Family Involvement in School and Low-Income Children's Literacy: Longitudinal Associations Between and Within Families

Abstract

Longitudinal data from kindergarten to 5th grade on both family involvement in school and children's literacy performance were examined for an ethnically diverse, low-income sample (N = 281). Within families, increased school involvement predicted improved child literacy. In addition, although there was an achievement gap in average literacy performance between children of more and less educated mothers if family involvement levels were low, this gap was nonexistent if family involvement levels were high. These results add to existing evidence on the value of family involvement in school by demonstrating that increased involvement between kindergarten and 5th grade is associated with increased literacy performance and that high levels of school involvement may have added reward for low-income children with the added risk of low parent education. As such, these results support arguments that family involvement in school should be a central aim of practice and policy solutions to the achievement gap between lower and higher income children. (PsycINFO Database Record (c) 2012 APA, all rights reserved)

Full text

Family Involvement in School and Low-Income Children’s Literacy:
Longitudinal Associations Between and Within Families
Eric Dearing
Boston College
Holly Kreider
Harvard University
Sandra Simpkins
Arizona State University
Heather B. Weiss
Harvard University
Longitudinal data from kindergarten to 5th grade on both family involvement in school and children’s
literacy performance were examined for an ethnically diverse, low-income sample (N⫽281). Within
families, increased school involvement predicted improved child literacy. In addition, although there was
an achievement gap in average literacy performance between children of more and less educated mothers
if family involvement levels were low, this gap was nonexistent if family involvement levels were high.
These results add to existing evidence on the value of family involvement in school by demonstrating that
increased involvement between kindergarten and 5th grade is associated with increased literacy perfor-
mance and that high levels of school involvement may have added reward for low-income children with
the added risk of low parent education. As such, these results support arguments that family involvement
in school should be a central aim of practice and policy solutions to the achievement gap between lower
and higher income children.
Keywords: educational involvement, child literacy, longitudinal methods, low income, maternal educa-
tion
Families’ involvement in their children’s schools is central to
most public efforts aimed at reducing the achievement gap be-
tween children living in low-income families and their wealthier
peers (e.g., U.S. Department of Health and Human Services,
2005). There is, in fact, increasing evidence that high levels of
family involvement in school are associated with high levels of
child achievement (for meta-analytic reviews, see Fan & Chen,
2001; Jeynes, 2003, 2005a). Nonetheless, the empirical knowledge
base on this topic is limited in two regards. First, few studies have
examined the relations between involvement in school and child
achievement longitudinally. Second, few studies have moved be-
yond main effect associations between involvement and achieve-
ment to consider for whom involvement-achievement relations
may be most meaningful.
Family Involvement in School and Child Literacy as
Developmental Processes
Families may be involved in their children’s education in a
variety of ways, including involvement in the home (e.g., help with
homework) and in the school (e.g., attending open houses) as well
as through parent–teacher communication and parent-to-parent
communication. With regard to involvement practices in the
school context, meta-analytic reviews have repeatedly documented
that children whose families are more involved in school (e.g., via
attending parent–teacher conferences and parent meetings, visiting
and volunteering in the classroom, and participating in social
events in the school) display higher levels of achievement than
children whose families are less involved in school (Fan & Chen,
2001; Jeynes, 2003, 2005a). Links between family involvement in
school and children’s literacy are particularly noteworthy given the
developmental importance of literacy skills. Indeed, the academic,
social, and economic sequelae of childhood literacy problems
include an increase in school dropout rates, juvenile delinquency,
and welfare costs (Autor, Levy, & Murnane, 2003; Hart & Risley,
1995; Snow, Burns, & Griffin, 1998).
Although family involvement in school may have some direct
effects on children’s literacy through achievement socialization
processes such as reinforcement and modeling, its benefits may
also be realized indirectly through children’s feelings, attitudes,
and self-efficacy (e.g., Bandura, Barbaranelli, Caprara, & Pas-
torelli, 1996; Dearing, McCartney, Weiss, Kreider, & Simpkins,
2004; Frome & Eccles, 1998; Hoover-Dempsey et al., 2001).
Specifically, family involvement in the school may promote pos-
itive feelings and attitudes towards education as well as increased
self-efficacy, which in turn may then promote literacy perfor-
mance. Through such mechanisms, family involvement in school
may have long-lasting effects on children’s life chances via posi-
tive effects on the development of literacy. There is some evi-
dence, in fact, that high levels of family involvement during the
Eric Dearing, Department of Counseling, Developmental, and Educa-
tional Psychology, Lynch School of Education, Boston College; Holly
Kreider and Heather B. Weiss, Graduate School of Education, Harvard
University; Sandra Simpkins, Department of Family and Human Develop-
ment, Arizona State University.
Correspondence concerning this article should be addressed to Eric
Dearing, Department of Counseling, Developmental, and Educational Psy-
chology, Lynch School of Education, Campion Hall, Boston College,
Chestnut Hill, MA 02467. E-mail: eric.dearing@bc.edu
Journal of Educational Psychology Copyright 2006 by the American Psychological Association
2006, Vol. 98, No. 4, 653–664 0022-0663/06/$12.00 DOI: 10.1037/0022-0663.98.4.653
653

elementary school years are associated with a decreased probabil-
ity of high school dropout and an increased probability of on-time
high school completion (Barnard, 2004).
Empirical work on family involvement in school and children’s
achievement, however, has been limited primarily to cross-
sectional studies. There are particularly few longitudinal studies on
links between family involvement in school and literacy during the
elementary school years, despite the developmental importance of
this time when children build the literacy skills necessary for later
life (Snow et al., 1998). There are also few longitudinal studies of
involvement in school focused on low-income samples. This lim-
itation to the knowledge base is particularly salient when consid-
ering recent work highlighting the need to disentangle the effects
of family involvement from those of socioeconomic status (e.g.,
Jeynes, 2005b). Furthermore, low-income children are at excep-
tional risk to develop literacy problems (e.g., Hart & Risley, 1995;
Snow et al., 1998), and these children are the focus of public policy
aimed at increasing family educational involvement (e.g., U.S.
Department of Health and Human Services, 2005).
It is worth noting, however, that parents’ motivational practices
and educational expectations have been addressed longitudinally,
both during the elementary school years and among low-income
families (e.g., Entwisle & Alexander, 1996; Gottfried, Fleming, &
Gottfried, 1994). Recently, literacy-related practices in the home
environment and communication with teachers have also been
studied longitudinally and in the context of low family income
(Serpell, Baker, & Sonnenschein, 2005). With regard to longitu-
dinal studies of low-income families’ involvement in the school
environment, two studies are notable.
Englund and colleagues (Englund, Luckner, Whaley, & Ege-
land, 2004) used path analysis to examine parents’ involvement in
their children’s schools (e.g., attending parent–teacher confer-
ences) and overall levels of academic achievement at first and third
grades. These authors found that both higher involvement and
higher achievement at first grade predicted higher involvement at
third grade and, in turn, higher achievement at third grade. Izzo
and colleagues (Izzo, Weissberg, Kasprow, & Fendrich, 1999),
also examining involvement in elementary school longitudinally
for low-income families, used residual change analyses (i.e., pre-
dicting later levels of an outcome while controlling for earlier
levels of that outcome) to examine involvement and achievement
between kindergarten and third grade. These authors found that
third grade achievement levels were related to earlier involvement
levels, while controlling for kindergarten achievement.
These results extend cross-sectional work by demonstrating
links over time between family involvement and child achieve-
ment. As Rogosas (1995) has detailed, however, between-waves
path analyses do not address processes of stability and change, per
se. Further, residual change analyses can result in estimates that
are imprecise and unreliable (Rogosa, 1995). In short, although
these methods have strengths (e.g., path analysis provides an
elegant means of simultaneously estimating multiple relations and
pathways of mediation), they are less than ideal if the goal of the
study is to examine developmental patterns of stability and change
over time. Growth modeling is preferable in this regard, primarily
because individual patterns of stability and change, between-
families differences in these patterns, and within-families associ-
ations among time-varying constructs may be simultaneously es-
timated (Raudenbush & Bryk, 2002; Rogosa, 1995; Singer &
Willett, 2003).
The potential contribution of research directly modeling within-
families patterns of stability and change on involvement in school
and literacy achievement is at least threefold. Consider first that
theory highlights the dynamic nature of family involvement in
school, emphasizing involvement as a process that varies over time
within families (Epstein, 1995; Hill & Taylor, 2004; Hoover-
Dempsey & Sandler, 1995, 1997). Families, for example, may
increase involvement following decreases in life demands, im-
provements in relationships with teachers and other school per-
sonnel, and/or increases in invitations and opportunities to become
involved (Henderson & Mapp, 2002; Hoover-Dempsey & Sandler,
1997). Thus, examining within-families patterns of stability and
change in involvement may more fully capture children’s educa-
tional experiences than analytic methods that ignore these
dynamics.
The second contribution of analyzing within-families patterns of
involvement is that such analyses can help determine whether
increased involvement is associated with improved achievement.
Although previous cross-sectional analyses have demonstrated that
lower levels of involvement are associated with lower levels of
achievement, scientists and policymakers must also know whether
change matters. If, for example, involvement increases within a
family, then does child achievement improve? Within-families
analyses of longitudinal data on both family involvement and child
achievement are necessary to answer such questions.
Third, within-families analyses of longitudinal data can help
control for potential omitted variable bias associated with fixed
characteristics of children and their families. Most studies of
family involvement in school are between-families comparisons
(i.e., comparing achievement levels across children with varying
levels of family involvement) using nonexperimental data. Esti-
mated associations between involvement and child achievement in
these types of studies are potentially biased by time-varying or
time-invariant omitted variables. That is, unmeasured characteris-
tics of children, their families, and their environments may be
causally related to both involvement and achievement, and as such,
associations between these two may be spurious.
Within-families estimates, however, are not susceptible to omit-
ted variable bias caused by between-families heterogeneity that is
fixed over time (Angrist & Krueger, 1999; Duncan, Magnuson, &
Ludwig, 2004; Hsaio, 2003). If, for example, changes over time in
involvement are associated with changes over time in child
achievement within families, characteristics of children, their fam-
ilies, and their environments that are stable over time (e.g., genet-
ics) may be ruled out as potential sources of bias. Although
within-families estimates are not a panacea for potential sources of
bias in nonexperimental studies (e.g., like between-families esti-
mates of nonexperimental data, within-families estimates may be
biased by reciprocal causation or time-varying omitted variables),
they are a useful and recommended means of dealing with bias
associated with potentially omitted variables that are time invariant
(Duncan et al., 2004).
Involvement in School Within Context
Despite advances in developmental science emphasizing the
physical and psychosocial milieu in which children’s growth is
654 DEARING, KREIDER, SIMPKINS, AND WEISS

