Introduction

Stepfamilies—cohabiting or married couples in which at least one partner has a child from a previous relationship, regardless of where children live—are now a common family type in the United States (Sweeney 2010). Many new stepfamilies go on to have at least one shared child (Stewart 2002; Thomson and Li 2002), even if the individuals in the union are at parities that are generally at low risk for further childbearing (Thomson et al. 2012). To explain this phenomenon, researchers have argued that shared children serve a unique purpose, drawing from the “value of children” perspective (Griffith et al. 1985; Holland and Thomson 2011). This perspective implies that having a birth with a new partner is a deliberate behavior and thus that births in new unions, regardless of stepfamily status, are likely to be intended births. The value of children perspective applies to all unions, but stepfamilies may be more motivated than their counterparts in other unions to have a shared child in order to publicly demonstrate commitment and create social capital (Stewart 2002). However, although stepfamily fertility has been well studied in Europe (e.g., Holland and Thomson 2011; Lappegård and Rønsen 2013; Vanassche et al. 2015), only two studies have explicitly examined U.S. stepfamily fertility (Li 2006; Stewart 2002)— a major limitation given differences in family and fertility behaviors across countries.

In addition to the lack of stepfamily fertility research in the United States, there is also another notable limitation to past work: are births in stepfamilies actually intended, as suggested by the value of children perspective? There is reason to think this is not necessarily the case in light of the emerging U.S. research on multipartnered fertility (MPF), defined as individuals who have children with more than one partner (Carlson and Furstenberg 2006; Guzzo 2014; Guzzo and Furstenberg 2007; Manlove et al. 2008). Despite the conceptual overlap, the MPF literature has largely developed with little attention to stepfamily research, perhaps because MPF is a characteristic of the individual (focusing on whether a person has children with two different partners), whereas stepfamily fertility is a characteristic of the union (focusing on whether a union includes children from either partner’s prior relationships, regardless of coresidence, and on whether the current partners also have a shared child). Although some MPF occurs entirely outside coresidential unions and thus cannot be formally classified as stepfamily fertility, the most common pattern is a nonmarital birth followed by marriage with a different partner and a subsequent birth (Dorius 2011)—that is, stepfamily fertility. Although not all stepfamilies have a shared child together (and thus no stepfamily fertility or MPF), viewing stepfamily fertility through the lens of MPF paints a very different story than that suggested by the value of children perspective. The MPF literature has largely characterized childbearing across partnerships as part of a pattern of disadvantage and instability (Carlson and Furstenberg 2006; Klerman 2007; Turney and Carlson 2011) in which unintended fertility is common (Guzzo 2014; Guzzo and Furstenberg 2007).

Drawing from the stepfamily, multipartnered, and unintended fertility literatures, this research examines fertility behaviors among cohabiting and married women by stepfamily status. Specifically, using birth and union data from the 2006–2010 National Survey of Family Growth (NSFG), I examine differences in the transition to a first shared birth, and the intendedness of that birth, between women in unions in which either partner has a child from a past relationship and women in unions in which both partners are childless at the start. I examine stepfamily status overall before considering different union types and stepfamily configurations.

This research fills an important gap in understanding both stepfamily fertility and MPF. Much of the research on MPF has stemmed from the Fragile Families and Child Wellbeing Study data, which represent a fairly disadvantaged sample disproportionately comprising individuals with nonmarital fertility. As such, it is unclear how generalizable those findings are to stepfamilies more broadly. Still, if stepfamily fertility is not necessarily the result of deliberate plans to have a shared child, then unintended fertility’s negative association with family stability and function (Boden et al. 2015; Guzzo and Hayford 2012, 2014) may exacerbate other issues that stepfamilies face (Cherlin 1978; Cherlin and Furstenberg 1994). Unintended fertility, regardless of union status, also has implications for maternal, child, and family behaviors and well-being, although the evidence is more correlational than causal (Gipson et al. 2008). For instance, research has suggested that unintended fertility increases the risk for poorer mental health outcomes among women (Herd et al. 2016; McCrory and McNally 2013) and lowers the rate of early prenatal care and breast-feeding (Kost and Lindberg 2015; Lindberg et al. 2015).

Childbearing Across Partnerships

As Thomson et al. (2014) noted, childbearing across partnerships (i.e., MPF) has become common in industrialized societies. In recent decades, the underlying mechanisms of MPF differ from the past in that MPF is now driven primarily by union instability rather than the death of a parent and partner. In the modern context, then, parents experience relationship dissolution (including nonresidential relationships as well as cohabitations and marriages), repartner, and have children with their new partner. High rates of instability among parents also increase the chances that a childless individual forms a union with someone who already has children, given that the pool of potential mates increasingly includes parents. New unions, in turn, represent new opportunities for childbearing for both childless couples and those in which one or both partners have a child already, and many stepfamily couples do indeed go on to have at least one shared child together (Holland and Thomson 2011; Stewart 2002).

What is interesting is that the general association between a given parity and subsequent childbearing seems to be weaker in stepfamilies than in other unions. Most individuals and couples desire to have children (Hagewen and Morgan 2005), with the ideal number in the United States hovering around 2.5 children since the 1980s (Newport and Wilke 2013). The majority of women have at least one child: the percentage of childless women aged 40–44 has not exceeded 20 % over the past four decades (Livingston 2015). Of those with children, the majority (approximately 80 %) have more two or more children, with the modal category being two children since the mid-1990s (Livingston 2015). After two children, however, parity progression declines sharply (Frejka and Sardon 2007), as does intended fertility. Even among married couples, who are most likely to intend births, more than one-third of third births are unintended (Mosher et al. 2012). In a stepfamily, though, childbearing by individual parity seems to differ because both intentions to have another child and actual fertility are higher than for women at the same parity in an intact family (Holland and Thomson 2011; Ivanova et al. 2014). Put differently, women with two children are more likely to have a third birth if they have a new partner than if they remain in the same partnership as their first two births.

