psychosocial well-being and the relationship between divorce and children's academic...

14
DANIEL POTTER University of Virginia Psychosocial Well-Being and the Relationship Between Divorce and Children’s Academic Achievement As an unprecedented number of children live in families experiencing divorce, researchers have developed increasingly complex expla- nations for the consequences associated with marital dissolution. Current accounts focus on changes to family finances, destabilized parent- ing practices, elevated parental conflict, and deterioration of the parent – child relationship, to explain the impact of divorce. A less stud- ied explanation draws attention to children’s diminished psychosocial well-being following divorce. Using data from the Early Child- hood Longitudinal Study—Kindergarten cohort (ECLS-K) (N = 10,061), I examined the role of psychosocial well-being in the relationship between divorce and children’s outcomes. The results suggest that divorce is associated with diminished psychosocial well-being in children, and that this decrease helps explain the con- nection between divorce and lower academic achievement. It is estimated that nearly half of all first marriages in the United States end in divorce, and of those divorces half involve a child under 18 years old (Cherlin, 1992; Popenoe, 1996). Department of Sociology, University of Virginia ([email protected]). This article was edited by Jay Teachman. Key Words: child school achievement, divorce, mental health, middle childhood, well-being. Children can respond differently to their parents’ divorce depending on circumstances, such as the age of a child at the time of divorce or level of conflict in the family (Amato, 2001; Stacy, 1996). Nevertheless, divorce generally has a negative association with children’s outcomes (Amato, 2001; Amato & Keith, 1991). In lieu of this detrimental effect, many scholars have sought to identify the mechanisms responsible for this relationship. Researchers have subsequently reconceptualized divorce as a process rather than an event, wherein the changes and experiences accompanying divorce are believed to be the mechanisms primarily responsible for the negative outcomes associated with marital separation (Amato, 2001; Cherlin et al., 1991; Sun, 2001; Sun & Li, 2002). Despite the multitude of possible influences involved in the divorce process, most theory and research has focused on four explanations: changes to family economic resources, declines in parenting quality, high levels of parental conflict, and disruption to the parent – child relationship. Some scholars have theorized a fifth explanation, which emphasizes children’s psychosocial well-being as an additional mech- anism of divorce (Amato, 2000; Hetherington, Bridges, & Insabella, 1998), but research has yet to empirically validate this link. In this study, I examine the connection between divorce and children’s psychosocial well-being, and I test whether well-being serves as a mechanism within the divorce process. In doing so, I address two research questions. Journal of Marriage and Family 72 (August 2010): 933 – 946 933 DOI:10.1111/j.1741-3737.2010.00740.x

Upload: daniel-potter

Post on 21-Jul-2016

331 views

Category:

Documents


18 download

TRANSCRIPT

Page 1: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

DANIEL POTTER University of Virginia

Psychosocial Well-Being and the Relationship

Between Divorce and Children’s Academic

Achievement

As an unprecedented number of children livein families experiencing divorce, researchershave developed increasingly complex expla-nations for the consequences associated withmarital dissolution. Current accounts focus onchanges to family finances, destabilized parent-ing practices, elevated parental conflict, anddeterioration of the parent – child relationship,to explain the impact of divorce. A less stud-ied explanation draws attention to children’sdiminished psychosocial well-being followingdivorce. Using data from the Early Child-hood Longitudinal Study—Kindergarten cohort(ECLS-K) (N = 10,061), I examined the roleof psychosocial well-being in the relationshipbetween divorce and children’s outcomes. Theresults suggest that divorce is associated withdiminished psychosocial well-being in children,and that this decrease helps explain the con-nection between divorce and lower academicachievement.

It is estimated that nearly half of all firstmarriages in the United States end in divorce,and of those divorces half involve a child under18 years old (Cherlin, 1992; Popenoe, 1996).

Department of Sociology, University of Virginia([email protected]).

This article was edited by Jay Teachman.

Key Words: child school achievement, divorce, mentalhealth, middle childhood, well-being.

Children can respond differently to their parents’divorce depending on circumstances, such asthe age of a child at the time of divorce orlevel of conflict in the family (Amato, 2001;Stacy, 1996). Nevertheless, divorce generallyhas a negative association with children’soutcomes (Amato, 2001; Amato & Keith, 1991).In lieu of this detrimental effect, manyscholars have sought to identify the mechanismsresponsible for this relationship. Researchershave subsequently reconceptualized divorce asa process rather than an event, wherein thechanges and experiences accompanying divorceare believed to be the mechanisms primarilyresponsible for the negative outcomes associatedwith marital separation (Amato, 2001; Cherlinet al., 1991; Sun, 2001; Sun & Li, 2002).

Despite the multitude of possible influencesinvolved in the divorce process, most theoryand research has focused on four explanations:changes to family economic resources, declinesin parenting quality, high levels of parentalconflict, and disruption to the parent – childrelationship. Some scholars have theorized afifth explanation, which emphasizes children’spsychosocial well-being as an additional mech-anism of divorce (Amato, 2000; Hetherington,Bridges, & Insabella, 1998), but research has yetto empirically validate this link.

In this study, I examine the connectionbetween divorce and children’s psychosocialwell-being, and I test whether well-being servesas a mechanism within the divorce process.In doing so, I address two research questions.

Journal of Marriage and Family 72 (August 2010): 933 – 946 933DOI:10.1111/j.1741-3737.2010.00740.x

Page 2: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

934 Journal of Marriage and Family

First, is divorce associated with children’s psy-chosocial well-being; second, if so, do changesto psychosocial well-being help explain theassociation between divorce and children’s aca-demic achievement? To answer these questions,I used data from four waves of the Early Child-hood Longitudinal Study—Kindergarten cohort.I focus on academic achievement, because edu-cation predicts a multitude of later life out-comes (Kingston, Hubbard, Lapp, Schroeder, &Wilson, 2003), and early school performancebegets later performance (Entwisle, Alexander,& Olson, 2005). Therefore, investigation intothe relationship between divorce and children’searly academic careers provides insight into thebeginning stages of a process that can haveramifications throughout the life course.

Results from this study support a multifacetedexplanation of divorce’s effects: Changes toeconomic resources, parenting practices, andchild’s psychosocial well-being each helpexplain the impact of divorce on children.In this complex process, psychosocial well-being explains a statistically and substantivelysignificant part of the relationship betweendivorce and academic success, which suggeststhat future research should continue to considerwell-being as part of the explanation for howdivorce influences children’s outcomes.

