is also included. Given changes in cohabiting union stability over time, analyses also include an indicator when the cohabitation started, in roughly five-year increments (1984 or earlier, 1990–94, 1995–99, 2000–04, 2005–09, 2010 or later). Finally, the event history models include a control for union duration, specified as a time-varying piecewise nonlinear spline (less than 6 months, 7–12 months, 13–24 months (omitted), 25–48 months, and 49 or more months).

Analytical approach

I begin by presenting bivariate information demonstrating the prevalence of stepfamily status/configuration by cohabitation order before showing descriptive information for cohabiting unions by stepfamily status. Analyses proceed with graphs from multiple-decrement life tables, which explore whether dissolution and marriage risks vary by stepfamily configuration; these show the hazard curves within 48 months of coresidence, longer than the average duration of cohabitation (Goodwin, Mosher, & Chandra, 2010). Since neither descriptive statistics nor life tables can simultaneously account for socioeconomic, demographic, and union factors, especially shared childbearing, I turn to multivariate analyses to examine differences in outcomes using event history models accounting for exposure and duration.

Multivariate analyses occur as a series of nested multinomial logistic models, presenting the relative risk ratios (RRRs). Model 1 is a baseline model with socioeconomic and demographic characteristics, past and current union indicators, and the dichotomous indicator of stepfamily status (along with period and duration indicators), testing Hypothesis 1. Model 2 replaces the dichotomous stepfamily status variable with the stepfamily configuration variable to test Hypothesis 2. Model 3 adds shared intended and unintended childbearing to Model 2 to explore whether the link between configuration and outcomes varies by shared childbearing to test Hypothesis 3, using the twelve-category interaction term. All analyses were conducted in Stata 14.1 using the “svy” commands to account for the complex sample design of the NSFG and the multiple cohabitations contributed by some respondents.

Descriptive results

Figure 1 shows the proportion of cohabiting unions by stepfamily status and cohabitation order. The first set of bars details the overall distribution, showing that the majority of all cohabitations ever formed among women aged 15–44 are not stepfamilies – 57% of all cohabitations involve two childless individuals. First cohabitations have the lowest proportions of stepfamilies, at just under a third. Among second and third cohabitations, however, the majority are stepfamilies – 65% of second cohabitations and 76% of third cohabitations. Higher-order stepfamily cohabitations are also more complex than first cohabitations. For first cohabitations, the modal category of stepfamily is one in which a childless woman is partnering with someone who has children from a prior union. For second and higher stepfamily cohabitations, however, the modal type of stepfamily is one in which both partners have children from prior unions.

Figure 1.

Stepfamily Type and Composition by Cohabitation Order

Table 1 details the socioeconomic, demographic, and family characteristics of cohabiting unions by the stepfamily status; significant differences between the distributions or means between stepfamily and non-stepfamily cohabitations are indicated by asterisks in the far right column. The general story here is that stepfamily cohabitations are comprised of less advantaged individuals. For instance, individuals in stepfamily cohabiting unions are disproportionately minority, with a lower proportion having grown up in an intact family or with a college-educated mother and a lower proportion with at least a high school degree. While women in stepfamily cohabitations tend to be older than their counterparts, the age difference is likely driven by time spent in prior unions. Roughly three-fourths of women in stepfamily cohabitations report that either they and/or their partner had been married previously, but this is true for only a tenth of non-stepfamily cohabitations. Similarly, the majority (82%) of non-stepfamily cohabitations are the first cohabitation for the respondent, but only half of stepfamily cohabitations are the first cohabitation. Fewer stepfamily cohabitations begin with plans to marry or an engagement than among non-stepfamily cohabitations. Significantly more stepfamily cohabitations had at least one birth while cohabiting, and just over half (16.8%/32.2% = 51.2%) of stepfamily cohabitations with a shared child had only intended births compared to 43.7% (11.1%/25.4%) of non-stepfamily cohabitations. Stepfamily cohabitations last, on average, about 4.5 months longer than non-stepfamily cohabitations. Finally, there seem to be differences in the outcomes of stepfamily cohabitations compared to non-stepfamily cohabitations. Fewer cohabiting stepfamilies have transitioned to marriage, by 8.4 percentage points, and slightly more stepfamily cohabitations are intact at the last month of observation. Dissolution is higher among stepfamily cohabitations than in non-stepfamily cohabitations.

