early family transitions and depressive symptom changes from adolescence to early adulthood

ALAN BOOTH AND ELISA RUSTENBACH Pennsylvania State University

SUSAN MCHALE Pennsylvania State University*

Early Family Transitions and Depressive Symptom

Changes From Adolescence to Early Adulthood

Some research and theory suggest that earlycohabitation, parenthood, and marriage havenegative long-term implications. Nevertheless,in the context of their resources and opportuni-ties, early transitions may represent positivechoices for some individuals. We studied thefamily and personal characteristics of youngadults (N ¼ 8,172) who did, versus those whodid not, make early family transitions. We as-sessed changes in their depressive symptomsfrom adolescence to young adulthood. Individu-als who made early family transitions were dis-advantaged in many respects, but differed littlefrom those who did not with respect to changesin depressive symptoms. That they stay ‘‘even’’with those who do not make transitions suggeststhat some young adults make positive choicesfrom among limited options.

From a developmental perspective, emergingadulthood, the period from 18 to 25 years ofage, has been described as a time for identityexploration in the domains of worldviews, love,and work among youth in modern industrialized

societies (Arnett, 2000). From this perspective,identity exploration is made possible by delayingthe assumption of adult roles such as marriageand parenthood. An extended period of identityformation, in turn, serves as a foundation formaking positive choices and creating a stable, sat-isfying life structure (Arnett). By implication, in-dividuals who fail to postpone family roletransitions miss out on opportunities (e.g., foreducation and work experience), make poorchoices (e.g., about romantic partners), and con-sequently may experience adjustment problems.

Some demographic research is consistent withthis perspective. For instance, parenting duringthe years of emerging adulthood has beenlinked to socioeconomic disadvantage and mari-tal instability over the long term (e.g., Astone &Upchurch, 1994). Early marriage is linked toincreases in fertility that may foreclose socioeco-nomic opportunities (e.g., Teachman, Polonko,& Leigh, 1987) and decrease educational (e.g.,Marini, 1985) and occupational achievement(e.g., Otto, 1979). Little is known about the impli-cations of cohabitation during emerging adult-hood; this can be viewed as an alternative tomarriage (Brown & Booth, 1996), a step towardmarriage, or as an alternative to singlehood(Rindfuss & Vanden Heuvel, 1990).

In the face of the aforementioned theory andresearch, we explore an alternative scenario.We begin with the idea that opportunities to treatyoung adulthood as a time of exploration are notopen to all individuals given their personal,social, and economic resources. In these cases,early family transitions may not have the same

Sociology Department, 211 Oswald Tower, PennsylvaniaState University, University Park, PA 16802 ([email protected]).

*Human Development and Families Studies Department,604 Oswald Tower, Pennsylvania State University, Univer-sity Park, PA 16802.

Key Words: cohabitation, depression, emergent adulthood,life events, life transitions, marital status, transition to par-enthood, youth.

Journal of Marriage and Family 70 (February 2008): 3–14 3

negative implications as they do in the case ofmore privileged individuals. In this scenario,some young adults make early family transitionsbecause they expect to benefit from them. Indeed,research shows that marrying, cohabiting, andbecoming a parent, at least at older ages, providesa range of benefits (Coleman, 1988). Theseinclude increasing social capital that buffers indi-viduals from stressful life circumstances, foster-ing a sense of belonging and social connection,and providing more positive living arrangements.Increases in economic capital, such as incomefrom a partner, are also possible (e.g., Pears,Pierce, Kim, Capaldi, & Owen, 2005). Althoughsome theory and research support the propositionthat early family formation has long-term nega-tive consequences, these findings suggest thatsome young adults may benefit from family tran-sitions. To date, researchers have not examinedthe short-term outcomes of transitions, particu-larly in the domain of psychological adjustmentwherein positive implications of family transi-tions may be most apparent.

To test our alternative scenario, this study ad-dressed two questions. Using data from theNational Longitudinal Study of AdolescentHealth, we first asked whether individuals whomade family transitions in early adulthood, com-pared to those who did not, came from less advan-taged family backgrounds, had attributes andexperiences in adolescence that put them at riskfor poorer psychological well-being, or both.Second, we tested whether those who made earlyfamily transitions, as compared to those who didnot, exhibited different patterns of change indepression symptoms from adolescence to youngadulthood. The family background characteris-tics we studied were family income, parental edu-cation, and family structure (i.e., whether youthdid or did not live with two biological parents).We also examined mother-child relationshipquality, verbal ability, school attachment, anddelinquent behavior in adolescence as potentialrisk or protective factors for psychological well-being. Given the different meanings of familytransitions for women and men, we studied therole of gender in these processes.

We expected that those making transitionswould come from less advantaged families andbe more at risk for adjustment difficulties thanthose who do not. Also, we suggested three pos-sibilities regarding the link between family transi-tions and patterns of change in depressivesymptoms from adolescence to young adulthood.

First, stresses associated with family formation,above and beyond the presence of backgroundand risk factors, may mean that those who makeearly transitions are more likely to show increasesor remain at higher levels of depressive symp-toms from adolescence to young adulthood rela-tive to those who do not make transitions.Second, the benefits of family transitions couldameliorate the background and risk factors suchthat those who make early transitions are morelikely to show decreases in depressive symptomsrelative to those who do not marry, cohabitate, orbecome parents. Third, the benefits associatedwith family transitions and the debilitating aspectsof background and risk factors may balance eachother such that there are no differences betweenthose who do and do not make transitions.

