1 Introduction

Parental separation and divorce have become increasingly common experiences for young children in the UK. Between 1970 and 2007, for instance, the probability of a child experiencing parental union dissolution by age five roughly doubled, from 10 to 19% (Kiernan 2004; Panico et al. 2010). Furthermore, in 2001, just under 20% of children were born to an unmarried mother who was not cohabiting with the father (Panico et al. 2010). Repartnering subsequent to union dissolution is also common, with 70% of previously cohabiting and 43% of previously married adults finding a new partner within 5 years in the UK in the 1990s (Ermisch 2002). Thus, many children born to a single parent or who experience parental separation may see their mother form a new union.

Prior research has examined the likelihood of repartnering among mothers with coresident children as a function of mothers’ demographic and socioeconomic characteristics (de Graaf and Kalmijn 2003; Ermisch 2002; Pevalin and Ermisch 2004; Rowlingson and McKay 1998; Jansen et al. 2009; Turunen 2011; Theunis et al. 2015). However, few studies have examined whether the extent to which nonresident fathers are involved in children’s lives may influence maternal repartnering. The scarcity of studies partly reflects that information about nonresident father involvement has often been lacking in existing data sources (Bracke and Schoors 2013) or is limited to legal arrangements, which may not reflect “de facto” arrangements.

This study adds to a small but growing literature on the determinants of maternal repartnering by examining whether nonresident father involvement, both in terms of fathers’ de facto physical contact with children and child support (maintenance) payment, are associated with subsequent maternal repartnering. Our data were drawn from the UK Millennium Cohort Study (MCS), a representative, longitudinal study following approximately 19,000 children born between 2000 and 2002. This large sample size means that, relative to other studies, we can observe a large number of parental separations involving young children, as well as repartnering trajectories for parents. The MCS includes rich information on father contact and child support receipt. We employed discrete time event history models to estimate associations of the frequency of contact between children and their nonresident biological fathers, as well as the regularity of child support payments by these fathers, with the probability that single mothers repartnered into marriage or cohabitation during the first 7 years of a child’s life. We focus on single mothers with young children because having a young child increases the likelihood that father contact with the child necessitates involvement with the mother. For older children, visitation can take place without the mother being involved: fathers and children can directly arrange visits. Because parents with more formalized relationships may be more likely to maintain ties after breaking up, we also examined these associations separately for those who were married or cohabiting with the biological father at the time of the child’s birth, as well as for mothers who were not in a coresidential relationship at the time of the child’s birth. Given that economic constraints are relevant to the repartnering process, we pay particular attention to the socioeconomic background of mothers, and explore whether processes are heterogeneous across different socioeconomic groups.

2 Conceptual Framework and Prior Research

Maternal repartnering may have potential benefits for mothers and their young children. First, parental union dissolution is frequently accompanied by substantial declines in real income for mothers (Jarvis and Jenkins 1999); such declines appear to have increased in recent decades (Uunk 2004) and are associated with decreased child well-being (Amato 2005; Schoon et al. 2012). Repartnering may alleviate such adverse economic conditions (Dewilde and Uunk 2008; Shafer and James 2013), particularly for economically disadvantaged mothers (Bzostek et al. 2012; Shafer and Jensen 2013) and those who receive little or no child support.

In addition to increased economic resources, mothers who repartner tend to experience higher levels of well-being than those who remain single (Osborne et al. 2012), despite the fact that repartnering only partially makes up for the negative effects of union dissolution on maternal well-being (Hughes and Waite 2009). Each of these factors may be beneficial for child well-being to the extent that they are associated with higher-quality parenting and home environments.

Nonresident father involvement may influence the likelihood that a mother repartners by potentially influencing three main drivers of repartnering: opportunity, attractiveness, and need (Turunen 2011). First, father contact and child support payment are likely to influence the amount of time a mother spends in childrearing activities. A more equal share of child rearing responsibilities between parents, including the financial costs thereof, can increase the probability of the mother finding a new partner for several reasons (Goldscheider and Sassler 2006). To begin with, caring for children is time consuming. Therefore, greater nonresident father involvement should increase the time available to the mother—her opportunity—to meet potential partners and build new relationships (Theunis et al. 2015; Ivanova et al. 2013). In addition, mothers with primary custody or childrearing responsibility may be penalized on the repartnering market and be less attractive to potential partners, relative to mothers with shared custody. The physical custody (residential) status of children has indeed been shown to play a role in maternal repartnering (Ivanova et al. 2013). Likewise, mothers who receive no or little financial support from their children’s fathers may both have less time to seek and sustain prospective partnerships (e.g., because they must spend more time earning income) and be less attractive to potential partners, because of the prospect of having to support these children.

