Shifts in Higher-Order Unions and Stepfamilies among Currently Cohabiting and Married Women of Childbearing Age

Abstract

Shifts in union formation and childbearing have undoubtedly altered the prevalence and structure of higher-order unions and stepfamilies, but no study has examined trends over time. Comparing the 1988 and 2011-2013 cycles of the National Survey of Family Growth (NSFG), I produce estimates of repartnering and stepfamily formation among currently partnered women aged 15-44. The percentage of intact unions that are remarriages stayed stable (around 27-28%), but a growing proportion of currently married and cohabiting women had another cohabiting partner in the past. The percentage of intact unions that are stepfamilies increased from 24% to 31%, with an increase in cohabiting stepfamilies from 19% to 39% of all stepfamilies. Further, while the majority of remarriages are stepfamilies, the majority of women's stepfamilies are no longer remarriages due to union formation among never-married parents. Cohabiting (but not marital) stepfamilies also exhibited changes in which partner had children and in shared childbearing.

In the contemporary United States, shifting patterns of union stability and family formation have resulted in serial partnerships and growing numbers of stepfamilies formed through both marriage and cohabitation (Stewart 2007; Sweeney, 2010). Union dissolution is fairly common, and individuals usually go on to repartner (Cherlin, 2009); if the dissolved unions involved children, then the new union is a stepfamily. Stepfamilies are also formed when a parent whose child was born outside of a coresidential union cohabits with or marries someone other than their child's biological parent. Although scholars discuss the rising levels of higher-order unions and stepfamilies, our ability to accurately track such trends over time has been hampered by data issues (Stewart, 2007; Sweeney, 2010) due to limited resources which would allow researchers to incorporate cohabitation experiences and/or to measure stepfamilies.

In this report, I use the 1988 and 2011-13 cycles of the National Survey of Family Growth (NSFG) to examine the proportion of current cohabitations and marriages that are higher-order unions (e.g., remarriages, first marriages among those with prior cohabitations, and second or higher cohabitations) and stepfamilies among women aged 15-44. While higher-order unions and stepfamilies occur across the life course (Brown, Bulanda, & Lee, 2012), those formed among women of childbearing age are particularly relevant for children's well-being and future fertility (Thomson & McLanahan, 2012; Thomson, Winkler-Dworak, Spielauer, & Prskawetz, 2012). Additionally, adults' well-being and relationship quality are affected by their union status, type, and past experiences (Carr & Springer, 2010; Falke & Larson, 2007; Halford, Nicholson, & Sanders, 2007; Shapiro & Stewart, 2011).

Defining and measuring higher-order unions and stepfamilies

Repartnering – used here to mean the formation of any higher-order coresidential union – can be a complex process to measure and analyze. Remarriage has a fairly straightforward definition: one or both spouses have been previously married (Stewart, 2007). Repartnering more generally also includes higher-order cohabitations (i.e., serial cohabitation), first marriages among those with prior cohabiting partner(s), or post-marital cohabitations. Identifying remarriages – and other types of higher-order unions – in data can be difficult, as they require union histories from both partners. Further, due to budget cuts in the 1990s, the National Center for Health Statistics discontinued the detailed marriage and divorce registries that produced national estimates of remarriage. After the 1990s, researchers relied upon survey data with cohabitation and marriage histories, such as the Survey of Income and Program Participation (SIPP) and the National Survey of Family Growth (NSFG), to examine higher-order unions (e.g., Kreider 2005; McNamee & Raley, 2011). In 2008, the American Community Survey (ACS) also added a series of marital history questions (but no cohabitation history questions). In sum, shifts in data sources have made it difficult to track long-term trends in repartnering, and the lack of such data has forced scholars to rely on outdated or conservative estimates that ignore cohabitation history or partner's union history (Sweeney, 2010).

A related concern is the paucity of stepfamily data. As often noted in stepfamily research (e.g., van Eeden-Moorefield & Pasley, 2013; Stewart, 2007; Teachman & Tedrow, 2008), stepfamilies are usually defined by residential patterns: a stepfamily is a cohabiting- or married-couple household that contains at least one child from one of the partner's earlier relationships. This is problematic for two reasons. First, some data sources, such as the ACS and the Census, rely upon a single person in a household (referred to as the ‘householder’) to list all persons who live in the household and detail how they are related to the householder. This approach does not capture stepfamilies in which the householder has resident biological children and a partner who is not the other biological parent. As a result, estimates about the prevalence of stepfamilies seem fairly low, with little change over time. For instance, using the 1990 and 2000 Census, Teachman and Tedrow (2008) found that only 5% of married-couple households in 1990 and 4.9% in 2000 included a stepchild; the corresponding estimates for cohabiting-couple households were 2.9% and 4.8%, respectively. Not all datasets collect such limited household relationship information – some datasets (such as the redesigned SIPP and the Current Population Survey) use a household matrix to identify relationships to resident children, enabling researchers to examine stepfamilies from the perspective of children (i.e., what proportion of children live in a stepfamily). Brown, Manning, and Stykes (2015), for example, found that 10% of all children were living with a stepparent in 2008. However, one cannot identify other characteristics about stepfamilies, such as whether this is a remarriage or higher-order cohabitation.