FAMILY BACKGROUND AND RISK OR

PROTECTIVE FACTORS AND THEIR LINK TO

FAMILY TRANSITIONS

The family background and risk or protectivefactors included in this study have been investi-gated in relation to a wide range of adolescentorientations toward risky sexual behaviors aswell as problem behaviors, more generally(e.g., Kurdek & Fine, 1994; Steinberg, Lamborn,Darling, Mounts, & Dornbusch, 1994). To date,these factors have not been examined vis-a-visfamily transitions during early adulthood. Wereview possible mechanisms by which thesefactors may influence the decision to make anearly transition.

Low family income is known to lead to paren-tal conflict, which, in turn, is linked to poorerparent-child relationship quality (Conger, Ge,Elder, Lorenz, & Simons, 1994). Other workshows that poorly educated adults tend to be lessskilled parents (Amato & Booth, 1997). Growingup in families comprised of one or no biologicalparents is associated with problematic parent-childrelationships, specifically less affection, less con-sensus, and less perceived support (Amato &Booth). A divisive home environment and poorerparenting skills may encourage offspring to seeksocial support in early marriage or cohabitation.

Low levels of parental warmth and affectionand a lack of clear expectations and limits arelinked to poorer self-regulation and social com-petence in youth and, in turn, higher levels ofproblem behavior (Parke & Buriel, 1998). Earlyfamily transitions may provide individuals withan opportunity to escape from an unloving home

4 Journal of Marriage and Family

environment as well as the possibility of creatinga more positive and supportive family context.Verbal intelligence is an individual characteristicthat has been linked to educational and occupa-tional achievement (Woodward, Fergusson, &Horwood, 2001) and also has implications forthe development of coping strategies for dealingwith adversity, including soliciting assistanceand support from others (Masten, Best, & Garmezy,1990). Relatedly, high levels of school attach-ment indicate a sense of belonging with peersand teachers and may enhance school achieve-ment. In contrast, individuals who are unable togarner social support or succeed in school maydecide to have a child who will love them uncon-ditionally (Edin & Kefalas, 2005). A final per-sonal characteristic of interest is delinquency,which may be reflective of both a lack of supportand limit setting in the family as well as poorself-regulation and impulse control in the in-dividual (Parke & Buriel). Delinquent activitiesmarginalize youth from their families and socialinstitutions. For these individuals, early familyformation may be a way to make up for their lackof connectedness and achievement.

Of course the same qualities that lead to earlyfamily formation may also lead to stress in newfamily relationships. If underdeveloped socialskills and poor impulse control challenge thequality of interaction with a new partner or baby,an increase in depressive symptoms may ensue.

In addition to these family and personal char-acteristics, we also expected that gender wouldbe a factor in early family transitions and theirlink with depressive symptoms. Women tend tobe more selective in their involvement in roman-tic relationships than men: They prefer thestability of marriage to cohabitation (Bumpass,Sweet, & Cherlin, 1999) and are more likely toseek partners who will provide resources and helpwith rearing offspring (Daly & Wilson, 1983).Thus, the meanings of cohabitation, parenthood,and marriage differ for women and men. Substan-tial research also documents gender differencesin depressive symptoms: Starting in adolescence,girls report higher levels of depression than boys(Petersen, Sarigiani, & Kennedy, 1991). Giventhese patterns, we examined family transitionsseparately for women and men.

Control variables consisted of age and ethnic-ity. Given that the correlates of family transitionsare likely to vary as a function of their develop-mental timing, we included respondent’s age.Data showing that the timing of family transitions

varies across ethnic groups (Bianchi & Casper,2000) led to including ethnicity.

VARIATION IN THE IMPLICATIONS OF

EARLY FAMILY TRANSITIONS

Our study is grounded in the premise that someindividuals may benefit from early family transi-tions, whereas others may not. A study of youngwomen at risk who benefited from marriage isinstructive in this regard (Rutter & Quinton,1984). This classic study focused on a group ofwomen who were raised in institutions and a com-parison group who resided in the same geographicarea. Overall, the psychological development ofthe institution-reared women was considerablypoorer. Nevertheless, a significant number func-tioned as well in adulthood as those in the compar-ison group: These were women who had positiveeducational experiences and a sense of efficacyand, as a result of these attributes, were marriedto supportive and nondeviant spouses. The au-thors concluded that a sense of efficacy derivedfrom positive school experiences mediated thelink between institutional upbringing and achiev-ing a supportive marriage relationship, which, inturn, explained well-being in young adulthood.

A study of unmarried men and womenevaluated the extent to which the transition tomarriage was associated with changes in depres-sion (Frech & Williams, 2007). Although theydid not differentiate between those making earlyversus on-time transitions, findings suggested thatthe psychological benefits of marriage dependedon the respondent’s premarital depression: Menand women who were depressed prior to marriagereported greater psychological gains from mar-riage than those who were not depressed. Thestudy further supports the idea that people at riskmay have more to gain from family transitionsthan less disadvantaged individuals.

METHOD

This study drew on interviews from Waves 1 and3 of the National Longitudinal Study of Adoles-cent Health (Add Health). At the time of the firstinterview, youth ranged in age from 12 – 18years. Wave 3 took place 5 years later, when mostindividuals were young adults. For the purpose ofthis study, family background and risk or protec-tive factors were measured at Wave 1, and familytransitions were measured at Wave 3. We useddata on depressive symptoms collected at both

Early Family Transitions and Depressive Symptom Changes 5

Waves 1 and 3 to create groups that varied in leveland change in depression symptoms.