Table 1.

Weighted Descriptive Statistics for Women’s Cohabitations in the 2006–2013 NSFG

	Not a stepfamily	Stepfamily
Race-ethnicity			***
White	71.4%	57.2%
Black	6.9%	20.9%
Native-bornHispanic	9.6%	9.5%
Foreign-bornHispanic	6.4%	7.1%
Other	5.7%	5.3%
Mean age at start	21.9 yrs (.122)	25.6 yrs (.126)	***
High school degree at start	74.4%	61.4%	***
Family structure at age 14			***
Both biological parents	60.7%	49.3%
Stepfamily	12.7%	13.4%
Other family	26.6%	37.4%
Maternal education			***
<HS or missing	20.7%	31.8%
HS degree	33.8%	35.8%
Some college	26.2%	20.8%
College or higher	19.3%	11.7%
Prior & current union characteristics
Respondent previously married	4.9%	31.6%	***
Partner previously married	6.2%	40.8%	***
Engaged at start	42.1%	36.7%	***
Shared childbearing by last month of observation			***
No births	74.6%	67.3%
Only intended births	11.1%	16.8%
Any unintended births	14.3%	16.0%
Cohabitation order			***
First cohabitation	81.9%	51.0%
Second cohabitation	14.4%	35.0%
Third or higher cohabitation	3.6%	14.0%
Period & duration
Time period cohabitation started			***
1984 or earlier	1.7%	1.3%
1985–1989	7.0%	5.2%
1990–1994	15.8%	11.8%
1995–1999	19.7%	19.1%
2000–2004	25.1%	25.4%
2005–2009	22.5%	27.3%
2010 or later	8.3%	10.0%
Mean duration at transition/interview	30.1 mos (.723)	34.6 mos (.814)	***
Cohabitation outcome
Intact	15.4%	17.9%	***
Married	47.9%	41.5%
Dissolved	36.7%	40.6%
Cohabitations	7,043	6,374

^*p≤.05

^**p≤.01

^***p≤.001

Significance tests across stepfamily status

Figures 2 and 3 show the hazards of marriage and dissolution by stepfamily configuration within 4 years of the start of cohabitation, derived from multiple decrement life tables. The hazard of marriage is noticeably greater for those in non-stepfamily cohabitations relative to any configuration of stepfamily cohabitations (Figure 2), consistent with Hypothesis 1; within 4 years, more than half of the non-stepfamily unions had transitioned to marriage. However, the hazard graphs are also consistent with Hypothesis 2 that suggests stepfamily stability varies by configuration. Stepfamily cohabitations in which both partners had children from a prior relationship show the lowest hazards of marriage, with roughly a third having married after 4 years. Stepfamily cohabitations in which only the partner had prior children transition to marriage more often than those in which the only the female respondent had prior children, though the proportion married is similar for the first two years of cohabitation. Dissolution hazards exhibit much less variation for marriage across stepfamily type (Figure 3), which is less consistent with Hypotheses 1 and 2. Stepfamily cohabitations in which both partners had children have the lowest hazard of dissolution (contrary to Hypothesis 2) whereas cohabitations in which a woman is partnered with someone who has children from a prior relationship have the highest hazards, but the overall differences across types are fairly small.

Figure 2.

Women’s hazard of marriage by stepfamily type within 48 months of cohabitation start

Figure 3.

Women’s hazard of dissolution by stepfamily type within 48 months of cohabitation start

Multivariate Results

Descriptive statistics and hazard graphs suggest that stepfamily cohabitations exhibit different patterns of marriage and dissolution than non-stepfamily cohabitations, consistent with Hypotheses 1 and 2, but the descriptive statistics also demonstrate that stepfamily cohabitations differ from their counterparts on a range of characteristics likely associated with outcomes, and the role of shared childbearing has yet to be fully considered. As such, I turn to multivariate event history models to better examine marriage and dissolution across cohabitations.

Table 2 shows the relative risk ratios (RRRs) from a multinomial logistic regression model to test the premise that stepfamily cohabitations differ from non-stepfamily cohabitations, as indicated by a dichotomous measure, in marriage and dissolution risks (Model 1), controlling for socioeconomic and union characteristics linked to cohabitation outcomes in prior work. In this model, we can see that stepfamily cohabitations are significantly less likely, by about 20%, to transition to marriage relative to staying cohabiting compared to non-stepfamilies. However, there is no difference in the risk of dissolution, and thus Hypothesis 1 is only partially supported.

