Abstract
We document the incidence and evolution of family complexity from the perspective of children. Following a cohort of firstborn children whose mothers were not married at the time of their birth, we consider family structure changes over the first 10 years of the child’s life—considering both full and half-siblings who are coresidential or who live in another household. We rely on detailed longitudinal administrative data from Wisconsin that include information on the timing of subsequent births to the mother and father, and detailed information on earnings, child support, and welfare. We find that 60% of firstborn children of unmarried mothers have at least one half-sibling by age 10. Our results highlight the importance of having fertility information for both fathers and mothers: estimates of the proportion of children with half-siblings would be qualitatively lower if we had fertility information on only one parent. Complex family structures are more likely for children of parents who are younger or who have low earnings and for those in larger urban areas. Children who have half-siblings on their mother’s side are also more likely to have half-siblings on their father’s side, and vice versa, contributing to very complex family structures—and potential child support arrangements—for some children.
Introduction
Multiple-partner fertility has been the focus of recent interest in the academic literature. Increases in nonmarital fertility, as well as increases in divorce and remarriage, have contributed to higher proportions of mothers and fathers who have had children with more than one partner. Multiple parenting relationships result in growing family complexity; over the course of their childhood, many children will come to share a household and parent with half-siblings and to share a nonresident parent with other half-siblings who live elsewhere. Multiple-partner fertility has also been of concern to policymakers, particularly to those dealing with child support. Determining how much formal child support nonresident parents should provide for their children when these children are spread across multiple families is a thorny but important policy problem (Meyer et al. 2005). Accounting for fathers’ multiple-partner fertility substantially decreases estimates of the amount of child support that could potentially be collected (Sinkewicz and Garfinkel 2009).
In this article, we document the incidence and evolution of family complexity from the perspective of nonmarital children. Following a cohort of firstborn children whose mothers were not married at the time of their birth, we document changes in family structure, considering both full and half-siblings who are coresidential or who live in another household. We rely on detailed longitudinal administrative data that capture almost 90% of all nonmarital births in the State of Wisconsin. These data allow us to consider the timing of subsequent births to the mother and father, together or with new partners, and to account for siblings and half-siblings, even if they are not coresident. We are thus able to consider the evolving complexity of sibling relationships and the potential consequences for resources available to children. We focus on nonmarital children because they are a growing group, they are the subject of public policy concern, and they can be considered at risk of gaining a new half-sibling from the time of their birth (whereas children born inside marriage generally enter the risk set only if and when their parents separate).1 We are particularly interested in the extent to which parental multiple-partner fertility seems to be mutually reinforcing—that is, the extent to which children whose mothers have had multiple partners are more likely to have fathers who have had multiple partners.
After reviewing related literature, we describe the unique data used for our analysis, discussing their advantages and limitations. We then document changes in family complexity from the perspective of firstborn nonmarital children and analyze the observable characteristics of mothers and fathers that are related to their entering into new parenting partnerships following the mother’s first birth. We close with a discussion of implications for child support policy.
Literature
In 2009, 41% of U.S. children were born to unmarried mothers (Hamilton et al. 2010). This percentage has risen steadily in the last two decades, from 23% in 1986 to 32% in 1996 and 41% in 2009 (Hamilton et al. 2010; Martin et al. 2009). Some of the increase is related to increases in cohabitation (Wu and Wolfe 2001). Nonmarital fertility has also been linked to the declining economic prospects of men (e.g., Moffitt 2001; Willis 1999) and other economic variables (e.g., Aassve 2003), to changes in norms (e.g., Axinn and Thornton 2000), and to the extent to which children are seen as social capital (Schoen and Tufis 2003). While the increasing prevalence of nonmarital births and some of its covariates are clear, there is substantial debate regarding the consequences of nonmarital fertility for child well-being (Amato 2005; Amato and Maynard 2007; regarding nonmarital teen births, see Fletcher and Wolfe 2009; Furstenberg 2007). In part, the debate reflects the difficulty of identifying the causal effects of nonmarital fertility per se, since having a child outside of marriage is more likely for individuals who are already at higher risk for negative socioeconomic outcomes.
One aspect of nonmarital fertility that has yet to be fully documented is the relationship between nonmarital birth and the risk and timing of exposure to complex family organization. The emerging evidence suggests that nonmarital fertility increases the risk of having children with multiple partners, but the mechanisms and consequences of this relationship are not well understood. In this brief review, we build on a relatively new literature that considers multiple-partner fertility.2 We focus on estimates of the prevalence, correlates, and consequences of multiple-partner fertility.
Estimates of prevalence are complicated by the lack of necessary details on fertility history in many nationally representative data sets. Nonetheless, an emerging body of research shows that multiple-partner fertility is fairly common. Guzzo and Furstenberg (2007a), using the National Survey of Family Growth (NSFG), found that 17% of all fathers between the ages of 15 and 44 have had children with more than one partner. Estimates using data from the Fragile Families and Child Wellbeing Study (hereafter referred to as the Fragile Families Study), which is representative of a recent birth cohort in large cities, suggest that for most couples having a child outside of marriage, at least one of the parents already has, or will have, a child by another partner (Carlson and Furstenberg 2006). Recent research drawing on the National Survey of Family Growth has documented the increasing prevalence of nonmarital multiple-partner fertility in more recent cohorts (Guzzo and Furstenberg 2007a; Manlove et al. 2008). Even studies that considered married or formerly married partners have found that multiple-partner fertility is not rare; for example, such complex families characterize 22% of urban married-couple families in the Fragile Families Study (Carlson and Furstenberg 2006) and more than one-third of all divorcing families in the child support enforcement system in Wisconsin (Cancian and Meyer 2006). Overall estimates of prevalence are difficult, in part because broad national data tend to focus on the fertility of one or the other parent (rather than both parents). Data from the Fragile Families Study have the advantages that they can examine both parents and they are national, but they are available only for urban births. Moreover, the best information within this data source comes from mothers, and the accuracy of their reports of the fertility of the fathers of their children is unknown.