Participants

The Add Health Study consists of a nationallyrepresentative sample of 20,745 middle and highschool students who were first interviewed in1995 – 1996 and for a third time in 2000 – 2001when they were young adults. Response ratesfor Waves 1 and 3 were 78.9% and 77.4%,respectively. The study oversampled a numberof populations, and weighting was used in orderto make the sample nationally representative.The number of respondents who had weight val-ues and who participated in both Waves 1 and 3was N ¼ 14,086. A more detailed description ofthe study can be found in Bearman, Jones, andUdry (1997).

Individuals who were close in age to those whotypically make the transition to marriage and par-enthood (23 to 25 years of age) were eliminatedfrom the sample so that, for comparison purposes,those making early transitions were in the sameage range as those who did not. For the analysisof early transition to cohabitation, those whowere close in age to the national average (i.e.,21 and 25 years of age) were eliminated fromthe analysis. Individuals who made family transi-tions prior to Wave 1 were also eliminated. Thefinal sample consisted of 8,172 individuals.

The resulting sample included individualswhose average age of making a family transitionwas below national averages. As shown inTable 1, the mean ages of the women and menin our sample who made family transitions wereunder 20 years. In contrast, in a U.S. sample ofa household population 15 – 44 years of age(the National Survey of Family Growth), theaverage age for initiating cohabitation by menwas 22.9. (There was no figure available forwomen.) In a national sample of females ages10 – 49, Mathews and Hamilton (2002) estimated

that the average age for becoming a mother was24.9 years. A comparable figure for first father-hood is unavailable. Finally, the national estimatefor age at marriage for women was 25 and for menwas 27 years (Bianchi & Casper, 2000). The finalsample made it possible to compare those makingearly transitions with those who did not and isideal for exploring the factors associated withmaking early transitions as well as the conse-quences of those transitions.

Measures

All scales were unidimensional and reliable atacceptable levels. For each scale, we calculatedthe mean across all items. Fewer than 5% of thevalues were missing for any variable exceptprimary caregiver education and per capitahousehold income. For these two variables,approximately 25% of the data were missingbecause they were obtained from the parentalquestionnaire, which was not administered inall households. Missing cases were replacedwith imputed values using the expectationmaximization (EM) algorithm (Allison, 2001)in SPSS.

We selected depressive symptoms as the out-come measure because they are associated withmany types of adversity such as poor physicalhealth, unemployment, and harsh family relation-ships (Amato & Booth 1997). Depressive symp-toms were indexed as the mean of seven itemscollected at Waves 1 and 3. For example, re-spondents were asked if, during the last week,they ‘‘were bothered by things that usually don’tbother you, or felt that you could not shake off theblues, even with help from your family and yourfriends.’’ Responses ranged from 0 ¼ never orrarely to 3 ¼ most of the time or all of the time,with high scores signifying more depressivesymptoms. The depressive symptoms scale hadan a of .81 for Wave 1 and .82 for Wave 3 (seeTable 2 for descriptive statistics). In an effort to

Table 1. Descriptive Statistics on Family Formation for Men and Women

Men (n ¼ 3,418) Women (n ¼ 4,057)

Cohabitation Birth Marriage Cohabitation Birth Marriage

N 899 307 273 1,379 933 608

% making transition 26 9 8 34 23 15

Mean age at transition (SD) 18.6 (1.25) 19.3 (1.53) 19.8 (1.35) 18.5 (1.24) 19.1 (1.62) 19.6 (1.36)

Note: Data from Add Health.


capture both level and change in depressivesymptoms, we classified individuals as beinghigh in depressive symptoms if they scored inthe top 20th percentile, and we created two clas-sification schemas: First, participants werecoded 0 if they were high in depression at bothWaves 1 and 3 and they were coded 1 if theywere high in depression at Wave 1 but not atWave 3, indicating a decline in depressive symp-tomology. Second, individuals were coded 0 ifthey scored low in depressive symptoms at bothWaves 1 and 3 and 1 if they were low in depres-sive symptoms at Wave 1 but high at Wave 3,signifying an increase in symptomology.

Although cutoffs for grouping variables aresometimes arbitrary, we chose a 20% cutoffbased on data showing that a little more than20% of the population meet the criteria for clini-cally significant mental health problems involv-ing depression symptoms (National Institutes ofMental Health, 2006), including mood disorders(9.5% of the population), major depression(6.7%), dysthymic disorder (1.5%), and bipolar dis-order (2.6%). Other cutoff points and ways of con-structing the depressive symptoms variable wereexamined, but they all produced similar results.

To measure first birth, respondents were askedhow many times they or their partner had hada pregnancy. Then respondents were asked,‘‘Next, please indicate the outcome of this preg-nancy by selecting the appropriate response.’’We restricted our focus to live first births(0 ¼ never had a live birth, 1 ¼ had at least 1live birth). In Wave 3 of Add Health, respond-ents were asked about their pregnancies and

births within the context of a history of theirromantic and sexual relationships. As such, somerespondents failed to report on births in the con-texts of cohabitations and marriages that hadbeen reported in an earlier section of the inter-view. As a result, Add Health’s fertility history isincomplete. We used the household roster tocheck for incompleteness and to add births to thehistory that had been omitted. A detailed descrip-tion of the procedure that was used to correct thefertility history is available from the authors. Inseparate analyses, we differentiated births thatoccurred in a marriage, in a cohabitation relation-ship, or outside either type of relationship.