Secondly, the household approach to identifying stepfamilies ignores nonresident children. Partnered mothers and fathers who do not live with their children– or childless adults who are partnered with a parent who does not live with their child – are not considered stepfamilies using a household-based definition (Stewart, 2007), yet these new unions are certainly still stepfamilies. It also underestimates the complexity of stepfamilies; more than half of coresidential stepparents also have children living in other households (Stewart, Manning, & Smock, 2003). Because today's families – especially stepfamilies – increasingly exist across households, a more inclusive definition would provide a more accurate, and no doubt higher, estimate of the prevalence of stepfamilies.

Trends in family behaviors related to repartnering and stepfamily formation

In addition to providing updated estimates with a more expansive definition, the current project addresses the need to examine the characteristics of today's higher-order unions and stepfamilies in light of changes in remarriage, cohabitation, and nonmarital fertility. Figure 1 displays some of the pathways into higher-order unions and stepfamilies, with higher-order unions and stepfamilies indicated by shading. The traditional pathway into a higher-order union and stepfamily is seen in Panel A. Here, we have a married person who divorces and remarries, and this new union is both a remarriage and a stepfamily (if there were no children in the first marriage, of course, it would be remarriage but not a stepfamily). This traditional pathway, however, likely occurs among fewer and fewer adults over time as union and childbearing shifts have altered family formation patterns. Declines in marriage, and the growth of nonmarital childbearing, have been concentrated among minorities and the less educated, so those following a more traditional pathway are likely to be disproportionately white and/or well-educated.

Figure 1. Sample Pathways into Higher-Order Unions and Stepfamilies.

Of the changes in family behaviors affecting repartnering and stepfamily formation, perhaps the biggest has been the rise of cohabitation and concomitant shift in remarriage. The majority of adults have cohabited at least once, and serial cohabitation has risen (Lichter, Turner, & Sassler, 2010; Manning & Stykes, 2015). Remarriage rates have fallen (Payne, 2015), as divorced men and women are increasingly turning to cohabitation. (Xu, Hudspeth, & Bartkowski, 2006). Panel B shows an example of how cohabitation fits into repartnering and stepfamily formation. A married individual divorces but does not immediately remarry; instead, he or she forms a new cohabiting union – this is a higher-order union and a stepfamily, but it is not a remarriage (though he or she may later transition from cohabitation to a remarriage, as show in the last section of this pathway). Using data from the mid-1980s, Bumpass and colleagues (1995) noted cohabitation was already changing the way stepfamilies were formed, with nearly two-thirds of women and children entering a stepfamily via cohabitation rather than directly marrying. Further, they found that including cohabitation in the definition of a stepfamily widened race-ethnic differences, as blacks are less likely to marry or remarry. Over the past 25 years, cohabitation has continued to increase, particularly for whites and the less-educated, but it has become the modal first union for all adults (Manning & Stykes, 2015). Even childless adults who have never lived with a partner may form unions that are higher-order unions and/or stepfamilies if their partners have had prior cohabitations, marriages, or children. Panel C shows an example of a never-married, never-cohabiting childless adult forming a cohabitation with someone who has children from a past marriage. In this scenario, if we focused only on the childless respondent with no prior unions, we would overlook the fact that this is a higher-order union, and if the partner's children did not live in the household, we might miss the fact that this is also a stepfamily. Thus, taking into consideration the rise of cohabitation, we would expect more higher-order unions overall and that an increasing proportion of higher-order unions are cohabitations. Further, it seems likely that cohabiting stepfamilies have become an increasing proportion of all stepfamilies, with even more people entering stepfamilies through cohabitation that later transition to marriage.

Nonmarital fertility, which rose from about a quarter of all births in the late 1980s to 41% of births by 2013 (Curtin, Ventura, & Martinez, 2014; Ventura & Bachrach, 2000), also affects stepfamily formation, largely by increasing the pool of never-married parents in the “market” for a union. Panel D of Figure 1 is an example of the role of nonmarital childbearing in stepfamily formation – and demonstrates that not all stepfamilies are necessarily higher-order unions. In Panel D, we have a never-married, never-cohabited adult who has a child outside of a coresidential union, and when he or she first lives with a partner, this produces a first union that is a stepfamily. Nonmarital fertility is highest among race-ethnic minorities and the less-educated, but there are interesting behaviors and variations underlying overall trends and patterns, with cohabitation playing a role. In the early 1990s, only 40% of nonmarital births were to cohabiting women (Kennedy & Bumpass, 2008), but by 2006-2010, nearly 60% were to cohabitors (Curtin, Ventura, & Martinez, 2014). The increase in cohabiting births is largely concentrated among Hispanics and whites (Lichter, 2012), and the proportion of nonmarital births to cohabitors increases with education (Manning, Brown, & Stykes, 2015). Cohabiting unions with children, however, are highly unstable (McLanahan & Beck, 2010), so these parents (most of whom have never married) go on to form new unions and thus stepfamilies. Cohabiting unions are often the initial union for never-married parents (Lichter, Sassler, & Turner, 2014); even if the cohabitation transitions to marriage, the subsequent marital stepfamily will not be a remarriage unless their partner was previously married. This pathway is exhibited in Panel E in which a never-married cohabiting parent dissolves a cohabitation and cohabits with a new, never-married partner; this new cohabitation is a stepfamily and a higher-order union, and if the new cohabitation transitions to marriage, there is a marital stepfamily but not a remarriage. In sum, changes in remarriage, cohabitation, and nonmarital childbearing likely mean that, today, most remarriages are stepfamilies but that most stepfamilies are not remarriages.