Much of the research addressing the correlates of multiple-partner fertility for those with nonmarital births has considered important family transitions. Factors related to nonmarital fertility (e.g., length of period of risk, contraceptive behavior, economic prospects, norms, and the extent to which children are social capital) may be considered, as well as factors associated with relationship breakup, with repartnering, and with fertility in new relationships. The literature on multiple-partner fertility has focused particularly on the role of demographic characteristics (especially age and race, and sometimes parity), but also on economic capacities, family history, relationship status, and individual attitudes.
Rates of multiple-partner fertility are particularly high in samples of economically disadvantaged parents (Meyer et al. 2005); similarly, the risk of multiple-partner fertility is higher among those with less education (Carlson and Furstenberg 2006). The risk of multiple-partner fertility has been found to be greater for those who had a first child at a younger age (Carlson and Furstenberg 2006; Manlove et al. 2008). Blacks and, to a lesser extent, Hispanics are more likely to experience multiple-partner fertility (Carlson and Furstenberg 2006; Guzzo and Furstenberg 2007a; Manlove et al. 2008), though differences are substantially reduced (and, in some cases, no longer statistically significant) in multivariate models with extensive controls. Carlson and Furstenberg (2006) also found a higher risk of multiple-partner fertility for men with a history of incarceration and for women (but not men) whose previous partner was of a different race or ethnicity. Those growing up in a single-parent family have higher rates of multiple-partner fertility (Carlson and Furstenberg 2006; Guzzo and Furstenberg 2007b; Manlove et al. 2008). Unmarried parents are more likely to have multiple-partner fertility than married parents (Carlson and Furstenberg 2006; Guzzo and Furstenberg 2007b). Finally, multiple-partner fertility is lower among those having more than one child with an initial partner (Manlove et al. 2008). None of the published research has examined whether rates of multiple-partner fertility differ across urban, suburban, and rural areas.
Researchers have also considered the implications of multiple-partner fertility and resulting family complexity for the economic and time resources available to children (Carlson et al. 2008; Harknett and Knab 2007; Manning and Smock 1999, 2000), family conflict (Carlson and Furstenberg 2008; Jayakody and Seefeldt 2008), union formation (Carlson et al. 2004), child support payments (Manning and Smock 2000; Meyer et al. 2005), the risk of child maltreatment (Berger et al. 2009), and a range of child outcomes (Evenhouse and Reilly 2004; Ginther and Pollak 2004). As a whole, the literature identifies the risks and complications associated with mothers and fathers having multiple partners. These risks include fewer investments in children, greater conflict, and lower probability of marriage or continued cohabitation. However, there is also some evidence that multiple partners may provide a type of insurance—whereby one partner may compensate for another (see, e.g., Meyer et al. 2005; though see also Harknett and Knab 2007). Moreover, while much of the literature on children in blended families suggests that they have worse outcomes than those in traditional two-parent families, recent research suggests that for many outcomes, some (Evenhouse and Reilly 2004; Ginther and Pollak 2004) or all (Hofferth 2006) of the difference may be explained by selection—that is, by differences in the characteristics of parents involved in different family forms, rather than being the consequence of the family structure per se.
The current analysis contributes to the literature on multiple-partner fertility in several ways. Using data on a cohort of nonmarital births in Wisconsin,3 we trace the evolution of family complexity over a 10-year period. We take the perspective of the child and consider the risk and timing of half-siblings on the mother’s and father’s side. One of our contributions is the use of unique data derived from administrative records. Although administrative records include limited demographic covariates, and no information on fertility intentions, informal work, or transfers, they do have substantial advantages. For the full 10-year period, we have measures of subsequent nonmarital births to both parents and detailed, accurate, time-varying measures of formal earnings, formal child support paid and received, and public program participation. Some of the previous work has only considered one parent’s later fertility or used proxy reports for later fertility, in part because men’s nonmarital fertility is so difficult to measure accurately (see, e.g., Bachu 1996; Garfinkel et al. 1998; Rendall et al. 1999). In contrast, we have substantial information on both men’s and women’s nonmarital fertility, which allows us to examine whether the multiple-partner fertility patterns of the child’s parents are related. Although we have substantial coverage of fertility, we do not have complete coverage. Online Resource 1 provides alternative estimates of family complexity using different assumptions. One of the advantages of our data is that they include births statewide, enabling us to compare family complexity for children in rural, suburban, and urban areas. In contrast, some related research has only been able to examine urban births.
Methods
Data and Sample
We use a unique set of data derived from State of Wisconsin administrative systems, primarily from the child support enforcement data system (named KIDS). KIDS contains a record for every child for whom a referral to the child support agency was required (welfare cases) as well as for any child whose parent initiated contact with the child support agency for help with paternity establishment, locating a nonresident parent, establishing or changing a child support order, or collecting a child support order. It also includes divorce cases in which child support orders are issued, whether the parents initiated contact with the agency or not.4 Nearly all nonmarital children are in KIDS; a comparison of nonmarital cases in KIDS with birth records (Brown and Cook 2008) found that 86% of all nonmarital children born in Wisconsin had records in KIDS.5
From KIDS, we extracted records for all children born in 1997 and identified whether they were nonmarital or marital. We then identified the parents of the nonmarital children and merged the records for all siblings and half-siblings of the initial 1997 birth cohort found in the KIDS system as of June 2008. There are 16,039 children of unmarried mothers in 1997 in KIDS for whom both parents are known. Our focus is on 8,019 nonmarital children who were their mother’s first child.6 We do not restrict our sample to births that are both parents’ first because these births represent a more select sample: every child’s mother has had a first birth, but not every child’s mother has had a first birth with a father who was also becoming a father for the first time. Although we could, instead, limit our sample to fathers’ first births (because every child’s father has a first birth), our records of men’s fertility, while good, are less complete than those for mothers. More specifically, for some children, we are uncertain of who the father is, and for others we know the father but are uncertain of whether he has fathered other children (and the father himself may also be uncertain as to whether he has fathered other children). In our data, 81.5% of the nonmarital children who were their mother’s first birth have an identified father.7 Thus, one crude estimate of the undercount in fathers’ fertility is 18.5%. Although our base sample is of mother’s firstborn children, we also show selected results for alternative samples in a sensitivity test. We discuss the sensitivity of our estimates to sample definition and other data issues below.
