Psychosocial Adjustment among Children

Experiencing Persistent and Intermittent

Family Economic Hardship

Kerry E. Bolger, Charlotte J. Patterson,

and William W. Thompson
University of Virginia

Janis B. Kupersmidt
University of North Carolina

BOLGER, KERRY E.; PATTERSON, CHARLOTTE J.; THOMPSON, WILLIAM W.; and KUPERSMIDT, JANIS
B. Psychosocial Adjustment among Children Experiencing Persistent and Intermittent Family
Economic Hardship. CHILD DEVELOPMENT, 1995, 66, 1107-1129. Much research on Family eco-
nomic hardship and its eFFects on children has heen cross-sectional in nature, has focused primar-
ily on white children, and has assessed the impact oF episodic rather than chronic economic
difficulties. In contrast, the present research was designed to study outcomes associated with
persistent economic hardship among a heterogeneous group oF children over time. Results
showed that. For both black and white children, a broad range oF difficulties was associated with
enduring economic hardship. Children who experienced persistent Family economic hardship
were more likely than those who did not to have diFliculties in peer relations, show conduct
problems at school, and report low selF-esteem. Children who experienced intermittent Family
economic hardship Fell between the other 2 groups. Connections between persistent economic
hardship and psychosocial adjustment were more pronounced for boys than For girls. Consistent
with suggestions by a number of investigators, the impact oF economic hardship on children in
the present sample was mediated in part by its association with parental behavior.

Cbildren are the poorest age group in Slaughter, 1988; Zill & Coiro, 1992; Zill,
the United States today. In 1992, 21,9% of Moore, Smith, Stief, & Coiro, 1991), Family
American children under the age of 18 were economic hardship has been associated with
living below federally designated poverty a wide variety of negative outcomes for chil-
levels, as compared to 14,5% of tbe entire dren, including low self-esteem (Isralowitz
population of tbe United States and to 12.9% & Singer, 1986), heightened levels of both
of the population 65 years of age or older internalizing and externalizing behavior
(U.S, Bureau of tbe Census, 1993). In 1992, problems (Werner, 1989), and lack of accep-
over 14 million American cbildren were liv- tance by peers (Patterson, Vaden, Griesler,
ing in poverty. As an issue, poverty affects & Kupersmidt, 1991; Patterson, Vaden, &
significant numbers of American cbildren. Kupersmidt, 1991). Existing research sug-
gests tbat childhood difficulties in behavior
A large body of research demonstrates and peer relationships are associated with
associations between poverty and negative adjustment problems in adolescence and
outcomes for cbildren who experience it adulthood, including delinquency, school
(e.g.. Elder, 1974; Elder, Conger, Foster, & dropout, and psychopathology (see Parker &
Ardelt, 1992; Gecas, 1979; Huston, 1991; Asber, 1987, for a review); this association
Kohn, 1963; Lempers, Clark-Lempers, & Si- makes tbe relation between poverty and
mons, 1989; McLoyd, 1989, 1990; Patter- childhood socioemotional problems a cause
son, Griesler, Vaden, & Kupersmidt, 1992; of even greater concern.