Moreover, the de facto frequency of contact a child has with their nonresident father—regardless of the legal physical custody arrangement—may be a particularly salient indicator of both an ongoing tie between former partners and of the amount of time each parent spends childrearing. Higher levels of father contact may signal stronger ties to an ex-partner as well as greater time spent arranging and organizing children’s living arrangements and visitation schedules (particularly when a child is young). These factors might further reduce both mothers’ opportunities to repartner and their attractiveness to potential partners. In addition, when the nonresident father is not (or is less) involved or contributes no (or less) child support, mothers may be more motivated to seek a new partner not only for themselves, but also to provide a father figure for their child(ren), and/or to ease financial constraints. These factors may influence how much a mother prioritizes repartnering. Finally, the repartnering process may vary by marital and cohabitation status at the time of the birth. Specifically, parents with more formalized and institutionalized relationships may be more likely to maintain ties after breaking up—especially if they were married or if they and their joint child(ren) lived together. The repartnering process may also vary across different socioeconomic groups due to differences in financial constraints and social resources available. We consider each of these factors in more detail below.

2.1 Nonresident father involvement

Mother-primary physical custody continues to be the predominant physical custody arrangement in the advanced industrialized countries, although (equal) shared custody is becoming increasingly common in the UK (Peacey and Hunt 2009) and elsewhere (Bauserman 2002; Bender 1994; Berger et al. 2008; Cancian et al. 2014). Existing studies of shared physical custody have focused on the consistency of physical custody orders with children’s physical custody arrangements and the stability of those arrangements (Berger et al. 2008; Cloutier and Jacques 1998; Kline et al. 1989; Maccoby and Mnookin 1992; Pearson and Thoennes 1990), as well as on child well-being (Bauserman 2002; Spruijt and Duindam 2009). Whereas most prior studies have focused on legal physical custody arrangements, it is particularly important to recognize that the amount of time allocated to each parent in a physical custody order does not necessarily indicate the amount of time a child actually spends in the care of each parent. Thus, we prefer to focus on associations of the de facto amount of contact a child had with his or her father regardless of the formal arrangement in place. We refer to de facto father–child contact as the level of father involvement and analyze the association with the mother repartnering chances.

Our a priori expectations regarding the direction of these associations are ambiguous. As described above, more frequent father–child contact may reduce maternal time devoted to childrearing, thereby reducing time-related barriers to social interaction and integration (Schmiege et al. 2001) and repartnering (de Graaf and Kalmijn 2003). We would then expect a positive association between father involvement and maternal repartnering. Conversely, however, greater nonresident father involvement may be associated with reduced maternal repartnering because: (1) nonresident fathers may serve as gatekeepers to the extent that they can disrupt mothers’ new relationships either through explicit actions toward mothers and their potential partners or, indirectly, by influencing the transaction costs of scheduling time-sharing and consistency therein; (2) when fathers are more heavily involved, mothers may avoid disrupting a working coparenting relationship and/or risking the father spending less time with or making fewer investments in the child(ren), and therefore expend less effort toward (or avoid) repartnering; (3) the process of negotiating time-sharing arrangements with a nonresident father may be stressful and time consuming, thereby impeding relationship formation and progression of a new relationship; (4) mothers may have less need for a new partner to serve as a father figure or share childrearing responsibilities when the nonresident father is heavily involved; and (5) high levels of nonresident father involvement may suggest that there are unresolved issues regarding the former relationship (Shafer et al. 2013) and therefore be perceived as a threat by potential partners who may be reluctant to invest in a relationship with a woman whose former partner is still part of her life.

The few recent studies examining associations of nonresident father involvement with maternal repartnering have produced inconsistent results. Schnor et al. (2015), using Belgian data, found that the likelihood of maternal reparterning is lower when mothers have primary physical custody of children. However, they focused only on custody arrangements and not on the amount of time nonresident fathers actually spent with children. Furthermore, they examined repartnering only for previously married women who had divorced, rather than for all single mothers. McNamee et al. (2014), using US data, found that nonresident father contact is positively associated with maternal remarriage. However, they focused only on remarriage and not on repartnering (into marriage or cohabitation) more generally. This is important considering that repartnering into cohabitation is increasingly common. In contrast to these studies, Meggiolaro and Ongaro (2015), using cross-sectional Italian data, found mean differences suggesting that greater father contact (either in person or by phone) is associated with reduced maternal repartnering. Unfortunately, however, they were unable to test this association in their multi-variable models because of sample size limitations. Our analyses use a large recent UK birth cohort and consider all forms of maternal repartnering and previous parental relationship status (noncoresident, cohabiting, married).