embedded (e.g., Bronfenbrenner & Morris, 1998; Magnusson &
Stattin, 1998), the role of developmental context has received little
attention in studies of family educational involvement beyond
main effect comparisons across social addresses. There is good
evidence that levels of family educational involvement vary, on
average, across social and economic contexts (Bandura et al.,
1996; Fan & Chen, 2001; Hill & Taylor, 2004; Hoover-Dempsey
et al., 2001; Kohl, Lengua, & McMahon, 2000; Lareau, 1989;
Sanders, 1998). Yet, few researchers have considered how devel-
opmental context may also moderate the effects of involvement on
children’s lives.
Given that some children are at greater risk than others for
academic underachievement, it stands to reason that some children
may benefit more than others from family educational involve-
ment. Consider, for example, families that are characterized by
both low income and low parent education. There is overwhelming
evidence that children living in low-income families display lower
levels of academic self-efficacy and achievement relative to other
children (e.g., Bandura et al., 1996; Taylor, Dearing, & McCart-
ney, 2004). Further, within low-income families, low levels of
parental education place children at exceptionally high risk for
academic failure (e.g., Rauh, Parker, Garfinkel, Perry, & Andrews,
2003). Of importance, however, is recent evidence from the sam-
ple we analyze in the present study that educational involvement
may be particularly meaningful for these children, even leading to
achievement levels similar to those of children from more edu-
cated families (Dearing et al., 2004). A focus on interacting pro-
cesses between involvement and parent education is central to the
present study, primarily because answers to these types of ques-
tions can help guide intervention efforts by identifying children
who are most likely to benefit from intervention.
Present Study
In the present study, longitudinal data on family involvement
within children’s school environments and literacy performance
from kindergarten through fifth grade were examined for a sample
of ethnically diverse low-income children and their families. Spe-
cifically, we used individual growth modeling and latent growth
modeling to address three research questions: (a) Was average
family involvement level associated with average literacy perfor-
mance? (b) Was average family involvement associated with
changes in literacy performance? and (c) Were changes in family
involvement within families associated with changes in literacy
performance? For each of these questions, we also examined the
potential moderating effect of maternal education.
We expected both between-families and within-families associ-
ations between family involvement in school and literacy perfor-
mance. For example, we expected that children living in families
with higher average levels of family involvement in school would
display higher average levels of literacy performance as well as
larger gains in literacy performance between kindergarten and fifth
grade compared with children living in families with lower aver-
age levels of involvement. Beyond these between-families associ-
ations, we also expected that changes in involvement within fam-
ilies would be associated with changes in literacy performance,
such that increased family involvement would be associated with
improved literacy performance. In addition, we expected that these
positive associations between involvement and literacy, both the
between-families and within-families relations, would be largest
for children whose mothers were least educated. Finally, given
theory on the potential direct and indirect effects of involvement
on literacy as well as previous longitudinal work indicating that
involvement in kindergarten predicts achievement in later grades
(even when controlling for achievement at kindergarten; e.g.,
Dearing et al., 2004; Izzo et al., 1999), we expected that our results
would be most consistent with a path of influence leading from
involvement to literacy performance rather than from literacy
performance to involvement.
1
Method
Participants
Data for this study were drawn from the impact evaluation of the
Comprehensive Child Development Program (CCDP) and the School
Transition Study (STS). The CCDP was a federally funded early interven-
tion program at 21 sites across the United States for low-income children
and their families from birth to entry into kindergarten (for further descrip-
tion of the CCDP, see St. Pierre, Layzer, Goodson, & Bernstein, 1999). The
intervention included services aimed at children (e.g., high-quality pre-
school) and their families (e.g., education and job training), with the dual
goals of enhancing child development and family economic self-
sufficiency. For the impact evaluation, approximately half of children were
randomly assigned to receive these intervention services, and the other half
were included in a control group.
Beginning in the fall of 1995, the STS was a follow-up investigation for
children from three of the CCDP sites (for further description of the STS,
see Harvard Family Research Project, 2006; Weiss et al., 2005). The
primary aim of the STS was to examine the developmental implications of
family and school contexts for low-income children, including the devel-
opmental role of family educational involvement. STS study sites were
selected to provide a diverse sample with regard to geographic region and
ethnicity, including a Northeastern city with a primarily African American
population, a rural New England town with an almost entirely European
American population, and a Western city with a primarily Latino
population.
Of the 403 children attending these three CCDP sites, 329 were followed
from kindergarten through the fifth grade in the STS; the remaining
children at these three sites were already in first grade by the start of the
STS. Of the 329 children, 281 had at least one observation for variables
that were measured longitudinally in the STS as well as complete data for
study covariates; as such, these 281 children could be included in this
study. Although children who were missing data did not significantly differ
from those with complete data on most demographic indicators (i.e.,
African American ethnicity, gender, birth weight, birth order, maternal
education, maternal employment, maternal partner status, maternal age at
childbirth, family income, or family Aid to Families with Dependent
Children [AFDC] receipt), children who were Latino and children who
were non-English speakers were more likely ( p⬍.05) to have missing
1
Although children may evoke family involvement in school, past
empirical work has often indicated that parents display an increased like-
lihood of getting involved when they perceive their child as having
achievement problems (for a review of this work, see Hoover-Dempsey et
al., 2001). Thus, to the extent that we find positive associations between
involvement in school and literacy performance, we expected these asso-
ciations to be the result of the effects of involvement on literacy rather than
the opposite.
655
FAMILY INVOLVEMENT AND LITERACY

