Carolan Weiss Matthews AEFP submissionx - The Association ...

groanaberrantInternet and Web Development

Feb 2, 2013 (4 years and 7 months ago)

146 views

Running Head: MIDDLE SCHOOL MODEL





Which Middle School Model Works Best? Evidence from the Early Childhood Longitudinal
Study


Brian V. Carolan

Montclair State University


Christopher C. Weiss

Columbia University


Jamaal S. Matthews

Montclair State University






Author Note

All correspondence should be directed to Brian V. Carolan, Department of Educational
Foundations, Montclair State University, UN 2161, 1 Normal Avenue, Montclair, NJ 07043
carolanb@mail.montclair.edu
.


Manuscript
presented at the Association for Education Finance and Policy Annual Conference,
Boston, MA, March, 2012.
Please do not cite or quote without permission.
MIDDLE SCHOOL MODEL

2


Abstract


There are few areas of school organization that reflect more dissatisfaction than
how to structure
the education

of
early adolescents
in

the middle grades
.

This study uses multilevel models on
n
ationally representative data provided by the Early Longitudinal Study to investigate which
grade span model is best associated with academic achievement and whether certain models are
beneficial for certain types of students from differing socio
-
demograph
ic backgrounds. Results
indicate that there is no generalizable relationship between grade span configuration and
achievement in math and reading.
The results should give reflective pause to reformers
considering whole
-
scale changes to the ways in which
grade spans are organized for early
adolescents.


Keywords:

grade span, students’ achievement, early adolescents




MIDDLE SCHOOL MODEL

3

Which Middle School Model Works Best? Evidence from the Early Childhood Longitudinal
Study

There are few areas of school organization that reflect more dissatisfaction than how to
structure the education of early adolescents in the middle grades. From the rise of the junior high
school over seventy years ago, through the peak years of the middl
e school, and to the current
trend back towards K
-
8 schools, states and school districts across the United States are perplexed
about how to best organize schools’ grade spans for early adolescents. In addition, the research
that has informed the developm
ent and replication of these various models has not generated a
consensus as to which model, if any, works best overall, or which types of students would best
be served by certain configurations.

If there is a consensus in this research literature, it is

that early adolescence is a life
course stage during which differences among students’ educational trajectories greatly accelerate
(
Juvonen, et al., 2004)
. More specifically, for many early adolescents grades decline (Barber &
Olsen, 2004), anxieties inc
rease (
Grills
-
Taqueche,
Norton, & Ollendick, 2010), academic
motivation, school interest and sense of belonging decrease (Maehr & Midgley, 1996), and
behavioral troubles surface (Theriot, Craun, & Dupper, 2010). Circumstantial evidence suggests
these probl
ematic trends are likely more pronounced for traditionally marginalized students
(French, Seidman, Allen & Abner, 2006; Seidman, Aber, Allen & French, 1996). Not all
adolescents experience extreme declines in self
-
perceptions, motivation and performance;
however, these trends are indeed persistent and may be particularly salient for underprivileged
students. Thus, the middle school years are characterized by a set of negative outcomes that may
seriously jeopardize the likelihood of secondary and post
-
seco
ndary success. As a result of these
negative conditions,
states and school districts across the country, especially in urban areas, have
MIDDLE SCHOOL MODEL

4

been reconsidering the practice of educating early adolescents in stand
-
alone middle schools,
which typically span gra
des six through eight, and replacing them with K

8 schools (Hough,
2005.

The research that has informed these efforts, however, is limited by three factors. First,
although a small number of rigorous studies have a high degree of internal validity and hav
e thus
been able to detect an effect of one model versus another, these studies have been hampered by
limited external validity. Therefore, althoug
h this research literature has several

high quality
studies, the data used in these studies do not reflect t
he broad population of early adolescents in
the United States. Second, there is an even larger body of literature that has examined the effects
of middle school, but these studies have done so in the absence of comparative data. Therefore,
research studi
es done along these lines may be able to compare students who attend the same
middle school, but they are unable to infer what these students’ outcomes would be had they
attended a different model. Third, while both of these issues are related to research
design, the
final issue is that few, if any, studies have examined which model works best for certain types of
students. While attempting to discern the effects of one model versus another, policies that have
emanated from this research have resulted in a

one
-
size
-
fits all approach to finding the optimal
middle level grade span. So, perhaps an even narrower question needs to be asked.