To assess cohabitation, respondents wereasked ‘‘Have you ever lived with someone ina marriage-like relationship for one month ormore?’’ Responses were coded 0 ¼ no/neverand 1 ¼ yes/at least once. Because cohabita-tions tend to be of short duration and becausewe expected that relationship instability wouldbe linked to symptoms of depression, wefocused on the first cohabitation and createda measure of cohabitation stability using re-spondents’ answer to the question ‘‘Are you stillliving together?’’

Marriage was derived from the question ‘‘Howmany times have you been married?’’ Responsesranged from 0 to 3. The number of individualswho had been married more than once was lessthan 100, so we recoded this variable to 0 ¼never married and 1 ¼ married at least once.Given that the total number of marital disrup-tions in our sample was very small, we did notcreate an index of marital stability.

Table 2. Descriptive Statistics for Control, Family Background, and Risk and Protective Factors

Men (n ¼ 3,418) Women (n ¼ 4,057)

M SD M SD

Age (W 1; range ¼ 12 – 20) 14.69 1.12 14.65 1.15

Black (W 1; 1 ¼ Black) .14 .34 .14 .35

Hispanic (W 1; 1 ¼ Hispanic) .07 .26 .07 .25

Other (W 1; 1 ¼ Other race) .10 .30 .09 .29

Per capita family income (W 1; range ¼ 0 – 100,000) 14,475.30 11,407.10 15,164.70 12,123.90

Caregiver education (W 1; range ¼ 0 – 9) 5.55 2.17 5.52 2.16

Family structure (W 1; 0 ¼ 2 biological parents) .41 .49 .40 .49

Mother-child relationship (W 1; range ¼ 0 – 5) 4.38 .58 4.24 .07

Vocab. skill (W 1; range ¼ 17 – 146) 102.49 14.08 100.9 14.09

School attachment (W 1; range ¼ 1 – 5) 3.80 .81 3.84 .84

Delinquency (W 1; range ¼ 0 – 3) .32 .39 .25 .31

Depression decrease (1 ¼ Decrease) .59 .49 .61 .49

Depression increase (1 ¼ Increase) .18 .38 .18 .38

Note: W 1 ¼ Wave 1.


The control variables of age and ethnicity wereincluded in every regression. Age was indexed asa continuous variable, measured at Wave 1. Eth-nicity consisted of four categorical variables(Black, Hispanic, and Other), with White as thereference category.

Family per capita income was obtained fromthe parent questionnaire by taking the total house-hold income and dividing by the number ofhousehold members. Primary caregiver educa-tion was measured by a single item: ‘‘How fardid you go in school?’’ from the parental ques-tionnaire; responses ranged from 0 ¼ neverwent to school to 9 ¼ professional trainingbeyond a 4-year college or university. Familystructure was obtained from a household rosterat Wave 1 and coded as a dichotomous variablewhere 0 ¼ two biological parents and 1 ¼ oneor no biological parents.

Mother-offspring relationship quality wasmeasured using a four-item scale (e.g., ‘‘Mostof the time, your mother is warm and lovingtoward you.’’). Respondents used a 5-point scale(1 ¼ strongly agree to 5 ¼ strongly disagree) torate how strongly they agreed or disagreed withstatements describing their experiences. Thescale is coded so that high values signify a posi-tive mother-child relationship, and Cronbach’sa was .84. Vocabulary skill was measured usingthe Add Health Picture Vocabulary Test, whichwas adapted from the Peabody Picture Vocabu-lary Test administered in Wave 1. Schoolattachment was measured at Wave 1 usinga three-item scale (e.g., ‘‘You feel close to peo-ple at your school.’’). Items were coded so thathigher values indicate higher levels of schoolattachment, and Cronbach’s a was .79. Delin-quency was measured using a 14-item measureon which respondents were asked to rate howoften they engaged in the past 12 months ina range of activities (e.g., ‘‘How often did youdeliberately damage property that didn’t belongto you?’’) using a 4-point scale (0 ¼ never orrarely to 3 ¼ 5 or more times). Items werecoded so that higher values indicate more delin-quency, and Cronbach’s a was .85.

RESULTS

The results are organized in two steps. First, wecompared the risk and protective characteristicsand experiences of those who did versus did notreport early cohabitation, parenthood, and mar-riage. Second, we tested whether individuals

who did versus did not make early family transi-tions differed in their patterns of stability andchange in depressive symptoms.

How Are Family Background and Risk orProtective Factors Linked toEarly Family Transitions?

We conducted a series of logistic regressions totest the probability of making each family transi-tion (cohabitation, birth, or marriage) as a func-tion of control, family background, and risk orprotective factors, conducting separate analysesfor women and men. In these analyses we usedevent history person-year files in which each indi-vidual has a record per year until he or she expe-riences a family formation event, and then nolonger contributes to the file. Separate person-year files were constructed for first birth, firstcohabitation, and first marriage. The resultingsample consisted of 8,172 cases from 1996 to2002: 53,459 person-years for the birth file,49,480 person-years for cohabitation, and54,730 person-years for the marriage file. Eventhistory analysis involves measuring a continuousrecord over time on a variable, in this case, lengthof time until the first family formation event ofeach type. In this way we take into account twofactors: (a) whether the individual had a familyformation transition or not and (b) the timing ofthe transition.

