Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

P SY CH OL OG I C AL S CIE N CE

Research Report

Child-Care History, Classroom


Composition, and Children’s
Functioning in Kindergarten
Julia Dmitrieva,1 Laurence Steinberg,1 and Jay Belsky2
1
Department of Psychology, Temple University, and 2Institute for the Study of Children, Families and Social Issues,
Birkbeck, University of London, London, United Kingdom

ABSTRACT—Previous research indicates that exposure to demic functioning (Belsky et al., 2007; Hill, Brooks-Gunn, &
nonparental child care at an earlier age and for more Waldfogel, 2003; NICHD Early Child Care Research Network,
hours predicts, at the individual level, more aggressive and 2006).
disobedient behavior and greater school preparedness. Work to date has focused on how a child’s personal experi-
The present study extended child-care research by inves- ences with child care affect his or her development. However,
tigating classroom-composition effects, that is, effects of these effects on individual children may aggregate at the
the combined child-care histories of all the children in the classroom level and influence children—even those who have
classroom. Analyses showed that effects at the individual not been in nonparental child care—in ways that cannot be fully
level aggregate at the classroom level: Kindergarten chil- accounted for by their own individual child-care histories (see
dren with limited or no child-care history exhibited more Belsky, 2001; NICHD Early Child Care Research Network,
externalizing behavior (e.g., fighting and arguing) and 2003, 2006). This possibility has not been investigated. In the
higher academic achievement when schooled in class- present study, we attempted to fill this gap in child-care research
rooms with more peers who had extensive child-care histories. by investigating such classroom-composition effects. We asked
whether children in kindergarten classrooms with a large
number of classmates who have extensive histories of child care,
Some 63% of American children under 5 years old receive some especially in centers, manifest more externalizing problems or
form of nonparental child care; in most cases, this care begins higher achievement than would be expected on the sole basis of
during the first year of life, though typically the earliest care is in their personal child-care histories. Our study’s longitudinal
homes, rather than child-care centers (Overturf Johnson, 2005). design also enabled us to investigate whether classroom com-
For more than three decades, the effects of child care have been position produces developmental change. The possibility that
intensively studied and hotly debated (Belsky, 1986; Clarke- classroom-composition effects might emerge, at least with re-
Stewart & Fein, 1983). Evidence indicates clearly that care gard to problem behavior, was suggested by research showing
initiated early in life and experienced for many hours, especially that children’s individual levels of problem behavior are exac-
in child-care centers, is associated with somewhat elevated erbated by their classmates’ aggression (Gifford-Smith, Dodge,
levels of externalizing behavior problems (e.g., aggression and Dishion, & McCord, 2005) and that exposure to many aggressive
disobedience; Bates et al., 1994; Belsky & Eggebeen, 1991; classmates increases the frequency of problem behavior in
Vandell & Corasaniti, 1990), and that these effects are not middle childhood (Bennett, Elliott, & Peters, 2005; Kellam,
simply a function of low-quality care (Belsky et al., 2007; Ling, Merisca, Brown, & Ialongo, 1998).
NICHD Early Child Care Research Network, 2003, 2006).
Research also indicates that high-quality and center-based care
are associated with enhanced linguistic, cognitive, and aca- METHOD

Participants
Address correspondence to Julia Dmitrieva, Department of Psy-
chology, Temple University, 713 Weiss Hall, 1701 N. 13th St., Phil- The participants were selected from the Early Childhood Lon-
adelphia, PA 19122, e-mail: jdmit@temple.edu. gitudinal Study, Kindergarten Cohort (ECLS-K; National Center