Background

Divorce is more than the act of marital dissolu-tion; it is a process often marked by instability.As such, many of the changes coinciding withdivorce are believed to be the mechanisms thatcontribute to the negative effect often associatedwith marital dissolution (Cherlin et al., 1991).Divorced families tend to have lower annualincomes (McLanahan & Sandefur, 1994) andfewer material resources (Sun, 2001) than non-divorced families. In addition, parenting qualityis often lower and less consistent in divorcedfamilies (Amato & Booth, 1996; for a recentexception, see Strohschein, 2007). Moreover,parental separation is sometimes accompaniedby premature independence for children whenhousehold rules and regulations break down, andparents befriend rather than parent their children(Dolgin, 1996; Wallerstein & Blakeslee, 1989).Divorce may also result in residential relocationor school change because of custodial arrange-ments, which contributes additional instabilityduring a time already marked with ongoing

uncertainty and potentially upsets a child’s abil-ity to learn (Mehana & Reynolds, 2004). Previ-ous research also has suggested that the elevatedlevels of parental conflict typically precedingmarital dissolution are independently associatedwith children’s outcomes (Davies & Cummings,1994; Kincaid & Caldwell, 1995). Furthermore,divorce has been shown to affect children bydamaging the parent – child relationship (Amato& Sobolewski, 2001). In light of all these factorsassociated with divorce, the relative importanceof any particular mechanism is debatable (Thom-son, Hanson, & McLanahan, 1994), yet mostresearchers agree that each is important forunderstanding the relationship between parentalseparation and children’s outcomes (Amato,2000, 2001; Hetherington et al., 1998).

In addition to the factors outlined above,there is another mechanism that has receivedless attention: psychosocial well-being. Scholarshave proposed that psychosocial well-being maybe a mechanism through which divorce influ-ences children’s outcomes (Amato, 2000; Het-herington et al., 1998), but this claim has gonelargely untested. The majority of researcherslooking at the association between divorce andwell-being have examined well-being as merelyan outcome of divorce (Cherlin et al., 1991;Strohschein, 2005; Woodward, Fergusson, &Belsky, 2000). Lansford et al. (2006) tracedchildren’s psychosocial well-being from theyear prior to three years following their par-ents’ divorce, and they compared internalizingand externalizing trajectories of children fromdifferent age groups at the time of their par-ents’ separation. For elementary school children,divorce was associated with a decline in psy-chosocial well-being that began the year priorto the actual divorce and persisted for two yearsbefore recovering. For high school children,their psychosocial well-being actually improvedleading up to and following divorce, althoughthis upward trend declined within two years.Regardless of age or trajectory, both groupsof divorce children had poorer psychosocialwell-being than did the nondivorce comparisongroup. Divorce appears to be related to children’spsychosocial well-being, but it remains unclearwhether this relationship can help explain theassociation between divorce and other outcomesin childhood.

Divorce is associated with numerous out-comes in children’s lives, one of which isacademic achievement. Children from divorced

Page 3: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

Divorce and Children’s Psychosocial Well-Being 935

families tend to do less well in school than chil-dren from nondivorced families (Amato, 2001;Martinez & Forgatch, 2002; Tillman, 2007). Instudying the connection between divorce andchildren’s schooling, the role of psychosocialwell-being may be particularly relevant giventhe role of socioemotional health in academicsuccess (Becker & Luthar, 2002; Lane, Carter,Pierson, & Glaeser, 2006). Roeser, Eccles, andFreedman-Doan (1999) used cluster analysis tocategorize high school students according topsychosocial well-being and academic motiva-tion. The vast majority of students were labeled‘‘well-adjusted,’’ but a segment of students ratedlow for psychosocial well-being and high foracademic motivation. Students in that group,despite academic motivation, had the largestdecline in grade point average from eighth toninth grade. McLeod and Fettes (2007) per-formed a similar analysis and found that childrenwith poorer psychosocial well-being typicallyhad lower rates of high school graduation.Psychosocial well-being is central to educa-tional achievement (Blackorby & Wagner, 1996;Needham, Crosnoe, & Muller, 2004); therefore,the impact of divorce on children’s psychoso-cial well-being may be especially important forthe association between divorce and academicoutcomes.

Divorce comprises several processes, suchas elevated parental conflict and dimin-ished economic resources, parenting practices,parent – child relations, and child’s psychoso-cial well-being. In combination, these factorshave sometimes been referred to as the gen-eral theory of divorce (Amato, 2000). Priorresearch has provided strong evidence for howchanges to economic resources and parentingpractices help explain the consequences associ-ated with divorce; however, much less attentionhas been given to the potential role of psy-chosocial well-being. I examine the role ofpsychosocial well-being in the divorce process,and in doing so, I provide a more complex, andarguably more complete, understanding of howdivorce affects children’s outcomes.

METHOD

Data

Data for this analysis come from the fifth-gradepublic use longitudinal data file of the EarlyChildhood Longitudinal Study – Kindergarten

cohort (ECLS-K). Founded by the NationalCenter for Education Statistics (NCES), theECLS-K is a seven-wave panel study of morethan 20,000 children begun in the fall of 1998(Tourangeau, Nord, Le, Pollack, & Atkins-Burnett, 2006). The sample was designed asa stratified random sample with students drawnfrom schools, which were drawn from counties.Information was collected from child, parent,teacher, and school administrator to create anencompassing view of children’s family andeducational status and progress. The ECLS-Khas seven waves of data; however, I use fourwaves (Spring Kindergarten, Spring Grade 1,Spring Grade 3, and Spring Grade 5) in thisstudy because of variable availability and sampleattrition in the other three waves.

Although the base year of the ECLS-K hasinformation on more than 20,000 students, thefifth grade public use data file did not includeevery child. As a result, the initial number ofrespondents in the sample was 17,565. From thisbeginning sample, cases were dropped if theyhad a nonpositive value for the sampling weight(n = 472), and in line with the selection criteriaof previous studies (e.g., Cherlin et al., 1991),cases were excluded if the child’s two biologicalparents were not married at the analytic baseline(n = 6,847), or if the parent reported beingmarried at the baseline but in a subsequentwave reported being never married (n = 108)or widowed (n = 77). These selection filters arestringent but suitable for isolating predivorceconditions, and they most accurately estimatethe effect associated with the timing of divorce.The final sample consists of 10,061 children, andof those children, 870 experienced their parents’divorce between spring 1999 and spring 2004.

Variables

Descriptive statistics are provided separately forthe time-varying and time-invariant variables.Table 1 contains the descriptive characteristicsfor all the time-varying measures at eachwave, and Table 2 contains the time-invariantmeasures.