Results are shown in Table 3. With respect tothe control variables, we found that those whobecame a parent or married by Wave 3 were old-er. Furthermore, compared to Whites, Blackswere less likely to marry and cohabit. In addition,compared to Whites, Hispanics were less likely tocohabit. Finally, those in the ‘‘other’’ category(mostly Asians and Native Americans) were lesslikely than Whites to cohabit or marry.

Turning to family background characteristics,those who made early family role transitions weremore likely to come from low income families.With the exception of men who married, thosewho made early transitions were more likely tohave parents with lower levels of education andwere more likely to have lived in a householdwith one or no biological parents. Similarpatterns have been reported in prior research(Bianchi & Casper, 2000).

Risk and protective factors also were related tofamily formation in emerging adulthood. First,women who had poor relationships with theirmothers in adolescence were more likely to


cohabit, but mother-child relationship qualityhad no association with becoming a parent orwith marriage. With the exception of men whomarried or became a parent, those with poorvocabulary skills were more likely to make fam-ily transitions. Both men and women who hadexperienced low school attachment were morelikely to cohabit, but only the women weremore likely to become parents. Both men andwomen who had reported high levels of delin-quency in adolescence were more likely tocohabit and become parents, but only the menwere more likely to marry.

Interaction terms were computed (Gender 3Risk Variable) for all of the predictor variableslisted in the table and added to regressions exam-ining the factors that have the potential for beingassociated with making an early transition (notshown). Of all gender differences reported inTable 3, five were statistically significant withrespect to factors that influenced family transi-tion decisions. Women who had a poor relation-ship with their mother were more likely tocohabit or become a parent compared to men insuch a relationship. Similarly, women who hadbeen involved in delinquent behavior weremore likely to cohabit than men with a historyof delinquent behavior. Finally, women whohad poor verbal ability or lived with one or nobiological parents, compared to men in thesecategories, were more likely to become a parent.In summary, there is a slight tendency forwomen with potential risk factors to make early

transitions more than for men to do so. It isimportant to note, though, that the data onmen’s transition to parenthood are less reliablethan the data for women.

What Are the Links Between EarlyTransitions and Patterns of

Change in Depression Symptoms?

As a preliminary step, we examined overall pat-terns of change in depressive symptoms by familytransition status. Table 4 shows the mean scores atWave 3 as a function of level, change group, andfamily formation event. Consistent with priorresearch (e.g., Petersen et al., 1991) females re-ported higher levels of depression symptoms thanmales. T tests indicated that gender differences indepression symptoms were statistically signifi-cantly at Wave 1 and at Wave 3. Nevertheless,as indicated in Table 2, there were no gender dif-ferences in patterns of change in depressionsymptoms: Although men and women had dif-ferent levels of depressive symptomology, theyshowed similar patterns of change.

Approximately two thirds of the respondentshad consistently low levels of depressive symp-toms, and similar proportions experienced de-creases and increases in symptoms (i.e., outof or into the top 20%) over a 5-year periodspanning from adolescence to early adulthood.Less than 10% were in the high depressivesymptoms group across the period of thestudy. Also, levels of depression symptoms

Table 3. Odds Ratios for Effects of Control, Family Background, and Risk and Protective Factors on Family Transitions for

Men (M) and Women (W)

Cohabitation Birth Marriage

M W M W M W

Age .97 .86** 1.16* 1.27*** 1.47*** 1.52***

Black .54*** .45*** 1.13 1.15 .48* .30***

Hispanic .46** .61* 1.41 .78 .99 1.16

Other .42*** .61* .92 .72 .51* .42**

Income .99 .99*** .99* .99** .99* .99**

Education .87*** .91*** .84*** .89*** .95 .92**

Family structure (0 ¼ 2 Biological parents) 1.41* 1.93*** 1.54* 2.04*** 1.38 1.34*

Mother-child relationship .92 .72*** 1.25 .88 1.16 1.04

Vocabulary skill .99** .99* .99 .97*** .98 .99*

School attachment .82** .81*** .93 .86* 1.00 .90

Delinquency 1.70*** 2.59*** 1.72** 1.83*** 1.52* .92

Total person-years 21,401 24,156 23,075 25,578 23,234 26,764

*p, .05. **p, .01. ***p , .001.


seem to be generally similar across transitiontype and whether or not any transition wasmade. To test whether or not making a familytransition was linked to an increase or decreasein depressive symptoms or made no differencein symptoms, we conducted a series of logisticregression analyses separately for women(Table 5) and men (Table 6) and for each typeof transition. In addition, because we expectedthat depressive symptoms would be lower inthe context of stable relationships (i.e., ongoingvs. terminated cohabitations; parenthood in thecontexts of marriage), we examined the depres-sive symptom change patterns for these sub-groups versus those who did not makea transition. For each analysis, the comparisongroup was those who made no family transition.Coefficients expressing the association between

early family transition experiences for womenwhose depressive symptoms were high initiallyand stayed high or decreased are shown in thefirst three columns and for those who were ini-tially low and stayed low or increased depressivesymptoms in the second three columns. The co-efficients without family background and risk orprotective variables in the equation are shown inthe first column. The second column includes co-efficients after family background variables areadded to the equation, and the third columnshows coefficients when both family backgroundand risk or protective variables are in the equa-tion.