For the mothers and fathers of the children in our sample, we use matched data from the state administrative systems for public assistance programs to determine their participation in food stamp and public health insurance programs, and matched data from the Unemployment Insurance (UI) system to ascertain their formal earnings over the entire observation period. The data on program participation are both accurate and complete except for those parents receiving benefits out of state. UI wage records are also highly accurate measures of formal earnings, though about 10% of those with formal earnings are excluded (for a discussion of the advantages of these data relative to survey reports, see Wallace and Haveman 2007).8
These data present several advantages for this analysis. Starting with a cohort of children born in 1997 allows us to observe the multiple-partner fertility experience from each focal child’s perspective from birth through the age of 10. We are able to observe the frequency and timing of multiple-partner fertility over a longer period than most previous research, including at least 10 years of the child’s life, in addition to observing additional paternal fertility that may have occurred before the child’s birth. By using a sample of mother’s firstborn children, we are able to consider the mother’s first 10 years of fertility experience.
These data differ from the more typical survey analyses and have a number of distinct advantages and limitations. First, nearly all (86%) nonmarital children born in Wisconsin in 1997 had records in KIDS (Brown and Cook 2008), so our coverage of nonmarital births (both in forming the sample and in identifying later nonmarital fertility) is not perfect, but it is quite good. Second, subsequent marital children will usually be recorded in KIDS only if their parents divorce and there is a child support order, so we are missing some half-siblings; however, other research has found relatively low rates of marital fertility with a new partner after a first nonmarital birth, suggesting that our interests—in focal child’s half-siblings—are likely to be well covered in our data.9 A third issue is that for less than 20% of the nonmarital births, paternity is not established, so we are unable to examine the (unknown) father’s multiple-partner fertility. Our approach to this issue is to document whether our key results are sensitive to alternative assumptions about these fathers’ fertility. A final issue that arises from using state records is that births that occur after the parent has moved out of state may also be under-observed. We find, however, that 88.4% of mothers in our sample have open KIDS records, public assistance participation, or UI earnings in Wisconsin through the end of our observation period, so the rate of these mothers leaving the state appears to be quite low. Unfortunately, we have very little information on the extent to which fathers move out of state or have children with women who live outside Wisconsin.10 Note that two of the factors in which the direction of bias is known (missing marital fertility and out-of-state fertility) lead us to underestimate multiple-partner fertility. In contrast, the likely disproportionate underrepresentation in our main sample of mothers who do not apply for welfare programs nor want child support probably leads us to overestimate complexity. Notwithstanding these potential biases, alternative estimates, discussed in Online Resource 1, suggest a fairly narrow range and confirm high rates of family complexity. One final note is that our focus is on the half-siblings that result from multiple-partner fertility. We do not have good data for (nor do we incorporate) any stepsiblings that result from a child’s parent forming a union with a new partner who has had previous children; however, if the union produces new children, these half-siblings are considered if they are nonmarital or if a marital union ends in divorce.
Analytical Approach and Factors Related to Acquiring a New Half-Sibling
We follow the mother and father of each of the 1997 birth cohort children, recording any full siblings and half-siblings born over the first 10 years of the focal child’s life. We then create measures reflecting the dynamics of children’s siblingship (no sibling, or full or half-sibling[s] from the mother, the father, or both) from birth through age 10. Children may have half-siblings from a given parent’s partnership with one or more new partners, so we also consider the number of additional parenting partners and how each of the two parents’ fertility is related. We then examine economic and demographic characteristics of the child and parents and their association with the evolution of family complexity. Finally, we utilize an event history model to predict the timing of each parent having a subsequent child with a new partner. We consider separate models for the risk of having a new half-sibling on the father’s side and a new half-sibling on the mother’s side; in both models, we begin at the child’s birth and follow parents until they make the transition we are examining or until the data end (when the child turns 10 years old).
What characteristics might be related to the risk of a new half-sibling (multiple-partner fertility)? We are particularly interested in the relationship between the addition of a new half-sibling on the father’s side and a child’s risk of having a half-sibling on the mother’s side, and vice versa. If these are positively correlated, it will mean there is a risk of very high levels of complexity for some children. If the relationship between mother’s and father’s fertility persists even when we control for a variety of socioeconomic and demographic characteristics, this relationship raises the possibility that multiple-partner fertility may be mutually reinforcing. This pattern would also highlight the importance of measures of complexity that consider both parents’ fertility.
We consider two main types of control variables in examining the risk of a mother having a child with another partner: economic status and demographic characteristics. (We do not have information on the parents’ own family background, attitudes, or fertility intentions.) Based on the prior literature, we anticipate that a new half-sibling on the mother’s side will be more likely for mothers with lower economic status. We incorporate several time-varying measures of economic status, such as mother’s annual earnings and whether she was consistently employed, both measured 10 months prior to the period being considered, at about the potential time of conception for a new half-sibling. We also include whether the mother received food stamps or was covered by publicly subsidized health insurance, and the amount of formal child support the focal child’s father paid the mother. We anticipate that a new half-sibling will be more likely for those with lower earnings, less employment, or more government benefits. The relationship with child support is unclear: those who receive child support may be more economically independent and less likely to partner for economic reasons (and therefore less likely to have a child with a new partner); on the other hand, this economic support may mean that these mothers are more attractive in the partnering market and are therefore more likely to have new children (see Cancian and Meyer 2009; Gassman-Pines and Yoshikawa 2006; Gibson-Davis et al. 2005).11
We also include baseline demographic characteristics. Based on the previous literature, we expect younger women, women of color, and those who were partnered with someone of another race to be more likely to have a child with a new partner. Finally, we include child’s gender, whether the parents lived in an urban area, and parity. We also allow for the baseline hazard to vary over time.
