Associations between Parental Relocation Following Separation in Childhood and Maladjustment in Adolescence and Young Adulthood

Abstract

Petitions by custodial parents to relocate children away from non-custodial parents present difficult choices for family courts. In the current study, the sample (N = 81) was randomly recruited through the children’s schools according to the following criteria: Children were 12 years old and at the time resided primarily with their mothers; mothers had been living with a male partner “acting in a father role” for at least the previous year. Thirty-eight children had been separated by more than an hour’s drive from their biological fathers due to either their mothers or fathers relocating. The data were collected from two reporters (children and mothers) at five time points (child ages 12.5, 14, 15.5, 19.5, and 22) by trained interviewers using standardized measures with adequate reliability and validity. Long-distance separation from biological fathers prior to age 12 was linked in adolescence and young adulthood to serious behavior problems, anxiety and depression symptoms, and disturbed relationships with all three parental figures (i.e., biological fathers, mothers, and step-fathers). These associations held after controlling for mother-stepfather conflict and domestic violence, mothers’ family income, and mother-biological father relationship quality. These longitudinal findings over time replicated the cross-sectional findings of Braver, Ellman, and Fabricius (2003) and Fabricius and Braver (2006). Policy implications for parental long-distance relocation following separation are discussed.

Following their parents’ divorce or separation, some children experience an additional separation due to the relocation of one of their parents. Hetherington and Kelly (2002) suggested that relocation that separates the parents’ homes by an hour’s driving distance creates substantial interference with the child’s access to both parents. Braver and O’Connell (1998) found that within two years following separation, 17% of custodial parents relocated at least one hour away, and Ford (1997) found that within four years, the rate of custodial parent relocation was 25%. Braver, Ellman, and Fabricius (2003) found that by the time children were in college, 29% of custodial parents had relocated at least one hour away from the original family home. Legal disputes arise when the custodial parent wishes to move a substantial distance away with the child and the other parent objects. In those cases, courts are faced with a considerable dilemma (Kelly & Lamb, 2003) that places them at the interdisciplinary intersection of social science, family law policy, and legal scholarship.

Social Science

Austin (e.g., 2008; 2015) has noted that state courts, policy makers, and advocates have long ignored the large, nationally representative studies of family mobility in the fields of sociology and demography. Family mobility has been shown to adversely impact several domains of child adjustment including school behavior, academic success, teen pregnancy and sexual activity, drug and alcohol use, and well-being. These findings have been explained from a “social capital” perspective (Coleman, 1988; 1990) as due to disruptions in family and community networks that provide support for children. Significantly, data from the Child Health Supplement to the 1988 National Health Interview Survey revealed that the loss of “community capital” entailed by family mobility was not associated with impaired school performance unless the child also experienced the loss of “social” and “human capital” entailed by not living with both biological parents (Tucker, Marx, & Long, 1998). Each of these two sources of family instability – residential moves and separations from either mothers or fathers – have recently been shown to pose risks for adolescent problems across a variety of developmental domains over and above family demographics and the quality of the adolescents’ current environments (Adam & Chase-Lansdale, 2002). Relocation of the child by the custodial parent away from the non-custodial parent thus exposes the child to both potential sources of family instability (i.e., moving and separation). However, the sociological and demographic literatures provide limited direct comparisons of relocation to non-relocation within samples of divorced families.

Saini, Allan-Ebron, and Barnes (2015) found 11 studies of relocation in the context of divorce or separation, but eight of those studies did not investigate children’s outcomes. In one of the three studies (Verropoulou, Joshi, & Wiggins, 2002) that did investigate children’s outcomes, 17% of the participants were single-mothers at the child’s birth, and the researchers did not examine relocation separately in the single-mother versus divorced families. In addition, relocation occurred within six months of a change in “family status,” which could have been parental separation, the new inclusion of a biological father, or the arrival of a stepfather. Finally, the Verropoulou (2002) study did not distinguish between short- and long-distance moves, and the data suggested that a high proportion of moves may not have even required children to change schools. Only the remaining two studies located by Saini et al. (2015) were empirical studies (Braver et al., 2003; Fabricius & Braver, 2006) that limited the comparison to divorced, non-relocating families versus divorced, relocating families. These studies (discussed in more detail below) drew from the same dataset and reported several long-term negative child outcomes associated with both custodial and non-custodial parent relocation.

Relocation Policy

In the United States, policies regarding relocation of children after parental separation are governed by statutory or case law. Prior to 1996, state appellate courts did not consider the empirical sociological and demographic literature on family mobility for insights about the child’s best interests in relocation cases, and there was no ruling by a state Supreme Court that could serve as a precedent to give direction to lower courts. In that environment, many trial courts employed conditional change-of-custody orders, in which courts would order, conditions permitting, that residential custody would revert to the non-moving parent if the petitioning parent chose to relocate (Braver et al., 2003). Evidence suggests that such orders usually worked as intended to deter relocations (Braver, Cookston, & Cohen, 2002).

In 1996, the California Supreme Court set a precedent when it ruled (In re the Marriage of Burgess, 1996) that there was no ground for permitting courts to test the relocating mother’s commitment to parenting the child by means of the “bluff” that custody would revert to the non-moving father. Instead, the court ruled that the primary custody parent has a presumptive right to move with the child. Similar rulings were handed down by the New York State Supreme Court in Tropea v. Tropea (1996), and later by the New Jersey State Supreme Court in Baures v. Lewis (2001). Richards (1999) credited an amica curiae brief filed by Wallerstein (1995; see also Wallerstein & Tanke, 1996) with influencing both Burgess and Tropea. The Wallerstein brief, which argued that the child’s best interests would be served by being allowed to relocate with the custodial parent, did not reference the sociological and demographic literature on family mobility available at the time, but relied instead on a limited selection of the empirical literature on children’s adjustment to divorce, composed mostly of Wallerstein’s own studies (Warshak, 2000).

In 2004, however, the California Supreme Court stepped back from its 1996 position in In re the Marriage of LaMusga, ruling that relocation cases required a detailed examination of whether a proposed move was in the child’s best interests, particularly in regard to the child’s future relationship with the non-moving parent. It is widely held (e.g., McElroy, 2004; Eaton, 2004) that an amici curiae brief (Warshak et al., 2003) that cited Braver et al. (2003) and that was signed by nearly 30 social scientists and mental health practitioners influenced the LaMusga court to reverse Burgess. The LaMusga ruling has been influential, most recently in the New Jersey Supreme Court ruling in Bisbing v Bisbing, 2017. In Bisbing the court “abandoned” (p. 35) its own earlier standard in Baures, citing two reasons. First, the social science that it relied upon at the time has since diverged to reveal that “a relocation far away from a parent may have a significant adverse effect on a child” (p. 29; citing, among other references, Braver et al., 2003). Second, “the progression in the law toward recognition of a parent of primary residence’s presumptive right to relocate with children, anticipated by this Court in Baures, has not materialized” (p. 29).

Legal Scholarship

There has been much on-going legal scholarship on the issue of how relocation statutes should be written. Two recent examples are Parkinson and Cashmore (2015) and Thompson (2015), who both agree that parental relocation is likely to harm the nonresident parent-child relationship. The issue on which they differ is whether courts should operate under legal presumptions that direct their evaluation of the appropriateness of the relocation, or whether courts should simply continue to use the Best Interest of the Child Standard (BIS) without presumptions. Parkinson and Cashmore (2015) eschewed presumptions “for” or “against” parental relocation, and suggested (as did Elrod, 2006), that the BIS is still the best approach when combined with a decision tree to help courts reach a relocation decision. Parkinson and Cashmore recommended that the starting point for determining whether to allow or prohibit relocation ought to be a recognition that nonresident parents play a developmentally important role in children’s lives and that children’s relationships with nonresident parents are likely to be significantly diminished if parents live far apart. Next, if the nonresident parent-child relationship is important to the child, viable alternatives to the parents living a long distance apart should be considered. If no such options are available, then courts should be allowed to determine which alternative is the least detrimental: (1) either allowing the resident parent to relocate with the child away from the nonresident parent, or (2) ordering a contingent change-of-custody, in which case, if the petitioner moves, custody reverts to the nonmoving parent and the child does not move.