The value of children perspective offers the primary explanation for fertility in unions with and without children from past relationships: shared children serve a unique purpose (Griffith et al. 1985) that is specific to the new relationship (Friedman et al. 1994). One purpose is that a shared child signals a couple’s commitment to each other and solidifies their status as a family unit (Thomson and Li 2002; Vikat et al. 1999). Stepfamilies may be particularly motivated by this signaling function if they feel greater pressure to symbolize their commitment. Stewart (2002) suggested that stepfamilies have to worker harder to create social capital for themselves than families without stepchildren, and the creation of social capital and the desire to express their commitment can be met by a shared birth (Astone et al. 1999; Coleman 1988). Stepfamilies might also have a higher risk of having a first birth, or at least having a first birth sooner, than couples in which both partners are childless because one partner (or both partners) has already entered a family-building stage. Preferences to have siblings close in age is likely a factor as well.

To date, the research on stepfamily fertility has relied almost entirely on European data, with two exceptions (Li 2006; Stewart 2002), yet family behaviors in the United States and Europe are quite different. The percentage of nonmarital births in the United States (41 % in 2011) is considerably higher than some European countries, such as Italy (28 %) and Greece (8 %), but considerably lower than others, such as Norway (55 %) and Iceland (67 %) (Livingston and Brown 2014). The union context of these births also differs; generally, the proportion of nonmarital births to cohabiting women is higher in Europe than the United States (Thomson et al. 2014). Moreover, the cohabiting unions in which children are born are much more stable in Europe than in the United States (Heuveline and Timberlake 2004). The limited research on stepfamily behaviors in the United States, where union instability is particularly high (Cherlin 2009; Kennedy and Ruggles 2014; Lundberg et al. 2016), has used data from the early 1990s and thus is unable to speak to the current family environment, with substantially higher levels of nonmarital fertility and cohabitation. Further, the two studies that have examined stepfamily fertility in the United States have produced mixed findings: Stewart (2002) found that stepchildren have a weak negative effect on shared childbearing relative to those with no stepchildren, whereas Li (2006) found no differences in women’s first births between marriages in which at least one partner already has children and marriages involving two childless individuals. Despite these mixed findings in the United States, the theoretical arguments about the value of children lead to the following hypothesis:

  • Hypothesis 1: Women in unions in which either partner already has a child (i.e., in a stepfamily) are more likely have a shared first birth than women in unions in which neither partner has children (i.e., in a nonstepfamily union).

Fertility Intendedness

The value of children perspective not only provides a compelling reason why couples have a child together, regardless of stepfamily status, but it also implies that births—especially in stepfamilies—are the result of deliberate behavior (Thomson 2003). This notion, however, has not been empirically tested. Instead, most work on fertility intentions in stepfamilies has looked at prospective intentions for future childbearing (e.g., Hohmann-Marriott 2015; Stewart 2002; Thomson 2003). In general, it seems that couples with no shared births (i.e., have a couple-level parity of 0) tend to both plan and have a first birth regardless of their individual parities, although second births in a particular union do seem to be more strongly and negatively affected by prior fertility (Stewart 2002; Thomson and Li 2002; Vikat et al. 2004). These studies suggest that stepfamily fertility (for at least the first shared birth) is both common and intended, although it is worth noting two things. First, no studies have matched prospective intentions (i.e., future fertility plans) with the actual intendedness of subsequent births in stepfamilies. Second, most studies have not actually compared stepfamily fertility with fertility in other unions but have only looked at fertility behaviors within a stepfamily (e.g., Hohmann-Marriott 2015; Holland and Thomson 2011; Thomson 2003). Still, if stepfamilies are deliberately having children to symbolize their commitment and more strongly feel a need to do so than unions that do not include children from prior relationships (Stewart 2002), this suggests the following hypothesis:

  • Hypothesis 2A: Women’s first shared birth in a stepfamily union is more likely to be intended than unintended compared with women’s first shared birth in a nonstepfamily union.

However, the growing body of literature on MPF—of which stepfamily fertility is a subtype—presents an alternative possibility, largely situated in the notion of MPF as characteristic of disadvantaged individuals and families (Klerman 2007). Women with MPF are much more likely than their counterparts whose children are all with the same partner to have started childbearing with an unintended birth (Guzzo 2014). In fact, having an unintended first birth increases the risk of union instability and thus exposure to forming a new relationship and having a birth with a new partner (Guzzo and Hayford 2012). Having an unintended birth of any parity, in turn, carries an elevated risk that subsequent births are also unintended (Guzzo and Hayford 2011). In general, MPF is linked to economic disadvantage and poor relationship quality (Carlson and Furstenberg 2006, 2007), as is ineffective contraceptive use and unintended childbearing (Manlove et al. 2014; Musick et al. 2009). Together, this suggests the following competing hypothesis:

  • Hypothesis 2B: Women’s first shared birth in a stepfamily union is less likely to be intended than unintended compared with women’s first shared birth in a nonstepfamily union.