Purpose

The purpose of this research is to address these limitations and ask a specific question: Is
there a particular
form of middle grades schooling that works best for students and does this
model work equally well for different types of students? In addition, by using data from a
nationally representative sample of early adolescents, the Early Childhood Longitudinal S
tudy

Kindergarten Class (ECLS
-
K), the answer to this question will have implications for a much
MIDDLE SCHOOL MODEL

5

larger population, thereby providing greater external validity. Finally, because the data from the
ECLS
-
K are longitudinal and follow the same cohort of childr
en across grades Kindergarten
through eight, it is possible to compare the outcomes for similar types of students in different
middle school models. These data address the two limitations noted above and enable one to
determine whether any particular mode
l works best for certain early adolescents.

Background

Different School Models for Early Adolescents

T
he period of the middle grades in the educational sequence has seen a number of
educational reforms that have sought to better tailor instruction and im
prove student outcomes in
these years. First among these reforms was the creation of the junior high school model

roughly grades 7 to 9

that rapidly replicated over 60 years ago (Goldin, 1999). Driven in part
by the perceived need to house adolescents in

separate school structures, as well as the need to
better prepare students for high school
-
level work, the junior high became a standard form of
schooling, however the effectiveness of this configuration has been questioned since its inception
(Cuban, 199
2).
Subsequently, beginning in the early 1970s, public school districts have been
moving away from junior high schools and toward middle schools
, typically spanning grades 6
to 8. These efforts were driven by a perceived need to better attend to the developmental needs
of early adolescents and to refine the schooling structure designed for the needs of this age
group.

Like the junior high school

whose wrongs it was intended to right, the middle school,
despite its tremendous growth and marked prevalence, is generally believed to have not lived up
to its potential (
Weiss & Kipnes, 206
). Research has generally argued that schools of this form
als
o do not adequately attend to the needs of adolescents as typically implemented (Cuban,
MIDDLE SCHOOL MODEL

6

1992). Like the junior high, the middle school has been the focus of reform efforts practically
since its inception (Juvonen et al., 2004). Based in part on this litera
ture, several districts,
including the large urban districts of Baltimore, Cleveland, and Philadelphia have initiated
reforms over the last decade to dismantle their middle schools and educate students in the middle
grades through other school models (Cook

et al., 2008).

Along these lines, the current trend in many districts is to disband junior highs and
middle schools and educate the middle grades in K
-
8 schools. This trend is motivated by
research that asserts that these models are associated with small
er, more personal instruction,
tighter social connections, and as a result, gains in students’ cognitive and non
-
cognitive
outcomes. For example, research has attributed a number of negative changes in the middle
grades years to the middle school, includi
ng academic achievement (Alspaugh, 2001; Hanushek,
Kain, & Rivkin, 2004) and motivation (Rudolph et al., 2001). Moreover, a number of studies
have reported that students in K
-
8 schools have higher levels of self
-
esteem and perceptions of
school safety (e.g
., Way, Ready, & Rhodes, 2007;
Weiss & Kipnes
, 2006). A recent study using
administrative record data for the state of North Carolina found that sixth graders attending
middle schools are substantially more likely to experience disciplinary trouble at scho
ol, relative
to those attending elementary schools (Cook et al., 2008).

Fewer studies have reported on the cognitive benefits of K
-
8 schools; however, a few are
noteworthy. In Philadelphia, Mac Iver and Mac Iver (2006) found that students in
long
-
establis
hed K
-
8 schools generally outgained middle school students in math, but these gains were
not as large in newly
-
established K
-
8 schools, substantiating an earlier finding from Byrnes and
Ruby (2007). A study of schools in Cleveland found higher reading and
math scores for sixth
grade students attending K
-
8 schools, compared to peers in middle schools (Poncelet & Metis
MIDDLE SCHOOL MODEL

7

Associates, 2004). Rockoff and Lockwood (2010) report substantial cognitive benefits for those
students in New York City who attend K
-
8 schoo
l
.
Taken together, the evidence on the relative
effectiveness of middle schools and K
-
8 schools is rather mixed. However, public dissatisfaction
with middle schools continues to grow.

