Literature review . Articles enclosed in PDF as well as instructions.. NO OTHER OUTSIDE SOURCES FOR LIT REVIEW just the Articles enclosed
ryamar1
school_readiness_study.pdf
School Readiness and Later Achievement
Greg J. Duncan Northwestern University
Chantelle J. Dowsett University of Texas at Austin
Amy Claessens Northwestern University
Katherine Magnuson University of Wisconsin–Madison
Aletha C. Huston University of Texas at Austin
Pamela Klebanov Princeton University
Linda S. Pagani Université de Montréal
Leon Feinstein University of London
Mimi Engel Northwestern University
Jeanne Brooks-Gunn Columbia University
Holly Sexton University of Michigan
Kathryn Duckworth University of London
Crista Japel Université de Québec à Montréal
Using 6 longitudinal data sets, the authors estimate links between three key elements of school readiness—school-entry academic, attention, and socioemotional skills—and later school reading and math achievement. In an effort to isolate the effects of these school-entry skills, the authors ensured that most of their regression models control for cognitive, attention, and socioemotional skills measured prior to school entry, as well as a host of family background measures. Across all 6 studies, the strongest predictors of later achievement are school-entry math, reading, and attention skills. A meta-analysis of the results shows that early math skills have the greatest predictive power, followed by reading and then attention skills. By contrast, measures of socioemotional behaviors, including internalizing and exter- nalizing problems and social skills, were generally insignificant predictors of later academic perfor- mance, even among children with relatively high levels of problem behavior. Patterns of association were similar for boys and girls and for children from high and low socioeconomic backgrounds.
Keywords: school readiness, socioemotional behaviors, attention, early academic skills
Supplemental materials: http://dx.doi.org/10.1037/[0012-1649.43.6.1428].supp
Greg J. Duncan, Amy Claessens, and Mimi Engel, School of Education and Social Policy, Northwestern University; Chantelle J. Dowsett and Aletha C. Huston, Department of Human Ecology, University of Texas at Austin; Katherine Magnuson, Department of Social Work, University of Wisconsin–Madison; Pamela Klebanov, Center for Research on Child Wellbeing, Princeton University; Linda S. Pagani, Department of Psycho- education, Université de Montréal, Québec, Canada; Leon Feinstein and Kathryn Duckworth, Department of Quantitative Social Science, Institute of Education, University of London, London, England; Jeanne Brooks- Gunn, Department of Pediatrics, Columbia University; Holly Sexton, Re- seach Center for Group Dynamics, University of Michigan; Crista Japel, Département d’éducation et formation spécialisées, Université de Québec à Montréal, Québec, Canada.
A preliminary version of this article was presented at the biennial
meetings of the Society for Research on Child Development, Atlanta, Georgia, April 2005. We are grateful to the National Science Foundation-supported Center for the Analysis of Pathways from Child- hood to Adulthood (CAPCA; Grant 0322356) for research support. We thank Larry Aber, Mark Appelbaum, Avshalom Caspi, David Cordray, Herbert Ginsburg, David Grissmer, Mark Lipsey, Derek Neal, Cybele Raver, Arnold Sameroff, Robert Siegler, Ross Thompson, Sandra Jo Wilson, Nicholas Zill, and other members of CAPCA and the MacArthur Network on Families and the Economy for helpful com- ments.
Correspondence concerning this article should be addressed to Greg J. Duncan, School of Education and Social Policy, Northwestern University, 2046 Sheridan Road, Evanston, IL 60208. E-mail: greg- [email protected]
Developmental Psychology Copyright 2007 by the American Psychological Association 2007, Vol. 43, No. 6, 1428–1446 0012-1649/07/$12.00 DOI: 10.1037/0012-1649.43.6.1428
1428
Early childhood programs and policies that promote academic skills have been gaining popularity among politicians and re- searchers. For example, President George W. Bush (2002) en- dorsed Head Start reforms in 2002 that focus on building early academic skills, observing that “on the first day of school, children need to know letters and numbers. They need a strong vocabulary. These are the building blocks of learning, and this nation must provide them” (p. 12). The National Research Council’s Commit- tee on the Prevention of Reading Difficulties in Young Children recommends providing environments that promote preliteracy skills for all preschool children (Snow, Burns, & Griffin, 1998). Similarly, the National Association for the Education of Young Children and the National Council of Teachers of Mathematics (2002) issued a joint statement that advocated for high-quality mathematics education for children ages 3–6.
Others, however, maintain that a broad constellation of behav- iors and skills enables children to learn in school. Asked to identify factors associated with a difficult transition to school, kindergarten teachers frequently mentioned weaknesses in academic skills, problems with social skills, trouble following directions, and dif- ficulty with independent and group work (Rimm-Kaufman, Pianta, & Cox, 2000). Researchers too have made this point. The National Research Council and Institute on Medicine argued that “the elements of early intervention programs that enhance social and emotional development are just as important as the components that enhance linguistic and cognitive competence” (Shonkoff & Phillips, 2000, pp. 398–399).
These two views have emerged in the current debate about what constitutes school readiness and in particular about what skills predict school achievement. Many early education programs, in- cluding Head Start, are designed to enhance children’s physical, intellectual, and social competencies on the grounds that each domain contributes to a child’s overall developmental competence and readiness for school. However, if early acquisition of specific academic skills or learning-enhancing behaviors forecasts later achievement, it may be beneficial to add domain-specific early skills to the definition of school readiness and to encourage inter- ventions aimed at promoting these skills prior to elementary school. Thus, understanding which skills are linked to children’s academic achievement has important implications for early edu- cation programs.
We adopted a child-centered model of school transition, which is nested within a broader ecological framework but focuses on the set of individual skills and behaviors that children have acquired prior to school entry (Rimm-Kaufman & Pianta, 2000). A child’s individual characteristics contribute to the environments in which the child interacts and the rate at which the child may learn new skills; in turn, the child receives feedback from others in the environment (Meisels, 1998). Thus, because they affect both the child and the social environment, early academic skills and socio- emotional behaviors are linked to subsequent academic achieve- ment because they provide the foundation for positive classroom adaptation (Cunha, Heckman, Lochner, & Masterov, 2006; Entwisle, Alexander, & Olson, 2005).
For example, a child who enters kindergarten with rudimentary academic skills may be poised to learn from formal reading and mathematics instruction, receive positive reinforcement from the teacher, or be placed in a higher ability group that facilitates the acquisition of additional skills. Similarly, a child who can pay
attention, inhibit impulsive behavior, and relate appropriately to adults and peers may be able to take advantage of the learning opportunities in the classroom, thus more easily mastering reading and math concepts taught in elementary school. For these reasons, the skills children possess when entering school might result in different achievement patterns in later life. If achievement at older ages is the product of a sequential process of skill acquisition, then strengthening skills prior to school entry might lead children to master more advanced skills at an earlier age and perhaps even increase their ultimate level of achievement.
Although there are strong theoretical reasons to expect that individual differences in children’s early academic skills and be- havior are linked to subsequent behavior and achievement, sur- prisingly little rigorous research has been conducted to test this hypothesis. Consequently, the purpose of this article is to assess as precisely as possible, using six longitudinal, nonexperimental data sets, the association between skills and behaviors that emerge during the preschool years and later academic achievement. As with Robins’s (1978) classic study of adult antisocial behavior, our approach consists of comparable analyses of a number of different longitudinal developmental studies.1 We are especially interested in identifying academic, attention, and socioemotional skills and behaviors that may be learned or improved through experiences prior to school entry. In the following sections, we draw from developmental literature to identify key dimensions of school readiness and to derive theoretical predictions about how chil- dren’s school-entry skills and behaviors contribute to short- and long-term academic success.
Associations Between Early Skills and Later Achievement
Academic achievement is a cumulative process involving both mastering new skills and improving already existing skills (Entwisle & Alexander, 1990; Pungello, Kuperschmidt, Burchinal, & Patterson, 1996). Information about how children acquire read- ing and math skills points to the importance of specific academic skills but also indicates that more general cognitive skills, partic- ularly oral language and conceptual ability, may be increasingly important for later mastery of more complex reading and mathe- matical tasks. Basic oral language skills become critical for un- derstanding texts as the level of difficulty of reading passages increases (NICHD Early Child Care Research Network, 2005b; Scarborough, 2001; Snow et al, 1998; Storch & Whitehurst, 2002; Whitehurst & Lonigan, 1998). Likewise, mastery of foundational concepts of numbers allows for a deeper understanding of more complex mathematical problems and flexible problem-solving techniques (Baroody, 2003; Ferrari & Sternberg, 1998; Hiebert & Wearne, 1996).
Although children’s academic achievement is largely stable throughout childhood, children do demonstrate both transitory fluctuations and fundamental shifts in their achievement trajecto- ries (Kowaleski-Jones & Duncan, 1998; Pungello et al., 1996). Nonexperimental data show that children’s achievement test
1 Robins (1978) justified her approach as follows: “In the long run, the best evidence for the truth of any observation lies in its replicability across studies. The more the populations studied differ, the wider the historical eras they span; the more the details of the methods vary, the more convincing becomes that replication” (p. 611).
1429SCHOOL READINESS AND LATER ACHIEVEMENT
scores are related to prior cognitive functioning and the attainment of basic skills in math and literacy such as number and letter recognition (Stevenson & Newman, 1986). In their meta-analysis, La Paro and Pianta (2000) found middle-range correlations in cognitive/academic skills both from preschool to kindergarten (.43) and from kindergarten to first or second grade (.48).
Attention-related skills such as task persistence and self- regulation are expected to increase the time during which children are engaged and participating in academic endeavors. Research has shown that signs of attention and impulsivity can be detected as early as age 2.5 but continue to develop until reaching relative stability between ages 6 and 8 (Olson, Sameroff, Kerr, Lopez, & Wellman, 2005; Posner & Rothbart, 2000). Studies linking atten- tion with later achievement are less common, but consistent evi- dence suggests that the ability to control and sustain attention as well as participate in classroom activities predicts achievement test scores and grades during preschool and the early elementary grades (Alexander, Entwisle, & Dauber, 1993; Raver, Smith- Donald, Hayes, & Jones, 2005). These attention skills, which are conceptually distinct from other types of interpersonal behaviors, are associated with later academic achievement, independent of initial cognitive ability (McClelland, Morrison, & Holmes, 2000; Yen, Konold, & McDermott, 2004) and of prior reading ability and current vocabulary (Howse, Lange, Farran, & Boyles, 2003). Ex- amining attention separately from externalizing problems has clar- ified the role of each in achievement, suggesting that attention is more predictive of later achievement than more general problem behaviors (Barriga et al., 2002; Hinshaw, 1992; Konold & Pianta, 2005; Ladd, Birch, & Buhs, 1999; Normandeau & Guay, 1998; Trzesniewski, Moffitt, Caspi, Taylor, & Maughan, 2006).
Children’s socioemotional skills and behaviors are also ex- pected to affect both individual learning and classroom dynamics. Inadequate interpersonal skills promote child–teacher conflict and social exclusion (Newcomb, Bukowski, & Pattee, 1993; Parker & Asher, 1987), and these stressors may reduce children’s participa- tion in collaborative learning activities and adversely affect achievement (Ladd et al., 1999; Pianta & Stuhlman, 2004). Cor- relational evidence linking problem behaviors to academic achievement is found in the Beginning School Study. First-grade ratings on items describing a cheerful, outgoing temperament (roughly the opposite of internalizing problems) predicted adult educational attainment better than preschool or first-grade achieve- ment scores (Entwisle et al., 2005). Other studies yield similar results. For example, children with consistently high levels of aggression from ages 2–9 were more likely than other children to have achievement problems in third grade (NICHD Early Child Care Research Network, 2004).
Experimental Evidence and Crossover Effects
Many nonexperimental studies find associations between early achievement, attention, and behavior and later achievement, yet few of these studies are designed to determine which of these skills can be modified prior to school entry or whether these changes predict later achievement. In theory, intervention research should shed light on this gap by demonstrating ways to improve children’s skills and by testing whether improvements in early skills are associated with better adjustment in the long term. Indeed, a small number of experimental interventions provide encouraging evi-
dence that high-quality programs for preschool children “at risk” for school failure produce gains in cognitive and academic skills and reduce behavior problems (Conduct Problems Prevention Re- search Group, 2002; Karoly, Kilburn, & Cannon, 2005; Love et al., 2003). Early educational interventions have also been found to result in long-term reductions in special education services, grade retention, and increases in educational attainment (Campbell, Ramey, Pungello, Sparling, & Miller-Johnson, 2002; Lazar et al., 1982; Reynolds & Temple, 1998).
As is the case with nonexperimental studies, few intervention studies are designed to isolate the relative contributions of changes in achievement, attention, and behavior to later school achieve- ment. A first problem is that behavioral interventions tend to measure behavioral but not achievement outcomes, whereas read- ing and math interventions tend to measure achievement but not behavioral outcomes. Interesting exceptions are a small number of experimental behavior-based interventions that tested for achieve- ment impacts (Coie & Krehbiel, 1984; Dolan et al., 1993). For example, a random-assignment evaluation of a behavioral inter- vention targeting both aggressive and shy behaviors among first graders found short-run improvements in both teacher and peer reports of aggressive and shy behavior but no crossover impacts on reading achievement (Dolan et al., 1993; Kellam, Mayer, Rebok, & Hawkins, 1998). Given evidence, albeit limited, that behavioral interventions succeed at improving behavior but not achievement, behavior would appear to play a limited role in academic success.