Changes in childbearing, in turn, have also made the child composition (i.e., “his, hers, and their” children) of stepfamilies more complicated. Nonmarital fertility tends to occur primarily among younger individuals, who are more likely to repartner than their older counterparts. As such, this could mean more women are bringing their own children into stepfamilies (as opposed, or at least in addition, to childless young women partnering with older men who are fathers), and overall increases in nonmarital fertility and high levels of union instability could translate into a higher proportion of stepfamilies in which both partners have children. The fertility histories of today's stepfamilies also affect whether there are shared children; stepfamilies in which both partners have children already are less likely to have a joint child (Stewart, 2002). However, stepfamilies often go on to have a shared child together, and the growing acceptability of cohabitation as a site for childbearing may translate into more cohabiting stepfamilies having a shared child.

In the current study I examine shifts in the proportion of intact unions that are higher-order unions and/or stepfamilies between two points in time, roughly 25 years apart, among nationally representative samples of women of childbearing age. To the extent that shifts in remarriage, cohabitation, and nonmarital fertility have occurred differentially across race-ethnic and education groups, changes in the prevalence and structure of stepfamilies may contribute to widening gaps in children's living experiences and well-being across socioeconomic status (McLanahan, 2004) as well as growing health gaps across adults. As such, I also explore key indicators of higher-order unions and stepfamilies across race-ethnic and educational groups.

Data

The analyses use the 1988 and the 2011-13 cycles of the National Survey of Family Growth (NSFG), a repeated cross-sectional, nationally representative survey of women aged 15-44, to analyze a nearly 25-year timespan. The NSFG surveys have collected marital and birth histories in each cycle but did not collect cohabitation histories until 1988. To produce prevalence estimates, the analyses focus on unions intact at the time of the interview. The data to produce lifetime experience is unavailable due to limited information about past partners in the 1988 NSFG, though some information is available in the 2011-13 cycle; overall experiences for the 2011-13 cohort of women 15-44 are discussed very briefly in the results section.

Of the 8,450 women in the 1988 NSFG, 4,031 women were married and 416 were cohabiting at the time of the interview (total N = 4,447). Of the 5,601 women in the 2011-2013, 1,711 women were married and 763 were cohabiting at the time of the interview (total N = 2,474). The NSFG is the only data source that has maintained consistent sampling techniques and questions, making it uniquely suited to track trends in repartnering and stepfamilies. Various cycles of the NSFG have been used widely to compare trends over time across a wide range of fertility and union behaviors, such as unintended fertility (Mosher, Jones, & Abma, 2012), cohabitation (Reinhold, 2010), and age at first union (Manning, Brown, & Payne, 2014).

Higher-order unions were fairly straightforward to identify, with virtually identical questions across surveys. For the women themselves, whether their union at the time of the interview was a remarriage is taken from a direct question about the number of marriages; any currently married woman who reported two or more marriages in total is by definition in a remarriage. These questions also permitted identifying postmarital cohabitations among those cohabiting at the time of the interview. Additionally, women were asked how many cohabiting partners they had other than their current partner or husband at the time of the interview. The NSFG also collected information on whether women's partner at the time of the interview, as well as former husbands and cohabiting partners, had ever been married; information about the partner's past cohabitation experiences was not collected. Combining this information, women married at the time of the interview were defined as being in a remarriage if they or their husband had been married previously, and women cohabiting at the time of the interview were defined as being in a higher-order cohabitation if either they or their partner had ever been married or they themselves had previously cohabited with a different partner.

Classifying intact unions as stepfamilies was a two-step process. The first step was identifying whether married or cohabiting women had children from a prior union. Dates of birth are collected separately from union dates in the NSFG, and women were not directly asked whether each birth was fathered by a particular cohabiting or marital partner. Instead, dates of births were compared to the start date of the intact union. In the 1988 NSFG, women in a union at the time of the interview were directly asked when they started living with their partner; this identified the start date of cohabitations for women in a cohabiting union, the start date of cohabitation for married women who cohabited with their future spouse prior to marriage, and the start date of marriage for women who entered marriage directly. In the 2011-2013 NSFG, marital information for women who were married at the time of the interview are part of the marriage history, which required first matching the current union to a particular marriage to identify the start date; a separate series of questions within each marriage ‘loop’ identified the date of premarital cohabitation for those who lived with their partner prior to marriage. For women cohabiting at the interview in the 2011-13 NSFG, there was a direct question about the start date. Once a start date was established for marriages and cohabitations intact at the time of the interview, births from prior relationships (i.e., stepchildren for the respondent's partner) are defined as those occurring more than 6 months prior to the start of the current union; alternate specifications (3 months vs. 9 months) yielded virtually identical estimates (not shown).

The second step was identifying whether their current partner had any children from a prior union. In 1988, women were asked whether their partner had any children under the age of 18. In 2011-13, information for husbands was again part of the marital histories, with married women asked whether their spouse had children from prior relationships and how many of these children were under 18; cohabiting women were asked the same questions for their partner. For comparability, I only consider a partner to have children if the children are under 18 in the 2011-13 NSFG. There was missing information on the partner's children for five women partnered at the time of the interview in the 2011-2013 NSFG, reducing the sample size to N = 2,469.