Our model of the risk of the father having a child with a new partner (the focal child having a new half-sibling on the father’s side) is generally parallel to the mother’s model: we consider the focal mother’s multiple-partner fertility, the father’s economic status (expecting higher risk for fathers with lower earnings or employment), and the father’s demographic characteristics (expecting higher risk for younger fathers, fathers of color, and those who partnered with someone of another race). In addition, the model of the risk of half-siblings from the father includes prior half-siblings from the father (because the focal child may not have been a first child for the father) and two measures of father’s child support payments: the amount paid to the focal child’s mother and the amount paid to other mothers. These models allow us to consider how multiple economic and demographic characteristics are related to the likelihood of firstborn children having new half-siblings from their mother or father with a different partner.
Results
The Incidence and Timing of Half-Siblings
Figure 1 shows the evolving family complexity from the perspective of these firstborn children. At birth, 78% are only-children; the other 22% have half-siblings with whom they share a father. At birth, these firstborn-to-mother children by definition have no siblings or half-siblings with the same mother. Family complexity increases over the child’s first 10 years, especially during the four years from age 2 through age 5, when the proportion who are only-children or who have only full siblings falls by at least 5 percentage points in each year.
By age 10, 60% of firstborn children of unmarried mothers have a half-sibling. Twenty-three percent have half-siblings only on their father’s side, 18% have half-siblings only on their mother’s side, and 19% have half-siblings from each of their parents—that is, they share their mother (and usually their household) with children who have different fathers, and they share their father with children who have different mothers. The figure also highlights the importance of including data on both parents in order to understand children’s exposure to family complexity. For example, relying only on information on mothers’ fertility, our estimate of the proportion of children in “simple” families (that is, without half-siblings) at age 5 would be 76% rather than the 52% we see in the figure. By age 10, the percentage in simple families without data on fathers’ fertility is 63%, compared with the 40% apparent when we consider the fertility of fathers as well. Alternatively, if we had data only on fathers’ fertility, we would estimate that at age 10, 58% of children would be in simple families, compared with our baseline 40%.
While Fig. 1 shows the proportion of children with half-siblings from their mother, their father, or both, it does not distinguish the number of half-siblings. The first panel of Table 1 shows the number of half-siblings on the father’s and mother’s side. By construction, at first birth, there are no half-siblings on the mother’s side. However, 22% of children have at least one half-sibling on their father’s side, and a total of about 10% have two or more half-siblings at birth. The number of half-siblings on both sides increases over time. By the time the child is 10, a majority of children still have only one (22%) or no (40%) half-siblings, almost evenly split between the father’s side and mother’s side. In contrast, 38% have two or more half-siblings. Higher numbers of half-siblings are not uncommon: almost 1 in 4 children have three or more half-siblings, and more than 1 in 10 have four or more half-siblings.
The second panel of Table 1 shows the distribution of other partners for the focal child’s mother and father at birth and age 10. Clearly, the number of parents’ other partners can only be less than or equal to the number of half-siblings. At birth, 16% of fathers have had children with one other mother, 4% with two other mothers, and 2% with three or more other mothers. Thus, about a quarter of the children with half-siblings from their father share their father with children of at least two other mothers at birth. After 10 years, 26% of fathers have children with one other partner, 10% have children with two other partners, and 7% have children with three or more additional partners.
By construction, focal children’s mothers have no other partners beyond the child’s father at birth. However, by age 10, 28% of focal children’s mothers have had children with one other partner; 8%, with two other partners; and 1.5%, with three or more other partners. Because most children born to unmarried parents live with their mothers, half-siblings from a mother are likely to have a greater impact on a child’s living situation and daily family interactions. At the age of 10, many firstborn children of unmarried mothers could be connected to several adults through their parents’ partnering: about 15% have half-siblings from three or more adults, in addition to their own parents.
Finally, it is noteworthy that those who have half-siblings by one parent are more likely to have them by the other. Examining the distributions at age 10, we see from the first row of the first panel of Table 1 that among those who have no half-siblings on their mother’s side, less than 10% have three or more half-siblings on the father’s side (5.6/63.1). In contrast, in the bottom row, one-quarter of those who have three or more half-siblings on their mother’s side also have three or more half-siblings on their father’s side (1.3/5.2). Similarly, the second panel shows that those who have one parent with three or more additional partners are more likely to have another parent with three or more additional partners.12 In sum, children who have the most complex relationships through one parent are also more likely to have complex relationships through their other parent.
The results in Table 1 demonstrate the importance of accounting for the fertility histories of both parents. For example, consider especially complex families in which the focal child has at least two half-siblings by age 10. Our results show this level of family complexity for more than one-third of focal children (37.3%, or 100 – 40.2 – 11.3 – 11.2). Were we to rely only on data on mothers, we would estimate only 15.2% (10.0 + 5.2) of focal children to have these highly complex relationships, or less than half as many as we observe with data on both parents.