observations than were other children.
2
Further details on patterns of
missing data for variables of primary interest (i.e., family school involve-
ment and children’s literacy) are provided along with the descriptions of
these measures.
Measures
Demographic data. Demographic data on child, maternal, and family
characteristics that were collected during recruitment into the CCDP were
used as covariates in analyses. Child characteristics included ethnicity,
gender, birth weight (i.e., low birth weight vs. other), biological risk at
birth (i.e., 7-point index assessing risk factors such as use of alcohol,
cigarettes, or drugs during pregnancy; higher scores indicated more risk
factors), and birth order (i.e., firstborn vs. later born). Maternal character-
istics included age at childbirth (i.e., teenager vs. other), partner status,
primary language (i.e., English vs. other), employment (i.e., employed vs.
unemployed), and depressive symptoms (i.e., Center for Epidemiological
Studies—Depression Scale; Radloff, 1977). Family characteristics in-
cluded CCDP status (i.e., control group vs. experimental group), AFDC
receipt (i.e., yes/no), per capita income, and study site (i.e., two dummy
codes for which Study Site 1 was the rural New England site and Study Site
2 was the Northeastern city).
Maternal education, used as a covariate and as a potential moderator of
family involvement in school, was collected during the STS at kindergar-
ten, third grade, and fifth grade. Maternal education was assessed on an
8-point scale ranging from 1 (no formal education)to8(graduate school).
In our analyses, we used the mean level of maternal education across these
assessments. Descriptive statistics for these demographic indicators are
presented in Table 1.
Families also reported on their total income and maternal hours of
employment at kindergarten, third grade, and fifth grade. There was,
however, considerable missing data (i.e., approximately 50% of families
had incomplete data) for these time-varying demographic indicators, and as
such, they were not included in our primary analyses. Nonetheless, it is
noteworthy that although most mothers were likely to be employed be-
tween kindergarten and fifth grade (e.g., averaged across this time period,
the 25th percentile was approximately 30 hr/ week, and the 75th percentile
was 40 hr/week), family income remained low (e.g., averaged across this
time period, the 50th percentile was less than $15,000/year, and the 75th
percentile was less than $30,000/year).
Family involvement in school. Mothers reported on family involve-
ment in children’s school at kindergarten, at third grade, and at fifth grade.
Specifically, eight dichotomous (yes/no) items were used to assess involve-
ment at school during the year (i.e., “Did you attend parent–teacher
conferences?”, “Did you visit your child’s classroom?”, “Did you attend
any school performances?”, “Did you attend any social events at your
child’s school?”, “Did you attend any field trips?”, “Did you volunteer in
your child’s classroom?”, “Did you attend meetings, like PTO or PTA?”,
“Did you attend classroom open houses?”, and “Did you volunteer in the
classroom?”).
For our within-families estimates of associations between involvement
and child literacy performance, the eight items were averaged within each
year so that yearly involvement scores ranged from 0 to 1, with higher
scores indicating higher levels of involvement within each year. Thus,
scores indicated the proportion of school involvement activities in which
families were involved at each year (e.g., a score of .50 indicated that
families were involved in four of the eight activities). Across children in
the study, the mean for yearly family involvement was 0.56 (SD ⫽0.26)
at kindergarten, 0.61 (SD ⫽0.25) at third grade, and 0.54 (SD ⫽0.26) at
fifth grade. On this indicator, 49.8% of sample children had no missing
data, 26.1% were missing one observation, 15.2% were missing two
observations, and 8.8% were missing all three observations.
Yearly involvement indicators were moderately reliable at each grade
(i.e., alpha [␣] ranged from .65 to .73) and were moderately intercorrelated
across grades (i.e., r⫽.33, p⬍.01 for kindergarten and third grade; r⫽
.33, p⬍.01 for kindergarten and fifth grade; and r⫽.41, p⬍.01 for third
and fifth grade). Approximately 45% of families in the STS displayed
decreases in yearly school involvement between kindergarten and fifth
grade, and another 10% of families displayed stable involvement levels
over time, results that are consistent with past research indicating that
educational involvement decreases during the early school years for many
low-income families (Serpell et al., 2005). Yet, the remaining 45% of
families in the study displayed increases in yearly involvement, with an
average increase of .29 points (i.e., 2.32 involvement activities) and a
maximum increase of .75 points (i.e., 6 involvement activities).
To assess average involvement across the study period, which was used
for our between-families estimates of associations between involvement
and child literacy performance, we averaged the three yearly involvement
indicators (i.e., the average of kindergarten, third grade, and fifth grade
involvement; ␣⫽.79). Thus, this variable provided an indicator of the
average proportion of school involvement activities in which families were
involved across kindergarten, third grade, and fifth grade.
Literacy performance. Children’s literacy performance at kindergar-
ten, at third grade, and at fifth grade was assessed using the Letter–Word
Identification subscale from the Woodcock–Johnson Psycho-Educational
Battery—Revised (WJ–R; Woodcock & Johnson, 1989). Specifically,
when completing this subscale, children are asked to (a) identify isolated
letters and words and (b) match pictographic representations of words with
pictures of the corresponding objects. At kindergarten, however, children
in the present study displayed a floor effect. To correct for this problem and
ensure variability in children’s scores, the subscale was adjusted at this
assessment time to include more letters and more short words of two to
four letters. This adjusted measure included 26 letters and 26 short words;
these 52 items proved to be internally consistent (␣⫽.85).
2
The Latino American ethnicity and primary language groupings were
largely overlapping. Specifically, of all non-English-speaking children,
92.8% were Latino American.
Table 1
Sample Descriptive Statistics
Variable M(SD)/%
African American ethnicity 36.4%
Latino American ethnicity 21.2%
Child is a boy 48.8%
Low birth weight 15.2%
Biological risk 1.07 (1.13)
Birth order was firstborn 35.7%
Teenage childbirth 29.3%
Mother had resident partner 44.4%
Primary language was English 82.2%
Mother was employed 13.8%
Maternal depressive symptoms 15.12 (9.98)
Maternal education
a
4.23 (1.13)
CCDP experimental group 49.2%
Received AFDC 66.7%
Income (per person) $2,003.37 ($1,095.23)
Study Site 1 42.8%
Study Site 2 23.1%
Note. CCDP ⫽Comprehensive Child Development Program; AFDC ⫽
Aid to Families With Dependent Children.
a
Maternal education was assessed on an 8-point scale ranging from 1 (no
formal education)to8(graduate school). The average level of education
(4.23) was slightly greater than high school completion, which was indi-
cated by a value of 4.
656 DEARING, KREIDER, SIMPKINS, AND WEISS