A second problem is that many intervention programs target both children’s academic skills and their socioemotional behav- iors, rendering it impossible to assess their separate impacts through simple experimental contrasts. For example, the Fast Track prevention program provided a number of services to chil- dren who were identified as disruptive in kindergarten, including direct tutoring in reading skills in first grade (Conduct Problems Prevention Research Group, 1992; 2002). It is possible to estimate nonexperimental mediated models to determine whether program effects are more likely to be due to children’s improved achieve- ment, attention, or behavior skills (e.g., Reynolds, Ou, & Topitzes, 2004). This is rarely done, however.
The Present Study
This study builds on previous school readiness research in several ways. First, the scope of the study is unprecedented. We estimated a carefully specified set of models with data from six large-scale longitudinal studies, two of which were nationally representative of U.S. children, whereas two were drawn from multisite studies of U.S. children, with one each focusing on children from Great Britain and Canada. Second, we included as predictors a wide representation of school readiness indicators used in previous research and carefully distinguished between related but conceptually distinct skills (e.g., oral language vs. preliteracy skills, attention vs. externalizing problems) wherever possible. Third, we examined multiple dimensions of academic achievement outcomes, including grade completion and math and reading achievement as assessed by both teacher ratings and test scores. Fourth, we implemented rigorous analytic methods that attempted to isolate the effects of school-entry academic, attention, and socioemotional skills by controlling for an extensive set of prior child, family, and contextual influences that may have been
1430 DUNCAN ET AL.
related to children’s achievement. Finally, we assessed whether the predictive power of school readiness components differs by gender or socioeconomic status, which would indicate that some children are at heightened risk of low achievement.
We tested a number of hypotheses related to how school-entry academic, attention, and socioemotional skills are associated with later school achievement. Developmental theory suggests that chil- dren’s informal, intuitive knowledge of early language and math concepts plays an important role in the acquisition of more com- plex skills formally taught in elementary school (Adams, Treiman, & Pressley, 1998; Baroody, 2003; Griffin, Case, & Capodilupo, 1995; Tunmer & Nesdale, 1998). Theoretically, children’s atten- tion and socioemotional skills should also affect achievement because they influence children’s engagement in learning activities and facilitate (or disrupt) classroom processes (Ladd, Birch, & Buhs, 1999; Pianta & Stuhlman, 2004). However, some scholars point out that it is important to distinguish between behaviors that are directly relevant for learning, such as attention, and those that may be correlated with attention but are less likely to be directly linked with achievement, such as interpersonal skills and problem behavior (Alexander et al., 1993; Cooper & Farran, 1991; McClel- land et al., 2000; McWayne, Fantuzzo, & McDermott, 2004). Therefore, we expected early academic and attention-related skills to predict subsequent test scores and teacher achievement ratings, and we expected attention skills to predict achievement more consistently than do socioemotional behaviors.
In seeking a better understanding of the extent to which our broad set of early skills is associated with later achievement, it is important to consider how outcomes are being measured. Although test performance provides a key independent assessment of aca- demic achievement, teacher ratings also lend insight into chil- dren’s everyday performance in the classroom. Teachers’ evalua- tions are probably based on a broad picture of children’s accomplishments, which include their academic skills as well as whether they complete assignments on time, work independently, get along with others, and show involvement in the learning agenda of the classroom. Moreover, previous research has found that children’s behavior can play a role that is equal to, if not greater than, prior cognitive ability in predicting teacher-rated attainment or achievement (Lin, Lawrence, & Gorrell, 2003; Schaefer & McDermott, 1999) and academic skills (National Cen- ter for Education Statistics, 1993). Consequently, we expected a stronger relationship between school-entry socioemotional behav- iors and subsequent teacher-rated achievement than with subse- quent test scores.
Although many previous studies have examined the association between early academic, attention, and socioemotional skills and subsequent achievement, few have systematically considered the extent to which these associations differ by gender (Trzesniewski et al., 2006). On average, boys receive poorer grades and have more difficulties related to school progress (e.g., grade retention, special education, and drop out) than do girls (Dauber, Alexander, & Entwisle, 1993; McCoy & Reynolds, 1999), and these differ- ences are especially pronounced among low-income children (Hin- shaw, 1992). Children from low-income families enter school with lower mean academic skills, and the gap tends to increase during the school years (Lee & Burkam, 2002). These groups also have higher rates of problems with attention and externalizing behavior
(Entwisle et al., 2005; Miech, Essex, & Goldsmith, 2001; Raver, 2004).
Despite differences in children’s behavior linked to gender and family socioeconomic status, few studies have considered whether gender and socioeconomic status moderate the association be- tween these early skills and behaviors and subsequent achieve- ment. We expected early academic skills, attention, and socioemo- tional behaviors to matter more for these subgroups, particularly when children enter school with very low levels of these skills.
To estimate the associations between early academic skills and socioemotional behaviors and later school achievement, we sum- marize results from a coordinated series of analyses across six longitudinal data sets in two ways. First, we relate early academic, attention, and socioemotional skills to later achievement in each of the six data sets and provide a basic summary of these results. Second, we formally summarize the findings from these studies in a meta-analysis, again focusing on the extent to which this collec- tion of early skills predicts later achievement.
Method
In this section, we describe the data sets used in this study and the common analytic procedures that were implemented across studies. Detailed information about the measures, descriptive sta- tistics, and regression results from each study is presented in Appendices A–F, which can be found online. As the goal of our study was to relate early academic, attention, and socioemotional skills and behaviors to later achievement, each data set has mea- sures of these constructs, although there is variation across the studies with respect to when and how each skill or behavior is assessed.
Table 1 provides an overview of data sources and measures available in each study. All six data sets provide measures of children’s academic skills as well as assessments of attention and socioemotional behaviors at about age 5 or 6. Because most children enter elementary school at this age, we refer to the timing of these measures as school entry but alert the reader that the precise timing varies considerably across studies. To facilitate comparison of findings across studies, we standardized all mea- sures to have a mean of 0 and standard deviation of 1.
We measured achievement outcomes using teachers’ reports, test scores, and grade retention in early elementary school and, in some studies, middle childhood. In terms of the timing of the measurement of achievement outcomes, the children of the Na- tional Longitudinal Survey of Youth (NLSY) measures are as- sessed as late as early adolescence, the National Institute of Child Health and Human Development Study of Early Child Care and Youth Development (NICHD SECCYD) as late as fifth grade, and the 1970 British Birth Cohort Study (BCS) at age 10, whereas none of the other studies measures achievement beyond third grade. As for measurement methods, two studies have both test-score-based and teacher reports of reading and mathematics achievement (the Early Childhood Longitudinal Study–Kindergarten Cohort [ECLS- K] and NICHD SECCYD).
We measured attention and socioemotional behaviors on the basis of mothers’ reports, teachers’ reports, and observation. Table 1 provides an overview of the similarities and differences in measurement across the six studies. One of our data sets, the Infant Health and Development Program (IHDP), has observer reports of
1431SCHOOL READINESS AND LATER ACHIEVEMENT
T ab
le 1
St ud
y M
ea su
re s
of O
ut co
m es
, Sc
ho ol
E nt
ry an
d P
re sc
ho ol
A ch
ie ve
m en
t, A
tt en
ti on
an d
So ci
oe m
ot io
na l
B eh
av io
rs
M ea
su re
E C
L S-
K G
ra de
3 N
L SY
ag e
13 –1
4 N
IC H
D SE
C C
Y D
G ra
de 5
IH D
P ag
e 8
M L
E PS
G ra
de 3
B C
S ag
e 10
O ut
co m
es R
ea di
ng A
ch ie
ve m
en t
T es
t R
ea di
ng ite
m re
sp on
se th
eo ry
(I R
T )
“a dv
an ce
d” su
bs ca
le s:
E xt
ra po
la tio
n, E
va lu
at io
n A
ca de
m ic
R at
in g
Sc al
e (�
� .9
5) :
T ea
ch er
R ep
or t
Pe ab
od y
In di
vi du
al A
ch ie
ve m
en tT
es ts
(P IA
T )
R ea
di ng
R ec
og ni
tio n:
M at
ch in
g le
tte rs
, na
m in
g na
m es
,a nd
re ad
in g
si ng
le w
or ds
ou tl
ou d
(te st
–r et
es t�
.8 9)
W oo
dc oc
k– Jo
hn so
n Ps
yc ho
- E
du ca
tio na
lB at
te ry
— R
ev is
ed (W
J- R
) R
ea di
ng :
L et
te r-
W or
d, W
or d
A tta
ck A
ca de
m ic
Sk ill
s R
at in
gs Sc
al e:
T ea
ch er
R ep
or t
W oo
dc oc
k– Jo
hn so
n T
es ts
of A
ch ie
ve m
en t—
R ev
is ed
:B ro
ad R
ea di
ng (�
� .9
0s )
V er
ba lS
ki lls
:U se
/ un
de rs
ta nd
in g
of Fr
en ch
, ch
ar ac
te ris
tic s
of or
al co
m m
un ic
at io
n (�
� .9
4) :
T ea
ch er
R ep
or t
E di
nb ur
gh R
ea di
ng T
es t:
Se lf
-c om
pl et
io n,
w or
d re
co gn
iti on
,u se
an d
un de
rs ta
nd in
g of
sy nt
ax ,s
eq ue
nc in
g, co
m pr
eh en
si on
,a nd
re te
nt io
n (�
� .9
6)
M at
h A
ch ie
ve m
en t
T es
t M
at h
IR T
“a dv
an ce
d” su
bs ca
le s:
M ul
tip lic
at io
n/ di
vi si
on ,
Pl ac
e- va
lu e,
W or
d pr
ob le
m s
A ca
de m
ic R
at in
g Sc
al e
(� �
.9 4)
: T
ea ch
er R
ep or
t
PI A
T M
at h:
A pp
lic at
io n
of m
at he
m at
ic al
co nc
ep ts
,n um
be r
re co
gn iti
on ,c
ou nt
in g,
m ul
tip lic
at io
n, di
vi si
on ,f
ra ct
io ns
,a nd
ad va
nc ed
al ge
br a
an d
ge om
et ry
(te st
/re te
st �
.7 4)
W J-
R M
at h:
A pp
lie d
pr ob
le m
s A
ca de
m ic
Sk ill
s R
at in
gs Sc
al e:
T ea
ch er
R ep
or t
W oo
dc oc
k– Jo
hn so
n T
es ts
of A
ch ie
ve m
en t—
R ev
is ed
: B
ro ad
M at
h (�
� .9
0s )
N um
be r
K no
w le
dg e
T es
t: N
um be
r se
qu en
ce ,
ad di
tio n,
su bt
ra ct
io n,
m ul
tip lic
at io
n, di
vi si
on ,
fr ac
tio ns
,d ec
im al
s
U ni
ve rs
ity of
B ris
to lM
at h
T es
t: R
ul es
of ar
ith m
et ic
, pl
ac e
va lu
e, fr
ac tio
ns ,
m ea
su re
m en
t, al
ge br
a, ge
om et
ry ,a
nd st
at is
tic s
(� �
.9 3)
Sc ho
ol en
tr y
sk ill
s Fa
ll of
ki nd
er ga
rt en
A ge
5– 6
A ge
4. 5
A ge
5 Ju
ni or
an d
se ni
or ki
nd er
ga rte
n A
ge 5
A ch
ie ve
m en
t R
ea di
ng A
ch ie
ve m
en tT
es tR
ea di
ng IR
T “e
ar ly
” su
bs ca
le s:
L et
te r
re co
gn iti
on ,
be gi
nn in
g an
d en
di ng
w or
d so
un ds
PI A
T R
ea di
ng R
ec og
ni tio
n W
J- R
R ea
di ng
:L et
te r-
W or
d Id
en tif
ic at
io n
(� �
.8 4)
L an
gu ag
e/ ve
rb al
ab ili
ty Pr
es ch
oo l
L an
gu ag
e Sc
al e–
3: E
xp re
ss iv
e co
m m
un ic
at io
n
W ec
hs le
r Pr
es ch
oo l
an d
Pr im
ar y
Sc al
e of
In te
lli ge
nc e
(W PP
SI ):
V er
ba l
IQ (�
� .9
4)
Pe ab
od y
Pi ct
ur e
V oc
ab ul
ar y
T es
t( PP
V T
), Fo
rm s
A an
d B
,F re
nc h
ad ap
ta tio
n: Sp
lit -
ha lf
re lia
bi lit
y .6
6 an
d .8
5 fo
r A
an d
B ,r
es pe
ct iv
el y;
te st
–r et
es ta
t1 w
ee k
� .7
2
E ng
lis h
Pi ct
ur e
V oc
ab ul
ar y
T es
t: V
er ba
l in
te lli
ge nc
e
M at
h A
ch ie
ve m
en t
T es
t M
at h
IR T
“e ar
ly ”
su bs
ca le
s: C
ou nt
in g,
or di
na lit
y an
d re
la tiv
e si
ze
PI A
T M
at h
W J-
R M
at h:
A pp
lie d
pr ob
le m
s (�
� .8
4) N
um be
r K
no w
le dg
e T
es t:
In fo
rm al
nu m
be r
kn ow
le dg
e, co
un tin
g, ad
di tio
n, nu
m be
r se
qu en
ce A
tte nt
io n
sk ill
s A
tte nt
io n
sk ill
s A
pp ro
ac he
s to
L ea
rn in
g (�
� .8
9) :T
ea ch
er R
ep or
t C
on tin
uo us
Pe rf
or m
an ce
T as
k: A
tte nt
io n
A tte
nt io
n: C
hi ld
co nc
en tr
at es
, lis
te ns
at te
nt iv
el y,
et c.