The longitudinal administrative data on which we rely exclude some births. In supplemental analyses, we estimate upper and lower bounds on family complexity, addressing the potential sensitivity of our estimates to a number of data limitations. In particular, as previously discussed, our data include only nonmarital children whose births are recorded in Wisconsin administrative records and for whom a legal father has been established, and include subsequent marital births only if the marriage dissolves and there is a child support order. As reported earlier, using our base assumptions, we find that 60% of nonmarital first births have at least one half-sibling by the time they are 10 years old. Even under a range of fairly extreme alternative assumptions, estimates of the proportion of children who have at least one half-sibling by age 10 remain between 50% and 73%, with most alternative estimates suggesting higher, rather than lower, rates of family complexity (see Online Resource 1).
Despite likely being conservative, our calculations suggest high levels of multiple-partner fertility that challenge conventional notions of more circumscribed family relationships. Most firstborn children born to unmarried parents will face a range of potential issues associated with half-sibling relationships, which may affect their biological parents’ willingness and ability to provide for them and may bring them into contact—for good or ill—with other adults who are the new partners of their parents. When resident mothers have new partners, these men may introduce both resources and threats to these children (Berger 2007). Nonresident fathers’ new partners may be less likely to have direct contact with children who are not living in their household, but may nonetheless create new challenges and opportunities.
Family Complexity for Alternative Samples
In Table 2, we consider the levels of family complexity for alternative samples. As described earlier, our base sample includes nonmarital children who were their mother’s first birth but not necessarily their father’s. The table shows selected results for three alternative samples of nonmarital children: those who were their father’s first birth but not necessarily their mother’s, those who were both parents’ first birth, and a sample of nonmarital children that is not limited to first births. Not surprisingly, restricting the sample to fathers’ first children results in lower levels of complexity on the fathers’ side and higher levels on the mothers’ side. Restricting the sample to children who are the firstborn to both parents yields lower rates of complexity, and including all nonmarital births in 1997 results in higher rates of complexity. Nonetheless, comparing the percentage in simple families at birth and age 10 shows that in each sample, the proportion still in simple families at age 10 is about half what it was at birth. Comparing the age-10 distributions, in three of the four sample definitions, half-siblings on the mother’s side are more common than on the father’s. As we have described, this may be because our measures of fertility are more complete for mothers than for fathers (in that a portion of the children in our data have known mothers but unknown fathers). Regardless of the sample, levels of family complexity are high, ranging from 50% to 64% at age 10 for firstborn children and even higher (74%) for the sample that is not limited to firstborn children.
Family Complexity by Demographic and Socioeconomic Characteristics
Children’s likelihood of experiencing family complexity varies by the demographic and socioeconomic characteristics of their parents. In Table 3, we show the age and race of the unmarried parents of mothers’ firstborn children. We also show differences in subsequent complexity. Focusing first on characteristics at the child’s birth (column 1), the mother’s first nonmarital birth generally occurs when she is young. More than half (53.3%) of all these unmarried mothers are teens at first birth, with most of the others between the ages of 20 and 25. Fathers are only slightly older: 30% are teens, and 43% are 20 to 25 years old. The third panel of Table 3 shows the relative age of parents. Almost 3 out of 4 fathers are within four years of mothers’ age, with fathers tending to be older. Few fathers (8%) are at least 10 years older than the mother.
Table 3 also shows the association between parents’ ages and new half-siblings. The younger the mother at first birth, the more likely the focal child is to have a half-sibling. For example, 63% of focal children born to teen mothers had a new half-sibling by the age of 10, compared with only 40% of those born to mothers aged 20 to 25 and 20% of those born to mothers aged 26 to 30. The patterns are similar between fathers and mothers: focal children with younger fathers are more likely to have new half-siblings than are focal children with older fathers.
Sixty percent of mothers are identified as white; 21%, as black; and 6%, as Hispanic. We are missing information on race and ethnicity for 11% of the mothers. Among those with information on race/ethnicity, fathers are less likely than mothers to be white and more likely to be black or Hispanic. However, information on father’s race is missing in 19% of the cases. The sixth panel of Table 3 shows the distribution of partnerships by race. Among those in which race is identified, there are high levels of racial/ethnic homogamy—including couples with two white (41%), two black (17%), or two Hispanic (4%) parents.13 Nonetheless, 9% of the children have a white mother and a black or Hispanic father.
Patterns of family complexity vary substantially with the race and ethnicity of the child’s parents. Children with two white or two Hispanic parents are less likely to have half-siblings over time—especially from their fathers. At age 10, almost half of these children have no new half-siblings. Most likely to have half-siblings are children born to two black parents or to a white mother and a black father.
Finally, the table shows differences between urban and rural areas, based on the county in the KIDS case record. We categorize cases into Milwaukee County (the only large urban area in the state, and thus more like the sample of nonmarital births found in the Fragile Families data); “other urban,” which includes 24 counties that are part of Metropolitan Statistical Areas; “rural” (all other counties); and “multiple” (children whose cases appear in more than one county in our time frame). Nearly half the children lived in urban counties outside Milwaukee County, with more than one-fourth of children in Milwaukee County (16.7% of Wisconsin’s overall population lived in Milwaukee County in 2007). Milwaukee County cases are about twice as likely to have new half-siblings on both sides (22%, compared with 11% to 14% for the other areas) and correspondingly less likely to have no half-siblings at age 10 (29%, compared with 42% to 46% for the other areas).
We present information on parents’ employment and earnings and mothers’ program participation in Table 4. Nearly two-thirds of the mothers were not fully employed in the year prior to pregnancy,14 and those who did work generally had low earnings—only 12% earned more than $10,000 (in 2007 dollars). Earnings rose over time, so that when the first child was age 10, 28% of mothers had earnings of $10,001 to $25,000 and 16% had earnings over $25,000 (not shown in table). Fathers’ employment and earnings are also quite low, though again there is some earnings growth over time. Only about a third of fathers were working all four quarters of the year prior to the pregnancy, and only 19% earned more than $10,000. By the time the focal child was age 10, 16% of fathers had earnings of $10,001 to $25,000 and 24% had earnings over $25,000 (not shown in table).
