Вы находитесь на странице: 1из 17

Child Development, January/February 2003, Volume 74, Number 1, Pages 127143

Reciprocal Influences Between Stressful Life Events and Adolescent


Internalizing and Externalizing Problems
Kee Jeong Kim, Rand D. Conger, Glen H. Elder Jr., and Frederick O. Lorenz

Investigated in this study were hypothesized reciprocal influences between stressful life events and adolescent
maladjustment using data from a 6-year, prospective longitudinal study. Stressful life experiences, internalizing
symptoms, and externalizing behaviors were assessed for a sample of adolescents (215 males, 236 females)
living in the rural Midwest. From 7th to 12th grades, autoregressive analyses showed that stressful life events
and these two forms of maladjustment were reciprocally interrelated over time. For example, stressful life
events at one point significantly predicted delinquent behaviors 1 year later, which, in turn, significantly
predicted stressful life events 1 year later, and vice versa. The findings provide evidence for the accumulating
disadvantage for adolescents that results from the mutual reinforcement of problematic situations and
adjustment problems over time.

Research evidence suggests that cumulative life Aseltine, Gore, & Gordon, 2000). And although the
stresses increase risk for emotional and behavioral association between life stress and adolescent de-
problems (Aneshensel, 1992; Cohen, Kessler, & pressed and anxious mood is fairly well established
Gordon, 1995; Compas & Phares, 1991; Jackson & (e.g., Compas & Phares, 1991; Menaghan, 1999),
Warren, 2000; Kessler, Gillis-Light, Magee, Kendler, there is less evidence that stressful life events
& Eaves, 1997). Consistent with the social-causation actually predict change in internalizing problems
hypothesis, these studies suggest that social condi- over time. To further understanding of the link
tions (i.e., stressful life events) influence variability between life stress and adolescent maladjustment, a
in subsequent developmental problems (Caspi, 1998; first goal of the present study was to examine the
Conger, 1995, 1997; Dohrenwend & Dohrenwend, degree to which stressful life events predict change
1969). With regard to conduct problems or antisocial in both externalizing and internalizing problems
behavior, some studies have demonstrated a link during early, middle, and late adolescence.
between specific, serious traumatic events such as A singular focus on social causation, however,
child abuse or environmental catastrophes and ignores the important competing hypothesis that
subsequent behavioral problems (e.g., Stouthamer- emotional or behavioral problems lead to increasing
Loeber, Loeber, Homish, & Wei, 2001; Thornberry, life stress (Caspi, Elder, & Bem, 1987; Compas &
Ireland, & Smith, 2001); however, research using Phares, 1991; Ensel & Lin, 1991). That is, contem-
general additive indices of social stress to predict porary theory suggests that emotional and behavior-
externalizing problems over time has been rare (e.g., al problems may reduce social and instrumental
competencies and thus increase risk for crises in
social relationships and instrumental activities. This
Kee Jeong Kim and Rand D. Conger, Department of Human and
Community Development, University of CaliforniaDavis; Glen proposed direction of effects has been called the
H. Elder Jr., Department of Sociology, University of North social-selection hypothesis (Caspi, 1998; Conger, 1995,
CarolinaChapel Hill; Frederick O. Lorenz, Department of 1997; Ensel & Lin, 1991). For example, adolescents
Sociology, Iowa State University. who are extremely sad and withdrawn (internaliz-
During the past several years, support for this research has
ing symptoms) may be unsuccessful in social rela-
come from multiple sources, including the National Institute of
Mental Health (MH00567, MH19734, MH43270, MH48165, and tions because they are unrewarding companions.
MH51361), the National Institute on Drug Abuse (DA05347), the Moreover, aggressive adolescents (externalizing
Bureau of Maternal and Child Health (MCJ-109572), the Ma- problems) may be too mercurial to be reliable
cArthur Foundation Research Network on Successful Adolescent friends. These types of symptoms would also jeo-
Development Among Youth in High-Risk Settings, and the Iowa pardize school achievement, which requires focus,
Agriculture and Home Economics Experiment Station (Project No.
3320). energy, and the regulation of competing emotions.
Correspondence concerning this article should be addressed to When evidence for both social causation and social
Kee Jeong Kim, Department of Human and Community Devel-
opment, University of CaliforniaDavis, 202 Cousteau Place, Suite
100, Davis, CA 95616. Electronic mail may be sent to keekim@uc- r 2003 by the Society for Research in Child Development, Inc.
davis.edu. All rights reserved. 0009-3920/2003/7401-0009
128 Kim, Conger, Elder, and Lorenz

selection is evaluated in a single study, as in the Leadbeater, Kuperminc, Blatt, & Hertzog, 1999;
present report, one can determine whether none, Swearingen & Cohen, 1985a). These findings are
one, or both of these hypotheses are consistent with consistent with the fact that adolescence is a very
the data. challenging developmental stage involving signifi-
Unfortunately, very few studies provide empirical cant biological, psychological, and social changes
evidence that can evaluate the possible mutual (Compas & Wagner, 1991; Hauser & Bowlds, 1990).
influence between stressful life events and either The evidence suggests that both nonnormative
internalizing or externalizing problems, a process we stressors (e.g., parents divorce) and normative
predict would increase risk for both stressful stresses (e.g., the transition from middle school to
experiences and maladjustment across the years of high school) increase risk for internalizing symp-
adolescence. The importance of this issue is under- toms (Hauser & Bowlds, 1990; Swearingen & Cohen,
scored by Caspis (1998) suggestion that clarifying 1985a). Most earlier research has shown that a high
the direction of influence between social conditions number of uncontrollable, negative life events
and individual characteristics is a central concern for predicts depression and anxiety, especially among
research on life-course development. As a second adolescent girls (Dornbusch, Mont-Reynaud, Ritter,
goal, then, this study examined possible reciprocal Chen, & Steinberg, 1991; Gore, Aseltine, & Colten,
influences between stressful life events and adoles- 1992; Leadbeater et al., 1995, 1999; Swearingen &
cent internalizing and externalizing problems (the Cohen, 1985a). For example, a cross-sectional study
reciprocity or mutual influence hypothesis). by Swearingen and Cohen (1985a) showed that
Another concern in the present study involves the seventh graders who reported a higher number of
methodological limitations of previous research. negative life events displayed a higher level of
Earlier research related to these phenomena has depressed and anxious mood.
either been cross-sectional or has involved only two Although some researchers, especially those
waves of data that include measures of both working with clinical samples, have argued that
stress and adolescent adjustment problems (e.g., parental psychopathology primarily accounts for
Hoffmann & Su, 1997; Jackson & Warren, 2000; child and adolescent emotional problems (e.g.,
Paternoster & Mazerolle, 1994; Vaux & Ruggiero, Strober & Carlson, 1982), the research evidence
1983). As Kessler and Greenberg (1981) have suggests that other life stresses in the lives of young
demonstrated, two waves of data may provide people also play an important role in the develop-
misleading information about how two variables ment of psychopathology. For instance, Gore et al.
interrelate with one another over time. The third (1992) found that parental mental illness did not
goal of this study was to overcome this limitation in relate to adolescents depressed mood when un-
previous research by prospectively investigating the desirable life events, which were significantly
hypothesized reciprocal influence process across the associated with distress, were taken into account.
years of adolescence. In the following sections, we The study by Gore et al. also extended previous
review the evidence for an association between research findings on gender differences in depres-
stressful life events and adolescent adjustment sive symptoms. Although the girls in their study
problems. This review also considers methodologi- reported a higher number of negative events than
cal limitations in this earlier research. We then did boys, there was no difference in the emotional
consider the theoretical underpinnings for the impact of stressful life events by adolescent gender
present study and provide our analytical model for (Gore et al., 1992). Leadbeater et al. (1995) also
the investigation. reported that boys were as vulnerable as girls in
response to negative life events.
Externalizing behaviors. Although life stress may
Stressful Life Events and Adolescent Internalizing reasonably lead to an angry, acting-out response as
and Externalizing Problems well as to an introspective and internalized response,
empirical examination of the possible linkage be-
From Life Stress to Adolescent Emotional and Behavioral
tween adversity in life circumstances, as measured
Maladjustment
by additive scales of negative life events, and
Internalizing problems. Earlier research indicates conduct problems or delinquent acts is limited.
that adolescent risk for emotional problems in- Rather, several studies have shown a connection
creases in response to the experience of stressful life between single significant traumas (e.g., child abuse)
events (Aseltine, Gore, & Colten, 1994; Compas & and conduct problems (e.g., Stouthamer-Loeber
Phares, 1991; Leadbeater, Blatt, & Quinlan, 1995; et al., 2001; Thornberry et al., 2001). Because our
Stressful Life Events and Adolescent Maladjustment 129