Results shown in Table 5 indicate that womenwho cohabited early were less likely to be in thedecreasing depressive symptom group and morelikely to be in the increasing depressive symptom

Table 4. Meansafor Men’s and Women’s Depression Symptoms at Wave 3 by Family Transition and Depression Change

Group

No Family Formation Cohabitation Birth Marriage Any Transition % by Category

Men

Constantly high 1.11 1.11 1.21 1.09 1.13 7%

Decrease 0.33 0.31 0.29 0.31 0.33 12%

Constantly low 0.24 0.23 0.25 0.21 0.24 66%

Increase 1.02 1.10 1.02 0.99 1.07 15%

Women

Constantly high 1.28 1.37 1.30 1.34 1.33 8%

Decrease 0.38 0.38 0.38 0.36 0.38 14%

Constantly low 0.29 0.34 0.34 0.32 0.34 64%

Increase 1.29 1.20 1.16 1.19 1.21 14%

aScores can range from 0 – 3.

Table 5. Odds Ratios for Effects of Family Transitions on Decreases (0 ¼ Stable/High, 1 ¼ Decrease) and Increases

(0 ¼ Stable/Low, 1 ¼ Increase) in Depressive Symptoms of Women With and Without Family Background and

Risk and Resilience Characteristics

Depression Decrease Depression Increase

FF

Only

With Family

Backgound

With Background

and R&R

FF

Only

With Family

Background

With Background

and R&R

Cohabitationa

.64* .63* .66 1.77*** 1.59*** 1.35*

Stable cohabitationa

.94 .91 .96 1.13 .99 .89

Unstable cohabitationa

.50** .50** .51** 2.55*** 2.31*** 1.92***

Birtha

.68 .69 .69 1.28 1.13 .98

Birth within a cohabitationa

.50 .47 .49 1.67 1.43 1.25

Birth within a marriagea

.90 .89 .84 .86 .75 .73

Birth outside any uniona

.74 .77 .79 1.18 1.05 .91

Marriagea

.79 .82 .84 1.24 1.12 1.04

Note: FF ¼ Family formation. R&R ¼ Risk and resilience. Controls for age and race or ethnicity included in all regressions.aThe reference group is no family formation.

*p , .05. **p, .01. ***p, .001.


group as compared to women who made no familytransition. After the inclusion of the risk and pro-tective variables, the chance of remaining in thedepressed category was no longer statistically sig-nificant. Verbal ability accounted for the largestdecline in statistical significance (analysis notshown). Comparison with women who were ina stable cohabitation versus an unstable one re-vealed that, although increases in symptoms andremaining depressed were more common forwomen who had experienced unstable cohabita-tion, women who experienced stable cohabitationdid not differ in their pattern of stability and changein symptoms from those who did not cohabit.

Compared to women who did not experiencea transition, those becoming a parent were nomore likely to experience a change in depressivesymptoms, regardless of whether the birth wasinside or outside of a union. Also, there was noevidence that women who married differed fromthose who did not with respect to stability andchange in depressive symptoms.

For men, compared to those who made no fam-ily transition, cohabitation was linked to a lowerchance of decline in depressive symptoms butnot to a greater chance of increasing depressivesymptoms when the relationship was unstable.Those who experienced an unstable cohabitation(row 3) were more likely to show an increase insymptoms. Once school attachment and delin-quency were in the equation, the latter coefficientwas no longer statistically significant. Likewomen, being in a stable cohabiting relationshipwas not linked to change in symptoms comparedto those who did not make a transition. The key tounderstanding the link between cohabitation and

depressive symptoms was whether the relation-ship was stable: Those in a stable cohabitationrelationship were no different from those whowere not in terms of change in depression symp-toms. Approximately half of women and menwho cohabited had a stable relationship.

Men who married early were no more likely toexperience a change in depressive symptomsthan those who made no family transition. Menwho became parents were less likely to experi-ence a decrease in depression symptoms overtime. When per capita income was included inthe equation, they were no more likely to bedepressed than those who had not experienceda family formation event. We did not exploreunion status at the time of birth for men becauseof the limited number of cases.

In summary, those who made early family tran-sitions in young adulthood differed little fromthose who did not with respect to changes indepressive symptoms. The only exception is thatwomen who experienced the dissolution of a co-habiting relationship, compared to women whodid not, were less likely to experience a decreasein depressive symptoms and more likely to experi-ence an increase in symptomology. The same find-ing applies to men, but the effects were weaker.

A possible explanation for these findings is thatcertain family background or risk or protectivefactors may keep individuals from exhibiting in-creases in depressive symptoms over time. Theseeffects could be misinterpreted as resulting frompositive consequences of a family formationevent. For example, having a warm mother-childrelationship could offset the negative effects oflow family income, and the lack of an increase

Table 6. Odds Ratios for Effects of Family Transitions on Decreases (0 ¼ Stable/High, 1 ¼ Decrease) and Increases

(0 ¼ Stable/Low, 1 ¼ Increase) in Depressive Symptoms for Men With and Without Family Background and

Risk and Resilience Characteristics

Depression Decrease Depression Increase

FF Only

With Family

Background

With Background

and R&R FF Only

With Family

Background

With Background

and R&R

Cohabitationa

.48** .48** .54** 1.14 1.08 1.03

Stable cohabitationa

.91 .92 1.04 .74 .68 .65

Unstable cohabitationa

.30*** .28*** .28*** 1.44* 1.4* 1.34

Birtha

.55* .57 .58 1.34 1.24 1.20

Marriagea

.59 .63 .65 .97 .90 .87

Note: FF ¼ Family formation. R&R ¼ Risk & Resilience. Controls for Age and Race/Ethnicity included in all regressions.aThe reference group is no family formation.