(� �
.8 2)
: T
ea ch
er R
ep or
t A
tte nt
io n
pr ob
le m
s H
yp er
ac tiv
ity (d
if fi
cu lty
co nc
en tr
at in
g, re
st le
ss ,e
tc .):
M at
er na
l R
ep or
t
(I )
C hi
ld B
eh av
io r
C he
ck lis
t (C
B C
L ),
A tte
nt io
n pr
ob le
m s
(� �
.9 3)
:F al
l of
K in
de rg
ar te
n T
ea ch
er R
ep or
t (I
I) C
on tin
uo us
Pe rf
or m
an ce
T as
k: Im
pu ls
iv ity
A ch
en ba
ch C
hi ld
B eh
av io
r Pr
of ile
, A
tte nt
io n
pr ob
le m
s (�
� .6
1) :M
at er
na l
R ep
or t
H yp
er ac
tiv e:
C hi
ld se
em s
ag ita
te d,
im pu
ls iv
e, et
c. (�
� .9
0) :
T ea
ch er
R ep
or t
R ut
te r
Sc al
e: In
at te
nt io
n (�
� .6
7) :
M at
er na
l R
ep or
t
1432 DUNCAN ET AL.
T ab
le 1
(c on
ti nu
ed )
Sc ho
ol en
tr y
sk ill
s Fa
ll of
ki nd
er ga
rt en
A ge
5– 6
A ge
4. 5
A ge
5 Ju
ni or
an d
se ni
or ki
nd er
ga rt
en A
ge 5
So ci
oe m
ot io
na l
be ha
vi or
s E
xt er
na liz
in g
pr ob
le m
s E
xt er
na liz
in g
Pr ob
le m
s (�
� .9
0) :
T ea
ch er
R ep
or t
(I )
H ea
d St
ro ng
(s tu
bb or
n, st
ro ng
te m
pe r,
et c.
): M
at er
na l
R ep
or t
(I I)
A nt
is oc
ia l
(c he
at s,
bu lli
es ,e
tc .):
M at
er na
l R
ep or
t
C B
C L
,A gg
re ss
iv e
be ha
vi or
(� �
.9 3)
: Fa
ll of
K in
de rg
ar te
n T
ea ch
er R
ep or
t
A gg
re ss
io n:
C hi
ld fi
gh ts
, bu
lli es
ot he
rs ,e
tc .(
� �
.7 2)
: T
ea ch
er R
ep or
t
R ut
te r
Sc al
e, E
xt er
na liz
in g:
C hi
ld bu
lli es
ot he
rs ,i
s di
so be
di en
t, et
c. (�
� .7
2) :M
at er
na lR
ep or
t
In te
rn al
iz in
g pr
ob le
m s
In te
rn al
iz in
g Pr
ob le
m s
(� �
.8 0)
: T
ea ch
er R
ep or
t A
nx io
us /d
ep re
ss ed
(f ee
ls un
lo ve
d, sa
d) :
M at
er na
l R
ep or
t
C B
C L
: In
te rn
al iz
in g
(� �
.9 3)
: Fa
ll of
K in
de rg
ar te
n T
ea ch
er R
ep or
t
A nx
io us
/D ep
re ss
ed :
C hi
ld w
or ri
es ,c
ri es
of te
n, et
c. (�
� .8
0) :
T ea
ch er
R ep
or t
R ut
te r
Sc al
e, In
te rn
al iz
in g:
C hi
ld w
or ri
es ,s
ee m
s m
is er
ab le
,e tc
.( �
� .5
4) :
M at
er na
l R
ep or
t So
ci al
sk ill
s (I
) Se
lf C
on tr
ol (�
� .7
9) :
T ea
ch er
R ep
or t
(I I)
In te
rp er
so na
l Sk
ill s
(� �
.8 9)
: T
ea ch
er R
ep or
t
So ci
al Sk
ill s
R at
in g
Sy st
em (S
SR S)
:C oo
pe ra
tio n,
as se
rti on
,s el
f- co
nt ro
l( �
� .9
3) :F
al lo
f K
in de
rg ar
te n
T ea
ch er
Pr os
oc ia
l: C
hi ld
is he
lp fu
l, sy
m pa
th et
ic to
ot he
rs ,e
tc .
(� �
.9 2)
: T
ea ch
er R
ep or
t
Pr io
r ch
ild co
nt ro
ls A
ge 3–
4 A
ge 3
A ge
3 Ju
ni or
an d
se ni
or ki
nd er
ga rt
en A
ge 42
m on
th s
Pr io
r co
gn iti
ve /
ac hi
ev em
en t
Pe ab
od y
Pi ct
ur e
V oc
ab ul
ar y
T es
t— R
ev is
ed (P
PV T
-R ;
sp lit
-h al
f re
lia bi
lit y
of .8
0)
B ra
ck en
B as
ic Sk
ill s
(c ol
or s,
le tte
rs ,n
um be
rs ,
et c.
,� �
.9 3)
R ey
ne ll
D ev
el op
m en
ta l
L an
gu ag
e Sc
al e
(v oc
ab ul
ar y
co m
pr eh
en si
on ,�
� .9
3; ex
pr es
si ve
la ng
ua ge
,� �
.8 6)
St an
fo rd
-B in
et IQ
T es
t Pe
ab od
y Pi
ct ur
e V
oc ab
ul ar
y T
es t(
PP V
T ),
Fo rm
s A
an d
B ,F
re nc
h ad
ap ta
tio n:
Sp lit
- ha
lf re
lia bi
lit y
.6 6
an d
.8 5
fo r
A an
d B
,r es
pe ct
iv el
y; te
st –r
et es
ta t1
w ee
k �
.7 2
N um
be r
K no
w le
dg e
T es
t: In
fo rm
al nu
m be
r kn
ow le
dg e,
sh ap
es ,c
ol or
s, co
un tin
g, nu
m be
r se
qu en
ce ,
ad di
tio n
C ou
nt in
g (�
� .9
5) Sp
ea ki
ng &
V oc
ab ul
ar y
(� �
.7 9)
C op
yi ng
de si
gn s
(� �
.8 3)
A ge
22 m
on th
s: C
ub e
st ac
ki ng
,L an
gu ag
e (�
� .9
3) ,C
op yi
ng de
si gn
s te
st (�
� .4
6)
Pr io
r at
te nt
io n
sk ill
s/ so
ci oe
m ot
io na
l be
ha vi
or
C om
pl ia
nc e
(� �
.5 9)
; M
at er
na l
R ep
or t
So ci
ab ili
ty (�
� .9
3) :
In te
rv ie
w er
ra tin
gs of
ch ild
’s co
op er
at io
n du
ri ng
th e
as se
ss m
en t
C B
C L
,E xt
er na
liz in
g, in
te rn
al iz
in g:
M at
er na
l R
ep or
t Fo
rb id
de n
to y
la bo
ra to
ry ta
sk :
Im pu
ls iv
ity
C B
C L
,O ve
ra ll
be ha
vi or
pr ob
le m
s (�
� .9
4) :
M at
er na
lR ep
or t
A tte
nt io
n (�
� .9
4) :
T es
to bs
er ve
r’ s
re po
rt
C oo
pe ra
tio n:
In te
rv ie
w er
ra tin
gs of
ch ild
’s co
op er
at io
n du
ri ng
th e
as se
ss m
en t
A ge
22 m
on th
s: Pe
rs on
al de
ve lo
pm en
t (�
� .8
2)
N ot
e. E
ar ly
C hi
ld ho
od L
on gi
tu di
na l
St ud
y- K
in de
rg ar
te n
C oh
or t
(E C
L S-
K )
in cl
ud es
a sc
ho ol
en tr
y te
st of
ge ne
ra l
kn ow
le dg
e. N
at io
na l
In st
itu te
of C
hi ld
H ea
lth an
d H
um an
D ev
el op
m en
t St
ud y
of E
ar ly
C hi
ld C
ar e
an d
Y ou
th D
ev el
op m
en t
(N IC
H D
SE C
C Y
D )
in cl
ud es
ag e
4. 5
W J-
R C
og ni
tiv e
A bi
lit y
m ea
su re
.I nf
an t
H ea
lth an
d D
ev el
op m
en t
Pr og
ra m
(I H
D P)
in cl
ud es
ag e
5 W
ec hs
le r
Pr es
ch oo
l an
d Pr
im ar
y Sc
al e
of In
te lli
ge nc
e (W
PP SI
) Pe
rf or
m an
ce IQ
(� �
.9 3)
an d
ag e
5 A
ch en
ba ch
C hi
ld B
eh av
io r
Pr of
ile M
at er
na l
R ep
or t:
O ve
ra ll
be ha
vi or
pr ob
le m
s (�
� .9
3) .B
ri tis
h B
ir th
C oh
or t
St ud
y (B
C S)
in cl
ud es
ag e
5 H
um an
Fi gu
re D
ra w
in g,
C op
yi ng
D es
ig ns
T es
t, an
d Pr
of ile
D ra
w in
g. R
om an
nu m
er al
s in
de x
m ea
su re
s co
ns is
te nt
ly in
T ab
le s
1 an
d 2.
N L
SY �
N at
io na
l L
on gi
tu di
na l
Su rv
ey of
Y ou
th ;
M L
E PS
� M
on tr
ea l
L on
gi tu
di na
l- E
xp er
im en
ta l
Pr es
ch oo
l St
ud y.
1433SCHOOL READINESS AND LATER ACHIEVEMENT
attention, another (NICHD SECCYD) has both test-based and teacher-rated measures of attention, and three (NLSY, IHDP, and BCS) have parent rather than teacher reports of socioemotional behaviors. In addition, two of the studies (NICHD SECCYD and the Montreal Longitudinal-Experimental Preschool Study [MLEPS]) measure both attention skills and problems, whereas three (NLSY, IHDP, and BCS) have measures of attention prob- lems but not skills, and one study (ECLS-K) has a measure of attention skills but not attention problems. In addition, with one exception, all of the studies provide measures of academic, atten- tion, and socioemotional skills prior to the point of school entry, which we used as key control variables in our analyses.
The Studies and Samples
The Early Childhood Longitudinal Study–Kindergarten Cohort (ECLS-K). The ECLS-K follows a nationally representative sam- ple of 21,260 children who were in kindergarten in 1998–1999. We used data from kindergarten, first grade, and third grade. Data were collected from multiple sources, including direct achieve- ment tests of children and surveys of parents, teachers, and school administrators (see Table 1; National Center for Education Statis- tics, 2001).
Achievement tests were administered in the fall of kindergarten and in the spring of kindergarten, first grade, and third grade. We used teacher reports of children’s “approaches to learning” (which measure both attention skills and achievement motivation) and socioemotional behaviors, including internalizing and externaliz- ing problems, self-control with peers, and interpersonal skills, collected in the fall and spring of kindergarten.
The battery of achievement tests given as part of the ECLS-K kindergarten and first-grade assessments covered three subject areas: language and literacy, mathematical thinking, and general knowledge. For third grade, the achievement tests included math- ematics, reading, and science. We used item response theory scores for the first two of these as key dependent variables. These third-grade assessments required students to complete workbooks and open-ended mathematics problems. As detailed in Appendix A, a host of family- and some child-level controls are available in the data.
The children of the National Longitudinal Survey of Youth (NLSY). The NLSY is a multistage stratified random sample of 12,686 individuals age 14 to 21 in 1979 (Center for Human Resource Research, 2004). Black, Hispanic, and low-income youth were overrepresented in the sample. Annual (through 1994) and biennial (between 1994 and 2000) interviews with sample mem- bers and very low cumulative attrition in the study contribute to the quality of the study’s data.
Beginning in 1986, the children born to NLSY female partici- pants were tracked through biennial mother interview supplements and direct child assessments. Given the nature of the sample, it is important to note that early cohorts of the child sample were born disproportionately to young mothers. With each additional cohort, the children become more representative of all children, and NLSY children younger than age 14 in 2000 share many demographic characteristics of their broader set of age mates.
The sample used in the present analysis consists of 1,756 chil- dren whose academic achievement was tracked from age 7–8 to age 13–14 and whose achievement and behavior was assessed at
age 5–6. Consequently, our sample comprises children who were age 5 or 6 in 1986, 1988, 1990, or 1992. The age 13–14 achieve- ment and behavior of these children were assessed in the respec- tive 1994, 1996, 1998, and 2000 interviews.
School readiness measures, including math and reading test scores (Peabody Individual Achievement Test; Dunn & Mark- wardt, 1970) and maternal reports of children’s behavior problems (adapted from the Achenbach Behavior Problems Checklist; Baker, Keck, Mott, & Quinlan, 1993) were collected at age 5 or age 6. Academic achievement outcome measures were collected biennially for children between the ages of 5 and 14. In addition, key control variables include children’s receptive vocabulary (Pea- body Picture Vocabulary Test—Revised; Dunn & Dunn, 1981) and children’s temperament (compliance and sociability) at age 3 or 4. Additional family- and child-level control variables are described in Appendix B.