interest focused on the examination of indices of nalizing and externalizing problems has reflected a
multiple life stressors, such as those typically used in unidirectional viewpoint, the social-causation hy-
studies of emotional distress, the following literature pothesis. Despite the potential importance of the
review is limited to studies that have examined social-selection perspective, there exists almost no
externalizing problems in relation to additive indices empirical evidence regarding the theoretically inter-
of negative life events. esting possibility that antisocial or depressed youth
Two cross-sectional studies have provided pre- may generate above average levels of stressful life
liminary evidence for the relationship between experiences. We could locate only three studies
stressful life events and externalizing behaviors. In providing preliminary evidence for the social-selec-
a study of preadolescents from 7 to 13 years old, tion hypothesis. In terms of internalizing problems,
Jackson and Warren (2000) found a positive relation- findings from a study by Swearingen and Cohen
ship between life stresses and externalizing beha- (1985a) support the notion that emotional problems
viors measured by aggression, hyperactivity, and exacerbate stressful life conditions. Based on data
conduct problems. A study by Vaux and Ruggiero collected at two points in time, adolescent depres-
(1983) also demonstrated a link between stressful life sive symptoms and anxiety at Time 1 predicted
events and delinquency. One of the strengths of their stressful life events 4 months later, after controlling
study was explicitly defining delinquency by break- for the initial level of life events. Stressful life events
ing it down into five dimensions: violence, property at Time 1 did not predict emotional distress at
damage, drug use, theft, and nonserious delin- Time 2. These findings support the social-selection
quency. Their regression analyses showed that hypothesis and are in need of replication.
stressful life events significantly predicted delin- With respect to externalizing behaviors, Leadbea-
quent behaviors regardless of the type of delinquent ter et al. (1999) found that only girls delinquent
act. They did not find any significant gender behaviors were significantly associated with their
differences except for violent behaviors. Only boys level of stressful life events 1 year later. Additionally,
violence was positively related to stressful life Aseltine and colleagues (2000) found that delinquent
events. Except for violence, however, the relation- behaviors among 9th, 10th, and 11th graders
ship between stressful life events and delinquency predicted higher levels of life stress and conflict
was stronger for girls than boys (Vaux & Ruggiero, with family members 1 year later. A significant
1983). Although these are promising findings, the limitation in both of these studies, however, was that
cross-sectional research design of these two studies they did not control for earlier levels of stress in the
seriously limits inferences about the possible causal equations predicting later stress from earlier delin-
relationship between delinquency and stressful life quency. That is, from their findings we cannot infer
events. that delinquency is associated with change or
A two-wave longitudinal study conducted by increased risk for stressful events over time. Never-
Hoffmann and Su (1997) also investigated stressful theless, these studies provide preliminary evidence
life events as a risk factor for delinquency. They for the social-selection hypothesis.
demonstrated that stressful life events significantly
and positively predicted delinquent acts 1 year later
The Present Investigation
for both males and females after controlling for the
previous level of delinquency. These results are The present study derived from contemporary
consistent with other evidence of a positive associa- theory that suggests that developmental trajectories
tion between stressful life events and delinquency involve an interconnected accumulation of indivi-
over time (Leadbeater et al., 1999; Paternoster & dual actions and social reactions in which an
Mazerolle, 1994). These longitudinal studies suggest exogenous variable at one time becomes an endo-
that stressful life events influence the delinquent genous variable at another time and vice versa
behaviors of adolescents after controlling for the (Conger, 1997; Conger & Simons, 1997; Elder, 1998;
initial level of delinquent behavior. That is, stressful Simons, Johnson, Conger, & Elder, 1998). That is,
events appear to predict change or increasing risk social stress at one point in time is predicted to
for conduct problems. increase later maladjustment and maladjustment is
expected to increase stress in a reciprocal process.
Thus, controlling for earlier levels of these variables,
From Problem Behaviors to Stressful Life Events
they each should explain significant variance in later
As noted earlier, most previous research on the levels. What remains to be established, however, is
link between stressful life events and both inter- whether this proposed process of increasing
130 Kim, Conger, Elder, and Lorenz

personal and social disadvantage through the mother, a target adolescent in the seventh grade
mutual reinforcement of emotional or behavioral (215 boys and 236 girls, M age 5 12.7 years), and a
problems and adversity can actually be observed in sibling within 4 years of the target adolescents age
the lives of adolescents. For example, research participated in the project. The IYFP continued for 5
evidence shows that engaging in various delinquent years and was renamed the Family Transitions
acts may increase several domains of life stress for Project in 1994. The Family Transitions Project is an
adolescents, such as conflicts with parents and ongoing longitudinal study focusing on the original
disrupted academic performance (Elliott, Huizinga, target adolescents transition from adolescence to
& Ageton, 1985; Herrenkohl et al., 2000; Sampson & adulthood. Because the original project focused on
Laub, 1993). However, previous research has not family financial hardship and family processes
clearly shown either that problematic behavior among families living in rural Iowa where almost
increases life stress in general or that these adver- no minority families live, the study sample included
sities generate additional risk for behavioral and only White families. For the purposes of the present
emotional problems. We evaluated this hypothe- study, we used data from the target adolescents self-
sized reciprocal process in this study and deter- report of delinquency (externalizing problems),
mined whether it operates in a similar fashion symptoms of depression and anxiety (internalizing
during early, middle, and late adolescence. problems), and stressful life events at five points in
By using a five-wave autoregressive model cover- time (i.e., 1989, 1990, 1991, 1992, and 1994), the
ing the full span of adolescence, the current period from early (7th grade) to late (12th grade)
prospective longitudinal study sought both to adolescence.
establish whether the proposed reciprocal relation- A common difficulty with multiyear studies,
ship between stressful life events and adolescent however, involves cases with missing data. Incom-
maladjustment operates for this sample and to plete information from a participant may occur if the
determine whether it exists during each of the three subject leaves the study, is unavailable for one or
stages of adolescence (early, middle, and late). If the more waves of data collection, or fails to answer
latter finding is obtained, subsequent evidence of some questions during a specific interview. In the
mutual influence during middle and late adoles- present study we used recently developed strategies
cence can be thought of as replications of positive for handling this difficult problem.
results during early adolescence. In this situation, Two common methods for dealing with missing
the resulting path coefficients between stress and cases involve either deleting the incomplete cases
adolescent maladjustment can be interpreted as from the entire data set (i.e., listwise deletion) or
reflecting the degree of change in the dependent imputing values based on average scores for missing
measure resulting from its association with the variables. Both of these procedures have serious
predictor variable (Kessler & Greenberg, 1981). limitations. First, deleting incomplete cases forces
Consistent with findings discussed earlier, we also the researcher to discard a large amount of already
hypothesized that there would be no gender observed information (Schafer, 1997). Second, using
differences in the association between stressful life only completely observed cases tends to introduce a
events and either delinquency or emotional distress. bias in the estimation of statistical parameters if
Finally, this repeated measure research design also missing cases do not occur at random (Schafer, 1997).
overcame the methodological limitations noted ear- If there is a systematic pattern of missing cases, it
lier for the cross-sectional and two-wave studies indicates that the incompletely observed cases
typically found in investigations of adolescent behave differently from the completely observed
maladjustment and negative life events. ones. Third, imputing scores from variable means
may preserve the overall sample means but it does
not consider covariance structures with other vari-
Method ables in the data (Schafer, 1997). That is, mean
imputation homogenizes the distribution of scores
Participants
for a specific variable and ignores the fact that these
Data for the present study came from the Iowa scores may vary systematically in relation to other
Youth and Families Project (IYFP), which was characteristics of the respondents.
launched in 1989 to investigate family economic To deal with the problem of missing data and to
hardship and life-course development (Conger & avoid the limitations of these earlier approaches, we
Elder, 1994). In the first year of data collection, 451 imputed missing data in this research using the
intact Caucasian families including a father, a expectation maximization (EM) algorithm, which
Stressful Life Events and Adolescent Maladjustment 131