Children of higher-earning mothers are less likely to have new half-siblings from either the mother or the father. For example, 50% of the children whose mothers had reported earnings of $10,000 or less have new half-siblings at age 10, compared with 29% for those whose mothers earned $10,001 to $25,000 and 17% for those who earned more than $25,000. Although earnings at birth are related to mother’s age, even if we consider the final measure of mother’s earnings (10 years after the conception of focal child), children of mothers with lower earnings are more likely to have half-siblings from their mother as well as half-siblings from their father.15 When we consider fathers’ earnings, we find a similar pattern. Those whose fathers have lower earnings are much more likely to have half-siblings on either their mother’s or father’s side.
In the year prior to first birth, mothers generally did not participate in public programs (not shown), and they did not receive formal financial support from the child’s father.16 But five years later, almost a quarter received food stamps, almost half had children enrolled in public health insurance, and more than 40% had child support paid by the father. Children whose mothers either participated in a public program or received child support from that father in the year before the children were 5 years old were more likely to have new half-siblings from the mother and from the father by the time they were 10 years old.
Multivariate Analysis of Half-Sibling Risk Due to Mother’s Subsequent Fertility
The first columns of Table 5 show estimates of the risk of a half-sibling due to the child’s mother having a child with a new partner. The first rows examine the relationship between having half-siblings on the father’s side and the risk of a half-sibling on the mother’s side. Father’s prior partnerships at birth have no discernible relationship with the mother having a child with a new partner, but the focal child’s father having children with a new mother is associated with a higher risk of a half-sibling from the focal child’s mother. Thus, gaining a half-sibling on one side is related to gaining one on the other.
The next panels focus on economic status. Children whose mothers worked all four quarters of the year, or who had higher earnings, were less likely to have a half-sibling on their mother’s side. These findings are consistent with other research showing that multiple-partner fertility is more likely for lower-income parents. While food stamps and public health insurance provide resources, both programs are means-tested. Thus, it is not surprising that, even controlling for earnings, participation in these programs is associated with a higher likelihood of having a half-sibling on the mother’s side. Child support income has relationships more like those of public programs than those of earnings: more child support is associated with an increased likelihood of a half-sibling on the mother’s side.
The next panels in Table 5 examine demographic characteristics. Consistent with the bivariate descriptive statistics and prior literature, firstborn children whose mothers are younger at the time of birth are more likely to have a half-sibling from their mother. Relative to those whose fathers are about the same age as their mothers, children whose mothers are at least 5 years younger or at least 10 years older than their fathers are significantly more likely to have a half-sibling from their mother. Considering the race and ethnicity of the focal child’s parents, relative to children with two white parents, a half-sibling from the mother is more likely for any other combination except a white mother and Hispanic father, for which there was no discernible difference. Compared with those in Milwaukee County, children living in other urban and rural counties were less likely to have half-siblings from their mother. Full siblings are associated with a reduction in the likelihood of a half-sibling, a result consistent with previous research.
Multivariate Analysis of Half-Sibling Risk Due to Father’s Subsequent Fertility
The last columns in Table 5 examine the risk of having a new half-sibling on the father’s side (the father having a child with another partner). Children who are not their father’s firstborn are more likely to have additional half-siblings, especially if the father has had previous children with more than one other woman (in addition to the focal child’s mother). Similarly, half-siblings on the father’s side are more likely when there are also half-siblings on the mother’s side, especially when there are children from more than one other father. Combined with the results from the first columns, this suggests that children who have half-siblings on one side are also more likely to have them on the other, leading to substantial relationship complexity.
Similar to the results for half-siblings on the mother’s side, children whose fathers have higher earnings are less likely to have half-siblings from their fathers. If the father pays child support of $1,000 or more per year to the prior-born children, there is a reduced risk of subsequent half-siblings from the father. In contrast, any child support paid to the focal child’s mother increases the risk of a new half-sibling from the father.
As was the case on the mother’s side, the focal child is more likely to have a half-sibling on the father’s side if the father is younger. A half-sibling from the father is also more likely if the mother is younger than the father (though in contrast to the results for mothers, there is no discernible relationship for mothers who are much older than fathers). Again, children with two white parents are less likely to have a half-sibling from the father than children with either two black parents or a white mother and black father. However, in contrast to the results for mothers, the results for children with two Hispanic parents show that they are no more likely to have a half-sibling from the father, but children with a white mother and a Hispanic father are at higher risk. Similar to the results for mothers, children in Milwaukee County have the highest risk. Finally, in contrast to some results in the previous literature (e.g., Manlove et al. 2008) and the results for mothers’ new parental partnerships, there is no discernible relationship between the number of full siblings and the risk of a new half-sibling on the father’s side once we control for child support paid and the other covariates.
Discussion
Following a birth cohort of firstborn children of unmarried mothers, we document the evolution of family complexity from the perspective of the child, accounting for new partnerships that result in half-siblings on the mother’s side or the father’s side. Our results show very high levels of half-siblings that result from multiple-partner fertility. The proportion with half-siblings increases steadily over a child’s first 10 years, but especially when the child is ages 2 to 5. By the time firstborn-to-mother nonmarital children are 10 years old, 60% have at least one half-sibling. Moreover, some children’s lives are quite complicated: more than 15% are potentially connected to at least five adults who are either their parents or the parents of half-siblings.
Our results highlight the importance of having information on fertility for both fathers and mothers. Estimates of the proportion of children with half-siblings would be qualitatively lower if we were to have fertility information on only one parent. Moreover, the multivariate analyses show that new fertility on one parent’s side is associated with an increased risk of new fertility on the other parent’s side. This could reflect the ending of the parents’ relationship and the fact that both parties are then free to move on to new relationships. Other explanations are also possible, including communities in which multiple partnering is the norm or assortative mating on characteristics associated with relationship stability—in which those most likely to have children with multiple partners have children with others who have similar characteristics.