The measure of psychosocial well-being wascalculated using teacher and parent or childreports of children’s internalizing problems,externalizing problems, and social skills. Parentand teacher reports were used in kindergartenand Grade 1, and child and teacher reports were

Page 4: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

936 Journal of Marriage and Family

Table 1. Divorce and Children’s Psychosocial Well-Being: Time-Varying Variables and Descriptive Statistics

Kindergarten Grade 1 Grade 3 Grade 5

Level 1 Mean SD Mean SD Mean SD Mean SD

Math Divorce 32.49 10.21 57.71 15.66 91.49 20.32 112.56 20.36Nondivorce 35.25 11.99 60.54 17.10 96.13 21.14 117.34 20.82

Reading Divorce 39.33 11.26 70.10 20.86 116.54 25.35 137.69 23.49Nondivorce 42.58 14.50 74.89 22.83 122.20 24.74 142.82 22.58

Psychosocial Divorce 3.23 0.33 3.21 0.35 3.06 0.43 3.09 0.42Nondivorce 3.27 0.31 3.27 0.32 3.18 0.39 3.21 0.39

Economic resourcesIncome Divorce 10.53 0.92 10.53 0.82 10.46 0.78 10.38 0.83

Nondivorce 10.79 0.96 10.89 0.75 10.96 0.75 11.01 0.77Books Divorce 4.02 0.91 4.14 0.98 4.27 0.94 4.02 0.98

Nondivorce 4.12 0.97 4.32 1.00 4.46 0.96 4.28 1.02

Parenting practicesFamily routine Divorce 5.61 1.77 5.02 1.71 4.68 1.71 4.50 1.61

Nondivorce 5.91 1.73 5.39 1.60 5.17 1.58 4.92 1.56School change Divorce 0.05 0.21 0.10 0.30 0.27 0.45 0.38 0.49

Nondivorce 0.02 0.14 0.10 0.30 0.25 0.43 0.41 0.49Parental school Divorce 4.28 1.67 4.56 1.66 4.76 1.58 4.55 1.54

Nondivorce 4.58 1.59 4.94 1.55 5.12 1.50 4.95 1.52Child activities Divorce 1.58 1.31 1.87 1.45 2.00 1.43 2.04 1.40

Nondivorce 1.82 1.36 2.17 1.51 2.34 1.44 2.36 1.48

Note: All estimates are weighted using the spring kindergarten sampling weight.

used in Grade 3 and Grade 5. Internalizing prob-lems (e.g., child is sad, lonely), externalizingproblems (e.g., child is impulsive, quick tem-pered), and social skills (e.g., child gets alongwell with peers) are indices provided by theECLS-K based on a copyrighted survey instru-ment. As a result, the individual questions usedby the ECLS-K to construct the indexes were notavailable. The ECLS-K user’s manual describesthe variable construction process and reports thateach index has relatively high internal validity(α > 0.70) (for discussion of index construction,see Tourangeau et al., 2006).

To compute the measure of well-being,the internalizing and externalizing problemvariables were inversely coded so that a higherscore indicated better well-being (e.g., lowerlevels of problems). Then the six indices (threefrom each informant) were averaged to generatewave-specific measures of psychosocial well-being. The psychosocial well-being measuredoes not replicate anything previously used,but its underlying components reflect many ofthe traits examined in earlier studies (Cohen,Kamarck, & Mermelstein, 1983; Lansford et al.,2006; Strohschein, 2005; Sun, 2001). The

term is considered psychosocial, because itdoes not represent a psychometric indicator ofinternal state—nor does it provide an entirelybehaviorist measure. The concept synthesizesindividuals’ psychological state with theirbehaviors and actions forming a psychosocialfactor (Aronson, 1984).

One possible limitation of the psychosocialwell-being measure in this study is its relianceon different informants across waves; however,the internal reliability of the measure remainedhigh for all time points (α = 0.78 – 0.82).Furthermore, a bivariate comparison acrosswaves revealed a moderate to strong relationshipover time (analyses not shown), and the strengthof relationship diminished as the time betweenwaves increased. Assuming that children’s well-being changes with time (regardless of divorce),I would expect this pattern of moderate tostrong association between time points, with anautoregressive structure over time. On the basisof these conditions, it does not appear that the useof different informants over time compromisedthe comparability of the psychosocial well-beingmeasure.

Page 5: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

Divorce and Children’s Psychosocial Well-Being 937

Table 2. Divorce and Children’s Psychosocial Well-Being:Time Invariant Variables and Descriptive Statistics

Divorce Nondivorce

Level 2 (n = 870) (n = 9,190)

Demographics Mean SD Mean SD

Female 0.48 0.50 0.49 0.50

African American 0.09 0.28 0.05 0.22

Hispanic 0.18 0.38 0.16 0.36

Other 0.08 0.28 0.11 0.32

Age 74.75 4.48 74.73 4.42

Parental depressive 1.52 0.49 1.38 0.39

Parent’s education

Some college 0.37 0.48 0.31 0.46

Bachelor’s degree 0.16 0.37 0.25 0.43

Graduate work 0.11 0.31 0.21 0.40

Divorce mechanisms (kindergarten value reported for

time-varying covariates)

Income 10.58 0.94 10.83 0.92

Books 4.04 0.90 4.13 0.96

Family routine 5.60 1.78 5.93 1.74

Parental schoolinvolvement

4.39 1.64 4.63 1.59

Child activities 1.60 1.31 1.83 1.36

Psychosocial well-being 3.23 0.33 3.27 0.31

Divorce mechanisms (time-invariant from kindergarten

wave)

Marital quality 3.04 0.63 3.27 0.44

Marital conflict 1.98 0.46 1.79 0.41

Challenges 1.84 0.36 1.84 0.36

Warmth 0.74 0.44 0.74 0.44

Like 0.80 0.40 0.80 0.40

Show 0.62 0.49 0.54 0.50

Express 0.92 0.27 0.91 0.29

Level 3 (n = 1,026)

Mean SD

Region

Northeast 0.18 0.39

Midwest 0.25 0.43

West 0.25 0.43

Area

Suburban 0.38 0.49

Rural 0.21 0.41

Public school 0.76 0.43

School size 465.65 250.82

% minority students 36.87 32.16

Note: All estimates are weighted using the springkindergarten sampling weight.

Reading scores and math scores were usedto measure children’s academic achievementfrom kindergarten through Grade 5. Scores werebased on tests administered as part of the studentassessment in the ECLS-K. The NCES used itemresponse theory to compute a score at each wave,which could be compared between students andacross time (Tourangeau et al., 2006).

Divorce was coded as a time-varyingindicator variable. At the baseline wave, eachchild was coded as 0, as the child’s biologicalparents had to be married for inclusion inthe analysis. For each wave that the parentsremained married, the child received a valueof 0. If a child’s parents divorced, the childreceived a value of 1 for the first wave in whichthe divorce was recorded and every subsequentwave (Singer & Willett, 2003).

To isolate the influence of psychosocialwell-being on children’s academic achievement,other divorce mechanisms were included inthe model (e.g., economic resources, parentingpractices, parental conflict, parent – child rela-tionship). Both time-varying and time-invariantmeasures of the other divorce mechanisms wereincluded by using the score from the kinder-garten wave whenever possible. Having bothtime-varying and time-invariant covariates dis-tinguished between the effects associated withchanges to those factors (e.g., decreases infamily income) and effects attributable to differ-ences in predivorce status (e.g., family incomeat kindergarten) (Fitzmaurice, Laird, & Ware,2004). Unfortunately, data for the divorce mech-anisms were not always collected at multipletime points; thus, family economic resourcesand parenting practices have both time-varyingand time-invariant measures of divorce, whereasparental conflict and parent – child relationshipare time-invariant only.