Support for this work from the Alton Jones Foundation, the United States Department oF
Education, and the Institute for Research on Poverty is gratefully acknowledged. Kerry Bolger
was supported as a predoctoral Fellow by the National Institute oF Mental Health. We wish to
thank Robert McCall For technical assistance and Robert Hauser, Deborah Phillips, and Nicholas
Zill For helpful suggestions about an earlier draft oF this report. Correspondence should be
directed to Kerry Bolger, Department oF Psychology, Gilmer Hall, University oF Virginia, Char-
lottesville, VA 22903.
[Child Development, 1995,66,1107-1129, © 1995 by the Society for Research in Child Development, Inc.
All rights reserved. 0009-3920/95/6604-0020$01.00]
1108 Child Development
Longitudinal studies of family income comes at only one time point. However, if
such as the Panel Study of Income Dynam- psychosocial risk is a transaction between
ics have revealed that, while many families the individual and the environment, ratber
with children move into and out of poverty than a fixed characteristic of tbe individual,
over time, others remain poor for long pe- then individuals may move into and out of
riods of time (Duncan et al., 1984; Duncan & risk status over time (Bell & Pearl, 1982;
Rodgers, 1988). The former group of families Sameroff, 1977), At any given time, such fac-
experiences intermittent economic hard- tors as tbe school and peer environment help
ship, whereas the latter group experiences determine whether risk manifests itself in
persistent economic hardship (Huston, poor developmental outcomes. Research
1991). Despite this, most studies of pov- that charts the sequelae of economic depri-
erty's effects on children have been cross- vation across cbildhood is essential to under-
sectional. For this reason, it has not yet been standing the effects of persistent and inter-
possible in most cases to study the effects of mittent family economic hardship.
persistent versus intermittent economic
hardship on children s development (Feath- A number of writers have suggested that
erman, Spenner, & Tsunematsu, 1988). the effects of family economic pressures are
felt by children and adolescents at least in
part through their infiuence on parental
Persistent poverty may exert a particu- emotional state and behavior (Clark, 1983;
larly strong negative infiuence on children's Conger et al,, 1992; Elder, 1979; Lempers et
adjustment. Cross-sectional research on the al,, 1989; McLoyd, 1990), McLoyd (1990)
impact of chronic and acute life stresses sug- has described a model in which conditions
gests that their effects on children's develop- of poverty increase parents' psychological
ment are likely to be cumulative. For exam- distress, which diminishes parents' capacity
ple, Patterson, Vaden, and Kupersmidt for sensitive, involved parenting, which in
(1991) found that tbe more chronic family turn leads to impaired socioemotional func-
adversities cbildren experienced (e.g., low tioning among children. Living in poverty
family income, parental divorce), the more may expose parents to a multitude of stress-
likely they were to experience difficulties in ful life events and conditions (e.g., marital
peer relationships. Shaw and Emery (1988) discord, family illness, criminal victimiza-
reported tbat multiple life stressors had cu- tion, and inadequate housing) over which
mulative effects on child behavior problems. they have little or no control. Under condi-
Such cumulative effects bave also been re- tions of poverty and concomitant psychologi-
ported by Rutter (1979; Rutter & Garmezy, cal distress, parents are more likely to use
1983). harsh discipline and physical punishment
and less likely to be supportively and af-
Although it seems likely that the effects fectionately involved witb their children
of persistent economic hardship on children (Gecas, 1979; McLoyd, 1990).
are more negative than those of intermittent
or no economic stress, little information has Harsh discipline appears to be a power-
been available to evaluate this hypothesis. ful mediator between parental psychological
Recently, however, two studies have exam- difficulties and child adjustment problems
ined tbe infiuence of persistence of poverty (Patterson, 1986). In his work on children
on child adjustment. McLeod and Shanahan of the Great Depression, Elder showed that
(1993) found tbat, among 4-8-year-old chil- effects of family economic hardship were
dren, current poverty was related to exter- conveyed to children at least in part through
nalizing symptoms wbereas persistence of increased punitive and arbitrary discipline
poverty was related to internalizing symp- on the part of their fathers (Elder, 1979; El-
toms. Duncan, Brooks-Gunn, and Klebanov der, Nguyen, & Caspi, 1985). Lempers and
(1994) found tbat low family income was as- his colleagues (1989) found that tbe efFect
sociated with internalizing and externalizing of family economic hardship on adolescent
bebavior problems at age 5; persistent eco- depression and delinquency was mediated
nomic hardship added to these negative ef- by inconsistent parental discipline. Our
fects. cross-sectional research bas also confirmed
the role of parental involvement as a mediat-
Even studies that have measured family ing variable between family background and
economic status longitudinally (e.g., Duncan child outcomes (Kupersmidt, DeRosier, Pat-
et al., 1994; McLeod & Shanahan, 1993) terson, & Griesler, 1990). Related findings
have generally assessed the effects of persis- have also been reported by Duncan et al.
tent and intermittent hardship on child out- (1994) and McLeod and Shanahan (1993).
 Bolger etal. 1109
In addition to tbeir involvement in child more likely than white children to grow
care and discipline in the home, parents also up in economically disadvantaged fami-
play a more direct role in preparing their lies (Duncan & Rodgers, 1988). Afriean-
children for success in school and peer envi- American children are also more likely to
ronments by monitoring and managing tbeir grow up in single-parent homes and to live
children's school and peer group activities in neighborhoods cbaracterized by poverty
(Ladd, Profilet, & Hart, 1993). Baker and Ste- (Wilson, 1987). Although a number of stud-
venson (1986) found that parents' manage- ies have reported tbat African-American
ment of tbeir children's scbool careers had children are at higher risk for psychosocial
direct consequences for their educational adjustment problems (Rutter & Garmezy,
achievement. Mothers of low socioeconomic 1983), furtber research is needed to clarify
status knew less about their children's the roles of ethnicity, income, and other fac-
school performance, had less contact with tors. Whether because tbe school environ-
teachers, and were less likely to manage ment requires different skills than those re-
children's school achievement in an ac- quired for success in other community and
tive fashion. Low parental monitoring of family settings they encounter, or because
children's peer activities has been associ- they suffer the effects of stratification based
ated with externalizing bebavior problems on both race and economic status (Ogbu,
in childhood (Wilson, 1980; Patterson & 1990), African-American children growing
Stouthamer-Loeber, 1984) and with lower up in economic hardship may be especially
child competence in classroom peer activi- likely to experience difficulties in school.
ties (Ladd & Goiter, 1988). Low levels of On the other band, poor white children may
parental monitoring are especially likely to experience effects of more negative social
occur in families under stress (McCord, comparisons as a result of living among
1979). Thus, parental bebavior linking the wealthier neighbors (McLeod & Shanaban,
family to tbe scbool and peer environments 1993). Researcb is needed to clarify whether
appears to be an important pathway from a single model of the effects of economic
family poverty to negative child outcomes. hardship can describe the experiences of
both African-American and wbite children.
In addition to family economic hard-
ship, personal qualities of the cbild such as In this context, the present research
gender are important predictors of psycboso- seeks to identify the effects of persis-
cial outcomes during childhood. It is widely tent family economic hardship on personal
agreed tbat greater difficulties are experi- and social development during cbildhood.
enced by boys during childhood (Elder, Drawing on the archives of tbe Charlottes-
1979; Hartup, 1983; Hetherington, Camara, ville Longitudinal Study, a study of psycho-
& Featherman, 1983). It is also believed tbat social risk and resilience among children in
boys may be more vulnerable than girls to a the Charlottesville (Virginia) Public Schools
variety of stressors (Rutter & Garmezy, during the years 1986-1989, the present re-
1983). In response to stressors sucb as family search compares psychosocial change and
confiict, boys are more likely than girls to growth among children from families that
react with aggressive and disruptive behav- experienced significant economic hardship
ior; this aggressive behavior may elicit puni- during all, part, or none of tbe period of time
tive responses from botb parents and peers. under study. Tbe study tbus compares de-
Boys may actually be exposed to more pa- velopment among children growing up in
rental conflict tban girls because parents are families that can be cbaracterized as experi-
more likely to fight in front of them (Hether- encing persistent, intermittent, or no eco-
ington et al., 1983). The association of family nomic hardship during the period of the re-
economic hardship and negative outcomes search.
for children seems to be more pronounced
for boys than for girls (Patterson, Kuper- In the present study, we tested a cumu-
smidt, & Vaden, 1990; Patterson, Vaden, lative risk model of the effects of family eco-
Griesler, & Kupersmidt, 1991). Thus, the ef- nomic hardship on children. We hypothe-
fects of family economic circumstances on sized that children experiencing family
children may depend, at least in part, on economic hardship would exhibit more be-
gender. havior problems, be less well accepted
among their peers, and have lower self-
In tbe United States today, African- esteem than tbeir socioeconomically advan-
American children are particularly likely to taged peers. We also tested the hypothesis
be exposed to the effects of family economic that children who experienced persistent
hardship. African-American children are economic hardship would show the most
1110 Child Developnient
negative effects in each of these domains. conducted with an intervention program that
We bypothesized tbat, within the age range was part of the curriculum of the Charlottes-
of our study, boys who experienced eco- ville Public Schools, Parents received a let-
nomic hardship, especially persistent hard- ter explaining the study and were given the
ship, would show particularly severe diffi- option of notifying the school if they pre-
culties in tbeir behavior, peer relationships, ferred not to have tbeir cbild participate.
and self-esteem. We also tested whether any Fewer than 2% of parents so refused. In
differences in psychosocial adjustment be- 1986, data were collected on 1,154 students,
tween African-American and wbite children who represented over 95% of all children
could be accounted for by greater exposure registered in these grades in the Charlottes-
to family economic hardship among African- ville Public Scbools during the years of the
American children. study. Using similar procedures, data were
collected each year for 3 subsequent years.
In addition to testing tbe direct link be-
tween patterns of economic hardship over Participating Children
time and child outcomes, we also wanted to The participants in the present research
examine the extent to whicb such linkages were 534 students who took part in the CLS
might be mediated by parent behavior. To during each of the four annual assessments.
this end, we tested the hypothesis that the Using criteria described below, each child
effects of persistent and intermittent eco- was classified as having lived in a low-
nomic hardship on child behavior problems, income family during all 4 years of the study
peer acceptance, and self-concept are medi- (persistent family economic hardship); dur-
ated, in part, by maternal involvement in the ing 1, 2, or 3 years of the study (intermittent
child's schooling. family economic hardship); or during none
of the years from 1986 to 1989 (no family
To extend previous cross-sectional re- economic hardship).
search, we wished to examine the ef-
fects of persistent and intermittent family Data for approximately half (575) of the
economic hardship longitudinally. There- original 1,154 subjects were available for all
fore, we tested not only whether children four time points of tbe study. To determine
who experienced persistent and intermit- whether subject attrition was random, we
tent hardship started off behind their peers compared the children who participated at
on measures of psychosocial adjustment, but all four time points (identified in Table 1 as
also whether these children caught up with, "stable") to those who did not participate at
stayed behind, or fell even farther be- one or more time points (identified in Table
hind their wealthier peers over time. Our 1 as "mobile"). As shown in Table 1, wbite
research is the first to examine differences children were more likely than black chil-
in the developmental trajectories from dren to be lost to attrition. Children in the
childhood to early adolescence of children mobile group were less popular witb their
who experienced persistent, intermittent, peers than children in the stable group. The
or no family economic hardship. stable and mobile groups did not differ sig-
nificantly on any other target variables, in-
Method cluding gender, economic hardship, and
self-esteem, as well as teacher ratings of ma-
Source of Data ternal involvement and internalizing and
Data were drawn from the archives externalizing behavior problems. Despite
of the Charlottesville Longitudinal Study some attrition over the 3-year period of the
(CLS) (Kupersmidt et al., 1990; Kupersmidt research, the longitudinal sample was repre-
& Patterson, 1991; Patterson et al., 1992; sentative in many respects of the original
Patterson, Kupersmidt, & Griesler, 1990; group from wbich it was drawn.
Patterson, Kupersmidt, & Vaden, 1990; Pat-
terson, Vaden, Griesler, & Kupersmidt, Several additional criteria were used for
1991; Patterson, Vaden, & Kupersmidt, inclusion in the present study. First, be-
1991), The GLS employed a cohort- cause we wished to examine the role of ma-
longitudinal design to follow a large, hetero- ternal involvement in the child's schooling,
geneous group of public school students we excluded children who did not live witb
over the years 1986—1989, In the spring of their motber during any year of the study.
1986, when the study began, the three co- Based on this criterion 12 children who
horts of children were in second, third, and lived witb a single father and 13 children
fourth grades, and their modal ages were 8, who lived witb neither parent were ex-
9, and 10 years, respectively. The study was cluded from the analyses. We also limited
 Bolger et al. 1111
TABLE 1
COMPARISON OF MOBILE AND STABLE SUBJECT GROUPS ON DEMOGRAPHIC CHARACTERISTICS
AND RISK CRITEMA

Stable Mobile
Children Children Significance
Measures at Time 1 (n = 575) (n = 579) Test
Demographic characteristics:
Percent male 48 46
Percent white 55 67 <=^ = 17.36**
Percent experiencing economic hardship 39 36
Mean scores on risk criteria:
Maternal involvement 3.58 3.67
Externalizing behavior problems 8.42 8.55
Internalizing behavior problems 5.57 5.69
Popularity 09 -.10 l, 1077) = 4.13*
Self-concept 3,58 3.67
* p < .05.
" p < .01.

our sample to children who spent all 4 years economic hardship were strongly related in
of the study in either a single mother house- this sample. For example, during the period
hold or a two parent household. Five chil- of this study, African-American children in
dren who spent part of the time period of this sample were more than four times as
the study in a two parent household and part likely as white children to have experienced
in a single mother household were ex- persistent family economic hardship. Afri-
cluded. Also, among the 575 children who can-American children made up only 20% of
participated in all four waves of data collec- those experiencing no family economic
tion, only 11 were identified as Asian, His- hardship, but they made up 60% of those
panic, or Native American. Because we experiencing intermittent family economic
lacked a sufficiently large sample to general- hardship and 79% of those experiencing per-
ize our findings to children from these eth- sistent family economic hardship.
nic groups, we did not include these chil-
dren in the analyses. Material and Procedures
The numbers of children in the sample In each year of the CLS, core assess-
in each category of family economic hard- ments focused on three areas of children's
ship, gender, and ethnicity are shown in Ta- competence at school: behavior problems,
ble 2. Although the distribution of children peer relations, and self-concept (Patterson,
was uneven across categories, there were Kupersmidt, & Griesler, 1990; Patterson,
sufficient numbers of children in each of the Kupersmidt, & Vaden, 1990). Measures of
major categories to allow analysis. peer relations were drawn from peer nomi-
nations. Group sociometric testing was con-
As expected on the basis of national ducted in each participating classroom by an
figures, ethnicity and persistence of family adult experimenter, with one or two aides.