Union Type, Fertility, and Intendedness

Yet another wrinkle emerges when thinking about the formation and stability of stepfamily unions, as well as changes in union formation overall. For the past two decades, the modal union after the demise of a cohabitation or marriage has been a new cohabitation (Coleman et al. 2000; Wu and Schimmele 2005). This trend mirrors rising overall levels of cohabitation, which is now the modal first coresidential union and precedes the majority of marriages (Rose-Greenland and Smock 2013). Rates of remarriage have declined substantially since 1990 (Brown and Lin 2013), as have first marriages after a nonmarital birth (Gibson-Davis 2011). Higher-order unions and marriages involving stepchildren have a higher risk of dissolution than first unions and unions without stepchildren (Sweeney 2010), but cohabitations are generally unstable.

With marriage the preferred setting for childbearing and childrearing (Pew Research Center 2010; Thorton and Young-DeMarco 2001), marriages are more likely to involve shared childbearing than cohabitations, regardless of stepfamily status. If the value of children perspective is true and stepfamilies feel a stronger need to establish themselves as a family, marital stepfamilies may be more likely to have a first shared birth—and to intend that birth—than marital unions involving two childless individuals. Most married women have children (Wu et al. 2016), though, so the differences by stepfamily status may not be large, especially because the majority of all marriages have to remain intact (and perhaps transition from a cohabitation) long enough to have a child, thus creating a select pool of married couples (Lillard and Waite 1993).

Despite the fact that Americans believe marital childbearing is ideal, cohabitation has increasingly become the site of childbearing (Manning et al. 2014), especially as nonunion conceptions are more often followed by pre-birth cohabitation (Lichter et al. 2014). Cohabiting stepfamilies may be more likely to have a birth because at least one member of the couple is already in a family-building stage, whereas cohabitations involving nonparents may be more transitory and involve those at earlier stages in the family life course. Births in cohabiting unions are less likely to be intended than births in marriages (Mosher et al. 2012), but this may be more so the case for cohabiting unions in which neither partner has children from past relationships than cohabiting unions in which at least one partner has a child from a past relationship. In sum, I propose the following hypothesis:

  • Hypothesis 3: Women’s risk of having a first shared birth, and intending that birth, is highest among marital stepfamily unions, followed by marital nonstepfamily unions, cohabiting stepfamily unions, and cohabiting nonstepfamily unions.

Stepfamily Configuration, Fertility, and Intendedness

Stepfamilies, of course, can take many forms depending on the number of children from past relationships and whether one or both partners has children: these aspects of configuration seem to influence childbearing decisions. Stepfamilies in which one partner is childless are more likely to have a shared child than when both partners already have children (Stewart 2002), perhaps stemming from desires for biological parenthood among the childless partner. Conversely, births to couples in which both partners already have children would elevate the couple’s overall childrearing responsibilities to a greater degree in a stepfamily than in a union in which only one partner has prior children because combined parity is likely to be higher; thus, I would expect these births are less likely to be intended. Combined and individual parity are important in general, given that intendedness is lower for third and later births (Mosher et al. 2012), and the greater obligations and costs of larger families suggests that additional births might also be unintended. Of course, it is not only whether partners have children from prior relationships that influence childbearing decisions in new unions; factors such as child coresidence, involvement, and financial support are also relevant because they tap into the intensity of the parent and stepparent roles (Hohmann-Marriott 2015; Vikat et al. 2004). Unfortunately, the current research is unable to account for these factors. Focusing on the configuration of stepfamilies as potentially influential for fertility behaviors leads to the following hypothesis:

  • Hypothesis 4: Fertility differences between unions with and without prior children depend on the configuration of the stepfamily, with women’s risk of a birth, and the intendedness of the birth, declining across a rough continuum of complexity from the simplest (only one partner has children) to the most complex (both partners have children, particularly if the female partner already has two or more children).

Data and Methods

The analyses use the 2006–2010 National Survey of Family Growth (NSFG). The NSFG is a nationally representative cross-sectional survey of 22,682 men and women. The data contain a detailed birth history as well as the start and end dates of all different-sex coresidential unions; however, the NSFG collected birth intendedness among men’s births only in the three years preceding the survey. As such, the analysis is restricted to women (n = 12,279) and further restricted to women who had at least one different-sex cohabiting or marital union (n = 8,304) with valid dates on union starts and ends (n = 8,248).

The primary independent variables are stepfamily status (i.e., whether the union is a stepfamily), configuration (which partner has how many children), and union type (cohabitation or marriage). The latter is straightforward to measure, but the first two variables are more challenging. An advantage of the NSFG is that for each cohabitation or marriage, the respondent is asked whether their partner had any children from prior unions (although the number, age, and coresidence of each partner’s children are not available). Partner’s children information is missing for nine women, reducing the sample size to 8,239. For women’s own prior fertility, I draw from the birth history. As with prior work on fertility and unions (Thomson et al. 2014), births occurring more than six months prior to the start of a union are considered to be with a different partner, and births occurring within the six-month period are considered to be with the cohabiting or marital partner. This widespread assumption and practice is supported by recent work demonstrating that the odds of union formation following a nonunion birth are quite low in the first six months and that the odds of forming a union with a new partner, rather than the child’s other biological parent, increase over time (Guzzo 2016). Those women who had a birth prior to coresidence but within the six months preceding union start (and thus with their cohabiting or marital partner) (n = 162) are excluded because the analytical strategy (discussed later) models the risk of a birth beginning with the start of coresidence because information on the start of dating is unavailable. This reduces the sample size to 8,077 women. Results, available upon request, are virtually identical when using more or less restrictive assumptions: no births prior to union start are considered to be with that partner, and thus all children from prior unions versus births up to a year prior to union start are considered to be with that partner, and only births more than a year earlier are children from prior unions. Only 288 women in total (of more than 8,000) had a birth in the year prior to coresidence, so changing assumptions recategorizes births and changes sample sizes for a very small proportion of the analytical sample.