Table 2.

Relative Risk Ratios from Multinomial Logistic Regression Event History Models Predicting Women’s Cohabitation Outcomes (Model 1)

	Married vs. intact		Dissolved vs. intact
Stepfamily status (ref=no stepchildren)
Any stepchildren	0.80	***	1.07
Socioeconomic & demographic characteristics
Race-ethnicity(ref=white)
Black	0.56	***	1.15	*
Native-bornHispanic	0.66	***	0.85	*
Foreign-bornHispanic	0.59	***	0.43	***
Other	1.02	*	0.93
Age (time-varying)	1.44	***	0.97	***
HS grad (time-varying)	1.43	***	0.98
Family structure at age 14 (ref=Two bio parents)
Stepfamily	0.92		1.20	*
Other family	0.96		1.09
Maternal education (ref=HS)
<HS or missing	0.98		0.93
Some college	1.03		1.15	*
College or higher	0.94		1.13	*
Prior & current union characteristics
Respondent previously married	1.24	***	0.94
Partner previously married	1.09		0.89
Engaged at start	2.36	***	0.76	***
Cohabitation order (ref=first cohabitation)
Second cohabitation	0.93		1.00
Third or higher cohabitation	1.07		1.18
Period & duration
Time period (ref=1995–1999)
1984 or earlier	1.33	*	0.84
1985–1989	1.18		0.84	*
1990–1994	1.18	*	0.96
2000–2004	0.88		1.08
2005–2009	0.81	*	1.58	***
2010 or later	0.81		2.17	***
Duration (ref=13–24 months, time-varying)
≤ 6 months	0.79	***	0.83	**
7–12 months	0.93		1.06
25–48 months	0.86	*	1.17	*
≥ 49 months	0.54	***	1.02
Constant	0.01	***	0.02	***
Person-months	434,653
Cohabitations	13,417
Women	9,367

^*p≤.05

^**p≤.01

^***p≤.001

The other covariates largely function consistently with prior work. Minority women’s cohabitations are generally less likely to transition to marriage than those among white women, though women in the “other” category are slightly more likely to transition to marriage than remain cohabiting. Age is positively associated with the risk of marriage, as is having at least a high school degree. Not surprisingly, cohabitors who were engaged or had definite plans to marry at the start of cohabitation are 2.4 times as likely to marry as remain cohabiting relative to those who were not engaged. The risk of marriage is higher for unions formed in earlier time periods relative to those formed in the late 1990s whereas unions formed later are less likely to transition to marriage. Cohabitations of any duration other than 13–24 months are less likely to transition to marriage than remain cohabiting. Looking at dissolution, blacks are more likely to experience dissolution than remain cohabiting compared to whites by about 15%, whereas Hispanics are less likely to experience dissolution (RRR = 0.85 for native-born Hispanics and RRR = 0.43 for foreign-born Hispanics). Increasing age reduces the risk of dissolution relative to remaining cohabiting. The respondent’s family background characteristics matter; growing up in a non-intact family raises the risk of dissolution relative to staying in an intact cohabitation, as does (counter intuitively) having a mother with more than a high school education compared to having a less educated mother. Engagement at the start of cohabitation reduces the risk of breaking up rather than remaining cohabiting by about 25%, and unions started in 2005 or later have a higher risk of dissolution relative to those started in the late 1990s. Cohabitations of 6 months or less duration are less likely (RRR = 0.83) to dissolve than stay intact compared to those of 1–2 years but those of 2–4 years are more likely to dissolve.

In Table 3, we see the results from Model 2 (testing Hypothesis 2) in which stepfamily configuration is included. The RRRs shown here are from variations of Model 2 in which the omitted category of stepfamily configuration changes to test Hypothesis 2’s supposition that there is a gradient across stepfamily configuration in the risk of marrying and dissolution. This is a full multivariate model containing the same covariates as in Table 2, but the covariate RRRs are not displayed; they are virtually unchanged from Model 1 and are available upon request.

Table 3.