Economic resources comprise family incomeand number of books in the household.To compute income, the categorical ECLS-Kmeasure was converted to dollars by taking themiddle value of each category. For example, if anindividual reported a Category 9 income level,‘‘between $40,000 and $50,000,’’ the middlevalue of $45,000 was substituted. Books wasbased on parents’ estimates of the number ofbooks in the household or borrowed from thelibrary. The distribution of both variables waspositively skewed; therefore, the natural log ofeach was used in the analysis.

Page 6: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

938 Journal of Marriage and Family

Four measures of parenting practice areincluded in the models: family routines, schoolmobility, parental school involvement, andchild’s activities. Family routine and schoolmobility are two measures of parenting con-sistency in the daily operations of the family.Family routine is a count variable (Law, Wong,& Mobley 1998) derived from eight ques-tions related to structured home environment(dinner together, breakfast together, dinner atsame time, breakfast at same time, regular bed-time, television programming rules, amount oftelevision watched rules, same house). The sec-ond measure of parenting consistency is schoolmobility, which is an indicator variable if thechild switched schools from one wave to thenext (1 = changed schools; 0 = same school).

Parental school involvement captured parentparticipation with the child’s school. In total,seven questions were used to create the countvariable (contacted school, parent – teacher con-ference, open house, parent – teacher associationmeeting, volunteered at school, attend schoolevent, and fund-raising efforts). Child activitiesreflected the different activities a child couldbe involved with outside of school. Given thechild’s age, any and all activities in which thechild participated were most likely at the insis-tence and mercy of the parent; therefore, changeto a child’s level of participation most likelyreflects change in parenting. The ECLS-K askedabout children’s involvement in seven activities(visited library, dance lessons, athletic team, artlessons, member of club, music lessons, per-forming arts club).

Parental conflict and parent – child relation-ship are time-invariant divorce mechanismsmeasured during the kindergarten wave. Parentalconflict was based on two measures: mari-tal quality and marital conflict. Marital qualitymeasured the strength of the parents’ relation-ship and was based on six items (α = .737)(tell troubles, laugh together, calm conversa-tions, work together, exchange ideas, subjectivemarriage quality rating). Marital conflict mea-sured parents’ relationship problems on the basisof 13 questions (α = .759) (keep opinion toself, heated arguing, physical aggression, argueabout chores, argue about children, argue aboutmoney, argue about affection, argue about sex,argue about religion, argue about leisure time,argue about alcohol, argue about infidelity, argueabout in-laws).

Parent – child relationship variables werebased on questions asked of parents regardingattitudes toward their child. Individual variableswere included in the models for level of parentalwarmth, parents making time to be with theirchild, child’s enjoyment of parent’s company,and expressed love between parent and child.These variables were originally coded from 1to 4 (1 = completely true; 4 = not true at all);however, because of the skewedness of theirdistributions, each was converted to a dummyvariable with any answer other than ‘‘completelytrue’’ as the reference. In addition, a compositevariable of parenting challenges was used tomeasure parent – child relationship and is basedon questions related to the child-rearing strainsthat parents face. In total, the challenges measurecomprised nine questions (α = .690) (too busyfor child, hard to always show child affection,being parent harder than expected, child doesthings that bother parent, sacrifice own needs tomeet child’s needs, feel trapped as parent, angrywith child, child harder to care for than expected,parenting more work than pleasure). Because theparental conflict and parent – child relationshipvariables are time-invariant, I cannot estimatethe impact that changes to those measureshave on children’s academic achievement.This limitation notwithstanding, time-invariantcontrols are better than no controls, and theirinclusion provides a more complete modelwithin which to study the role of psychosocialwell-being in the divorce process.

Control Variables

A set of control variables was also includedin the analysis. Control variables includedbasic demographic information such as gen-der (1 = female), race (1 = African American,1 = Hispanic, 1 = other, and 0 = White), andage at spring semester of kindergarten (inyears). Controls were also included for par-ent’s education and parental depression becauseof their relationship with divorce and children’seducational attainment (Kincaid & Caldwell,1995; Morrison & Cherlin, 1995). Last, kinder-garten school control variables include type (1 =public), size, percentage of minority students,region (1 = Northeast, 1 = Midwest, 1 = West,and 0 = South), and urbanicity (1 = suburban,1 = rural, and 0 = city).

Page 7: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

Divorce and Children’s Psychosocial Well-Being 939

Time

The ECLS-K assessed children across a rangeof dates throughout the spring semester of eachwave. To model this sample design feature,child-specific time measures were computed bysetting the grand mean test date of the springkindergarten assessment to 0 and subtractingall waves from that time point. The timevariable thus adjusts for variability in initialassessment date (e.g., a child assessed priorto the mean assessment date in kindergartenreceived a negative value, and a child assessedafter received a positive value) and the differencein the number of days between assessmentoccasions (e.g., the number of days betweenchild i’s assessment occasion at a later wave andthe mean assessment date for spring kindergartenis unique to child i). Time was initially computedin days, but to increase the relative size of theestimated coefficient, it was converted to years.

Missing Values

Missing values were dealt with using multipleimputation via the ICE command (Royston,2007) in STATA. Five data sets were imputed,using all the variables contained in the modelto estimate missing values. Because of thelongitudinal nature of the data, imputationwas performed in the person-file format andthen transposed to a person-period layout. Toensure that certain variables took only positivevalues during the imputation process (e.g.,age), the natural log of all nonnegative intervalvariables was taken prior to imputing and thenexponentiated to original scale for analysis. Thefinal sample included 40,244 observations, from10,061 children, nested in 1,026 kindergartenschools.

Analytic Strategy

Multilevel mixed-effect modeling (Singer &Willett, 2003) was used to assess whetherdivorce was associated with children’s psy-chosocial well-being and whether changes topsychosocial well-being helped explain the asso-ciation between divorce and academic achieve-ment. Mixed-effect modeling was chosen overalternative strategies, such as fixed-effect mod-els, because not all divorce mechanisms weremeasured over time, and mixed-effect modelsare able to distinguish between effects associated

with time-invariant (e.g., predivorce) factors,such as marital conflict or parent – child rela-tionship, and the effects associated with time-varying factors, such as economic resources andpsychosocial well-being. In addition, the three-level design corrects for the clustered natureof the data (e.g., time nested within child andchild nested within school). Finally, in line withprevious research (Votruba-Drzal, Li-Grining,& Maldonado-Carreno, 2008), I performed myanalyses with HLM software, because it handlesmultiply imputed data and can apply samplingweights to the analysis.

Children’s time-varying measures were esti-mated at Level 1 (n = 40,244), which werenested within children at Level 2 (n = 10,061),who were nested within kindergarten schoolsat Level 3 (n = 1,026). Preliminary analysesrevealed a nonlinear trend in children’s psy-chosocial well-being, as well as their math andreading scores from kindergarten to Grade 5. Asa result, TIME and TIME2 were included in eachmodel to estimate the functional form of theoutcome measures. The full model predictingchildren’s math and reading is as follows:

Level 1:

Ytij = π0ij + π1ij(TIMEtij) + π2ij(TIMEtij)2

+ π3ij(DIVORCEtij) + π4ij(PSYCtij)

+ π5ij(DIV MECHtij) + etij.

