Developmental Psychology Copyright 2003 by the American Psychological Association, Inc.

2003, Vol. 39, No. 4, 657– 669 0012-1649/03/$12.00 DOI: 10.1037/0012-1649.39.4.657

Variations in Patterns of Developmental Transitions in the

Emerging Adulthood Period

Patricia Cohen, Stephanie Kasen, and Henian Chen Claudia Hartmark and Kathy Gordon
New York State Psychiatric Institute and Columbia University New York State Psychiatric Institute

The assumption of adult roles has largely been examined as status changes in school attendance, leaving
the parental home, and marriage. Nevertheless, levels of autonomy and individuation vary considerably
within these states. This study obtained such information through narrative behavioral descriptions within
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

financial, residential, romantic, and family formation domains covering ages 17 to 27 years. Analyses of
data from 240 members of a community-based longitudinal study investigated the association of
trajectories in these domains with family socioeconomic status, parental divorce, gender, and race.
Findings indicated that subgroup differences were not generally attributable to educational enrollment.
Both within and between domains, many individuals showed dramatic changes in the assumption of adult
roles, returning to more dependent, other-determined roles for short or even extended periods.

Recent theoretical and empirical work has identified the transi- of the parental home. The current study addressed this develop-
tion from adolescence to adulthood as a period with distinct mental period with measures that reflect both these status variables
characteristics and importance for the understanding of human and the level of independence and responsibility assumed within
development. Arnett (2000) labeled the period from ages 18 to 25 new roles and settings. Consisting of narrative descriptions of
emerging adulthood, making the case for the view that this period behavior from age 17 to age 27, the data permit close examination
is theoretically and empirically distinct from both adolescence and of individual developmental trajectories and average differences
young adulthood. Within Western developed societies, those in associated with parental socioeconomic status (SES), sex, and
this age group have left the dependency of childhood and adoles- race.
cence but have not yet assumed the enduring responsibilities that For example, people ages 18 to 25 have the highest rates of
are more normative in adulthood. They face a formidable array of residential change of any age group (Arnett, 2000; Goldscheider &
new experiences and tasks requiring the development of new Goldscheider, 1994). By age 25, only a little more than one fifth
knowledge and skills in order to experience a successful transition still live with parents (Mogelonsky, 1996). However, this devel-
into adulthood (Arnett, 1997, 2000; Arnett & Taber, 1994; George, opmental transition, like the other transitions taking place during
1993; Hogan & Astone, 1986; Mogelonsky, 1996; Shanahan, this period, is potentially conditioned on the norms and opportu-
2000). nity structure experienced differently by the demographic sub-
Despite the strong theoretical case for viewing this age span as groups of which young people are part. Therefore, determining
a distinct developmental period, much of our current knowledge is demographic variations in these transitions is an essential aspect of
based on changes in developmental “markers” such as moving out understanding this life stage.

Historical Changes in Social Roles During the

Patricia Cohen, Epidemiology of Mental Disorders Research Unit, New Emerging Adulthood Period
York State Psychiatric Institute; Department of Psychiatry, College of
Physicians and Surgeons, Columbia University; and Department of Epi- Shanahan (2000) noted that modernization of societies has co-
demiology, Mailman School of Public Health, Columbia University. incided with both standardization and individualization of the life
Stephanie Kasen, Epidemiology of Mental Disorders Research Unit, New course. Standardization appears in increasing “compactness” in the
York State Psychiatric Institute, and Department of Psychiatry, College of ages of school completion, marriage, parenthood, and beginning a
Physicians and Surgeons, Columbia University. Henian Chen, Epidemiol- career, whereas individualization is found in increasingly diverse
ogy of Mental Disorders Research Unit, New York State Psychiatric sequences of these markers (Shanahan, 2000). Thus, the process of
Institute, and Department of Epidemiology, Mailman School of Public individuation and assumption of adult roles as well as adult status
Health, Columbia University. Claudia Hartmark and Kathy Gordon, Epi- positions may not be uniform across these life domains. In addi-
demiology of Mental Disorders Research Unit, New York State Psychiatric tion, the opportunities for separation from the more dependent
Institute.
roles of childhood are affected by economic fluctuations and
This study was supported by National Institute of Mental Health Grant
R01-MH54161 to Patricia Cohen.
discrete historical events (Elder, 1998; Elder & Conger, 2000).
Correspondence concerning this article should be addressed to Patricia Perhaps one of the most remarkable trends in the latter part of
Cohen, Department of Psychiatry, Suite 16D, College of Physicians and the 20th century has been the extension of this period of develop-
Surgeons, School of Public Health, Columbia University, 100 Haven ment, which suggests a societal tendency to delay processes of
Avenue, New York, New York 10032. E-mail: prc2@columbia.edu individuation and autonomy. Young people remain in the parental

657
 658 COHEN, KASEN, CHEN, HARTMARK, AND GORDON

home longer and are more likely to return temporarily (Gold- Childhood experience of parental divorce may impact the tran-
scheider & Goldscheider, 1994; Mogelonsky, 1996). They become sition to adulthood, particularly the assumption of romantic com-
financially independent more gradually and at a later age. They mitment, both by increasing a general wariness of the likelihood of
take longer to settle into occupational tracks and are more likely to long-term mutuality and by raising the offspring’s standards for
have extended and part-time educational careers (Arnett, 2000; relationship quality. Other evidence suggests that the lower family
Shanahan, 2000). Recent generations of men and, especially, of income in divorced families may also have an impact on offspring
women are much more likely to pursue higher education. This financial transitions (Cooney, 1989; McLeod, 1992). In a Finnish
tendency may account for some of the delays in the development sample, 22-year-olds from divorced families had less education
of independence from the family of origin. In past generations, the and an earlier transition to parenthood and working life (Aro &
early assumption of full-time employment and an occupational line Palosaari, 1992).
was a major influence on developmental course and psychological
change, at least for men (Kohn & Schooler, 1983). The extension Research Designs Used to Study the
of the educational period, and the frequent delay of career com-
Emerging Adult Period
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

mitment, may mean that such influences have declined, especially

This document is copyrighted by the American Psychological Association or one of its allied publishers.