For each union, I created two sets of variables. The first is a dichotomous indicator of whether the union was a stepfamily. The second is a categorical measure with six categories for stepfamily configuration: (1) neither partner has children from a past relationship (nonstepfamily union); (2) respondent has one child, partner has no children; (3) respondent has two or more children, partner has no children; (4) respondent has no children but partner has children; (5) respondent has one child, partner has children; and (6) respondent has two or more children, partner has children.

The analyses use event-history models to examine the outcomes of women’s unions, treating a first birth as an outcome; I discuss these models in more detail later. These outcomes are categorized in two different ways, both of which are time-varying. The first combines information from the birth histories and the union histories to create a three-category variable: (1) intact union with no birth; (2) intact union with a birth; or (3) union dissolved. The second dependent variable disaggregates births by intendedness. 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 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?” As has become standard with NSFG and other research on unintended births (Chandra et al. 2005; Lindberg et al. 2015), 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. Further, research has shown that births that are less than two years too early strongly resemble wanted births in terms of maternal behaviors and child outcomes, whereas births that are two years or more too early strongly resemble unwanted births (Lindberg et al. 2008; Pulley et al. 2002). As such, I dichotomize intendedness into wanted or moderately mistimed (intended) versus seriously mistimed or unwanted (unintended), which is an approach that has become increasingly common in unintended fertility research (e.g., Guzzo and Hayford 2011, 2012, 2014). Analyses keeping the categories separate revealed essentially no differences between wanted and moderately mistimed or between seriously mistimed and unwanted but showed differences between these two groupings (not shown). Intendedness here relies entirely on the woman’s report; her partner’s categorization may differ but is unavailable in the data. Incorporating intendedness resulted in the second dependent variable, with four categories: (1) intact with no birth; (2) intact with an intended birth; (3) intact with an unintended birth; or (4) dissolved. Intendedness was missing for all births for 13 women, leaving a final sample size of 8,064 women.

To examine fertility differences across stepfamily status while accounting for the differences in union type and stability, I first created a file for each of women’s cohabitations and marriages: every new union represents another opportunity to have a first shared birth with a new partner. That is, all unions have a shared parity of 0 at the start of coresidence, but the woman (or her partner) may not have an individual parity of 0 because a first shared birth with a particular partner is not necessarily the woman’s or her partner’s first birth as an individual. Women with more than one union thus contribute more than one record; this approach restructures the data from wide to long format (with an indicator of union number), and cohabitations that transition to marriage are counted as one union. Put differently, this approach reconfigures the data so that instead of one record per woman with all her union information listed as separate variables, the data now contain one record per union per woman: each of a woman’s unions becomes a unique data point for analysis. This created a file of 11,378 (6,662 unions in which both the respondent and her partner were childless and 4,716 unions in which either the respondent or her partner had a child from a past relationship). Next, I created a person-month file consisting of a separate observation for each month that a woman was at risk of a first birth with their partner in a specific union. A risk spell began the month when the coresidential union started, and ended when a birth occurred or the union dissolved, with women censored at the time of survey if they neither had a birth nor experienced dissolution. Thus, the final analytical sample consists of 11,378 risk spells (i.e., cohabitations and marriage) and 378,552 person-months to 8,064 women. Union duration is included as a time-varying piecewise nonlinear spline (less than 7 months, 8–18 months, 19–36 months, 37–48 months, and 40 or more months).

In addition to union characteristics, the analyses also take into consideration factors linked in prior research to both stepfamilies and birth intentions. Primary among these are union type, and so a time-varying dichotomous indicator of cohabitation versus marital status is included, along with age at union start and a categorical indicator of union number (first, second, third, or fourth or higher). Socioeconomic disadvantage is linked to unintended fertility (Mosher et al. 2012; Musick et al. 2009), and the analyses include time-invariant measures of race/ethnicity/nativity, family structure at age 14, and mother’s education. High school diploma is included as a time-varying characteristic; the date was missing for 10 respondents, and these cases were assigned the average graduation month for others their age (other approaches for imputation did not change the results). The control variables have no other missing data. Finally, prior unintended fertility is associated with higher-order unintended fertility (Guzzo and Hayford 2011), so prior unintended fertility is included as a dichotomous indicator; this is time-invariant within unions but can vary across unions.

Analytical Strategy

I begin by presenting the weighted descriptive characteristics of the analytical sample of women’s unions at risk of experiencing a first shared birth. Next, I show the weighted bivariate association between (1) stepfamily status; (2) stepfamily status by union type; and (3) stepfamily configuration and the dependent variable, looking at the first outcome variable (no birth, birth, dissolution) and differences between intendedness among those with a birth. I then turn to multivariate analyses. I begin by testing whether women in stepfamilies have a higher risk of a first shared birth than women in unions in which neither partner has children (Hypothesis 1), showing the relative risk ratios (RRRs) from multinomial logistic regression event-history models. I then test whether births to women in stepfamilies are more or less likely to be intended than births to women in nonstepfamily unions (competing Hypotheses 2A and 2B). Next, I test whether the risk of fertility and intendedness varies across both stepfamily status and union type (Hypothesis 3). Finally, I examine whether stepfamily configuration is associated with overall fertility and intendedness (Hypothesis 4). In the multivariate analyses, fertility and dissolution are competing risks, but the discussion of the results will focus primarily on fertility. 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 observations contributed by some respondents.