Level 2:

π0ij = β00j + β01jDEMOij + β02jPEij

+ β03jPDij + β04jDMkij

+ β05jDMij + r0ijπ1ij

= β10j + β11jDEMOij.

Level 3:

β00j = γ000+γ001SCHj + γ002Uj

+ γ003Rj + u2j.

TIME represented the number of yearssince the grand mean test date of the springkindergarten assessment, TIME2 was used toestimate the nonlinear change over time, andDIVORCE was a time-varying indicator ofparent’s marital status. In addition, the fullLevel 1 model included PSYC, which is the

Page 8: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

940 Journal of Marriage and Family

time-varying measure of psychosocial well-being, and DIV MECH, which is a vectorof the time-varying divorce mechanisms (e.g.,economic resources and parenting practices). AtLevel 2, DEMO is a vector of demographicvariables, PE is a vector of parent’s educationvariables, and PD is the measure of parentaldepression at kindergarten. In addition, DMkis a vector containing the time-invariant formsof the time-varying divorce measures (e.g.,kindergarten wave values), and DM is avector of divorce mechanisms collected onlyat kindergarten (e.g., parental conflict andparent – child relationship). For Level 3, SCHis a vector of kindergarten school characteristics(e.g., school type, school size, percentage ofminority students), R is a vector of regionvariables, and U is a vector of urbanicityvariables. The π0ij represents the score forthe ith individual at TIME = 0; π1i is theinstantaneous growth rate for the ith individualat TIME = 0; and π2ij captures curvature ofindividual growth trajectory. The π3ij is thedifference in the mean level of the outcomevariable for an individual whose parents divorceat wave t and each subsequent wave. The π4ijis the instantaneous change in the mean mathor reading score associated with a one-unitchange in children’s psychosocial well-being.Finally, π5ij is a vector of coefficients associatedwith the time-varying divorce mechanisms andrepresents the instantaneous change in themean math or reading score associated witha one-unit change in the divorce mechanismmeasures. A similar model was used toestimate children’s psychosocial well-being.Because of the large number of variables ineach of the models, the tables report onlyselected variables. The full list of variablesin each model is provided below each table.I used three models to estimate the associationbetween divorce and psychosocial well-being,and four models to estimate the relationshipbetween divorce and children’s reading andmath scores. The three psychosocial well-beingmodels represent increasingly stringent tests ofthe association between divorce and well-being.The four academic achievement models wereused to examine whether the introduction ofpsychosocial well-being helped explain the gapin children’s academic achievement typicallyassociated with parental divorce. Consequently,a Sobel test was used to estimate whether anyreduction in the divorce coefficient occurring

with the introduction of psychosocial well-beinginto the model was statistically significant.

RESULTS

Descriptive Differences

Examination of the descriptive statistics inTable 1 reveals important differences andsimilarities between children of divorced andnondivorced parents. Specifically, children ofdivorced parents have an average score about3 points lower than their nondivorced peers inmath and reading on the spring kindergartenassessment. This gap suggests that academicdifferences between divorced and nondivorcedchildren exist prior to the occurrence of theactual divorce. By fifth grade, the gap betweendivorce and nondivorce children increased to 5points, which indicates that divorce may slightlyexacerbate preexisting academic inequalities.Interestingly, there was no kindergarten gapbetween divorced and nondivorced childrenwith regard to their level of psychosocialwell-being. At kindergarten, children whoseparents eventually divorce had an average well-being score 0.04 points lower than nondivorcedchildren, a difference of about one tenthof a standard deviation. By Grade 5, thegap in psychosocial well-being expanded to0.12 points, which is nearly one third ofa standard deviation. These two descriptivepatterns suggest that divorce widens an alreadypresent gap in children’s academic achievementand may do so by increasing disparitiesin children’s psychosocial well-being. Thefollowing analyses test this proposition.

Psychosocial Well-Being

Three mixed-effect models helped answer thefirst research question: Does divorce impactchildren’s psychosocial well-being? Model 1presents the baseline relationship betweendivorce and psychosocial well-being (seeTable 3). The trend in children’s psychoso-cial well-being was downward throughout muchof the elementary school years, and divorcediminished children’s well-being even further.Divorce was negatively associated with psy-chosocial well-being: Children who experiencedparental separation had a decline in their well-being score of 0.07 points, a decrease of nearlyone fifth of a standard deviation. Model 2 con-trols for demographic characteristics, parental

Page 9: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

Divorce and Children’s Psychosocial Well-Being 941

Table 3. Association Between Divorce and Children’s Psychosocial Well-Being (40,244 Measurements, From 10,061Children, in 1,026 Schools)

Model 1a Model 2b Model 3c

b SE B SE b SE

Level 1 (time varying)Divorce −0.07∗∗∗ 0.01 −0.06∗∗∗ 0.01 −0.05∗∗∗ 0.01Intercept 3.32∗∗∗ 0.02 0.71∗∗∗ 0.06 0.73∗∗∗ 0.06Time −0.05∗∗∗ 0.00 0.00 0.02 0.00 0.02Time sq. 0.01∗∗∗ 0.00 0.01∗∗∗ 0.00 0.01∗∗∗ 0.00

Note: All estimates are weighted using the spring kindergarten sampling weight. aLevel 1 controls: none; Level 2 controls:none; Level 3 controls: region (1 = Northeast, 1 = Midwest, and 1 = West), area (1 = suburb, and 1 = rural), schooltype (1 = public), school size, and percentage of minority students. bLevel 1 controls: none; Level 2 controls, intercept:gender (1 = female), race (1 = African American, 1 = Hispanic, and 1 = other), age at kindergarten assessment, parentaldepressive score, parent’s education (1 = some college, 1 = bachelor’s degree, 1 = graduate work or more), householdincome at kindergarten (natural log), number of books in the household (natural log), family routines at kindergarten,parental school involvement at kindergarten, child activities at kindergarten, psychosocial well-being at kindergarten. Level2 controls, time slope: gender (1 = female), race (1 = African American, 1 = Hispanic, and 1 = other), age at kindergartenassessment. Level 3 controls: region (1 = Northeast, 1 = Midwest, and 1 = West), area (1 = suburb, and 1 = rural), schooltype (1 = public), school size, and percentage of minority students.cLevel 1 controls: household income (natural log), numberof books in household (natural log), family routines, school mobility (1 = changed schools), parental school involvement,child activities. Level 2 controls, intercept: gender (1 = female), race (1 = African American, 1 = Hispanic, and 1 = other),age at kindergarten assessment, parental depressive score, parent’s education (1 = some college, 1 = bachelor’s degree,1 = graduate work or more), household income at kindergarten (natural log), number of books in the household (natural log),family routines at kindergarten, parental school involvement at kindergarten, child activities at kindergarten, psychosocialwell-being at kindergarten. Level 2 controls, time slope: gender (1 = female), race (1 = African American, 1 = Hispanic,and 1 = other), age at kindergarten assessment. Level 3 controls: region (1 = Northeast, 1 = Midwest, and 1 = West), area(1 = suburb, and 1 = rural), school type (1 = public), school size, and percentage of minority students.