in the late-adolescent/early-adult period. Both men and women Status Indicators

have much-delayed marriages compared with earlier generations
(Bloom & Trussell, 1984; Chen & Morgan, 1991; Rindfuss, 1991; Typical markers of entry into adulthood are the establishment of
Rindfuss, Morgan, & Swicegood, 1988) as well as later family a stable residence, school completion, career selection and train-
formation (Conger, Cui, Bryant, & Elder, 2000; Wadsworth, Mac- ing, and marriage or a long-term commitment with a romantic
lean, Kuh, & Rodgers, 1990), and in increasing proportions, many partner. However, the characteristics that matter most to young
remain permanently childless (Rindfuss, 1991; Rindfuss & Parnell, people in their subjective sense of attaining adulthood are not these
1989). Perhaps one of the most interesting questions is whether the “status” markers. Their essential criteria include accepting respon-
increase in divorce rates over the past decades accounts in part for sibility for themselves, making independent decisions, becoming
these historical changes. With regard to parental attitudes toward financially independent, and becoming a parent (Arnett, 1997,
these changes, there is probably more ambivalence than ever about 1998, 2000; Greene, Wheatley, & Aldava, 1992). Nonetheless,
pushing offspring to attain full independence. It is not clear to these role changes have been little studied other than by status
parents if prolongation of partial dependency is likely to contribute indicators such as marriage, employment, leaving the parental
positively to offspring well-being, nor is it clear whether it will home, or starting a full-time job, in each of which the extent that
foster the well-being of society as a whole. a fully adult role is assumed may vary substantially. Research in
the past two decades has examined the variability of status mark-
ers, including predictors of different patterns, their sequencing,
Demographic Differences in the Transition to Adulthood their degree of co-occurrence, and the duration of intervals among
them (George, 1993; Rindfuss & Parnell, 1989; Rindfuss, Swice-
The experience of becoming an adult often differs by ethnic good, & Rosenfeld, 1987; Shanahan, 2000). Most such research
groups, by family SES, and for men compared with women. has used event-history models to examine the timing of and
Historically, persons of lower SES, African Americans and His- precursors to these transition markers.
panics, and rural residents tended to complete these transitions Reliance on these transition markers may be a major limitation
earlier than their peers. In general, women experience these tran- of prior work. First, unlike other transitions (e.g., birth, entering
sitions earlier than men (George, 1993). Young women leave school, and death), many aspects of the transition to adulthood are
home at a younger age than do young men. Some of this difference reversible; for example, young people may leave the home of
relates to women’s younger age at marriage, but even unmarried origin and later return. Thus, the markers fail to capture the
women are more likely than men to live away from parents within-individual variability over time in the transition process.
(Goldscheider, Thornton, & Young-DeMarco, 1993). The extent to Second, the transitions are more complex than discrete, clearly
which these departures actually signal earlier assumption of adult bounded occurrences, as the use of event-history or logistic re-
roles with regard to separation and independence is less clear. gression models would suggest. Most important for the current
Socioeconomic resources in families of origin may affect the investigation, status markers fail to capture the substantial vari-
transition to adult roles. Among youths from high-SES families, ability of functional roles—the level of autonomy and individua-
increased opportunities for postsecondary education may delay tion—within these categorical changes.
residential and financial independence. For example, youths who
attend college often continue to rely on parents for financial
The Current Study Design
support and, albeit to a lesser extent, residential support; however,
youths who enter full-time employment are more likely to be In this article, we report findings from a community-based
financially self-supporting and to live in a residence independent longitudinal study that included both qualitative and quantitative
of parents (Goldscheider & Goldscheider, 1994). On the other data during the transition from ages 17 to 27. The study as a whole
hand, although it is delayed somewhat, later residential and finan- was designed to increase understanding of this developmental
cial independence among college youths is facilitated by advanced period and of the influences on the processes of separation, indi-
education (Whittington & Peters, 1996). Details on whether and viduation, and autonomy over its course. Potential influences
how higher education accounts for these pattern differences in the include demographic subgroups. The major purposes of the anal-
assumption of other adult roles are lacking. yses reported here were (a) to examine individual variations in the
 PATTERNS IN EMERGING ADULTHOOD 659

patterns characterizing the assumption of adult roles in the areas of were consistently high in symptoms, those consistently low, those who
residence, finance, romance, and parenting; (b) to test differences changed from high to low, and those who changed from low to high, and
in the average or expected growth patterns in these trajectories to have each group approximately equally represented. For the purposes of
associated with sex, parental SES, race, and parental divorce; and this report, respondents are weighted back to reflect their prevalence in the
overall population. An additional 40 subjects were randomly chosen from
(c) to determine the extent to which these differences were attrib-
the younger members of the CIC cohort to participate in a coordinated
utable to participation in educational programs.
methodological study. For these subjects, the transition variables used here
Specifically, in keeping with prior theory and empirical work, are based on the prospective data collected in three interviews spaced
we investigated the following hypotheses with regard to demo- approximately 2 years apart. The data did not reveal any average differ-
graphic influences on trajectories over the period from age 17 to ences in role transitions between these two subsamples, and thus they were
age 27. We predicted earlier assumption by young women of more combined to better represent the population of origin. The combined
adult roles with regard to residential independence, romantic com- sample of 240 individuals has a slightly narrower age range than the decade
mitment, and child rearing and family formation. Youth from range of the original sample (SD ⫽ 2.49 years compared with 2.86 years);
higher parental SES families were expected to remain more de- on average, they may be considered comparable to a 1970 birth cohort.
pendent on parents with regard to residence and financial support Most of the participants were White (92.5%), and 52.1% were female.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Like the full cohort, study participants were from a broad spectrum of
This document is copyrighted by the American Psychological Association or one of its allied publishers.

for a longer period, although they were expected to show some

socioeconomic backgrounds and mixed rural, suburban, and urban resi-
catching-up as reflected in an increase in the trajectory in the latter
dences. We measured family SES by combining standardized measures of
years of this period. We advanced no specific hypotheses about years of maternal and paternal education, maternal and paternal highest
differences between Black young people and the White majority occupational status, and family income (categorized prior to standardiza-
beyond those that may be attributable to the SES differences in tion in order to moderate skew), which were based on our 1986 maternal
families. We examined the hypothesis that offspring of divorced interviews when the children were still living at home.
parents might delay the assumption of a more adult (committed) In this sample, 22.5% (n ⫽ 54) had experienced parental divorce when
romantic role. In keeping with recent empirical work, we predicted they were under age 13, an age selected to ensure residence in a parental
that earlier assumption of adult financial roles would also charac- home at the time. Although we did not make specific predictions about race
terize offspring of divorced parents. Finally, we explored the in these models, we included race of the participant as a variable despite the
small Black sample (18 participants) because of the often substantial effect
extent to which these differences might reflect differential ad-
of race on roles and opportunities in American society.
vanced education.
The youth cohort studied here was born over a 7-year period
(1967–1973). Analyses examining possible differences that might Narrative Interviews
be due to variations in the national economic conditions into which
these youth entered upon leaving high school showed no signifi- The narrative interviews, which typically took between 3 and 5 hr to
cant differences and are not discussed further. complete, were carried out by telephone between 1997 and 2000. Inter-
viewers were intensively trained and supervised, and a manual was used
that provided many illustrations of the match of concrete behavioral
Method examples to scale points. The data collection in this study was begun by
establishing a framework for the 10-year period from age 17 to age 27
Study Participants through asking the participants to complete a “life chart” (Lyketsos, Nes-
tadt, Cwi, Helthoff, & Eaton, 1994). This document, completed in advance
The sample consisted of 240 members of the Children in the Community of the telephone interview, charted the changes over these years in where
(CIC) cohort of 800 young adults who were studied since early childhood the participant lived, worked, and studied and the dates of significant
when they were randomly selected on the basis of residence in one of two experiences that occurred then, such as graduations, special vacations or
upstate New York counties (Kogan, Smith, & Jenkins, 1977). Families trips, engagements, marriages, births, major illnesses or victimizations,
ranged from affluent to living in poverty and resided in urban, suburban, special achievements or honors, or any other experiences that helped the
and rural settings (for further sample selection criteria and characteristics, participant reconstruct the period being covered by the interview. The life
see P. Cohen & Cohen, 1996). The cohort, 50% female and primarily chart was retained by the participant and was reviewed with the interviewer
White (91% vs. 8% Black and 1% other), ranged in age from 1 to 10 years in order to establish a common framework for the subsequent narrative.
old in 1975; at that time, their mothers were interviewed about a wide range The narrative in this study covered six domains of the person’s life:
of individual characteristics of the target study child and about demo- residence, finance, school, employment or career, romance, and parenting.
graphic and contextual risk factors. Mothers and cohort offspring were Interviews elicited the more concrete facts regarding where and with whom
interviewed four additional times with expanded parallel protocols that also respondents lived, worked, and studied and by what financial means they
included psychiatric diagnostic interviews. At all interviews, consent was were supported between ages 17 and 27. Participants were asked to report
obtained from all participants according to institutional review board as many behavioral descriptions as possible (e.g., “Who paid for
standards, and a National Institute of Health Certificate of Confidentiality that?” “What did you do there?” “Who signed the lease?”“How often did
exists for these data. you see each other?” “Who did the shopping?”) in order to allow the coders
For the present transition study, 200 subjects from the CIC cohort were to assess the level of independence and responsibility assumed within the
selected in a “four corners” design and were limited to those who would be domain.
at least 27 years old when interviewed. That strategy was in keeping with The interviews began by establishing the dwelling setting and the
the ultimate goal of this investigation, which was to determine the recip- behavior of the participant within that setting at the time of the 17th
rocal influences between the course of personality disorders from adoles- birthday with regard to concrete aspects including other occupants, routine
cence to adulthood and role transitions from ages 17 to 27. A scatterplot of chores, and responsibilities for household tasks. We found that having the
personality disorder symptom levels, obtained prospectively by structured participant describe and date changes that took place in one domain at a
interview at mean ages of 15.5 and 22.2 years allowed us to randomly time usually worked best. It also allowed the interviewer to check the
select from its quadrants (i.e., each of the “four corners”) those youths who various dates of change in one domain against dates provided for other
 660 COHEN, KASEN, CHEN, HARTMARK, AND GORDON