Thompson (2015) argued that Parkinson and Cashmore’s (2015) approach was “more of the same,” because their guided decision process is already the de facto approach taken by most courts when deciding relocation cases. Thompson contended that the BIS has failed because it allows too much judicial discretion and variation, and that specific presumptions are needed both “for” and “against” relocations. While acknowledging that there can be no universal presumptions, Thompson proposed four categories that warrant presumptions. The first category is interim moves that require a quick decision for a job assignment or an educational opportunity that will not result in a permanent move. Thompson argued for a presumption against relocating the child in such cases because the child’s interests are not served by moving back and forth. The second category is unilateral relocations in which a relocating parent has attempted to move without notice, or has even engaged in abduction, often for the clear purpose of separating or even alienating the non-custodial parent from the child. Thompson also argued for a presumption against relocation in those cases. The third proposed presumption against relocation applies to both parents when they share physical custody and are substantially active in the child’s life. Finally, Thompson argued that a presumption for relocation should be in place when there is an explicitly identified predominant primary caregiver. Thompson acknowledged, however, that not all the definitions of the four categories are clear and distinct, and examples of exceptions may include relocation for a primary parent job or needing to move for child or elder health care.

Neither Thompson (2015) nor Parkinson and Cashmore (2015) provided their readers with a discussion of the quantitative social science data directly related to relocation in divorce cases. Parkinson and Cashmore discussed only their own qualitative interview data obtained from 70 volunteering parents. Thompson provided only a cursory mention of the two empirical psychological studies (Braver et al., 2003; Fabricius & Braver, 2006), and suggested that those findings could be dismissed because they reflected only the bias of the researchers without informing his readers that both studies were peer-reviewed and that many commentators (e.g. (Austin, 2008; Pasahow, 2004; Taylor & Freeman, 2004; Mart & Bedard, 2005; Waldron, 2004) praised the papers, while a few criticized them (Horsfall & Kaspiew, 2010; Bruch, 2006; Ramsey & Kelly, 2004), as detailed below.

The Current Study

Crandall and Sherman (2016, p. 93) define “conceptual replication” as “the attempt to test the same theoretical process as an existing study, but that uses methods that vary in some way from the previous study.” Successful conceptual replication means that “the idea is substantially more likely to be correct than if it replicates using the exact same operationalizations….as such, conceptual replications are critical for establishing the generalizability of an initial observation and the theory it purports to support” (p. 94). The current study fits the definition of a conceptual replication of the two previous empirical studies (Braver et al., 2003; Fabricius & Braver, 2006). Below we describe the methods and findings of the two previous studies, discuss all the published criticism of those studies, and summarize the methodological differences that make the current study a conceptual replication.

Braver et al. (2003) studied a sample of 602 college students from divorced families. Using planned, one-tailed tests, they found that students who reported that they had relocated with their mothers showed several current negative outcomes compared to students who reported that their mothers and fathers had remained in the same community. These outcomes included more inner turmoil and distress about the divorce, poorer self-rated physical health, worse relationship quality between their parents, worse relationships with their fathers, and less financial support from parents. Very similar results were obtained with planned, one-tailed tests for the other group of students who reported that their fathers had relocated without them (the only differences were that they did not report poorer physical health but did worry more about college expenses). Thus, negative outcomes were not associated with the child moving, but rather with the separation from the father. Among the remainder of the 14 outcome variables (personal/emotional adjustment, hostility, relationship quality with mothers and with both parents, platonic and romantic relationship choices, worry about substance use and general life satisfaction), none showed any benefits associated with relocation (confirmed by re-analysis by the current authors with one-tailed tests in the opposite direction), despite the expectation that relocation will improve the parent’s and, by extension, the child’s lives.

The Braver et al. (2003) study was critiqued in amici curiae briefs filed with the California State Supreme Court by Bruch (2002), Navarro (2003), and Wallerstein et al. (2003). Those critiques were responded to in detail by Fabricius and Braver (2006). In summary, the Briefs made three main points that we address in the current study. First, the briefs claimed that the outcome variables on which there were no significant differences between those who moved with their mothers and those who stayed with both parents were the most important variables, and that there were only two important differences (“distress from the divorce … [and] overall physical health” Bruch, 2003, p. 3–4). The authors of the briefs gave no rationales for how they decided which outcome variables were most important, or for why they ignored the rationales given by Braver et al. (2003) for the importance of the quality of the inter-parental relationship and the quality of the father-child relationship for children’s well-being.

The second issue was that Braver et al. (2003) did not identify the children whose mothers were denied permission to relocate them. They could be “hidden” among the group whose parents did not move, and they could also be among the 8% of the sample whose mothers relocated without them. The Briefs argued that because the latter group often had some of the most negative scores, the California Supreme Court should uphold Burgess and re-affirm the presumptive right of custodial parents to relocate children. Fabricius and Braver (2006) responded that the most likely reason that denying permission to relocate the child would cause harm would be if the child was left in dangerous or difficult circumstances, and thus the more justifiable policy is that “courts have to balance the types of negative outcomes [Braver et al. (2003)] found were associated with separation from a parent, and the types likely to arise from staying in dangerous or difficult situations” (p. 11).

The third issue was the question of causality and the possibility that some other, unmeasured “third variables” could have caused both the relocation and the negative outcomes. The Briefs suggested some more or less plausible candidates, including younger child age at the divorce, remarriage, weaker ties to the community, and lower income. Braver et al. (2003) also discussed this issue, and suggested poor parent mental health and poor parenting. Both sets of writers suggested parent conflict. High levels of pre-existing parent conflict and lower income are the most plausible candidates for factors that could both substantially contribute to parents’ decisions to relocate and substantially harm children’s outcomes. Remarriage, for example, is not necessarily harmful, and poor parenting or mental health are not typical reasons for relocating.

Follow-up analyses of the same data set by Fabricius & Braver (2006) addressed the question of causality by controlling for (a) the frequency and severity (the highest category on this scale included physical violence) of parent conflict over the seven-year period from two years prior to the separation to five years post separation, (b) domestic violence between the parents measured with several variables that either left the time period unspecified or specified after the divorce, (c) child age at separation, and (d) number of years the parents lived an hour’s drive apart. Fabricius and Braver (2006) reported evidence for the reliability and validity of measures in (a) and (b). After controlling for all four sets of variables, relocation of either the mother with the child or the father without the child still predicted more lingering inner turmoil and distress about the divorce among the college students, worse current relationship quality between their parents, and worse current relationships with their fathers (but not less financial support from parents, more worry about college expenses, or poorer self-rated physical health). Effect sizes showed that the negative impact associated with relocation on each of these three variables was either comparable to or twice that of parent conflict and domestic violence. These findings indicated that pre-existing parent conflict and violence were not likely the explanation for the distress and worsened relationships associated with relocation. However, in only one of the analyses (financial support from parents) did the researchers also control for parents’ financial status (in which case relocation was no longer significant), and thus parents’ financial status remained a potential third variable. Finally, there was no evidence that relocation reduced the frequency or severity of parent conflict up to five years post-separation.