Together, the presence of respondent's children from a prior relationship and the partner's children from a prior relationship were combined to create two variables – (1) a dichotomous variable indicating whether the union at the time of the interview was a stepfamily (the respondent and/or or her partner already had child); and (2) a four-category variable indicating stepfamily configuration: no stepchildren; only the respondent had children from a prior relationship, only the partner had children from a prior relationship, and both had children from a prior relationship. This definition is not dependent on the children's coresidence in the household with the respondent. However, women who reported that their partner had children were asked if their partner's children lived in their (the respondent's) household; this information is used to produce an estimate similar to those of household-based reports of stepfamilies in which children in the household are indexed to one person and to various household members.

Additional analyses consider shared children and variation by key sociodemographic variables. The presence of shared children is defined as those who were pregnant at the time of the interview or reported any children born within six months prior to union start through the time of the survey (alternate specifications yielded virtually identical results, not shown). The availability of comparably measured socioeconomic and demographic variables in 1988 and 2011-13 was somewhat limited given changes in question wording (i.e., changes in family background questions, employment, etc.) and small sample sizes when disaggregating by groups, so the analyses focus on race-ethnicity (non-Hispanic white, non-Hispanic black, and Hispanic) and years of education (less than 12 years, 12 years, 1-3 years post-secondary, and 4 or more years of post-secondary education). The socioeconomic characteristics of the full NSFG samples and the analytical samples are presented by NSFG cycle in Appendix A.

All analyses are weighted. Both the 1988 and the 2011-13 cycles employed a stratified cluster design, but the 1988 NSFG did not provide stratum or cluster variables in the public data file. As such, analyses using the 1988 NSFG use only the final post-stratified weight to produce population estimates, whereas analyses using the 2011-13 NSFG account for the stratified cluster design with Stata's svy commands. This leads to an unfortunate problem: for the 1988 NSFG, the standard errors and test statistics are computed under an assumption that the design was a simple random sample, thus producing standard errors that are too small. As such, the estimates provided here do not include standard errors nor tests of statistical difference across surveys.

Analytical strategy

This paper documents trends in higher-order unions and stepfamilies between two points roughly twenty-five years apart using basic bivariate statistics, providing updated data for family scholars and demographers. I first briefly document changes between 1988 and 2011-13 in union status at the time of the interview for all women aged 15-44. Then, I focus on the prevalence and composition of higher-order unions and stepfamilies among women in intact unions, briefly discussing race-ethnic and education differences. I finish by examining changes in stepfamily configuration in terms of which partner has children and the presence of shared children.

Results

Figure 2 displays the weighted percentages of women aged 15-44 by union status at interview and previous union status for 1988 and 2011-13. At both time points, the percentage of women in a union is fairly similar – in 1988, 55.5% of women were in a coresidential union compared to 53.1% of women in 2011-13. These general similarities, however, mask differences in the type of unions. In 1988, 91% of all women in an intact union were married. By 2011-13, the percentage of partnered women in a marriage declined by nearly twenty percentage points, to 72%, reflecting the tripling of women cohabiting at the time of the interview from 5.2% to 15.0%. There are similar changes in the composition of those not in a union at the time of survey. The percentage of women who have never been in a coresidential union declined slightly between 1988 and 2011-13, as did the percentage previously married – largely due to the rise in those who had previously cohabited, which nearly doubled from 4.9% to 9.1%. Of those who were not currently in a union at the time of survey, the percentage who had two or more prior unions rose from 9.6% in 1988 to 15.6% in 2011-13 (not shown).

Figure 2. Union Status at Interview among Women 15-44 (weighted).

Table 1 displays selected characteristics about higher-order unions, remarriages, and stepfamilies for the unions of women who were partnered at the time of the survey. Panel A documents the prevalence of higher-order unions and stepfamilies. The percentage of the current marriages among women aged 15-44 that are remarriages for one or both partners has changed little over the past quarter century, at just over a quarter of marriages. However, there has been a large increase in the proportion of current cohabitations that involve previously married individuals. In 1988, less than a tenth of intact cohabitations among women aged 15-44 involved a previously married individual (either the female respondent or her partner), but by 2011-13, a fifth of intact cohabitations did. Additionally, women have increasingly lived with someone other than their partner at the time of interview among both the married (increasing from 8.6% to 21.0%) and the cohabiting (28.8% to 38.2%). Thus, the prior union histories of partnered women (and the union histories of their partners) became more complex over time.

Table 1. Characteristics of intact unions among women aged 15-44 in the 1988 and 2011-13 cycles of the National Survey of Family Growth, weighted.