domains to make sure that they were consistent. Often those checks helped
the respondent to sort out sequence uncertainties. Changes in study vari-
ables were dated to the month in which they occurred. Thus, coding of the
structured narrative generated 120 consecutive monthly records on each of
the defined variables, and the value of each code was retained until the
month when a new code was entered. When there were multiple changes
in a single domain within a month, the coders attributed the changes to
consecutive months, making sure to retain the sequence indicated by the
participant. In order to ensure reliability of measures, all interviewer-coded
interviews were tape recorded and blindly recoded by another interviewer,
and the scores were averaged across coders (for more about this narrative
interview, see P. Cohen, Chen, Hamigami, Gordon, & McArdle, 2000;
reliabilities are reported below).
This article examines the age patterns in adult role assumption for the
sample as a whole and the demographic effects on these trends over age 17
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

to age 27 as reflected in the transitions in residence, financial support,

This document is copyrighted by the American Psychological Association or one of its allied publishers.

romantic involvement and commitment, and family formation and child

rearing as well as in the timing of involvement in educational programs.
(Analyses of the qualities of these transitions [e.g., instrumentality, iden-
tification with role, external difficulties encountered, and quality of coping]
and of the influence of social support and conflict and life events during
this period will be the topics of separate reports.) In addition, the com-
Figure 1. Illustrative example of residential transition levels.
plexities of interrelationships among these transitions over the 120 months
are examined in the current analyses.

rearing. Interrater reliabilities were computed as correlations between

Monthly Measures of Domain-Specific Transition Level interviewer codes and blind recodes of taped interviews by another inter-
viewer. Reliability estimates are stepped up to reflect the higher reliability
As noted above, the central construct of the study was the extent to of the composite of two ratings. Because the means and variances of the
which, at each point in time, the participant’s behavior was more like that two sets of codes were necessarily approximately equal, these correlations
of a child or approximated fully adult role behavior, defined by indepen- are equivalent to intraclass correlations. Correlations were determined for
dence from parental control, expression of own goals and preferences, and the entire sample and were based on 28,800 ratings (120 months ⫻ 240
assumption of responsibilities. These ratings of transition level (TL) were subjects). Reliabilities for residential TL, financial TL, romantic TL, and
quantified in each of four domains—residence, finance, romance, and parental TL were .92, .91, .91, and .98, respectively.
parenting. Role behavior was rated on a 100-point scale, where 0 repre- In addition to the four TL variables examined here, we collected data on
sented a fully childlike role and 99 represented a fully adequate adult role. employment/career TL. We had little problem rating TL for the extent to
This scale enabled us to characterize changes in score units as percentage which the amount and type of employment or business venture represented
point changes (P. Cohen, Cohen, Aiken, & West, 1999). movement toward a long-term occupational career, which we considered to
In the area of residence, TL scores ranged from 0, which indicated living reflect more adult role behavior. However, we had much trouble deciding
in a parental home with no responsibilities, to 99, which indicated that the how to characterize months in which the individuals were not employed,
participant carried all the responsibilities of establishing and maintaining often because of full involvement in school or parenting. This variable was
an independent residence. To facilitate an understanding of how the par- also probably the one for which it was the most difficult to avoid the
ticipant’s narrative was coded into the TL scale for residence, we present influence on ratings of social class considerations regarding occupations.
in Figure 1 a fictitious participant. At age 17, this participant was living in For this reason, and because of the complexity of the material already
the family home, and the description of his responsibilities and rights with included in these analyses, we decided to limit the current study to the four
regard to this residence led the interviewer to code the TL at 25. At the age TL variables previously mentioned.
of 18 years 2 months, he left for a college dormitory, where his role in
selecting furnishings and maintaining his quarters led to a rating of 36. Models
Thirteen months later, he moved with friends to an apartment from which,
after 17 months, he moved to an apartment with his girlfriend, with We used the SAS PROC MIXED program to estimate random and fixed
increasing residential role responsibility in these settings. A romantic effects on growth patterns for each of the TL variables (Littell, Miliken,
breakup led him to take more responsibility for his now sole occupancy, Stoup, & Wolfinger, 1996; McArdle & Bell, 2000; Singer, 1998). In these
but in less than 1 year, he moved back home. At about age 24, he remained models, longitudinal data on individuals are considered the basic “random”
in his parental home but assumed more responsibilities than at the younger data, just as in cross-sectional data the individual subject variables are the
age, and at about age 26, he established a new residence with his fiancee. basic units of analyses. Thus, the first step in these analyses examined the
Financial TL was similarly rated from 0, indicating that the participant between-subjects differences (“random effects”) in mean level; in linear,
was completely supported by others, usually family, to partial self-support quadratic, and cubic slope over age; and in the autoregressive error. For our
(scores under 50, depending on the degree), to 99, a score that indicated TL variables, the linear changes over age were not sufficient to describe the
stable, secure, and adequate (although not necessarily high-income) pro- basic growth shape. We found substantial random variance (individual
vision for the financial needs of the self and potential dependents. For differences) in mean TL and in the linear and quadratic (but not cubic)
romantic TL, 0 indicated an absence of interest in romantic involvement, changes over age. Furthermore, we found, as expected, that residuals were
and 99 reflected a sustained, reciprocal romantic commitment, usually not independent over adjacent months and added autocorrelation to the
including marriage. Parental TL was rated from 0, indicating the partici- random model. This random model was then the basis for examination of
pant had no children and no interest in conceiving a child in the foreseeable the fixed effects of sex, SES of family of origin, childhood parental
future, to 99, which reflected a highly committed involvement in child divorce, and race of respondent on the mean TL and linear and quadratic
 PATTERNS IN EMERGING ADULTHOOD 661