2.2 Child Support Receipt

Our a priori expectations regarding associations between child support receipt and maternal repartnering are ambiguous. Given that, on average, single mothers are socioeconomically disadvantaged, they should have an economic incentive to repartner (Shafer and James 2013), particularly if they are receiving little, inconsistent, or no child support. As noted above, repartnering may assist mothers in overcoming the adverse financial consequences associated with union dissolution and single parenthood (Dewilde and Uunk 2008; Shafer and James 2013; Shafer and Jensen 2013). Indeed, despite that concerns about the financial well-being of single-mother families have led to public policies in many countries to provide public support, single-mother families remain at a high risk of poverty. As such, we might expect consistent child support receipt, by reducing financial need (and uncertainty), to be negatively associated with maternal repartnering. At the same time, women who are receiving child support may be more attractive to potential partners than those who are not. That is, men may be skeptical of partnering with a woman if they believe they will have to financially support her children. If so, child support receipt should be positively associated with maternal repartnering.

Empirical results linking child support receipt and maternal repartnering have also been mixed. Whereas there is some evidence that receiving public (sometimes linked to single parenthood status) or private (child support) transfers is associated with a reduced likelihood of repartnering (de Graaf and Kalmijn 2003), other work has found no association of receiving financial support from the nonresident father with maternal remarriage among divorced women (McNamee et al. 2014). We add to this literature by estimating associations of the regularity of child support received with the likelihood of maternal repartnering.

2.3 Relationship Status and Socioeconomic Background

Children are more likely to be brought up by single parents in Britain than in any other major country in Europe (Ruggeri and Bird 2014), and single parents are disproportionately disadvantaged (Culliney et al. 2014). Entry into single parenthood can happen through two paths: mothers may be single when they give birth or they may have a child within a cohabiting or marital relationship, which subsequently dissolves (Berrington and McGowan 2014). Single parenthood from birth is an important phenomenon in the UK (characterizing 20% of children born in 2000–2001), particularly among young and disadvantaged women (Panico et al. 2010).

Prior research has highlighted the importance of distinguishing mothers’ pathways into single parenthood when studying maternal repartnering (Le Bourdais et al. 1995). Parental ties are likely to be weakest if parents were not cohabitating or married at the birth, and strongest if they were married at the birth or shortly thereafter. These initial levels of commitment, investment, and legal and social ties are likely to be associated with future parental behaviors (Cherlin 2004; Nock 1995), including investments in children and repartnering. Marriage involves a formal legal and social commitment to jointly invest in a family unit, including children (England and Farkas 1986; Furstenberg and Cherlin 1991; Townsend 2002). Cohabitation implies weaker ties and lesser joint commitment than marriage, but greater ties and shared commitment than that of parents who never lived together or were not in a serious relationship. Also, fathers who spent time living with their children may have developed stronger attachments to them and, therefore, may engage in greater future involvement or provision of child support than those who did not. On the whole then, we may expect nonresident father investments to be highest when they were formerly married to the mother and lowest when they have never cohabited with the child. As such, we conducted all of our analyses separately for mothers who were not in a coresident relationship, were cohabiting, or were married at the time of the focal child’s birth.

Parental socioeconomic background might also play a role in repartnering. Turunen’s (2011) theoretical framework regarding the drivers of repartnering (attractiveness, opportunity, and need) suggests that mothers with fewer socioeconomic resources might have a higher perceived “need” to repartner in order to improve their living standards and benefit from economies of scale associated with coresidence (Zabel 2012). Conversely, more advantaged mothers have less economic need to repartner but may be more attractive to potential partners. We examine such potential heterogeneity by estimating separate models by poverty status and maternal education at the time of first interview (when the focal child was approximately 9 months of age).

2.4 Social Selection and Reverse Causality

Associations of nonresident father involvement and child support payment with maternal repartnering may be partially or fully driven by social selection and/or reflect reverse causality. With regard to social selection, other characteristics and circumstances of mothers and fathers may jointly determine nonresident father involvement, child support payment, and the probability that a mother repartners, such that links between them are spurious. To adjust for this possibility, we controlled for a host of observable characteristics that are likely associated with each. These included child age and sex, the mother’s family structure as an adolescent, maternal age at the child’s birth, the mother’s education, whether the mother had a disability (limiting physical or mental health condition), maternal employment, the biological relationships among children in the household, whether a grandparent was present in the household, and household income (de Graaf and Kalmijn 2003; Ermisch 2002; Pevalin and Ermisch; 2004; Rowlingson and McKay 1998). Turning to reverse causality, prior research (Berger et al. 2012; Tach et al. 2010) suggests that the negative association between maternal repartnering and father involvement could reflect the influence of maternal repartnering on father behaviors rather than vice versa. To address this, we used lagged indicators of father involvement to predict subsequent maternal repartnering, and, in supplemental analyses, tested a cross-lagged structural equation model.