	1988	2011-13
PANEL A: PREVALENCE OF HIGHER-ORDER UNIONS AND STEPFAMILIES
Higher order unions
Percentage of intact marriages that are remarriages for one or both partners	28.3%	26.7%
Percentage of intact cohabitations involving previously married individual(s)	8.8%	20.4%
Percentage of intact marriages in which respondent had cohabited w/different partner	8.6%	21.0%
Percentage of intact cohabitations in which respondent had cohabited w/different partner	28.8%	38.2%
Stepfamilies
Percentage of intact unions that are stepfamilies	23.6%	31.3%
Cohabitations	47.6%	43.3%
Marriages	21.2%	26.7%
Percentage of intact unions in which respondent has children that are not biologically related to her partner OR has a resident stepchild	17.8%	24.8%
Percentage of intact unions in which the resident has a resident stepchild	1.6%	1.8%
Percentage of stepmothers with a resident stepchild	13.8%	13.2%
PANEL B: COMPOSITIONAL SHIFTS IN STEPFAMILIES
Percentages of intact stepfamilies by union type
Cohabitations	18.8%	38.9%
Marriages	81.2%	61.1%
Prior marital experiences and stepfamilies
Percentage of intact remarriages that are stepfamilies	60.6%	62.9%
Percentage of 2nd or higher intact cohabitations (for respondent) that are stepfamilies	60.8%	66.0%
Percentage of all intact stepfamilies that are remarriages	65.8%	37.5%
Percentage of intact marital stepfamilies that are remarriages	81.0%	61.4%
Percentage of intact cohabiting stepfamilies involving previously married individual(s)	15.6%	37.6%
Percentage of intact stepfamilies with two never-married individuals	15.9%	24.3%
Cohabitation and stepfamilies
Percentage of intact stepfamilies cohabiting at interview or begun by cohabitation	64.9%	84.4%
Percentage of intact marital stepfamilies begun by cohabitation	56.8%	74.5%
N (unions at time of interview)	4,447	2,469

The growing complexity of past union behaviors (along with changes in fertility) among those in a union at the interview translates into changes in stepfamily prevalence and composition. In 1988, just under a fourth of intact unions among women aged 15-44 were those in which the woman, her partner, or both had children from a prior relationship (i.e., a stepfamily). Twenty-five years later, the proportion that were stepfamilies rose to nearly a third of all intact unions among women of childbearing age (a roughly 33% increase). Further, among women not in a union at the time of the 2011-13 survey, 23.5% had been in at least one stepfamily union in the past (not shown; a comparable estimate for 1988 cannot be ascertained). The percentage of cohabitations that were stepfamilies decreased slightly, from 48% to 43%. Intact marriages, however, exhibited the opposite pattern, with 21% of intact marriages in 1988 involving those with children from prior relationships, rising to 27% in 2011-13.

The last few lines of Panel A demonstrate how alternative operationalizations would underestimate the prevalence of stepfamilies among women's current unions. First, using the type of information available from a household matrix approach to identify a stepfamily (i.e., a mother is living with a man who is not the biological father of at least one of her children or who reports that a partner's child lives in their household), fewer stepfamilies are identified. Here, 17.8% of intact unions are identified as stepfamilies in 1988 and 24.8% in 2011-13; these underestimate the prevalence by 25% and 21%, respectively, compared to a definition that includes a partner's nonresident children. Even if we expand the definition to include partner's children who split time between households (and so may or may not show up on household rosters and matrices), the estimate remains the same in 1988 and rises only to 25.6% (not shown) in 2011-13 (reflecting the changes in shared custody over the past few decades).

If we use the definition in which household relationships were indexed to the single person – in this case, the female respondent – stepfamilies become virtually non-existent among women. That is, when women are reporting only their relationships to individuals in the household, the lack of a biological connection between her biological children and her partner would be missed; only resident non-biological (or adopted) children would be the marker of a stepfamily. Of all women who were partnered at the interview, less than 2% report having a resident stepchild at either time point. Even when focusing on women who report that their partner has a biological child from a prior relationship, only about 13-14% of stepmothers report that their stepchild lives in their household. Including stepchildren for whom the respondent reports that her partner's child splits time between parental homes would increase the percentage by two percentage points in 1988 and by twelve percentage points in 2011-13 (not shown). Thus, without information relating all children in and out of the household to both partners in the union, we underestimate – sometimes drastically – the prevalence of stepfamilies among those in intact cohabitations and marriages among women of childbearing age.

In addition to changes in prevalence, composition has also shifted (Panel B), though the majority of stepfamilies are marital stepfamilies. In 1988, 81% of intact stepfamilies among women aged 15-44 were marriages (first or remarriages). This declined by twenty percentage points, to 61%, in 2011-13; that is, nearly 40% of women's intact stepfamilies were cohabiting stepfamilies in the most recent time period. The decline in remarriage and rise of cohabitation documented above, combined with rising rates of nonmarital childbearing, have changed the linkages between remarriage and stepfamilies. In general, the majority of higher-order unions (both remarriages and second or higher cohabitations) are stepfamilies. However, although the majority of remarriages are stepfamilies at both time points (about 60%), the majority of stepfamilies are no longer remarriages due to nonmarital childbearing among either the women themselves or their partners. Just twenty-five years ago, two-thirds of all stepfamilies among women of childbearing age were remarriages. By 2011-13, the percentage had been almost halved, as only 37.5% of intact stepfamilies among women were remarriages.

As more never-married parents enter into their first marriages with children from prior relationships, even today's marital stepfamilies are not necessarily remarriages, as the percentage of marital stepfamilies that was a second or later marriage for at least one partner declined from 81% to 61% over the twenty-five year period. Conversely, the proportion of cohabiting stepfamilies in which one or both partners was previously married more than doubled, rising from 15.6% in 1988 to 37.6% in 2011-13. Rising rates of nonmarital childbearing and cohabitation, in fact, have led to an increase in the proportion of stepfamilies in which neither partner has ever been married. In 1988, 16% of stepfamilies involved two never-married individuals, but this increased to 24% by 2011-13 (an increase of over 50%).