† p < .10. ∗p < .05. ∗∗p < .01. ∗∗∗p < .001.

education, and divorce mechanisms in kinder-garten, and the coefficient for divorce wasreduced but remained statistically significant.Finally, Model 3 introduced the full set of time-varying and time-invariant divorce mechanismsto examine the robustness of the relationshipbetween divorce and children’s psychosocialwell-being. Other divorce mechanisms wererelated to psychosocial well-being, but theydid not explain away the negative associationbetween divorce and well-being. On the basisof these results, divorce appears to be associ-ated with a decline in children’s psychosocialwell-being. These findings corroborate theory(Amato, 2000) and prior research on the rela-tionship between parental divorce and children’sinternalizing and externalizing problems (Lans-ford et al., 2006; Strohschein, 2005).

Children’s Academic Achievement

Four mixed-effect models examined the relation-ship between divorce and children’s math and

reading scores. Model 1 estimated the averageinfluence of divorce on reading and math scores.During the first wave in which a divorce wasexperienced and each subsequent wave there-after, children of divorced parents scored 1.3points and 1.7 points lower than children fromnondivorced parents in math and reading, respec-tively; a difference of about one tenth of astandard deviation. This decline in math andreading scores is associated with the timing ofthe divorce and does not reflect any predivorcedifferences in children’s test scores that mayhave existed between the two groups of children.As such, the coefficients suggests that childrenlose ground relative to their peers immediatelyfollowing their parents’ divorce, and the gappersists throughout the elementary school years.To test whether psychosocial well-being canhelp explain the decline associated with divorce,Model 2 added psychosocial well-being, Model3 introduced the other divorce mechanisms, andModel 4 combined Model 2 and 3 to examine

Page 10: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

942 Journal of Marriage and Family

the role of psychosocial well-being within thebroader divorce process.

Model 2 addressed the second research ques-tion (i.e., does psychosocial well-being explainthe relationship between divorce and children’soutcomes). If divorce affects children’s psy-chosocial well-being, which then influenceschildren’s outcomes, then the inclusion of well-being in Model 2 should reduce the divorcecoefficient. Indeed, the inclusion of psychoso-cial well-being reduced the estimated effectof divorce by 17% in the math model (see

Table 4) and 18% in the reading model (seeTable 5). Although these are modest reduc-tions, the inclusion of psychosocial well-beingexplained enough of the gap to render the divorcecoefficient only marginally statistically signifi-cant in math and reading (p < .10). Moreover,a Sobel test of the reduction in the divorcecoefficient associated with the inclusion of psy-chosocial well-being was statistically significantin the math and reading models (p < .001).

Psychosocial well-being explained a nontriv-ial part of the divorce coefficient; however,

Table 4. Divorce and Psychosocial Well-being HLM Results for Children’s Reading Scores (40,244 Measurements, From10,061 Children, Within 1,026 Schools)

Model 1 Model 2 Model 3 Model 4

b SE b SE b SE b SE

Level 1 (time varying)Divorce −1.67∗ 0.74 −1.37† 0.74 −1.50∗ 0.74 −1.23† 0.74Psychosocial well-being 5.82∗∗∗ 0.41 5.80∗∗∗ 0.41Divorce mechanisms

Income (ln) 0.46∗ 0.21 0.41† 0.21Books (ln) 0.47∗∗ 0.16 0.46∗∗ 0.16Family routine −0.15† 0.08 −0.16† 0.08School change −0.83∗ 0.34 −0.86∗ 0.34Parental school involvement −0.02 0.09 0.00 0.09Child activities 0.36∗∗ 0.10 0.35∗∗ 0.10

Level 2 (kindergarten)Divorce mechanisms

Parental conflictMarital quality −1.56∗∗ 0.45 −1.55∗∗ 0.44Marital conflict 0.24 0.50 0.21 0.49

Parent-child relationshipChallenges 2.04∗∗ 0.56 2.06∗∗∗ 0.56Warmth −0.45 0.46 −0.42 0.46Make time −2.21∗∗∗ 0.39 −2.13∗∗∗ 0.39Company 0.16 0.50 0.15 0.50Express 0.05 0.68 0.09 0.67

Intercept −64.02∗∗∗ 5.28 −68.71∗∗∗ 5.24 −66.28∗∗∗ 5.62 −70.70∗∗∗ 5.57Time 41.60∗∗∗ 0.92 41.60∗∗∗ 0.92 41.22∗∗∗ 0.94 41.23∗∗∗ 0.94Time sq. −3.18∗∗∗ 0.04 −3.22∗∗∗ 0.04 −3.14∗∗∗ 0.04 −3.18∗∗∗ 0.04

Note: All analyses are weighted using the spring kindergarten sampling weight. All reported coefficients are Level 1variables. Each model includes the following control variables. Level 1 controls: none; Level 2 controls, intercept: gender(1 = female), race (1 = African American, 1 = Hispanic, and 1 = other), age at kindergarten assessment, parental depressivescore, parent’s education (1 = some college, 1 = bachelor’s degree, 1 = graduate work or more), household income atkindergarten (natural log), number of books in the household (natural log), family routines at kindergarten, parental schoolinvolvement at kindergarten, child activities at kindergarten, psychosocial well-being at kindergarten. Level 2 controls, timeslope: gender (1 = female), race (1 = African American, 1 = Hispanic, and 1 = other), age at kindergarten assessment. Level3 controls: region (1 = Northeast, 1 = Midwest, and 1 = West), area (1 = suburb, and 1 = rural), school type (1 = public),school size, and percentage of minority students.

†p < .10. ∗p < .05. ∗∗p < .01. ∗∗∗p < .001.

Page 11: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

Divorce and Children’s Psychosocial Well-Being 943

Table 5. Divorce and Psychosocial Well-being HLM Results for Children’s Math Scores (40,244 Measurements, From10,061 Children, Within 1,026 Schools)

Model 1 Model 2 Model 3 Model 4

b SE b SE b SE b SE

Level 1 (time varying)Divorce −1.27∗ 0.55 −1.05† 0.54 −1.07∗ 0.54 −0.86 0.54Psychosocial well-being 4.39∗∗∗ 0.41 4.38∗∗∗ 0.42Divorce mechanisms

Income (ln) 0.59∗∗ 0.19 0.56∗∗ 0.19Books (ln) 0.17 0.14 0.16 0.14Family routine −0.16∗ 0.06 −0.17∗∗ 0.06School change −0.48∗ 0.24 −0.50∗ 0.24Parental school involvement 0.09 0.08 0.10 0.08Child activities 0.38∗∗ 0.10 0.38∗∗ 0.10