TABLE 2
NUMBERS Of- CHILDREN IN THE SAMPLE AS A FUNCTION O F GENDER, ETHNici-rY,
AND FAMILY ECONOMIC HARDSHIP

No INTERMII-TENT PERSISTENT
ECONOMIC ECONOMIC ECONOMIC
HARDSHIP HARDSHIP HARDSHIP

White Black White Black White Black Totals

Cirls 117 29 28 40 15 46 275
Boys 99 25 23 35 14 63 259
Totals 216 54 51 75 29 109 534
1112 Child Development
according to the procedures described by the internalizing scale, .64 (.63 and .64 for
Goie, Dodge, and Goppotelli (1982). Ghii- black and white children, respectively).
dren were presented with an alphabetized
hst of children in their grade and were asked Teachers were also asked to answer the
to nominate three children whom they liked question, "How involved is this mother
most and three whom they liked least. God- in her child's educational development?"
ing of the sociometric data was accom- Teachers rated mother involvement using a
plished using the criteria and procedures de- five-point scale; higher scores indicated
veloped by Goie and his colleagues (1982). greater involvement (for further details, see
As a measure of children's acceptance by Kupersmidt et al., 1990). These ratings were
their peers, liked-most and liked-least scores relatively stable over time (mean test-retest
were computed to derive values for social correlation = .46) and so were averaged
preference (the difference between liked- across time to create the mother involve-
most and liked-least nominations). Because ment scores used in the present study. Previ-
we wanted to preserve any changes in vari- ous research by Stevenson and Baker (1987;
ance over time, we did not standardize these Baker & Stevenson, 1986) has shown that
scores. teacher ratings similar to the item we used
predict child performance in school, even
Following the completion of the socio- after controlling for maternal education and
metric testing, each child was asked to com- child's age and gender.
plete the Self-Perception Profile (Harter,
1985), a self-report instrument designed to Interviewers were white; therefore,
assess children's perceptions of their own race of the interviewer was unfortunately
competence in various domains. We used not matched with race of the participant for
the subscale for global self-worth, which as- African-American children and teachers.
sessed the extent to which children felt good Median correlations among variables
about themselves and were happy with how are shown in Table 3.
they were leading their lives (Harter, 1985).
Mean Gronbach's alpha of the scale across Categorization of Family Income Levels
all time points of the study was .76 (.71 and Ghiidren were identified as coming
.79 for black and wbite children, respec- from a low-income family in any year that
tively). they were receiving federally mandated free
or reduced price school lunches. To be eligi-
Each teacher was individually inter- ble for free lunches, children's family in-
viewed in a separate room while the socio- comes could not exceed 130% of the feder-
metric testing was conducted in his or her ally mandated poverty level; to be eligible
classroom each year. For the teacher in- for reduced price lunches, children's family
terview, the interviewer read each item incomes could not exceed 185% of the pov-
aloud and recorded the teacher's response. erty level. During the period of this study,
Teacher ratings of behavior problems for the poverty line for a family of three ranged
each child were collected at this time. As- from $9,120 to $10,060 per year; the same
sessments of behavior problems were made cutoff for free lunches ranged from $11,856
by classroom teachers for each child, using to $13,078 and for reduced price lunches
a standardized rating scale adapted from the from $16,872 to $18,611 per year. Our crite-
Glassroom Adjustment Rating Scales devel- rion thus identified not only children living
oped by Gowen and his colleagues (Gowen, below the poverty line but also those living
Lorion, & Galdwell, 1975; Patterson, Kuper- just above it.
smidt, & Vaden, 1990). This scale, devel-
oped as a screening instrument to identify Use of this criterion has a number of ad-
adjustment problems among elementary vantages. First, receipt of free or reduced
school children, has the advantage of being price lunches represents an ecologically sa-
relatively brief and hence feasible for a lient indicator of the economic dimensions
teacher to complete for every child in a class. of children's lives in their school environ-
Factor analysis yielded summary scores for ments. Second, the criterion represents fed-
internalizing (i.e., shy, anxious) and for ex- erally mandated income cut-off points, just
ternalizing (e.g., aggressive, acting out) be- above poverty levels, that are adjusted for
havior problems (see Patterson, Kupersmidt, family size, and it yields results that are
& Vaden, 1990). Averaged across all time readily comparable with those of research
points of the study, Gronbach's alpha for the conducted in other parts of the country.
externalizing scale was .91 (.92 for black Third, this indicator is available in school
children and .89 for white children) and, for records and could be collected for every
 Bolger etal. 1113
TABLE 3
MEDIAN CORRELATIONS AMONG VARIABLES

1
1. Economic hardship
2. Persistent hardship 03
3. Gender 05 .07
4. Ethnicity 50 .15 .06
5. Maternal involvement -.56 -.12 -.08 -.42
6. Externalizing 24 .08 .30 .20 -.29
7 Internalizing 14 .08 .12 .06 -.17 .30
8. Popularity -.24 -.04 -.08 -.05 .21 -.30 -.20
9. Self-esteem -.16 -.12 .01 -.10 .22 -.12 -.12 .18

child participating in the study. Thus, chil- lowing three cohorts of children, we were
dren who received public assistance in the able to model change across a 5-year rather
form of free or reduced price lunches in any than a 3-year interval, while at the same time
given year were categorized as having low increasing the power to detect significant
family incomes for that year. differences.
A latent growth curve model is depicted
Results in F^igure 1 (after McArdle & Hamagami,
Growth Curve Models 1992). Observed variables are shown here as
Numerous methods are available for the squares and latent variables as circles. In
study of longitudinal data (Baltes & Nes- this model, an individual's score on the de-
selroade, 1977). One recently developed pendent variable Y at each time point is as-
method (McArdle, Hamagami, Elias, & Rob- sumed to be a linear combination of three
bins, 1991) has combined the latent growth unobserved variables: a Level component, a
curve model (McArdle & Epstein, 1987) and Slope component, and an Error component.
the convergence (or accelerated longitudi- The Level function (labeled £„ in Fig. 1)
nal) approach (Bell, 1953, 1954). The latent embodies differences among individuals; its
growth curve model describes both individ- effect on the dependent variable Y is held
ual differences and patterns of change over constant over time. The Slope function (la-
time (McArdle & Hamagami, 1992). Thus, beled S^) represents change over time; B,
the latent growth curve model allows for es- represents the effect of this Slope compo-
timation not only of mean differences on nent on dependent variable Y, at time t. The
measures of psychosocial adjustment but effect of the Slope function can change over
also of differences in patterns of change over time, in a linear or nonlinear fashion, as a
time, as a function of family economic hard- function of B,. The variances of Level and
ship, gender, and ethnicity. Slope are represented in Figure 1 as L^* and
Sj,*, respectively. A constant (labeled 1 in
Growth curve models have usually re- Fig. 1) is used to scale the other variables.
quired measurement of individuals over The Error component represents error of
long periods of time (McArdle & Hamagami, measurement. Gonsistent with the princi-
1992). However, the convergence approach ples of the classical true score model (see
(Bell, 1954) allows information on individu- Grocker & Algina, 1986; Spearman, 1904),
als from multiple cohorts, measured at over- measurement error is assumed to be random,
lapping time points, to be linked. The con- with a mean of zero and no autocorrelation
vergence approach assumes that the same over time. (For further details on the statisti-
model is invariant across cohorts. This as- cal assumptions of the latent growth curve
sumption was considered appropriate for model, see McArdle & Epstein, 1987, and
this study because few, if any, cohort effects McArdle & Hamagami, 1992.)
are expected for groups of children who are
very close in age. By making this assump- Figure 2 (after McArdle & Hamagami,
tion, we take full advantage of the cohort- 1992) shows a multiple group latent growth
sequential design (Schaie, 1986), which is curve model for our cohort-sequential de-
intended to measure developmental trends sign, with each of three cohorts measured at
while controlling for secular effects. By fol- four overlapping occasions of measurement.
1114 Child Development