The NICHD Study of Early Child Care and Youth Development (SECCYD). Longitudinal data from the NICHD SECCYD are drawn from a multisite study of births in 1991 (NICHD Early Child Care Research Network, 2005a). Participants were recruited from hospitals located at 10 sites across the United States. During 24-hr sampling periods, 5,265 new mothers met the selection criteria and agreed to be contacted after returning home from the hospital. At 1 month of age, 1,364 healthy newborns were enrolled in the study. Although it is not nationally representative, the study sample closely matches national and census tract records with respect to demographic variables such as ethnicity and household income. The majority of children in the sample are White, 12% are African American, and 11% are Hispanic or of another ethnicity. About 30% of mothers had a high school education or less, and 14% were single parents (NICHD Early Child Care Research Network, 1997). The analysis sample had valid data on the achievement outcome measures and at least three sources of in- formation on the key independent variables (approximately 981 at first grade, 928 at third grade, and 907 at fifth grade).
School readiness measures, including achievement tests and attention/impulsivity tasks, were administered in a controlled lab- oratory setting at age 4.5, and attention problems, aggression, internalizing behavior, and social skills were measured by teacher report in the fall of the kindergarten year. Outcomes at first, third, and fifth grades include achievement in math and reading accord- ing to teacher ratings and Woodcock–Johnson Tests of Achieve- ment—Revised test scores (Woodcock & Johnson, 1990; see Table 1). Key control variables at age 3 include children’s cognitive ability, language skills, impulsivity, externalizing problems, and internalizing problems. The NICHD SECCYD also collects infor- mation from infancy about children’s early environments, includ- ing child-care type and quality, home environment, and parenting; these and other child- and family-level covariates are described in Appendix C.
The Infant Health and Development Program (IHDP). The IHDP is an eight-site randomized clinical trial designed to evaluate the efficacy of a comprehensive early-intervention program for low birth weight (LBW) premature infants. Infants weighing 2,500 g (5.51 lb) or less at birth were screened for eligibility if their postconceptional age between January and October 1985 was 37 weeks or less and if they were born in one of eight participating medical institutions. A total of 985 infants was randomly assigned either to a medical follow-up only or to a comprehensive early
1434 DUNCAN ET AL.
childhood intervention group immediately following hospital dis- charge.
Infants in both the comprehensive early childhood intervention and medical follow-up only groups participated in a pediatric follow-up program of periodic medical, developmental, and famil- ial assessments from 40 weeks of conceptional age (when they would have been born if they had been full term) to 36 months of age corrected for prematurity. The intervention program, lasting from hospital discharge until 36 months, consisted of home visits, child-care services, and parent group meetings. A coordinated educational curriculum of learning games and activities was used both during home visits and at the center.
The primary analysis group consisted of 985 infants. Of these 985 infants, cognitive assessments are available for 843 children at age 3, 745 children at age 5, and 787 children at age 8. In addition, 76 children who were born at an extremely low birth weight (ELBW; 1,000 g [3.27 lb] or less) were excluded from the sample because ELBW children differ markedly from other LBW children in cognitive and behavioral functioning (Klebanov, Brooks-Gunn, & McCormick, 1994a, 1994b). Thus, this study focuses on a subsample of 690 children who were not born ELBW and for whom cognitive assessment and family background data were available.
Data come from a variety of sources: questionnaires, home visits, and laboratory tests (see Table 1). School readiness mea- sures include preschool performance and verbal test scores, paren- tal reports of children’s mental health and aggressive behavior, and observer reports of children’s attention and task persistence. We assessed reading and math achievement using the Woodcock– Johnson Tests of Achievement—Revised broad reading and math tests and the Wechsler Intelligence Scale for Children—Third Edition (Wechsler, 1991) performance and verbal tests at 8 years of age. Key control variables include cognitive ability, sustained attention, and behavior problems at age 3. Additional family- and child-level control variables are described in Appendix D.
The Montreal Longitudinal-Experimental Preschool Study (MLEPS). The MLEPS comprises several consecutive cohorts launched from 1997 to 2000. The original sample of 4- and 5-year-old children (N � 1,928), representing one third of its population base, was obtained after a multilevel consent process involving school board administrators, local school committees, parents, and teachers. Given that its final cohort (2000) does not meet all the data requirements for the research objective examined here, we limited ourselves to the sample of children beginning kindergarten in the fall of 1998 and the fall of 1999.
Incomplete data reduced the sample from 1,369 to 767 children. Students in the final sample had a valid value on any of the four outcome measures of interest (first- and third-grade achievement measures) and on at least four of the six socioemotional measures. Of the 767 participants in the final sample, 439 began kindergarten in 1998 and 328 began kindergarten in the fall of 1999. Addition- ally, for 350 of the 767 students, initial data were collected during the fall of junior kindergarten (332 who began junior kindergarten in 1997 and 18 who began junior kindergarten in 1998).
Initial and follow-up data were collected from multiple sources, including direct cognitive assessments of children and surveys of parents and teachers. Early academic assessments include individ- ually administered number knowledge and receptive vocabulary tests at the end of senior kindergarten. Teachers rated children’s
behavioral development, including physically aggressive, anxious, depressive, hyperactive, inattentive, and prosocial behavior. Third- grade assessments include a group-administered math test and teacher ratings of children’s French language skills (see Table 1). Key control variables include number knowledge and vocabulary measured on entry into junior kindergarten (age 4) for Cohort 1 and on entry into senior kindergarten (age 5) for Cohort 2. Addi- tional family- and child-level control variables are detailed in Appendix E.
The 1970 British Birth Cohort Study (BCS). The U.K. 1970 BCS, a nationally representative longitudinal study, has followed into adulthood a cohort of children born in Great Britain during 1 week in 1970 (Bynner, Ferri, & Shepherd, 1997). The birth sample of 17,196 infants was approximately 97% of the target birth population. Attrition has reduced the original sample to 11,200 participants. Nevertheless, the representativeness of the original birth cohort has largely been maintained, although the current sample is disproportionately female and highly educated (Ferri & Smith, 2003). Missing data on key variables reduce the sample size for most analyses to between 9,000 and 10,000 cases.
At each wave, cohort members were given a battery of tests of intellectual and behavioral development (see Table 1). School readiness measures include vocabulary and copying skills tests, and maternal reports of attention, externalizing behavior, and internalizing behavior were collected when the children were 5 years of age. Reading and mathematics achievement tests were administered at age 10. Key control variables include measures of basic skills and behavior at ages 22 and 42 months for a 10% subsample of the data. Additional family- and child-level controls are described in Appendix F.
Analysis Plan
We begin our analysis by estimating a similar set of regression models across all six studies, in which school-entry academic, attention, and socioemotional skills are related to later academic achievement. For example, in ECLS-K data, the school-entry skills and behaviors are measured in the fall of kindergarten (referred to hereafter as FK), whereas math and reading achievement are measured in the spring of third grade (referred to hereafter as 3rd). The resulting equation is as follows:
ACH i3rd � a1 � �1ACADiFK � �2ATTNiFK � �3SEiFK
� �1FAMi � �2CHILDi � eit, (1)
where ACHi3rd is the math or reading 2 achievement of child i in
the spring of third grade; ACADiFK is the collection of math, reading, and general knowledge skills that child i has acquired at school entry, assessed by achievement tests in the fall of the kindergarten year; ATTNiFK is a teacher-reported measure of attention; SEiFK is the collection of socioemotional skills that child i’s teacher reports; FAMi and CHILDi are sets of family back- ground and child characteristics, respectively, included in analyses to control for individual differences that might influence child achievement before and after school entry; a1 is a constant; and eit is a stochastic error term.
2 We use reading as shorthand for the set of reading, language, and verbal ability skills measured in our data sets.
1435SCHOOL READINESS AND LATER ACHIEVEMENT
Our interest is in estimating �1, �2, and �3, which, if correctly modeled, can be interpreted as the impact of school-entry aca- demic, attention, and socioemotional skills on subsequent achieve- ment. A key challenge in this approach is ensuring that we have accounted for the possibility of omitted variable bias, which is likely to arise if unobserved family or child characteristics are correlated with both children’s school entry skills and their later achievement. Our principal strategy for securing unbiased estima- tion of �1, �2, and �3 is to estimate a model of the form of Equation 1 that includes as many prior measures of relevant child and family characteristics as possible.
All of the studies contain important measures of child and family characteristics that may be confounded with children’s achievement, attention, and behavior. Although the specific set of covariates varies across studies, most studies include measures of the child’s race and ethnicity, maternal education, family structure, and family income or economic well-being. In some studies, measures of child health, maternal depressive symptoms, parent- ing, and quality of the home environment, as well as children’s participation in early child care and education during early child- hood were also included as controls.3 Details about the specific controls used in each study are provided in the appendices, and a complete list of covariates for each study can be found in Tables A6, B6, C6, D5, E6, and F5.
Our analysis was designed to examine the relations between early skills and later achievement, irrespective of the characteris- tics of the classroom/school the child attends. The ECLS-K is an exception, owing to a sample design that selects an average of 4 students per kindergarten classroom to be enrolled in the study. We took advantage of this classroom clustering by adjusting our ECLS-K estimates for classroom fixed effects. Thus, all of the variation used in the regression stems from within-classroom dif- ferences, which holds constant school and classroom characteris- tics.
Of course, we cannot be certain that even a comprehensive set of control variables captures all of the important confounds, which leaves open the possibility that this approach will still produce biased estimates of �1, �2, and �3. For example, an obvious bias of this sort would arise if scores on a kindergarten mathematics test reflected both math skills and underlying cognitive ability.
To further reduce the possibility of biases, we include measures of a child’s attention and socioemotional behaviors and either cognitive ability or preacademic skills assessed prior to school- entry, which are available in all but two studies (ECLS-K and MLEPS).4 With these prior measures, our model becomes
ACH i3rd � a1 � �1ACADiFK � �2ATTNiFK � �3SEiFK
� �4ACADiPre-FK � �5ATTNiPre-FK � �6SEiPre-FK
� �1FAMi � �2CHILDi � eit. (2)
ACADiPre-FK, ATTNiPre-FK, and SEiPre-K refer to child i’s respec- tive achievement, attention, and socioemotional behavior prior to school entry, respectively. This constitutes a particularly powerful version of Equation 1, because controlling for the child’s cognitive and behavioral skills before school entry should reduce, if not eliminate, omitted-variable bias in �1, �2, and �3.
One concern about Equation 2 is that by controlling for school entry achievement, we might reduce the deserved explanatory
power of attention and socioemotional skills. This would occur if one of the ways in which attention and socioemotional skills affected later achievement were to raise children’s school entry academic skills. We investigate this possibility by estimating ver- sions of Equation 2 that omit ACADiFK.
In the second step of our analysis, we use meta-analytic tech- niques to summarize coefficients obtained from our six studies’ estimates of Equation 2 and seek to determine whether particular study characteristics are associated with larger (or smaller) coef- ficients. More specifically, the meta-analysis treats the standard- ized regression coefficients from Equation 2 as observations in a regression predicting academic achievement measured as late in childhood as possible. Independent variables in the meta-analytic regression include (a) the type of school-entry measure,5 (b) elapsed time (scaled in years) between measurement of school- entry characteristics and the outcome, (c) whether the outcome is math or reading achievement, and (d) whether the outcome is based on a test or a teacher report. In keeping with standard meta-analytic practices, we weighted each regression coefficient observation by the inverse of its variance (Hedges & Olkin, 1985).
Results
Regression Results
To consider whether school entry6 achievement, attention, and socioemotional skills are predictive of subsequent achieve- ment, we first estimated a comparable set of regressions (Equa- tion 2) across all of the studies. For each study, reading and math outcomes measured as late in the data set as possible were regressed on school-entry achievement, attention, and socio- emotional behaviors, with controls for important family and child characteristics also included in the regression. In all but two cases, our regressions include measures of both cognitive ability and either attention or socioemotional behaviors.
3 Our list of child and family control variables is more extensive than in most developmental studies. In selecting these variables, we were careful to include only variables measured prior to or concurrently with our school-entry measures of achievement and behavior. We were also mindful that added controls might introduce multicolinearity into our regression estimation, but there was no indication that this might be the case. And finally, our appendix tables compare models run with and without our child and family controls and show that the results of our analyses depend little on adjustments for these factors; concurrent controls for the other achieve- ment and behavioral measures matter much more.
4 The MLEPS provides preschool cognitive measures but not attention or socioemotional behaviors (see Table 1).
5 The decision of which type of measure should serve as the omitted dummy variable category is noteworthy, because the coefficients on the included measure categories represent differences from the omitted cate- gory. We selected internalizing behavior problem coefficients as the omit- ted category because the simple average of their regression coefficients was very close to zero (–. 01 for reading outcomes and –. 01 for math outcomes).
6 We remind the reader we use the term school entry somewhat loosely. It refers to age 5 in four cases, age 5–6 in one case, and the fall of the kindergarten year in only one case.
1436 DUNCAN ET AL.
Standardized regression coefficients and standard errors from models predicting achievement from the school-entry academic, attention, and socioemotional behaviors are presented in Table 2. Complete regression results using all available reading and math outcomes are presented in appendix tables and are sum- marized below in our meta-analysis.
As expected, the regression results indicate that school-entry read- ing and math skills are almost always statistically significant predic- tors of later reading and math achievement, with standardized coef- ficients ranging from .05 to .53. Not surprisingly, school-entry reading skills predict subsequent reading achievement better than subsequent math achievement, just as early math skills are more predictive of later math than reading achievement.