1032 Copyright r 2007 Association for Psychological Science Volume 18—Number 12


Julia Dmitrieva, Laurence Steinberg, and Jay Belsky

for Educational Statistics, 2001), a nationally representative TABLE 1


sample of 17,2121 U.S. kindergarteners. To ensure that we had a Demographic Characteristics of the Sample
sufficient number of children within each classroom to estimate
Variable Value
classroom-level effects, we included a classroom in our study
only if at least 50% of its students were in the ECLS-K study. The Age (years) M 5 5.7, SD 5 0.36
Ethnicity (%)
final sample consisted of 3,440 children (50% male, 50% fe-
European American 75
male; mean age 5 5.7 years) enrolled in 282 kindergarten Hispanic American 9
classrooms (see Table 1 for demographic information on the African American 8
sample). Most participants were of European descent, lived with Asian American 5
both biological parents, and had parents with at least some Other, mixed 3
college education. Most of the mothers had been employed at Family composition (%)
some time since their child’s birth. On average, participants Both biological parents 76
Biological mother only 14
entered nonparental care at 20 months of age and spent 25 hr/
Biological mother and stepfather 6
week in child care during the year before kindergarten. Other 5
Half of participants’ kindergarten programs were publicly Parental educational attainment (%)
supported, and the other half were privately funded. Twenty Did not complete high school 4
percent of the schools were located in rural areas, 30% in small High school diploma, general
cities or towns, and 47% in large cities or suburbs of large cities equivalency diploma 19
(this information was missing for 3% of the participants). On Some college or vocational school 35
Bachelor’s degree 23
average, participating schools had 17 kindergarteners per
Graduate or professional training 20
classroom. Total enrollment of the schools ranged from 150 to Mother employed since the child’s birth (%) 75
500 students, and 12% of the students qualified for free lunch. Age at first child care (months) M 5 20, SD 5 19.25
Forty-three percent of the schools had some type of admission Hours/week in child care during the
requirement (e.g., an interview or an aptitude test) for kinder- prekindergarten year M 5 25, SD 5 20.83
garten. In the preceding year, 10% of the schools had experi-
enced problems with students carrying weapons and robberies,
4% had implemented extreme security measures (e.g., security
Procedure and Measures
guards, metal detectors), and 54% had maintained locked
At both Time 1 (fall 1998) and Time 2 (spring 1999), parents
school grounds.
participated in a telephone interview, teachers completed a self-
To explore selection effects, we compared the study sample
administered questionnaire, and each child was assessed di-
and full sample on multiple characteristics: demographic vari-
rectly. Teacher-reported externalizing behavior was measured
ables (age; gender; socioeconomic status, or SES; parental ed-
using an abbreviated version of Gresham and Elliott’s (1990)
ucation; and ethnicity), school variables (private vs. public,
Social Skills Rating System. Five items assessed frequency of
urban vs. rural, percentage of students qualifying for free lunch,
arguing, fighting, getting angry, acting impulsively, and dis-
and class size), and child variables (age at first child-care ex-
turbing classroom activities (4-point scale from 1, never, to 4,
perience, number of hours in child care, externalizing behavior,
very often). Teacher reports and direct assessments of children’s
and academic achievement). Seven of the 13 comparisons were
competence in reading, mathematics, and general knowledge
significant, with moderate-to-large effect sizes. Compared with
were standardized and averaged to measure academic achieve-
participants in the full sample, those in the study sample had
ment. Parental reports provided demographic data that included
higher SES (d 5 0.38) and higher parental educational
family SES (parental education level, parental occupation, and
achievement (d 5 0.34), were less likely to belong to an ethnic
household income), parental marital status, and child-care
minority group (d 5 0.50), and had higher academic achieve-
history (age at first care, in months; hours per week in all kinds of
ment (d 5 0.40). Also, compared with the schools in the full
care and in specifically center care during the year before
sample, those in the study sample were less likely to be public
kindergarten).
schools (d 5 0.91), had slightly smaller class sizes (d 5 0.72),
and had fewer free-lunch students (d 5 0.34). Given these
differences, final models were tested in both the restricted RESULTS
sample and an expanded sample that included classrooms in
which data were available for at least 25% of the children. Our individual-level analyses replicated previous findings.
Analyses controlling for age, gender, family SES, parental
1
marital status, and concurrent child-care hours showed that
The complete kindergarten ECLS-K sample includes 17,212 U.S. children.
Most participants (92%; n 5 15,780) were recruited in the fall of 1998; the rest children placed in child care (any kind or solely center care) for
of the participants were added in the spring of 1999. longer hours (see Fig. 1a) and at earlier ages (see Fig. 1b)