0 0
FIG. 1.—Latent growth curve model

for a total of six time points. Each group in 1] , to specify a linear pattern of change over
this model represents a single cohort in our time. In the third model, a latent growth
design. The model parameters are con- model, coefficients B2 through B5 were esti-
strained to be invariant across cohorts. As in mated from the data. Coefficient B^ was fixed
MANOVA, the error variance for the depen- at zero and coefficient Bg ^^^ fixed at one,
dent variable Y is constrained to be equal in order to provide a scale for the other Slope
across time points. Where Yj is unmeasured loadings in this model. In this way, the la-
at a particular time point for a given cohort tent growth model could estimate any non-
(i.e.. Times 5 and 6 for Gohort 1, Times 1 linear pattern of change over time that is
and 6 for Gohort 2, and Times 1 and 2 for present in the data. To evaluate the relative
Gohort 3), Y, is represented as a latent vari- fit of alternative models, we used a likeli-
able, with variance fixed at {NIN - 1), where hood improvement percentage (LIP) (McAr-
N is the sample size for that cohort. dle & Hamagami, 1992), which was calcu-
lated as the ratio of the x^ fit index of the
To determine whether and how chil- linear and latent growth models compared
dren's psychosocial adjustment changed to the no-growth baseline model (computed
over time, we compared three alternative as the x^ fit index for the baseline model mi-
models for each of the four dependent vari- nus x^ for the linear or latent model, divided
ables (externalizing and internalizing behav- by the baseline x^). We chose a less parsimo-
ior problems, popularity with peers, and nious model (i.e., one with fewer degrees of
self-esteem). In the first model, a no-growth freedom) over a more parsimonious model
model, all Slope loadings (Bj) were fixed at only when the likelihood improvement in-
zero. This model specified that children's dex was at least 10%.
scores on the dependent variable were sta-
ble over time; that is, this model specified Before examining patterns of change
individual differences among children but over time, we wanted to verify that each of
no individual differences in change. In the the dependent variables was measuring the
second model, B^ was fixed at [0 .2 .4 .6 .8 same construct at each time point. We there-
 Bolger etal. 1115
fore tested for factor invariance over time by
comparing two models for each dependent
variable. In the first model (the invariant
model), the factor loadings for each item
were constrained to be equal at each time
point. In the second (variant) model, load-
ings were allowed to vary across time. To
the extent that the variant model improves
over the invariant model, the assumption of
measurement invariance is violated. For
three of the four dependent variables, the
variant model did not improve substantially
over the invariant model (externalizing: in-
variant x^ = 2,835 (1,349), variant x^ = 2,661
(1,324), LIP = 6%; popularity: invariant x^
= 1,303 (200), variant x^ = 1,258 (190), LIP
= 3%; self-concept, invariant x^ = 1,368
(1,929), variant x^ = 1,302 (1,902), LIP =
5%). For internalizing behavior problems,
however, the variant model fit substantially
better than the invariant model (invariant x^
= 1,514 (850), variant x^ = 1,352 (830), LIP
= 11%). Examination of the factor loadings
revealed that loadings for the internalizing
items were lower at grade 7 than at other
time points. When this last time point was
dropped from the model, the invariant
model fit well, as compared to the variant
model (invariant x^ = 1,208 (857), variant x^
= 1,133 (841), LIP = 6%). The grade 7 items
for internalizing were therefore excluded
from further analysis.

Table 4 shows the results of compari-

sons among the no slope, linear growth, and
latent growth models. For each dependent
variable, the linear growth model provided
a substantial improvement in fit as compared
to tlie no-grow£b model. The linear model
improved over the no-growth model by 72%
for externalizing behavior problems, 91% for
internalizing behavior problems, 59% for
peer acceptance, and 86% for self-concept.
For externalizing behavior problems, the la-
tent growth model improved substantially
(15%) over the fit of the linear growth model.
For internalizing behavior problems and
Y,,, popularity, the latent growth model im-
proved only slightly (1%) and 4%, respec-
tively) over the more parsimonious linear
model. For self-esteem, the latent growth
model offered no improvement in fit over
the linear model. Thus, children's scores on
FIC. 2.—Multiple cohort growth curye model each of the dependent variables changed
substantially over time. Ghanges in exter-
nalizing behavior problems followed a non-
linear pattern, increasing between grade 2
and grade 3 and decreasing thereafter.^ In-

^ The measure of externalizing problems used here includes disruptive, aggressiye, and
attention-seeking behavior; thus it does not directly tap delinquent or risk-taking behayior of the
1116 Child Development
TABLE 4
GOODNESS OF F I T FOR LATENT GROWTH, LINEAR GROWTH,
AND No SLOPE BASELINE MODELS

Latent Linear No Slope

Externalizing behayior problems:
Likelihood ratio (degrees of freedom) 148(71) 307(75) 1,115(78)
Likelihood improvement % 87 72
Internalizing behavior problems.-
Likelihood ratio (degrees of freedom) 79(51) 89(54) 1,033(57)
Likelihood improvement % 92 91 ...
Popularity:
Likelihood ratio (degrees of freedom) 59(71) 64(75) 159(78)
Likelihood improyement % 63 59 ...
Self-concept:
Likelihood ratio (degrees of freedom) 245(71) 252(75) 1,789(78)
Likelihood improyement % 86 86 ...
NOTE.—Degrees of freedom = LISREL df - 3 (due to fixed unit for each cohort). Likeli-
hood ratio test (LRT) = LISREL x^ value. Likelihood improvement % = [(LRT,, - LRT)/
* 100 where LRTp = baseline.

ternalizing behavior problems, popularity, children who experienced 1 year of hardship

and self-esteem changed in a linear pattern. to those who experienced 2 or 3 years, and
the fourth contrast compared children who
Economic Hardship, Gender, Ethnicity, lived in hardship for 2 years versus 3 years.
and Children's Psychosocial Adjustment We also created effect codes for gender and
The main data analyses were designed ethnicity and for their interactions with fam-
to describe the course of psychosocial devel- ily economic hardship.
opment among cbildren experiencing per-
sistent, intermittent, or no family economic A separate model was estimated for each
hardship over time. To this end, we added of the four main dependent measures (exter-
economic hardship, gender, ethnicity, and nalizing behavior problems, internalizing
their interactions to the multiple cohort behavior problems, popularity with peers,
growth model described above as exogenous and self-esteem). For externalizing behavior
variables affecting Level and Slope. Tests of problems, we specified a latent growth pat-
significance were one-tailed when hypothe- tern, as indicated by the results of the alter-
ses were directional; otherwise, they were native model comparisons. For internalizing
two-tailed. problems, popularity, and self-esteem, we
specified a linear pattern of change over
To compare the effects of intermittent time.
and persistent hardship, we created four or-
thogonal contrast variables: the first contrast Externalizing behavior problems.—
compared children who experienced family Results of the latent growth curve model for
economic hardship, either intermittent or externalizing behavior problems are shown
persistent, to those who experienced no in Table 5. Using the latent growth model
hardship. The second contrast compares parameter estimates from this model, we
children who experienced intermittent hard- plotted estimated developmental functions
ship to those who experienced persistent for externalizing behavior problems as a
hardship. Within the intermittent hardship function of family economic hardship, gen-
group, we compared children who experi- der, and ethnicity. These functions appear
enced 1, 2, and 3 years of economic hard- in Figure 3 (for boys) and Figure 4 (for girls).
ship: the third orthogonal contrast compared Results (see Table 5) revealed main effects

sort that is likely to increase during adolescence. The decline in childhood forms of externalizing
behavior evident in these results may be offset by an increase in adolescent problem behaviors
that are unmeasured in our dataset. However, the measure employed here is developmentally
appropriate for the age range of our sample because attention problems in childhood are a risk
factor for delinquency (Maguin, Loeber, & LeMahieu, 1993).
 Bolger etal. 1117
TABLE 5
LATENT GROWTH MODEL OF EFFECTS OF FAMILY ECONOMIC HARDSHIP, GENDER, AND ETHNICITY ON
EXTERNALIZING BEHAVIOR PROBLEMS
A. EFFECTS OF FAMILY ECONOMIC HARDSHIP, PERSISTENT HARDSHIP, GENDER, AND ETHNICITY

EFFECT ON EFFECT ON
MEAN LEVEL SLOPE

M (SE) (SE) B (SE)

Main effects:
Family economic hardship -.01 (.04) .27** (.06) .20** (.08)
Persistent hardship .02 (.03) .25** (.09) .23** (.11)
1 versus 2-3 years of hardship .10 (.02) .34** (.12) .18 (.15)
Gender 03 (.04) .25** (.08) .25** (.09)
Ethnicity - .11 (.04) .13 (.06) .04 (.08)
Interactions:
Hardship x gender .01 (.04) .00 (.21) .23 (.26)
Persistent hardship x gender .03 (.03) .58** (.22) .06 (.28)
1 versus 2-3 years x gender -.03 (.02) .38 (.24) .10 (.29)
Hardship x ethnicity .19 (.04) .37* (.22) .36 (.27)
Persistent hardship x ethnicity .15 (.03) .53* (.24) .41 (.29)
1 versus 2-3 years x ethnicity .05 (.02) .18 (.24) .24 (.29)

B. PARAMETER ESTIMATES FOR FUNCTION MEANS, SLOPE LOADINGS, AND FUNCTION DEVIATIONS

B (SE)
Function means:
l-> level 3.11 (.08)
1 —f slope .59 (.10)
Slope loadings;
Slope-* Yi 00 (==)
S l o p e s Y2 58 (.18)
Slope-* y3 44 (.16)
S l o p e s Y4 28 (.14)
Slope-^ Y5 02 (.13)
Slope -^ Ye -1.00 (==)
Function deviations:
L* -^ level 94 (.09)
S* -» slope 34 (.18)
L* « S* 1.00 (==)
E,^Y, 1.19 (.02)
NOTE.—The symbol " = = " indicates a fixed parameter.
• p < .05.
**p< .01.