Our results confirm previous findings that multiple-partner fertility is more common when parents are younger, when they have lower incomes, when they have fewer children together, or when one or both parents are black. Our data and approach allow us to consider additional factors not considered in most previous research. We find that children whose mothers have children with more than one additional father are more likely to have fathers who have children with more than one additional mother (and vice versa)—leading some children to have very complex family structures. We find that multiple-partner fertility is substantially more common in urban than in suburban or rural areas, suggesting that analyses using data only on urban families, including the Fragile Families Study, may overestimate the level of multiple-partner fertility in the country as a whole.
Our analyses are of nonmarital births, in part because our data are more complete for these families. Children born to married couples are also at risk of acquiring half-siblings. Although our data are not as complete for children born inside marriage, and the period at risk for acquiring new half-siblings does not generally start with a child’s birth, we can use these data for rough estimates of half-siblings among the marital children who experience their parents’ divorce. These calculations show that 31% of mothers’ firstborn marital children whose parents divorce have half-siblings by age 10, compared with the 61% in our nonmarital sample. In our view, this crude calculation demonstrates high levels of family complexity for all children, marital as well as nonmarital.
One interesting finding is that formal child support is associated with an increased risk of a mother having a new parental partner. There are at least two explanations for this association that are not causal. In particular, because the receipt of formal child support is more likely when the mother is not cohabiting with the child’s father (Meyer et al. 1997; Nepomnyaschy and Garfinkel 2010), mothers receiving support are less likely to be living with the focal child’s father and therefore are more likely to have a new sexual partner. Second, while our model includes a measure of recent receipt of public benefits, it may be that more-disadvantaged mothers are both at greater risk for new partnerships and more likely to receive child support because receipt of public benefits is associated with higher child support enforcement. Some prior literature (e.g., Wallace 2007) has suggested that mother’s fertility is relatively insensitive to child support receipt. That said, the positive relationship between child support receipt and new partners is consistent with child support income making the mother a more attractive partner or with mothers being more willing to try a new relationship given that child support may result if the relationship breaks down, as others have noted (Curtis and Waldfogel 2008).
Paying child support to the child’s mother is also associated with an increased risk of a father having a new parental partner. As noted earlier, this may simply reflect that fathers paying support are not cohabiting with the focal child’s mother, increasing their risk of new partnerships. However, paying support to a previous partner is associated with a decreased risk of new parenting partnerships. Fathers in this situation may be more aware of the financial consequences of additional children and therefore may be less likely to partner with another woman; this would be consistent with some other research suggesting that father’s fertility is sensitive to child support enforcement (Aizer and McLanahan 2006; Garfinkel et al. 2003; Plotnick et al. 2004, 2007). An alternative possibility is that, all else equal, fathers who pay support to their most recent partner (focal child’s mother) may be more attractive partners to other women, increasing the risk of a new birth and a new half-sibling for the child.
The extent to which multiple-partner fertility is a problem for the child support system depends partly on its frequency (Cancian and Meyer 2006; Meyer et al. 2005). These results demonstrate that half-siblings are quite common and illustrate the dynamic nature of children’s siblingships and, therefore, of resident and nonresident parents’ obligations and sources of support. The timing of these changes raises serious issues for child support policy.
There is no consensus regarding the optimal approach to setting child support orders for complex families (e.g., Cancian and Meyer 2006; Meyer et al. forthcoming). Current policy in most states (and some developed countries) treats each couple individually, taking into account the obligor’s previous obligations in setting new orders, but leaving the previous order unchanged.17 Some policy alternatives call for a child support order to be adjusted downward whenever a noncustodial parent has a new obligation, in order to treat all children equally and keep the total support owed manageable.18 Other approaches call for reductions in orders whenever a custodial parent has a new child, so that mothers receive the same support for a given number of children, whether they all have the same father or have different fathers.19 If the child support order system included provisions like these that allow for an adjustment every time one of the parents had another child, adjustments for this reason would be fairly common. Among the cohort of firstborn nonmarital children of 1997, almost half the fathers had other children by the time the focal child was 10 years old. Similarly, in nearly 40% of these cases, a custodial mother had a child with another person. Thus, any system that allows for adjustments whenever there is a new birth could result in frequent revisions.
Our results highlight the frequency and evolution of family complexity. The implications for policy and research are potentially profound. Policies that were designed for simple families, with parents who had children only with one another, often are not well adapted to complex families. This raises issues not only for child support policy and marriage promotion policy but also for tax and income support policies for which family structure is important in determining eligibility. The prevalence of multiple-partner fertility and the resulting complex relationships within and across families also raise important questions for research and measurement. As we have shown, complex families look different when viewed from the perspective of parents and children, and measures that account for both mothers’ and fathers’ partnership also differ markedly from those that rely on only one parent’s history. New data sources that are less tied to a given household and less sensitive to living arrangements are necessary. Although administrative data have other limitations, we show here that they can provide important insights into the evolution of complex families.
Acknowledgments
The research reported here was supported in part by the Wisconsin Department of Children and Families and the Institute for Research on Poverty; any views expressed are those of the authors and not necessarily those of the sponsoring institutions. A previous version of this article was presented at the 2009 annual meeting of the Population Association of America, and at the 2009 annual meeting of the Association for Public Policy Analysis and Management. The authors thank meeting participants and three anonymous referees for helpful comments, Lynn Wimer for expert assistance with data construction, Eunhee Han for research assistance, and Deborah Johnson and Dawn Duren for assistance with the preparation of this manuscript.
Notes
Moreover, as we discuss later, our data are more complete for nonmarital than marital births.
Stewart (2002) provides a relatively recent overview of the literature on stepfamily fertility. While new fertility within stepfamilies is, by definition, multiple-partner fertility, we are interested in a different (and broader) construct, since we are concerned about all half-siblings, not just those born into coresidential unions.