Volume 18—Number 12 1033


Classroom-Composition Effects of Child Care

Fig. 1. Associations between externalizing behavior and academic achievement in kindergarten and (a)
hours in child care and (b) age at first child care. Results are shown separately for all kinds of child care
and care specifically at child-care centers.

manifested significantly more externalizing problems. Effects of vidual-level analyses were used to test classroom-level effects;
center care on academic achievement were curvilinear: More that is, our predictors were the proportion of students placed in
hours in center care generally predicted greater achievement, care before the age of 2 and the proportion of students placed in
although when hours of center care exceeded 30 per week, care for more than 30 hr/week. To avoid multicollinearity, we
longer hours predicted lower achievement (see Fig. 1a). Earlier centered individual-level child-care-exposure variables around
initiation of center care up to the age of 2 predicted poorer classroom means (Kreft, de Leeuw, & Aiken, 1995). For all
achievement, whereas earlier initiation of center care after the subsequent models, gender, age, family SES, and marital status
age of 2 predicted higher achievement (see Fig. 1b). served as covariates for Time 1 and Time 2 outcomes. Because
Using Mplus Version 4.1 (Muthén & Muthén, 1998–2006), we Time 1 outcomes were controlled in evaluating Time 2 out-
created and tested multilevel path models (Heck, 2001) that comes, classroom-composition effects pertaining to Time 2
evaluated individual- and classroom-level effects of child care outcomes reflect collective effects of child-care history on
on externalizing behavior and academic achievement at Times 1 changes in children’s externalizing behavior and academic
and 2. The curvilinear inflection points detected in the indi- achievement across the two measurement occasions.

1034 Volume 18—Number 12


Julia Dmitrieva, Laurence Steinberg, and Jay Belsky

Baseline unconditional models predicting externalizing be- Classroom-level effects were significant for three of the four
havior and academic achievement indicated that variations in models predicting externalizing behavior and for one model
classroom characteristics accounted for 15% (Time 1) and 19% predicting academic achievement. As shown in Table 2, being in
(Time 2) of the variance in externalizing behavior and for 35% a class with more students with high child-care hours (i.e., more
(Time 1) and 31% (Time 2) of the variance in academic than 30 hr/week of any care or of center care specifically) was
achievement. Thus, further investigation of classroom-level ef- associated with more externalizing behavior at Time 1 and with
fects was warranted. Individual- and classroom-level effects of increases in such behavior over time. Being in a class with
each of the four child-care-exposure variables (i.e., age at first more students who had experienced early center care (i.e., be-
care, hours per week in any care, age at first center care, hours fore age 2), but not being in a class with more students who
per week in center care) were tested for each outcome. At the had experienced early care of any type, was associated with
individual level, greater child-care exposure (both general care more externalizing behavior at Time 1 (see Table 2). Just
and center care) was associated with higher externalizing be- as noteworthy, having more classmates with high center-care
havior, and greater center-care exposure was associated with hours was associated with greater increases in academic
higher academic achievement (see Table 2). Individual-level achievement over time. Classroom-level effects accounted for
effects accounted for less than 1% to 3% of the variance in Time 1% to 8% of the variance in outcome measures at Time 1 and
1 outcome measures and from less than 1% to 3% of the variance 2% to 5% of the variance in outcome measures at Time 2
in Time 2 outcome measures (see Table 3). (see Table 3).

TABLE 2
Standardized Coefficients for the Multilevel Path Models of Externalizing Behavior and
Academic Achievement

Predictor variable Time 1 outcome measure Time 2 outcome measure


Model 1: predicting externalizing behavior from hours in any care
Individual-level effects
Hours in any care .13nn .05nn
2
(Hours in any care) .03 .01
Outcome measure at Time 1 .69nn
Classroom-level effects
Proportion with long hours .21n .20n

Model 2: predicting externalizing behavior from age at first center care


Individual-level effects
Age at first center care .14nn .03
(Age at first center care)2 .03 .00
Outcome measure at Time 1 .71nn
Classroom-level effects
Proportion in early care .21n .14

Model 3: predicting externalizing behavior from hours in center care


Individual-level effects
Hours in center care .17nn .05n
(Hours in center care)2 .02 .00
Outcome measure at Time 1 .68nn
Classroom-level effects
Proportion with long hours .29nn .23n

Model 4: predicting academic achievement from hours in center care


Individual-level effects
Hours in center care .09nn .01
(Hours in center care)2 .06n .00
Outcome measure at Time 1 .86nn
Classroom-level effects
Proportion with long hours .11 .19n

Note. Time 1 and Time 2 outcome measures were tested simultaneously in a single model.
n
p < .01. nnp < .001.