Braver et al. (2003) noted that the generalizability of their findings to the population of divorced families might be limited because their participants were college students. If the children of divorce who go to college are the “success stories” (Garfinkel, McLanahan & Wallerstein, 2004, p. 343), then a college sample should be less likely to be negatively affected by relocation and more likely to underestimate negative outcomes. However, there are several sources of evidence suggesting that what is learned from college students can generalize to the larger population of children of divorce. Braver et al. (2003) reported evidence that among college students, the percentage of children from divorced families is similar to the percentage of children of married parents who are expected to experience their parents’ divorce (about 30%; Bumpass & Sweet, 1989; National Center for Health Statistics, 1990). Fabricius and Braver (2004) reported evidence that among divorced families, the percentage of children who attend college is similar to the percentage among intact families (about 50%; McLanahan & Sandefur, 1994). In addition, the associations between parenting time and child outcomes are similar in samples of convenience, including college students, and samples obtained from court records (Bauserman, 2002; Laumann-Billings & Emory, 2000); painful feelings about divorce are similar in college students and low-income community samples of adolescents (Laumann-Billings & Emory, 2000); and endorsement of shared parenting is similar among college students and the general public (Braver, Ellman, Votruba, & Fabricius, 2011; Fabricius & Hall, 2000).

The current study qualifies as a conceptual replication of Braver et al. (2003) and Fabricius and Braver (2006) because the new methods included a community sample, (versus a college sample) oversampling of Mexican-American families, only families in which the mother had re-partnered by child age 12 (versus both types of families), new control variables (income, domestic violence, and parent conflict in the mother’s home), data obtained by trained interviewers from mothers and youth (versus only youth self-report), an expanded list of outcome measures, appropriate to age, and five waves of assessment (versus one wave at age 19).

As part of a larger, longitudinal project, a community sample of 175 families was obtained in which the mother was the primary residential parent and a stepfather also resided in the home; thus, the mother was separated or divorced from the biological father. These families were randomly recruited through the children’s schools when they were 12 years old. Approximately half of the children fit the criteria for this study (described in the Method section; e.g., the birth parents had lived together when the child was born). Prior to being sampled when they were 12, half of the children had experienced a separation from their biological fathers of at least one hour’s drive away either because their mothers relocated with them (N=17) or their fathers moved away from them (N=17). The other portion of the sample did not experience any separation from their biological fathers because of a move (N=43). This was the control/comparison group. The data were collected from two reporters (children and mothers) at each of five distinct time points (child ages 12.5, 14, 15.5, 19.5, and 22) by trained interviewers. It is important to note that the data set did not contain an important relevant group, divorce cases in which the mother had re-partnered, which accordingly poses a limit on the generalizability of these findings. No two children in the sample were in the same family.

Youth outcomes were assessed from age 12 to age 22, and targeted serious behavioral and mental health problems, including involvement with the juvenile justice system, risky sexual behaviors, drug use, involvement with delinquent peers, aggression, and delinquency, and symptoms of depression and anxiety. Additionally, we were interested in the links between relocation and how much adolescents felt they mattered to their parents (Elliott, 2009; Marshall, 2001; Rosenberg & McCullough, 1981; Schenck et al., 2009). Stevenson et al. (2014) found that increased daily interactions with the same-sex parent predicted increases in adolescents’ perceived mattering to that parent one to two years later. Thus, we expected that the decrease of daily father-child interactions resulting from relocation could threaten children’s belief that they mattered to their biological fathers.

Research has shown that harmful consequences may arise when children doubt how much they matter to their parents. When children perceive they matter to their parents then they can trust that the parents will be there when needed, which allows them to feel emotionally secure in the relationship. Emotional security theory (Cummings & Davies, 1996) and attachment theory (Bowlby, 1969), both hold that children’s emotional security in the family is necessary for optimal adjustment. Indeed, Rosenberg and McCullough (1981) found that adolescents’ doubts about mattering to their parents predicted symptoms of depression, anxiety, and delinquency even after controlling for adolescents’ self-esteem. Schenck, Braver, Wolchik, Saenz, Cookston, and Fabricius (2009) found that adolescents’ perceptions of how much they mattered to their residential stepfathers as well as their non-residential biological fathers, controlling for perceived mattering to mothers, were each significantly related to adolescents’ mental health problems. Most recently, Suh et al. (2016) found that mattering to residential fathers and stepfathers in early adolescence, rather than mattering to mothers, was the significant predictor of changes in adolescents’ mental health symptoms two years later.

In the present study, we controlled for family factors available in the data set that are known to be associated with child adjustment problems; namely, mothers’ family income, mother-stepfather non-violent conflict, mother-stepfather intimate partner violence, and mother-biological father relationship quality, as reported by mothers. Recent work has shown that nonviolent and violent parent conflict, as well as the emotional quality of the parents’ relationship, each contributes independently to adolescent adjustment (Suh et al., 2016). Controlling for all these family factors allowed us to test whether relocation predicted adolescent and young adult behavior problems, symptoms of depression and anxiety, and perceived mattering to parents, above and beyond any impact on these outcomes from exposure to violent and non-violent parental conflict in the residential home, poor relationships between the biological parents, and stress due to lower family income.

Method

Participants

Participants came from the Parents and Youth Study (PAYS), a longitudinal investigation of the role of fathers and stepfathers in adolescent and young adult development conducted in two sites: the greater Phoenix, AZ and Riverside, CA areas. The variables for the current study were a small part of the overall project, and participants were not made aware that one study goal might be the evaluation of outcomes following relocation. All procedures were approved by the University Institutional Review Boards at both sites. Prior to the interview, youth gave assent, and mothers and stepfathers gave informed consent.

At both sites families were recruited from the children’s schools according to the same ethnic background and family composition criteria. The ethnic background criterion required the child, mother, and father to all self-identify as either European-Anglo or Mexican-American. The family composition criterion required either that the child’s birth parents were still married or cohabitating, or that the child lived primarily with the mother and her male partner “acting in a father role” who had resided with the child’s mother for at least the previous year. The sampling protocol provided approximately equal numbers of families in the four ethnicity by family composition groups. No data were collected about the number of stepfather figures a child might have had, nor whether the child had a stepmother residing with the biological father. As noted, single-mother (i.e., mother not repartnered) families were excluded from the sample. No two children in the sample were in the same family.

In Arizona, a total of 2,459 families appeared eligible from a school survey, and 640 were randomly selected for additional screening and consent. From these, 204 (32%) families were both eligible and initially agreed to participate. In California, researchers were not allowed to call families without their prior consent. 540 families responded to school personnel that they were willing to be contacted and were eligible for additional screening and consent. From these, 192 (36%) families were both eligible and initially agreed to participate. The total number of families from the two sites that participated in Wave 1 was 392. Of these, 217 of them were intact families (110 European-American and 107 Mexican American) and 175 were stepfather families (89 European American and 86 Mexican American). More sample recruitment details are reported in Schench et al. (2009). The 217 intact, biological families of both ethnicities were ultimately excluded from the current analysis because the parents were not separated.