Relative Risk Ratios from Multinomial Logistic Regression Event History Models Predicting Women’s Cohabitation Outcomes by Stepfamily Configuration (Model 2)

	Contrast 1 (Ref=Not a stepfamily)		Contrast 2 (Ref=Only respondent had children)		Contrast 3 (Ref= Only partner had children)
	Married vs. intact	Dissolved vs. intact	Married vs. intact	Dissolved vs. intact	Married vs. intact	Dissolved vs. intact
Configuration
Not a stepfamily	--	--	1.15	0.94	1.20	0.91
Only respondent had children	0.87	1.06	--	--	1.04	0.97
Only partner had children	0.83 *	1.10	0.96	1.03	--	--
Both had children	0.62 ***	1.01	0.72 ***	0.95	0.75 ***	0.92

^*p≤.05

^**p≤.01

^***p≤.001

Note: All models include controls for race-ethnicity, age, high school graduate status, family structure at age 14, maternal education, respondent and partner prior marital status, engagement, cohabitation order, time period, and duration.

These results suggest that not all stepfamily configurations are equally associated with the risk of marriage relative to remaining cohabiting, generally consistent with Hypothesis 2. In the columns labeled Contrast 1, in which the reference category is a non-stepfamily cohabitation, stepfamily cohabitations in which the respondent’s partner had prior children – regardless of whether the respondent herself also had children – are significantly less likely to marry than remain cohabiting. Childless women partnered with a father are 17% less likely, and mothers partnered with fathers are 38% less likely, to marry than remain cohabiting. Relative to a non-stepfamily cohabitation, the risk of dissolution relative to remaining cohabiting does not vary by configuration among stepfamily cohabitations. Moving to Contrast 2, in which the reference category is only the respondent had prior children, there are no differences in either marriage or dissolution risks for non-stepfamily cohabitations nor those in which only the male partner had children. However, stepfamily cohabitations in which both partners had children are about a quarter less likely (RRR = 0.72) to marry than remain cohabiting compared to stepfamily cohabitations in which only the female respondent had children. Finally, in Contrast 3, in which only the women’s partner had children is the omitted group, the results largely mirror those for the previous model. Relative to stepfamily cohabitations in which only the male partner had children, only stepfamily cohabitations in which both partners have children are significantly less likely to marry than remain cohabiting, again by a quarter (RRR = 0.75). In sum, Hypothesis 2 is only somewhat supported: the most complicated stepfamilies – those in which both partners have children – are significantly less likely to marry (but no more likely to dissolve) than remain cohabiting relative to all other types of cohabitations.

Table 4 displays the results from Model 3 interacting stepfamily configuration with shared intended and unintended childbearing to explore whether shared childbearing is differentially association with the stability of cohabitation (Hypothesis 3). As with Model 2 in Table 3, this table shows the results from a series of models switching the omitted category, and the RRRs for the full set of covariates are not displayed (but are virtually unchanged from Model 1, available upon request). Four sets of contrasts are shown to test whether stepfamily configuration is differently associated stability. It is worth noting, however, that in preliminary models in which shared childbearing was entered as a dichotomous variable (any shared childbearing or not) or as a three-category variable (no shared, only intended, any unintended) separate from stepfamily configuration (not shown), shared childbearing is predictive of stability and marriage. In the model with the dichotomous indicator, cohabitations with any shared children were 15% less likely to either marry or dissolve rather than remain intact compared to those with no shared childbearing. However, adding in intendedness changed this basic association. Relative to cohabitations in which there was no shared childbearing, any unintended births reduced the risk of marriage (by about 15%) whereas only intended births reduced the risk of dissolution (by about 25%). Among those with a shared birth, intendedness did not significantly predict the risk of marriage relative to remaining cohabiting, but those with any unintended births were 23% more likely to break up than those with only intended births. The main effects of stepfamily configuration did not change with the inclusion of these variables.

Table 4.

Relative Risk Ratios from Multinomial Logistic Regression Event History Models Predicting Women’s Cohabitation Outcomes by Stepfamily Configuration and Intended and Unintended Shared Childbearing (Model 3)