Volume 18—Number 12 1035


Classroom-Composition Effects of Child Care

These classroom-composition effects were over and above the The center-care model (see Fig. 3) also revealed individual-
individual effects of child care and indicated that children who level effects of early placement and long hours on Time 1 ex-
experienced limited or no child care nevertheless were affected ternalizing behavior and of long hours on Time 2 externalizing
by their exposure to classmates with early and substantial child- behavior. At the individual level, the model accounted for 8.9%
care experience. None of the individual-level effects of child (Time 1) and 54.1% (Time 2) of the variance in externalizing
care varied across classrooms, which indicates that the indi- behavior and 22.9% (Time 1) and 77.0% (Time 2) of the vari-
vidual-level effects were not moderated by the classroom-level ance in academic achievement. At the classroom level, being
effects. Thus, children in classrooms with extensive child-care schooled in a classroom containing more students with histories
exposure were just as affected by their own individual child-care of long center-care hours was associated with both greater Time
experiences as were children in classrooms with low exposure. 1 externalizing behavior and greater increases in academic
Subsequent analyses modeled the combined effects of age at achievement over time. The model accounted for 10.5% (Time
first placement and hours in care on externalizing behavior and 1) and 67.3% (Time 2) of the classroom-level variance in ex-
academic achievement. Separate models were fit for nonparental ternalizing behavior and 35.4% (Time 2) of the classroom-level
care overall (Fig. 2) and center care (Fig. 3). At an individual variance in academic achievement.
level, children with longer or earlier nonparental care showed In order to explore possible bias due to selection effects, we
more externalizing behavior at Time 1; children who had been performed additional analyses on an expanded sample that in-
placed in child care for longer hours were also more likely to cluded classrooms in which data were available for at least 25%
exhibit increases in externalizing behavior over time (see Fig. of the children. The final models illustrated in Figures 2 and 3
2). Earlier nonparental care was also associated with increases were retested with this expanded sample. All individual- and
in academic achievement over time. The model accounted for classroom-level effects remained significant, although the co-
6.3% (Time 1) and 52% (Time 2) of the individual-level variance efficient for the classroom-level path from center-care hours to
in externalizing behavior and for 22.9% (Time 1) and 77.1% Time 2 academic achievement declined in magnitude (from b 5
(Time 2) of the individual-level variance in academic achieve- .23, prep 5 .96, to b 5 .13, prep 5 .92).
ment. (Stability in Time 1 and Time 2 measures of the same
outcome accounts for the large amount of variance explained at DISCUSSION
the individual level at Time 2.) At the classroom level, having
more classmates with histories of long nonparental care hours These findings indicate that both the beneficial (increased
was associated with greater externalizing behavior at Time 1. achievement) and the adverse (increased problem behavior)
The model accounted for 2.6% (Time 1) and 67.1% (Time 2) of effects of children’s early or extensive experience in nonparental
the variance in externalizing behavior. care are not restricted to the children who experienced this care

TABLE 3
Proportion of Variance Explained by the Child-Care Variables

Models predicting externalizing behavior . . . Model predicting


From hours in From age at From hours in academic achievement
Outcome and effect size measure any care first center care center care from hours in center care
Individual-level variance
Time 1 outcome measure
R2 – Rbaseline 2 .010 .024 .030 .005
d 0.200 0.310 0.350 0.140
Time 2 outcome measure
R2 – Rbaseline 2 .004 .028 .006 —a
d 0.130 0.340 0.160 —

Classroom-level variance
Time 1 outcome measure
R2 .045 .044 .082 .012
d 0.430 0.430 0.600 0.220
Time 2 outcome measure
R2 .040 .019 .052 .034
d 0.410 0.280 0.470 0.380

Note. The baseline model included control variables: age, gender, family socioeconomic status, and parental marital status. For
all effect sizes (d) reported in this table, the prep values were greater than .995.
a
The individual-level child-care-exposure variables did not contribute to a reduction in variance for academic achievement at Time 2.