Table 2 shows selected stepfamily characteristics by race-ethnicity and years of education. Although changes over the past twenty-five year in repartnering and stepfamily formation were universal, the magnitude of change and absolute levels varied quite dramatically across subgroups. Panel A displays information by race-ethnicity, showing that both the largest absolute increase and the largest proportional increase in the percentage of intact unions that were stepfamilies occurred among Hispanics. Between 1988 and 2011-13, the percentage of stepfamilies among Hispanic women aged 15-44 who were partnered at the time of the interview rose fifteen percentage points from 27% in 1988 to 42% in 2011-13, a 55% increase. The highest levels of stepfamilies were found among partnered black women, at roughly half of intact unions at both time points, with only a small absolute increase of seven percentage points between 1988 and 2011-13. For white women, the percentage of stepfamilies among intact unions increased modestly (by about 28%) over the past twenty-five years, from 22% to 28%.

Table 2. Selected stepfamily characteristics by race-ethnicity and years of education among women aged 15-44 currently in a union, weighted.

PANEL A: Race-ethnicity	White		Black		Hispanic
	1988 (n=3073)	2011-13 (n=1340)	1988 (n=906)	2011-13 (n=297)	1988 (n=341)	2011-13 (n=685)
Percentage of intact unions that are stepfamilies	21.5%	27.8%	46.4%	53.3%	27.2%	42.2%
Percentage of intact stepfamilies by type
Cohabitations	17.1%	35.3%	19.3%	42.7%	28.8%	48.5%
Marriages	82.9%	64.7%	80.7%	57.3%	71.2%	51.5%
Percentage of all intact stepfamilies that are remarriages	74.9%	44.7%	32.7%	25.0%	50.9%	24.8%
PANEL B: Education	Less than 12 years		12 years		1-3 years college		4+ years college
	1988 (n=694)	2011-13 (n=434)	1988 (n=1713)	2011-13 (n=544)	1988 (n=1141)	2011-13 (n=745)	1988 (n=899)	2011-13 (n=746)
Percentage of intact unions that are stepfamilies	38.7%	48.0%	25.3%	41.7%	22.5%	33.0%	11.7%	17.0%
Percentage of intact stepfamilies by type
Cohabitations	25.3%	49.6%	17.7%	37.7%	19.1%	40.2%	7.7%	26.0%
Marriages	74.7%	50.4%	82.3%	62.3%	80.9%	59.8%	92.3%	74.0%
Percentage of all intact stepfamilies that are remarriages	57.5%	25.8%	67.2%	32.7%	65.2%	39.2%	80.2%	55.8%

Perhaps more interesting, though, are the differences in the shift away from marriage, particularly remarriage, as the site of stepfamilies involving women of childbearing age. By 2011-13, although the majority of stepfamilies were marriages overall, marital stepfamilies were far less common among Hispanics (52%) and only marginally higher among blacks (57%). For whites, just under two-thirds of stepfamilies were marital stepfamilies. Further, whites and especially Hispanics experienced dramatic declines in the proportion of all stepfamilies that were remarriages. For whites, this declined about 40%, from 75% to 45%; for Hispanics, the proportion was halved, falling to 25% by 2011-13. The proportion of stepfamilies involving remarriages was a quarter for blacks as well by 2011-13, down from a third in 1988.

Turning to changes by years of education (Panel B), there is a fairly linear pattern, with the least educated having the highest percentages living in a stepfamily and the lowest proportions living in a marital stepfamily and in a stepfamily formed via remarriage. Nearly half of those with less than 12 years of schooling who were in an intact union were in a stepfamily by 2011-13. Conversely, for those with 4 or more years of post-secondary education, the proportion was less than a fifth (though the proportional increase was fairly large). The proportion of stepfamilies that were cohabiting unions doubled for all groups except those with four or more years of college. For this group, it actually more than tripled from 1988 to 2011-13, yet the majority (74%) of stepfamilies among the highest-educated women were marital stepfamilies, and this group also reported the highest proportion of stepfamilies that involve remarriage.

Finally, it is worth noting changes in the configuration of stepfamilies and shifts in the presence of shared children. As seen in Figure 3, there were modest shifts in which partner had children among intact stepfamilies overall. In 1988, just over a fifth of women of childbearing age living in a stepfamily reported that both they and their partner had children from a prior union, with 29% reporting that only their partner had children from a prior union. Half of stepfamilies in 1988 were formed when the respondent partnered with a childless man but brought her own child(ren) into the union. Twenty-five years late, slightly more mothers were partnered with a childless man (rising from 49.7% to 54.3%) and fewer childless women were partnered with a father (falling from 28.6% to 22.8%). Changes among married stepfamilies were similar to those for stepfamilies overall, but cohabiting stepfamilies exhibited greater changes over time in configuration. In particular, the percentage of cohabiting stepfamilies in which both partners had children increased from 29% to 45%, along with a large increase in the percentage of mothers partnering with childless men from 35% in 1988 to 53% in 2011-13.