In the case of attention skills and attention problems, coeffi- cients are usually smaller than those for math skills, but they are statistically significant for more than half of the coefficients. In contrast, coefficients for socioemotional behaviors—externalizing and internalizing behavior problems and social skills—rarely pass the threshold of statistical significance.
This general pattern—relatively strong prediction from school-entry reading and math skills, moderate predictive power for attention skills, and few to no statistically significant coefficients on socioemotional behaviors—is also found for reading and math achievement measured at earlier points in the studies and in logistic regressions in which grade retention is the dependent variable (results shown in Tables A3, B3, C3, D3, E3, and F3 but not in Table 2).
To consider whether the effects of school-entry skills differ by children’s gender or socioeconomic status (SES), we ran regressions (Equation 2) but also included gender interactions with school-entry achievement, attention, and socioemotional skills for all six data sets and SES interactions for all but the NICHD SECCYD and MLEPS (results shown in Appendices A–F).7 In the case of gender interac- tions, 10 of 76 relevant interaction coefficients were .05 or larger and statistically significant, but there was no consistent pattern in the direction of effects. In the case of SES interactions, only 2 of the 30 interaction coefficients were .05 or greater and statistically significant. It appears that the influences of school-entry achievement, attention, and socioemotional skills are broadly similar for both boys and girls and for children from both low- and high-SES families.8
Meta-Analytic Results
To summarize findings across the six studies more system- atically, we first averaged the 102 bivariate correlations be- tween school-entry achievement, attention, and socioemotional skills and the latest available reading and math achievement available in each of the data sets. (Detailed correlation tables are shown in Tables A2, B2, C2, D2, E2, and F2.) As shown in the first column of Table 3, the absolute value of these corre- lations average between .40 and .50 for school-entry reading and math achievement, average .25 for the collection of atten- tion measures, and average between .10 and .21 in absolute value for the three sets of socioemotional measures.
Next, we conducted a formal meta-analysis of the standardized regres- sion coefficients emerging from the individual study regressions. We used two sets, with the first comprising the 102 coefficients shown in Table 2, and drew from regressions based on Equation 2 and achievement
outcomes measured as late in childhood as possible. These results are shown in the second column of Table 3. The second meta-analytic regression is based on the 228 coefficients taken from regressions with outcomes measured at all possible points in a given study. These coeffi- cients are shown in the appendix tables and produce the results shown in the third, fourth, and fifth columns of Table 3.
A clear conclusion from the first meta-analytic regression is that only three of the school-entry skill categories predict subsequent reading and math achievement: reading/language, math, and attention. Moreover, rudimentary mathematics skills appear to matter the most, with an average standardized coefficient of .33.9 The association of reading skill with later achievement was less than half as large (.13), and, at .07, the average standardized coefficients on the attention-related measures was less than one quarter the size of the mean math-skills coefficient. As expected from Table 2, the meta- analysis results confirm that behavior problems and social skills are not associated with later achievement, holding constant achievement as well as child and family characteristics. Indeed, none had a stan- dardized coefficient that averaged more than .01 in absolute value.
Turning to the other coefficients listed in the second column of Table 3, one can see that the school-entry skills coefficients decreased a little (.010 per year) with each additional year between school entry and the point of assessment of the math or reading outcome. As for whether teacher-report outcomes or direct skill assessments are more likely to be predicted by early skills, our meta-analytic results suggest that both types of measures performed about the same.
Our appendices (Tables A3, B3, C3, D3, E3, and F3) provide standardized coefficients from regressions of achievement outcomes measured at different ages, which constitute the 228 observations used for the meta-analytic regression results shown in the remaining col- umns of Table 3. The advantage of using outcomes at several ages is that it enables us to control for the study from which a given stan- dardized coefficient was estimated, which we do by including a set of study indicator variables.10 The third column of Table 3 summarizes the results across domains, and the fourth and fifth columns show
7 Coefficients were excluded from our summary calculations if the standard errors for the gender or SES interactions were too large to detect differences of .15.
8 The ECLS-K, NLSY, and IHDP studies provided enough observa- tions on Black and White children to enable us to test for race inter- actions as well. We found no consistent evidence of race-based inter- actions.
9 Technically speaking, the .33 coefficient reflects the regression- adjusted difference between the average school-entry math and the omitted-group internalizing problem behavior standardized coefficient. Recall that the simple average of regression coefficients on internaliz- ing behavior problems was –.01 for both math and reading outcomes. The large sample sizes available in the ECLS-K push weighted meta- analytic results closer to the ECLS-K study coefficients than when the coefficient-variance weights are not used. Unweighted, the coefficients on school-entry reading and math are .12 and .20, respectively; the coefficient on attention is .05, the coefficient on externalizing is .00, and the coefficient on social skills is –.01.
10 With study dummies in the regression, we have what amounts to a fixed-effects regression in which coefficients are averaged within rather than across studies. Dropping the study dummies produced few changes in the remaining coefficients.
1437SCHOOL READINESS AND LATER ACHIEVEMENT
T ab
le 2
C oe
ff ic
ie nt
s F
ro m
R eg
re ss
io ns
of In
di vi
du al
St ud
y A
ch ie
ve m
en t
O ut
co m
es on
Sc ho
ol E
nt ry
A ch
ie ve
m en
t, A
tt en
ti on
, an
d So
ci oe
m ot
io na
l B
eh av
io rs
F ro
m M
od el
s W
it h
F ul
l C
on tr
ol s In
de pe
nd en
t va
ri ab
le
E C
L SK
G ra
de 3
N L
SY ag
e 13
–1 4
ac hi
ev em
en tt
es t
sc or
e
N IC
H D
SE C
C Y
D G
ra de
5 IH
D P
ag e
8 ac
hi ev
em en
t te
st sc
or e
M L
E PS
G ra
de 3
te ac
he r-
ra te
d ac
hi ev
em en
t te
st
B C
S ag
e 10
ac hi
ev em
en t
te st
sc or
e A
ch ie
ve m
en t
te st
sc or
e T
ea ch
er -r
at ed
ac hi
ev em
en t
A ch
ie ve
m en
t te
st sc
or e
T ea
ch er
-r at
ed ac
hi ev
em en
t
R ea
di ng
M at
h R
ea di
ng M
at h
R ea
di ng
M at
h R
ea di
ng M
at h
R ea
di ng
M at
h R
ea di
ng M
at h
V er
ba l
M at
h R
ea di
ng M
at h
Sc ho
ol en
tr y
A ch
ie ve
m en
t R
ea di
ng .1
8* *
*
(. 01
) .0
5* *
*
(. 01
) .1
5* *
*
(. 02
) .0
9* *
*
(. 02
) .1
6* *
(. 02
) .1
2* *
(. 02
) .1
9* *
*
(. 04
) .1
1* *
(. 04
) .0
9*
(. 04
) .1
1* *
(. 04
) L
an gu
ag e/
V er
ba l
A bi
lit y
.1 1*
*
(. 04
) .1
2* *
(. 04
) .2
0* *
*
(. 04
) .1
2*
(. 05
) .3
0* *
*
(. 05
) .3
4* *
*
(. 05
) .1
9*
(. 09
) .0
5 (.
08 )
.1 3*
* *
(. 03
) .0
9* *
(. 03
) M
at h
.2 7*
* *
(. 01
) .5
3* *
*
(. 01
) .3
1* *
*
(. 02
) .3
4* *
*
(. 02
) .1
1* *
(. 02
) .2
2* *
(. 02
) .0
7 (.
04 )
.1 9*
* *
(. 04
) .1
2* *
(. 04
) .1
2*
(. 05
) .0
7 (.
06 )
.2 9*
* *
(. 05
) A
tte nt
io n
sk ill
s .0
4* *
(. 01
) .1
0* *
*
(. 01
) .1
4* *
*
(. 02
) .1
2* *
*
(. 02
) �
.0 2
(. 03
) .0
0 (.
03 )
.0 5
(. 03
) .0
8*
(. 03
) .2
0* *
*
(. 06
) .1
4* *
(. 05
) A
tte nt
io n
pr ob
le m
s (I
) �
.0 5*
(. 02
) �
.0 9*
*
(. 03
) �
.1 1*
*
(. 04
) �
.1 7*
* *
(. 04
) �
.1 1*
*
(. 04
) �
.1 7*
* *
(. 05
) �
.0 5
(. 03
) �
.0 9*
(. 04
) .1
2 (.
07 )
� .1
1*
(. 06
) �
.0 8*
*
(. 03
) �
.0 9*
*
(. 03
) A
tte nt
io n
pr ob
le m
s (I
I) .0
5 (.
03 )
.0 6*
(. 03
) .0
3 (.
04 )
.0 1
(. 04
) So
ci oe
m ot
io na
l be
ha vi
or s
E xt
er na
liz in
g pr
ob le
m s
(I )
.0 0
(. 01
) �
.0 1
(. 01
) .0
0 (.
02 )
.0 1
(. 02
) .0
3 (.
02 )
.0 0
(. 03
) .0
9*
(. 04
) .0
5 (.
04 )
.1 1*
* *
(. 04
) .1
1* *
*
(. 04
) �
.0 3
(. 06
) �
.0 6
(. 05
) �
.0 3
(. 03
) .0
0 (.
03 )
E xt
er na
liz in
g pr
ob le
m s
(I I)
� .0
5*
(. 02
) �
.0 2
(. 03
) In
te rn
al iz
in g
pr ob
le m
s .0
0 (.
01 )
.0 0
(. 01
) �
.0 1
(. 01
) �
.0 2
(. 01
) .0
1 (.
02 )
.0 2
(. 03
) .0
3 (.
03 )
.0 5
(. 03
) .1
2* *
*
(. 03
) .0
4 (.
04 )
.0 0
(. 05
) .0
3 (.
04 )
.0 1
(. 03
) .0
2 (.
03 )
So ci
al sk
ill s
(I )
.0 1
(. 02
) .0
0 (.
02 )
.0 1
(. 02
) .0
1 (.
02 )
.0 6
(. 04
) .0
3 (.
04 )
.1 2*
* *
(. 04
) .0
6 (.
04 )
.0 4
(. 05
) .0
0 (.
04 )
So ci
al sk
ill s
(I I)
.0 2
(. 02
) �
.0 2
(. 01
) .0
1 (.
02 )
� .0
1 (.
02 )
Pr io
r co
nt ro
ls
C on
tr ol
s fo
r pr
io r
co gn
iti ve
/ ac
hi ev
em en
t In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
.
C on
tr ol
s fo
r pr
io r
at te
nt io
n/ so
ci oe
m ot
io na
l be
ha vi
or s
In cl
. In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
. In
cl .
O th
er ba
ck gr
ou nd
m ea
su re
s In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
. In
cl .
In cl
. O
bs er
va tio
ns 10
,7 79
10 ,8
33 8,
77 6
8, 64
7 1,
75 6
1, 75
6 90
7 90
7 85
9 85
1 69
0 69
0 47
7 56
5 1,
77 8
1, 75
3 N
um be
r of
cl as
sr oo
m s
(f ix
ed ef
fe ct
s) 2,
57 9
2, 58
6 2,
29 2
2, 28
0 R
2 .4
4 .5
0 .3
9 .3
2 .6
4 .3
6 .4
2 .4
3 .3
6 .2
9 .5
7 .5
4 .2
3 .2
8 .4
5 .4
4
N ot
e. St
an da
rd er
ro rs
ar e
in pa
re nt
he se
s. E
ar ly
C hi
ld ho
od L
on gi
tu di
na lS
tu dy
–K in
de rg
ar te
n C
oh or
t( E
C L
S- K
) re
gr es
si on
s in
cl ud
e a
sc ho
ol en
tr y
te st
of ge
ne ra
lk no
w le
dg e.
N at
io na
lI ns
tit ut
e of
C hi
ld H
ea lth
an d
H um
an D
ev el
op m
en t
St ud
y of
E ar
ly C
hi ld
C ar
e an
d Y
ou th
D ev
el op
m en
t (N
IC H
D SE
C C
Y D
) re
gr es
si on
s in
cl ud
e ag
e 4.
5 W
oo dc
oc k–
Jo hn
so n
Ps yc
ho -E
du ca
tio na
l B
at te
ry —
R ev
is ed
C og
ni tiv
e A
bi lit
y m
ea su
re .I
nf an
tH ea
lth an
d D
ev el
op m
en tP
ro gr
am (I
H D
P) re
gr es
si on
s in
cl ud
e ag
e 5
W ec
hs le
r Pr
es ch
oo la
nd Pr
im ar
y Sc
al e
of In
te lli
ge nc
e Pe
rf or
m an
ce IQ
an d
ag e
5 A
ch en
ba ch
C hi
ld B
eh av
io r
Pr of
ile M
at er
na l
R ep
or t:
O ve
ra ll
be ha
vi or
pr ob
le m
s. B
ri tis
h B
ir th
C oh
or t
St ud
y (B
C S)
re gr
es si
on s
in cl
ud e
ag e
5 H
um an
Fi gu
re D
ra w
in g,
C op
in g
D es
ig ns
T es
t, an
d Pr
of ile
D ra
w in
g. In
cl .
� se
t of
m ea
su re
s in
cl ud
ed in
th e
re gr
es si
on ;
N L
SY �
N at
io na
l L
on gi
tu di
na l
Su rv
ey of
Y ou
th ;
M L
E PS
� M
on tr
ea l
L on
gi tu
di na
l- E
xp er
im en
ta l
Pr es
ch oo
l St
ud y.