generates maximum likelihood estimates for para- found. The rates of stressful life events measured in
metric models when data are missing (for a complete 1989 were higher for the missing or excluded
review, see Schafer, 1997). Earlier research has adolescents. No statistically significant differences
shown that the EM procedure produces more were found between the two groups in terms of
accurate and less biased estimates of statistical depression, anxiety, and delinquency. These results
coefficients than either listwise deletion or mean indicate that the final sample was slightly truncated
imputation even when cases are not missing in terms of the highest levels of negative life events.
completely at random and even when a large This truncation should produce a conservative bias
portion of study participants have some missing in the analyses in the sense that limited variation in
data (Schafer, 1997). life stress will decrease the probability that it will
Complete data by listwise deletion were available predict or be predicted by adolescent maladjust-
for up to 348 adolescents (77% of the original ment.
sample) over the 6-year period. Using the SPSS
10.0 (2000) missing value analysis, we imputed data Procedures
for missing cases by using the EM algorithm
In the first year of the project, 34 public and
(Dempster, Laird, & Rubin, 1977; Schafer, 1997),
private school districts in eight counties in rural
which recovered all the missing cases (N 5 451).
Iowa agreed to identify potential participants. In the
Schafer (1997) has demonstrated that EM generates
eight counties, names and addresses of families with
less biased statistical findings than listwise deletion,
7th graders were obtained from all schools in
even when a large proportion of cases has missing
communities with populations of 6,500 or less.
data. In the results section, we compare the findings
Families received a letter including an introductory
using either the EM or listwise deletion procedure.
summary of the project and then were recruited by
In terms of sociodemographic information on
telephone or in person if they had no phone.
participating families, at study initiation 34% of the
Approximately 78% of the eligible families agreed
families lived on farms, 12% lived in a rural area but
to participate in the project. Each participating
not on farms, and 54% lived in a town with a
family member received about $10 per hour for his
population no greater than 6,500. The average ages
or her time in the study.
of fathers and mothers at the first year of contact
For each years data collection, each family was
were 40 and 38 years, respectively. In 1989, the
visited twice in their own home by a trained
parents had been married for an average of 18 years.
interviewer. Each visit lasted approximately 2 hr
Fathers averaged 13.4 years of education and
on average. During the first home visit, the four
mothers averaged 13.2 years of education. Occupa-
family members separately completed a set of
tions for fathers were distributed as follows: 20%
questionnaires, some of which addressed stressful
farmers, 30% professionals or managers, 40% skilled
life events and internalizing as well as externalizing
workers such as technicians, and the rest were either
problems. Within 2 weeks of the first visit, the
office workers or unskilled workers. Mothers
second visit was made to videotape structured
occupations were distributed as follows: 30% clerks
discussions among the four family members. Be-
and secretaries; 27% service and household workers;
cause data from the second visit were not used in the
26% professionals, including teachers; 10% full-time
present report, we do not discuss the procedures
homemakers; and the rest were technicians or skilled
related to that visit.
workers. Family median income in 1988 was $33,399,
and the average family size was 4.95 in 1989.
Measures
We examined mean differences in several socio-
demographic factors and the study variables for the Stressful life events. During each year of data
adolescents who remained in the study and for those collection, adolescents completed a stressful life
who dropped out by 1994. First, among a set of events measure. The index was adapted from the
group comparison analyses on family background Junior High Life Experiences Survey developed by
variables background variables (i.e., parents age, Swearingen and Cohen (1985b). The original survey
education, and family income at the initiation of the included items regarding deviant behaviors (e.g.,
study), differences arose only in parents education running away from home) and positive life events
levels. Both fathers and mothers education levels (e.g., outstanding academic achievement). Items in
were higher for the target adolescents who remained these two domains were excluded because the
in the study. Second, the study variables were present investigation focused on negative rather
compared and one other statistical difference was than positive life events and because the deviant
132 Kim, Conger, Elder, and Lorenz

behaviors in the original stress scale overlap with the $25, attacking someone with a weapon, and purpo-
externalizing problems expected to predict and be sely damaging or destroying property. Minor activ-
predicted by negative life events. ities included items such as stealing something
Each item in the index was scored 1 if the target worth less than $25, throwing objects such as rocks
adolescent indicated a specific event had occurred to scare someone, and trying to break into a building
during the past year and 0 if the event had not just for fun or to look around. The major and minor
occurred. Examples of items appearing in the index dimensions of delinquency were used as two
were financial crises, death of a parent, parents separate indicators for the construct.
divorce, physical illness, breaking up with a boy- Internalizing problems. Adolescent emotional dis-
friend or girlfriend, changing schools, and getting tress was measured in 1989, 1990, 1991, 1992, and
into trouble with classmates at school. Because 1994 using the depression and anxiety subscales
adolescent maladjustment was predicted to affect from the Symptom Checklist90Revised (SCL90
crises in social relations and social events, we did not R; Derogatis, 1983). These scales have been validated
attempt to delete potentially controllable events with samples of adolescents and adults (Derogatis,
from the checklist. That is, we were interested both 1983). Using a 5-point scale (1 5 not distressed,
in potentially uncontrollable traumas as well as 5 5 extremely distressed), we asked adolescents to
crises that may be exacerbated by the adolescents report how much they had been distressed by each
own behaviors. Indeed, the distinction between symptom during the past week. The depression
controllable and uncontrollable life events is difficult subscale (12 items) demonstrated adequate reliabil-
to make. For example, parents may divorce, in part, ity in the present study, providing alpha coefficients
because of family difficulties influenced by an of .88, .91, .92, .93, and .90, in 1989, 1990, 1991, 1992,
adolescent with emotional or behavioral problems. and 1994, respectively. The reliability of the anxiety
A total of 25 items was used each year to measure subscale (10 items) also was satisfactory (a 5 .83, .83,
stressful life events. The exact same stressful events .86, .90, and .89). Examples of the 12 items for
were used in the index each year. Again, any items depression included crying easily, feeling blue,
from the original index primarily addressing delin- feeling everything is an effort, feeling hopeless about
quent behaviors were removed. Based on the final the future, and thoughts of ending his or her life.
25-item measure, each adolescent had a possible Examples of the 10 items constituting the anxiety
score of from 0 to 25 negative events for each year of subscale included symptoms of nervousness, trem-
the study. Because each event may occur indepen- bling, feeling fearful, and feeling restless. Two
dently from other events, a traditional estimate of indicators for a latent construct of internalizing
internal consistency is not appropriate for this problems were generated by summing the 12 items
measure (Hoffmann & Su, 1997; Newcomb & for depression and 10 items for anxiety.
Harlow, 1986). The index has demonstrated con-
current and predictive validity in earlier research
Results
(e.g., Ge, Lorenz, Conger, Elder, & Simons, 1994;
Swearingen & Cohen, 1985a). Preliminary data analyses did not show any statisti-
Externalizing behaviors. Adolescents reported their cally significant differences in the results due to the
externalizing problems in the form of delinquent gender of target adolescents. Two different analyses
behaviors using a measure adapted from the were used to test for gender differences. First, we
National Youth Survey (Elliott et al., 1985; Elliott, estimated the proposed autoregressive model of
Huizinga, & Menard, 1989). Each year the target reciprocity separately for boys and girls. The results
adolescents indicated how often they engaged in any were almost identical regardless of the type of
of 20 different delinquent activities during the past maladjustment (i.e., internalizing or externalizing
12 months. Items were rated on a 5-point Likert-type problems). Second, a group comparison was per-
scale (0 5 never, 4 5 six or more times). The delin- formed (Bollen, 1989). The test involved the compar-
quency scale demonstrated good internal consis- ison of a null model, which held constant the values of
tency; alpha coefficients ranged from .69 (Year 1, the path coefficients for boys and girls, with a model
1989) to .77 (Year 5, 1994). that allowed these coefficients to vary by gender of
To form indicators of a latent construct for adolescent. If there were substantial gender differ-
externalizing behaviors, we divided the 20 items ences in the hypothesized model, the chi-square value
into two dimensions: major and minor delinquent should improve when the paths are allowed to vary.
activities. Examples of major activities (10 items in With regard to the relationship between stressful
total) included stealing something worth more than life events and externalizing problems, the null
Stressful Life Events and Adolescent Maladjustment 133