3 Empirical Analysis

3.1 Data

Our data were drawn from the MCS, a nationally representative prospective birth cohort study in the UK (Dex and Joshi 2005). The total sample includes 19,517 children in 19,244 households living in the UK. The study mainly consisted of interviews with the primary caregiver (the mother in 98% of cases in wave 1), and a separate interview with the primary caregiver’s resident partner (if applicable). Respondents were asked about multiple dimensions of family circumstances, socioeconomic and demographic characteristics, and, if applicable, nonresident parents’ involvement with the focal child and provision of child support. Data for our analyses were collected when focal children were about 9 months of age, as well as 3, 5, and 7 years of age. We organized our data in child-wave observations. Our analyses focused on waves 2 through 4, when children were aged 3–7, with lagged information drawn from wave 1 (when children were approximately 9 months old). A child entered our analysis sample in the first wave in which his or her mother was neither coresident (cohabiting or married) with, nor in a noncoresident romantic relationship with, the child’s biological father. Our analysis sampleFootnote 1 consisted of 8745 child-wave observations of 4568 children.

Children born to mothers who were neither married to nor cohabiting with their father at the time of their birth are overrepresented in our analysis sample because these mothers are disproportionately likely to be observed at a greater number of interview waves given our sample inclusion criteria. Indeed, whereas nearly 60% of children born to single (noncohabiting and unmarried) mothers met our sample inclusion criteria at one or more interview waves, this was true of only 26% of children born to cohabiting parents and 11% of children born to married parents. As such, whereas children born to married, cohabiting, and single mothers account for about 57, 23, and 20% of children in the full MCS sample, respectively, they account for 26, 25, and 49% of the children in our analysis sample.

Missing data in our analysis sample were relatively limited on most variables (2–5%), with the greatest amount of missing data for household income (10%). Moreover, the proportion of cases with missing values (see Table 1) was similar for mothers who did and did not repartner, with the exception of two variables: child race/ethnicity (5 vs. 1% missing) and whether the mother had a limiting condition (2 vs. 1% missing). As such, we replaced missing data with either the mean (for continuous variables) or zero (for dichotomous variables), and we included in our regression models an additional missing data indicator for initially missing values for each relevant variable. We estimated supplemental models in which list-wise deletion was used to construct the analysis sample and results did not differ.

3.2 Measures

3.2.1 Maternal Repartnering

Our key outcome variable was an indicator (1 = yes) of whether the child’s mother had repartnered (for the first time) in a given wave. This variable was constructed using the information collected in the household roster, which lists all resident members of the household at the time of the interview, as well as through questions posed to the mother regarding whether she had a cohabiting partner and, if so, his relationship to both her and the focal child. We defined maternal repartnering to have occurred when a mother formed a new coresidential relationship by living with a partner. We did not distinguish between married or unmarried coresidential relationships as most new relationships in our sample began with a period of nonmarital cohabitation. Like most previous studies, we do not include noncoresidential relationships in our operationalization of repartnering because noncoresident couples are less likely to share resources such as income and childcare. However, we estimated an alternative specification of our models in which we included noncoresident couples (mothers who report being in a new romantic, but not cohabitating, relationship), often referred to as couples who are living apart but are together (LAT), in our definition of repartnering.

3.2.2 Father Involvement and Child Support

Our key predictors were lagged measures of mothers’ reports of the frequency of contact between the biological father and the focal child, as well as the regularity of child support payments received from the biological father. The father contact measure was drawn from a single item in which mothers were asked “How often does [the absent parent] see [the cohort child]?” This represents de facto father contact regardless of the parents’ physical custody arrangement. The seven response categories for this item included every day, 5–6 times a week, 3–4 times a week, once or twice a week, less often but at least once a month, less often than once a month, and never. We present descriptive statistics for each of these categories in Table 1. However, as cell sizes for several of the categories were small, we collapsed the original seven father involvement categories into fewer categories for our regression analyses.

We operationalized father contact to include no father contact, father contact between once per month and twice per week, father contact 3–4 times per week, and father contact five or more times per week. An indicator for missing father contact data (about 4% of the sample) is also included. This categorization roughly approximates de facto physical custody arrangements of mother sole, mother primary, equal shared, and father primary. We also tested three alternative coding schemes.Footnote 2

We used a lagged measure of child support receipt, drawn from a single item in which mothers were asked “Does [the absent parent] make regular payments?” Response categories included “makes regular payments,” “makes irregular payments,” and “makes no payments”. In addition to these categories, we also included an indicator that child support payment information was missing (9% of the analysis sample).