	Contrast 1A (Ref=Not step, no shared)		Contrast 1B (Ref=Not step, only intended		Contrast 2A (Ref=Resp. only, no shared)		Contrast 2B (Ref: Resp. only, only intended)
	Married vs. intact	Dissolved vs. intact	Married vs. intact	Dissolved vs. intact	Married vs. intact	Dissolved vs. intact	Married vs. intact	Dissolved vs. intact
Configuration
Not step, no shared children	--	--	1.18	1.47 ***	1.26 **	0.96	1.30	0.99
Not step, only intended children	0.85	0.68 ***	--	--	1.07	0.65 ***	1.10	0.68 **
Not step, any unintended children	0.74 **	0.96	0.87	1.42 **	0.93	0.92	0.96	0.96
Respondent only, no shared children	0.79 **	1.04	0.94	1.53 ***	--	--	1.03	1.04
Respondent only, only intended children	0.77	1.01	0.91	1.48 **	0.97	0.97	--	--
Respondent only, any unintended children	0.95	0.87	1.12	1.28	1.20	0.84	1.24	0.86
Partner only, no shared children	0.85	1.11	1.00	1.63 ***	1.07	1.06	1.10	1.10
Partner only, only intended children	0.65 *	0.86	0.77	1.26	0.82	0.82	0.84	0.85
Partner only, any unintended children	0.69 **	0.99	0.81	1.45 *	0.87	0.95	0.89	0.98
Both, no shared children	0.59 ***	1.04	0.70 *	1.53 **	0.75 **	1.00	0.77	1.03
Both, only intended children	0.57 ***	0.71 *	0.67 *	1.04	0.72	0.68 *	0.74	0.70
Both, any unintended children	0.58 *	0.94	0.69	1.37	0.74	0.90	0.76	0.93
	Contrast 3A (Ref=Part. only, no shared)		Contrast 3B (Ref=Part. only, only intended)		Contrast 4A (Ref=Both, no shared)		Contrast 4B (Ref=Both, only intended)
	Married vs. intact	Dissolved vs. intact	Married vs. intact	Dissolved vs. intact	Married vs. intact	Dissolved vs. intact	Married vs. intact	Dissolved vs. intact
Configuration
Not step, no shared children	1.18	0.90	1.54 *	1.16	1.69 ***	0.96	1.75 ***	1.42 *
Not step, only intended children	1.01	0.61 ***	1.30	0.79	1.43 *	0.64 **	1.48 *	0.97
Not step, any unintended children	0.87	0.87	1.14	1.12	1.25	0.93	1.30	1.37
Respondent only, no shared children	0.94	0.94	1.22	1.21	1.34 **	1.00	1.39	1.48 *
Respondent only, only intended children	0.91	0.91	1.18	1.73	1.30	0.97	1.35	1.43
Respondent only, any unintended children	1.12	0.78	1.46	1.01	1.61 ***	0.84	1.67 *	1.23
Partner only, no shared children	--	--	1.30	1.29	1.43 ***	1.07	1.48 *	1.57 **
Partner only, only intended children	0.77	0.77	--	--	1.10	0.82	1.14	1.22
Partner only, any unintended children	0.81	0.90	1.05	1.15	1.16	0.95	1.20	1.40 *
Both, no shared children	0.70 **	0.94	0.91	1.21	--	--	1.04	1.48 *
Both, only intended children	0.67 *	0.64 **	0.88	0.82	0.96	0.68 *	--	--
Both, any unintended children	0.69	0.84	0.90	1.08	0.99	0.90	1.03	1.32

^*p≤.05

^**p≤.01

^***p≤.001

Note: All models include controls for race-ethnicity, age, high school graduate status, family structure at age 14, maternal education, respondent and partner prior marital status, engagement, cohabitation order, time period, and duration.

The results in Table 4 suggest there is even more nuance to the role of shared intended and unintended childbearing, as the association is not uniform across stepfamily configuration. A number of contrasts are shown here, as indicated in the labels across the top. More specifically, this table shows two sets of omitted categories for each stepfamily type (those with no shared children and those with only intended children), and the main focus is whether intended and unintended childbearing are associated with dissolution and marriage in the same manner across stepfamily configuration. In Contrasts 1A (not a stepfamily, no shared births) and 1B (not a stepfamily, only intended births), we see that, among non-stepfamily unions, any unintended births decrease the risk of marriage relative to those with no births (RRR = 0.74), but among those with a birth, unintended fertility does not affect the risk of marriage. However, intended fertility is protective against dissolution for non-stepfamily cohabitations – compared to those with only intended births, both childless couples and those with any unintended births are more likely to break up (RRR = 1.47 and RRR = 1.42, respectively). As we move to the various stepfamily configurations, though, shared intended and unintended childbearing is largely unrelated to dissolution and marriage risks. In stepfamily cohabitations in which only the respondent has children (Contrasts 2A and 2B) and those in which only the respondent’s partner has children (Contrasts 3A and 3B), shared intended and unintended fertility are not significant predictors of outcomes. Only for the most complicated cohabiting stepfamilies (both partners have children from prior relationships) does shared childbearing matter. Compared to cohabiting stepfamilies in which both partners have prior children but have no shared children, having only intended births decreases the risk of dissolution (RRR = 0.68). For this group, shared childbearing, regardless of intendedness, is unrelated to the risk of marriage, nor are there differences in the risk of dissolution by intendedness among those with at least one shared birth.