Results

Descriptive Results

Table 1 shows the weighted descriptive statistics for the analytical sample, disaggregating by the dichotomous indicator of stepfamily status. Overall, more than one-third of the unions of women of childbearing age are stepfamilies; of those, the modal category is for a childless woman to be partnered with a father (34 %). Nearly as common, though, is for mothers to be partnered with a father: 32 % of stepfamilies involve women with one child and a father with children from a prior union (15 %) or women with two or more children and a father (17 %). The majority of all unions are cohabitations at the start of coresidence, although there is variation, at 65 % of nonstepfamily unions and 84 % of stepfamilies; by the last month of observation, only 39 % of nonstepfamily unions are still cohabitations compared with 57 % of stepfamily unions (not shown).

Table 1 Weighted descriptive statistics for women’s unions with no shared children (number of women = 8,064)
Full size table

In general, women in stepfamily unions are less advantaged than those in unions in which neither partner has children from a prior relationship. A higher percentage of women’s stepfamily unions involve those who are nonwhite, whose mothers have lower levels of education, and who did not grow up living with both biological parents. Fewer women in stepfamily unions are high school graduates at the start of coresidence than their peers in nonstepfamily unions (63 % vs. 76 %, respectively). Women in stepfamily unions are significantly older, by about 3.5 years, than their counterparts as well, which is likely due to the time spent in prior relationships. The vast majority of women in nonstepfamily unions are living with a partner for the first time (85 %); for women in a stepfamily union, more than one-half (56 %) are in a second or higher union. In addition, although 77 % of first unions involve two childless individuals, this is far less likely the case for higher-order unions: only 34 % of second unions involve two individuals with no children from prior relationships, declining to 22 % for third unions and to 18 % for fourth and higher unions (not shown). Finally, just under one-half of women’s stepfamily unions involve women who had a past unintended birth with a different partner; by definition, none of the nonstepfamily unions include a history of prior unintended fertility.

Table 2 shows the outcome of the unions by stepfamily type. Recall that births and dissolution are competing risks, such that some of the unions that experience a birth may later go on to dissolve, but none of the dissolutions in this analysis involved a birth prior to dissolution (and, of course, any union still intact at the time of survey may experience either a birth and/or dissolution that is not observed). Looking at the first set of figures, which are not disaggregated by intendedness, approximately one-half (52 %) of the full analytical sample of unions exited the analysis via a first birth with their partner. Less than 30 % dissolved before the survey without having a birth, and 18 % were still intact and had no shared children. These percentages varied by stepfamily status: 56 % of those in nonstepfamily unions had a birth, and 27 % dissolved; fewer stepfamily unions had a birth (46 %), and more dissolved (34 %). Outcomes also varied by stepfamily status and union type: the percentage with a shared first birth was considerably higher among marital unions in which neither partner had prior children (70 %) than other groups, and dissolution was highest among cohabiting unions in which neither partner had prior children. Having a child also differed by stepfamily configuration. Stepfamily unions in which only the female respondent had a single child and her partner had no previous children had the highest percentage having a shared birth, at 56 %. At the other end of the spectrum, only 27 % of stepfamily unions in which the respondent already had two children and her partner also had children experienced a first shared birth.

Table 2 Weighted bivariate distribution of first birth, intendedness, and stability by stepfamily status, stepfamily status by union type, and stepfamily configuration (8,064 women with 11,378 unions)
Full size table

The latter two columns of Table 2 examine the distribution of intendedness among those with a first shared birth. Of births within all unions, 75 % are intended, but this varies considerably by the categories considered here. The majority of births in both nonstepfamily and stepfamily marriages were intended, at 86 % and 83 %, respectively. For births in cohabiting stepfamilies, approximately 59 % were intended, higher than for births to cohabiting nonstepfamily unions (50 %). There is also variation in intendedness by stepfamily configuration among those who had a birth. For stepfamily unions in which women have only one child, regardless of whether her partner also has children, approximately 75 % of births are intended. However, the percentage of births that are intended falls substantially for other stepfamily types, ranging from 69 % (respondent has two or more children but her partner has no children) to 59 % (respondent has two or more children and her partner also has children).

These results do indeed support the overall hypothesis that fertility behaviors differ between stepfamily and nonstepfamily unions in addition to the hypotheses that stepfamily union type and configuration matters. However, socioeconomic and demographic factors as well as union-specific factors and prior fertility differ by stepfamily status, as shown in Table 1, and these differences may play into differences in fertility behaviors (and stability). To disentangle these relationships, I turn to multivariate analyses.

Multivariate Results

Fertility by Stepfamily Status

Results from multinomial logistic regression using event-history models are presented in Table 3. This table presents two sets of RRRs. Model 1 examines whether there are differences in the risk of a first birth (and dissolution) relative to remaining intact with no birth across stepfamily status. Model 2 disaggregates fertility by intendedness. Both models include the full set of control variables.

Table 3 Relative risk ratios multinomial logistic event-history models predicting fertility and stability in women’s unions by stepfamily status
Full size table

Model 1 shows support for Hypothesis 1: women in stepfamily unions are 15 % more likely to have a first shared birth with their partner than women in nonstepfamily unions. Stepfamily unions and nonstepfamily unions do not differ in stability. Although this finding may seem counterintuitive, it is largely a function of having fertility as a competing risk in which individuals leave the risk set after a birth. In analyses that follow unions through to dissolution or the survey date and treating fertility as a time-varying covariate (not shown), stepfamily unions do have a significantly higher risk of dissolution than nonstepfamily unions, by approximately 15 %.