Finally, a substantial proportion of mothers both dissolved their relationship with the focal child’s biological father and repartnered within the same 2-year period between data collection waves. Lagged father involvement and child support data were therefore unavailable for these mothers because they were still living with (or, in some cases, romantically involved with) the biological father at the prior interview. We thus controlled for whether the father was in the household at the prior interview wave in all of our models. As a robustness check, we also estimated models in which these cases were excluded from the sample. However, we retain this relatively large groupFootnote 3 in our primary analyses because these families experienced the most recent family structure transitions and the fastest maternal repartnerings following union dissolution with the biological father. Excluding them would considerably bias our analysis sample.

3.2.3 Covariates

We controlled for a host of child characteristics, mother characteristics, and household characteristics that may be associated with father contact, child support payment, and maternal repartnering. Child characteristics included the child’s sex, age in months at the time of each interview, and whether the child was white (versus of another race/ethnicity). Mother characteristics included the mother’s age at the birth of the focal child, whether her parents had divorced or separated by the time she was 15, whether she reported having a limiting (physical or mental health) condition at the 9-month interview, and a lagged indicator (1 = yes) that she was employed at the time of each interview, as well her level of educational attainment. Categories for educational attainment were less than level three National Vocational Qualifications (NVQ 3), NVQ 3, and more than NVQ 3.Footnote 4 Household characteristics included lagged indicators that full, half, and step siblings of the focal child resided in the household, a lagged indicator that a grandparent resided in the household, and a continuous measure of lagged equivalized total household income, using the OECD modified equivalence scale (Hagenaars et al. 1994) to adjust for household size and composition.

3.3 Analytic Approach

We used a series of discrete time event history models to estimate associations of father involvement and child support receipt with maternal repartnering. Specifically, we estimated pooled logistic regressions in which an indicator that the mother repartnered (for the first time) in the current period was regressed on lagged father contact, lagged child support payment, and the control variables. Standard errors were corrected for intracluster correlation due to multiple observations per mother. We estimated these models separately by family structure at birth (single mothers, cohabiting parents, married parents). The reference groups in these models were no father contact and no child support payments. However, we also tested the statistical equivalence of each pair of father contact and child support payment coefficients. Finally, for each family type, we conducted subgroup analyses by household poverty status at the time of the birth and maternal education (measured at the 9-month interview).

4 Results

4.1 Descriptive Statistics

Descriptive statistics by repartnering status are presented in Table 1. Just over a quarter of the mothers in our analysis sample repartnered during the observation period. Fathers had daily contact with children in about 7% of these families, weekly contact in 26%, monthly contact in 9%, less than monthly (but some) contact in 4%, and no contact at all in 27%. Mothers whose children had daily contact with their father were much less likely to repartner than those whose children had less contact with their father. Other differences in father–child contact by subsequent maternal repartnering were relatively small. Approximately 38% of mothers received regular child support payments from the nonresident father, about 6% received irregular payments, and 20% received no payments. Mothers who received regular child support payments were less likely to repartner than those who did not.

Table 1 Descriptive statistics by maternal repartnering
Full size table

The father having been present in the household at the previous interview was associated with a greater likelihood of maternal repartnering.Footnote 5 This may reflect a greater propensity for some mothers to repartner in the first few years after union dissolution or that these mothers have unobserved characteristics that put them in an advantageous position on the repartnering market. Furthermore, some may have exited their relationship with the focal child’s father either because they met someone new or hoped to find a new partner.

Beyond father involvement and child support receipt, there were also differences in many of the covariates by maternal repartnering status. For example, mothers who were cohabiting at the focal child’s birth were disproportionately likely to repartner, whereas those who were single at the birth were more likely to remain single. White mothers, younger mothers, more highly educated mothers, and mothers with fewer children were also more likely to repartner.

Table 2 shows a number of significant differences in both father contact and child support payment between family types at the focal child’s birth. First, mothers who were married and, in particular, those who were cohabiting at the focal child’s birth were more likely to repartner than those who were single. Second, there was more variation in father contact for children whose mothers were single at their birth than for those whose mothers were married. That is, children whose mothers were single at their birth were both most likely to have daily father contact and most likely to have no father contact. This still holds when we select only the sample of children not living with their father at the prior wave. Third, very frequent father contact tended to be negatively associated with maternal repartnering for mothers who were single or cohabiting at the birth, but not for mothers who were married at the birth. Finally, mothers who were married at the birth and received no child support were considerably more likely to repartner than those who were married at the birth and received child support, whereas this was not the case for mothers who were single or cohabiting at the birth. These patterns further support our decision to estimate our regression models separately by family type at birth.