For the current analysis, a subset of the 175 stepfather families was selected according to the following criteria: (1) the birth parents had lived together when the child was born (n = 86), (2) the birth parents had separated after the child was born and prior to Wave 1 (n = 86), (3) if relocation occurred, it happened before Wave 1 (n = 83) and (4) children lived more than half time with the mother and stepfather when the relocation occurred (n = 81). The final sample thus included 81 adolescents (38 male, 43 female) from families where the mother had remarried or partnered with a step-father figure. The sample was generally upper middle class (Wave 1 mean adjusted income = $74,528) according to mothers’ reports at Wave 1. Data about father income was not reported. Of these 81 families, 38 had one parent (father or mother) relocate an hour or more away and 43 did not relocate prior to Wave 1. Of the relocated families, 17 children moved away with their mothers, and 17 children remained with their mothers while their fathers moved away. Four families did not provide data on which parent moved but were retained in the analyses because we did not plan to test for differences in which parent moved given the small cell size. Data were collected at five waves: Wave 1 (Mean child age = 12.5 years), Wave 2 (M = 14 years), Wave 3 (M = 15.5 years), Wave 4 (M = 19.5 years), and Wave 5 (M = 22 years). The mean age of children when the children were separated from their biological father due to a move was 5.3 years of age.

Eight families dropped out of the final sample (N = 81) by Wave 5 (remaining N =73). No significant differences were found on any demographic variables between families that dropped from the study by Wave 5 and families that remained in the study. Missing data were handled using Full Information Maximum Likelihood (FIML) estimation in all analyses. This method of handling data does not delete families that have data on at least some of the predictors and outcomes and enabled us to retain the maximum amount of families.

Procedures

In Arizona, three interviewers travelled to each home, and the three family members (mother, stepfather, and child) were interviewed in different rooms. In California, the family traveled to the research site and family members were similarly interviewed in separate rooms. The data reported here were a small part of the overall project, and so no mention was made during recruitment about relocation. At Wave 1, mothers reported family income, mother-stepfather non-violent conflict, mother-stepfather intimate partner violence (IPV), and mother-non-resident father relationship quality. At Wave 3 mothers reported the age of the child when the parents separated, family relocation status, age of the child at relocation, and which parent relocated. At all five waves, adolescents reported their perceptions of mattering to their mother, stepfather, and non-resident, biological father. At Wave 3, 4 and 5 adolescents reported via a private, self-administered paper-and-pencil survey their involvement with the juvenile justice system, risky sexual behavior, drug use, and internalizing and externalizing symptoms (Mothers also reported some of the IPV items in the self-administered format for privacy). At Waves 3 and 4 adolescents also reported their involvement with delinquent peers. Involvement with delinquent peers was not measured at Wave 5 because the measure was more appropriate for teenagers.

Measures

Demographic Variables

Mothers reported child age at separation (M = 3.96 years, SD = 2.66, range = 11.00); relocation status “After the two of you permanently separated, did you or the child’s birthfather ever move (or live) more than a one-hour drive away from each other?” (38 yes, 43 no); child age at relocation (M = 5.29, SD = 3.69, range = 12.00); and which parent relocated (17 children moved away with mothers, 17 children remained with mothers while fathers moved away; in four cases we could not determine which parent moved).

Parent Relational Variables

Mothers reported the following variables at Wave 1.

Mother-stepfather non-violent conflict

Five items were selected from the Intensity and Frequency subscales of the Children’s Perception of Interparental Conflict (CPIC; Grych, Seid, & Fincham, 1992). Sample items include “Child never saw you and partner arguing or disagreeing,” and “When you and partner had an argument in front of child, you yelled a lot.” Mothers responded for the past year about her and the stepfather. The CPIC response scale is 1 (true), 2 (sort of true), and 3 (false). Three items were reverse coded so that higher scores indicated higher levels of conflict and were averaged to form a composite score. The Cronbach coefficient alpha for the current sample was .76 (M = 2.31, SD = 0.30, range = 1.60).

Mother-stepfather intimate partner violence (IPV)

Items from the Physical Assault Subscale of the Conflict Tactics Scale −2 (CTS-2; Straus, Hamby, Boney-McCoy, & Sugarman, 1996) were combined into one question that assessed how many times any types of physical assault (i.e., “shoved, slapped, punched, kicked, bit, choked you, threw things at you, or any similar physical act”) had been committed in the past year. Mothers were self-administered this question twice: once for her violence toward the stepfather and once for the stepfather’s violence toward her. The Injury Subscale items from the CTS-2 were combined into one question asking how many times injury (“bleeding or bruised”) had occurred from physical assault in the past year. Mothers were self-administered this question twice: once about her injury from the stepfather and once about stepfather’s injury from her. The response scale was the same as for the original scale items, 0 (never), 1 (once), 2 (twice), 3 (three to five times), 4 (six to 10 times), 5 (11 to 20 times), and 6 (more than 20 times).

Mothers also responded (in the interview format) to two items from the Intensity subscale of the CPIC that assessed other types of physically intimidating violent behaviors between the mother and stepfather that the child may have witnessed over the past year. These were: “Child saw that you and partner sometimes pushed or shoved each other during arguments,” and “You and partner broke or threw things during arguments in front of the child.” The CPIC response scale is 1 (true), 2 (sort of true), and 3 (false). Items were recoded such that higher scores indicated higher levels of violent behaviors.

All six items were z-scored and averaged to form one IPV composite (α =.80 for the current sample). IPV was non-normally distributed; thus, we log transformed this variable, as recommended (Rasmussen & Dunlap, 1991). Suh et al. (2016) reported evidence for reliability of these six items in that mothers’ reports correlated significantly with fathers’ reports, and for validity in that the six items formed a latent construct which predicted changes in children’s mental health symptoms from ages 12.5 to 15.5.

Mother-biological father relationship quality

Two items were constructed for the PAYS data set: “How well do you get along?” (responses were scored on a 5 point scale and ranged from 1= “extremely well” to 5= “not well at all”), and “What kind of relationship do you have?” (responses were scored on a 7-point scale and ranged from 1= “the worst” to 7= “the best.”). The first item was reverse scored so that higher scores indicated a better relationship. Both items were z-scored and averaged (α = .77 for the current sample). This measure was collected at Wave 1 when the children were a mean age of 12.5 years old.

Child Outcome Variables

Youth reported the following variables:

Juvenile justice involvement

At Waves 3, 4, and 5, five items were administered based on the Services and Assessment for Children & Adolescents (SACA; Ascher, Farmer, Burns, & Angold, 1996). The response scale was 0 = “no” and 1 = “yes” Items were: “In the past 12 months, have you been placed in a police car, picked up or brought home by the police?” (W3: 15 “yes”, 65 “no”, 1 missing; W4: 14 “yes”, 50 “no”, 17 missing; W5: 6 “yes”, 57 “no”, 18 missing); “… gotten a ticket or citation from the police for any reason, other than a parking ticket?” (W3: 8 yes, 70 no, 3 missing; W4: 31 yes, 33 no, 17 missing; W5: 17 yes, 46 no, 18 missing); “… been referred to a diversion program because of an offence you committed?” (W3: 10 yes, 69 no, 2 missing; W4: 8 yes, 56 no; 17 missing; W5: 2 yes, 61 no, 18 missing); “… been arrested for any reason?” (W3: 9 yes, 69 no, 3 missing; W4: 11 yes, 53 no, 17 missing; W5: 3 yes, 60 no, 18 missing); “… been found guilty or convicted of any crime?” (W3: 4 yes, 74 no, 3 missing; W4: 10 yes, 54 no, 17 missing; W5: 2 yes, 61 no, 18 missing). The five items at each wave were averaged to form one composite score. Alphas for the current sample were W3 = .78, W4 = .83, and W5 = .72. The composite was non-normally distributed and was log transformed.