model yielded a chi-square of 779.69 with 166 of major delinquency (kurtosis values ranged from
degrees of freedom. The chi-square statistic of the 9.41 to 18.47). However, simulation studies indicate
comparison model that allowed the structural paths that parameter estimates and statistical inferences in
in the model to vary by gender was 754.75 (df 5 150). structural equation models using maximum like-
The change in chi-square was not statistically lihood estimation, as used in the present analyses,
significant. Similarly, a difference in chi-square are robust even when the assumption of multivariate
between the null model and the gender comparison normality is violated (Bollen, 1989). Therefore, we
model for internalizing problems was not large did not transform the data to reduce skewness for
enough to conclude that the model behaved differ- any measures in the study.
ently by adolescent gender (Dw2 with 16 df 5 25.24).
Given the absence of gender differences in relation-
Reciprocal Influences Between Life Stress and
ships among stressful events and maladjustment, the
Externalizing Problems
following results are based on the combined sample
of boys and girls. Figure 1 shows the results of the cross-lagged
Additionally, preliminary analyses demonstrated estimation of the relationship between stressful life
almost exactly the same results using either listwise events and adolescent externalizing behaviors using
deletion or EM imputation. Because earlier simula- LISREL 8.30 and maximum likelihood estimation
tion studies demonstrated less bias with EM than (Joreskog & Sorbom, 1999). A covariance matrix was
with listwise deletion (Schafer, 1997), the results used to estimate the cross-lagged autoregressive
from the imputed data are presented in this article. model. All factor loadings for externalizing pro-
Means, standard deviations, and intercorrelations blems, ranging between .64 and .92, indicate a good
among all variables used in this investigation are correspondence between our measures of problem
presented in Table 1. There were significant correla- behaviors and the latent construct. Both stressful life
tional relationships between stressful life events and events and externalizing problems demonstrated
internalizing problems with rs ranging from .11, reasonably high stability from one measurement
po.05 (stressful life events at Time 1 and depression point to the next. The lowest stability path coefficient
at Time 5, and events at Time 1 and anxiety at Time for stressful life events was found between 10th and
4), to .43 , po.01 (events and anxiety at Time 2). 12th grades (b 5 .48, po.01). Stability coefficients for
Similar findings occurred for stressful life events and externalizing problems ranged from .43 (7th8th
externalizing problems. Furthermore, correlations grade) to .61 (9th10th grade).
between the variables were significant not only Most important, the findings supported the hy-
within the same measurement period but also across pothesis that stressful life events and delinquent
several points in time. Moreover, earlier stressful behaviors would be reciprocally interrelated over
events were significantly correlated with later time. For example, delinquent behaviors occurring
maladjustment and vice versa, consistent with the during the 7th grade predicted an increase in stressful
reciprocal influence hypothesis. And, as is usually life events from 7th to 8th grade (b 5 .07, po.01), and
observed, internalizing and externalizing problems stressful life events during 7th grade predicted an
also were intercorrelated (e.g., Aseltine, Gore, & increase in delinquent behaviors from 7th to 8th
Colten, 1998). These correlational findings suggestd grade (b 5 .21, po.01). It is worth noting that these
that more formal tests of the mutual influence reciprocal influences also were found between the last
hypothesis would be appropriate. two waves in which there was a 2-year lag instead of
Because including both forms of maladjustment the 1-year lag during the earlier years of the study.
in a single model would produce too many para- Except for the period from 10th to 12th grade, the
meter estimates per available cases (1.5), we esti- magnitudes of the path coefficients for the paths from
mated the models separately for externalizing and stressful life events to externalizing problems ap-
internalizing problems. It also should be noted that peared larger than those for the paths from externa-
descriptive statistics showed that some of the lizing problems to stressful life events.
measures were skewed. Stressful life events, minor We performed additional analyses to examine
delinquency, and depression scores were not highly whether systematic increases over time in the path
skewed (average kurtosis value 5 3.06). Anxiety coefficients from externalizing problems to stressful
scores appeared to be more seriously skewed than life events and the reversed pattern for the paths
depression scores (average kurtosis value 5 7.01). from stressful life events to externalizing problems
Considering the nature of the measure, it was not were statistically significant. First, a test for detecting
surprising to detect severe skewness in the measure the significance of these patterns was performed by
134
Table 1

Kim, Conger, Elder, and Lorenz


Means, Standard Deviations, and Intercorrelations Among the Study Variables (N 5 451)

Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Stressful life events


1. Time 1 (7th grade)
2. Time 2 (8th grade) .52nn
3. Time 3 (9th grade) .46nn .67nn
4. Time 4 (10th grade) .39nn .54nn .65nn
5. Time 5 (12th grade) .31nn .33nn .35nn .42nn
Internalizing problems
Depression
6. Time 1 (7th grade) .31nn .28nn .20nn .19nn .18nn
7. Time 2 (8th grade) .25nn .37nn .24nn .33nn .20nn .49nn
8. Time 3 (9th grade) .22nn .24nn .31nn .34nn .24nn .41nn .54nn
9. Time 4 (10th grade) .13nn .18nn .18nn .39nn .30nn .38nn .53nn .61nn
10. Time 5 (12th grade) .11n .17nn .15nn .20nn .24nn .31nn .39nn .53nn .51nn
Anxiety
11. Time 1 (7th grade) .33nn .31nn .17nn .19nn .15nn .76nn .43nn .34nn .30nn .21nn
12. Time 2 (8th grade) .25nn .43nn .31nn .34nn .20nn .39nn .72nn .51nn .47nn .33nn .42nn
13. Time 3 (9th grade) .22nn .27nn .38nn .33nn .19nn .35nn .45nn .76nn .50nn .44nn .33nn .58nn
14. Time 4 (10th grade) .11n .18nn .22nn .35nn .23nn .30nn .43nn .51nn .78nn .46nn .27nn .48nn .59nn
15. Time 5 (12th grade) .12nn .22nn .24nn .23nn .30nn .27nn .31nn .49nn .47nn .76nn .23nn .39nn .53nn .57nn
Externalizing problem behaviors
Major activities
16. Time 1 (7th grade) .29nn .30nn .24nn .16nn .11n .18nn .10n .18nn .04 .10n .24nn .13nn .25nn .08 .14nn
17. Time 2 (8th grade) .25 .39 .28 .14 .10
nn nn nn nn n
.11n
.18nn
.05 .03 .02 .14 nn
.20nn
.07 .03 .03 .39nn
18. Time 3 (9th grade) .28nn .46nn .49nn .38nn .15nn .16nn .13nn .25nn .11n .11n .14nn .29nn .33nn .17nn .19nn .38nn .42nn
19. Time 4 (10th grade) .19nn .43nn .40nn .43nn .18nn .01 .19nn .12nn .10n .03 .02 .26nn .19nn .12n .09 .26nn .40nn .59nn
20. Time 5 (12th grade) .24 .25 .26 .28 .25
nn nn nn nn nn
.03 .03 .12n
.08 .00 .02 .14nn .15nn .15nn .08 .13nn .31nn .28nn .42nn
Minor activities
21. Time 1 (7th grade) .36nn .43nn .35nn .24nn .13nn .25nn .08 .14nn .06 .04 .30nn .19nn .18nn .10n .09 .50nn .36nn .34nn .20nn .23nn
22. Time 2 (8th grade) .29nn .45nn .36nn .24nn .11n .07 .17nn .05 .01 .05 .12n .25nn .15nn .05 .02 .31nn .62nn .41nn .44nn .33nn .47nn
23. Time 3 (9th grade) .33 .43 .52 .38 .17
nn nn nn nn nn
.11n
.09 .20nn
.10n
.05 .14 nn
.28nn
.33 nn
.18nn
.17nn .38nn .40nn .66nn .47nn .38nn .45nn .55nn
24. Time 4 (10th grade) .28nn .38nn .43nn .49nn .24nn .03 .14nn .16nn .16nn .07 .08 .25nn .28nn .21nn .10n .29nn .30nn .53nn .64nn .42nn .31nn .44nn .63nn
25. Time 5 (12th grade) .21nn .27nn .29nn .33nn .35nn .02 .08 .10n .11n .10n .07 .16nn .14nn .14nn .17nn .15nn .32nn .23nn .35nn .38nn .28nn .41nn .37nn .48nn
Mean 5.79 4.47 4.81 4.95 3.48 19.43 17.86 18.25 18.68 19.86 15.02 13.89 14.04 13.70 14.12 .28 .33 .36 .46 .74 .44 .59 .87 1.08 1.66
Standard deviation 3.51 3.32 3.36 3.22 2.60 7.07 5.66 6.36 7.06 7.62 5.08 4.51 4.55 4.96 5.12 .61 .67 .74 .86 1.14 .84 .98 1.25 1.37 1.59
po.05.
n

po.01.
nn
Stressful Life Events and Adolescent Maladjustment 135

Stressful .50 Stressful .57 Stressful .50 Stressful .48 Stressful


Life Events Life Events Life Events Life Events Life Events
7th Grade 8th Grade 9th Grade 10th Grade 12th Grade
R2 = .28 R2 = .43 R2 = .42 R2 = .19

.07 .08 .11 .15

.42 .27 .23 .20 .20


.21 .16 .14 .10

Externalizing Externalizing Externalizing Externalizing


Externalizing Problems Problems Problems
Problems .43 Problems .52 .61 .46
8th Grade 9th Grade 10th Grade 12th Grade
7th Grade R2 = .46
R2 = .30 R2 = .56 R2 = .33
.64 .81 .70 .83 .77 .83 .75 .84 .74 .92

Major Minor Major Minor Major Minor Major Minor Major Minor
Activity Activity Activity Activity Activity Activity Activity Activity Activity Activity

2 = 162.91
D.F. = 52
GFI = .96
AGFI = .90
RMSEA = .06
CN = 218.15
Figure 1. Results of the structural equation analysis evaluating the theoretical model of reciprocity between stressful life events and
adolescent externalizing problems. Although not shown, residuals for the same indicators of delinquent behaviors at different points in
time were correlated. Maximum likelihood estimation of the theoretical model (N 5 451) using completely standardized coefficients,
po.01, for all coefficients. GFI 5 goodness-of-fit index; AGFI 5 adjusted goodness-of-fit index; RMSEA 5 root mean square error of
approximation; CN 5 Hoelters critical N.