First-birth risks are higher among nonwhites, those whose mother had less than a high school diploma, and those who grew up in a stepfamily themselves. The risk of having a first birth relative to remaining intact and having no birth is twice as high (RRR = 1.91) among women in marriages than in cohabiting unions and is highest 8–18 months after the start of coresidence (RRR = 1.17) but lower in very early durations or durations longer than 36 months. Relative to women whose mother had only a high school diploma, women whose mother had a college degree or more are less likely to have a birth (RRR = 0.86), as are women with at least a high school diploma themselves relative to those with less than a high school diploma (RRR = 0.87).

Intendedness by Stepfamily Status

Model 2 of Table 3 tests Hypotheses 2A and 2B. Here, the omitted category is an unintended birth to demonstrate whether first births are more or less likely to be intended; the first column is the key column of interest. In support of Hypothesis 2B, births in stepfamily unions are significantly less likely to be intended than unintended, by 27 %, relative to births in unions in which neither partner had a child from a past relationship. When the omitted category is switched to “intact, no birth” (not shown), stepfamily unions are 42 % more likely to have an unintended birth than have no birth, with no differences between an intended birth relative to no birth across stepfamily status. Stepfamily unions are also less likely to remain intact or dissolve relative to having an unintended birth. In this model, the association between the covariates and the outcome categories also varies from Model 1, but this is partially due to the shift in the reference category from “intact, no birth” to “unintended birth.” In this model, blacks are marginally less likely (RRR = 0.81) to have an intended birth than an unintended birth relative to whites, whereas foreign-born Hispanics are about 60 % more likely. Having lived in an “other” family type during adolescence or having a prior unintended birth decreases the risk that a birth is intended rather than unintended, and births occurring in the first 18 months of coresidence are less likely to be intended relative to births occurring later in the union. Births in marital unions are more than twice as likely to be intended than unintended compared with cohabiting unions (RRR = 2.44), and age is positively associated with intended fertility.

Fertility and Intendedness by Stepfamily Status and Union Type

Models testing Hypothesis 3 are presented in Table 4. These models parallel those seen in Table 3 but use a variable combining stepfamily status and union type and vary the omitted category of this new variable. The reference group for each model is indicated in the table subheadings as “contrasts.” These models are full models with all covariates, but only the RRRs for the stepfamily-union type variable are displayed in this table because the RRRs for the other covariates are virtually identical to those in Table 3. Because the interest lies in how union type may be differentially related to both the risk of a birth overall and the intendedness of any births, only the contrasts showing a birth versus remaining intact (corresponding to the first column in Model 1, Table 3) and showing an intended birth versus an unintended birth (corresponding to the first column in Model 2, Table 3) are presented. These results only partially support Hypothesis 3. Compared with cohabiting nonstepfamily unions (shown in the columns labeled Contrast 1 in Table 4), cohabiting stepfamily unions are 60 % more likely to have a birth, although there is no difference in intendedness. However, compared with marital stepfamily unions, cohabiting stepfamily unions (Contrast 2) are less likely to have a birth and less likely to intend a birth if they do have one. Marital stepfamilies (Contrast 3) do not differ in their risk of having a birth relative to marital unions in which neither partner had a child from a previous relationship, but any births that do occur are less likely to be intended than unintended (RRR = 0.59).

Table 4 Relative risk ratios from multinomial logistic event-history models predicting a first shared birth and intendedness, with varying reference groups for stepfamily union type
Full size table

Fertility and Intendedness by Stepfamily Configuration

Finally, models testing Hypothesis 4 regarding stepfamily configuration are shown in Table 5. As with Table 4, these are the results from a series of full models replicating Models 1 and 2 in Table 3 but showing only the contrast between birth versus remaining intact and intended versus unintended, varying the omitted category to explore differences in stepfamily configuration. The reference group is indicated in the table subheadings as “contrasts.” The results in this table provide partial support for Hypothesis 4: there are differences in the risk of having a first birth overall and in the intendedness of any births between nonstepfamily unions and different types of stepfamilies. As shown in the first column, only stepfamily unions in which the woman has one child have an elevated risk of a birth relative to nonstepfamily unions (Contrast 1, Table 5), regardless of whether her partner has children (RRR = 1.37) or not (RRR = 1.48). Further the next column shows no differences in the intendedness of a birth between stepfamily unions in which the woman has only one child (again regardless of whether their partner has children) and nonstepfamily unions. Births that occur in any other type of stepfamily are significantly less likely to be intended than unintended relative to births in a union in which neither partner had children from prior relationships. Moving across the table, which shows models with different omitted categories for stepfamily configuration, fertility behaviors vary across these groups. For instance, stepfamily unions in which the female partner has only one child and is partnered with a childless man (Contrast 2) are more likely to have a birth, and to intend that birth, than stepfamily unions involving childless women partnered with a father or those involving women with two or more children of their own. When the respondent has two or more children and her partner has at least one child (Contrast 5), the risk of having a birth, and intending a birth, is significantly lower than when the respondent only has one child and her partner has at least one child.