Risky sexual behavior

At Waves 3, 4, and 5, seven items were selected from the Youth Risk Behavior Survey (YRBS; Brener, McManus, Kinchen, Sundberg, & Ross, 2002). Youth were asked if they had ever had sexual intercourse (W3 35 yes, 43 no, 3 missing; W4 60 yes, 4 no, 17 missing, W5 60 yes, 3 no, 18 missing); if yes, at what age they first had intercourse (W3 M = 14; W4 M = 15.78; W5 M = 15.84), how many partners they had over the past year (W3 M = 2.11; W4 M = 4.02; W5 M = 5.45), and how often they used a form of contraception over the past year (1=never to 5=always; W3 M = 4.12; W4 M = 3.83; W5 M = 3.70); if they had ever engaged in oral sex (W3 30 yes, 50 no, 1 missing; W4 56 yes, 8 no, 17 missing; W5 56 yes, 7 no, 18 missing); if yes, at what age they first engaged W3(M = 14.50, W4 M = 17.42; W5 M = 17.24), how many partners they had over the past year (W3 M = 1.86; W4 M = 2.92; W5 M = .41). Items were scored such that higher scores reflect having had sex, at a younger age, with more partners, and with riskier contraceptive behavior. Items were z-scored, averaged to form one composite score. Alphas for the current sample were W3 = .88, W4 = .74, and W5 = .69.

Drug use

At Waves 3, 4, and 5, six items were constructed for the PAYS data set asking “How many times during the past 30 days have you used cigarettes; … alcohol; …marijuana; … inhalants; … “other illegal drugs such as cocaine, crack, ecstasy, speed, PCP, LSD, heroin, crystal meth, crank, ice or pills?” (1 = 0 days to 7 = “all 30 days”), and “How many times during your life have you taken steroid pills or steroid shots without a doctor’s prescription?” (1 = 0 times to 6 = 40 or more times). We were interested in the degree to which youth engaged in polysubstance use. Thus, responses to each item were dichotomized into 0 (no use) or 1 (at least one use), and the six items were summed to create a composite variable such that higher scores represented more polysubstance use (W3: 66 no use, 14 at least one substance, 1 missing; W4: 47 no use, 17 at least one substance, 17 missing; W5: 18 no use, 63 at least one substance, none missing). At Wave 3 and Wave 4 this composite was non-normally distributed, and was log transformed.

Involvement with delinquent peers

At Waves 3 and 4, 15 items included the following: Items 1 to 5 (see below) derived from the Denver Youth Survey with adolescents (Mason, Cauce, Gonzales, & Hiraga, 1996); items 6 and 9 to 13 adapted from Dishion, Patterson, Stoolmiller, & Skinner’s (1991) measure of delinquent peers with slight word modification; item 8 drawn from Huizinga, Esbensen, and Weiher (1991); and items 7, 14, and 15 constructed for the PAYS data set based on a review of the literature. All items began, “During the past year, how many of your friends have…” 1. “used force (e.g., threats or fighting) to get things from people?” 2. “been in gang fights?” 3. “gotten drunk or high? 4. “lied about their age to buy or do things?” 5. “started rumors or told lies?” 6. “cheated on school tests? 7. “gotten suspended from school?” 8. “missed school without an excuse?” 9. “gotten in trouble at school?” 10. “stolen something worth less than $50? 11. “stolen something worth $50 or more?” 12. “ruined or damaged other people’s things on purpose (including tagging/graffiti)?” 13. “started a fight with someone?” 14. “used a weapon (e.g. rocks, bottles, knives, guns)?” 15. “hurt animals on purpose?” The response scale for all items ranged from 1 (“none”) to 5 (“almost all”). Items were summed to form one composite score. Alphas for the current sample were W3 = .94 (M = 29.21, SD = 12.80, range = 57), and W4 = .93 (M = 26.45, SD = 9.84, range = 58). At Wave 4 this variable was non-normally distributed and was log-transformed.

Internalizing symptoms

At Wave 3, six items were selected from the Revised Children’s Manifest Anxiety Scale (RCMAS; Reynolds & Richmond, 1985). Examples are: “In the past month you worried about what was going to happen” and “In the past month you woke up scared some of the time.” Item responses were 1 (yes) or 2 (no). All items were reverse scored and summed such that higher scores reflected greater anxiety symptoms. Alpha for the current sample was .69 (M = 8.93, SD = 1.81, range = 6).

At Wave 3 seven items were selected from the Children’s Depression Inventory (CDI; Kovacks, 1992). The CDI asks children to answer with respect to the last month and to pick one response from three responses for each item; for example, 0 (“I did not feel alone”) 1 (“I felt alone many times”) 2 = “I felt alone all the time.” All items were scored and summed such that higher scores reflected greater depression symptoms. Alpha for the current sample was .68 (M = 10.09, SD = 2.40, range = 9.00).

RCMAS and CDI scores were correlated, r = .60, p < .001, and were z-scored and averaged to form a composite variable representing general internalizing behavior with higher scores representing more internalizing symptoms (α = .80)

At Waves 4 and 5, young adults answered the Adult Self Report (ASR; Achenbach, 1991). The ASR incorporates many items of the 1997 editions of the Young Adult Self-Report (YASR), plus new items and national norms that span ages 18–59. The subscale for internalizing emotions was used for the current study. The profiles represent scale scores in relation to norms for each gender at ages 18 to 35 and 36 to 59, based on national probability samples. The response scale is 0 (not true), 1 (somewhat or sometimes true), and 2 (very true or often true). Higher scores reflect more internalizing symptoms. Alphas for the current sample were W4 = .90 (M = 54.05, SD = 9.92 range = 46) and W5 = .88 (M = 86.76, SD = 11.59, range = 21).

Externalizing symptoms

At Wave 3, eight items were selected from the Aggression Subscale and four items were selected from the Delinquency Subscale of the Behavior Problems Index (BPI; Peterson & Zill, 1986). The response scale is 1 (not true), 2 (somewhat true), and 3 (very true). Items were scored and summed such that higher scores indicated more externalizing symptoms. Alpha for the present sample was .69 (M = 15.73, SD = 2.87, range = 12.00).

At Wave 4 and 5, as described above for Internalizing, we used young adults’ scaled scores for externalizing on the ASR. Higher scores reflected more externalizing symptoms. Alphas for the current sample were W4 = .88 (M = 57.83, SD = 11.59, range = 60.00) and W5 = .88 (M = 91.52, SD = 2.97, range = 85.00).

Perceived mattering to parents

At all waves, a 7-item scale adapted from Rosenberg and McCullough’s (1981) review of correlates of mattering to parents was used to assess perceived mattering to mothers, residential stepfather, and non-resident biological father. Sample items included, “My (parent) really cares about me,” and “I’m not that important to my (parent)” (see Stevenson et al., 2014 for complete list of items). The response scale was 1 (strongly agree) to 5 (strongly disagree). Items were reverse scored so that higher scores reflected higher perceived levels of mattering, and items were summed to create an overall mattering score. The following describes the mattering to stepfather scale: Wave 1 α = .88, M = 29.60, SD = 5.48, Range = 25.00; Wave 2 α = .88, M = 28.85, SD = 5.84, Range = 23.00; Wave 3 α = .93, M = 27.56, SD = 7.63, Range = 28.00; Wave 4 α = .94, M = 27.53, SD = 8.05, Range = 28.00; Wave 5 α = .93, M = 29.49, SD = 6.31, Range = 27.00). The following describes the mattering to mother scale: Wave 1 α = .70, M = 32.94, SD = 3.09, Range = 14; Wave 2 α = .75, M = 33.04, SD = 2.84, Range = 11.00; Wave 3 α =.84, M = 32.71, SD = 3.55, Range = 20.00; Wave 4 α = .83, M = 33.38, 3.14, Range = 20.00; Wave 5 α = .80, M = 33.97, SD = 2.92, Range = 17.00). The following describes the mattering to non- resident father scale: Wave 1 α = .93, M = 28.80, SD = 7.34, Range = 28.00; Wave 2 α = .95, M = 29.91, SD = 7.15, Range = 27.00; Wave 3 α =.95, M = 28.63, SD = 7.70, Range = 28.00; Wave 4 α = .93, M = 24.43, SD = 8.35, Range = 28.00; Wave 5 α = .94, M = 25.50, SD = 7.79, Range = 25.00.