Level 2 (kindergarten)Divorce mechanisms

Parental conflictMarital quality −1.84∗∗∗ 0.39 −1.83∗∗∗ 0.38Marital conflict −0.35 0.46 −0.37 0.45

Parent – child relationshipChallenges 1.36∗∗ 0.45 1.38∗∗ 0.44Warmth −0.40 0.40 −0.38 0.40Make time −2.09∗∗∗ 0.34 −2.03∗∗∗ 0.34Company −0.19 0.44 −0.20 0.43Express −0.37 0.56 −0.35 0.56

Intercept −74.04∗∗∗ 4.05 −77.58∗∗∗ 4.00 −72.81∗∗∗ 4.61 −76.16∗∗∗ 4.56Time 35.01∗∗∗ 0.85 35.01∗∗∗ 0.86 34.60∗∗∗ 0.88 34.61∗∗∗ 0.89Time sq. −1.96∗∗∗ 0.03 −2.00∗∗∗ 0.03 −1.93∗∗∗ 0.03 −1.96∗∗∗ 0.03

Note: All analyses are weighted using the spring kindergarten sampling weight. All reported coefficients are Level 1variables. Each model includes the following control variables. Level 1 controls: none; Level 2 controls, intercept: gender(1 = female), race (1 = African American,1 = Hispanic, and 1 = other), age at kindergarten assessment, parental depressivescore, parent’s education (1 = some college, 1 = bachelor’s degree, 1 = graduate work or more), household income atkindergarten (natural log), number of books in the household (natural log), family routines at kindergarten, parental schoolinvolvement at kindergarten, child activities at kindergarten, psychosocial well-being at kindergarten. Level 2 controls, timeslope: gender (1 = female), race (1 = African American, 1 = Hispanic, and 1 = other), age at kindergarten assessment. Level3 controls: region (1 = Northeast, 1 = Midwest, and 1 = West), area (1 = suburb, and 1 = rural), school type (1 = public),school size, and percentage of minority students.

†p < .10. ∗p < .05. ∗∗p < .01. ∗∗∗p < .001.

it is possible that the explanation attributedto psychosocial well-being is merely the by-product of other co-occurring factors involvedin the divorce process. To test this possibil-ity, Model 3 removed psychosocial well-beingand introduced the divorce mechanisms typ-ically considered in previous analyses (e.g.,economic resources, parental practices, parentalconflict, parent – child relationship). These fac-tors reduced the divorce coefficient by 16%in the math model and 10% in the readingmodel (relative to Model 1), although in both

models, divorce remained statistically signifi-cant. Model 4 provided a more stringent testof the second research question by estimat-ing whether psychosocial well-being predictedacademic achievement and helped explain theassociation between divorce and children’s mathand reading scores controlling for other divorcemechanisms. Indeed, psychosocial well-beingremained statistically significant (p < .001) andreduced the divorce coefficient by an additional19% and 18% in the math and reading models,respectively (relative to Model 3). In the math

Page 12: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

944 Journal of Marriage and Family

models, divorce was no longer marginally signif-icant (p > .10) and was only barely marginallysignificant in the reading model (p = .098).The coefficients of the other divorce mecha-nisms changed little from Model 3 to Model4, which suggests that the relationship betweenpsychosocial well-being and the other divorcefactors described in Table 3 does not appearto explain the association between well-beingand children’s academic outcomes. To thatend, psychosocial well-being has an importantand complementary role in understanding howdivorce influences children’s math and readingscores, and together with other mechanisms inthe divorce process, it explained the divorce gapin math and reading scores over time.

DISCUSSION

Divorce alters the lives of children, and thosechanges are largely responsible for the effectstypically associated with parental separation.The findings reported in this study indicatethat divorce diminishes children’s psychosocialwell-being, and the decline in well-being helpsexplain the poorer performance of divorcedchildren at school. In addition, psychosocialwell-being was an important predictor ofacademic achievement independent of otherdivorce mechanisms typically considered inprevious research. Moreover, the orthogonalrelationship between the different divorcemechanisms suggests that the consequences ofdivorce are not primarily driven by a singlefactor but result from an array of independent,yet equally consequential processes.

The findings are based on data from theECLS-K, and despite the richness of the dataset, the ECLS-K focused on elementary schoolchildren only. In the future, researchers maywant to look at the role of psychosocial well-being in the divorce process specifically foradolescents, as evidence suggests that childrenmay react differently to divorce accordingto their age (Lansford, 2009; Lansford et al.,2006). Moreover, this study examined academicachievement, and a different set of outcomescould reveal a different set of explanatorypatterns.

Finally, a caution must be posited regardingthe causal ordering proposed in this study.There is the potential of reverse ordering in thedescribed relationships. It could be that changesto children’s academic performance happen

prior to changes in their psychosocial well-beingand that changes to well-being happen priorto the actual divorce. The longitudinal designof the analysis and the time-varying measuresof divorce, psychosocial well-being, and testscores alleviate some potential for endogenouseffects (Duncan, Magnuson, & Ludwig, 2004),but, as is the case with all observational data,complex modeling does not fully eliminatethe possibility. A cross-lagged analysis wasperformed on children’s math and reading scoresand psychosocial well-being as a final check oftemporal ordering. The cross-lagged analysessupported the pattern of relationships outlinedin this article (e.g., changes to well-beingprecede changes to academic achievement).Causal ordering aside, the main contribution ofthis study is its highlighting of psychosocialwell-being as part of the complex divorceprocess and its relation to children’s academicachievement. As research continues to moveforward, consideration of the many factorsinvolved in the divorce process will providethe best opportunity for continued growth inknowledge and understanding.

NOTE

I would like to thank Josipa Roksa, Liz Gorman, BradWilcox, David Morris, participants in the Works in Progressmeeting at the University of Virginia, and the late SteveNock for their comments and insight throughout this project.During this research the author was funded by the Institute ofEducation Sciences, U.S. Department of Education, throughGrant R305B040049 to the University of Virginia. Theopinions expressed are those of the author and do notrepresent views of the funding agencies.

REFERENCES

Amato, P. (2000). The consequences of divorce foradults and children. Journal of Marriage and theFamily, 62, 1269 – 1287.

Amato, P. (2001). Children of divorce in the 1990s:An update of the Amato and Keith (1991)meta-analysis. Journal of Family Psychology, 15,355 – 370.

Amato, P., & Booth, A. (1996). A prospective studyof divorce and parent – child relationships. Journalof Marriage and the Family, 58, 356 – 365.

Amato, P., & Keith, B. (1991). Parental divorceand the well-being of children: A meta-analysis.Psychological Bulletin, 110, 26 – 46.

Amato, P., & Sobolewski, J. (2001). The effects ofdivorce and marital discord on adult children’spsychosocial well-being. American SociologicalReview, 66, 900 – 921.

Page 13: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

Divorce and Children’s Psychosocial Well-Being 945

Aronson, E. (1984). The social animal (4th ed.). NewYork: Freeman.

Becker, B. E., & Luthar, S. S. (2002). Social-emotional factors affecting achievement out-comes among disadvantaged students: Closing theachievement gap. Educational Psychologist, 37,197 – 214.