Table 5 Relative risk ratios from multinomial logistic event-history models predicting a first shared birth and intendedness, with varying reference groups for stepfamily configuration
Full size table

Discussion

The past few decades have witnessed remarkable changes in fertility behaviors as well as in union formation and stability. These shifts, in turn, mean that many parents are in the relationship market, and many childless individuals will form unions with parents. When either (or both) partner(s) has a child from a past relationship, a stepfamily is formed, and when parents form new unions or childless individuals partner with parents, these individuals are at risk of having a birth with their new partner—that is, childbearing across partnerships.

Of course, any union is potentially at risk of having a shared birth. Couples have children for many reasons (Morgan and King 2001), but stepfamilies present a unique situation because one or both members of the couple already have children, and higher parities are generally associated with lower fertility risks. Stepfamilies seem to have a different orientation toward childbearing than unions in which both partners are childless at the start. Because of the growth in stepfamily formation over the past few decades, and spurred by potential theoretical differences, stepfamily fertility has become a focused area of research in recent years, although the bulk of this work has used European data (e.g., Henz and Thomson 2005; Holland and Thomson 2011; Thomson 2003; Thomson et al. 2014; Vikat et al. 1999, 2004).

Two lines of research have examined childbearing across partnerships. The first line has explicitly focused on stepfamilies, looking at planned and actual fertility. Stepfamilies often report that they plan to have children in the future in their new union (Hohmann-Marriott 2015; Stewart 2002; Thomson 2003). And in fact, childbearing in stepfamilies is common, with approximately one-half of stepfamilies having a shared child (Holland and Thomson 2011), even at individual parities that are usually at low risk for subsequent childbearing (Thomson et al. 2012). The primary explanation for both planned and actual stepfamily fertility is that, like fertility in other unions, having a shared child serves a purpose by linking a couple, expressing their commitment, and creating social capital (Astone et al. 1999; Coleman 1988; Friedman et al. 1994; Ganong and Coleman 1988). However, stepfamilies could be especially motivated to have a shared child: the need to establish themselves as family unit and create shared capital may be higher than in nonstepfamilies (Stewart 2002). If this is true, stepfamilies would be more likely to have not only a shared birth but also an intended rather than unintended first birth, yet no prior research has explicitly examined the intendedness of births in stepfamily unions relative to births in other unions. The second line of research has looked at the newly recognized (but not actually new) phenomenon of MPF, of which stepfamily fertility is a subtype; this literature has largely developed separately from work on stepfamilies. MPF is generally considered to be a deleterious behavior (Klerman 2007), stemming from problematic fertility and relationship behaviors. According to this perspective, MPF does not result necessarily result from decisions to have a child to create social capital or express commitment but is more commonly associated with social and economic disadvantage, with unintended fertility playing a large role (Guzzo 2014). Putting these two frameworks together—and in light of the fact that one-third of all births are unintended (Finer and Zolna 2014)—unintended fertility could play a role in stepfamily fertility. To date, however, the intendedness of births in stepfamily unions has not been explored.

Drawing from these two literatures as well as unintended fertility research, this study examined four sets of hypotheses to compare the risk of a first birth, and its intendedness, across stepfamily status. The first hypothesis tested the value of children perspective—that stepfamilies are more likely to have a shared child than unions in which neither partner had children from a prior relationship (i.e., a nonstepfamily union). The second hypothesis examined whether births in stepfamily unions are more or less likely to be intended than births to nonstepfamily unions. Prior research, at least in Europe, has interpreted high levels of fertility in stepfamilies as consistent with the value of children perspective, but I suggest that without determining whether such births are actually intended, any such conclusion is premature. The third hypothesis tested whether the risk of having a shared birth together and its intendedness varies by marital status. Much of the existing European work has not distinguished between cohabiting and marital unions (e.g., Henz and Thomson 2005; Holland and Thomson 2011), but these unions are far more similar in Europe than the United States, where cohabiting unions are considerably more unstable and disadvantaged. Finally, the fourth hypothesis examined whether the risk of a first shared birth and its intendedness varies by stepfamily configuration—that is, which partner had a child and how many children the female respondent had.

The results here support Hypothesis 1: compared with unions in which neither partner had children from a prior relationship, stepfamily unions are more than 15 % likely to have a shared birth together. At first glance, then, this supports the value of children perspective. However, the models testing Hypothesis 2A and 2B challenge this interpretation: births in stepfamily unions are significantly less likely (by approximately one-quarter) to be intended than unintended relative to births in nonstepfamily unions. Further, and in contrast to Hypothesis 3, birth intendedness varies in unexpected ways across stepfamily status and union type. Although first-birth risks do not differ across marital unions, first births in marital stepfamily unions are 40 % less likely to be intended than unintended relative to the first shared birth in marital unions in which neither partner had children from a prior relationship. It is unclear why women in marital stepfamilies have an elevated risk of an unintended birth compared with other married women. Perhaps these women have more reservations about their unions (given either their own and/or their partner’s prior relationship failures), which could affect contraceptive behavior. Alternatively, married women in stepfamilies may view and characterize their births differently, perhaps with less certainty or more concern over the practical implications of additional children. Certainly, though, it does not appear that marital stepfamilies are more likely to engage deliberately in shared childbearing as a way of legitimizing their family unit compared with other married couples, as suggested in other work (Stewart 2002).