Discussion

Cohabitation is now the modal pathway into marriage (Guzzo, 2014), and most stepfamilies begin as cohabitations (Guzzo, forthcoming). These changes, along with more childbearing in cohabiting unions (Manning, Brown, & Stykes, 2015), mean that many cohabitations include stepchildren, shared children, or both. Research on cohabitation has not yet fully examined outcomes across cohabiting unions with different types of children nor has research on stepfamilies fully considered cohabitation. I address these gaps by looking at the stability of cohabitations, comparing unions in which neither partner had children to those in which at least one partner already had children. I go beyond this basic analysis, though, to examine whether it mattered (in terms of marriage and dissolution risks) which partner had children and whether shared children played a similar role across stepfamily and non-stepfamily cohabitations.

Specifically, the current project tested three hypotheses. The most general, Hypothesis 1, posited that when either partner had children from a past relationship, it reduced the likelihood of marriage and increased the likelihood of dissolution. This was partially supported – stepfamily unions had a lower risk of marriage relative to remaining cohabiting compared to non-stepfamily unions. Dissolution, however, did not vary by stepfamily status. Hypothesis 2 suggested a gradient across stepfamily configurations: the most complex stepfamilies would be the most negatively associated with marriage and the most positively associated with dissolution, followed by those in which only the female partner had children, and then by those in which only the male partner had children. This, too, was only partially supported. Relative to non-stepfamily cohabitations, stepfamily cohabitations in which the male partner had children were less likely to marry than remain cohabiting; this is true regardless of whether the female respondent had children, but the likelihood was lower when she was a parent as well. Moving to other contrasts, though, showed that marriage and dissolution risks were not significantly different when only one partner had children, but the most complex stepfamilies continued to exhibit lower chances of marriage. Returning to the family systems perspective, then, it seems that the boundary ambiguity present in cohabiting stepfamilies in which both partners have children – and thus likely have variation not just in relationships within the household but outside the household as well, with members perhaps literally moving in out and of the physical family system – presents a particular challenge for marital transitions. The lack of differences in dissolution risks, though, suggests whatever issues such stepfamilies face are not any more disruptive than those of non-stepfamily cohabitations. The latter point is especially important if not all cohabiting couples –particularly those in which both members have past family experiences already – want to marry.

Hypothesis 3 dealt with shared childbearing, positing that intended childbearing decreased the risk of marriage or dissolution relative to remaining cohabiting while unintended childbearing increased the risk of dissolution and decreased the risk of marriage. Prior literature provided conflicting evidence as to whether this would vary by stepfamily status and configuration. Results suggest that a shared child affected the chances of marrying or breaking up but largely only for non-stepfamily cohabitations. For cohabiting couples in which neither partner had prior children, intended fertility was protective against dissolution compared to those with no births and those with any unintended births; unintended fertility increased the risk of dissolution among non-stepfamily unions relative to those with no shared birth. But among stepfamily cohabitations, shared childbearing, regardless of intendedness, was generally unrelated to dissolution and marriage relative to remaining cohabiting, with one exception (among the more complex stepfamilies, having only intended births reduced dissolution risks compared to those with no shared births). The lack of an association between a new child and outcomes suggests that for many cohabiting stepfamilies, a shared child (intended or not) is not indicative of a highly committed couple choosing to stay cohabiting rather than marry. On a more positive note, a shared child does not seem to cause additional problems for stepfamilies, either, or at least not problems severe enough to lead to dissolution or inhibit marriage.