Results

Preliminary Analyses

Table 1 shows the frequencies of categories of child age at relocation and categories of distance between the parents’ homes after the relocation. The mean age of the child when either parent relocated more than an hour away was 5.3 years.

Table 1.

Frequencies of Categories of Distance of Relocation and Age at Relocation.

Variable	Category	Frequency
Distance of relocation	About a 1 hour drive	12
	2 or 3 hour drive	8
	4 or 5 hour drive	1
	6 or 7 hour drive	3
	More than a 7 hour drive	14
Age at relocation
	0- to 2 years old	12
	3–5 years old	8
	6–10 years old	14
	11–12 years old	4

We tested for differences on the Wave 1 control variables between the non-relocated and relocated families. There were no significant differences in family income between families that did not relocate (M = $70,816) and families that did relocate ($78,730; t(79) = −.69, p = .49); child’s age at separation (Ms = 3.84, 4.11, respectively; t(79) = −.45, p = .65); mother-stepfather IPV (Ms = 0.48, 0.47, respectively; t(79) = .99, p = .33); mother-biological father relationship quality (Ms = −0.05, 0.11, respectively; t(62) = −.43, p = .67); or child sex (20 males and 23 females did not relocate, 20 males, 18 males relocated; χ²(1, N = 81) = .01, p = .94). The only significant difference was that families that relocated had slightly lower levels of mother-stepfather non-violent conflict (M = 2.23 vs. M = 2.38), t(79) = 2.34, p = .02). Mother-stepfather non-violent conflict was nonetheless retained as a control. Given the importance in the literature of family income, child exposure to violence in the home of residence, and the residential parents’ relationship quality to children’s outcomes, those variables were also retained as controls in all future analyses.

Associations of relocation with adjustment and mattering to parents

Multiple regression analyses were conducted using Mplus 7.2 separately for each outcome (e.g. juvenile justice involvement, risky sexual behavior, internalizing, etc.). All regression analyses controlled for Wave 1 mother’s family income, mother-stepfather nonviolent conflict, mother-stepfather IPV, and mother-non-resident father relationship quality. Table 2 shows the beta coefficients for relocation status and control variables in the multiple regressions predicting measures of adjustment. In addition, it reports the overall R², which indicates how much of the total variability in the adjustment measure is accounted for by the overall regression model that includes all controls; e.g., R² = .20 means that 20% of the variability was accounted for by all variables in the model. Standardized betas are reported here in the text to capture the effect of relocation on each outcome, after partialling out the effects of control variables. At Wave 3 (mean age 15.5), relocation was associated with more juvenile justice involvement, β = .25, p = .04, more involvement with delinquent peers, β = .30, p = .01, and higher externalizing problems, β = .33, p = .003. By Wave 4 (mean age 19.5), relocation was associated with more drug use, β = .32, p = .01, internalizing, β = .33, p = .01, and externalizing, β = .41, p < .001. At Wave 5 (mean age 22), the associations with relocation had faded.

Table 2.

Beta Coefficients and R² for Relocation Status and Control Variables in Multiple Regressions Predicting Measures of Adjustment.

		Control Variables Wave 1
Adjustment	Relocation Status β	Mother- Stepfather Non- Violent Conflict β	Mother- Stepfather Intimate Partner Violence β	Mother-Non- Resident Father Relationship β	Mother’s Family Income β	R2
Wave 3
Juvenile Justice	.25*	.25*	--	--	−.35**	.18*
Risky Sexual	--	.29*	--	--	--	.14
Behavior Drug Use	--	--	--	--	--	.02
Delinquent Peers	.30**	.34**	.40***	--	−.22†	.29**
Internalizing	--	--	.26*	--	--	.11
Externalizing	.33**	.23*	.39***	--	−.22†	.27**
Wave 4
Juvenile Justice	--	--	--	--	--	--
Risky Sexual	--	--	--	--	--	--
Behavior Drug Use	.32*	--	--	--	−.26*	.16†
Delinquent Peers	--	.31*	.25*	--	−.28*	.19*
Internalizing	.33**	--	.38***	--	−.26*	.29*
Externalizing	.41***	--	.26*	--	−.29*	.28**
Wave 5
Juvenile Justice	--	--	--	--	--	.08
Risky Sexual	--	--	--	.25†	--	.08
Behavior Drug Use	--	--	--	--	.34**	.14†
Internalizing	--	--	--	--	−.28*	.16
Externalizing	--	.21†	--	--	−.45***	.27**

Note.

^†p < 09

^*p < .05.

^**p < .01

^***p < .001.

W3 = Wave 3. W4 = Wave 4. W5 = Wave 5. All analyses controlled for mother’s family income, Wave 1 mother-stepfather non-violent conflict, mother-stepfather intimate partner violence, and mother-non-resident father relationship quality. All betas presented are standardized betas. N ‘s ranged from 52 to 64.

Table 3 shows the analogous multiple regressions predicting perceptions of mattering to mother, mattering to stepfather, and mattering to non-resident father at Waves 1–5. At Wave 1, relocation was found to be associated with perceptions of less mattering to non-resident father, β = −.29, p = .02, and also, at a trend level, less mattering to stepfathers, β = −.22, p = .06. At Wave 2, relocation was associated with perceptions of less mattering to mother, β = −.32, p = .003, less mattering to stepfather, β = −.27, p = .02, and less mattering to non-resident father, β = −.29, p = .02. At Wave 3, relocation was associated with perceptions of less mattering to mother, β = −.25, p = .04, less mattering to stepfather, β = −.28, p = .05, and, at a trend level, less mattering to nonresident father, β = −.18, p = .09. At Waves 4 and 5, the associations with relocation had faded with the exception of perceptions of less mattering to stepfather at Wave 5, β = −.29, p = .03.

Table 3.

Beta Coefficients and R² for Relocation Status and Control Variables in Multiple Regressions Predicting Perceived Mattering to Parents.

		Control Variables Wave 1
Perceived Mattering to Parents	Relocation Status β	Mother- Stepfather Non- Violent Conflict β	Mother- Stepfather Intimate Partner Violence β	Mother-Non- Resident Father Relationship β	Mother’s Family Income β	R2
Wave 1
Mattering Mom	--	--	--	--	.25*	.08
Mattering StepF	−.22†	−.25*	--	--	.37***	.20*
Mattering NRF	−.29*	--	--	.22†	--	.18*
Wave 2
Mattering Mom	−.32**	--	--	--	.42***	.26**
Mattering StepF	−.27*	--	--	--	.31**	.14†
Mattering NRF	−.29*	--	--	.25*	--	.18*
Wave 3
Mattering Mom	−.25*	--	--	--	.25*	.12
Mattering StepF	−.28*	--	−.21†	--	--	.12
Mattering NRF	−.18†	−.28**	--	.49***	.18†	.37***
Wave 4
Mattering Mom	--	--	--	--	.45***	.20*
Mattering StepF	--	--	--	--	.25†	.10
Mattering NRF	--	--	--	.38***	.27*	.38***
Wave 5
Mattering Mom	--	−.31**	−.44**	.25*	.30**	.47***
Mattering StepF	−.29*	−.28*	−.62***	--	--	.37**
Mattering NRF	--	--	--	.36*	--	.20†

Note.