At third and fifth grades, we administered the complete Letter–Word Iden-
tification subscale (57 items) using standardized administration procedures,
including basal and ceiling levels to establish subtest scores. Because complete
subtests were administered only at third and fifth grades, we calculated
children’s literacy performance using the percentage of items that children
answered correctly, such that a higher proportion of items correct indicated
better literacy performance. At kindergarten, this proportion was calculated by
dividing the number of items answered correctly by 52 (M⫽0.39, SD ⫽
0.23). At third and fifth grades, this proportion was calculated by dividing
children’s raw scores by 57 (M⫽0.61 and 0.71, respectively; SD ⫽0.15 and
0.14, respectively). The third and fifth grade raw scores were the total number
of questions answered correctly, including items below the basal level (i.e.,
raw scores were derived by subtracting item failures from ceiling levels).
Although we did not examine internal consistency for the third and fifth
grade assessments because children completed different item sets as a
result of the use of basal and ceiling levels, this WJ–R subscale demon-
strated excellent split-half reliability and validity in the standardization
sample (Flanagan & Alfonso, 1995; McGrew & Knopik, 1993; Woodcock
& Johnson, 1989). In addition, children’s literacy scores in the present
study were significantly correlated over time, helping validate the use of
the adjusted measure at kindergarten; specifically, correlations for chil-
dren’s literacy performance across grades were .45 ( p⬍.001) for kinder-
garten and third grade, .44 ( p⬍.001) for kindergarten and fifth grade, and
.63 ( p⬍.001) for third and fifth grades. Of the 329 children in the sample,
46.2% had no missing data for literacy assessments, 31.6% had missing
data for one of the three assessments, 14.6% had missing data for two of
the three assessments, and 7.6% were missing all three assessments.
Results
Analysis Plan
Multilevel models (Raudenbush & Bryk, 2002; Singer & Willett,
2003) were used to estimate individual growth curves for children’s
literacy performance from kindergarten through fifth grade, as well as
to examine patterns of association between family involvement in
school and child literacy performance. Analyses were conducted in
four steps. First, as is generally recommended for growth modeling
(Raudenbush & Bryk, 2002; Singer & Willett, 2003), we estimated an
unconditional growth model to help determine the appropriate param-
eter specifications for subsequent models. Second, we estimated two-
level models in which average school involvement across the study
was specified as a time-invariant (Level 2) predictor of literacy
performance growth curves, and average school involvement at each
year was specified as a time-varying (Level 1) predictor of literacy
performance.
3
Third, we estimated two latent growth models predict-
ing changes in literacy performance from school involvement and
changes in school involvement from literacy performance.
The second analytic step was taken to address two questions, the
first focused on between-families differences and the second fo-
cused on within-families associations. Specifically, we addressed
(a) whether children living in families with higher average levels
of involvement in school displayed higher average levels of
achievement and greater gains in achievement levels over time and
(b) whether changes in involvement were associated with changes
in child literacy achievement such that increased involvement
predicted increased achievement. In answering these questions, we
also examined the moderating effect of maternal education to
determine whether involvement in school was more strongly re-
lated to literacy performance for children whose mothers were
relatively less educated compared with children whose mothers
were relatively more educated.
It is important to note that in all models, child grade and the
time-varying school involvement predictor were centered within fam-
ilies. Although other centering options for time-varying predictors in
multilevel growth models (e.g., uncentered or grand-mean centered)
result in estimates that contain a mix of within-families and between-
families information, by using within-families centering (alternatively
referred to as “group-mean” centering) we ensured that our estimated
associations between the time-varying predictors and literacy perfor-
mance were within-families estimates and were not biased by
between-families heterogeneity (Raudenbush & Bryk, 2002; Singer &
Willett, 2003). Yet, despite this advantage of within-families esti-
mates compared with between-families estimates, both can be af-
fected by reciprocal causation. Our third analytic step was taken to
help address this issue. Specifically, for our third analytic step, we
were interested in whether associations between family involvement
and literacy performance would be most consistent with a pathway of
influence leading from involvement to performance, from perfor-
mance to involvement, or both.
After completing these three analytic steps, we examined the ro-
bustness of our results to respecification. First, we reestimated our
models using only those items from the kindergarten literacy assess-
ment that were taken directly from the WJ–R. We used this respeci-
fication to help examine whether the additional items, which were
included to adjust for floor effects, biased our results. Next, we
reestimated our models using multiple imputation for missing values
to help ensure that our results were not biased by missing data. Of
import, in the models using multiple imputation, we were also able to
include two additional covariates to help control for contemporaneous
family circumstances: average family income and maternal hours of
employment between kindergarten and fifth grade.
4
Multiple imputation replaces missing data with values computed
from multivariate analyses of participants’ nonmissing data on
other variables plus random variation (Rubin, 1987; Schafer &
Graham, 2002). For the present study, we used multiple imputation
by chained equations (i.e., MICE, Royston, 2004) to generate five
complete data sets that combined observed and imputed values.
We then reestimated our multilevel and latent models from these
five complete data sets and combined estimates according to
“Rubin’s rules” (Rubin, 1987).
3
The following equation corresponds to the two-level model that in-
cluded family involvement as a time-varying predictor of literacy
performance:
yit ⫽关␤00 ⫹␤01AvgInvolvementi⫹␤10共ChildGradeit
⫺ChildGradei)⫹␤11AvgInvolvementi⫻共ChildGradeit
⫺ChildGradei)⫹␤20共AvgInvolvementit ⫺AvgInvolvementi兲]
⫹关␨00 ⫹␨10共ChildGradeit ⫺ChildGradei兲兴 ⫹␮it.
Level 2 time-invariant covariates have been excluded from the equation for
brevity. In addition, only the fixed effect for child grade was estimated in
models presented here, although all statistically significant effects were also
present when fixed and random effects were estimated for this parameter.
4
These covariates were not included in the first set of analyses because
they would have reduced the analysis sample size by 30% as a result of
missing observations.
657
FAMILY INVOLVEMENT AND LITERACY

Unconditional Growth Model for Children’s Literacy
Performance
To help determine appropriate parameter specifications for sub-
sequent growth models, we first estimated an unconditional growth
model for which child grade was the only predictor specified for
children’s literacy performance. Results from this unconditional
growth model are summarized in Table 2. The fixed-effect coef-
ficients noted in Table 2 provide estimates for the average child
literacy performance across the study period and the average
association between child grade and literacy performance (i.e., the
average rate of change in literacy performance from kindergarten
to fifth grade). The random effect estimates provide information
regarding individual variations around the sample averages.
When we averaged results across the 3 years of assessment, we
found that children responded correctly to 56% of the literacy
items, and this percentage of correct answers increased by approx-
imately 7% each year, as indicated by the fixed-effect coefficients
of .56 and .07, respectively. Thus, the average proportion of items
correct at kindergarten was 39%, and the average proportion
correct at fifth grade was 74%. As indicated by the significant
random effects (i.e., standard deviations), there were statistically
distinguishable variations around these averages as well. Although
estimates for average performance were more reliable than those
for child grade (i.e., .85 vs. .36), both were well above the
minimum computational criterion of .05, as recommended by
Raudenbush and Bryk (2002).
Average performance and rate of change in performance, how-
ever, were very strongly and negatively associated with one an-
other (⫺.83), such that children with lower levels of average
performance experienced greater increases in performance over
time than did children with higher levels of average performance.
This is justification for constraining the residual variance of one or
more parameter estimates to be zero (i.e., only estimating model
fixed effects). Considering this, we estimated two sets of models
for our questions of interest: (a) models with fixed and random
effects for child grade and (b) models with only fixed effects for
child grade (i.e., constraining the residual parameter variance for
child grade to zero). Given the strong association between param-
eters, we present models that included only the fixed effect for this
parameter; however, all significant results we report hereafter were
also significant when both fixed and random effects were esti-
mated for child grade.
Average Family Involvement and Average Literacy
Performance: Between-Families Associations
As a second analytic step, we estimated a two-level individual
growth model. In the first level of this model, yearly measures of
family involvement in school were specified as a time-varying
predictor of literacy performance to determine whether changes in
involvement were associated with changes in child literacy per-
formance such that increased involvement predicted increased
performance. In the second level of this model, average family
involvement between kindergarten and fifth grade was specified as
a predictor of average literacy performance and change in literacy
performance (i.e., child grade) to determine whether children liv-
ing in families with higher average levels of involvement in school
displayed higher average levels of achievement and greater gains
in achievement levels over time.
To control for potential omitted variable bias when examining
between-families differences at Level 2, we also included the
following child and family covariates as predictors of average
performance and change in performance: child gender, ethnicity,
biological risk at birth, birth order, and birth weight; maternal
education, age at childbirth, partner status, employment status,
primary language, and depression; family income, AFDC receipt
status, and CCDP group; and study site. Results for the family
involvement predictors in this model specification are displayed in
Table 3. Results for maternal education are also displayed because
of its inclusion in later models as a potential moderator.
Although the model coefficient for average family involvement
in school was positive for children’s average literacy performance
(i.e., .09) and for change in literacy performance (i.e., .01), these
trends were not statistically significant (i.e., p⫽.09 and .46,
respectively). Similarly, maternal education was positively asso-
ciated with average literacy performance, but not significantly so
(p⫽.09). Within-families changes in yearly involvement in
school, however, were significantly associated with changes in
literacy performance (coefficient ⫽.07) such that increased in-
volvement predicted increased performance. The effect size for
this association, as gauged by the partial correlation (i.e., pr ⫽
.13), was small in absolute terms. Relatively, however, it was
similar to the partial correlation for income (i.e., pr ⫽.11), the
largest effect size among the covariates.
When considering both the statistical and practical significance
of family involvement in this model, it is also important to note
that by estimating only the main effect, we have constrained
associations between involvement and literacy performance to be
equal across children. To move beyond the main effect results, we
added the interaction of average family involvement and maternal
education as a Level 2 predictor of both average literacy perfor-
mance and change in literacy performance. In addition, maternal
education was added as a Level 2 predictor of the within-families
association between yearly family involvement and literacy per-
formance. By so doing, we examined whether the between-
families and within-families associations between involvement
and literacy varied as a function of maternal education.
A summary of results from the model that included involvement
by education interactions is displayed in Table 4. Neither the
between-families association for average involvement and change
in literacy performance nor the within-families association for
change in yearly involvement and change in literacy performance
Table 2
Unconditional Growth Model for Children’s Literacy
Performance
Parameter
Fixed effect Random effect
Coefficient (SE)␶
a
Reliability SD
b
Average performance
(intercept)
.56*** (.01) ⫺.83 .85 .16***
Child grade (slope) .07*** (.003) .36 .03***
a
Tau is an estimate of the correlation between the random effects for the
model intercept and slope.
b
The statistical significance of random-effect
variance was estimated with a chi-square test statistic.
*** p⬍.001.
658 DEARING, KREIDER, SIMPKINS, AND WEISS