Middle School Effects and Differences among Early Adolescents


Some po
rtion of the dissatisfaction with either the junior high or middle school

can be
traced to what developmental psychologists have referred to as the “developmental mismatch”
between early adolescents’ developmental needs and the social and academic configur
ations of
the middle schools they typically attend. A developmental extension of person
-
environment fit
theory, stage
-
environment fit focuses on how the myriad of changes early adolescents experience
are not well served by the middle school environments i
n which they are typically educated
(Eccles, Lord, & Midgely, 1991;
Seidman, Aber, Allen, & French
, 1994).


In addition to significant psychological and physical changes associated with this life
course stage (see, e.g., Eccles & Roeser, 2009 or
Anderman

& Maehr, 1994

for a review of this
literature), early adolescents also experience major changes in relationships with peers, parents,
and adults at school, particularly teachers, in this phase of life (Eccles and Roeser, 2009). Early
adolescents desire m
ore autonomy, greater decision making ability, an increased need for
competence and relatedness, and a shift from family
-
centered orientation to peer culture (Deci &
Ryan, 1994; Giordano, 2003). For early adolescents, relations with teachers and other adu
lts at
school become less personal, less positive, and more punitive (
Feldlaufer, Midgley, & Eccles,
1988
). Middle schools’ relations with parents become more formal and less frequent, thereby
reducing the amount and quality of information that flows betw
een the school and home.

MIDDLE SCHOOL MODEL

8

These changes in the social lives of early adolescents are in some ways exacerbated by
middle schools and the transitions into them (Eccles, Wigfiled, Midgley &
Reuman
, 1993;
Rudolph et al. 2001). For example, middle school and
junior high administrations tend to place a
greater emphasis on control and discipline, giving students less academic autonomy during a
critical period where adolescents actually desire more autonomy. Further, in middle schools the
quality of student
-
teach
er relationships decrease, becoming less personal and less positive
(Eccles, et al., 1993). There is also an increased emphasis on performance, competition and
comparison, which undermines intrinsic motivation and mastery goal orientation. Grading
practice
s also become more stringent, which can undermine student self
-
efficacy and feelings of
competence (Eccles, et al., 1993; Friedel, Cortina, Turner, & Midgley, 2010). These foci are the
result of middle schools relying more on between
-
classroom ability grou
ping, whole
-
class
instruction, and social comparison of grades. Finally, the transition to middle school disrupts
social networks at a time when adolescents are increasingly concerned with peer relations. These
environmental changes, coupled with the norma
l course of adolescent development, result in a
developmental mismatch between the needs of adolescents and the school environment.

However, some adolescents may be able to negotiate this transition better than others.
More specifically, students from disa
dvantaged backgrounds may be even more adversely
affected by the transition to middle school (Rumberger, 1995). These students may benefit from
K
-
8 models that do not require a transition between schooling forms, thereby providing a more
seamless educati
onal experience and an important buffer from the developmental changes they
are experiencing. Moreover, making the transition to a new school requires one to
simultaneously cope with developmental change and school change. And since coping with
multiple ch
anges is more challenging that coping with only one, these students are more likely to
MIDDLE SCHOOL MODEL

9

experience negative academic outcomes than those students who only have to cope with
developmental change (
Eccles, Lord, & Midgley, 1991).

Generally, students from disad
vantaged backgrounds, including minorities, those from
single parent families, and those with limited English proficiency, are likely to have a more
difficult time in and out of school (Sirin, 2005). Therefore, the continuity provided by the K
-
8
model all
ows for greater stability with peer relations, more informal contact between home and
school, and increased opportunities for informal relations with teachers and other adults at
school. While some may be better able to successfully transition to middle s
chool models, early
adolescents from disadvantaged backgrounds may benefit from the types of stable, consistent,
and predictable relations with peers, parents, and teachers that are afforded by the K
-
8 model. It
is important for research to evaluate wheth
er and to what extent different middle grades models
work best for certain students, especially those from disadvantaged backgrounds.