* p
� .0
5. *
* p
� .0
1. *
* * p
� .0
01 .
1438 DUNCAN ET AL.
results separately for reading and math outcomes. To provide a visual representation of the data underlying the meta-analysis, we plot the 114 coefficients for reading outcomes in Figure 1 and the 114 coef- ficients for math outcomes in Figure 2.
The general pattern of results found for outcomes measured later in childhood also holds when we also consider the broader set of regression coefficients, which includes outcomes measured earlier in childhood (column 3 of Table 3). With an average standardized coefficient of .34, school-entry math skills are most predictive of subsequent achievement outcomes, followed by reading skills (.17) and attention-related measures (.10). None of the socioemo- tional behavior categories show predictive power.
Results in columns 4 and 5 of Table 3 confirm that early reading skills are stronger predictors of later reading achievement than of later math achievement. Less expected are the fourth column’s results showing that early math skills are as predictive of later reading achievement as are early reading skills. Children’s attention skills appear to be equally important (and socioemotional behaviors equally unimportant) for reading and math achievement. The separate regres- sions for reading and math skills also show that the association between school-entry skills and later achievement declines more quickly over time for reading than for math outcomes.
As before, we find no overall difference in the size of the standard- ized coefficients depending on whether the outcome being predicted
is based on teacher reports or direct skill assessments. To answer the more specific question of whether shared method variance might lead to more pronounced associations between school-entry achievement test scores and later achievement (as opposed to school-entry teacher- reported achievement) outcomes, we added to the meta-analytic re- gressions interactions between the categories of school entry skills and dummy variables for type of achievement measure (results not shown). Surprisingly, we found no evidence that the impact of early reading and math skills mattered more for test-based than for teacher- reported outcomes.11 Thus, shared method variance does not appear to be biasing our results.
Because some of our skills groupings are quite broad, we explored whether they might be concealing systematic differences among more specific skills. For example, our reading category
11 None of the interactions between type of achievement assessment and school-entry reading and math skills was statistically significant. We did find one statistically significant interaction—between the school-entry assessments of attention and the mode of outcome assessment. The average coefficient on attention skills was nearly twice as large for teacher reports of reading and math achievement (.13) as for reading and math test scores (.07). However, this result does not bear on the issue of whether the explanatory power of school-entry achievement test scores is artificially high owing to shared method variance.
Table 3 Average Correlations and Meta-Analytic Regression Results for the Standardized Coefficients From the Six Data Sets
Independent variable
Zero-order correlation coefficients
Most recent reading and
math outcomes Reading and math
All observed outcomes
Reading Math
School-entry measure Reading .44 .13*** (.01) .17*** (.03) .24*** (.03) .10*** (.02) Math .47 .33*** (.06) .34*** (.04) .26*** (.02) .42*** (.04) Attention skills .25 .07* (.02) .10*** (.01) .08*** (.02) .11*** (.02) Externalizing problems �.14 .01 (.00) .01 (.01) .01 (.02) .01 (.01) Internalizing problemsa �.10 — — — — Social skills .21 �.01 (.01) �.01 (.01) �.00 (.02) �.01 (.01)
Time (years between school entry measure and outcomes)
�.010*** (.001) �.009 (.005) �.012** (.005) �.005 (.005)
Outcome source Test score .00 (.01) �.00 (.02) �.01 (.02) .01 (.02) Teacher reporta — — — —
Outcome subject Math .01** (.00) �.00 (.02) Readinga — — — —
Data set ECLS-Ka — — — NLSY �.01 (.03) .00 (.03) �.02 (.03) NICHD SECCYD �.01 (.02) �.01 (.02) �.01 (.02) IHDP .03 (.05) �.01 (.03) .08 (.09) MLEPS �.03 (.02) �.05 (.03) �.03 (.02) BCS �.00 (.02) �.01 (.03) .01 (.02)
Observations 102 102 228 114 114 R2 .75 .74 .80 .86
Note. All coefficients used in these analyses come from the individual study regressions that include full controls. Column 1 shows the simple average correlation between the given measure and the most recent math and reading outcomes. Model 2 standard errors are corrected for within-study clustering using Huber–White methods. Regression coefficient observations are weighted by the inverse of their variances. Robust standard errors are in parentheses. Dashes indicate that this category was omitted. ECLS-K � Early Childhood Longitudinal Study–Kindergarten Cohort; NLSY � National Longitudinal Survey of Youth; NICHD SECCYD � National Institute of Child Health and Human Development Study of Early Child Care and Youth Development; IHDP � Infant Health and Development Program; MLEPS � Montreal Longitudinal-Experimental Preschool Study; BCS � British Birth Cohort Study. a Omitted in regression models. * p � .05. ** p � .01. *** p � .001.
1439SCHOOL READINESS AND LATER ACHIEVEMENT
includes measures of school-entry reading achievement as well as language and verbal ability. When we reran the meta-analytic regression in the third column of Table 3 with separate groups for reading achievement and the collection of other language-related measures, we found that we could not reject the hypothesis of equal effects ( p � .11).
Extensions
Beyond shared method variance, there are a number of other reasons to worry that we may have stacked the deck in favor of our school-entry achievement measures: (a) Attention and socioemo- tional skills may be more difficult to measure than achievement- related skills; (b) maternal reports available in three of our data sets may be less predictive than teacher reports of later academic achievement, in part because parents do not observe their children in school settings; (c) our models may overcontrol for achievement-related impacts of attention and socioemotional skills; (d) socioemotional skills may matter more for important school-related outcomes such as drop out than do test scores, because drop out reflects some combination of achievement and behavior; (e) most of our outcomes are measured in middle child- hood, and the relative importance of school-entry factors may change as schools encourage older children to become independent learners; (f) a number of our socioemotional measures are counts of students’ problems, which restricts their range and perhaps explanatory power relative to the full-scale achievement measures; and (g) substantial attrition in some of our studies may bias results.
A first potential threat to our general conclusion is that chil- dren’s behavior is more difficult to measure than their early achievement. Perhaps the lower reliability or validity of the be- havioral measures accounts for their weaker explanatory power. It is certainly true that school-entry tests have high internal consis- tency (e.g., the alphas were at least .74). But the internal consis- tency of most of the attention and socioemotional skills was also
fairly high, particularly in the case of teacher reports, which were all .79 or higher.
To investigate the potential impact of unreliable measurement on our results, we used the reported internal consistencies in the ECLS-K and NLSY data to estimate regression models, using the errors-in-variables reliability adjustment in the EIVREG procedure in Stata (Stata Corporation, 2004). To accord the behavioral mea- sures maximum explanatory power, we included in our regressions school-entry academic test scores as well as family and child control variables but only the given measure of attention or socio- emotional behaviors.
For third-grade reading outcomes and with no reliability adjust- ments in the ECLS-K, the respective standardized coefficients on approaches to learning, self-control, interpersonal skills, external- izing problems, and internalizing problems were .05, .02, .02, –.03, and .00, respectively. Reliability adjustments produced very sim- ilar coefficients: .06, .02, .03, –.04, and .00, respectively. Reliability-related changes to the coefficients on these measures predicting math achievement were similarly modest.
For the NLSY early-adolescence reading test score outcome, respective coefficients associated with hyperactivity, headstrong, antisocial behavior, and anxiety/depression were –.06, –.02, –.05, and –.01, respectively. Adjusting for reliability generally increased (the absolute value of) these coefficients somewhat to –.09, –.03, –.08, and –.01, respectively. Reliability-related changes in coeffi- cients predicting early adolescents’ math scores were similar. Although the proportionate increases in these coefficients are substantial in some cases, none of the reliability-adjusted coeffi- cients begins to rival the size of the coefficients on early reading and math skills. In sum, it is unlikely that lower internal reliability explains the low explanatory power of our attention and socio- emotional behavior measures. Although test–retest correlations may provide a richer understanding of the reliability of these measures, these data were not readily available. Nevertheless, with
Figure 1. Standardized coefficients from models of reading achievement estimated from six data sets. Filled squares indicate statistically significant coefficients.
1440 DUNCAN ET AL.
average effect sizes ranging from –.01 to .01 (see Table 3), it is unlikely that even substantial reliability adjustments to our behav- ior and social skills measures would change our conclusions.
The overall validity of the attention and socioemotional behav- ior measures is much more difficult to assess. Correlations shown in the first column of Table 3 between later achievement and the attention and socioemotional behavioral measures have the ex- pected signs and range from .10 to .25 in absolute value, suggest- ing at least some degree of validity. However, there remains the possibility that low validity might lead us to underestimate their predictive power. Of course, the validity of kindergarten-level achievement tests has also been questioned (Hirsh-Pasek, Kochanoff, Newcombe, & de Villiers, 2005; Meisels & Atkins- Burnett, 2004), so validity-based downward bias is also a concern with respect to the coefficients on the early achievement measures.
A second concern is that we relied on maternal reports of socioemotional behaviors in three studies (NLSY, IHDP, and BCS). Because maternal ratings are comparatively (siblings vs. classmates) and contextually (family vs. school) different from teacher ratings, it is possible that our reliance on maternal reports in these data sets leads to a downward bias in the estimated effects of the attention and socioemotional behavior measures (Gagnon, Vitaro, & Tremblay, 1992).
To investigate whether this might be the case, we took data from the ECLS-K and NICHD SECCYD, both of which gathered com- parable ratings from parents and teachers of several components of socioemotional behaviors, and substituted parent reports for the teacher-report-based measures of these skills. Using our full set of controls and averaging across latest reading and math outcomes, we found that standardized coefficients on externalizing and in- ternalizing problems and social skills averaged, respectively, .05, .03, and .03 for teacher reports and –.00, .01, and .01 for parent reports. Noting the unexpected direction of effects for the teacher reports of problem behaviors and the very modest coefficients in general, reporter bias does not appear to be driving our results.
The third issue, overcontrol, is complicated. Our regression models control for school-entry achievement, but if early attention and socioemotional skills affect later achievement primarily by affecting school-entry achievement skills, are we not robbing the school-entry nonachievement measures of some of their explana- tory power?12 The pattern of average correlations presented in the first column of Table 3 bears on this issue. Bivariate associations between school-entry attention and socioemotional skills are con- siderably larger than their regression-adjusted partial associations, suggesting that this might possibly be the case.
To investigate this possibility more systematically, we reesti- mated our full-control models, using the latest outcomes in each study and omitting our school-entry measures of reading and math skills (but retaining all other control variables). Standardized co- efficients on attention, externalizing problems, internalizing prob- lems, and social skills averaged .13, .03, .02, and .03, respectively, without school-entry reading and math skills. Corresponding av- erages for coefficients from models that included school-entry academic skills were .07, .03, .02, and .02, respectively. Thus, even without including school-entry reading and math skills, only attention skills appear to relate to later reading and math achieve- ment, and this may be due in part to their correlation with the omitted school-entry achievement measures rather than to a true mediation through achievement. That said, it remains the case that our analysis is focused on behavior during the years just before and at the point of school entry. If some types of socioemotional skills are well established before the preschool years, and unchanging during these years, then we will not be able to detect their effects.
Fourth, attention and socioemotional skills may matter more for outcomes such as special education classification or dropping out
12 Note that an overcontrol argument applies equally to the achievement as to the nonachievement measures, because early success in learning reading and math skills may alter preschool behavior.
Figure 2. Standardized coefficients from models of math achievement estimated from six data sets. Filled triangles indicate statistically significant coefficients.
1441SCHOOL READINESS AND LATER ACHIEVEMENT
of school than for the test scores and teacher-reported achievement outcomes used in our studies. The outcomes of our analyses are indeed limited, and it may well be that these types of measures of school completion and success are more strongly linked to chil- dren’s socioemotional behavior and attention skills than to aca- demic skills. Our test for this possibility was to estimate models of the effects of early academic and self-regulatory skills on grade retention, an outcome that includes elements of both academic and behavioral competence. Results in Tables A3, B3, C3, D3, E3, and F3 were quite similar to those from models with purer achievement-related outcomes.13 Nevertheless, the possibility re- mains that the predictive power of school-entry skills may differ for other, even more behavior-based educational outcomes.
A fifth concern is that most of the outcomes were measured during children’s elementary-school years. This is important for two reasons. First, teachers and classrooms differ across the extent to which they support learning academic, attention, and behavior skills. Our analysis does not consider how the associations among these skills may differ as a function of classroom and teacher contexts. Moreover, the associations among these skills may change over time as the contexts of classrooms change. Achieve- ment in the middle- and high-school years involves increasingly complex reading and mathematical tasks, and it may be that general cognitive skills, particularly oral language and conceptual abilities, are crucial for comprehension and advanced problem solving (Baroody, 2003; Ferrari & Sternberg, 1998; Hiebert & Wearne, 1996; NICHD Early Child Care Research Network, 2005b; Scarborough, 2001; Snow et al., 1998; Storch & White- hurst, 2002; Whitehurst & Lonigan, 1998). It is also possible that once children are past learning the basics in the early grades, the relative importance of early attention and socioemotional skills grows as children are increasingly called on to be independent learners, allocate their own time, and complete group work and assignments.