What do these findings say about the institutionalization of cohabitation and stepfamilies? The lower risk of marriage – but not dissolution – relative to staying cohabiting among some stepfamily cohabitations suggests that such stepfamilies accept cohabitation as a union in which to raise children, even if children are not biologically related to both partners. Childless cohabitors, conversely, may view their union as a more transitory relationship – either moving forward to marriage or ending fairly quickly (Rose-Greenland & Smock, 2013). Perhaps cohabitors who have children from a past relationship are more wary or less desirable of marriage, since at least one partner has had a failed past relationship; that the lower risk of marriage is strongest among cohabiting stepfamilies in which both partners already had children provide some support for the notion that some cohabitors may not desire marriage. Alternatively, despite the view from a family systems perspective that clearly defined ties are beneficial, cohabitors with children may prefer to maintain less formal ties and connections between new partners and their children. Parents might be concerned that formal ties will reduce involvement from the other biological parent, or new partners may worry about the legal and financial obligations for nonbiological children that marriage entails. Another explanation, given that the lower marriage risks were concentrated among women partnered with a father (whose children are likely nonresident), is that something about the stepmother role influences cohabiting women’s marital transitions. Perhaps the ambiguity of the stepmother role stemming from fewer chances to interact and build ties with stepchildren leads to problems that discourage marriage but are not substantial enough to increase the risk of dissolution. In any case, if unpartnered parents are increasingly forming – and staying in – cohabiting unions, cohabitation may be becoming institutionalized among stepfamilies but stepfamilies themselves are likely becoming less institutionalized because roles, relationships, and obligations are less clear in such unions.

Though it has been argued that childbearing in cohabiting unions is suggestive of the institutionalization of cohabitation (Kiernan, 2001), the results here suggests this is only true if births are intended and, further, seems only to be especially the case for non-stepfamily cohabitations (and perhaps stepfamily cohabitations in which both partners have children already). From a family systems perspective, the more clearly defined ties and obligations when a biological child is introduced in a stepfamily in which only one partner had a prior child may complicate relationships – one partner is both a parent and a stepparent but the other partner is only a biological parent. When both partners are already parents, though, a new shared child does not introduce any new roles and may serve as a common – and deliberate, at least for intended births – connection for each partner’s prior children (i.e., both sets of children will have a new step-sibling), and this connection may reduce instability.

Limitations

A major limitation is the exclusion of men due to limited birth intendedness information, potentially missing information about some children, and concerns over the comparability of stepfamily categories (i.e., a respondent who had children from a prior relationship would mean something different for female and male respondents). Even for women, I cannot determine coresidence explicitly since the NSFG lacks child coresidence histories. Another major limitation is that the NSFG lacks information about the partner’s children (number, age, coresidence, involvement); changes in custody arrangements, visitation, and child support mean nonresidential parents’ obligations and involvement are higher in more recent years (Cancian, Meyer, Brown, & Cook, 2014; Meyer, Cancian, & Chen, 2015). Another limitation is that this research may have missed or mis-identified some pre-union births as with a different partner (or with this partner) as it relied on dates rather than direct partner identification for births (which are not available in the NSFG for female respondents). Finally, the data have an upper age limit of 44; union dissolution and repartnering increases with age, so many cohabiting stepfamilies are not included in this sample (though those over 44 would have be unlikely to have shared births). And although this is not exactly a limitation, the findings cannot speak to how cohabiting stepfamilies compare to directly married stepfamilies in terms of stability.

Conclusion

As union formation and childbearing behaviors have changed, today’s unions are now more complex. The majority of adults cohabit at some point (Manning & Stykes, 2015), and many cohabiting unions involve children with the current partner as well as from past relationships, especially among those who have cohabited more than once. These shifts in the complexity of cohabiting unions mean that cohabitation does not fill a single function or role in the relationship spectrum (Rose-Greenland & Smock, 2013); for never-married childless adults, for instance, cohabitation may be part of the marriage process or simply a form of coresidential dating in which the union ends quickly. Childbearing within cohabitation for such couples does seem indicative of the institutionalization of cohabitation (Kiernan, 2001). For those with children from past relationships, there are few differences in stability and transitions compared to those with no children. This suggests that these most cohabitors with prior children are, like childless cohabitors, trying to figure out their new relationship and where it is heading. Whether this exploration process has consequences for their children remains to be seen, and so the new complexities of today’s unions need to be better understood. Additional work focusing on the role of involvement and coresidence with stepchildren as well as interactions across households (with biological and stepchildren as well as with former partners) would provide insight into the processes and functioning of both stepfamilies and cohabitations. More work is also needed to study union formation, stability, and functioning from men’s perspectives.