Figure 3. Stepfamily configuration (weighted).

Larger shifts in cohabiting than in marital stepfamilies are also seen for shared children (Figure 4). In 1988, 47% of women aged 15-44 in a stepfamily had a shared child or were pregnant at the time of the interview, with little change by 2011-13 (46%). Compared to cohabiting stepfamilies, a larger proportion of marital stepfamilies reported a shared child or pregnancy, at just over half at both time points (53%). However, cohabiting stepfamilies with shared children or a pregnancy increased by 64% between 1988 and 2011-13, rising from just over a fifth (21.2%) of intact cohabiting stepfamilies to just over a third (34.8%).

Figure 4. Proportion of Stepfamilies with Shared Children or Pregnant at Interview.

Discussion

Over the past few decades, the United States has witnessed remarkable changes in family behaviors. Marriage has become increasingly delayed, and the proportion of individuals who have never married has risen (Anderson & Payne, 2012; Wang & Parker, 2012). The growth of cohabitation, though, has largely offset the changes in the levels and timing of marriage (Manning, Brown, & Payne, 2014). Cohabitation is similarly linked to the growth of nonmarital childbearing; the proportion of nonmarital births to women not living in a union has stayed remarkably stable (Manning, Brown, & Stykes, 2015). Although divorce rates have stabilized among those under 35, they have risen for those over 35 (Kennedy & Ruggles, 2014). Further, the emergence of cohabitation also plays a role here – young adults' declining marital instability is almost certainly due to the substitution of cohabitation for marriage and the growing selectivity of early marriages. The increasing instability of cohabitation (Guzzo, 2014a) suggests that young adults' unions are not more stable than in the past.

These shifts, in turn, affect not only initial union formation but higher-order unions and family behaviors. In essence, unions have become more complicated over time as a growing proportion of those currently in a union have some prior family experiences. This paper provided new estimates of the prevalence of higher-order unions and stepfamilies among currently partnered women of childbearing age over the past twenty-five years, filling a major gap. As Sweeney (2010) noted in a review of remarriage and stepfamily research, the appropriate data to study and document prevalence and trends has become increasingly limited, and no study had yet used the few datasets capable of identifying higher-order unions and stepfamilies to document shifts over time. As such, we know little about the prevalence or form of today's higher order unions and stepfamilies even as researchers are increasingly trying to understand complex family processes (see, for instance, the July 2014 ANNALS volume on family complexity). Additionally, this report used a broad definition of stepfamilies to incorporate nonresidential children, producing much higher estimates than seen in other work (e.g., Brown, Manning, & Stykes, 2015; Teachman & Tedrow, 2008). Using definitions that do not incorporate a partner's nonresident children substantially reduced the apparent prevalence of stepfamilies, especially when using a householder-based method; household matrices seem to do a better job of identifying stepfamilies but still underestimate stepfamily prevalence. For instance, when producing a similar estimate as would be obtained from a household matrix with the NSFG data, between a fifth and a quarter of stepfamilies in this data were not identified.

The descriptive statistics here revealed several interesting shifts over time. There has been a modest growth in the proportion of intact unions among women aged 15-44 that are stepfamilies, rising to just under a third of unions. More interesting, though, is that cohabitation now plays a much bigger role in today's stepfamilies. Almost 40% of women in stepfamilies are cohabiting, and when we expand this to consider premarital cohabitation among marital stepfamilies, 84% of intact stepfamilies are cohabitations or entered their marital stepfamily through cohabitation. And while the majority of stepfamilies are marriages, fewer of these marital stepfamilies involve remarriage, as more never-married parents enter their first marriages with someone other than their child's biological parent. In total, just over a third of all stepfamilies today are remarriages.

Moreover, cohabiting stepfamilies show the largest increase in complexity in terms of child composition, with 45% of women in a cohabiting stepfamily reporting that both they and their partner have children from a prior union. And although just over half of marital stepfamilies have a shared child, cohabiting stepfamilies showed the biggest change over time, with a third now reporting a shared child. Shared children in a stepfamily represent a form of multipartnered fertility, which has become an area of growing interest for family demographers (for a review, see Guzzo (2014b)). That cohabitation plays such a large role in the shifts in higher-order unions and stepfamilies over the past twenty-five years suggests that the role cohabitation plays in the American family system continues to evolve past simple notions of cohabitation as an alternative to dating or a trial marriage (Rose-Greenland & Smock, 2013).

The changes discussed above have not been experienced equally across groups. Black and Hispanic women have considerably higher levels of nonmarital childbearing than white women, although both white and Hispanic women have more nonmarital births in cohabiting unions than black women (Manning, Brown, & Stykes, 2015). The proportion of births outside of marriage is inversely associated with education; more than half of births to women without additional education after high school are outside of marriage, but only 11% of births to college-educated women are nonmarital (Manning, Brown, & Stykes, 2015). College-educated women form their unions, on average, nearly six years later than those without a high school degree, and they are more often entering directly into marriage than their less educated peers as well (Manning, Brown, & Payne, 2015). If disadvantaged women's initial family experiences – their first unions and their first births – occur in much less stable circumstances than more advantaged women, it is not surprising that there are large race-ethnic and educational differences in stepfamily prevalence among women who are partnered at any given point in time. These differences in stepfamily experiences – and the form and prior history of these unions – have implications for the functioning and stability of women's unions.