Longitudinal Comparative Framework

The research literature, however, has yet to firmly establish the relationship between
middle level grade span and relevant outcomes, and, more importantly, how this relationship
varies by student characteristics. On a more general note,
perhaps what is m
ost notable with
respect to the literature on the effects of the transition to stand
-
alone middle schools

is

the very
limited number of studies that directly compare middle schools with alternative schooling forms.
Stated differently, although there are ma
ny arguments about the effects of particular schooling
forms (such as the middle school) on students’ outcomes, surprisingly little research has directly
examined the effects of particular schooling forms in a comparative framework. In part, this may
refle
ct a limitation of data and of how reforms are implemented. Because most districts have
only one configuration of schooling forms (e.g., elementary school for grades K

5, middle
MIDDLE SCHOOL MODEL

10

school for grades 6

8, and high school for grades 9

12), comparisons are probl
ematic, since it
can be difficult to disentangle district
-
level differences from school
-
level differences.


A majority of past studies on the effects of middle school have used two primary
strategies: 1) within
-
cohort analyses and, 2) cross
-
sectional analy
ses. Within
-
cohort analyses
examine a group of adolescents as they move from one schooling form to another. This design
is useful for providing rich accounts of the changes adolescents experience as they move from
one school to another during a developm
entally turbulent time period. While this research is
able to compare students who all experienced the same transition, it is unable to infer what these
students’ outcomes would be had they not made the transition. Cross
-
sectional analyses address
this l
imitation as they can indirectly compare similar types of students in different schooling
forms at one point in time (e.g., Eccles, et al. 1991;
Weiss & Kipnes,

2006).
However, w
hile this
design can compare different types of students and schools, it lacks

the ability to measure change
over time.


A small but influential number of studies have addressed the limitations of both of these
strategies by using a longitudinal comparative framework to investigate how the transition to
middle school affects early a
dolescents. Four of these studies are noteworthy. First, in one of
the early studies to undertake such a comparison over time, Blyth et al. (1978) found that
seventh
-
grade males in junior high school do not experience the growth in self
-
esteem of the K

8

seventh graders, nor do they experience the decline in self
-
esteem of the females in junior high
school. Second,
Byrnes and Ruby (2007) use data from over 40,000 students collected over five
years from 95 different schools serving eight grade students in

Philadelphia. Controlling for
prior achievement, their multilevel models revealed that 1) established K
-
8 schools had higher
levels of achievement than new K
-
8 schools or middle schools and 2) that this advantage
MIDDLE SCHOOL MODEL

11

decreased when controlling for students’
demographics. Third, also using data from Philadelphia,
Weiss and Baker
-
Smith

(2010) found that, controlling for prior achievement, students who
attended middle school in eighth grade performed worse in ninth grade, relative to peers who
attended K
-
8 schoo
ls. However, a substantial portion of this K
-
8 advantage is due to differential
likelihood of attendance at the district’s magnet high schools.
Fourth and most recently, Rockoff
and Lockwood (2010)
examined ten years of data made available for New York Cit
y school
children

and

concluded that when students move to a middle school, their academic achievement
fell substantially relative to that of their counterparts who continued to attend K
-
8 schools.
Though all four studies have high degrees of internal val
idity,
none of these studies employed
data from nationally representative samples of children and adolescents, thereby limiting the
extent to which these results are generalizable to larger populations.

Research Questions and Hypotheses

This research addresses this shortcoming and builds on the existing literature by asking
two questions using longitudinal data derived from a nationally representative sample of
children. First, what is the relationship between schools’ middle level grade

span and students’
achievement? Second, how does this relationship vary across different types of students, defined
in terms of their social and academic backgrounds?

It is expected that, after controlling for select student
-

and school
-
level characteri
stics,
there will be no main effect of middle level grade span on students’ achievement. However, in
reference to the second question, it is expected that students with disadvantaged social and
academic backgrounds will academically benefit from attending

a K
-
8 school.