on Level for family economic hardship, R These main effects were qualified by
= .27, p < .01, persistent family economic two significant interactions influencing the
hardship, B = .25, p < .01, 1 versus 2-3 Level function. Gender interacted with per-
years of hardship, B = .34, p < .01, gender, sistent family economic hardship, B = .58,
B = .25, p < .01. Children who experienced p < .01. As shown in Figure 3 (for boys) and
1 year of family economic hardship showed Figure 4 (for girls), gender differences in ex-
greater externalizing behavior problems ternalizing behavior problem scores were
than did those experiencing no family eco- greatest for children experiencing persistent
nomic hardship. Children with 2 versus 3 hardship. For girls, differences in externaliz-
years of hardship did not differ significantly ing problems between those who did and
from one another, B = - .09, but fared worse did not experience persistent family eco-
than those with 1 year of hardship. Children nomic disadvantage were relatively small.
who experienced persistent hardship had For boys, however, externalizing behavior
the most difficulties. Boys had higher scores problems were greater among those experi-
for externalizing problems than did girls. encing persistent family economic hardship
1118 Child Development
55r

50

,45

I 40

LU

§35

25 Persistent hardship
2 - 3 years ot hardship x
1 year of hardship +

20- No hardship o
White
Black

15
4
Grade

FIG. 3.—Externalizing behayior problems as a function of grade, family economic hardship, and
race for boys.

than among those who experienced no hard- and decreased more in the later years, as
ship or intermittent hardship. Thus, the con- compared to children who experienced no
nection of economic hardship, especially economic hjirdship. The main effect of gen-
persistent economic hardship, with chil- der on Slope was also significant, B = .25,
dren's externalizing behavior in this sample p < .01. As shown in Figures 3 and 4, girls'
was more important for boys. levels of externalizing behavior problems
were relatively stable across time, wbereas
There was also a significant interaction boys' externalizing behavior problems in-
of persistent family economic hardship and creased more in the first year of the study
ethnicity, B = - . 5 3 , p < .05. Among chil- and decreased more in subsequent years.
dren experiencing no economic hardship
or intermittent hardship, African-American
children received higher externalizing Internalizing behavior problems.—
scores than white children. However, among Results of the linear growtb model for inter-
children experiencing persistent family eco- nalizing behavior problems (shown in Table
nomic hardship, white children scored 6) revealed main effects on Level for family
higher on externalizing problems than Afri- economic hardship, B = .20, p < . 01, and
can-American children. persistent family economic hardship, B =
.24, p < .01. Children who experienced per-
The Slope function for externalizing be- sistent family economic hardship had higher
havior problems was infiuenced by a main scores for internalizing behavior problems
efFect of family economic hardship, B = .20, than did those who experienced intermittent
p < .01, by persistent hardship, B = .23, p hardship, who in turn had higher scores for
< .01. Among children who experienced internalizing problems than those who did
economic hardship, especially persistent not experience family economic problems.
hardship, externalizing behavior problems Children in the intermittent group did not
increased more in the first year of the study differ significantly from each other as a func-
 Bolder etal. 1119
55

Persistent hardship
2 • 3 years of hardship x
50
1 year of hardship +
No hardship o
White
,45
Biaci<

1*0
I
UJ
• -X

.£
£35

30

25

20

15
4
Grade

FIG. 4.—Externalizing behavior problems as a function of grade, family economic hardship, and
race for girls.

tion of number of years of economic hard- — .53, p < .01, and for persistent family eco-
ship (1 vs. 2—3 years: Level, B = .01, Slope, nomic hardship, B = —.24, p < .05. Chil-
B = - . 1 3 ; 2 vs. 3 years: Level B = .04, dren who experienced no family economic
Slope B = -.31). Cender affected Slope, B hardship were more popular among their
= .16, p < .05. Over time, internalizing peers than were children who experienced
problems decreased among girls but in- intermittent family economic difficulties.
creased among boys. Children who experienced persistent family
These main effects were qualified by a economic hardship were the least popular.
significant effect of persistent hardship x Children in the intermittent group did not
ethnicity affecting both Level, B - - .45, p differ significantly from one another based
< .01, and Slope, B = .56, p < .01. As shown on their number of years of economic hard-
in Figures 5 (for boys) and 6 (for girls), ship (1 vs. 2-3 years: Level, B = .02, Slope,
among children who experienced no family B = -.24; 2 vs. 3 years: Level, B = - . 2 3 ,
economic hardship or intermittent hardship, Slope, B = .13).
there were no significant differences in the
internalizing scores of African-American The main effect of family economic
versus white children. However, among hardship on Slope was also significant, B =
those experiencing persistent family eco- .29, p < .05. As Figure 7 shows, children
nomic hardship, white cbildren's scores de- who experienced no economic hardship be-
clined over time while African-American came less popular over time, while those
children's scores increased. who experienced economic hardship be-
came more popular. Nevertheless, at every
Peer relations.—Results of the linear time point, children who experienced eco-
growth model for acceptance among peers nomic hardship were less popular than those
(shown in Table 7) revealed main effects on who did not and children who experienced
Level for family economic hardship, B = persistent hardship were least popular.
1120 Child Development
TABLE 6
LiNEAH GROWTH MODEL OF EFFECTS OF FAMILY ECONOMIC HARDSHIP, GENDER, AND ETHNICITY ON
INTERNALIZING BEHAVIOR PROBLEMS
A. EFFECTS OF FAMILY ECONOMIC HARDSHIP, PERSISTENT HARDSHIP, GENDER, AND ETHNICITY

E F F E C T ON EFFECT O N
MEAN LEVEL SLOPE

M (SE) B (SE) B (SE)

Main effects:
Family economic hardship --.01
.01 (.05) 20** (.07) -.04 (.10)
Persistent hardship .02 (.03) .24** (.09) -.18 (.14)
Gender -.03 (.05) .09 (.08) .16* (.09)
Ethnicity -.11 (.04) ,10 (.10) -.09 (.10)
Interactions:
Hardship x gender .01 (.03) .05 (.14) -.10 (.22)
Persistent hardship x gender .03 (.02) .17 (.16) -.16 (.25)
Hardship x ethnicity .19 (.03) .02 (.05) .05 (.24)
Persistent hardship x ethnicity .15 (.02) - .45** (.18) .56* (.28)

B. PARAMETER ESTIMATES FOR FUNCTION MEANS, SLOPE LOADINGS, AND FUNCTION DEVIATIONS

B (SE)
Function means:
1 -^ level 2.78 (.07)
1 -* slope 00 (.10)
Slope loadings:
Slope-^ Yi 00 (==)
Slope-* Yj 25 (==)
S l o p e s Y3 50 (==)
Slope-^ Y4 75 (==)
Slope -^ Y5 1.00 (==)
Function deviations:
L* -^ level 52 (.09)
S* -^ slope 79 (.16)
L* <-» S* -.38 (.22)
E , ^ Yt 94 (.02)
NOTE.—The symbol = = " indicates a fixed parameter.
• p < .05.
**p < .01.

Self-esteem.—Results of the linear experiencing differing levels of poverty, we

growth model for self-esteem scores yielded also wanted to examine factors that might
main effects on Level for family economic link family economic hardship to child out-
hardship, B = - .22, p < .05, and persistent comes. To test the hypothesis that parental
family economic hardship, B = — .42, p < behavior links family economic hardship to
.01 (see Table 8). Children in the intermit- child outcomes, we adapted the three-part
tent group did not differ significantly from data analysis strategy advocated by Baron
each other as a function of number of years and Kenny (1986) to establish mediation.
of economic hardship (1 vs. 2 - 3 years: First, according to this strategy, the indepen-
Level, B = - 0 6 , Slope, B = - . 1 4 ; 2 vs. 3 dent variable must affect the proposed medi-
years: Level, B = .03, Slope, B = -.22). As ator. Second, the independent variable must
illustrated in Figure 8, children who experi- affect the dependent variable. Third, when
enced family economic hardship reported the effect of the mediator on the dependent
lower self-esteem than other children, and variable is added to the model, the mediator
those who experienced persistent hardship must affect the dependent variable, and the
were lower than any others. direct effect of the independent variable on
the dependent variable must be reduced.
Role of Maternal Involvement For each dependent variable, we tested for
In addition to assessing the differences each of these relationships by adding the
in psychosocial outcomes among children specified variables and effects to the multi-
 Bolger etal. 1121
38r

36

Persistent •

34 Intermittent X
£
No hardship 0

Black __

White _

30

26
-o— •
-O

24

4
Grade

FIG. 5.—Internalizing behavior problems as a function of grade, family economic hardship, and
race for boys.