varied by level of maternal education. However, the association
between average level of family involvement and average literacy
performance between kindergarten and fifth grade was signifi-
cantly different for children with less educated versus more edu-
cated mothers. The direction of this interaction is illustrated in
Figure 1 (i.e., for graphing purposes, the percentage of literacy
items answered correctly are displayed at levels of family involve-
ment that were either one standard deviation above or below the
mean and maternal education levels that were either one standard
deviation above or below the mean).
For children whose mothers had above-average levels of edu-
cation, there was no association between average level of family
involvement in school and average literacy performance. On the
other hand, for children whose mothers had below-average levels
of education, higher levels of family involvement were associated
with better literacy performance. In fact, although children with
less educated mothers had, on average, approximately 10% fewer
items correct on the literacy scales than children with more edu-
cated mothers if family involvement in school was low, this
achievement gap was nonexistent if family involvement in school
was high.
5
Latent Growth Models
As a third analytic step, we estimated two latent growth models
examining associations between family involvement in school and
children’s literacy performance. In the first model, four latent
constructs were estimated from the observed data: kindergarten
status for literacy performance, change in literacy performance
between kindergarten and fifth grade, average school involvement
between kindergarten and fifth grade, and change in school in-
volvement between kindergarten and fifth grade. In this model,
change in literacy performance was then regressed on the other
three latent variables as well as the Level 2 control variables
specified in the multilevel growth models. In other words, we
estimated the association between change in literacy and the in-
volvement indicators while controlling for initial literacy perfor-
mance levels as well as child and family covariates.
5
For this interaction, we estimated the range of values on maternal
education for which the association between involvement and literacy was
significant (i.e., the region of significance). At maternal education levels of
3.65 and lower, family involvement in school was positively and signifi-
cantly associated with child literacy performance. In other words, the
positive effects of family involvement in school were significant for
children whose mothers had less than a high school level of education.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Family School Involvement
Percent Correct on
Woodcock-Johnson
Literacy Scales
Above Average Maternal Education
Below Average Maternal Education
Below Average Above Average
Figure 1. The association between family school involvement and liter-
acy performance for children whose mothers had above- and below-
average levels of maternal education. For both family involvement and
maternal education, above- and below-average predicted scores were cal-
culated with values one standard deviation above and below the mean,
respectively.
Table 3
Two-Level Growth Models for Child Literacy Performance
Model parameter
Literacy performance
Coefficient (SE)t(262)
Average literacy performance (i.e., Level 1
intercept)
Average family involvement in school .09 (.06) 1.66
†
Maternal education .02 (.01) 1.68
†
Change in literacy performance (i.e., child
grade)
Average family involvement in school .01 (.02) .72
Maternal education .00 (.00) .26
Yearly family involvement in school .07 (.03) 2.27*
Note. For brevity, coefficients are displayed only for predictors of pri-
mary interest; however, the following covariates were estimated for both
average performance and child grade (i.e., change in performance): child
gender, ethnicity, biological risk at birth, birth order, and birth weight;
maternal age at childbirth, partner status, employment status, primary
language, and depression; family income, Aid to Families With Dependent
Children receipt status, and Comprehensive Child Development Program
group; and study site.
†
p⬍.10. * p⬍.05.
Table 4
Summary of Family Involvement by Maternal Education
Interactions
Model parameter
Literacy performance
Coefficient (SE)t(261)
Average literacy performance (i.e., Level 1
intercept)
Average family involvement in school .61 (.18) 3.39***
Maternal education .11 (.03) 3.34***
Family Involvement ⫻Maternal Education ⫺.13 (.04) ⫺3.01**
Change in literacy performance (i.e., child grade)
Average family involvement in school .01 (.07) .16
Maternal education ⫺.00 (.01) ⫺.31
Family Involvement ⫻Maternal Education .00 (.02) .01
Yearly family involvement in school
Maternal education .02 (.03) .87
Note. For brevity, coefficients are displayed only for predictors of pri-
mary interest; however, the following covariates were estimated for both
average performance and child grade (i.e., change in performance): child
gender, ethnicity, biological risk at birth, birth order, and birth weight;
maternal age at child birth, partner status, employment status, primary
language, and depression; family income, Aid to Families With Dependent
Children receipt status, and Comprehensive Child Development Program
group; and study site.
** p⬍.01. *** p⬍.001.
659
FAMILY INVOLVEMENT AND LITERACY