Research Methods

Sample

MIDDLE SCHOOL MODEL

12

The data used to address these questions are drawn from the Early Childhood
Longitudinal Study, Kindergarten Class 1998
-
99 (ECLS
-
K). The ECLS
-
K focuses on children’s
early school experiences beginning with kindergar
ten and ending with eighth grade. ECLS
-
K is a
multi
-
method and multisource study that includes: interviews with parents, the collection of data
from principals and teachers, and student records, as well as direct child assessments. In the
eighth
-
grade data

collection, a student questionnaire was also included.

To reduce the dimensionality of the research questions to a manageable level and to fully
exploit the study’s panel design, child
-
level data from the spring
-
fifth
-
grade (round six), and
spring
-
eigh
th grade (round seven) waves are used. The final analytic sample was derived from
those students who: 1) had a non
-
zero longitudinal weight; 2) had valid scores on the spring
-
eighth grade assessments; and 3) attended a comprehensive public school in the sp
ring
-
eighth
grade, which excludes magnet and charter schools. This subsampling strategy resulted in a
final
analytic sample that includes 5,585 children nested within 876 public schools.
The final
analytical sample can be generalized to
children who atten
ded kindergarten in the United States
in the 1998
-
99 school year or attended first grade in the United States in the 1999
-
2000 school
year. All analyses include the appropriate longitudinal child
-
level weight.

Variables


Dependent variables.
While much research has examined the relationship between
middle school and early adolescents’ non
-
cognitive outcomes, such as self
-
esteem and locus of
control, this study exclusively focuses on cognitive outcomes, specifically achievement in two
school
-
sp
ecific subject areas.

The dependent variables are students’ scores on the mathematics
and reading assessments administered in the spring of the eighth grade year. Although ECLS
-
K
includes information on both cognitive and non
-
cognitive outcomes, achieveme
nt in
MIDDLE SCHOOL MODEL

13

mathematics and reading were chosen for two reasons. First, both subjects are important to
students’ future success. In addition, they are very different from one another and therefore
differentially influenced by school effects.

These assessments
maximized the accuracy of measurement that could be achieved in a
limited amount of testing time while minimizing floor and ceiling effects by matching sets of test
questions to initial estimates of students’ achievement (
Tourangeau, et al., 2009)
. The tes
ts’
specifications were derived from a variety of sources, including national and state performance
standards in each domain. The scope and sequence of the materials from state assessments, as
well as from major publishers was also considered. As opposed
to using the
tests’
scale scores,
which are most useful in cross
-
sectional analyses, this study uses theta scores (Reardon, Cheadle,
& Robinson, 2009)
, which
have three advantages: they are 1) approximately interval scaled; 2)
normally distributed at each
assessment wave and; 3) less dependent on the particular test items
included on the assessment. Checks on the reliability and validity of both assessments are
reported in
Najarian, Pollack, Sorongon, and Hausken (2009).

Independent variables.

There are
several student
-
level socio
-
demographic and academic
background covariates employed as predictors.
Children’s socio
-
economic status is measured
with a composite of parents’ income, education, and occupational prestige (a z
-
score [M=0,
SD=1]). The analyses
also employ a dummy
-
coded gender measure (boys=1, girls=0) and a
measure indicating whether the child was a member of a traditionally under
-
performing
racial/ethnic group (White and Asian children=1, Hispanic, African
-
American, Native
-
American
and multi
-
ra
cial children=0). In addition, the models also include a measure of whether the child
lived in a single
-
parent home (yes=1, no=0); and whether English was the student’s native
MIDDLE SCHOOL MODEL

14

language
(yes=1, no=0). The final variable related to students’ academic backgro
und is whether
the child has a disability (yes=1, no=0).

At the school
-
level, the models also employ variables that reflect the schools’ socio
-
demographic and academic characteristics. In terms of the schools’ social environment, the
primary covariate of

interest is the grade span of students’ spring
-
eighth grade school.