constraining the corresponding paths to be equal to at each point in time were statistically significant. We
one another for different periods. For example, the constrained the two paths to be equal to one another
path from externalizing problems at Time 1 to for each point in time. During the period from 7th to
stressful life events at Time 2 was set to be equal to 8th grade, the test indicated that the magnitudes of
the path from externalizing problems at Time 2 to the two path coefficients were statistically different
stressful life events at Time 3, and so on. The (Dw2 with 1 df 5 13.09). This result suggests that the
constraints did not reduce the model fit. This result path coefficient of .21 (po.01) from stressful life
indicates that the gradual increments in the path events to externalizing problems indicates a greater
coefficients from externalizing problems to stressful effect of life stress on externalizing problems than of
life events over time and decreases in the paths from externalizing problems on stressful life events
stressful life events to externalizing problems were (b 5 .07, po.01). The same pattern was detected
not statistically significant. We speculate that the from 8th to 9th grade. The same test did not yield
apparent changes in the magnitude of the coeffi- any significant difference in the size of the path
cients might relate to the changing distributions in coefficients from 10th to 11th grade. However,
these variables across time. Table 1 demonstrates during the last two waves (10th and 12th grades),
that the standard deviations for stressful events the size of the path coefficient from externalizing
decline from early to late adolescence, whereas the problems to life stress turned out to be larger than
reverse pattern occurs for externalizing behaviors. that from life stress to externalizing problems (Dw2
Because a greater distribution in a predictor variable with 1 df 5 8.27). These findings indicate that there
increases its effect as a predictor, these patterns in may be important developmental changes in the
the variables may account for the nonsignificant relationship between stress and delinquency across
changes in path coefficients in Figure 1. the years of adolescence. We return to this issue in
A second test was performed to examine whether the discussion section.
differences in the size of path coefficients from Despite the large number of parameters that were
stressful life events to externalizing problems and estimated, the proposed autoregressive model
from externalizing problems to stressful life events showed a reasonably good fit with the data. With
136 Kim, Conger, Elder, and Lorenz

52 degrees of freedom, the chi-square was 162.91 current article, we also examined the contempora-
(po.01). If the ratio between the chi-square statistic neous relationships between stressful events and
and the degrees of freedom is less than 3, the tested externalizing problems and found almost identical
model is considered a good fit (Cole, 1987). Because results to those from the cross-lagged model
the chi-square statistic is extremely sensitive to presented in Figure 1 (results are available from
sample size (e.g., Bentler, 1990; Bentler & Bonett, the first author). However, as Lorenz et al. suggested
1980), however, the large chi-square value is not should be the case, the path coefficients between
unusual for a sample of this size. Thus, recommen- stressful life events and externalizing problems were
dations on evaluations of structural equation models larger in the contemporaneous analysis (bs ranging
suggest the use of more than one fit index (e.g., from .17 to .38, all pso.01). The contemporaneous
Bentler, 1990; Hu & Bentler, 1999). The other fit model also demonstrated a good fit with the data.
indices reported in Figure 1, including goodness-of- The findings from this alternative analytic approach
fit (GFI; .96), the adjusted goodness-of-fit (AGFI; .90), indicate that the cross-lagged results may under-
the root mean square error of approximation estimate the degree of reciprocity between stress and
(RMSEA; .06), and Hoelters critical N (CN; 218.15), delinquency. Because the cross-lagged findings
indicated that the model fit the data reasonably well. provide the more conservative estimates, however,
When the GFI and AGFI are .90 or more, the RMSEA we chose to emphasize them in this report.
is less than .06, and CN exceeds 200, then the
hypothesized model fits the observed data well
Reciprocal Influences Between Stressful Life Events and
(Bentler, 1990; Bentler & Bonett, 1980; Cole, 1987;
Internalizing Problems
Cramer, 1999; Hoelter, 1983; Hu & Bentler, 1999;
Joreskog & Sorbom, 1999). As described earlier, internalizing problems were
Although these various fit indices reached the measured by a series of items asking how distressed
same conclusion that there was a reasonable fit of the the adolescents were during the past week. How-
proposed model to the observed data, they are ever, stressful life events were assessed by the total
considered a point estimate of fit. MacCallum, number of events that occurred to each target
Browne, and Sugawara (1996) argued that a point adolescent during the past year. This difference in
estimate of fit might not be a precise test to estimate measurement invalidates the use of a standard cross-
model fit in the population. They suggested con- lagged model (e.g., Figure 1) to investigate reciprocal
structing a confidence interval around a specific fit influences between stressful life events and inter-
index, the RMSEA, to assure that the true value of nalizing problems. Because of the measurement
the fit index for a proposed model in a given procedures, the time lag between stressful life events
population exists within the confidence interval. If a at Time 1 and internalizing problems at Time 2 can
constructed confidence interval contains .05, the test be as much as 2 years. For example, stressful life
fails to reject the null hypothesis that a proposed events reported by the targets in the 7th grade could
model demonstrates a good fit. MacCallum et al. actually have occurred during the 6th grade,
also argued that a confidence interval including .08 whereas depressive and anxious moods measured
indicates a fair fit of a tested model. Following their in 8th grade indicated adolescents emotional dis-
suggestion, we formed a 90% confidence interval for tress during the previous week. For that reason, we
the RMSEA; the interval was between .054 and .078. used negative events reported in 8th grade, which
Because the interval did not include either .05 occurred in the past year including during the 7th
(marker of good fit) or .08 (marker of fair fit), the grade, to predict internalizing symptoms reported in
test result indicate that the fit of our model from the the 8th grade. Moreover, we used internalizing
observed sample was reasonable. problems from one year to predict negative life
Finally, it should be noted that an alternative events during the next year. For example, depressive
analytic strategy for addressing a reciprocal link symptoms reported in 7th grade were used to
between stressful life events and externalizing predict negative life events during the period from
problems is a contemporaneous estimation of the 7th to 8th grade as reported in the 8th grade. Our
relationship (Compas & Phares, 1991). Indeed, analytic strategy followed from these methods of
Lorenz, Conger, Simons, and Whitbeck (1995) measurement. The results are provided in Figure 2.
demonstrated that a contemporaneous model and a As shown in Figure 2, all factor loadings for the
cross-lagged model are mathematically equivalent to indicators of internalizing problems were statisti-
each other, at least for a two-wave, two-variable, cally significant ranging from .81 (depression, Time
two-indicator model. Although not presented in the 5) to .95 (anxiety, Time 5). The stability coefficients
Stressful Life Events and Adolescent Maladjustment 137

Stressful .47 Stressful .55 Stressful .51 Stressful .29 Stressful


Life Events Life Events Life Events Life Events Life Events
7th Grade 8th Grade 9th Grade 10th Grade 12th Grade
R2 = .28 R2 = .44 R2 = .43 R2 = .19

.13 .10 .12 .16

.36

.26 .24 .21 .16

Internalizing Internalizing Internalizing Internalizing Internalizing


Problems .47 Problems .49 Problems .41 Problems .42 Problems
7th Grade 8th Grade 9th Grade 10th Grade 12th Grade
R2 = .07 R2 = .37 R2 = .44 R2 = .41 R2 = .35
.83 .91 .84 .85 .84 .90 .85 .91 .81 .95
Depression Anxiety Depression Anxiety Depression Anxiety Depression Anxiety Depression Anxiety

2 = 120.39
D.F.= 56
GFI = .97
AGFI = .93
RMSEA = .05
CN = 313.18
Figure 2. Results of the structural equation analysis evaluating the theoretical model of reciprocity between stressful life events and
adolescent internalizing problems. Although not shown, residuals for the same indicators of internalizing problems at different points in
time were correlated. Maximum likelihood estimation of the theoretical model (N 5 451) using completely standardized coefficients,
po.01, for all coefficients. GFI 5 goodness-of-fit index; AGFI 5 adjusted goodness-of-fit index; RMSEA 5 root mean square error of
approximation; CN 5 Hoelters critical N.