As expected, however, marital stepfamilies are more likely than cohabiting stepfamilies to have a first shared birth and to intend that birth. This finding suggests that even among stepfamilies, marriage is a more acceptable context for childbearing and childrearing. Cohabiting stepfamilies are more likely to have a birth than cohabiting nonstepfamilies, with no differences in intendedness. The higher fertility in cohabiting stepfamilies could be due to countervailing fertility behaviors. On the one hand, they may experience more unintended fertility because of the same processes occurring in marital stepfamilies and with MPF in general. On the other hand, cohabiting stepfamilies could also be more likely to have an intended first birth because at least one member is already in a family-building stage. Also possible is that cohabitors with children from prior relationships may view cohabitation as a more acceptable site for childbearing than their childless counterparts, given prior failed relationships. Fertility and intendedness also differ by configuration (Hypothesis 4): when the female partner has only one child, stepfamily unions are more likely than other unions to have a first shared child, perhaps because of the strong two-child norm in the United States. Although intendedness does not differ when the female partner has only one child (regardless of whether her partner has any children), any births that do occur for other types of stepfamily unions (e.g., women with two or more children or childless women partnered with a father) are significantly less likely to be intended than unintended relative to births in nonstepfamily unions.

In sum, then, the results here align more strongly with the perspective that childbearing across partnerships in the United States is part of a story of economic and relational disadvantage, as suggested by the growing body of work on MPF. No evidence supports the notion that fertility is higher in stepfamilies than nonstepfamily unions because stepfamilies are deliberately engaging in shared childbearing to establish themselves as a family unit, make symbolic statements about their union, or create shared social capital. Instead, although stepfamily unions often do involve a shared birth, these births are often unintended and thus may represent an additional source of strain in these more complicated family situations.

Limitations

Several limitations in this work can be identified. A primary problem is that the information gathered comes exclusively from the female partner, who was reporting about her partner’s prior childbearing information. In the NSFG, for women’s partner’s children, information on the number of children (Ivanova et al. 2014) and whether they lived with the respondent and her partner (Vikat et al. 1999, 2004) was available only for married stepfamilies at the time of the interview and was thus not included in the analysis. In addition, I was unable to include information about visitation and financial support. Residence information for women’s own children is also not included because of the lack of time-varying data—an omission that is problematic given evidence suggesting that coresident children may influence fertility decisions more strongly than nonresident children (Hohmann-Marriott 2015). Coresident biological children may provide more returns to the parental role (in the form of love and affection), but they may also entail more financial and time obligations; stepparents may also experience these costs and benefits to a greater extent when stepchildren are coresident. The data also include no information about whether the partner had a past unintended birth or about the partner’s characterization of the focal birth as intended. One partner may characterize the birth as intended while the other partner does not, and this situation could be more common in stepfamilies than unions in which neither partner had children from a prior relationship because the partners do not necessarily share the same individual-level parity. To the extent that stepfamilies are more likely to have disagreement about the intendedness of shared births, then unintendedness is conservatively estimated for stepfamilies.

As is the case for most research on unintended fertility, the analyses use retrospective reports of pregnancy wantedness, which are potentially subject to social desirability bias or retrospective rationalization (Musick 2002; Trussell et al. 1999; Williams et al. 1999). In general, though, birth intendedness reports have high face validity (Bachrach and Newcomer 1999; Joyce et al. 2000). There is little reason to suspect that social desirability bias differs by stepfamily status. If anything, stepfamilies might be more motivated to categorize a birth as intended if they have concerns over how their current behaviors are interpreted in light of their past relationship instability and fertility behaviors; in this case, these are conservative estimates of unintended fertility in stepfamilies. Because the data are not longitudinal, the analysis could not match prospective fertility plans with actual fertility behavior and birth intendedness; most prior work has focused only on prospective fertility plans or did not match fertility plans with actual intendedness of subsequent births. These analyses excluded pregnancies that did not end in a live birth because abortions are underreported in survey data (Jones and Kost 2006), which likely underestimates unintended pregnancies in both stepfamilies and cohabitations overall to a greater extent than married families in which neither partner had children from a prior relationship. Finally, women who never cohabit or marry but who have children are not included in this analysis; this applies to only 2.5 % of the full NSFG sample, and the majority (73 %) of these women had at least one unintended birth.

Conclusion

The conditions driving childbearing across partnerships—nonmarital and unintended childbearing, along with parental union instability and repartnering—show little sign of abating. As such, families will likely become even more complicated in the future, warranting attention to the conditions in which complex families occur. As shown in this study, when stepfamilies have a shared child, these births are often unintended, indicating that family processes work differently in stepfamilies than in more traditional families. To the extent that unintended births can strain relationships (Boden et al. 2015; Guzzo and Hayford 2012, 2014), these births may be an additional source of stress and a potential cause of instability among stepfamilies, which already face issues of navigating complex and unclear family roles and expectations (Cherlin 1978; Cherlin and Furstenberg 1994).

These findings also present a picture of diverging destinies (McLanahan 2004). Couples with no prior children are more likely to plan their births and have their children in more stable and normative conditions, such as marriage. Stepfamilies with shared children, conversely, are often in less stable unions, and many of the births are unintended. Cohabiting unions with children are less stable than marriages with children (Manning et al. 2004), and unions with an unintended birth are more unstable than those with only intended births (Guzzo and Hayford 2011). Thus, children in stepfamilies face an elevated risk of experiencing the dissolution of this union, and this could be especially problematic for nonshared children, who have already experienced their biological parents’ relationship dissolution. Children who are deliberately planned and wanted in committed relationships, who experience stable childhoods, are becoming an increasingly select and advantaged group. Their counterparts who grow up in less stable circumstances will likely face more challenges with fewer familial resources, making it a challenge for society to identify ways to provide more support.