^†p <09

^*p < .05.

^**p < .01

^***p < .001.

W3 = Wave 3. W4 = Wave 4. W5 = Wave 5. StepF = stepfather. NRF = non-resident father. All analyses controlled for mother’s family income, Wave 1 mother-stepfather non-violent conflict, mother-stepfather intimate partner violence, and mother-non-resident father relationship quality. All betas presented are standardized betas. N ‘s ranged from 41 (W5) to 64 due to missing data.

Associations of control variables with adjustment and mattering to parents

Table 2 also shows that both mother-stepfather non-violent conflict and mother-stepfather IPV at Wave 1 had unique associations with most of the aspects of youth adjustment, including more juvenile justice involvement, risky sexual behavior, involvement with delinquent peers, internalizing, and externalizing; although, as with the associations with relocation, these associations were concentrated at Waves 3 and 4. We found no evidence for an association between mother-non-resident father relationship at Wave 1 and youth adjustment. Higher mother-stepfather family income at Wave 1 was associated with lower rates of involvement with the juvenile justice system, delinquent peers, internalizing and/or externalizing at Waves 3, 4, and 5, lower rates of drug use at Waves 4 and higher rates of drug use at Wave 5.

Table 3 shows that both mother-stepfather non-violent conflict and mother-stepfather IPV at Wave 1 had unique associations with perceptions of less mattering to mother and less mattering to stepfather at Wave 5. A better mother-non-resident father relationship at Wave 1 was significantly associated with perceptions of more mattering to non-resident father at all waves after Wave 1. Mother’s family income at Wave 1 was associated with perceptions of more mattering to mother at all waves, and more mattering to stepfather at Waves 1 and 2.

Discussion

In this community sample of 81 parents who had lived together when the child was born and who subsequently permanently separated before the child was 12 years old, and in which the mother had re-partnered, about half of the children had experienced the relocation of one of their parents at least an hour’s drive away from the other parent before they were 12. In about half of these 38 relocating cases, the mother relocated with the child, and in half the father relocated without the child. Relocation was associated with a range of negative child outcomes over the next 10 years. The findings provide a conceptual replication (Crandall & Sherman, 2016) of the earlier findings of Braver et al. (2003) and Fabricius and Braver (2006) across substantially different designs, samples, procedures, and analyses. Successful conceptual replications provide important evidence that the findings are generalizable and are not artifacts of one methodology.

Braver et al. (2003) found that relocation of either parent that resulted in the separation of child and father by more than an hour’s drive appeared to pose a long-term risk to children’s relationships with their parents as well as to their mental and behavioral adjustment. The current study yielded similar findings. Pre-adolescent children from stepfather families in which either parent relocated, in comparison to those whose parents did not relocate, were later more likely to report multiple signs of maladjustment. From ages 12.5 to 15.5 they were significantly more likely to harbor doubts about how much they mattered to their non-resident biological fathers, as well as to their mothers and stepfathers with whom they lived. From ages 15.5 to 19.5 they were significantly more likely to engage in high-risk behaviors with potentially serious consequences, including involvement with the juvenile justice system, association with delinquent peers, and drug use. They were also significantly more likely to experience symptoms of depression and anxiety, and to act out in ways symptomatic of aggression and delinquency.

Fabricius and Braver (2006) found that the associations with relocation remained significant after controlling for conflict and violence in the biological parents’ relationships with each other. This is important because a poor inter-parental relationship can potentially function as a “third variable;” i.e., as a cause of both the relocation and the harmful outcomes. In the current longitudinal design, we also controlled for earlier family factors (at age 12.5) in predicting later adolescent and young adult adjustment. The family factor controls included overall quality of the biological parents’ relationships with each other (although not specifically violent and non-violent conflict between them), non-violent conflict and intimate violence between mothers and stepfathers, and mothers’ family income, each of which was associated with several of our measures of harm to children’s mental and behavioral health or relationships with parents. Thus, our findings add to evidence that, over and above these well-known family risks, relocation of either parent poses additional, similar risks to children.

We did not have measures that assessed the motives and circumstances surrounding relocation. There are many valid reasons for relocation, including job and educational opportunities, health, etc. But assessing the reasons to relocate is only important in so far as they might function as “third variables” that caused both the relocation and the negative outcomes. For example, it is hard to see how having a better job opportunity in a distant location could also be the cause of children’s negative outcomes. Fabricius and Braver (2006) found that violence between mother and biological father was related to relocation, but that it did not account for all of the negative outcomes associated with relocation. If parents who relocated faced difficult personal and family circumstances necessitating a move, those circumstances could also account for the negative outcomes, such as a mother moving to care for a sick or dying parent with the accompanying stress to the mother herself, and, presumably, less positive parenting. We found no evidence in the current study that the following control variables were related to parents’ relocation: mother-biological father relationship quality, mother’s family income, mother-stepfather non-violent conflict, mother-stepfather intimate partner violence, age of child at relocation, or age of child at separation. The only difference was that families that did relocate had lower levels of mother-stepfather non-violent conflict. Conceivably, different factors could account for both relocation and negative outcomes in the two different relocation scenarios, mother relocating with the child, and father relocating without the child. In this study, we did not have sufficient sample size to test whether different control variables were related to relocation and outcomes in the two scenarios, but Fabricius and Braver (2006) did, and found that for both scenarios, relocation was associated with negative outcomes over above their control variables.

These three studies give some insight into how relocation might harm children. Relocation of either parent when the child resides with the mother logically should result in less father-child parenting time, and there is evidence of links between parenting time with fathers and father-child relationships, and between father-child relationships and child adjustment (Fabricius, Sokol, Diaz & Braver, 2012; 2016). At Wave 1, adolescents whose parents did not move reported that they had 35% of yearly parenting time with fathers, while those who moved with their mothers reported 28% and those whose fathers moved away reported only 21% (see Fabricius, Braver, Diaz, & Velez, 2010, footnotes 4 and 6 for how parenting time during childhood and adolescence was calculated in the current data set). Studies of parenting time (Fabricius & Luecken, 2007; Fabricius & Suh, 2017) reveal positive associations between more parenting time with fathers, up to and including equal time, and better long-term father-child relationships. Even among non-separated parents, Stevenson et al. (2014) found that the amount of interaction adolescents had with each parent in daily (shopping, cooking) and leisure (playing games, attending events) activities predicted how much they felt they mattered to that parent. Stevenson et al. (2014) hypothesized that such discretionary activities hold meaning for the child regarding the parent’s desire to be with the child and, by extension, the importance of the child to the parent. Thus, one way that relocation can harm the father-child relationship may be by removing a daily source of reassurance and security for the child that the biological father cares and will be there when needed. In the absence of daily reassurance, doubts about the strength of the father-child relationship may enter from many sources, including misinterpreting father’s intentions and hostile framing by others, and doubt can fester from lack of communication. Even without relocation, the non-resident father-child relationship can be at risk, and thus it is noteworthy that in the current study the quality of the relationship between the mother and the biological father was consistently positively associated with mattering to the father regardless of relocation status. This corroborates and extends previous findings with non-separated parents of a positive association between the quality of the parents’ relationships and adolescents’ perceptions of mattering to fathers (Suh et al., 2016), as well as findings that maternal gatekeeping mediated the link between the quality of the parents’ relationships and the amount of interaction adolescents had with their fathers (Stevenson et al., 2014).