For a general look at the evidence about whether any of the impacts of academic, attention, and socioemotional skill measures are growing over time, we reran our meta-analytic regressions including interactions between each of the school-entry measures and the time between school entry and the outcome assessment. Coefficients on the interactions with early reading, early math, attention, and socioemotional behavior were all negative and ranged from –.023 to –.039 annual decrements in effect sizes, providing no support for hypotheses of increasing importance.
Because most of our outcomes are assessed in elementary school, the interactions shed relatively little light on what would happen if more outcomes were measured during middle or high school. The NLSY data are most telling on this point, because the same skill assessment was given at both age 7–8 and age 13–14. In regressions of outcomes based on these two time points, we find that although school-entry hyperactivity retained its modest ex- planatory power (effect sizes around –.07) between these two points, the explanatory power of reading and math fell. In the case of NICHD SECCYD outcomes measured in third and fifth grade, there was inconclusive evidence on the direction of coefficient change.
A sixth worrisome methodological concern is that a number of our socioemotional measures are counts of students’ problems, which restricts the range of behaviors they capture and might therefore reduce their predictive power. We explored this possi-
bility by estimating spline regressions, which allow for nonlinear effects (results not shown). In these analyses, two linear segments per measure were fit to the data. The first segment was estimated for the most problematic third of the sample distribution, and the second segment was estimated for the other two thirds. If, say, externalizing behavior problems matter a great deal for the chil- dren with very high levels of problem behavior but owing to the restricted range, much less for the others, then the slope of the line fit to the most problematic group should be significantly larger (in absolute value) than the slope for the rest of the sample.
We found little systematic evidence that this was the case. In the ECLS-K data, there was some evidence that improving early math skills mattered more for low math achievers, whereas the NLSY showed that hyperactivity mattered more for the most hyperactive children. No significant nonlinearities emerged in the analysis of NICHD SECCYD data. These spline analyses confirm that there are few nonlinear associations between the socioemotional mea- sures included in this study and the outcomes they predict. It does not rule out, however, the possibility that other measures that capture a broader range of behaviors may be more strongly asso- ciated with later achievement.
A seventh and final concern is that sample attrition in some of our studies may bias our results. The extent of attrition is docu- mented in our appendices. All of the coefficients used in our meta-analyses come from models in which missing data are ac- counted for with missing data dummy variables. In the appendices, we present results for three of our data sets (ECLS-K, NICHD SECCYD, and MLEPS) in which missing data are handled with multiple imputation and listwise deletion.
In two data sets (ECLS-K and MLEPS), we also used nonre- sponse weighting adjustments. In the ECLS-K, both multiple im- putation and listwise deletion estimates are slightly smaller in magnitude than the results using missing data dummy variables, although the pattern of results is consistent across all the tech- niques. In the MLEPS, the pattern of results does not change across the different methods; however, listwise deletion produces few statistically significant estimates given the reduction in sample size from 500 to approximately 150. In the NICHD SECCYD, the overall pattern of results is similar across the two missing data techniques; however, the coefficient estimates are most consistent between missing data dummy estimates and multiple imputation. Across the three data sets, the respective coefficients on school entry reading, math, attention, externalizing, internalizing, and social skills averaged .13, .23, .06, .04, .02, and .02, respectively, when we used missing data dummies. Corresponding coefficient averages for multiply-imputed models were .12, .17, .06, .01, .01, and .02, respectively. Corresponding coefficient averages for list- wise deletion models were .09, .27, .07, .00, .02, and .00, respec- tively. Corresponding coefficient averages for nonresponse weighted models were .12, .31, .10, –.00, .01, and –.00, respec- tively.
13 Only the BCS has followed study participants long enough to measure their completed schooling and early-career labor market earning and, in results not reported in the appendices, we found that school-entry attention problems were a significant predictor of school completion but not labor market success.
1442 DUNCAN ET AL.
Discussion and Conclusions
We have presented results from a coordinated analysis of six longitudinal data sets relating school-entry skills to later teacher ratings and test scores of reading and math achievement, holding constant children’s preschool cognitive ability, behavior, and other important background characteristics. Our meta-analytic results indicate that such early math concepts as knowledge of numbers and ordinality were the most powerful predictors of later learning (the average effect size of school-entry math skills was .34 and every bit as large as early reading skills in predicting later reading achievement). Less powerful, but also consistent, predictors across studies were early language and reading skills such as vocabulary; knowing letters, words, and beginning and ending word sounds (the average effect size across our studies was .17); and attention skills (average effect size .10). The average effect sizes of exter- nalizing and internalizing problem behaviors and social skills were close to zero.
Despite our extensive investigation of the robustness of our key results, any nonexperimental analysis using imperfectly measured cognitive, achievement, and behavioral constructs such as ours cannot rule out all threats to its conclusions. First, although shared method variance, reporter bias, overcontrol, restricted range, and measurement reliability cannot account for the differential predic- tive power of school-entry achievement and socioemotional mea- sures, we are unable to rule out bias from the lower validity of socioemotional measures. Second, despite our ability to control for cognitive ability prior to school entry in five of our six studies, and despite our controls for concurrent reading skills in all six studies, it remains possible that our surprisingly large school-entry math coefficients overstate causal impacts.
One of our noteworthy results is that early math is a more powerful predictor of later reading achievement than early reading is of later math achievement. Despite our controls for cognitive ability, it remains possible that some of the apparent effects of early math skills is spurious. To the extent that the effects are real, it is important to discover why. Math is a combination of both conceptual and procedural competencies. Although our data do not allow us to examine these competencies separately, future research could focus on this direction. Another productive avenue of re- search would be to examine efforts to improve math skills prior to school entry. Random-assignment evaluations of early math pro- grams that focus on the development of particular mathematical skills and track children’s reading and math performance across elementary school could help to illuminate missing causal links between early skills and later achievement.
Another finding from our analysis is that attention skills are modestly but consistently associated with achievement outcomes. One explanation for this predictive power is that attention skills increase the time children are engaged and participating in aca- demic endeavors and learning activities. Other studies have shown that attention skills have important associations with school suc- cess, independent of cognitive and/or language ability (Alexander et al., 1993; Howse et al., 2003; McClelland et al., 2000; Yen et al., 2004), but few of these studies have controlled for prior levels of academic skills as well as prior levels of behavior. Our results suggest that attention skills, but not problem behavior or social skills, predict achievement outcomes, even after the effects of
early achievement knowledge and cognitive ability have been netted out.
Although all of the studies we analyzed were drawn from normative populations, all contain at least some children falling in the clinical ranges of behavior problems. We were surprised that our spline-regression models produced no consistent evidence of nonlinear effects of problem behaviors on later achievement. We caution, however, that it remains possible a more focused analysis, perhaps with clinical samples, might reach different conclusions.
Given that teachers emphasize the importance of attention skills and socioemotional behavior for school readiness and the possi- bility that these skills shape classroom learning processes, it might be expected that these early skills would have crossover effects on subsequent reading and math achievement. With the important exception of attention skills, we did not find evidence that changes in these skills during the preschool years predict later achievement. However, as noted earlier, academic skills are only one facet of educational success, and improvements in problem behavior or social skills may better predict other important school outcomes, such as a child’s engagement in school and motivation for learn- ing, relationships with peers and teachers, and overall self-concept and school adjustment (Greenberg et al., 2003). It might also be the case that early-grade teachers are somehow able to prevent prob- lem behaviors from interfering with student learning but that problem behaviors would be linked with lower achievement if teachers were less capable. Despite the uniformly small and often insignificant coefficients on these measures in our regressions, we caution against completely dismissing the potential academic ben- efits of environments or programs that promote positive socioemo- tional development.
An additional caveat is that any one child’s socioemotional behavior, in particular externalizing problem behaviors, may affect other students’ achievement more than the child’s own individual achievement. For example, problem behaviors may disrupt class- room activities such that even well-behaved children spend less time engaged in instructional and learning activities. Our analyses do not consider this possibility because it requires more complete data about classmates’ behavior than the studies provide. We raise this point, however, because we believe that the topic of peer effects deserves further attention in future research on socioemo- tional behavior.
Our analyses focus on skills and behaviors that emerge at the time of school entry and not on the effects of socioemotional behaviors that emerge after children enter school. This is impor- tant, because it may be that reading achievement and problem behavior develop in tandem during the early elementary years (Trzesniewski et al., 2006). Additional research is necessary to further elucidate the potentially complex and reciprocal relation- ships between children’s socioemotional behaviors and their aca- demic achievement.
Our conclusions about the importance of early academic and attentions skills are consistent with the recommendations endorsed by the National Association for the Education of Young Children and the National Council of Teachers of Mathematics (2002) and the National Research Council’s Committee on the Prevention of Reading Difficulties in Young Children (Snow et al., 1998). How- ever, our results say nothing about the types of curricula that would be most effective in promoting these skills. Play-based, as opposed to “drill-and-practice,” curricula designed with the developmental
1443SCHOOL READINESS AND LATER ACHIEVEMENT
needs of children in mind can foster the development of academic and attention skills in ways that are engaging and fun. Taking early math skills as an example, the Big Math for Little Kids program has been designed to capitalize on children’s interest in exploring and manipulating numbers (Greenes, Ginsburg, & Balfanz, 2004). In addition, play-based curricula may also have the added benefit of fostering attention-related skills (Berk, 1994).
Our findings support three key conclusions for developmental research. First, math and reading skills at the point of school entry are consistently associated with higher levels of academic perfor- mance in later grades. Particularly impressive is the predictive power of early math skills, which supports the wisdom of exper- imental evaluations of promising early math interventions. Second, among attention-related and socioemotional behaviors, only the attention-related skills predicted later academic achievement with any consistency. We find no noteworthy regression-adjusted asso- ciations between either interpersonal skills (or problems) or ag- gression and later achievement. Finally, all of our data sets suggest that reading and math tests that were individually administered to children by trained personnel around the point of school entry can be a highly reliable way of assessing early skills. That said, it was also the case that we could not attribute most of the variation in later school achievement to our collection of school-entry factors, so the potential for productive interventions during the early school grades remains.
References
Adams, M. J., Treiman, R., & Pressley, M. (1998). Reading, writing and literacy. In I. Sigel & A. Renninger (Eds.), Handbook of child psychol- ogy: Child psychology and practice (5th ed., Vol. 4, pp. 275–355). New York: Wiley.
Alexander, K. L., Entwisle, D. R., & Dauber, S. L. (1993). First grade classroom behavior: Its short- and long-term consequences for school performance. Child Development, 64, 801–514.
Baker, P. C., Keck, C. K., Mott, F. L., & Quinlan, S. V. (1993). NLSY Child Handbook (Rev. ed.). Columbus: The Ohio State University, Center for Human Resources Research.
Baroody, A. J. (2003). The development of adaptive expertise and flexi- bility: The integration of conceptual and procedural knowledge. In A. J. Baroody & A. Dowker (Eds.), The development of arithmetic concepts and skills: Constructing adaptive expertise studies (pp. 1–34). Mahwah, NJ: Erlbaum.
Barriga, A. Q., Doran, J. W., Newell, S. B., Morrison, E. M., Barbetti, V., & Robbins, B. D. (2002). Relationships between problem behaviors and academic achievement in adolescents: The unique role of attention problems. Journal of Emotional and Behavioral Disorders, 10, 233–240.
Berk, L. E. (1994). Vygotsky’s theory: The importance of make-believe play. Young Children, 50, 30–39.
Bush, G. W. (2002, April 3). President, Mrs. Bush promote early childhood education initiative. Speech posted by the Office of the Press Secretary. Retrieved July 6, 2005, from http://www.whitehouse.gov/news/releases/ 2002/04/20020403–6.html
Bynner, J., Ferri, E., & Shepherd, P. (Eds.). (1997). Twenty something in the 1990s: Getting on, getting by, getting nowhere. Aldershot, England: Ashgate.
Campbell, F. A., Ramey, C., Pungello, E. P., Sparling, J. J., & Miller- Johnson, S. (2002). Early childhood education: Young adult outcomes from the Abecedarian Project. Applied Developmental Science, 6, 42– 57.
Center for Human Resource Research. (2004). A guide to the 1979–2002
National Longitudinal Survey of Youth. Washington, DC: U.S. Depart- ment of Labor, Bureau of Labor Statistics.
Coie, J. D., & Krehbiel, G. (1984). Effects of academic tutoring on the social status of low-achieving, socially rejected children. Child Devel- opment, 55, 1465–1478.
Conduct Problems Prevention Research Group. (1992). A developmental and clinical model of antisocial development and prevention. Develop- ment and Psychopathology, 4, 509–527.
Conduct Problems Prevention Research Group. (2002). Evaluation of the first three years of the Fast Track Prevention Trial with children at high risk for adolescent conduct problems. Journal of Abnormal Child Psy- chology, 30, 19–35.
Cooper, D., & Farran, D. (1991). The Cooper–Farran Behavior Rating Scale. Brandon, VT: Clinical Psychology.
Cunha, F., Heckman, J., Lochner, L., & Masterov, D. (2006). Interpreting the evidence on life cycle skill formation. In E. Hanushek & F. Welch (Eds.), Handbook of the economics of education (pp. 307–451). North Holland: Elsevier.
Dauber, S. L., Alexander, K. L., & Entwisle, D. R. (1993). Characteristics of retainees and early precursors of retention in grade: Who is held back? Merrill-Palmer Quarterly, 39, 326–343.