1036 Volume 18—Number 12


Julia Dmitrieva, Laurence Steinberg, and Jay Belsky

Fig. 2. Unstandardized coefficients and critical ratios (in parentheses) for the multilevel path model of the
relation between nonparental child-care exposure and externalizing behavior and academic achievement at
Times 1 and 2. Model-fit statistics are as follows: w2(6) 5 10.47, p 5 .11; comparative fit index 5 .999; Tucker-
Lewis index 5 .995; root-mean-square error of approximation 5 .016. Nonsignificant paths are not displayed.

directly, but extend to their classmates as well. Moreover, may play a role (Dishion, McCord, & Poulin, 1999), as aggres-
classroom-composition effects are independent of individual sive incidents can lead to increased group-level encouragement
children’s personal child-care histories. That is, being in a of aggression (DeRosier, Cillessen, Coie, & Dodge, 1994),
classroom with a high proportion of students who have extensive and highly aggressive children may be accepted—rather than
child-care histories affects students with little or no early child- rejected—in aggressive classrooms (Stormshak et al., 1999).
care experience just as much as students with extensive child- Teachers’ behavior (e.g., behavior-management strategies) may
care experience. also mediate the relation between individual child-care ex-
Unfortunately, the ECLS-K data did not permit a fine-grained periences and classroom-level effects. Thus, future research
examination of the within-classroom processes that might have should examine whether patterns of peer or student-teacher
been responsible for the observed classroom-composition ef- interaction vary across classrooms as a function of the number of
fects. Previous research suggests that direct peer contagion students with extensive child-care experience. Future research

Volume 18—Number 12 1037


Classroom-Composition Effects of Child Care

Fig. 3. Unstandardized coefficients and critical ratios (in parentheses) for the multilevel path model of the
relation between center-care exposure and externalizing behavior and academic achievement at Times 1 and 2.
Model-fit statistics are as follows: w2(8) 5 5.83, p 5 .67; comparative fit index 5 1.00; Tucker-Lewis index 5
1.00; root-mean-square error of approximation 5 .000. Nonsignificant paths are not displayed.

should also determine whether our findings generalize to eth- care experiences—are affected by child care, both for better
nically diverse populations, as our sample was not very diverse (i.e., increased academic achievement) and for worse (i.e., in-
(75% European American). creased externalizing problems). Future study should determine
Although statistically significant, the classroom-composition whether the classroom-composition effects observed here are
effects identified in the present study were small in magnitude— also seen at other grade levels, and whether the effects are
just as are individual-level effects of child care (see Belsky et amplified or attenuated as children progress through their years
al., 2007; NICHD Early Child Care Research Network, 2006). of schooling.
But the effects identified here reflect just 1 year in each child’s
school career. Even a small effect can have large consequences
REFERENCES
if it accumulates or is compounded over time. The fact that early
and extensive nonparental child care is virtually the norm in the Bates, J.E., Marvinney, D., Kelly, T., Dodge, K.A., Bennett, D.S., &
United States today means that thousands of classrooms—and Pettit, G. (1994). Child care history and kindergarten adjustment.
all of the children in them, irrespective of their personal child- Developmental Psychology, 30, 690–700.