Table 1
Percentages of Participants in Marker States Between Age 17 and Age 27

Age

Marker state 17 18 19 20 21 22 23 24 25 26 27

In school 93.3 67.0 49.6 40.9 36.3 31.6 16.2 15.4 13.0 13.7 16.3
Employed
Part time 53.8 45.1 29.6 21.2 20.8 16.7 12.5 10.9 9.6 10.8 9.5
Full time 19.2 27.4 45.4 53.8 56.7 63.4 71.3 76.4 77.9 74.7 71.0
Living with family 96.7 84.2 59.2 47.9 41.3 41.7 33.3 27.1 23.3 20.0 15.0
Supported by family 96.3 87.5 69.6 60.4 52.5 42.1 32.1 24.6 18.3 15.1 9.5
Ever pregnant 6.7 11.7 15.1 19.2 22.5 26.7 30.0 33.3 36.3 42.9 47.5
Biological offspring 1.3 2.5 6.6 9.2 12.1 14.6 19.2 20.0 22.9 29.3 37.9
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

age changes in TL. Each model was tested for improvement of fit to the participants were strikingly less likely to have full-time jobs
data with chi-square. Following linear and quadratic fixed effects of age, (44.4%) than were White participants (80.6%).
the main effects for the demographic variables were added, followed by At age 25, more than twice as many men as women were still
selected demographic interactions. Next, interactions of demographic vari- living with family (32.2% vs. 15.2%), and nearly twice as many
ables with the changes over age were added to test the demographic
men as women were primarily supported by their families (24.3%
impacts on slope. These began with the linear slopes as a set of predictors
vs. 12.8%). Race differences in living with family and family
and were followed by the quadratic slope component interactions. Finally,
we tested triple interactions among the demographic variables and slope support were not striking, but differences by parental SES were
effects. Because of the large number of variables and, in particular, the large. Those from lower SES backgrounds were more likely to be
large number of interactive and curvilinear relationships tested, the esti- living with (33.6%) and supported by (23.9%) family than were
mates reported here are those from the simplest models consistent with the those from higher SES backgrounds (14.2% and 13.4%,
data according to these tests. respectively).
In the next analyses we added the time-varying measure of educational Women were more likely to be sexually active, to have been
involvement (coded here as yes or no because of the distributional prob- married, to have been in a relationship that resulted in a pregnancy,
lems caused by the sparse number of participants reporting part-time and to have living biological offspring than were men. Nearly half
involvement in any given month). The extent to which higher education
of the women had been married by age 25, compared with only just
accounted for other demographic influences tested our hypothesis about
over a quarter of the men. Comparable differences were present for
this variable as a mediator.
a history of having conceived a child, and about three times as
many women (33.6%) as men (11.3%) had biological offspring.
Results Lower SES background was associated with much higher rates of
being sexually active (93.8% vs. 77.2%), married (45.1%
Transition Status Markers: Age 17 to Age 27 vs. 31.5%), and having biological offspring at age 25 (35.4%
Table 1 shows status on transition markers from age 17 to vs. 11.8%) than was higher SES background. Black participants
age 27. At age 17, most respondents were in high school, living in were much less likely to have been married than were White
a parental home, and supported primarily by family. About one participants (22.2% vs. 39.2%), but they were more likely to have
fifth of the respondents were working full time at the time of their offspring (55.6% vs. 20.2%).
17th birthdays, primarily those with birthdays during the summer.
Over two fifths were sexually active, but fewer than 1% had been Demographic Differences in Educational Enrollment
married and fewer than 2% were actively parenting. When these
The first analyses of the longitudinal change data examined the
young people were 27, about 16% were still attending school, over
demographic differences in school attendance over the 120 months
80% were employed, mostly full time, and fewer than 10% were
from age 17 to age 27 with a multilevel logistic regression model.
still financially supported primarily by their families. At age 27,
There were significant sex and SES differences in the proportion
15% were still living with family, but nearly all were sexually
of the sample enrolled in educational programs over this period.
active and nearly half were or had been married and had a history
Almost all participants were attending school at age 17, but men
involving pregnancy. Nearly 40% had living biological offspring.
were 21⁄2 months older when they graduated than were women, on
In Table 2, one can see demographic differences in status at
average. Over the next 4 years, women were in school in higher
age 25, including a comparison of the youth above and below the
proportions than were men. By age 22, most of the sample were
average on the family SES index. Current school attendance was
not enrolled in educational programs, and the sex differences in the
lower for those from low-SES families (8.8%) compared with
later period were no longer significant. SES effects on school
those from higher SES families (16.5%) but did not differ by sex
nor substantially by race.1 Participants from low-SES families
were more likely to have full-time employment (86%) than were 1
Note that the Black sample of 18 participants comprised men and
those from high-SES families (71%). Women were less likely to be women equally. Because this sample is small, we do not present individual
working full time than were men (73.6% vs. 82.6%), and Black age curves, nor do we focus heavily on these differences.
 662 COHEN, KASEN, CHEN, HARTMARK, AND GORDON

Table 2
The Percentages of Demographic Subgroups in Marker States at Age 25

Total Males Females White Black Low SES Higher SES

Marker state (N ⫽ 240) (n ⫽ 115) (n ⫽ 125) (n ⫽ 222) (n ⫽ 18) (n ⫽ 127) (n ⫽ 113)

In school 13.0 13.0 12.8 13.1 11.1 8.8 16.5

Employed
Part time 9.6 10.4 8.8 7.2 38.9 11.5 7.9
Full time 77.9 82.6 73.6 80.6 44.4 85.8 70.9
Living with family 23.3 32.2 15.2 23.0 27.8 33.6 14.2
Family support primary 18.3 24.3 12.8 18.5 16.7 23.9 13.4
Sexually active 85.0 80.9 88.8 86.0 72.2 93.8 77.2
Ever pregnant 36.3 24.3 47.2 34.2 61.1 47.8 26.0
Biological offspring 22.9 11.3 33.6 20.3 55.6 35.4 11.8
Ever married 37.9 26.1 48.8 39.2 22.2 45.1 31.5
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Note. SES ⫽ socioeconomic status.

enrollment began as soon as the high school years were completed Correlations Among the TL Variables and School
and continued throughout the remaining years, although the pro- Attendance
portion of the entire sample in school dropped to levels below 10%
in all of the later years (see Figure 2). There were no race School attendance was moderately negatively correlated with
differences that were independent of SES effects. TL across the decade for the sample as a whole, with correlations
ranging from ⫺.26 with financial TL to ⫺.33 with residential TL.
Transition Level Scales Of course, much of this relationship simply reflects the temporary
role that school attendance plays in this developmental period. For
Individuals varied enormously in the shape of their role assump- the period as a whole, the correlations across participants between
tion curves for residential TL over these ages. Some individuals average school attendance and average TL ranged from ⫺.47 with
remained in parental homes with little or no change in the level of parenting to ⫺.23 and ⫺.19 with romantic TL and residential TL,
responsibility and independence assumed in that setting over the respectively, to .10 with financial TL, which suggests that ulti-
entire decade. Others moved rapidly toward complete residential mately school “pays off.” For the second 5-year period, when
independence, sometimes returning one or more times to a past school attendance was low, the correlations of monthly attendance
role in the family residence. Yet others followed the age group with transition variables ranged from ⫺.01 with financial TL to
average, moving in incremental steps toward more independence ⫺.18 with parenting TL.
and responsibility. Such individual variability in change over time The TL variables generally correlated positively with one an-
characterized all of the TL data. other, although financial TL and parental TL were essentially
uncorrelated (r ⫽ ⫺.03). Financial TL also correlated modestly
with romantic TL (.16). The remaining correlations ranged from
.43 to .47, except for the more substantial relationship between
residential TL and romantic TL (.54). These correlations also
reflect the general tendency for developmental changes in TL
across the decade as well, and correlations for shorter periods of
time, such as the first 5 years or the second 5 years, were generally
much lower.