Dolan, L., Kellam, S., Brown, C., Werthamer-Larsson, L., Rebok, G., Mayer, L., et al. (1993). The short-term impacts of two classroom-based preventive interventions on aggressive and shy behaviors and poor achievement. Journal of Applied Developmental Psychology, 14, 317– 345.
Dunn, L. M., & Dunn, L. M. (1981). Manual for Peabody Picture Vocab- ulary Test—Revised. Circle Pines, MI: American Guidance Service.
Dunn, L. M., & Markwardt, F. C. (1970). Peabody Individual Achievement Test Manual. Circle Pines, MN: American Guidance Service.
Entwisle, D. R., & Alexander, K. L. (1990). Beginning school math competence: Minority and majority comparisons. Child Development, 61, 454–471.
Entwisle, D. R., Alexander, K. L., & Olson, L. S. (2005). First grade and educational attainment by age 22: A new story. American Journal of Sociology, 110, 1458–1502.
Ferrari, M., & Sternberg, R. J. (1998). The development of mental abilities and styles. In W. Damon (Ed.), Handbook of child psychology: Child psychology and practice (5th ed., Vol. 2, pp. 899–946). New York: Wiley.
Ferri, E., & Smith, K. (2003). Family life. In E. Ferri, J. Bynner, & M. Wadsworth (Eds.), Changing Britain, changing lives: Three generations at the turn of the century (pp. 105–132). London: University of London, Institute of Education.
Gagnon, C., Vitaro, F., & Tremblay, R. E. (1992). Parent–teacher agree- ment on kindergarteners’ behavior problems: A research note. Journal of Child Psychology and Psychiatry, 33, 1255–1261.
Greenberg, M. T., Weissberg, R. P., O’Brien, M. U., Zins, J. E., Fredericks, L., Resnik, H., et al. (2003). Enhancing school-based prevention and development through coordinated social, emotional, and academic learn- ing. American Psychologist, 58, 466–474.
Greenes, C., Ginsburg, H., & Balfanz, R. (2004). Big math for little kids. Early Childhood Research Quarterly, 19, 159–166.
Griffin, S., Case, R., & Capodilupo, A. (1995). Teaching for understand- ing: The importance of the central conceptual structure in the elementary mathematics curriculum. In A. McKeough, J. Lupart, & A. Marini (Eds.), Teaching for transfer (pp. 123–151). Mahwah, NJ: Erlbaum.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. Orlando, FL: Academic Press.
Hiebert, J., & Wearne, D. (1996). Instruction, understanding, and skill in multidigit addition and subtraction. Cognition and Instruction, 14, 251– 283.
Hinshaw, S. P. (1992). Externalizing behavior problems and academic
1444 DUNCAN ET AL.
underachievement in childhood and adolescence: Causal relationships and underlying mechanisms. Psychological Bulletin, 111, 127–155.
Hirsh-Pasek, K., Kochanoff, A., Newcombe, N. S., & de Villiers, J. (2005). Using scientific knowledge to inform preschool assessment: Making the case for “empirical validity.” Society for Research in Child Development Social Policy Report, XIX(I), 3–19.
Howse, R. B., Lange, G., Farran, D. C., & Boyles, C. D. (2003). Motivation and self-regulation as predictors of achievement in economically disad- vantaged young children. Journal of Experimental Education, 71, 151– 174.
Karoly, L. A., Kilburn, M. R., & Cannon, J. S., (2005). Early childhood interventions: Proven results, future promise (Rep. No. MG-341). Santa Monica, CA: RAND.
Kellam, S. G., Mayer, L. S., Rebok, G. W., & Hawkins, W. E. (1998). Effects of improving achievement on aggressive behavior and of im- proving aggressive behavior on achievement through two preventative interventions: An investigation of causal paths. In B. P. Dohrenwend (Ed.), Adversity, stress, and psychopathology (pp. 486–505). New York: Oxford University Press.
Klebanov, P. K., Brooks-Gunn, J., & McCormick, M. C. (1994a). Class- room behavior of very low birth weight elementary school children. Pediatrics, 94, 700–708.
Klebanov, P. K., Brooks-Gunn, J., & McCormick, M. C. (1994b). School achievement and failure in very low birth weight children. Journal of Developmental and Behavioral Pediatrics, 15, 248–256.
Konold, T. R., & Pianta, R. C. (2005). Empirically-derived, person- oriented patterns of school readiness in typically-developing children: Description and prediction to first-grade achievement. Applied Devel- opmental Science, 9, 174–187.
Kowaleski-Jones, L., & Duncan, G. J. (1999). The structure of achievement and behavior across middle childhood. Child Development, 4, 930–943.
Ladd, G. W., Birch, S. H., & Buhs, E. S. (1999). Children’s social and scholastic lives in kindergarten: Related spheres of influence? Child Development, 70, 1373–1400.
La Paro, K. M., & Pianta, R. C. (2000). Predicting children’s competence in the early school years. A meta-analytic review. Review of Educational Research, 70, 443–484.
Lazar, I., Darlington, R. B., Murray, H., Royce, J., Snipper, A., & Ramey, C. T. (1982). Lasting effects of early education: A report from the Consortium for Longitudinal Studies. Monographs of the Society for Research in Child Development, 47(2–3, Serial No. 195).
Lee, V., & Burkam, D. (2002). Inequality at the starting gate: Social background differences in achievement as children begin school. Wash- ington, DC: Economic Policy Institute.
Lin, H., Lawrence, F. R., & Gorrell, J. (2003). Kindergarten teachers’ views of children’s readiness for school. Early Childhood Research Quarterly, 18, 225–237.
Love, J. M., Harrison, L., Sagi-Schwartz, A., van Ijzendoorn, M. H., Ross, C., Ungerer, J. A., et al. (2003). Child care quality matters: How conclusions may vary with context. Child Development, 74, 1021–1033.
McClelland, M., Morrison, F. J., & Holmes, D. L. (2000). Children at risk for early academic problems: The role of learning-related social skills. Early Childhood Research Quarterly, 15, 307–329.
McCoy, A. R., & Reynolds, A. J. (1999). Grade retention and school performance: An extended investigation. Journal of School Psychology, 37, 273–298.
McWayne, C., Fantuzzo, J., & McDermott, P. A. (2004). Preschool com- petency in context: An investigation of the unique contribution of child competencies to early academic success. Developmental Psychology, 40, 633–645.
Meisels, S. J. (1998). Assessing readiness (Rep. No. 3–002). Ann Arbor, MI: Center for the Improvement of Early Reading Achievement. Re- trieved September 15, 2001, from http://www.ciera.org/library/reports/ inquiry-3/3–002/3–002.pdf
Meisels, S. J., & Atkins-Burnett, S. (2004). The Head Start National Reporting System: A critique. Beyond the Journal: Young Children on the Web. Retrieved July 6, 2005, from http://journal.naeyc.org/btj/ 200401/meisels.pdf
Miech, R., Essex, M. J., & Goldsmith, H. (2001). Socioeconomic status and the adjustment to school: The role of self-regulation during early childhood. Sociology of Education, 74, 102–120.
National Association for the Education of Young Children and National Council of Teachers of Mathematics. (2002). Math experiences that count! Young Children, 57, 60–62.
National Center for Education Statistics. (1993). Public school kindergar- ten teachers’ views on children’s readiness for school. (Rep. No. NCES 93–410). Washington, DC: U.S. Department of Education, Office of Educational Research and Improvement.
National Center for Education Statistics. (2001). User’s manual for the ECLS-K public-use data files and electronic codebook. Washington DC: U.S. Department of Education.
Newcomb, A. F., Bukowski, W. M., & Pattee, L. (1993). Children’s peer relations: A meta-analytical review of popular, rejected, neglected, con- troversial, and average sociometric status. Psychological Bulletin, 113, 99–128.
NICHD Early Child Care Research Network. (1997). Poverty and patterns of child care. In J. Brooks-Gunn & G. Duncan (Eds.), Consequences of growing up poor (pp. 100–131). New York: Russell Sage Foundation.
NICHD Early Child Care Research Network. (2004). Trajectories of phys- ical aggression from toddlerhood to middle childhood: Predictors, cor- relates, and outcomes. Monographs of the Society of Research in Child Development, 69(4, Serial No. 278).
NICHD Early Child Care Research Network. (2005a). Child care and development: Results from the NICHD Study of Early Child Care and Youth Development. New York: Guilford Press.
NICHD Early Child Care Research Network (2005b). Pathways to reading: The role of oral language in the transition to reading. Developmental Psychology, 41, 428–442.
Normandeau, S., & Guay, F. (1998). Preschool behavior and first-grade achievement: The mediational role of cognitive self-control. Journal of Educational Psychology, 90, 111–121.
Olson, S. L., Sameroff, A. J., Kerr, D. C. R., Lopez, N. L., & Wellman, H. M. (2005). Developmental foundations of externalizing problems in young children: The role of effortful control. Development and Psycho- pathology, 17, 25–45.
Parker, J. G., & Asher, S. R. (1987). Peer relations and later personal adjustment: Are low-accepted children at risk? Psychological Bulletin, 86, 357–389.
Pianta, R. C., & Stuhlman, M. W. (2004). Teacher–child relationships and children’s success in the first years of school. School Psychology Re- view, 33, 444–458.
Posner, M. I., & Rothbart, M. K. (2000). Developing mechanisms of self-regulation. Development and Psychopathology, 12, 427–441.
Pungello, E. P., Kuperschmidt, J. B., Burchinal, M. R., & Patterson, C. (1996). Environmental risk factors and children’s achievement from middle childhood to adolescence. Developmental Psychology, 32, 755– 767.
Raver, C. C. (2004). Placing emotional self-regulation in sociocultural and socioeconomic contexts. Child Development, 75, 346–353.
Raver, C. C., Smith-Donald, R., Hayes, T., & Jones, S. M. (2005, April). Self-regulation across differing risk and sociocultural contexts: Prelim- inary findings from the Chicago School Readiness Project. Paper pre- sented at the biennial meeting of the Society for Research in Child Development, Atlanta, GA.
Reynolds, A., Ou, S., & Topitzes, D. (2004). Paths of effects of early childhood intervention on educational attainment and delinquency: A confirmatory analysis of the Chicago Child–Parent Centers. Child De- velopment, 75, 1299–1328.
1445SCHOOL READINESS AND LATER ACHIEVEMENT
Reynolds, A., & Temple, J. A. (1998). Extended early childhood interven- tion and school achievement: Age thirteen findings from the Chicago Longitudinal Study. Child Development, 69, 231–246.
Rimm-Kaufman, S. E., & Pianta, R. C. (2000). An ecological perspective on the transition to kindergarten: A theoretical framework to guide empirical research. Journal of Applied Developmental Psychology, 21, 491–511.
Rimm-Kaufman, S. E., Pianta, R. C., & Cox, M. J. (2000). Teachers’ judgments of problems in the transition to kindergarten. Early Childhood Research Quarterly, 15, 147–166.
Robins, L. N. (1978). Sturdy childhood predictors of adult antisocial behaviour: Replications from longitudinal studies. Psychological Medi- cine, 8, 611–622.
Scarborough, H. S. (2001). Connecting early language and literacy to later reading (dis)abilities: Evidence, theory, and practice. In S. B. Neuman & D. K. Dickinson (Eds.), Handbook of early literacy research (pp. 97– 110). New York: Guilford Press.
Schaefer, B. A., & McDermott, P. A. ( 1999). Learning behavior and intelligence as explanation for children’s scholastic achievement. Jour- nal of School Psychology, 37, 299–313.
Shonkoff, J., & Phillips, D. (Eds.). (2000). From neurons to neighbor- hoods: The science of early childhood development. Washington, DC: National Academy Press.
Snow, C. E., Burns, M. S., & Griffin, P. (Eds.). (1998). Preventing reading difficulties in young children. Washington DC: National Research Coun- cil, National Academy Press.
Stata Corporation. (2004). Stata Reference Manuals and User Guide (Re-
lease 10) [Computer software and manual]. College Station, TX: Stata Press.
Stevenson, H. W., & Newman, R. S. (1986). Long-term prediction of achievement and attitudes in mathematics and reading. Child Develop- ment, 57, 646–659.
Storch, S. A., & Whitehurst, G. J. (2002). Oral language and code-related precursors to reading: Evidence from a longitudinal structural model. Developmental Psychology, 38, 934–947.
Trzesniewski, K. H., Moffitt, T. E., Caspi, A., Taylor, A., & Maughan, B. (2006). Revisiting the association between reading achievement and antisocial behavior: New evidence of an environmental explanation from a twin study. Child Development, 77, 72–88.
Tunmer, W. E., & Nesdale, A. R. (1985). Phonemic segmentation skill and beginning reading. Journal of Educational Psychology, 77, 417–427.
Wechsler, D. (1991). Manual for the Wechsler Intelligence Scale for Children—Third Edition. San Antonio, TX: Psychological Corporation.
Whitehurst, G. J., & Lonigan, C. J. (1998). Child development and emer- gent literacy. Child Development, 69, 848–872.
Woodcock, R. W., & Johnson, M. B. (1990). Manual for the Woodcock– Johnson Tests of Achievement—Revised. Allen, TX: RCL Enterprises.
Yen, C., Konold, T. R., & McDermott, P. A. (2004). Does learning behavior augment cognitive ability as an indicator of academic achieve- ment? Journal of School Psychology, 42, 157–169.
Received April 18, 2006 Revision received April 30, 2007
Accepted May 16, 2007 �
1446 DUNCAN ET AL.