*p, .05. **p, .01. ***p , .001.


in depression could have nothing to do with fam-ily formation. As a final step in the analysis, there-fore, we examined whether adolescent risk andprotective factors moderated the link betweenmaking family transitions and symptoms ofdepression. Here we tested whether the effectsof early transitions on young adults’ depressivesymptom change varied as a function of familybackground or adolescent risk.

We explored this possibility by creating aninteraction term for each type of transition andeach risk or protective variable (Transition3Riskor Protective) and entered the term into an equationin which depression symptom change (high anddecrease or low and increase) was regressed ontype of transition, risk or protective variable, andthe interaction term along with the controls ofage and ethnicity. This involved 112 regressions(2 indicators of depression3 8 types of transitions3 7 risk factors) for females, but only 70 for malesbecause of the unreliability of the fatherhood data.Only one interaction term was statistically signifi-cant for males, whereas six were significant for fe-males. Because this number of interactions wouldoccur by chance, we concluded that the risk char-acteristics that characterized many of those whomade transitions did not moderate the associationsbetween early family transitions and depressivesymptom changes.

DISCUSSION

Some research and theory suggests that makingfamily transitions — to cohabitation, parenthood,and marriage — in early adulthood may be linkedto poorer well-being. We argue, in contrast, thatfor some individuals, family transitions duringthis period may be beneficial or at least benign.Although early assumption of adult roles andresponsibilities may limit opportunities for iden-tity exploration in the domains of work and edu-cation, individuals may take on family rolesbecause they represent possibilities for positivechange in other ways. Analysis of data from theNational Longitudinal Study of AdolescentHealth confirmed that individuals who madeearly family transitions were more disadvantagedwith respect to family background and risk orprotective characteristics. In spite of these disad-vantages, individuals who made early familytransitions were rarely different from those whodid not with respect to stability and change intheir depressive symptoms from adolescencethrough early adulthood. The vast majority

(86%) did not differ from those who did not makea transition with respect to changes in symptoms.The exceptions were women and men whobecame involved in unstable cohabiting unions.That many individuals who make early familyformation decisions come from less advantagedbackgrounds but do not differ in their depressivesymptoms from those who do not make transi-tions may mean that early transition decisionscan be rational and sound.

It is worth noting that there was a tendency forwomen with risk characteristics to report earlytransitions more so than men with such character-istics: Women, more than men, who had prob-lematic relations with their mothers were morelikely to cohabit and become mothers; womenwho were delinquent were also more likely thandelinquent men to cohabit. In addition, womenwith poor verbal ability and those who did notlive with both biological parents were more likelythan men with these characteristics to become pa-rents. The reasons are not clear, but may be attrib-uted to the fact that women generally make familytransitions at a younger age than men. Impor-tantly, most of these gender differences involveparenthood transitions, and data on early father-hood are less reliable than data on early mother-hood. As such, the role of gender in the linksbetween risk and protective factors and early fam-ily transitions requires further study.

This study is not without limitations. First, theanalyses would have benefited from longitudinaldata on the risk and protective indices so that wecould examine how changes in these were linkedto changes in depression symptoms. More detailon family transition experiences, including indi-viduals’ reasons for their choices and subjectiveevaluations of their family formation experiencesalso would have allowed us to better understandthe conditions under which early family forma-tion is advantageous or disadvantageous to indi-vidual well-being. Finally, as we suggested,problems with the data on births with respect tomen mean that our findings on this family transi-tion must be viewed with caution.

Future research should reexamine the notion thatearly family formation decisions have negativeconsequences in the long term. It may be that, forwomen and men who come from less advantagedbackgrounds, some family formation choicesmay take them out of harsh or unsupportive livingenvironments and have positive consequences. Animportant question is whether benefits stemmingfrom early family transitions are sustained long


term among those with fewer resources and oppor-tunities. Future work also should be directed atidentifying the early characteristics and experien-ces that distinguish those who make early transi-tions and stay on positive trajectories from thosewho do not. These represent important questionsfor research on an understudied population:women and men in early adulthood.

Staying ‘‘even’’ with those who did not makeearly family transitions, given differences in fam-ily backgrounds and risk or protective character-istics and experiences, is consistent with resultsof research focused on low income women show-ing that parenthood can be a source of validation,companionship, sense of accomplishment, identity,and meaning (Edin & Kefalas, 2005), althoughnot always less depression (Evenson & Simon,2005). Our findings also are consistent withresearch showing that marriage can be especiallybeneficial for the psychological well-being ofdisadvantaged individuals (Frech & Williams,2007; Rutter & Quinton, 1984). More generally,our study highlights the importance of studyingearly family formation in context, that is, ofexamining choices about marriage, parenthood,and cohabitation in light of the range of oppor-tunities open to an individual. Emerging adult-hood has been described as an important periodfor exploration for women and men in modernindustrialized societies, but most research onthis developmental period has been conductedwith college students (Arnett, 2000). AddHealth data provided an opportunity to studythe experiences of individuals who have quitedifferent life trajectories from the typical collegestudent; our findings suggest that the post-high-school experiences of individuals in the UnitedStates are diverse, and that their implicationsare best understood in the context of individu-als’ preexisting resources and constraints. Earlyfamily transitions may be a viable and produc-tive option for some young adults.