Table 2 Descriptive statistics for maternal repartnering and father involvement by family type and maternal repartnering
Full size table

4.2 Discrete Time Event History Results

Table 3 presents results from our discrete time event history analyses. For children whose mothers were single at their birth, those who had very frequent father contact (5 or more times per week) were about 70% less likely to experience maternal repartnering than those who had no father contact. They were also 15% less likely to experience maternal repartnering than those who had father contact 3–4 times per week, 34% less likely than those who had father contact ranging from less than once a month to 2 times per week. These are relatively large differences given that the repartnering rate in a given wave was about 8% for single mothers at birth, and that the repartnering rate over the observation period (during first 7 years of the focal child’s life) was about 24%. We found no associations between father involvement and maternal repartnering for children whose mothers were cohabiting or married at the time of their birth. We also found no associations between child support payment and maternal repartnering for any groups. Results for the covariates suggest that white children were considerably more likely to experience maternal repartnering, and that maternal repartnering tended to be more common for younger and more highly educated mothers, as well as those with fewer coresident children.

Table 3 Discrete time analyses of the likelihood of maternal repartnering
Full size table

4.3 Socioeconomic Subgroup Analyses

Table 4 presents results from two sets of subgroup analyses. First, we estimated separate models for children whose families were below the poverty line (based on having an income of 60% or less than the OECD equivalized-scale median income) at the time of the 9-month interview and those whose families were not. On the whole, we found that the negative association between very frequent father contact and maternal repartnering among mothers who were single at the focal child’s birth was relatively consistent for both groups, although it was somewhat smaller in magnitude among poor families.

Table 4 Discrete time analyses of the likelihood of maternal repartnering, subgroup analyses using collapsed specification 1
Full size table

Second, we estimated separate models by maternal educational attainment. Results suggest that, for mothers who were single at the focal child’s birth, the negative association between very frequent father contact and repartnering was relatively consistent among the low and high education groups. In addition, we found an extremely large positive association between very frequent father involvement and maternal repartnering for more educated mothers who were married at the focal child’s birth. These mothers were five times more likely to repartner than their counterparts whose children had no father contact. For households in which the biological father sees the child less frequently (3 or 4 times a week, or 1 or 2 times a week), mothers are also about two times more likely to repartner.

4.4 Robustness Checks

We tested the robustness of our primary results to several alternative model specifications. First, we estimated regressions using three alternative coding schemes for father contact. Each of the alternative coding schemes for father involvement (coding father contact in the full set of eight categories initially collected and presented in Tables 1, 2; coding high-frequency contact as three or more times per week; and coding low-frequency contact as less than once per week) yielded substantively consistent results as those of our primary coding scheme, again suggesting that high-frequency father contact is negatively associated with maternal repartnering among mothers who were single at the focal child’s birth, but not among other mothers (see Appendix Table 5A–C). Of particular note, in models using the full range of father contact categories, we found no significant difference between the association of daily father contact and that of father contact 5–6 times per week with maternal repartnering and no significant difference between the association of father contact 5–6 times per week and that of father contact 3–4 times per week with maternal repartnering. Thus, our findings do not appear to be unduly influenced by our coding scheme choice.

Second, we estimated regressions using an alternative definition for maternal repartnering, which defined mothers who had a noncoresident partner (were in a LAT relationship) as having repartnered. These results were substantially and interestingly different from our primary results (see Appendix Table 5D). Here, we found that mothers who were single at the focal child’s birth and whose children had contact with their father two times per week or less were more likely to repartner than those whose children had no contact with their fathers, as well as those whose children had greater levels of contact with their fathers. We found no association between highly frequent father contact and maternal repartnering. Taken together with our primary specification results, this suggests that, whereas very frequent father contact is associated with a decreased likelihood that a mother forms a new coresident (marital or cohabiting) union, it is not associated with a decrease in forming a noncoresident union. We also find that, among children whose mothers were cohabiting at their birth, those who had contact with their father two times per week or less were more likely to experience maternal repartnering (including LAT) than those who had no contact with their father. These results suggest that the association between father contact and maternal repartnering differs with respect to coresident and noncoresident maternal repartnerships.Footnote 6

Third, we estimated supplemental models (available upon request) using complementary log–log regressions instead of standard logit regressions. Whereas the standard logit model is symmetrical (the probability of experiencing an event is assumed to be of equal magnitude but opposite sign as that of not experiencing the event), the complementary log–log model is asymmetrical such that the probabilities of experiencing and not experiencing an event may differ. We also estimated Cox proportional hazards models in which the outcome variable was the number of days between the biological parents’ separation and the mother’s repartnering (censored at the interview date for mothers who had not repartnered). Consistent with our primary results, both analyses indicated that very frequent father involvement is associated with a lower likelihood of maternal repartnering relative to infrequent and no father involvement (results not shown).