for stressful life events were all statistically signifi- Statistical Power
cant, ranging from .29 (between 10th and 12th grades)
We also considered the issue of statistical power
to .55 (between 8th and 9th grades). The stability in
because some researchers might question whether
internalizing problems appeared to be more uniform
the results from our two autoregressive models
compared with stressful life events (bs ranging from
would be reliable given a sample size of 451 and 105
.41 to .49; all pso.01). All eight paths between life
covariances generated for evaluating the models.
stress and internalizing problems were statistically
MacCallum et al. (1996) found that power increases
significant and positive, providing substantial em-
as either degrees of freedom or sample size
pirical support for the mutual influence or reciprocity
increases. In their study, a minimum sample size
model. For example, internalizing symptoms during
for power of .80 was 243 with 50 degrees of freedom.
7th grade predicted higher levels of stressful life
Each of our two autoregressive models contained 52
events during 7th and 8th grades (b 5 .13, po.01).
degrees of freedom with a sample size of 451.
Equally, adverse life events during 7th and 8th grades
According to MacCallum et al., the estimated power
predicted an increase in internalizing problems from
for 50 degrees of freedom with a sample size of 400
7th to 8th grade (b 5 .26, po.01).
was .979 in covariance structure modeling. Based on
The autoregressive model also showed a good fit
these statistics, we concluded that our models were
with the data. With 56 degrees of freedom, the chi-
estimated with adequate statistical power.
square was 120.39. The other fit indices such as GFI
(.97), AGFI (.93), RMSEA (.05), and CN (313.18)
Comparisons of Alternative, Hierarchically Related
indicated that the model fit the data well. A 90%
Models
confidence interval around the RMSEA was also
formed for internalizing problems (MacCallum et al., Although our proposed reciprocal causation
1996). The lower bound of the confidence interval models for both externalizing and internalizing
was .037 and the upper bound was .062. Because the problems demonstrated reasonable fit with the data,
entire confidence interval was below the indicator of it is possible that theoretically interesting alternative
fair fit (.08), results suggest that the fit was reasonable. models would also fit the data well. Indeed, there
138 Kim, Conger, Elder, and Lorenz

Table 2
Comparisons of Theoretically Competing Models

Externalizing problems Internalizing problems

w2/df Ddf Dw2 w2/df df w2

1. Null model: stability and measurement model 223.97/60 256.55/64


2A. Social-causation model 177.50/56 4 46.47 149.41/60 4 107.14
2B. Social-selection model 195.87/56 4 28.10 217.41/60 4 39.14
3. Reciprocal-influence model 162.91/52 120.39/56
(1) Comparison with Model 1 8 61.06 8 136.16
(2) Comparison with Model 2A 4 14.59 4 29.02
(3) Comparison with Model 2B 4 32.96 4 97.02

Note. For all Dw2/Ddf, po.01.

often will be alternative equivalent models assessing model reduced the amount of unexplained variance
different path relationships from those evaluated in in the model. In that sense, the substantively and
an originally proposed model (e.g., MacCallum, statistically significant change in chi-square com-
Wegener, Uchino, & Fabrigar, 1993). The important pared with the null model for each of the three
issue here was that structural equation models using theoretical models was expected.
a covariance structure can generate a reasonably The most interesting question was whether the
good fit to the data from either the proposed model reciprocal influence model would significantly im-
or alternative equivalent models or even both. Thus, prove the fit between the model and the observed
goodness-of-fit tests cannot distinguish the original data compared with the social-selection or social-
model from equivalent models (MacCallum et al., causation models. For externalizing problems, the
1993). Rather, theoretical meaningfulness and inter- change in chi-square from the social-causation
pretability of path coefficients should be used to model to the reciprocal-influence model was 14.59
differentiate the original model from alternative with 4 degrees of freedom. The chi-square change
equivalent models (MacCallum et al., 1993). In our from the social-selection model to the reciprocal-
study, either the social-causation model or the social- influence model was 32.96, also with 4 degrees of
selection model might yield a fit that is as good as freedom. Both of these chi-square change statistics
that of the reciprocal-causation model. However, if were significant (po.01). A very similar pattern
the reciprocal-causation model could improve the emerged in the chi-square statistics for internalizing
overall model fit compared with either the social- problems. The reciprocal influence model signifi-
causation or the social-selection models, the argu- cantly improved model fit compared with both the
ment for reciprocity in the relationship between social-causation model (Dw2 with 4 df 5 29.02) and
stressful life events and adolescent externalizing and the social-selection model (Dw2 with 4 df 5 97.02).
internalizing problems would be strengthened. These results indicate that our hypothesized reci-
Table 2 presents the results of theoretically procal influence model demonstrated a substan-
competing model comparisons. As proposed by tively and statistically better correspondence with
Bollen (1989), direct model comparisons were made the observed data than either the social-causation
among alternative, hierarchically related models by model or the social-selection model.
interpreting the change in chi-square statistics per
change in degrees of freedom. Using the stability
Discussion
and measurement model as a null model, we found
that all three theoretically competing models yielded Three goals guided the present inquiry. The first goal
a significant improvement in chi-square compared was to extend earlier research by examining the
with the null model. For example, the chi-square relationship between life stress and both internaliz-
statistic for the social-causation model for externa- ing and externalizing problems of adolescents.
lizing problems was 177.50, which improved the chi- The second goal was to evaluate both the influence
square statistic compared with the null model by of stressful events on adolescent maladjustment
46.47. The improvement was statistically significant (the social-causation hypothesis) and the influence
at po.01. Because the null model did not include any of maladjustment on stressful events (the social-
structural relationships, the fit of the model was selection hypothesis). We predicted that both
expected to be poor. Adding structural paths to the hypotheses would be supported, thus revealing a
Stressful Life Events and Adolescent Maladjustment 139

reciprocal process of increasing risk over time. behaviors and that these dimensions of maladjust-
The third goal was to examine the hypothesized ment would lead to increasing life stress. We have
reciprocal influence process between life stress and been able to identify only three previous studies that
internalizing and externalizing problems across provided preliminary evidence showing that either
the full span of adolescence. In turn, we consider depressive mood (Swearingen & Cohen, 1985a) or
the results of the study in relation to each of these delinquent behaviors (Aseltine et al., 2000; Leadbea-
objectives. ter et al., 1999) predict stressful life events. The
primary limitation in these earlier reports was that
they did not control for earlier events in predicting
Connecting Stressful Life Events to Internalizing and
later events; thus, they could not determine whether
Externalizing Problems
adolescent maladjustment actually predicted change
Regarding the two dimensions of adjustment in life stress. Our findings did show that maladjust-
problems, earlier research has demonstrated an ment predicts increases in life stress over time and
association between stressful events and depressive provided evidence for a mutually reinforcing pro-
symptoms (Aseltine et al., 1994; Leadbeater et al., cess in the relationship between stressful life events
1995, 1999; Swearingen & Cohen, 1985a). The results and adolescent emotional and behavioral problems.
from the current study replicated the findings from We found consistent evidence across the years of
this previous body of research by demonstrating that adolescence that both earlier delinquent behaviors
life stresses adolescents experience have a similar and depressed or anxious mood predicted later
effect on their overall emotional distress. The stressful life events after controlling for the previous
analysis of our five-wave autoregressive model level of stressful life events. Thus, both the social-
showed that a higher number of negative life events selection and the social-causation hypotheses were
predicted higher levels of depressed and anxious supported in this investigation.
mood among the adolescents in this study. Previous More intriguing is the fact that the results were
studies showed this adverse association between consistent with our hypothesis of reciprocal causa-
negative life events and depressive symptoms tion in which the two competing hypotheses are
among adolescents primarily using cross-sectional integrated. Support for the reciprocity hypothesis
data (e.g., Swearingen & Cohen, 1985a). Our study suggests a complex interconnection between stress-
adds to this research by showing that stressful ful life events and adolescent emotional or behavior-
events in the lives of adolescents predict increasing al problems that cannot be explained by either the
risk for internalizing problems over time. social-causation hypothesis or the social-selection
Regarding externalizing problems, there exists hypothesis alone. Specifically, the test of the theore-
research evidence for a connection between summa- tical model in this investigation demonstrated a
tive indices of stressful life events and angry, acting- reciprocal process through which stressful life events
out behaviors by adolescents (Aseltine et al., 2000, and adolescent maladjustment can be thought of as
Hoffmann & Su, 1997; Paternoster & Mazerolle, both causes and effects over time.
1994; Vaux & Ruggiero, 1983). The results from our Although the results of the present investigation
five-wave study were in line with previous findings. were consistent with the hypothesized reciprocal
Our findings demonstrated that stressful life events process of mutual influence between stressful events
at each point significantly and positively predicted and adolescent maladjustment, the results also
delinquent behaviors at the next measurement point suggest that this dynamic may change over time,
after controlling for the previous level of delinquent at least for externalizing behaviors. One of the
behaviors. These results significantly increased con- statistical tests that we performed revealed that
fidence in the hypothesis that stressful events will there were differences in cross-lagged influences
exacerbate both internalizing and externalizing from early to late adolescence. More specifically, the
problems in the lives of adolescents. effect of life stress on externalizing problems during
early adolescence (7th9th grade) turned out to be
greater than the effect of externalizing problems on
Examining Reciprocal Influences
life stress. There was no difference in the effect size
Our second goal was to test a reciprocal causation of either life stress or externalizing problems on the
hypothesis that takes the place of simpler notions of other study variable during middle adolescence.
either social causation or social selection. From However, during late adolescence, the effect size of
this perspective, we expected that life stressors externalizing problems to life stress was statistically
would increase emotional distress and delinquent larger than from life stress to externalizing
140 Kim, Conger, Elder, and Lorenz