We also estimated four latent constructs for the second model:
average literacy performance between kindergarten and fifth
grade, change in literacy performance between kindergarten and
fifth grade, kindergarten status for school involvement, and change
in school involvement between kindergarten and fifth grade. In this
model, change in school involvement was regressed on the other
three latent variables as well as the Level 2 control variables from
the multilevel growth models. In other words, we estimated the
association between change in school involvement and the perfor-
mance indicators while controlling for initial school involvement
levels as well as child and family covariates.
These two latent growth models were estimated to help deter-
mine whether associations between family involvement and liter-
acy performance were most consistent with a pathway of influence
leading from involvement to performance, from performance to
involvement, or both. If involvement influences performance, then
high average levels of involvement between kindergarten and fifth
grade and increasing involvement across this time period should be
associated with increased performance after kindergarten (i.e.,
when controlling for children’s performance levels at kindergar-
ten) in the first latent growth model. Alternatively, if performance
influences involvement, then high average levels of performance
and increasing performance should be associated with increased
involvement after kindergarten in the second growth model. Note,
however, that although the results of the latent models could prove
to be consistent or inconsistent with one or both of these pathways
of influence, causality could not be entirely disentangled on the
basis of any of our analyses, because the present study relies on
nonexperimental data.
Results from the first latent model are summarized in Table 5,
and results from the second are summarized in Table 6. Changes
in family involvement in school between kindergarten and fifth
grade predicted changes in children’s literacy performance during
this time, controlling for children’s literacy performance at kin-
dergarten. Specifically, children whose families increased their
involvement over time had greater gains in literacy performance
after kindergarten than children whose families were stable or
decreased their involvement over time.
6
Although not large in
absolute terms, the effect size for this association ( pr ⫽.23) was
the second largest in the model; only the effect size for kindergar-
ten literacy performance ( pr ⫽.38) was larger. On the other hand,
controlling for family involvement at kindergarten, neither average
literacy performance nor increased literacy performance was sig-
nificantly associated with changes in literacy performance.
Replication Models
To help ensure that the adjusted kindergarten assessment and
missing data were not biasing our results, we reestimated all of our
statistical models. First, we reestimated our models using only
those items from the kindergarten assessment that were taken
directly from the WJ–R Letter–Word Recognition subscale. Sec-
ond, we reestimated our models using multiple imputation for
missing values. In the latter, we also added two additional control
variables: families’ average income and mothers’ average hours of
employment between kindergarten and fifth grade. The inclusion
of these additional covariates allowed us to control for contempo-
raneous economic and employment circumstances, in addition to
the baseline controls used in our original analyses.
All of our significant results for family involvement in school
were replicated in these additional analyses. In fact, when using
the smaller set of literacy items for the kindergarten assessment,
we also found a significant association between average level of
school involvement and changes in literacy performance, such that
higher average levels of involvement were predictive of more
rapid increases in literacy between kindergarten and fifth grade.
This additional result, however, was not evident when we replaced
missing data using multiple imputation.
Summary of Results
Our results indicated that both between-families differences in
school involvement and within-families changes in school involve-
ment were associated with child literacy. Although the main effect
6
In latent growth models, we also examined the possibility that kinder-
garten achievement was indirectly associated with later achievement via
family involvement. That is, high levels of kindergarten achievement may
lead to high levels of family involvement and, thereby, positively influence
later achievement. However, we found no evidence that this was the case.
Table 5
Latent Growth Curve Analysis Predicting Changes in Literacy
Performance
Predictor Coefficient (SE)t(226)
Kindergarten literacy performance ⫺.13 (.02) ⫺6.81***
Change in literacy performance
Average family involvement in school .01 (.01) 0.75
Change in family involvement in school .24 (.06) 3.95***
Note. For brevity, coefficients are displayed only for predictors of pri-
mary interest; however, the following covariates were estimated for both
average performance and child grade (i.e., change in performance): child
gender, ethnicity, biological risk at birth, birth order, and birth weight;
maternal age at childbirth, partner status, employment status, primary
language, and depression; family income, Aid to Families With Dependent
Children receipt status, and Comprehensive Child Development Program
group; and study site.
*** p⬍.001.
Table 6
Latent Growth Curve Analysis Predicting Changes in Family
Involvement
Predictor Coefficient (SE) t(226)
Kindergarten family involvement in school ⫺.16 (.06) ⫺2.65**
Change in family involvement in school
Average literacy performance .29 (.77) 0.37
Change in literacy performance 2.27 (5.26) 0.43
Note. For brevity, coefficients are displayed only for predictors of pri-
mary interest; however, the following covariates were estimated for both
average performance and child grade (i.e., change in performance): child
gender, ethnicity, biological risk at birth, birth order, and birth weight;
maternal age at childbirth, partner status, employment status, primary
language, and depression; family income, Aid to Families With Dependent
Children receipt status, and Comprehensive Child Development Program
group; and study site.
** p⬍.01.
660 DEARING, KREIDER, SIMPKINS, AND WEISS

of average involvement in school across the study was not a
statistically significant predictor, its effect did vary by maternal
education. Specifically, despite having little association with the
literacy performance of children whose mothers were relatively
more educated, average involvement was positively associated
with average literacy performance for children whose mothers
were relatively less educated, such that these children displayed
higher levels of performance when their families were more in-
volved. In addition, increased involvement in school within fam-
ilies was associated with improved child literacy performance, in
both our multilevel and latent growth models.
Discussion
Family educational involvement is often touted by policymakers
and educators alike as one way of improving the achievement of
low-income children (U.S. Department of Education, 2005). The
present study helps address these claims and extends previous
empirical work on family involvement in school and low-income
children’s literacy by examining longitudinally both between-
families and within-families associations for involvement and lit-
eracy achievement. Higher average levels of involvement were, for
example, positively associated with higher average levels of liter-
acy performance across the study for children from less educated
families. In addition, regardless of maternal education level, when
families increased their involvement in school between kindergar-
ten and fifth grade, children’s literacy performance improved.
Average Family Involvement in School and Literacy
Performance
In this study, the main effect of average family involvement in
school was not significantly associated with average literacy per-
formance or change in literacy performance; the former associa-
tion, however, did vary by maternal education. Specifically, al-
though there was no association between average involvement and
average literacy for children whose mothers were relatively more
educated, average involvement levels between kindergarten and
fifth grade were positively associated with average literacy per-
formance levels for children whose mothers were relatively less
educated. Indeed, although there was an achievement gap in av-
erage literacy performance across the study between children of
more and less educated mothers if family involvement levels were
low, this gap was nonexistent if family involvement levels were
high.
On the basis of our past analyses of this same data set (Dearing
et al., 2004), we reported similar moderating effects of maternal
education for associations between family educational involve-
ment at kindergarten and the development of low-income chil-
dren’s feelings about literacy and literacy achievement. We found,
for example, that although children of relatively less educated
mothers have more negative feelings about their literacy perfor-
mance at kindergarten than do children of relatively more educated
mothers, this difference diminishes over time if families are highly
involved. Thus, the present study adds to increasing evidence that
family involvement in school among low-income families may be
most likely to improve the achievement of children who are at
exceptional risk for academic failure because of both low income
and low parent education.
Within-Families Changes in Involvement and Changes in
Literacy Performance
Regardless of whether families had high or low average levels
of involvement between kindergarten and fifth grade, we also
found evidence that changes in involvement were associated with
changes in child literacy performance. Specifically, increases in
family involvement in school were associated with increases in
literacy performance. This result adds to existing studies in at least
two important ways. First, because most empirical work has fo-
cused solely on between-families differences using nonexperimen-
tal data (e.g., do children with higher levels of involvement have
higher levels of achievement?), it has remained unclear whether
changes in involvement within families have developmental reper-
cussions; our results suggest that they do.
Second, by using within-families analyses, our estimates were
not susceptible to bias caused by time-invariant omitted variables
(Duncan et al., 2004). Causal inferences are not appropriate when
associations between involvement and child achievement have
been estimated using nonexperimental data, primarily because we
cannot assume that different families provide different levels of
involvement randomly. Of importance, however, because the as-
sociation between increased involvement and improved achieve-
ment in the present study was estimated within families, charac-
teristics of children, their families, and/or their environments that
were fixed over time (i.e., did not vary) could be ruled out as
potential sources of bias.
Nonetheless, as is true for between-families analyses of nonex-
perimental data, within-families analyses of nonexperimental data
can be biased by reciprocal causation. For example, although
literacy performance was specified as the outcome in most of our
analyses, it may influence family involvement, or the two may
simultaneously influence one another. Our results, however, were
most consistent with a path of influence leading from involvement
to achievement. Specifically, although changes in achievement
were not predictive of changes in involvement when controlling
for initial levels of involvement, changes in involvement were
predictive of changes in achievement when controlling for initial
levels of achievement.
When considering these results, it is important to recognize
evidence that family involvement in school often decreases across
the middle and high school years for middle-income families (e.g.,
Adams & Christenson, 2000; Simon, 2004), and it begins decreas-
ing across the early school years for many low-income families
(e.g., Serpell et al., 2005). In fact, nearly half of families in the
present study displayed declines over time in their school involve-
ment between kindergarten and fifth grade. On the other hand,
most other families in the sample displayed increases in involve-
ment, and these increases predicted improved literacy above and
beyond average level of involvement. Thus, although low-income
families may experience many barriers to increasing, or even
sustaining, involvement in their children’s early schooling, our
results indicated that doing so has benefits for children’s literacy.
Relative Effect Sizes
The effect sizes for family school involvement were not large in
absolute terms. Yet, as McCartney and Rosenthal (2000) have
pointed out, the practical significance of effect sizes is often best
661
FAMILY INVOLVEMENT AND LITERACY