These
indicator variables were constructed from two different items on the
spring
-
eighth grade
administrator’s survey, which asked respondents to indicate 1) the school’s lowest grade fr
om
seven response options and 2) the school’s highest grade from three response options. These 21
possible combinations were collapsed into five non
-
overlapping categories that best reflect the
landscape of middle level grade configurations. These grade s
pan categories include: 1) K
-
8
schools with at least one grade lower than fifth and no higher than eighth; 2) 6
-
8 schools, which
includes schools with no grade lower than sixth and not higher than eighth; 3) 7
-
8 schools with
no grade lower than seventh and

no grade higher than eighth; 4) grades 7
-
12 includes those
schools with no grade lower than seventh and at least one grade higher than eighth and; 5) K
-
12
schools that have at least one grade lower than seventh and one grade higher than tenth. Grades
6
-
8

schools, the modal category and the standard configuration of the middle school, serve as the
referent.

In addition to this school
-
level covariate, a number of others are employed as statistical
controls. School size and cohort size are especially note
worthy. First, a series of indicator
variables are used to measure school size: small schools (<
150 children); small
-
medium schools
(150
-
299); medium
-
sized schools (300
-
499); medium
-
large schools (500
-
749) and; large schools
(>750). In the multivariate ana
lyses that follow, large schools are the referent category.

Second,
cohort size is measured using an indicator variable for whether there are more than 180 students
MIDDLE SCHOOL MODEL

15

in the eighth grade
(yes=1, no=0).

The school
-
level models also incorporate socio
-
demogra
phic
controls for high
-
minority enrollment (a dummy variable indicating Hispanic or African
-
American enrollments at or above 25%) and a standardized measure for the percentage of
students receiving free lunch
(a z
-
score [M=0, SD=1]).


Analytic Procedure


H
ierarchical linear models with random intercepts are used to estimate the relationship
among these variables and properly account for the nested nature of the data. They also include
controls for the ECLS
-
K stratified sampling design and for the probabili
ty of selection for
individuals. In these models, a common set of predictor variables is used consisting of
respondents’ and schools’ social and academic characteristics. The models also take advantage
of the panel design of the ECLS
-
K study by controllin
g for respondents’ status on the dependent
variables in the spring
-
fifth grade in predicting their eighth
-
grade outcomes. This control not
only allows one to specify more precisely transition effects, but also provides a control for any
effects that might
have resulted from previous school transitions. To address the issue of missing
values common to complex panel designs, p
ooled estimates derived from these models are based
on
twenty

imputed data sets that have been constructed using
multivariate normal
regression
(Rubin,

1987). This helps minimize bias due to missing values (ranging from 0
-
10% on the
independent variables) and retain the largest possible sample size.

Results

The results reflect a variety of relationships between grade span configurati
on and
achievement in reading and mathematics. While somewhat muddled, two broad conclusions can
be derived from this work. First, there is little evidence that the K
-
8 model

or any other grade
span configuration, for that matter

is associated with great
er gains in students’ ability in either
MIDDLE SCHOOL MODEL

16

mathematics or reading. Student growth in these cognitive areas is remarkably similar across
schooling forms. Second, it is also not evident that the K
-
8 model best serves early adolescents
from disadvantaged backgro
unds. To help disentangle these relationships, descriptive results for
all analytic variables are presented first, then the main effects of grade span configurations are
presented next. Finally, to test the second of two hypotheses, the results of the c
ross
-
level
interaction are reported.

Descriptive Results


Descriptive statistics for all analytical variables are presented in Table 1. Three points
are of interest. First, differences in average math and reading scores show that students in
middle sc
hools are higher than for those in the types of school forms current policy is promoting.
Students in the K
-
8 and K
-
12 models have the lowest mean scores for both the spring
-
fifth and
spring
-
eighth grade assessments in math and reading. However, these diff
erences may be related,
at least in part, to compositional differences. Second, compared to either the grades 6
-
8 or 7
-
8
models, grades K
-
8 schools have more students eligible for free lunch, and are more likely to be
a high minority school. In addition,
K
-
8 schools are also likely to be smaller in total enrollment
and have a cohort size less than 180. Third, mirroring these school
-
level descriptive statistics,
children in grades K
-
8 schools have lower mean SES scores and are more likely to have a
disabil
ity. Compared to their counterparts in the 6
-
8 or 7
-
8 models, they are also more likely to
come from single parent homes.