behaviors. The findings indicate that social influ- (Conger & Simons, 1997; Rosenfield, 1999), the
ences may be relatively stronger during early results from our study did not support a gender-
adolescence, when stressful life events were more specific hypothesis regarding the influence of stress-
predictive of increasing maladjustment than the ful life events. That is, both boys and girls in our
reverse. We expect this differential effect may result sample showed very similar responses to stressful
from the rapid pace of life changes during early life events, consistent with some previous research
adolescence. By late adolescence, externalizing pro- (Gore et al., 1992; Hoffmann & Su, 1997; Leadbeater
blems had a greater influence on stressful events et al., 1995; Vaux & Ruggiero, 1983).
than the reverse, consistent with the idea that
increasing mean levels of delinquent activities (see
Other Issues
Table 1) will have a relatively greater effect on social
circumstances. These issues need to be addressed in The present results also suggest several important
future research. issues for future research. First, this investigation
was specifically geared toward studying a hypothe-
sized reciprocal process through which adolescent
Investigating Reciprocity Over an Extended Period
maladjustment and stressful life events were ex-
Regardless of the dimension of problem behavior, pected to be connected. Hence, the current study did
our finding that reciprocal influences persisted over not address specific factors that have either buffering
time is consistent with the idea of cumulative or amplifying effects on adversity in the reciprocal
continuity proposed by Caspi and colleagues process between the two. A fair amount of research
(1987). As they suggested, problem behaviors should suggests that the effect of social stressors on mental
produce stressful consequences and these negative health will be reduced for people with greater
consequences will tend to reinforce problem beha- compared with fewer personal or social resources
viors over time. For example, earlier research (e.g., sense of control or social support; Ensel & Lin,
suggests that the likelihood of encountering conflict 1991; Pearlin, 1999; Vinokur & van Ryn, 1993).
situations with parents is higher for adolescents who Related to delinquency in adolescents, for example,
engage in various deviant behaviors and who a recent study showed that good parenting tactics
demonstrate poor academic performance (Elliott have beneficial and protective effects on adolescents
et al., 1985; Herrenkohl et al., 2000; Sampson & delinquent behaviors (Scaramella, Conger, &
Laub, 1993). These conflicts with parents, in turn, are Simons, 1999). Future research should identify
likely to exacerbate problem behaviors (Conger, factors that buffer as well as intensify the self-
1997; Conger & Simons, 1997). In general, our perpetuating cycle of stressful life situations and
findings are consistent with the proposal that adolescent maladjustment revealed in the present
problem behaviors will increase the risk that analyses.
adolescents will experience stressful life situations, Second, although the primary goal of the current
which, in turn, exacerbate these behaviors in a study was not to examine co-occurring externalizing
reciprocal process over time (Caspi et al., 1987; and internalizing problems, growing evidence for
Compas & Phares, 1991; Sampson & Laub, 1993). their co-occurrence should not be ignored. For
This risk involves problems in both instrumental example, recent studies have documented the
activities as well as social relationships. Because of co-occurrence of substance use and depression
the idiosyncrasies in their behaviors, for example, (Aseltine et al., 1998) in adolescent populations.
depressed, anxious, or delinquent youth might be One of the limitations of the present investigation
less desirable social partners and less able in was that we did not have a sufficient sample to
instrumental activities than their more well-adjusted evaluate both these related processes and the
peers (e.g., Pope & Bierman, 1999). Crises appear to reciprocal effects of primary interest in our five-
intensify emotional or behavioral problems, which, wave autoregressive models. These questions should
in turn, instigate further crises. Our confidence in be pursued in future investigations.
these results is increased by the fact that the Also, the findings need to be replicated across
reciprocal process was replicated during each stage diverse ethnic and socioeconomic populations. Be-
of adolescence. cause the original project from which the sample for
These findings were consistent for boys and girls this investigation came was designed to study the
in our study. Although girls and women experience effect of economic hardship on families living in the
higher levels of internalizing problems and boys and rural Midwest, and because there are almost no
men demonstrate more externalizing behaviors minority families living in this area, the sample was
Stressful Life Events and Adolescent Maladjustment 141

restricted to White, two-parent families. Future Caspi, A., Elder, G. H., Jr., & Bem, D. J. (1987). Moving
research is needed to determine whether the present against the world: Life-course patterns of explosive
findings generalize to families of different ethnic children. Developmental Psychology, 23, 308313.
groups, single-parent families, and urban or inner- Cohen, S., Kessler, R. D., & Gordon, L. U. (1995). Strategies
city families. Earlier attempts at replication of other for measuring stress in studies of psychiatric and
physical disorders. In S. Cohen, R. D. Kessler, & L. U.
findings from this study with urban and minority
Gordon (Eds.), Measuring stress: A guide for health and
populations have been successful, however, increas- social scientists (pp. 328). New York: Oxford University
ing confidence in the potential generalizability of the Press.
current results (Conger, Patterson, & Ge, 1995; Cole, D. A. (1987). Utility of confirmatory factor analysis in
Conger et al., 2002). test validation research. Journal of Consulting and Clinical
Despite remaining issues to be addressed in the Psychology, 55, 584594.
future, the findings from this investigation provided Compas, B. E., & Phares, V. (1991). Stress during childhood
important new evidence regarding the reciprocal and adolescence: Sources of risk and vulnerability.
process through which social and personal disad- In E. M. Cummings, A. L. Greene, & K. H. Karraker
vantage appear to be mutually reinforcing over time. (Eds.), Life-span developmental psychology: Perspectives
These results suggest that negative life events on stress and coping (pp. 111129). Hillsdale, NJ:
Erlbaum.
experienced during early adolescence intensify
Compas, B. E., & Wagner, B. M. (1991). Psychosocial stress
symptoms of sadness, fear, and antisocial conduct. during adolescence: Intrapersonal and interpersonal
These domains for maladjustment, in turn, increase processes. In M. E. Colten & S. Gore (Eds.), Adolescent
risk for future adversities and life crises. The results stress: Causes and consequences (pp. 6786). New York:
also suggest that the reciprocal process between Aldine de Gruyter.
negative life events and maladjustment has a Conger, R. D. (1995). Unemployment. In D. Levinson (Ed.),
developmental dynamic that unfolds in a mean- Encyclopedia of marriage and the family (pp. 731735). New
ingful and understandable fashion across the years York: MacMillan.
of adolescence. Conger, R. D. (1997). The social context of substance abuse:
A developmental perspective. In E. B. Robertson,
Z. Sloboda, G. M. Boyd, L. Beatty, & N. J. Kozel.
References
(Eds.), Rural substance abuse: State of knowledge and issues
Aneshensel, C. S. (1992). Social stress: Theory and research. (Research Monograph No. 168, NIH Publication No. 97-
Annual Review of Sociology, 18, 1538. 4177, pp. 636). Washington, DC: National Institute on
Aseltine, R. H., Jr., Gore, S., & Colten, M. E. (1994). Drug Abuse.
Depression and the social developmental context of Conger, R. D., & Elder, G. H., Jr. (1994). Families in troubled
adolescence. Journal of Personality and Social Psychology, times: Adapting to change in rural America. Hillsdale, NJ:
67, 252263. Aldine.
Aseltine, R. H., Jr., Gore, S., & Colten, M. E. (1998). The co- Conger, R. D., Patterson, G. R., & Ge, X. (1995). It takes two
occurrence of depression and substance abuse in late to replicate: A mediational model for the impact of
adolescence. Development and Psychopathology, 10, parents stress on adolescent adjustment. Child Develop-
549570. ment, 66, 8097.
Aseltine, R. H., Jr., Gore, S., & Gordon, J. (2000). Life stress, Conger, R. D., & Simons, R. L. (1997). Life-course
anger and anxiety, and delinquency: An empirical test of contingencies in the development of adolescent anti-
general strain theory. Journal of Health and Social Behavior, social behavior: A matching law approach. In T. P.
41, 256275. Thornberry (Ed.), Advances in criminological theory: Vol. 7
Bentler, P. M. (1990). Comparative fit indexes in structural (pp. 5599). New Brunswick, NJ: Transaction Publishers.
models. Psychological Bulletin, 107, 238246. Conger, R. D., Wallace, L. E., Sun, Y., Simons, R. L.,
Bentler, P. M., & Bonett, D. G. (1980). Significance tests and McLoyd, V. C., & Brody, G. H. (2002). Economic
goodness of fit in the analysis of covariance structures. pressure in African American families: A replication
Psychological Bulletin, 88, 588606. and extension of the family stress model. Developmental
Bollen, K. A. (1989). Structural equations with latent variables. Psychology, 38, 179193.
New York: Wiley. Cramer, K. M. (1999). Psychological antecedents to help-
Brown, G. W. (1989). Life events and measurement. In G. seeking behavior: A reanalysis using path modeling
Brown & T. O. Harris (Eds.), Life events and illness (pp. structures. Journal of Counseling Psychology, 46,
345). New York: Guilford Press. 381387.
Caspi, A. (1998). Personality development across the Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977).
life course. In W. Damon & N. Eisenberg (Eds.), Maximum likelihood from incomplete data via the EM
Handbook of child psychology (Vol. 3, pp. 311388). New algorithm. Journal of the Royal Statistical Society, Series B,
York: Wiley. 39, 138.
142 Kim, Conger, Elder, and Lorenz