pie cohort growth model. We included all of the variance in externalizing behavior
effects of economic hardship that had been problems, 31% of the variance in self-
shown, in our previous models, to infiuence esteem, and 14% of the variance in popular-
Level and Slope significantly. To assess the ity accounted for by the main and interac-
relative importance of the mediated path, we tive effects of persistent and intermittent
used path analysis (Loehlin, 1987; Wright, hardship.
1920) to compute, for each dependent vari-
able, the proportion of variance accounted Discussion
for by intermittent and persistent hardship,
along with their interactions with gender Considerable interest has focused in re-
and ethnicity, that was mediated by mater- cent years on the distinction between persis-
nal involvement. tent and intermittent economic hardship and
on the implications of both for child devel-
For internalizing behavior problems, opment (Duncan & Rodgers, 1988; Huston,
maternal involvement failed to predict 1991; McLoyd, 1990). Because most re-
Level (B = - .09) or Slope (B = - .08) sig- search to date has been cross-sectional in na-
nificantly. We therefore concluded that ma- ture, however, it has not allowed a full un-
ternal involvement did not mediate the ef- derstanding of the effects of family economic
fects of family economic hardship on hardship on children over time (Featherman
internalizing problems. Results of the medi- et al., 1988). Even previous research that has
ation models for externalizing, popularity, measured family economic hardship over
and self-esteem appear in Table 9. Table 9 time has not examined concomitant changes
shows the effects of economic hardship vari- in child outcomes over the same period. By
ables on matemal involvement, and the di- studying botb family economic hardship and
rect effects of hardship on Level and Slope child outcomes over time, the present study
after maternal involvement is added to the described associations between persistent
model. Maternal involvement mediated 34% and intermittent family economic hardship
1122 ChildDevelopment

Persistent •

Intermittent X

36 No hardship 0

Biack
White _
34

S32

30-

i
1
26-

24-

22
4
Grade

FIG. 6.—Internalizing behavior problems as a function of grade, family economic hardship, and
race for girls.

and difFerent aspects of children's school- hardship. Our results revealed that children
based competence over time. Combining exposed to persistent economic hardship
the latent growth curve and convergence were less well adjusted across a range of as-
models with a cohort-sequential design al- sessments collected in the school envi-
lowed us to examine both individual differ- ronment.
ences and change over time as a function of
family economic hardship, while reducing Because family economic circumstances
the infiuence of possible cohort effects on and ethnicity were themselves interrelated,
our results. we wanted to separate the effects of each
variable. We found that negative outcomes
Our results document a remarkable associated for children with persistent fam-
range of difficulties encountered at school ily economic hardship in our sample oc-
by children whose families experienced per- curred even after evaluating statistically the
sistent economic hardship. Children who ex- separate and interactive contributions of^ eth-
perienced persistent family economic hard- nicity. Thus, aspects of economic status over
ship started out behind other children on and above ethnicity were linked most
every measure of school-based competence strongly to child outcomes in our sample.
we collected and generally stayed behind Patterns of growth and decline across the
throughout all the years of the study. Over- years of the study were, however, moderated
all, the greatest difficulties in adjustment by both economic hardship and ethnicity.
were shown by children whose families ex-
perienced persistent economic disadvan- Consistent with findings from a large
tage, followed by those whose families expe- body of research, our findings revealed that
rienced intermittent economic hardship; the boys were more affected than girls by family
fewest difficulties were shown by those economic hardship, at least in terms of exter-
whose families did not undergo economic nalizing behavior problems (Elder, 1979;
 Bolger etal. 1123
TABLE 7
LINEAR GROWTH MODEL OF EFFECTS OF FAMILY ECONOMIC HARDSHIP, GENDER,
AND ETHNICITY ON POPULARITY
A. EFFECTS OF FAMILY ECONOMIC HARDSHIP, PERSISTENT HARDSHIP, GENDER, AND ETHNICITY

EFFECT ON EFFECT ON
MEAN LEVEL SLOPE

M (SE) B (SE) (SE)

Main effects:
Family economic hardship -.02 (.04) .53** (.09) .29* (.12)
Persistent hardship .01 (.03) .24* (.12) .18 (.17)
Gender -.04 (.04) .10 (.10) .04 (.15)
Ethnicity -.09 (.04) .12 (.13) .13 (.18)
Interactions:
Hardship x gender .05 (.03) .06 (.19) .09 (.26)
Persistent hardship x gender .05 (.02) .07 (.21) .04 (.30)
Hardship x ethnicity .51 (.03) .03 (.20) .22 (.28)
Persistent hardship x ethnicity .11 (.02) .44 (.25) .66 (.35)

B. PARAMETER ESTIMATES FOR FUNCTION MEANS, SLOPE LOADINGS, AND FUNCTION DEVIATIONS

B (SE)
Function means:
1 -* level 06 (.09)
1 -^ slope .07 (.12)
Slope loadings:
Slope ^ Y i 0 (==)
Slope ^ Y j 2 (==)
Slope ^ Y 3 4 (==)
Slope-^ Y4 6 (==)
Slope-^Yj 8 (==)
S l o p e s Yg 1.0 (==)
Function deviations:
L* -^ level 1.14 (.08)
S* -* slope 85 (.23)
L* « S* -.47 (.11)
£ , - ^ Y, 1.02 (.02)
NOTE.—The symbol "= =" indicates afixedparameter.
• p < .05.
**p< .01.

Hartup, 1983; Hetherington et al., 1983; Pat- findings from cross-sectional analyses (Ku-
terson, Kupersmidt, & Vaden, 1990; Pat- persmidt et al., 1990), we also found that the
terson, Vaden, Griesler, & Kupersmidt, linkage of economic hardship and children's
1991; Rutter & Garmezy, 1983). We found school-based competence was mediated in
that the association between persistent fam- part by parental involvement. We found that
ily economic hardship and externalizing be- scores for matemal involvement accounted
havior problems was greater for boys than for a significant portion of the linkages be-
for girls. This result held true for both white tween family economic hardship and aspects
and African-American boys. To the extent of children's competence at school. When
that gender difFerences emerged in response we controlled statistically for matemal
to persistent economic hardship, boys had involvement in children's educational prog-
more difficulties than girls. ress, the significant effects of economic vari-
ables on externalizing behavior problems,
In accord with suggestions from a num- peer relations, and self-esteem were attenu-
ber of investigators (Conger et al., 1992; El- ated. Thus, in the present case as in others
der, 1979; Elder et al., 1992; Lempers et al., (e.g.. Conger et al., 1992; Duncan et al.,
1989; McLoyd, 1990) and with our own 1994; Lempers et al., 1989; McLeod &
1124 Child Development

0.8

0.6

0.4

0.2

3
O--0.4

-0.6
No hardship o
Intermittent x
-0.8
Persistent

2 3 4 5 6 7
Grade
FIG. 7.—Popularity as a function of grade and family economic hardship

Shanahan, 1993), the impact of economic negative effects of family economic hardship
hardship on children appeared to be filtered, on children ought to be directed toward
at least in part, through the family. changing parental behavior. However, such
interpretations of our data should be viewed
Several caveats apply to our findings with caution for at least two reasons. First,
about mediation of the effects of economic most of the effect of economic hardship on
hardship by maternal involvement. First, child outcomes was unaccounted for by ma-
our measure of maternal involvement was ternal involvement in children's schooling.
based on teachers' response to a single item; Second, the mediation model we tested
these same teachers also rated children's specifies that maternal involvement infiu-
classroom behavior. Although our results are ences child outcomes, but also that maternal
consistent with those from studies that have involvement itself is affected by economic
used maternal self-reports and observer rat- hardship. In theory, the causal chain be-
ings to assess parental behavior (e.g., Dun- tween family economic hardship and child
can et al., 1994; McLeod & Shanahan, 1993), outcomes could be weakened or broken at
scores on maternal involvement may be af- either of these links. Thus our findings about
fected by reporter bias and limitations on re- mediation by maternal behavior do not ar-
liability. Also, our assessment was limited gue against the potential importance of in-
to mothers' behavior. Other family variables, terventions aimed at direct improvement of
including fathers' role in children's educa- family economic well-being.
tional development and other family pro-
cesses in the home, are clearly important in
children's development. Although the present results are fully
consistent with family mediation models
Based on our findings about mediation such as those proposed by Conger and his
by maternal involvement, it may be tempt- colleagues (1992) and by McLoyd (1990),
ing to argue that interventions to reduce the they by no means rule out other interpreta-
 Bolger etal. 1125
TABLE 8
LINEAR GROWTH MODEL OF EFFECTS OF FAMILY ECONOMIC HARDSHIP, GENDER,
AND ETHNICITY ON SELF-CONCEPT
A. EFFECTS OF FAMILY ECONOMIC HARDSHIP, PERSISTENT HARDSHIP, GENDER, AND ETHNICITY

EFFECT ON EFFECT ON
MEAN LEVEL SLOPE

M (SE) (SE) (SE)