Fourth, we used cross-lagged structural equation models to further investigate the direction of the associations of focus. Specifically, we tested the relative magnitude of associations of father contact and child support payment with subsequent maternal repartnering compared to those of maternal repartnering with subsequent father contact and child support payment. These results (available upon request) indicated that, for mothers who were single at birth, there was a stronger link from very frequent father contact to maternal repartnering than vice versa.

Fifth, we added controls for child disability and child behavior problems to the model. We did not include these variables in our primary analyses because, if measured in the concurrent (nonlagged) interview, they are potentially endogenous. We were unable to use lagged versions of these variables because they were not measured until wave 2 (approximately age 3). Sixth, we estimated models using alternative samples in which we excluded all observations in which the mother was living with the biological father in the prior interview wave. Finally, for mothers who were single at the focal child’s birth, we excluded observations in which the mother reported that she and the father were “closely involved” at the time of the birth (37% of single-mother families at birth), which we interpreted as potentially implying that they were in a romantic relationship at the time of the birth. Results (not shown) from each of these robustness checks were consistent with the results from our primary analyses.

5 Discussion

This paper adds to a small but growing literature on the determinants of maternal repartnering. Despite both high rates of single parenthood in the UK and increases in nonresident father involvement over time, few studies have examined associations of father involvement (contact with the child and the payment of child support) with maternal repartnering. Using data from the UK Millennium Cohort Study, we find that very high-frequency nonresident father contact is negatively associated with the likelihood that a mother will form a new cohabiting or marital union only for mothers who were neither married nor cohabiting with their child’s biological father at the time of the focal child’s birth. The size of this association is large: mothers who were single at the birth were 70% less likely to repartner if the nonresident father had contact with the child at least five times a week than if he had no contact at all, and were also more likely to repartner if he was in contact with the child but less frequently. However, this finding reflects repartnering into a residential union (marriage or cohabitation) only. In contrast, very frequent father contact was associated with a greater probability of a new noncoresident relationship for these mothers.

We found no association between father involvement and repartnering for mothers who were married to or cohabiting with the child’s father at the time of the birth. We also found no evidence that child support receipt is associated with maternal repartnering, regardless of parental relationship status at the time of the birth.

There are several potential explanations for why mothers whose children experience high levels of nonresident father involvement may refrain from repartnering into marriage or cohabitation. First, it is possible that these mothers may perceive less need to form a new cohabiting or marital relationship to provide a father figure for their child. A second possibility is that very frequent father contact is indicative of a close coparenting relationship, which mothers do not want to risk upsetting. Relatedly, highly involved nonresident fathers may implicitly discourage mothers from repartnering, particularly if a mother perceives the father’s involvement as being contingent on her remaining single. Involved fathers may also act as gatekeepers by attempting to disrupt mothers’ new relationships through their actions toward mothers and their (potential) partners. Third, men may not want to partner with mothers who are in very frequent contact with their ex-partner, which may suggest both fewer opportunities for and lesser attractiveness of these mothers vis-à-vis repartnering.

Each of these hypotheses should be relevant to all mothers, regardless of relationship type at their child’s birth. Why, then, do we only find a negative association between highly involved nonresident fathers and maternal repartnering for mothers who were neither married to nor cohabiting with their child’s biological father at the time of the child’s birth? The most likely hypothesis may be that, for these families, high levels of father involvement imply ongoing ambiguities or unresolved issues regarding the parents’ romantic relationship (Shafer et al. 2013). Specifically, whereas divorce or the dissolution of a cohabitating relationship may constitute a relatively clear and explicit end to a romantic relationship, a breakup in the context of a noncoresident relationship may be less clear-cut. If so, then despite that we excluded from our analyses mothers who reported being in a romantic relationship with their child’s biological father at any observation point, our results may still reflect that some of these mothers perceive that their relationship with their child’s biological father may not have completely ended. Furthermore, potential new partners may also perceive high levels of father involvement among these families—those who did not experience divorce or the formal, physical separation represented by the end of a cohabiting relationship—as a potential threat and may, therefore, be reluctant to heavily invest in a relationship. Whereas we could not directly test these hypotheses, they merit further exploration.