Derogatis, L. R. (1983). SCL-90-R: Administration, scoring, Stress and adversity over the life course: Trajectories and
and procedures manual-II. Towson, MD: Clinical Psycho- turning points (pp. 2949). New York: Cambridge
metric Research. University Press.
Dohrenwend, B. P., & Dohrenwend, B. S. (1969). Social Kessler, R. C., & Greenberg, D. F. (1981). Linear panel
status and psychological disorder: A causal inquiry. New analysis: Models of quantitative change. New York: Aca-
York: Wiley. demic Press.
Dornbusch, S. M., Mont-Reynaud, R., Ritter, P. L., Chen, Z., Leadbeater, B. J., Blatt, S. J., & Quinlan, D. M. (1995).
& Steinberg, L. (1991). Stressful events and their Gender-linked vulnerabilities to depressive symptoms,
correlates among adolescents of diverse backgrounds. stress, and problem behaviors in adolescents. Journal of
In M. E. Colten & S. Gore (Eds.), Adolescent stress: Causes Research on Adolescence, 5, 129.
and consequences (pp. 111130). New York: Aldine de Leadbeater, B. J., Kuperminc, G. P., Blatt, S. J., & Hertzog,
Gruyter. C. (1999). A multivariate model of gender differences in
Elder, G. H., Jr. (1998). The life course as developmental adolescents internalizing and externalizing problems.
theory. Child Development, 69, 112. Developmental Psychology, 35, 12681282.
Elliott, D. S., Huizinga, D., & Ageton, S. S. (1985). Lorenz, F. O., Conger, R. D., Simons, R. L., & Whitbeck, L.
Explaining delinquency and drug use. Beverly Hills, CA: B. (1995). The effects of unequal covariances and
Sage. reliabilities on contemporaneous inference: The case of
Elliott, D. S., Huizinga, D., & Menard, S. (1989). Multiple hostility and marital happiness. Journal of Marriage and
problem youth: Delinquency, substance use, and mental the Family, 57, 10491064.
health problems. New York: Springer-Verlag. MacCallum, R. C., Browne, M. W., & Sugawara, H. M.
Ensel, W. M., & Lin, N. (1991). The life stress paradigm and (1996). Power analysis and determination of sample size
psychological distress. Journal of Health and Social for covariance structure modeling. Psychological Meth-
Behavior, 32, 321341. ods, 1, 130149.
Ge, X., Lorenz, F. O., Conger, R. D., Elder, G. H., Jr., & MacCallum, R. C., Wegener, D. T., Uchino, B. N., &
Simons, R. L. (1994). Trajectories of stressful life events Fabrigar, L. R. (1993). The problem of equivalent models
and depressive symptoms during adolescence. Develop- in applications of covariance structure analysis. Psycho-
mental Psychology, 30, 467483. logical Bulletin, 114, 185199.
Gore, S., Aseltine, R. H., Jr., & Colten, M. E. (1992). Social Menaghan, E. G. (1999). Social stressors in childhood and
structure, life stress, and depressive symptoms in a high adolescence. In A. V. Horwitz & T. L. Scheid (Eds.), A
school-aged population. Journal of Health and Social handbook for the study of mental health (pp. 315327). New
Behavior, 33, 97113. York: Cambridge University Press.
Hauser, S. T., & Bowlds, M. K. (1990). Stress, coping, and Newcomb, M. D., & Harlow, L. L. (1986). Life events
adaptation. In S. S. Feldman & G. R. Elliott (Eds.), At the and substance use among adolescents: Mediating
threshold (pp. 388413). Cambridge, MA: Harvard effects of perceived loss of control and meaninglessness
University Press. in life. Journal of Personality and Social Psychology, 51,
Herrenkohl, T. I., Maguin, E., Hill, K. G., Hawkins, J. D., 564477.
Abbott, R. D., & Catalano, R. F. (2000). Developmental Paternoster, R., & Mazerolle, P. (1994). General strain
risk factors for youth violence. Journal of Adolescent theory and delinquency: A replication and extension.
Health, 26, 176186. Journal of Research in Crime and Delinquency, 31,
Hoelter, J. W. (1983). The analysis of covariance structures: 235263.
Goodness-of-fit indices. Sociological Methods & Research, Pearlin, L. I. (1999). Stress and mental health: A conceptual
11, 325344. overview. In A. V. Horwitz & T. L. Scheid (Eds.), A
Hoffmann, J. P., & Su, S. S. (1997). The conditional effects of handbook for the study of mental health (pp. 161175). New
stress on delinquency and drug use: A strain theory York: Cambridge University Press.
assessment of sex differences. Journal of Research in Crime Pope, A. W., & Bierman, K. L. (1999). Predicting adolescent
and Delinquency, 34, 4678. peer problems and antisocial activities: The relative
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes roles of aggression and dysregulation. Developmental
in covariance structure analysis: Conventional criteria Psychology, 35, 335346.
versus new alternatives. Structural Equation Modeling, 6, Rosenfield, S. (1999). Gender and mental health: Do
155. women have more psychopathology, men more, or both
Jackson, Y., & Warren, J. S. (2000). Appraisal, social the same (and why)? In A. V. Horwitz & T. L. Scheid
support, and life events: Predicting outcome behavior (Eds.), A handbook for the study of mental health (pp. 348
in school-age children. Child Development, 71, 14411457. 360). New York: Cambridge University Press.
Joreskog, K. G., & Sorbom, D. (1999). LISREL 8.30: Users Sampson, R. J., & Laub, J. H. (1993). Crime in the making:
reference guide. Chicago, IL: Scientific Software. Pathways and turning points through life. Cambridge, MA:
Kessler, R. C., Gillis-Light, J., Magee, W. J., Kendler, K. S., & Harvard University Press.
Eaves, L. J. (1997). Childhood adversity and adult Scaramella, L. V., Conger, R. D., & Simons, R. L. (1999).
psychopathology. In I. H. Gotlib & B. Wheaton (Eds.), Parental protective influences and gender-specific
Stressful Life Events and Adolescent Maladjustment 143

increases in adolescent internalizing and externalizing Swearingen, E. M., & Cohen, L. H. (1985a). Life events and
problems. Journal of Research on Adolescence, 9, 111141. psychological distress: A prospective study of young
Schafer, J. L. (1997). Analysis of incomplete multivariate data. adolescents. Developmental Psychology, 21, 10451054.
London: Chapman & Hall. Swearingen, E. M., & Cohen, L. H. (1985b). Measurement
Simons, R. L., Johnson, C., Conger, R. D., & Elder, G. H., Jr. of adolescents life events: The Junior High Life
(1998). A test of latent trait versus life-course perspec- Experiences Survey. American Journal of Community
tives on the stability of adolescent antisocial behavior. Psychology, 13, 6985.
Criminology, 36, 217242. Thornberry, T. P., Ireland, T. O., & Smith, C. A. (2001). The
SPSS for Windows. (2000). Base system users guide, release importance of timing: The varying impact of childhood
10.0. Chicago, IL: SPSS. and adolescent maltreatment on multiple problem
Stouthamer-Loeber, M., Loeber, R., Homish, D. L., & Wei, outcomes. Development & Psychopathology, 13, 957979.
E. (2001). Maltreatment of boys and the development of Vaux, A., & Ruggiero, M. (1983). Stressful life change and
disruptive and delinquent behavior. Development & delinquent behavior. American Journal of Community
Psychopathology, 13, 941955. Psychology, 11, 169183.
Strober, M., & Carlson, G. (1982). Bipolar illness in Vinokur, A. D., & van Ryn, M. (1993). Social support
adolescents with major depression: Clinical, genetic, and undermining in close relationships: Their indepen-
and psychopharmacologic predictors in a three- to four- dent effects on the mental health of unemployed
year prospective follow-up investigation. Archives of persons. Journal of Personality and Social Psychology, 65,
General Psychiatry, 39, 549555. 350359.

Вам также может понравиться