evaluated within context. Compared with the effect sizes for our
analysis covariates, those for school involvement were among the
largest in the models we estimated. Given the socioeconomics of
the sample, however, the range of values was restricted for covari-
ates such as family income. Nonetheless, within the empirical
context of the present study, family involvement in school was of
both statistical and practical significance for children’s literacy
achievement, particularly for low-income children whose parents
had relatively low levels of education.
Study Limitations
Our within-families analysis provided control for time-invariant
heterogeneity across families. Between-families differences,
whether observed or unobserved, could not have biased the within-
families association between increased involvement and improved
literacy. Our between-families results, on the other hand, were
susceptible to omitted variable bias associated with either time-
invariant or time-varying variables that were not observed. Fur-
ther, our within-families results were susceptible to omitted vari-
able bias associated with time-varying variables that were not
observed.
During the course of the present study, for example, families
may have experienced life changes (e.g., decreases in maternal
depressive symptoms or increases in income) that influenced both
their propensity to become involved in their children’s education
as well as their children’s literacy performance. Indeed, five of the
baseline covariates (i.e., maternal partner status, employment, and
depression; and family income and AFDC receipt), which were
assessed when children were born, likely varied over time. Al-
though our results were, in general, replicated in models when we
controlled for average income and maternal employment between
kindergarten and fifth grade (using multiple imputation for missing
data), time-varying data on all of these constructs would have
strengthened statistical control in the present study.
The present study was also limited to examining family involve-
ment in school. Beyond in-school activities, families can be in-
volved in their children’s education through involvement in the
home (e.g., help with homework), home–school communications
(e.g., phone conversations with teachers), and social networks
(e.g., parent-to-parent networks) as well as parental attitudes (e.g.,
educational expectations). Although parents’ educational expecta-
tions and involvement behaviors in the home have been of greatest
importance for children’s achievement compared with other areas
of involvement when examined using between-families analyses
(e.g., Fan, 2001; Izzo et al., 1999), future longitudinal studies that
examine multiple areas of educational involvement within families
could help further disentangle the unique and interactive effects of
these family investments in their children’s educations. This type
of future research is of particular importance for understanding the
involvement of low-income families who are working; these fam-
ilies experience unique barriers to involvement during the work
day (e.g., inflexible work schedules limit classroom volunteering),
but they may also develop unique strategies to stay involved (e.g.,
Weiss et al., 2005).
Higher rates of missing data for Latino and non-English-
speaking families compared with other families limited the present
study, particularly in light of these families’ potential cultural and
language barriers to family involvement in school. Although ap-
proximately 20% of the families in the present study were Latino
and/or non-English speaking, and although our results were, in
general, replicated when all families were included via multiple
imputation of missing data, our findings should be replicated in
samples with lower attrition rates before generalizing to these
children and families.
In addition, there are two measurement issues of note. First,
given that our estimates of family involvement in school were
based on parent report, social desirability is a concern (i.e., parents
may overreport their involvement in school). Because measure-
ment error in predictor variables downwardly biases estimated
associations, the results of the present study may underestimate
effect sizes for school involvement as a function of social desir-
ability. Second, the measure of child literacy performance at
kindergarten differed from the measures administered at later
grades. Although kindergarten scores were correlated with scores
at later grades, standardized administration across waves would
have strengthened the study.
Conclusion
Our results support the usefulness of family involvement in
schools as a means of improving the achievement of children
living in low-income families and underscore the value of empir-
ically modeling both family involvement and child achievement as
developmental phenomena. Given that higher average levels of
involvement and increases in involvement were both associated
with literacy performance, implications for practice include the
need for schools to view family involvement over the long term,
ideally creating an educational environment that increases the
involvement of families that are less involved, and then helps
sustain that involvement across elementary school. This need is
underscored by the fact that, on average, lower income families
display relatively low levels of involvement that may decrease
across the early schooling years (Jeynes, 2005a, 2005b; Serpell et
al., 2005).
Past research indicates that increases in family involvement are
most likely to occur among lower income families when schools
not only reach out to families and invite them to become involved
but also help these families overcome barriers to involvement,
such as child care and transportation needs (for a review, see
Henderson & Mapp, 2002). Beyond practice, the policy implica-
tions of the present study include the potential benefit of targeting
family involvement intervention toward children at exceptional
risk for educational failure, those living in families with low
income and little education. Considering that average level of
family involvement in school was most strongly associated with
literacy achievement for these children, both the child-level and
societal-level benefits of intervention may be maximized when
efforts are focused on these children and their families.
References
Adams, K. S., & Christenson, S. L. (2000). Trust and the family–school
relationship: An examination of parent–teacher differences in elemen-
tary and secondary grades. Journal of School Psychology, 38, 447– 497.
Angrist, J., & Krueger, A. (1999). Empirical strategies in labor economics.
In O. Ashenfelter & D. Card (Eds.), Handbook of labor economics (Vol.
3, pp. 1277–1366). Amsterdam: Elsevier Science.
Autor, D. H., Levy, F., & Murnane, R. J. (2003). The skill content of recent
662 DEARING, KREIDER, SIMPKINS, AND WEISS