Blackorby, J., & Wagner, M. (1996). Longitudinalpostschool outcomes of youth with disabilities:Findings from the National Longitudinal Transi-tion Study. Exceptional Children, 62, 399 – 413.

Cherlin, A. (1992). Marriage, divorce, remarriage.Cambridge, MA: Harvard University Press.

Cherlin, A., Furstenberg, F., Chase-Lansdale, L.,Kiernan, K. E., Robings, P. K., Morrison, D. R.,et al. (1991). Longitudinal studies of effects ofdivorce on children in Great Britain and the UnitedStates. Science, 252, 1386 – 1389.

Cohen, S., Kamarck, T., & Mermelstein, R. (1983).A global measure of perceived stress. Journal ofHealth and Social Behavior, 24, 385 – 396.

Davies, P. T., & Cummings, E. M. (1994). Maritalconflict and child adjustment: An emotionalsecurity hypothesis. Psychological Bulletin, 116,387 – 411.

Dolgin, K. G. (1996). Parents’ disclosure of their ownconcerns to their adolescent children. PersonalRelationships, 3, 159 – 169.

Duncan, G. J., Magnuson, K., & Ludwig, J. (2004).The endogeneity problem in developmental stud-ies. Research in Human Development, 1, 59 – 80.

Entwisle, D. R., Alexander, K. L., & Olson, L. S.(2005). First grade and educational attainmentby age 22: A new story. American Journal ofSociology, 110, 1458 – 1502.

Fitzmaurice, G. M., Laird, N. M., & Ware, J. H.(2004). Applied longitudinal analysis. Hoboken,NJ: Wiley.

Hetherington, E. M., Bridges, M., & Insabella, G. M.(1998). What matters? What does not? Fiveperspectives on the association between maritaltransitions and children’s adjustment. AmericanPsychologist, 53, 167 – 184.

Kincaid, S. B., & Caldwell, R. A. (1995). Maritalseparation: Causes, coping, and consequences.Journal of Marriage and Divorce, 22, 109 – 128.

Kingston, P. W., Hubbard, R., Lapp, B., Schroeder,P., & Wilson, J. (2003). Why education matters.Sociology of Education, 76, 53 – 70.

Lane, K. L., Carter, E. W., Pierson, M. R., & Glaeser,B. C. (2006). Academic, social, and behavioralcharacteristics of high school students withemotional disturbances or learning disabilities.Journal of Emotional and Behavioral Disorders,14, 108 – 117.

Lansford, J. E. (2009). Parental divorce and children’sadjustment. Perspectives on Psychological Sci-ence, 4, 140 – 152.

Lansford, J. E., Malone, P. S., Castellino, D. R.,Dodge, K. A., Pettit, G. S., & Bates, J. E. (2006).Trajectories of internalizing, externalizing, andgrades for children who have not experienced theirparents’ divorce of separation. Journal of FamilyPsychology, 20, 292 – 301.

Law, K. S., Wong, C., & Mobley, W. H. (1998).Toward a taxonomy of multidimensional con-structs. Academy of Management Review, 23,741 – 755.

Martinez, C. R., & Forgatch, M. S. (2002). Adjustingto change: Linking family structure transitions withparenting and boys’ adjustment. Journal of FamilyPsychology, 16, 107 – 117.

McLanahan, S., & Sandefur, G. (1994). Growing upwith a single parent: What hurts, what helps.Cambridge, MA: Harvard University Press.

McLeod, J. D., & Fettes, D. L. (2007). Trajectoriesof failure: The educational careers of childrenwith mental health problems. American Journal ofSociology, 113, 653 – 701.

Mehana, M., & Reynolds, A. J. (2004). Schoolmobility and achievement: A meta-analysis.Children and Youth Services Review, 26, 93 – 119.

Morrison, D. R., & Cherlin, A. (1995). The divorceprocess and young children’s well-being: Aprospective analysis. Journal of Marriage and theFamily, 57, 800 – 812.

Needham, B. L., Crosnoe, R., & Muller, C. (2004).Academic failure in secondary school: The inter-related role of health problems and educationalcontext. Social Problems, 51, 569 – 586.

Popenoe, D. (1996). Life without father. New York:Simon & Schuster.

Roeser, R. W., Eccles, J. S., & Freedman-Doan, C.(1999). Academic functioning and mental healthin adolescence: Patterns, progression, and routesfrom childhood. Journal of Adolescent Research,14, 135 – 174.

Royston, P. (2007). Multiple imputation of missingvalues: Further update of ICE, with an emphasison interval censoring. Stata Journal, 7, 445 – 464.

Singer, J., & Willett, J. (2003). Applied longitudinaldata analysis. New York: Oxford University Press.

Stacy, J. (1996). In the name of the family: Rethinkingfamily in the postmodern age. Boston: BeaconPress.

Strohschein, L. (2005). Parental divorce and childmental health trajectories. Journal of Marriageand Family, 67, 1286 – 1300.

Strohschein, L. (2007). Challenging the presumptionof diminished capacity to parent: Does divorcereally change parenting practice? Family Rela-tions, 56, 358 – 368.

Sun, Y. (2001). Family environment and adolescents’well-being before and after parents’ disruption:A longitudinal analysis. Journal of Marriage andFamily, 63, 697 – 713.

Page 14: Psychosocial Well-Being and the Relationship Between Divorce and Children's Academic Achievement

946 Journal of Marriage and Family

Sun, Y., & Li, Y. Z. (2002). Children’s well-beingduring parents’ marital disruption process: Apooled time-series analysis. Journal of Marriageand Family, 64, 472 – 488.

Thomson, E., Hanson, T. L., & McLanahan, S. (1994).Family structure and child well-being: Economicresources vs. parental behaviors. Social Forces,73, 221 – 242.

Tillman, K. H. (2007). Family structure pathwaysand academic disadvantage among adolescents instepfamilies. Sociological Inquiry, 77, 383 – 424.

Tourangeau, K., Nord, C., Le, T., Pollack, J. M., &Atkins-Burnett, S. (2006). Early Childhood Lon-gitudinal Study, Kindergarten Class of 1998 – 99(ECLS-K), combined user’s manual for the ECLS-K fifth-grade data files and electronic codebooks

(NCES 2006 – 032). Washington, DC: NationalCenter for Education Statistics.

Votruba-Drzal, E., Li-Grining, C. P., & Maldonado-Carreno, C. (2008). A developmental prospectiveon full- versus part-day kindergarten and children’sacademic trajectories through fifth grade. ChildDevelopment, 79, 957 – 978.

Wallerstein, J. S., & Blakeslee, S. (1989). Secondchanges: Men, women, and children a decadeafter divorce. New York: Ticknor and Fields.

Woodward, L., Fergusson, D. M., & Belsky, J. (2000).Timing of parental separation and attachment toparents in adolescence: Results of a prospectivestudy from birth to age 16. Journal of Marriageand the Family, 62, 162 – 174.