Wisconsin’s fertility rate in 2007 was 65.0; the U.S. rate was 69.5. In Wisconsin, 35.4% of births were to unmarried women; the U.S. rate was 39.7% (Hamilton et al. 2010). Thus, Wisconsin’s fertility rate is within 10% of the national rate, but the proportion of births to unmarried women is somewhat lower than most other states. Wisconsin also has fewer people of color than most other states: in 2007, 6.0% of Wisconsin’s population was African American (alone), and 4.9% was of Hispanic origin; percentages in the United States as a whole are 12.8% and 15.1%, respectively (U.S. Census Bureau 2009b). In 2006, Wisconsin’s median family income was 3.6% higher than the nation as a whole, and in 2000, its population was more rural (31.7%, compared with 21.0% for the United States; U.S. Census Bureau 2009b).
Data from Wisconsin’s court records suggest that about three-quarters of divorce cases with children have child support orders within a year of the final divorce decree (authors’ calculations). National data show that in 2008, about 70% of divorced custodial mothers reported that they had a child support award (U.S. Census Bureau 2009a).
In part, the high coverage is because many nonmarital births are to low-income mothers who eventually apply for some welfare benefit. Also, nonmarital relationships are frequently unstable, and many resident parents seek child support orders regardless of their own income level. Finally, a portion of the funding states receive from the federal government to administer their child support agencies is based on performance indicators; therefore, states have a financial interest in helping custodial mothers establish paternity and establish child support orders.
We exclude children whose parents had more than one child together in 1997 (N = 170)—mostly twins or other multiple births. The 8,019 children are associated with 8,019 mothers. Because 100 fathers had two children who were the mother’s firstborn child in 1997, and two fathers had three such children, there are 7,915 unique fathers in our sample.
We examine records through December 31, 2007, when the children were 10 years old, to see if paternity was legally established. Prior research shows that few paternities are established after the child is 7 years old (Brown and Cook 2008).
Those with no UI earnings include parents with no actual earnings and those whose earnings are not reported to the state UI office. Unreported earnings include earnings from out of state, the federal government, and self-employment. Although formal earnings are an incomplete record of economic resources, they are the primary measure used by the child support system.
Estimates of the extent to which those who have a nonmarital birth go on to have a birth inside marriage for recent cohorts are quite low, at less than 5% (Bzostek et al. 2007; Guzzo and Furstenberg 2007b), although these estimates are based on data that follow mothers for a relatively short period. Among women born in the 1950s and early 1960s, about one-quarter of those with a nonmarital first birth went on to have a second marital birth (Aassve 2003; Wu et al. 2001). However, some of the subsequent marital births were with the same father (and thus the child has a full sibling, not a half-sibling). Finally, we note that any marital births that follow a nonmarital birth will be captured in our data if the marriage ends before the focal child is 10 years old.
KIDS does record a mailing address for many fathers, but these are not consistently updated and their accuracy is unknown. Another issue related to fathers’ residence is that we do not have information on whether fathers have been (or are currently) in prison. In our view, prison information would be beneficial for predicting the risk of gaining a new half-sibling, but the lack of this information does not introduce any particular biases into our counts of multiple-partner fertility.
We have information only on formal child support. Although informal child support is more common and of higher value in the first year of a nonmarital child’s life, by their third year, fewer than 30% receive informal support, and it averages less than $500 per year (Nepomnyaschy 2007).
Of those whose mother has not had other partners by the time the focal child is age 10, less than 5% have fathers who have had three or more partners (2.9/63.1). In contrast, 13% of those whose mothers have three or more partners have fathers with three or more partners (0.2/1.5).
The table shows separate results only for the more common groups. Both parents are Native Americans in 1.0% of the cases, and both are Asian/Pacific Islander in 0.7% of the cases. Including these cases, racially homogamous couples account for 63% of all couples and 83% of couples for whom we have complete information on race and ethnicity.
We measure employment status and earnings in the four quarters prior to pregnancy. In particular, for a child born between January and March of 1997, we assume that the pregnancy began between April and June of 1996 and measure employment in the last three quarters of 1995 and the first quarter of 1996 (i.e., April 1995 to March 1996). Note that “fully employed” means having some earnings in each quarter, and thus we miss spells of unemployment unless they last at least a calendar quarter.
Of course, measures of mothers’ later earnings may, in part, reflect the consequences of additional births.
Few mothers are eligible for programs other than food stamps in the year prior to pregnancy. Many low-income pregnant women would be eligible for public health insurance coverage, but our lagged measure pre-dates pregnancy. Some young mothers may have qualified as children in another qualifying household; such coverage is not reflected here. Figures for 10 months prior to the child’s birth show participation rates of about 1% for both programs.
For example, in Wisconsin, a father who owes 17% of his income for a prior-born child should be ordered to pay 14.1% of his income (17% of the income remaining after paying the first order) for a subsequent child born to another mother. In contrast, a mother’s prior children, and any support owed to her from the father of those children, should not affect the support ordered for a subsequent child with a new father.
For example, rather than owe 17% of income for the first child and 14.1% for the second (see previous footnote), the first and second order amounts could be averaged so both children received the same support, requiring a downward revision to the first order and resulting in a higher second order. This type of policy was simulated in a recent article by Sinkewicz and Garfinkel (2009) and is the policy of some other countries (Meyer et al. forthcoming).
For example, a mother currently should receive 17% of the nonresident father’s income for one child or 25% for two children with the same father. However, if a mother with one child has a second child with a new father, she should receive 17% of the second father’s income for the second child. An alternative would be to reduce the first father’s order to 12.5%, and set the second father’s order at 12.5%. If both fathers have the same income, the mother would then be due the same support for two children, whether they had one or two fathers.