technological change: An empirical exploration. Quarterly Journal of
Economics, 118, 1279 –1333.
Bandura, A., Barbaranelli, C., Caprara, G. V., & Pastorelli, C. (1996).
Multifaceted impact of self-efficacy beliefs on academic functioning.
Child Development, 67, 1206 –1222.
Barnard, W. M. (2004). Parent involvement in elementary school and
educational attainment. Children & Youth Services Review, 26, 39 – 62.
Bronfenbrenner, U., & Morris, P. A. (1998). The ecology of developmental
processes. In W. Damon & R. M. Lerner (Eds.), Handbook of child
psychology: Vol. 1: Theoretical models of human development (5th ed.,
pp. 993–1028). New York: Wiley.
Dearing, E., McCartney, K., Weiss, H. B., Kreider, H., & Simpkins, S.
(2004). The promotive effects of family educational involvement for
low-income children’s literacy. Journal of School Psychology, 42, 445–
460.
Duncan, G. J., Magnuson, K. A., & Ludwig, J. (2004). The endogeneity
problem in developmental studies. Research in Human Development, 1,
59 – 80.
Englund, M. M., Luckner, A. E., Whaley, G. J. L., & Egeland, B. (2004).
Children’s achievement in early elementary school: Longitudinal effects
of parental involvement, expectations, and quality of assistance. Journal
of Educational Psychology, 96, 723–730.
Entwisle, D. R., & Alexander, K. L. (1996). Family type and children’s
growth in reading and math over the primary grades. Journal of Mar-
riage & the Family, 58, 341–355.
Epstein, J. L. (1995). Family/school/community partnerships: Caring for
the children we share. Phi Delta Kappa, 76, 701–712.
Fan, X. (2001). Parental involvement and students’ academic achievement:
A growth modeling analysis. Journal of Experimental Education, 70,
27– 61.
Fan, X., & Chen, M. (2001). Parental involvement and students’ academic
achievement: A meta-analysis. Educational Psychology Review, 13,
1–22.
Flanagan, D. P., & Alfonso, V. C. (1995). A critical review of the technical
characteristics of new and recently revised intelligence tests for pre-
school children. Journal of Psychoeducational Assessment, 13, 66 –90.
Frome, P. M., & Eccles, J. S. (1998). Parents’ influence on children’s
achievement-related perceptions. Journal of Personality and Social Psy-
chology, 74, 435– 452.
Gottfried, A. E., Fleming, J. S., & Gottfried, A. S. (1994). Role of parental
motivational practices in children’s academic intrinsic motivation and
achievement. Journal of Educational Psychology, 86, 104 –113.
Hart, B., & Risley, T. R. (1995). Meaningful differences in the everyday
experience of young American children. Baltimore: Brookes Publishing.
Harvard Family Research Project. (2006). School transition study: Project
description. Retrieved May 25, 2006, from http://www.gse.harvard.edu/
hfrp/projects/sts.html
Henderson, A. T., & Mapp, K. L. (2002). A new wave of evidence: The
impact of school, family, and community connections on student
achievement. Austin, TX: National Center for Family and Community
Connections with Schools.
Hill, N. E., & Taylor, L. C. (2004). Parental school involvement and
children’s academic achievement: Pragmatics and issues. Current Di-
rections in Psychological Science, 13, 161–164.
Hoover-Dempsey, K. V., Battiato, A. C., Walker, J. M. T., Reed, R. P.,
DeJong, J. M., & Jones, K. P. (2001). Parental involvement in home-
work. Educational Psychologist, 36, 195–209.
Hoover-Dempsey, K. V., & Sandler, H. M. (1995). Parental involvement in
children’s education: Why does it make a difference? Teachers College
Record, 97, 310 –332.
Hoover-Dempsey, K. V., & Sandler, H. M. (1997). Why do parents
become involved in their children’s education? Review of Educational
Research, 67, 3– 42.
Hsaio, C. (2003). Analysis of panel data. New York: Cambridge University
Press.
Izzo, C. V., Weissberg, R. P., Kasprow, W. J., & Fendrich, M. (1999). A
longitudinal assessment of teacher perceptions of parent involvement in
children’s education and school performance. American Journal of Com-
munity Psychology, 27, 817– 839.
Jeynes, W. H. (2003). A meta-analysis: The effects of parental involvement
on minority children’s academic achievement. Education and Urban
Society, 35, 202–218.
Jeynes, W. H. (2005a). A meta-analysis of the relation of parental involve-
ment to urban elementary school student academic achievement. Urban
Education, 40, 237–269.
Jeynes, W. H. (2005b). The effects of parental involvement on the aca-
demic achievement of African American youth. Journal of Negro Edu-
cation, 74, 260 –274.
Kohl, G. K., Lengua, L. J., & McMahon, R. J. (2000). Parent involvement
in school: Conceptualizing multiple dimensions and their relations with
family and demographic risk factors. Journal of School Psychology, 38,
501–523.
Lareau, A. (1989). Home Advantage: Social class and parental interven-
tion in elementary education. Philadelphia: Falmer Press.
Magnusson, D., & Stattin, H. (1998). Person-context interaction theories.
In W. Damon & R. M. Lerner (Eds.), Handbook of child psychology:
Vol. 1: Theoretical models of human development (5th ed., pp. 685–
759). New York: Wiley.
McCartney, K., & Rosenthal, R. (2000). Effect size, practical importance,
and social policy for children. Child Development, 71, 173–180.
McGrew, K. S., & Knopik, S. N. (1993). The relationship between the
WJ–R Gf-Gc cognitive clusters and writing achievement across the
life-span. School Psychology Review, 22, 687– 695.
Radloff, L. S. (1977). The CES–D Scale: A self-report depression scale for
research in the general population. Applied Psychological Measurement,
1, 385– 401.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models:
Applications and data analysis methods: Advanced quantitative tech-
niques in the social sciences (2nd ed., Vol. 1). Newbury Park, CA: Sage.
Rauh, V. A., Parker, F. L., Garfinkel, R. S., Perry, J., & Andrews, H. F.
(2003). Biological, social, and community influences on third-grade
reading levels of minority Head Start children: A multilevel approach.
Journal of Community Psychology, 31, 255–278.
Rogosa, D. (1995). Myths and methods: “Myths about longitudinal re-
search” plus supplemental questions. In J. M. Gottman (Ed.), The anal-
ysis of change (pp. 3– 66). Mahwah, NJ: Erlbaum.
Royston, P. (2004). Multiple imputation of missing values. Stata Journal,
4, 227–241.
Rubin, D. B. (1987). Multiple imputation for nonresponse in surveys. New
York: Wiley.
Sanders, M. (1998). The effects of school, family, and community support
on the academic achievement of African American adolescents. Urban
Education, 33, 385– 409.
Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state
of the art. Psychological Methods, 7, 147–177.
Serpell, R., Baker, L., & Sonnenschein, S. (2005). Becoming literate in the
city: The Baltimore Early Childhood Project. New York: Cambridge
University Press.
Simon, B. S. (2004). High school outreach and family involvement. Social
Psychology of Education, 7, 185–209.
Singer, J. D., & Willett, J. T. (2003). Applied longitudinal data analysis:
Modeling change and event occurrence. New York: Oxford University
Press.
Snow, C. E., Burns, M. S., & Griffin, P. (Eds.). (1998). Preventing reading
difficulties in young children. Committee on the Prevention of Reading
Difficulties in Young Children, National Research Council. Washington,
DC: National Academies Press.
663
FAMILY INVOLVEMENT AND LITERACY

St. Pierre, R. G., Layzer, J. I., Goodson, B. D., & Bernstein, L. S. (1999).
The effectiveness of comprehensive case management interventions:
Evidence from the national evaluation of the Comprehensive Child
Development Program. American Journal of Evaluation, 20, 15–34.
Taylor, B. A., Dearing, E., & McCartney, K. (2004). Incomes and out-
comes in early childhood. Journal of Human Resources, 39, 980 –1007.
U.S. Department of Education. (2005). No Child Left Behind: Facts and
terms every parent should know about NCLB. Retrieved October 19,
2005, from http://www.ed.gov/nclb/overview/intro/parents.html
U.S. Department of Health and Human Services. (2005). Head Start
Bureau. Retrieved October 19, 2005, from http://www.acf.hhs.gov/
programs/hsb/
Weiss, H. B., Dearing, E., McCartney, K., Kreider, H., Mayer, E., &
Simpkins, S. (2005). Family educational involvement: Who can afford it
and what does it afford? In C. R. Cooper, C. Garcia Coll, T. Bartko, H.
Davis, & C. Chatman (Eds.), Developmental pathways through middle
childhood: Rethinking context and diversity as resources (pp. 17–39).
Hillsdale, NJ: Erlbaum.
Woodcock, R. W., & Johnson, M. B. (1989). Woodcock–Johnson Psycho-
educational Battery—Revised. Chicago: Riverside Publishing.
Received August 31, 2005
Revision received June 15, 2006
Accepted June 27, 2006 䡲
Call for Nominations
The Publications and Communications (P&C) Board has opened nominations for the editorships
of Journal of Applied Psychology, Psychological Bulletin, Psychology of Addictive Behaviors,
Journal of Personality and Social Psychology: Interpersonal Relations and Group Processes
(IRGP), and Journal of Educational Psychology for the years 2009-2014. Sheldon Zedeck, PhD,
Harris Cooper, PhD, Howard J. Shaffer, PhD, Charles S. Carver, PhD, and Karen R. Harris, PhD,
respectively, are the incumbent editors.
Candidates should be members of APA and should be available to start receiving manuscripts in
early 2008 to prepare for issues published in 2009. Please note that the P&C Board encourages
participation by members of underrepresented groups in the publication process and would partic-
ularly welcome such nominees. Self-nominations are also encouraged.
Search chairs have been appointed as follows:
•Journal of Applied Psychology, William C. Howell, PhD and J Gilbert Benedict, PhD
•Psychological Bulletin, Mark Appelbaum, PhD and Valerie F. Reyna, PhD
•Psychology of Addictive Behaviors, Linda P. Spear, PhD and Robert G. Frank, PhD
•Journal of Personality and Social Psychology: IRGP, David C. Funder, PhD
•Journal of Educational Psychology, Peter A. Ornstein, PhD and Leah L. Light, PhD
Candidates should be nominated by accessing APA’s EditorQuest site on the Web. Using your
Web browser, go to http://editorquest.apa.org. On the Home menu on the left, find “Guests”. Next,
click on the link “Submit a Nomination,” enter your nominee’s information, and click “Submit.”
Prepared statements of one page or less in support of a nominee can also be submitted by e-mail
to Susan J.A. Harris, P&C Board Search Liaison, at sjharris@apa.org.
Deadline for accepting nominations is January 10, 2007, when reviews will begin.
664 DEARING, KREIDER, SIMPKINS, AND WEISS