On the whole, our findings lend no support to the hypothesis that mothers whose children have little contact with their biological father are more likely to seek a new relationship, potentially to provide a father figure for their child(ren). Nor do they suggest that mothers who are receiving little or no child support repartner to increase their economic stability. They provide no support for the hypothesis that greater father–child contact or regular child support payments offer mothers greater opportunities to seek new partners by easing time-related or financial burdens associated with childrearing. At the same time, our analyses of socioeconomic subgroups show a positive relationship between father contact and maternal repartnering for highly educated mothers who were married at the time of the focal child’s birth. A higher frequency of father involvement is associated with increased repartnering for this group. This may reflect that highly educated mothers benefit more on the repartnering market from father involvement than do less educated mothers because the former face greater opportunity costs of engaging in child rearing activities relative to the latter. The participation of fathers can increase their repartnering chances by freeing time for seeking a new relationship. Unfortunately, our data do not allow us to test why frequent father involvement is associated with increased maternal repartnering for more educated, previously married mothers but not for less educated never married mothers. We also caution that children of formerly married parents who experience very frequent nonresident father involvement represent only a small portion of our sample, particularly when we consider highly educated mothers. Nonetheless, this finding is in line with US research, which has found positive associations between father involvement and remarriage among previously divorced mothers (McNamee et al. 2014). It may, thus, imply that the time availability hypothesis (that mothers with an involved nonresident father can allocate more time away from childcare and toward establishing and maintaining a new relationship) is relevant only for relatively advantaged mothers. More generally, father involvement may influence repartnering among disadvantaged mothers differently than among advantaged mothers. These possibilities are ripe for examination in future work.

Our results also show that, for some groups, the repartnering process can be independent of father involvement after parental union dissolution, notably for the mothers in a cohabiting relationship at birth. This could be interpreted as greater autonomy among separated mothers or that the financial constraints and opportunities on the remarriage market remain the main determinants of repartnering after separation, regardless of father involvement.

There are several limitations to our analyses that should be considered when interpreting our findings. First, despite that our results were robust to a number of alternative specifications and operationalizations of key measures, and that we utilized lagged measures of father involvement and child support receipt, we cannot fully rule out the possibility of selection bias or be fully sure of the causal ordering of the associations of focus. Of particular note, the child-wave observation points in our data took place at roughly 2-year intervals, such that we could not determine whether, within each period, a maternal repartnership preceded or followed a change in father involvement or child support receipt for families in which the biological father and new partner were observed as being coresident in contiguous waves. However, our confidence in the causal ordering of these relations was increased in that our findings were robust to an alternative specification of our models in which we excluded all such families. Additionally, our supplemental cross-lagged results suggested that high-frequency father involvement had a larger association with subsequent maternal repartnering than vice versa. Another consequence of the 2-year interval between observations is that we may have underestimated maternal repartnering in that we would not have observed repartnerships that formed and dissolved between consecutive waves. To the extent that such partnerships are systematically associated with father involvement and child support receipt, this may have biased our estimates.

Second, our father involvement and child support measures were limited in that we had no information on legal physical custody arrangements. Nor did we know the length of father–child visits, where they took place, whether they were only during the day or also overnight, or whether the mother was present. We also were able to examine only the frequency of visitation, not the nature or “quality” of father–child (or mother–father) interactions. Each of these factors may influence the association of father involvement with maternal repartnering. Additionally, although child support receipt refers specifically to transfers on behalf of a child from the nonresident biological parent to the resident biological parent, no definition of “regular” or “irregular” payments was provided to mothers, and we had no information on the amounts of child support received, nor whether in-kind contributions were received from fathers. If mothers systematically differed in their reporting of father involvement and child support receipt in a way that was associated with subsequent maternal repartnering, our results will have been biased. However, even if limited, repeated information on father involvement and child support payments are still relatively rare in surveys; this study therefore represents a unique opportunity to examine associations between father involvement and maternal repartnering. Future studies should endeavor to include better measures.

Third, our analyses do not adjust for sample attrition over time. As is often the case with longitudinal studies, sample attrition in the MCS is greater among disadvantaged families and more unstable households than their more advantaged and stable counterparts. Given that such families are most likely to experience union dissolution and repartnering, coupled with the fact that our findings appear to be strongest among disadvantaged families, our findings might underestimate the strength of the relationships presented.

Despite these limitations, this study makes an important contribution to what we know about associations of nonresident father involvement and child support receipt with maternal repartnering. In sum, using a large, representative birth cohort study of UK mothers with a young child, we found that very high levels of father contact are negatively associated with maternal repartnering for mothers who were not married to or cohabiting with their child’s father at the time of the child’s birth. A possible explanation for this finding is that some parents may be engaged in relatively fluid relationships over a considerable number of years following a noncoresident birth. Future research should further examine this possibility and examine whether cross-national trends toward greater involvement by nonresident fathers (both voluntary and encouraged by public policy) may influence disparities in repartnering over time and, if so, how this might influence children’s well-being.