Results From the Two-Level Linear and Quadratic

Growth Models
The sequence of models tested in these analyses included ran-
dom effects2 and the fixed effect of intercept followed by (a) linear
and quadratic components of age slope, (b) main effects of sex,
race, parental SES, and parental divorce, (c) interactions between
sex and the other demographic variables, (d) different linear age
slopes by demographic variables, (e) different quadratic changes in
age by demographic variables, (f) different age changes associated

2
Simpler random effects models were also tested, but the included
Figure 2. Proportion of sample in school by age and family socioeco- components invariably added significantly, and thus these simpler models
nomic status (SES). are not presented here.
 PATTERNS IN EMERGING ADULTHOOD 663

with interactive effects of sex, race, and parental SES, and (g) the Residential transition level. On average, residential TL in-
effects of time-varying school attendance on the TL and on the creased linearly, about 4.25 percentage points (PP) per year, from
estimated effects of demographic variables. The chi-squares for the about 26.7 at age 17 to about 69.0 at age 27. A statistically
improvement in fit of each cumulative model to the data are significant mean difference across the age range was found be-
reported in Table 3. As can be seen, the combined linear and tween women and men, women being 10.83 PP higher on average
quadratic slope main effects were significant for all TL variables, and showing a 0.62-PP greater increase per year than men. How-
as were the collective main effects of the demographic variables. ever, the average curve shapes were not equal for men and women
Although the interactions of other demographic variables with sex (see Figure 3). Women assumed more adult residential roles earlier
were not significant in aggregate, individual interaction effects in than did men and then leveled off, whereas men began to “catch
these sets were significant for financial and romantic TL. There up” in their middle 20s. Other demographic variables, including
were also significant demographic differences in average age parental SES and divorce history, were not related to either the
curves for each TL variable. Finally, on romantic TL, there were mean residential TL or to changes from age 17 to age 27. Although
significantly different patterns by age for Black and White partic- there was no mean difference between White and Black partici-
ipants, depending on whether they were men or women. pants at age 22, White participants’ increase in residential TL
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Following selection of the best model on statistical grounds, was 1.31 PP per year more than that of Black participants, as tested
variables without significant effects were removed from the mod- by the Race ⫻ Age interaction. The absence of a race main effect
els unless they were part of significant higher order terms (qua- reflects the higher residential TL of Black participants in their
dratic or interaction terms) in order to simplify the presentation. early 20s followed by a somewhat lower residential TL in their
Table 4 presents the estimates generated in the final models by middle to late 20s. When we added the education variable as a
two-level linear and quadratic growth models for the four TL time-varying covariate, it made a significant prediction, with
variables. a 1.85-PP higher level of residential TL for those who were in

Table 3
Tests of the Improvement in Fit Over the Model With Random Effects Plus Intercept, as
Additional Fixed Predictor Sets Were Added

Fixed predictor set ⌬␹2 df p

Residence
Linear and quadratic age slope 401.2 2 ⬍.001
⫹ Main effects for demographic variables 14.1 4 ⬍.01
⫹ Demographic interactions 4.0 3 ns
⫹ Demographic ⫻ Linear Age 19.4 4 ⬍.001
⫹ Demographic ⫻ Quadratic Age 12.8 4 ⬍.05
⫹ Sex by SES ⫻ Age and Sex by SES ⫻ Quadratic Age 2.4 2 ns

Finance

Linear and quadratic age slope 343.7 2 ⬍.001

⫹ Main effects for demographic variables 11.7 4 ⬍.05
⫹ Demographic interactions 7.2 3 ⫽.12
⫹ Demographic ⫻ Linear Age 16.5 4 ⬍.001
⫹ Demographic ⫻ Quadratic Age 8.2 4 ⫽.12
⫹ Sex by SES ⫻ Age and Sex by SES ⫻ Quadratic Age 2.4 2 ns

Romance

Linear and quadratic age slope 215.0 2 ⬍.001

⫹ Main effects for demographic variables 32.5 4 ⬍.001
⫹ Demographic interactions 6.1 3 ⫽.15
⫹ Demographic ⫻ Linear Age 7.0 4 ns
⫹ Demographic ⫻ Quadratic Age 9.9 4 ⬍.05
⫹ Sex by Race ⫻ Age and Sex by Race ⫻ Quadratic 0.7 2 ns
Age

Parenting

Linear and quadratic age slope 130.2 2 ⬍.001

⫹ Main effects for demographic variables 3.2 4 ⫽.10
⫹ Demographic interactions 2.1 3 ns
⫹ Demographic ⫻ Linear Age 41.5 4 ⬍.001
⫹ Demographic ⫻ Quadratic Age 13.5 4 ⬍.01
⫹ Sex by SES ⫻ Age and Sex by SES ⫻ Quadratic Age 6.7 2 ⬍.05

Note. N ⫽ 240. SES ⫽ socioeconomic status.

 664 COHEN, KASEN, CHEN, HARTMARK, AND GORDON

Table 4
Estimated Covariance and Fixed Effects From Best-Fitting Linear and Quadratic Models of Domain Transition Levels Across
120 Months (N ⫽ 240)

Residential TL Financial TL Romantic TL Parenting TL

Model Estimate SE Estimate SE Estimate SE Estimate SE

Random effects

Intercept 147.26 18.51 174.89 26.28 416.11 48.76 3.14 64.25

Slope 1.84 0.41 2.61 0.69 5.77 1.04 3.16 1.06
Intercept/slope 7.53 1.89 7.19 2.73 17.49 4.99 32.48 5.09
Slope2 0.16 0.04 0.15 0.06 0.42 0.09 0.19 0.05
Slope/slope2 0.22 0.08 0.49 0.13 0.23 0.21 0.61 0.17
Intercept/slope2 ⫺4.23 0.75 ⫺4.18 1.03 ⫺9.31 1.79 ⫺3.47 1.29
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Autoregressive error 0.91 0.004 0.93 0.004 0.90 0.004 0.99 0.001
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Residual 159.44 7.25 240.06 13.54 358.76 15.30 533.47 51.77