Limitations

Though this report provided important new information on repartnering and stepfamily trends over time, there are several limitations to note. The biggest limitation of the analysis is the focus on unions that were intact at the time of the survey. Unfortunately, the 1988 NSFG collected only limited information about past unions. It did not collect any information about women's partner's children for prior unions (only collecting such information among women who were currently living with a partner), nor did it collect a complete cohabitation history (only identifying whether women had ever cohabited with someone other than with their current partner or current/past husbands). As such, it was not possible to produce estimates of whether women had ever lived in a stepfamily regardless of whether they were currently in a stepfamily at the time of survey for 1988. Put differently, these are not lifetime estimates of ever living in a stepfamily, which are certainly higher than those currently in a stepfamily. Another issue is that the cross-sectional nature of the analyses means that more stable stepfamilies are over-represented, such as marital stepfamilies or those in which only one partner had children. The age restriction of the NSFG to those aged 44 or less, along with the identification of only stepchildren under 18 in the 1988 NSFG (and in the 2011-13 NSFG, for comparability), means that many stepfamilies are missed – those with non-minor children as well as unions formed at older ages. The latter is particularly problematic – from a life course perspective, NSFG estimates are conservative given the rise of “gray divorce” and cohabitation among older adults (Brown & Lin, 2012; Brown, Lin, Hammersmith, & Wright, forthcoming). Another limitation, as discussed in the data section, is that I am unable to accurately produce standard errors for the 1988 cycle and thus unable to produce test statistics of differences between the sample estimates. Despite the consistency in question wording across cycles, differences in sampling procedures, question wording, and weighting procedures likely introduce some level of error.

Additionally, though one can identify whether partners had been previously married, partners' cohabitation history is unknown, so the overall proportion in a higher-order union (regardless of marital status) is an underestimate. Finally, although more recent cycles of the NSFG include men, documenting change over time required a focus on women only. Given evidence that, compared to mothers, fathers are more likely to repartner (and do so sooner) and more likely to partner with someone younger and childless (Livingston, 2014), trends and configuration may differ for men than women. However, it is worth noting that measures based on householder reports and those based on household matrices likely underestimate men's experiences of stepfamilies far less than for women, since men's partners are more likely to have resident children (if they are parents) than vice versa (Grall, 2013).

Conclusion

Today's unions have indeed become more complicated, as many partnered individuals have experiences – and children – from earlier relationships. These shifts could have major implications for family functioning. For instance, serial cohabitation may alter individuals' attitudes towards marriage and increase the risk of divorce (Lichter & Qian, 2008), and shifts away from remarriage towards cohabitation have implications for the social ties of stepparents and stepchildren (Manning, Smock, & Bergstrom-Lynch, 2009). Similarly, the rise in nonmarital childbearing has implications for child support and visitation when parents repartner – agreements are often informal for children born outside of marriage (Nepomnyaschy & Garfinkel, 2010), so new unions among unmarried parents may complicate more tenuous arrangements (Tach, Mincy, & Edin, 2010).

As more individuals have serial partnerships, their past experiences (and those of their partners) may influence how they relate to one another, their expectations for the future of their union, and their willingness to leave the union. Thus, more attention is needed to consider how the increasingly complex union histories of individuals and potential partners influence subsequent union formation and the quality and stability of current unions. Similarly, researchers need to think more critically about the varied pathways into a stepfamily. The growing diversity of stepfamily structures begs the question as to which types of stepfamilies are more or less stable and whether some stepfamily types (i.e., cohabiting vs. married, shared children vs. no shared children, etc.) are better or worse for couple, parent, and child well-being. The complexities documented here provide compelling evidence of the importance of collecting information on both partners, ideally for both cohabiting and marital prior unions as well as parenthood status. Moreover, given the linkages between union status and type and well-being (Carr & Springer, 2010; Falke & Larson, 2007; Shapiro & Stewart, 2011), additional work is needed to understand the mechanisms linking these increasingly less formal and more diverse unions to the health and well-being of adults. Finally, the existence of race-ethnic and educational disparities in the prevalence and composition of stepfamilies also suggests that social and economic disadvantage may be a double-edged sword, increasing the likelihood of having unstable unions and more complicated families in which family processes and social support may be weaker, contributing to what McLanahan (2004) called “diverging destinies.”

Appendix A. Race-ethnicity and education of women 15-44 in the 1988 and 2011-13 cycles of the National Survey of Family Growth, weighted

	All women		Analytical sample (intact unions)
	1988	2011-13	1988	2011-13
Race-ethnicity
Non-Hispanic white	73.5%	58.5%	79.5%	65.5%
Non-Hispanic black	12.8%	14.7%	7.6%	8.4%
Native-born Hispanic	5.7%	12.1%	4.4%	9.7%
Foreign-born Hispanic	3.9%	7.7%	4.6%	9.7%
Other	4.1%	7.1%	3.9%	6.7%
Years of education
Less than 12 years	23.3%	22.5%	15.0%	14.7%
12 years	33.4%	21.9%	38.2%	21.1%
1-3 years of college	25.4%	28.4%	25.1%	28.5%
4 or more years of college	17.9%	27.2%	21.7%	35.6%
N	8, 450	5,601	4,447	2,469