1038 Volume 18—Number 12


Julia Dmitrieva, Laurence Steinberg, and Jay Belsky

Belsky, J. (1986). Infant day care: A cause for concern? Zero to Three, Kellam, S.G., Ling, X., Merisca, R., Brown, C.H., & Ialongo, N. (1998).
6, 1–7. The effect of the level of aggression in the first grade classroom on
Belsky, J. (2001). Emanuel Miller lecture: Developmental risks (still) the course and malleability of aggressive behavior into middle
associated with early child care. Journal of Child Psychology and school. Development and Psychopathology, 10, 165–185.
Psychiatry and Allied Disciplines, 42, 845–859. Kreft, I.G.G., de Leeuw, J., & Aiken, L.S. (1995). The effect of different
Belsky, J., & Eggebeen, D. (1991). Early and extensive maternal forms of centering in hierarchical linear models. Multivariate
employment and young children’s socioemotional development: Behavioral Research, 30, 1–21.
Children of the National Longitudinal Survey of Youth. Journal of Muthén, L.K., & Muthén, B.O. (1998–2006). Mplus user’s guide (4th
Marriage & the Family, 53, 1083–1098. ed.). Los Angeles, CA: Muthén & Muthén.
Belsky, J., Vandell, D.L., Burchinal, M., Clarke-Stewart, A., McCart- National Center for Educational Statistics. (2001). Early Childhood
ney, K., Owen, M.T., & the NICHD Early Child Care Research Longitudinal Study: Kindergarten class of 1998–1999 [Data file].
Network. (2007). Are there long-term effects of early child care? Washington, DC: U.S. Department of Education, Office of Edu-
Child Development, 78, 681–701. cational Research and Improvement. (Distributed by the Inter-
Bennett, P., Elliott, M., & Peters, D. (2005). Classroom and family University Consortium for Political and Social Research, Ann
effects on children’s social and behavioral problems. Elementary Arbor, MI)
School Journal, 105, 461–480. NICHD Early Child Care Research Network. (2003). Does amount of
Clarke-Stewart, A., & Fein, G.G. (1983). Early childhood programs. In time spent in child care predict socioemotional adjustment dur-
M.M. Haith & J.J. Campos (Eds.), Handbook of child psychology: ing the transition to kindergarten? Child Development, 74, 976–
Vol. 2. Infancy and developmental psychology (pp. 917–999). New 1005.
York: John Wiley and Sons. NICHD Early Child Care Research Network. (2006). Child-care effect
DeRosier, M.E., Cillessen, A.H.N., Coie, J.D., & Dodge, K.A. (1994). sizes for the NICHD Study of Early Child Care and Youth De-
Group social context and children’s aggressive behavior. Child velopment. American Psychologist, 61, 99–116.
Development, 65, 1068–1079. Overturf Johnson, J. (2005). Who’s minding the kids? Child care ar-
Dishion, T.J., McCord, J., & Poulin, F. (1999). When interventions rangements: Winter 2002 (Current Population Reports No. 70–
harm: Peer groups and problem behavior. American Psychologist, 101). Washington, DC: U.S. Census Bureau.
54, 755–764. Stormshak, E.A., Bierman, K.L., Brushi, C., Dodge, K.A., Coie, J.D., &
Gifford-Smith, M., Dodge, K.A., Dishion, T.J., & McCord, J. (2005). the Conduct Problems Prevention Research Group. (1999). The
Peer influence in children and adolescents: Crossing the bridge relation between behavior problems and peer preference in
from developmental to intervention science. Journal of Abnormal different classroom contexts. Child Development, 70, 169–182.
Child Psychology, 33, 255–265. Vandell, D.L., & Corasaniti, M.A. (1990). Variations in early child
Gresham, F.M., & Elliott, S.N. (1990). Social Skills Rating System care: Do they predict subsequent social, emotional, and
manual. Circle Pines, MN: American Guidance Service. cognitive differences? Early Childhood Research Quarterly, 5,
Heck, R.H. (2001). Multilevel modeling with SEM. In G.A. Marcou- 555–572.
lides & R.E. Schumacker (Eds.), New developments and tech-
niques in structural equation modeling (pp. 89–127). Mahwah, NJ:
Erlbaum.
Hill, J.L., Brooks-Gunn, J., & Waldfogel, J. (2003). Sustained effects of
high participation in an early intervention for low-birth-weight (RECEIVED 2/16/07; REVISION ACCEPTED 5/22/07;
premature infants. Developmental Psychology, 39, 730–744. FINAL MATERIALS RECEIVED 5/24/07)

Volume 18—Number 12 1039

You might also like