Main effects:
Family economic hardship -.08 (.06) .22* (.12) .19 (.20)
Persistent hardship 00 1.04) .42** (.17) .04 (.28)
Gender -.12 (.06) .06 (.13) .17 (.22)
Ethnicity -.13 (.06) .15 (.17) .16 (.27)
Interactions:
Hardship x gender -.07 (.04) .28 (.25) .54 (.41)
Persistent hardship x gender -.02 (.03) .03 (.30) .59 (.49)
Hardship x ethnicity .20 (.04) .05 (.27) .08 (.44)
Persistent hardship x ethnicity .15 (.03) .11 (.34) .15 (.56)

B. PARAMETER ESTIMATES FOR FUNCTION MEANS, SLOPE LOADINGS, AND FUNCTION DEVIATIONS

B (SE)
Function means:
1 -^ level 7.81 (.05)
1 -^ slope .36 (.04)
Slope loadings:
Slope -^. Yi 0 (==)
Slope -^ Y2 2 (==)
Slope -^ Y3 .4 (==)
S l o p e s Y4 6
Slope ^ Y5 8
Slope ->• Ye 1.0
Function deviations:
L*-^ level 83 (.14)
S* -* slope 98 (.35)
L* « S* -.10 (.40)
£ , - * Y, 1.20 (.03)
NOTE.—The symbol "= =" indicates afixedparameter.
* p < .05.
•*p < .01.

tions. Low maternal involvement might be in the linkages between economic hardship,
a result of economic hardship and its se- parental behavior, and child outcomes is an
quelae, but its origins might also be else- important task for future research.
where. For instance, Zill, Moore, Nord, and
Stief (1991) found that most long-term recip- From a methodological standpoint,
ients of welfare scored low on tests of intelli- some concerns about our criterion for family
gence and that most had not graduated from economic hardship might be noted. Despite
high school. Maternal involvement scores its many strengths, our criterion might have
are predicted by mothers' own levels of edu- omitted children from some low-income
cational attainment, which are in turn pre- families who were eligible for but did not
dicted by intelligence test scores (Stevenson participate in the subsidized school lunch
& Baker, 1987). In short, maternal character- program. Inclusion of low-income children
istics may contribute to as well as result from in our no hardship group, however, would
family economic circumstances, and the in- have worked against botii our expectations
fluence processes are likely to be complex. and the main trends of our results. To the
Further specification of processes involved extent that this occurred, the present study
 85 r

:80

ti
UI

•5
;75-

No hardship 0
Intermitteni x
Persistent

70
2 3 4 5 6 7
Grade
FIG. 8.—Self-esteem as a function of grade and family economic hardship

TABLE 9
MEDIATION BY MATERNAL INVOLVEMENT OF THE EFFECTS OF FAMILY ECONOMIC HARDSHIP ON
GHILDREN'S EXTERNALIZING BEHAVIOR PROBLEMS, POPULARITY, AND SELF-ESTEEM

EFFECT ON EFFECT ON EFFECT ON

INVOLVEMENT LEVEL SLOPE

B (SE) (SE) (SE)