Fixed effects

Intercept (age 22) 47.84 0.92 48.75 1.07 45.98 1.63 10.89 1.40
Age 4.25* 0.14 3.91* 0.17 4.09* 0.22 2.93* 0.22
Age2 ⫺0.07 0.04 ⫺0.36* 0.05 ⫺0.07 0.07 0.18* 0.05
Sex ⫺10.83* 1.82 4.02* 1.25 ⫺20.62* 3.09 ⫺8.75* 2.07
Race ⫺3.38 2.32 2.61 2.55 ⫺9.43 5.93 14.56* 5.49
SES ⫺0.14 0.48 ⫺0.20 1.10 ⫺1.72 1.11 ⫺4.95* 1.11
Divorce — — — — — — — —
Sex ⫻ Race — — — — 35.01* 11.83 — —
Sex ⫻ SES — — ⫺3.35* 1.30 4.27* 2.20 — —
Sex ⫻ Age ⫺0.62* 0.26 — — ⫺0.001 0.44 ⫺1.72* 0.41
Race ⫻ Age ⫺1.32* 0.48 — — ⫺1.70* 0.83 ⫺0.24 0.87
SES ⫻ Age — — 0.72* 0.17 — — ⫺0.94* 0.22
Sex ⫻ Age2 0.29* 0.08 — — 0.51* 0.13 — —
SES ⫻ Age2 — — 0.14* 0.05 — — — —
Race ⫻ Age2 — — — — 0.06 0.25 ⫺0.65* 0.19
Sex ⫻ Race ⫻ Age — — — — ⫺3.26* 1.66 — —

Note. All entries are maximum-likelihood estimates; dashes indicate variables not included in the best-fitting model. Age was centered at 22 years; sex
was centered as .5 ⫽ male, ⫺.5 ⫽ female; race was coded 0 ⫽ White, 1 ⫽ Black. Divorce was coded 1 (vs. 0) but was not included in any final model.
Thus age main effects reflect age changes for the average White person. SES ⫽ socioeconomic status.
* p ⬍ .05.

school at the time. This reflects the fact that school attendance was no significant SES background difference in the average
rates were highest in the first 3 or 4 years of the period, when the financial TL over this period, young people from high-SES (1 SD
alternative was often to remain in the parental home. The inclusion above the mean) families had, on average a 1.44-PP (⫽ 2 ⫻ .72)
of school attendance in the model, however, did not alter the greater increase per year than those from low-SES (1 SD below the
demographic influences of race and sex. mean) families. This difference was not linear, however, and those
One may evaluate the extent to which these demographic pre- from higher SES families showed a gradually accelerated pattern,
dictors accounted for the individual variance in trajectories by whereas those from lower SES families showed a higher initial
comparing the variance components of the random model with level and a gradually decelerating trajectory (see Figure 4).
only linear and quadratic age changes as fixed predictors with Women were more affected by their socioeconomic backgrounds
those of the random model including the demographic variables in than were men: Women from high-SES families had a higher mean
order to obtain rough equivalents of R2 (J. Cohen, Cohen, West, & financial TL than did women from low-SES families by 6.58 PP
Aiken, 2002, chap. 15). When we did so, there was a 16.5% (⫽ 2 ⫻ 3.29). No statistically significant differences by race or
reduction in the variance of mean residential TL and an 11% parental divorce were found. When we added school attendance to
reduction in the variance of linear age changes. The random effects the model, we found that it predicted a 1.53-PP higher financial TL
in quadratic age changes were reduced by 12%, and, necessarily, but did not mediate the effect of family SES, for which the effects
other random effects were essentially unchanged. on the overall trajectory of financial independence for both men
Financial transition level. For the group, on average, the fi- and women remained unaltered.
nancial TL was 49.14 at age 22, moving from 21.4 at age 17 Again, when we investigated the extent to which the fixed
to 64.5 at age 27, with a linear increase of 3.89 PP per year predictors accounted for random effects, we found a 7% reduction
combined with a 0.35-PP deceleration in this increase. The mean in the variance of individual mean financial TL and a 12% reduc-
TL was higher for men than for women by 4.17 PP. Although there tion in the variance in linear age changes, whereas the quadratic
 PATTERNS IN EMERGING ADULTHOOD 665
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 3. Individual residence transition level (TL) trajectories of 10% of men (left panel) and trajectory curves
by sex (right panel).

age variance was reduced by 9%. Thus, on the whole, these fixed found between men from low-SES families and men from higher
effects accounted for a modest fraction of the original differences SES families (see Figure 5).
between participants in these parameters. At age 22, White women had higher romantic TL than did White
Romantic transition level. Romantic TL increased 4.09 PP per men, whereas Black women showed substantially lower mean
year, from an average of about 27 at age 17 to about 68 at age 27. values than did Black men (31 PP). For men and women com-
However, women were at a higher TL at age 17 and continued to bined, the linear increase in romantic TL per year was much lower
be until age 27. Because of deceleration in the women’s increase for Black participants, but there were also substantial differences
in romantic TL over time and acceleration in the men’s increase, in these patterns by sex (see Figure 6). Black men showed their
differences were greatest at about age 22. The romantic TL of maximum level of romantic TL in their early 20s, on average, and
women from low-SES families was higher than that of women moved toward less romantic commitment in the latter half of this
from higher SES families, signifying earlier romantic commitment developmental phase. Black women were slow to move toward
in these women. No statistically significant mean differences were romantic commitment but showed a tendency to catch up with the

Figure 4. Individual financial transition level (TL) trajectories of low socioeconomic status (SES) women (left
panel) and model-produced trajectories by sex and family SES (right panel).
 666 COHEN, KASEN, CHEN, HARTMARK, AND GORDON
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 5. Individual romantic transition level trajectories for low socioeconomic status (SES) men (left panel)
and model-produced trajectories by sex and family SES (right panel).

White women’s level by the end of the period. These findings must quadratic age changes were accounted for by the demographic
be generalized very cautiously given the small sample of Black fixed predictors.
participants. Parenting transition level. Interest and involvement in parent-
We had anticipated a possible effect of parental divorce on ing as reflected in the TL was generally very low at age 17,
young people’s readiness to make romantic commitments, but we although 4 participants were active parents. The mean score was
found none. We also found that school attendance had no impact about 11 at age 22 and about 30 at age 27; thus the increase greatly
on romantic TL, and no other estimated effects were altered by its accelerated in the later period. There were significant differences
inclusion in the model. in mean parenting TL by sex (men were lower), race (Blacks were
When we evaluated the extent to which the fixed predictors higher), and socioeconomic background (those from higher SES
accounted for components of the random model, we found that families were lower).
24% of the random variance in mean romantic TL, 12% of random In addition to a sex difference of nearly 9 PP in parenting TL at
model linear age changes, and 18.5% of the random variance in age 22, the linear increase was about 60% greater for women (see

Figure 6. Individual parenting transition level trajectories for low socioeconomic status (SES) women (left
panel) and model-produced trajectories by sex and family SES (right panel).
 PATTERNS IN EMERGING ADULTHOOD 667

Figure 6). Young people from high- and low-SES families were Transition Level Ratings
not significantly different at age 22, but the linear increase for the
high-SES (⫹1 SD) offspring was about 60% lower than that for the The transition level variables added richness to the picture
low-SES (⫺1 SD) offspring. The SES effects were not signifi- presented by these markers. They demonstrated that although the
cantly different for men and women (see Figure 6). Black partic- young people on average changed roles and assumed responsibil-
ipants had a much higher mean parenting TL at age 22 (28.35) than ities in accordance with developmental expectations, their patterns
did White participants (11.19) but showed much lower linear (2.42 of doing so were remarkably diverse. Many, perhaps most, young
vs. 2.85) and quadratic (⫺.49 vs. ⫹.18) age changes. This reflects people had periods during which they assumed great indepen-
the fact that although some Black participants were parents before dence, followed by periods in which a more childlike level of
the age of 20, those who were not parents by then were not likely function was characteristic, at least in one or two domains. Fur-
to have children before the age of 27. We did not find a significant thermore, development was not a unitary phenomenon. Although
effect of parental divorce on the readiness to assume a parental TL correlated across the different role domains examined here,
role, nor did current school attendance correlate with movement many young people moved at very different paces along their
toward or assumption of a parental role. various developmental paths.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

When assessed by transition level ratings, individual develop-

This document is copyrighted by the American Psychological Association or one of its allied publishers.