The current findings further showed that, in families that relocated, adolescents perceived, rather surprisingly, that they mattered less to their residential mothers and stepfathers as well. Previous studies on parental relocation have not considered that there may be an additional negative impact of relocation on the child’s relationship with the residential mothers and stepfathers with whom the child continued to live. Perhaps, perceiving that their mothers and stepfathers provide insufficient acceptance of and support for their continued relationships with their distant fathers could undermine children’s security about how much they mattered to those mothers and stepfathers. These disruptions in 12- to 15-year-olds’ relationships with their mother, father, and stepfather after they had been separated from their fathers by relocation when they were, on average, 5 years old, could reasonably have contributed to the harmful outcomes we found later at ages 15 to 19. Both emotional security theory (Cummings & Davies, 1996) and attachment theory (Bowlby, 1969) hold that children’s emotional security in the family supports their optimal adjustment. In line with theory, doubts about how much they matter to their parents have been found to predict adolescents’ mental health problems, including symptoms of depression, anxiety, aggression, and delinquency (Rosenberg & McCullough, 1981; Schenck et al., 2009; Suh et al., 2016).

Limitations

Several limitations must be acknowledged and considered in evaluating the findings. First, the sample excluded cases in which the mother did not re-partner by child age 12, and thus the findings cannot be assumed to generalize to those cases. Braver et al. (2003) undoubtedly included both types of cases, but they did not identify them. Relocation that separates children from their fathers might be a more severe risk factor for children with no stepfathers, because there is evidence that stepfather involvement is associated with improved adolescent well-being and decreased depression (Schenck et al., 2009; Suh et al., 2016; Yuan & Hamilton, 2006). In a sample of 1,472 children from a British national longitudinal study, those whose separated mothers had re-partnered fared better in terms of anxiety and aggression than those whose mothers had not re-partnered, and they did not differ from children in intact families (Verropoulou et al., 2002). Future research should include mothers who do not re-partner.

Second, the sample contained only 17 cases, each, in which mother or father moved (four were undetermined), which was too small to provide adequate statistical power to include which parent moved as an independent variable. The impact of relocation could differ in the two cases. When the child moves to a new community it involves changing schools and friends, but when the father moves away it could be experienced as abandonment. However, with a larger sample (about 150 per group), Braver et al. (2003) found essentially no differences in outcomes between the two groups. Future research should continue to explore differences associated with which parent moved.

Third, the dataset had several measurement limitations. Some of the measures included items selected or adapted from standardized scales rather than all of the original items. Using mothers’ reports for predictors and youths’ reports for outcomes is a strength because it eliminates concerns about shared method variance, which was a limitation of Fabricius and Braver (2006), but not including fathers or teachers meant we were limited to understanding only mothers’ and youths’ perspectives. There were also no data about the reasons for the moves. When courts are asked to decide contested cases of child relocation, the reasons, along with viable alternatives, become primary considerations (Parkinson & Cashmore, 2015; Thompson, 2015). Differentiating child outcomes based on reasons for custodial as well as non-custodial parents’ moves is essential for future research. Furthermore, the sample and the data did not allow us to examine cases in which the mother petitioned the court to relocate the child and was denied. Those mothers who nevertheless moved without the child would not be in the sample because one recruitment criterion for PAYS was primary mother residence. Those mothers who stayed would be among the group of non-relocating mothers, and there were no data to identify them. Differentiating child outcomes based on successful versus unsuccessful petitions to relocate the child, in conjunction with the reasons for the moves, will help courts better understand how to serve children’s best interests in relocation cases.

Finally, an inherent limitation in all relocation research is the threat of selection bias. Parents who choose to relocate might differ from those who do not in ways that predispose their children to negative outcomes. However, there are some sources of reassurance that selection bias does not account for all the negative outcomes. These include using covariates to control for parent and family characteristics (e.g., low income, parent conflict, domestic violence) that are known to adversely affect children. Also, the decision to relocate is at least partly determined by factors outside of one’s control, such as employment or educational opportunities, remarriage considerations, or location of extended family. Nevertheless, non-custodial parents who are less committed to their children might be more open to opportunities to move, and their ex-partners might be less worried about moving the children away from them when considering their own possible moves. In our sample, that would mean that relationships with fathers would be negatively impacted, but we found that relocation was associated with youth doubting how much they mattered to all three of their parents. In short, while selection bias cannot be ruled out, a plausible selection bias mechanism is not immediately apparent, while the alternative mechanism, that relocation undermines the child’s emotional security in the family, remains plausible.

Policy and Practice Implications

This conceptual replication of earlier empirical findings on relocation provides an empirical basis to help evaluate different policy alternatives. Under the Best Interests Standard that governs all states’ child custody laws, a rebuttable presumption in favor of relocation implies beneficial outcomes for children associated with permitting relocation, but the empirical findings do not provide such evidence. A different argument could be made that a rebuttable presumption in favor of relocation should require evidence of negative outcomes associated with not permitting relocation. None of the studies to date focused on parents who petitioned to move but were denied, so there are no data directly on this point.

A rebuttable presumption against relocation should require evidence of negative outcomes for children associated with permitting relocation, and the empirical findings consistently provide such evidence. The current findings raise to a new level of seriousness the risks to children posed by relocation. These risks include involvement with delinquent peers and the juvenile justice system, illicit drug use, symptoms of depression, anxiety, and aggression, and doubts about how much they matter to their parents. Furthermore, the empirical findings give policy makers new perspective on what types of factors should be relevant to rebutting the presumption. For example, risks might be mitigated if the father-child relationship has been chronically weak and characterized by sporadic parenting time prior to the move, and the family financial wherewithal exists to maintain long-distance visits that might provide more consistent parenting time. The empirical findings should dissuade courts from giving dispositive weight to the traditional factors that the child would be moving with the primary caregiver and likely to experience increased social capital in the new location. The findings do not include evidence of benefits to the child from the increased social capital that presumably was provided by the relocation cases that have been studied, but rather reveal the risk of harm associated with loss of the child’s “relationship capital” with all three parents. In sum, the absence of empirical findings of benefits to the child’s mental and behavioral health and relationships with parents associated with relocation reveals that the factors that have traditionally been considered in relocation cases, such as continuity of the primary caregiver, improvement to the parent’s life, and enhancement of the child’s opportunities, have not compensated for the risk of harm associated with relocation.

In those cases where the presumption against relocation is not rebutted, the empirical findings further imply that when the court can determine that the parent will not move without the child, the strategic use of conditional change-of-custody orders to discourage moves will result in the best outcome for the child, which is the conclusion also arrived at by Braver et al. (2003). Such orders would have to be contingent on the court finding that the nonmoving parent is not only capable of assuming the responsibilities associated with primary physical custody, but also is at least as likely as the petitioning parent to support the child’s positive relationship with the other parent (Austin, 2015; Warshak, 2013). The empirical findings also imply that in conditional change-of-custody orders courts should not be further required to find that the best interests of the child would be better met by remaining with one parent than by moving with the other, because the risk of harm would remain in either case. When the court can determine that the parent will be forced to move with or without the child, and that the other parent is not able to also move to the new location, there may be no obvious best solution.

Finally, we carry the legal implications one step further. Half of all the relocations in the current study and in Braver et al. (2003) involved the non-custodial parent moving without the child. The empirical findings imply that legal scholars and policy makers should consider how to discourage relocation by non-custodial parents, who are currently free to move away and disrupt the child’s parenting time without having to seek the courts’ permission, and without incurring societal disapproval. While legal prohibitions are almost certainly unconstitutional, courts could require petitions from parents to relocate without the child, accompanied by justifications and revised parenting plans. Some such procedures could allow the educative and normative function of the law and the bench to discourage non-custodial parents from moving away.