Externalizing:
Matemal involvement .23** (.02) .04* (.02)
Hardship -.61** (.14) .17** (.06) .18** (.08)
Persistent hardship -.54** (.22) .20** (.09) .25** (.11)
1 versus 2-3 years hardship 93* (.30) .26* (.12)
Persistent X gender 07 (.19) .19** (.08)
Hardship x ethnicity 1.66** (.14) .04 (.07)
Persistent x ethnicity 97** (.21) .21* (.09)
Popularity:
Matemal involvement .26** (.09) .13 (.13)
Hardship - .59** (.04) .40** (.10) .22 (.14)
Persistent hardship -.21** (.06) .21* (.12)
Self-esteem:
Matemal involvement . .. .25* (.13) .21 (.21)
Hardship 61* (.05) .09 (.14)
Persistent hardship -.22** (.07) .37** (.17)
• p < .05.
•*p < .01.
 Bolger etal. 1127
provided a conservative test of our main ex- References
pectations. The fact that our hypotheses Auletta, K. (1982). The underclass. New York:
were confirmed despite the conservatism in- Random House.
herent in our methods adds to confidence in Baker, D. P., & Stevenson, D. L. (1986). Mothers'
the results. strategies for children's school achievement:
Managing the transition to high school. Soct-
Another important issue of interpreta-
ology of Education, 59, 156-166.
tion rests on a distinction between the extent
of economic hardship, on the one hand, and Baltes, P. B., & Nesselroade, J. R. (1977). Life-
its persistence over time, on the other. Were span developmental psychology: Introduc-
our results attributable to greater hardship tion to research methods. Hillsdale, NJ:
(e.g., lower family incomes) among families Erlbaum.
coded as experiencing persistent economic Baron, R. M., & Kenny, D. A. (1986). The modera-
hardship, to the persistence of their difficul- tor-mediator distinction in social psychologi-
ties, or to both? Because detailed informa- cal research: Conceptual, strategic, and statis-
tion about family incomes was not available tical considerations. Jourrwl of Personality
to us, we were not able to examine this ques- and Social Psychology, 51, 1173-1182.
tion directly here. In many cases, however, Bell, R. Q. (1953). Convergence: An accelerated
the extent and persistence of family eco- longitudinal approach. Child Development,
nomic difficulties can be expected to go to- 24, 145-152.
gether (Duncan, 1991). Indeed, the idea of Bell, R. Q. (1954). An experimental test of the ac-
an underclass was proposed to describe the celerated longitudinal approach. Child Devel-
existence of a group characterized by sig- opment, 25, 281-286.
nificant economic difficulties extending over Bell, R. Q., & Pearl, D. (1982). Psychosocial
substantial periods of time (Auletta, 1982; change in risk groups: Implications for early
Wilson, 1987). In the present study, it seems identification. Prevention in Human Services,
likely that both the extent and persistence 1, 45-59.
of economic pressures are important factors Clark, R. M. (1983). Family life and school
underlying our results. To address this issue achievement. Chicago: University of Chicago
more directly, however, future research Press.
should include more detailed measures of Coie, J. D., Dodge, K. A., & Coppotelli, H. (1982).
family income (see Elder et al., 1992). Dimensions and types of social status in the
school: A cross-age comparison. Develop-
In summary, our findings provide an im- mental Psychology, 18, 557-570.
portant assessment of the psychosocial out- Conger, R. D., Conger, K. J., Elder, G. H., Lorenz,
comes associated with persistent family eco- F. O., Simons, R. L., & Whitheck, L. B. (1992).
nomic hardship for children. Whereas much A family process model of economic hardship
previous research in this area has been and adjustment of early adolescent boys.
cross-sectional in nature, has focused pri- Child Development, 63, 526-541.
marily on white children, has assessed the Cowen, E. L., Lorion, R. P., & Caldwell, R. A.
impact of episodic rather than chronic eco- (1975). Nonprofessionals' judgements about
nomic hardship (Duncan, 1991; McLoyd, clinical interaction problems./oumaZ of Con-
1989, 1990), and has measured child out- sulting and Clinical Psychology, 43,619-625.
comes at only one time point, the present Crocker, L. M., & Algina, J. (1986). Introduction
study evaluated outcomes associated with to classical and modem test theory (pp. 105-
persistent economic hardship among a het- 130). New York: Holt, Rinehart, & Winston.
erogeneous group of children over time. Re- Duncan, C. J. (1991). The economic environment
sults showed that for both white and African- of childhood. In A. C. Huston (Ed.), Children
American children a broad range of negative in poverty: Child development and public
outcomes was associated with enduring eco- policy (pp. 23-50). New York: Cambridge
nomic hardship. These linkages were par- University Press.
tially accounted for by variations in maternal Duncan, G. J., Brooks-Gunn, J., & Klebanov, P. K.
involvement among the groups. Consistent (1994). Economic deprivation and early child-
with suggestions made by a number of in- hood development. Child Development, 65,
vestigators (Clark, 1983; Conger et al., 1992; 296-318.
Elder, 1979; McLoyd, 1990), the association Duncan, G. J., with Coe, R. D., Corcoran, M. E.,
of economic hardship and children's school- Hill, M. S., Hoffman, S. D., & Morgan, J. N.
based competence was apparently mediated (1984). Years of poverty, years of plenty. Ann
in part by processes within the family. Arbor, MI: Institute for Social Research.
1128 ChildDevelopment
Duncan, G. J., & Rodgers, W. L. (1988). Longitudi- self-reported behavior problems in preadoles-
nal aspects of childhood poverty. Journal of cence. Journal of Abnormal Child Psychol-
Marriage and the Family, 50, 1007-1021. ogy, 19, 427-449.
Elder, G. H. (1974). Children of the Great Depres- Ladd, C. W., & Goiter, B. S. (1988). Parents' man-
sion. Chicago: University of Chicago Press. agement of preschoolers' peer relations: Is it
Elder, C. H. (1979). Historical change in life pat- related to children's social competence? De-
terns and personality. In P. B. Baltes & O. G. velopmental Psychology, 24, 109-117.
Brim (Eds.), Lifespan development and be- Ladd, G. W., Profilet, S. M., & Hart, C. H. (1993).
havior (Vol. 2, pp. 117-159). New York: Aca- Parents' management of children's peer rela-
demic Press. tions: Facilitating and supervising children's
Elder, G. H., Conger, R. D., Foster, E. M., & Ar- activities in the peer culture. In R. D. Parke &
delt, M. (1992). Families under economic G. W. Ladd (Eds.), Family-peer relationships:
pressure, yourno/ of Family Issues, 13, 5-37. Modes of linkage (pp. 215-253). Hillsdale,
Elder, C. H., Nguyen, T., & Caspi, A. (1985). Link- NJ: Erlbaum.
ing family hardship to children's lives. Child Lempers, J. D., Clark-Lempers, D., & Simons, R.
Development, 56, 361-375. L. (1989). Economic hardship, parenting, and
Featherman, D. L., Spenner, K. 1., & Tsunematsu, distress in adolescence. Child Development,
N. (1988). Class and the socialization of chil- 60, 25-39.
dren: Constancy, change, or irrelevance? In Loehlin, J. C. (1987). Latent variable models: An
E. M. Hetherington, R. M. Lerner, & M. Perl- introduction to factor, path, and structural
mutter (Eds.), Child development in life-span analysis. Hillsdale, NJ: Erlbaum.
perspective (pp. 67-90). Hillsdale, NJ: Maguin, E., Loeber, R., & LeMahieu, P. G. (1993).
Erlbaum. Does the relationship between poor reading
Gecas, V. (1979). The influence of social class on and delinquency hold for males of different
socialization. In W. R. Burr, F. I. Nye, & ages and ethnic groups? Journal of Emotional
1. L. Reiss (Eds.), Contemporary theories and Behavioral Disorders, 1, 88-100.
about the family (Vol. 1, pp. 365-404). New McArdle, J. J., & Epstein, D. (1987). Latent
York: Free Press. growth curves within developmental struc-
Harter, S. (1985). Manual for the Self-Perception tural equation models. Child Development,
Profile for Children. University of Denver, 58, 110-133.
Department of Psychology. McArdle, J. J., & Hamagami, F. (1992). Modeling
Hartup, W. W. (1983). The peer system. In E. M. incomplete longitudinal and cross-sectional
Hetherington (Ed.), P. H. Mussen (Series data using latent growth structural models.
Ed.), Handbook of child psychology: Vol. 4. Experimental Aging Research, 18, 145-166.
Socialization, personality, and social devel- McArdle, J. J., Hamagami, F., Elias, M. F., & Rob-
opment (pp. 103-196). New York: Wiley. bins, M. A. (1991). Structural modeling of
Hetherington, E. M., Camara, C , & Featherman, mixed longitudinal and cross-sectional data.
D. L. (1983). Achievement and intellectual Experimental Aging Research, 17, 29-52.
functioning of children in one-parent house- McCord, J. (1979). Some childrearing antecedents
holds. In J. T. Spence (Ed.), Achievement and of criminal behavior in adult men. Joumal
achievement motives (pp. 205-284). San of Personality and Social Psychology, 9,
Francisco: Freeman. 1477-1486.
Huston, A. (Ed.). (1991). Children in poveriy: McLeod, J. D., & Shanahan, M. J. (1993). Poverty,
Child development and public policy. New parenting, and children's mental health.
York: Cambridge University Press. American Sociological Review, 58, 351-366.
Isralowitz, R., & Singer, M. (1986). Unemploy- McLoyd, V. (1989). Socialization and develop-
ment and its impact on adolescent work val- ment in a changing economy: The efTects of
ues. Adolescence, 21, 145-158. paternal job and income loss on children.
Kohn, M. L. (1963). Social class and parent-child American Psychologist, 44, 293-302.
relationships: An interpretation. American McLoyd, V. (1990). The impact of economic hard-
Joumal of Sociology, 68, 471-480. ship on Black families and children: Psycho-
Kupersmidt, J. B., DeRosier, M. E., Patterson, logical distress, parenting, and socioemo-
C. J., & Griesler, P. C. (1990). Parental tional development. Child Development, 61,
involvement and children s peer relation- 311-346.
ships, behavior, self-concept, and academic Ogbu, J. U. (1990). Overcoming racial barriers to
adjustment. Unpublished manuscript. Uni- equal access. In J. 1. Goodlad & P. Keating
versity of North Carolina at Chapel Hill. (Eds.), Access to knowledge: An agenda for
Kupersmidt, J. B., & Patterson, C. J. (1991). Child- our nation's schools (pp. 59-89). New York:
hood peer rejection, aggression, withdrawal, College Entrance Examination Board.
and perceived competence as predictors of Parker, J., & Asher, S. (1987). Peer acceptance and
 Bolger et al.
later personal adjustment. Psychological Bul- Schaie, K. W. (1986). Beyond calendar definitions
letin, 102, 357-389. of age, time, and cohort: The general develop-
Patterson, C. J., Criesler, P. C , Vaden, N. A., & mental model revisited. Developmental Re-
Kupersmidt, J. B. (1992). Family economic view, 6, 252-277.
circumstances, life transitions, and children's Shaw, D. S., & Emery, R. E. (1988). Chronic fam-
peer relations. In R. D. Parke & C. W. Ladd ily adversity and school-age children's adjust-
(Eds.), Family-peer relationships: Modes ment. Joumal of the American Academy
of linkage (pp. 385-424). Hillsdale, NJ; of Child and Adolescent Psychiatry, 27,
Erlbaum. 200-206.
Patterson, C. J., Kupersmidt, J. B., & Criesler, Slaughter, D. (Ed.). (1988). Black children and
P. C. (1990). Children's perceptions of self poverty: A developmental perspective (New
and of relationships with others as a function Directions in Child Development, No. 42).
of sociometric status. Child Development, 61, San Francisco: Jossey—Bass.
1335-1349. Spearman, C. (1904). The proof and measurement
Patterson, C. J., Kupersmidt, J. B., & Vaden, N. A. of association between two things. American
(1990). Income level, gender, ethnicity, and Joumal of Psychology, 15, 72-101.
household composition as predictors of chil- Stevenson, D. L., & Baker, D. P. (1987). The fam-
dren's school-based competence. Child De- ily-school relation and tlie child's school per-
velopment, 61, 485-494. formance. Child Development, 58, 1348-
Patterson, C. J., Vaden, N. A., Criesler, P. C , & 1357.
Kupersmidt, J. B. (1991). Income level, gen- U.S. Bureau of the Census (1993). Current Popula-
der, ethnicity, and household composition as tion Reports, Series P60-185. Poverty in the
predictors of children's peer companionship United States: 1992. Washington, DC: U.S.
outside of school. Joumal of Applied Devel- Covemment Printing OiHce.
opmental Psychology, 12, 447-465. Werner, E. E. (1989). High-risk children in young
Patterson, C. J., Vaden, N. A., & Kupersmidt, J. B. adulthood: A longitudinal study from birth to
(1991). Family background, recent life events, 32 years. American Journal of Orthopsychia-
and peer rejection during childhood. Journal try, 59, 72-81.
of Social and Personal Relationships, 8, Wilson, H. (1980). Parental supervision: A ne-
347-361. glected aspect of delinc}uency. British Jour-
Patterson, C. R. (1986). Performance models for nal of Criminology, 20, 203-253.
antisocial boys. American Psychologist, 41, Wilson, W. J. (1987). The truly disadvantaged:
432-444. The inner city, the underclass, and public
Patterson, C. R., & Stouthamer-Loeber, M. (1984). policy. Chicago: University of Chicago Press.
The correlation of family management Wright, S. (1920). The relative importance of he-
and delinquency. Child Development, 55, redity and environment in determining the
1299-1307. piebald pattern of guinea pigs. Proceedings
Rutter, M. (1979). Protective factors in children's of the National Academy of Sciences, 6,
responses to stress and disadvantage. In M. 320-332.
Kent & J. Rolf (Eds.), Primary prevention of Zill, N., & Coiro, M. J. (1992). Assessing the condi-
psychopathology (Vol. 3, pp. 49-74). Han- tion of children. Children and Youth Services
over, NH; University Press of New England. Review, 14, 119-136.
Rutter, M., & Carmezy, N. (1983). Developmental Zill, N., Moore, K. A., Nord, C. W., & Stief, T.
psychopathology. In E. M. Hetherington (1991). Welfare mothers as potential employ-
(Ed.), P. H. Mussen (Series Ed.), Handbook ees: A statistical profile based on national
of child psychology: Vol. 4. Socialization, survey data. Washington, DC: Child Trends.
personality and social development (pp. 775- Zill, N., Moore, K. A., Smith, E. W., Stief, T., &
911). New York: Wiley. Coiro, M. J. (1991). The life circumstances
Sameroff, A. J. (1977). Concepts of humanity in and development of children in welfare fami-
primary prevention. In C. W. Albee & J. M. lies: A profile based on national survey data.
Joffee (Eds.), Primary prevention of psycho- Washington, DC: Child Trends.
pathology (Vol. 1, pp. 42-61). Hanover, NH:
University Press of New England.