NOTE

Support was provided by NICHD Grant 1 RO1 HD045309,Family Formation in an Era of Family Change. Supportservices were also provided by the Population Research Insti-tute, Pennsylvania State University. We are indebted to PaulAmato, David Eggebeen, Nancy Landale, and Robert Schoenfor helpful comments on an earlier version of this paper.This research uses data from the National Longitudinal Studyof Adolescent Health (Add Health), a project designed byJ. Richard Udry, Peter S. Bearman, and Kathleen MullanHarris, and funded by grant P01-HD31921 from NICHD, withcooperative funding from 17 other agencies. Persons inter-

ested in obtaining data files from Add Health should contactAdd Health, Carolina Population Center, 123 W. FranklinStreet, Chapel Hill, NC 27516-2524.

REFERENCES

Allison, P. (2001). Missing data. Thousand Oaks,

CA: Sage.

Amato, P., & Booth, A. (1997). A generation atrisk: Growing up in an era of family upheaval.Cambridge, MA: Harvard University Press.

Arnett, J. (2000). Emerging adulthood: A theory of

development from the late teens through the

twenties. American Psychologist, 55, 469 – 480.

Astone, N., & Upchurch, D. (1994). Forming a fam-

ily, leaving school early, and earning a GED: A

racial and cohort comparison. Journal of Marriageand Family, 56, 759 – 771.

Bearman, P. S., Jones, J., & Udry, J. R. (1997). Thenational longitudinal study of adolescent health:Research design. Retrieved March 15, 2006, from

http://www.cpc.unc.edu/projects/addhealth/design.

html.

Bianchi, S., & Casper, L. (2000). American Families.

Population Bulletin 55, No. 4 (December).

Brown, S., & Booth, A. (1996). Cohabitation versus

marriage: A comparison of relationship quality.

Journal of Marriage and Family, 58, 668 – 678.

Bumpass, L., Sweet, J., & Cherlin, A. (1999). The

role of cohabitation in declining rates of marriage.

Journal of Marriage and Family, 53, 913 – 927.

Coleman, J. (1988). Social capital in the creation of

human capital. American Journal of Sociology,94(Suppl. 95), S95 – S120.

Conger, R., Ge, X., Elder, G., Lorenz, F., & Simons,

R. L. (1994). Economic stress, coercive family

process, and developmental problems in adoles-

cents. Child Development, 65, 541 – 561.

Daly, M., & Wilson, M. (1983). Sex evolution andbehavior. Boston: Willard Grant Press.

Edin, K., & Kefalas, M. (2005). Promises I can keep:Why poor women put motherhood before mar-riage. Berkeley, CA: University of California

Press.

Evenson, R., & Simon, R. (2005). Clarifying the rela-

tionship between parenthood and depression. Jour-nal of Health and Social Behavior, 46, 341 – 358.

Frech, A., & Williams, K. (2007). Depression and the

psychological benefits of entering marriage. Jour-nal of Health and Social Behavior, 48, 149 – 163.

Kurdek, L., & Fine, M. (1994). Family acceptance

and family control. Child Development, 65, 1137 –

1146.

Marini, M. (1985). Determinants of the timing of adult

role entry. Social Science Research, 14, 309 – 350.


Masten, A., Best, K., & Garmezy, N. (1990). Resil-

ience and development: Contributions from the

study of children who overcome adversity. Devel-opment and Psychopathology, 2, 425 – 444.

Mathews, T., & Hamilton, B. (2002). Mean age ofmother, 1979–2000. National vital statistics reports(Vol. 51, No. 1, p. 2). Hyattsville, MD: National

Center for Health Statistics.

National Institute of Mental Health. (2006). The num-bers count: Mental disorders in America. National

Institutes of Health publication number 06–4584

2006. Washington, DC: Author.

Otto, L. (1979). Antecedents and consequence of

marital timing. In W. Burr, F. Hill, F. Nye, & I.

Reiss (Eds.), Contemporary theories about thefamily: Research based theories (Vol. 1, pp. 101 –

126). New York: Free Press.

Parke, R., & Buriel, R. (1998). Socialization in the fam-

ily: Ethnic and ecological perspectives. In W. Damon

(series ed.) & N. Eisenberg (volume ed.), Handbookof child psychology: Social and personality develop-ment (pp. 463 – 552). New York: Wiley.

Pears, K., Pierce, S., Kim, H., Capaldi, D., & Owen,

L. (2005). The timing of entry into fatherhood in

young, at risk men. Journal of Marriage and Fam-ily, 67, 429 – 447.

Petersen, A., Sarigiani, P., & Kennedy, R. (1991).

Adolescent depression: Why more girls? Journalof Youth and Adolescence, 20, 247 – 271.

Rindfuss, R., & Vandenheuvel, A. (1990). Cohabita-

tion: A precursor to marriage or an alternative to

being single. Population and DevelopmentalReview, 16, 703 – 726.

Rutter, M., & Quinton, D. (1984). Long-term follow-

up of women institutionalized in childhood:

Factors promoting good functioning in adult life.

British Journal of Developmental Psychology, 2,

191 – 204.

Steinberg, L., Lamborn, S., Darling, N., Mounts, N., &

Dornbusch, S. (1994). Over time changes in

adjustment and competence among adolescents

from authoritative, authoritarian, indulgent,

and neglectful families. Child Development, 65,

754 – 770.

Teachman, J., Polonko, K., & Leigh, G. (1987). Mar-

ital timing: Race and sex comparisons. SocialForces, 66, 239 – 268.

Woodward, L., Fergusson, D., & Horwood, L. J.

(2001). Risk factors and life processes associated

with teenage pregnancy: Results of a prospective

study from birth to 20 years. Journal of Marriageand Family, 63, 1170 – 1184.


early family transitions and depressive symptom changes from adolescence to early adulthood

Documents