Again, when we evaluated the model including the demographic

predictors for the proportionate reduction in random variance, mental patterns were influenced by gender, SES, and race but not
there was a 22% reduction in mean level variance, a reduction of by parental divorce.3 The absence of parental divorce effects,
24% in the linear age variance, and a 10% reduction in the variance particularly on the timing of the assumption of adult romantic
of individual quadratic age effects. roles, is contrary to frequent speculation. Although the presence or
absence of divorce had no significant effects, other related vari-
Discussion ables such as parental conflict may be more predictive of devel-
opmental trajectories (e.g., Fincham & Osborne, 1993). Alterna-
Representativeness of the Sample tively, parental divorce may exert more influence on
The larger sample used in this study was originally selected nondevelopmental qualitative aspects of romantic relationships
from a geographic area that was broadly representative of the U.S. and child rearing, such as role satisfaction and mastery, a possi-
population with regard to family income distribution, family struc- bility we are examining in related work.
ture, race, and urbanicity. At the time of the present assessments, Female/male differences. Large sex differences were seen for
conducted 20 years after the original sampling, members of this each of the transition trajectories. Women assumed adult roles with
cohort resided in nearly every state in the country. However, this regard to residence earlier than did men and continued with per-
sample cannot reflect the more recent immigration patterns of sistently higher levels in this domain. However, despite their
people of similar age with Hispanic and Asian backgrounds. Nev- earlier separation from parents as the primary source of income,
ertheless, our findings with regard to the markers of adult status women did not assume full financial self-support to the same
were generally consistent with those in the literature, showing a extent as did men, partly because of support from partners. Thus,
continuation of secular trends found in the general U.S. population. this financial TL pattern may be partly explained by sex differ-
In this 1970 birth cohort, 84% and 42% of participants lived with ences in romantic and parenting TL patterns. At age 22, women
parents at ages 18 and 22, respectively. At age 25, about one fifth had a 20-PP lead with regard to assuming more adult romantic
lived with a parent, whereas about three quarters had an indepen- roles and commitment. However, this age showed the maximum
dent residence, including about half who lived with a romantic sex discrepancy, and the male romantic TL pattern in the following
partner. By age 25, only a little more than half were fully finan- years showed a tendency to approach the average romantic TL in
cially independent of family. Thus, in comparison with previously women. Clearly, assumption of adult roles in some domains may
published data, our results suggest a further delay in the age when hinder parallel role development in other domains.
young people become financially independent, especially for men. SES effects. Surprisingly, we found no significant differences
In contrast to the 1962 cohort of White women studied by Chen in the transition to residential independence associated with the
and Morgan (1991), in which 44% had borne a child by age 25, SES of the family of origin. Apparently, the tendency of lower
only 34% of the women in this 1970 cohort had borne a child by SES offspring to live longer in parental homes was compensated
that age. for by their assuming more responsible roles, sometimes including
Differences associated with parental SES were marked, includ- financial contributions, in that setting, in comparison with the
ing less full-time employment, nearly doubled rates of school dormitories and shared quarters more frequently occupied by those
attendance, and less living with family and family financial sup- from higher SES families.
port for those from higher SES backgrounds. High rates of being In contrast, family SES was an important predictor of financial
sexually active, being married, having had or produced a preg- TL, and these disparities tended to widen with offspring age. It is
nancy, and having living biological offspring were much more
common among respondents from lower SES families. Race dif- 3
We also investigated two other potential influences, namely, birth
ferences were also apparent despite the small sample of Black
order of the participant and maternal birth year. The first of these variables
participants. White participants were more likely to be working reflects a historically significant influence on the life course, and the
full time and to be married and less likely to have offspring. Thus, second reflects a possible difference in parental expectations and behavior
these status marker variables tend to show that at the turn of the that may influence the transition process. We found neither of these
millennium, the social role differences associated with sex, race, variables to produce more significant findings than were consistent with
and SES persist. chance.
 668 COHEN, KASEN, CHEN, HARTMARK, AND GORDON

often implicitly assumed that these differences reflect access to patterns for the assumption of these adult roles over this critical
advanced education. However, including educational attendance in decade of life. Despite the gradual increase in the extent to which
the model as a time-varying covariate did not alter the influence of the average participant had assumed adult roles in these four
family SES on offspring financial independence. The financial role domains, the actual progression of individuals was much more
difference may reflect better access to nonfamily sources of sup- variable, moving back and forth between increasing and decreas-
port for higher SES offspring, especially better wages, or perhaps ing dependency. It is likely, although not demonstrable, that these
differences in family expectations for financial independence. Fur- patterns are increasingly complex compared with historical trajec-
thermore, the impact of parental SES on the assumption of adult tories, which may have had relatively fewer normative patterns and
financial responsibility was greater for women than for men. This more clearly delineated social and family expectations, both more
difference may have been particularly influenced by the earlier heavily conditioned by demographic context and sex-specific
marriages and childbearing of the women from lower SES norms. In our society, where individualism is highly prized, con-
backgrounds. straints on youth are much relaxed, and relative affluence has
Family SES effects were also apparent with regard to romantic reduced the risk of deviating from normative patterns. What the
involvement and parenting. For both of these transitions, partici- long-term outcomes of such substantial fluctuation in the course of
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

pants from higher SES backgrounds were much slower to assume development toward adult roles will be for society and for indi-
adult roles. Although SES-related delays in the full assumption of viduals remains to be seen.
adult roles in domains of romance and parenting might have been These data support the view of this age period as representing a
attributable to a greater focus on career development, educational true developmental stage. In addition, they show that status mark-
attendance was not related to either of these TLs. ers cannot capture the richness and diversity reflected in these
Race differences. Despite the small sample of Black partici- individual differences in developmental trajectories over the
pants, there were a number of significant differences in TL means emerging adulthood period. We had anticipated that participation
and trajectories by race that were independent of any family SES in higher education would mediate demographic differences in
effects. Black participants assumed more adult residential roles in trajectory mean or slope, but we found that they did not. Account-
the immediate post-high-school years but did not subsequently ing for these complex individual trajectories will require other
increase their independence and responsibility in this area to the time-varying predictors of the patterns of change within this tran-
same extent as did White participants. It is possible that full sition period.
residential independence of young Black adults is still hampered
by residential racial bias. We did not find significant race differ- References
ences in the assumption of adult financial roles that were indepen-
dent of family SES effects. However, Black participants, particu- Arnett, J. J. (1997). Young people’s conceptions of the transition to
larly men, were very late in making or unlikely to make full adulthood. Youth & Society, 29, 1–23.
romantic commitments and long-term plans. Child rearing began Arnett, J. J. (1998). Learning to stand alone: The contemporary American
early for some members of this subgroup, but there was little transition to adulthood in cultural and historical context. Human Devel-
opment, 41, 295–315.
movement toward family formation in the remainder of this small
Arnett, J. J. (2000). Emerging adulthood: A theory of development from
sample. the late teens through the twenties. American Psychologist, 55, 469 –
480.
The Influence of Educational Involvement on the Arnett, J. J., & Taber, S. (1994). Adolescence terminable and interminable:
Transition to Adult Roles When does adolescence end? Journal of Youth and Adolescence, 23,
517–537.
School attendance explained differences in early residential Aro, H. M., & Palosaari, L. K. (1992). Parental divorce, adolescence, and
separation from families and in average financial adult role attain- transition to young adulthood: A follow-up study. American Journal of
ment, in both cases being associated with higher TL. However, it Orthopsychiatry, 62, 421– 429.
Bloom, D. E., & Trussell, J. (1984). What are the determinants of delayed
did not mediate the effects of other demographic variables, includ-
childbearing and permanent childlessness in the United States? Demog-
ing family SES.
raphy, 21, 591– 611.
The demographic predictors included in these models were Chen, R., & Morgan, S. P. (1991). Recent trends in the timing of first births
necessarily constant over this decade and thus incapable of ac- in the United States. Demography, 28, 513–533.
counting for individual differences in trajectories not captured by Cohen, J., Cohen, P., West, S. G., & Aiken, L. (2002). Applied multiple
mean differences or systematic linear or quadratic age changes. regression/correlation analysis for the behavioral sciences (3rd ed.).
The demographic predictors accounted for nontrivial proportions Mahwah, NJ: Erlbaum.
of the individual “random” variance in mean TLs, ranging from Cohen, P., Chen, H., Hamigami, F., Gordon, K., & McArdle, J. J. (2000).
24% of romantic TL to 7% of financial TL, as well as for 26% (for Multilevel analyses for predicting sequence effects of financial and
parenting TL) to 7% (for romantic TL) of the linear age change employment problems on the probability of arrest. Journal of Quanti-
tative Criminology, 16, 223–235.
variance and for about 10% of the quadratic age change variance
Cohen, P., & Cohen, J. (1996). Life values and adolescent mental health.
for each TL.
Mahwah, NJ: Erlbaum.
Cohen, P., Cohen, J., Aiken, L. S., & West, S. G. (1999). The problem of
Individual Slopes units and the circumstance for POMP. Multivariate Behavioral Re-
search, 34, 315–346.
Despite the demographic influences, the most dramatic findings Conger, R. D., Cui, M., Bryant, C. M., & Elder, G. H., Jr. (2000).
of this study were the great variability in individual trajectory Competence in early adult romantic relationships: A developmental
 PATTERNS IN EMERGING ADULTHOOD 669

perspective on family influences. Journal of Personality and Social models for developmental data analysis. In T. D. Little, K. U. Schnabel,
Psychology, 78, 224 –237. & J. Baumert (Eds.), Modeling longitudinal and multilevel data: Prac-
Cooney, T. M. (1989). Young adults and parental divorce: Exploring tical issues, applied approaches, and specific examples (pp. 69 –107).
important issues. Human Relations, 41, 805– 822. Mahwah, NJ: Erlbaum.
Elder, G. H., Jr. (1998). Children of the Great Depression: Social change McLeod, J. D. (1992). Childhood parental loss and adult depression.
in life experience (25th anniversary ed.). Boulder, CO: Westview Press. Journal of Health and Social Behavior, 32, 205–220.
Elder, G. H., Jr., & Conger, R. D. (2000). Children of the land: Adversity Mogelonsky, M. (1996). The rocky road to adulthood. American Demo-
and success in rural America. Chicago: University of Chicago Press. graphics, 18, 26 –36.
Fincham, F. D., & Osborne, L. N. (1993). Marital conflict and children: Rindfuss, R. R. (1991). The young adult years: Diversity, structural
Retrospect and prospect. Clinical Psychology Review, 13, 75– 88. change, and fertility. Demography, 28, 493–512.
George, L. K. (1993). Sociological perspectives on life transitions. Annual Rindfuss, R. R., Morgan, S. P., & Swicegood, C. G. (1988). First births in
Review of Sociology, 19, 353–373. America: Changes in the timing of parenthood. Berkeley: University of
Goldscheider, F., & Goldscheider, C. (1994). Leaving and returning home California Press.
in 20th century America. Population Bulletin, 48, 1–35. Rindfuss, R. R., & Parnell A. M. (1989). The varying connection between
Goldscheider, F., Thornton, A., & Young-DeMarco, L. (1993). A portrait marital status and childbearing in the United States. Population and
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

of the nest-leaving process in early adulthood. Demography, 30, 683– Development Review, 15, 447– 470.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

699. Rindfuss, R. R., Swicegood, C. G., & Rosenfeld, R. A. (1987). Disorder in

Greene, A. L., Wheatley, S. M., & Aldava, J. F. (1992). Stages on life’s the life course: How common and does it matter? American Sociology
way: Adolescents’ implicit theories of the life course. Journal of Ado- Review, 52, 785– 801.
lescent Research, 7, 364 –381. Shanahan, M. J. (2000). Pathways to adulthood in changing societies:
Hogan, D. P., & Astone, N. M. (1986). The transition to adulthood. Annual Variability and mechanisms in life course perspective. Annual Review of
Review of Sociology, 12, 109 –130. Sociology, 26, 667– 692.
Kogan, L., Smith, J., & Jenkins, S. (1977). Ecological validity of indicator Singer, J. D. (1998). Using SAS PROC MIXED to fit multilevel models,
data as predictors of survey findings. Journal of Social Service Re- hierarchical models, and individual growth models. Journal of Educa-
search, 1, 117–132. tional and Behavioral Statistics, 24, 323–355.
Kohn, M. L., & Schooler, C. (1983). Work and personality: An inquiry into Wadsworth, M., Maclean, M., Kuh, D., & Rodgers, B. (1990). Children of
the impact of social stratification. Norwood, NJ: Ablex. divorced and separated parents: Summary and review of findings from a
Littell, R. C., Miliken, G. A., Stoup, W. W., & Wolfinger, R. D. (1996). long-term follow-up study in the UK. Family Practice, 7, 104 –109.
SAS system for mixed models. Cary, NC: SAS Institute. Whittington, L. A., & Peters, H. E. (1996). Economic incentives for
Lyketsos, C. G., Nestadt, G., Cwi, J., Helthoff, K., & Eaton, W. W. (1994). financial and residential independence. Demography, 33, 82–97.
The Life Chart Interview: A standardized method to describe the course
of psychopathology. International Journal of Methods in Psychiatric Received December 10, 2001
Research 4, 143–155. Revision received December 10, 2002
McArdle, J. J., & Bell, R. Q. (2000). An introduction to latent growth Accepted December 31, 2002 䡲