In sum, the descriptive statistics suggest that the five different grade span configurations
serve different types of students and
also vary on critical school
-
level characteristics. Therefore,
subsequent multilevel analyses must control for these
socio
-
demographic and academic
background characteristics.

MIDDLE SCHOOL MODEL

17

[INSERT TABLE 1 ABOUT HERE]

Main Effects


Table 2 presents four different mo
dels that test the first hypothesis, examining the
relationship between schools’ grade span and students’ learning in math and reading. Model 1
reports on the unconditional model with no predictors at either the student
-

or school
-
level. The
average stud
ent in the average school scores about 1.46 in math and 1.31 in reading. The derived
intraclass correlation from the math model is 0.17 and for the reading model is 0.15, which is
generally in line with estimates ranging from 10
-
30% in the school
-
effects l
iterature and
substantiating the use of multilevel models.


Model 2 incorporates a number of student
-
level
socio
-
demographic and academic
background characteristics. Unsurprisingly, scores from the spring
-
fifth grade are strong and
significant predictors of spring
-
eighth grade achievement in both math and reading. Note also
that the amount of between
-
school di
fferences have been reduced by about two
-
thirds. There are
two unexpected results from this model. First, Asian and White students are predicted to score
lower in math (ß =
-
0.02,
t
=
-
2.66,
p

= 0.008) and reading (ß =
-
0.05,
t
=
-
6.61,
p

< 0.001).
Althou
gh these unexpected relationships are statistically significant, their substantive
significance is minimal, as the coefficient indicates a difference of .02 or .05 standard deviations
in the outcomes. Second, the relationship between these characteristics
varies by subject area.
For example, while there is a significant negative relationship between single parent and math
scores, that relationship is weak and non
-
significant when it comes to reading scores. These
inconsistencies reflect the need to examin
e students’ learning in more than one cognitive domain.


Model 3 employs the same socio
-
demographic and academic background characteristics,
but includes a small number of school
-
level characteristics. The one school
-
level variable of
MIDDLE SCHOOL MODEL

18

note is
standardiz
ed measure for the percentage of students receiving free lunch. The
relationship of this variable to students’ scores also varies: it has a small, but significant
relationship with math scores
(ß =
-
0.01,
t
=
-
2.68,
p

= 0.007), but a weaker and non
-
significant
relationship with reading scores (ß =
-
0.01,
t
=
-
1.89,
p

= 0.059). Not reported in Model 3 are the
results for the variables indicating cohort and total school size, neither of which had a significant
relations
hip to either dependent variable. The intraclass correlation for both reading and math
models has decreased substantially with the addition of these predictors. For math, the level of
the intraclass correlation has decreased from 0.17 to 0.06, while for re
ading the change is from
0.15 to 0.056.


The final model reported in Table 2 directly tests the first hypothesis. This model
includes the student
-

and school
-
level controls incorporated into the two preceding models. As
hypothesized, after controlling fo
r these key characteristics, the K
-
8 model, compared to the
grades 6
-
8 model, is not significantly associated with gains in either math or reading. In fact,
only the grades 7
-
8 model has a slight association with increases in reading scores, though this
r
elationship is just short of conventional significance (ß = 0.02,
t
= 1.73,
p

= 0.084). The grades
K
-
12 model, conversely, has a larger association with decreases in reading scores, but this, too, is
non
-
significant (ß =
-
0.04,
t
=
-
1.96,
p

= 0.051). In
short, the results from this final model report
that compared to the grades 6
-
8 model, the other four models offer no advantage when it comes
to achievement in math or reading. Moreover, the inclusion of the school form predictors in
Model 4 yields no cha
nge in the amount of between
-
school variance. For example, in reference
to math scores, the proportion of school
-
level variance explained by the covariates is
approximately 87%, and the proportion of student
-
level variance explained is about 70%;
however,

this amount is identical to what was explained in the two previous models. Taken
MIDDLE SCHOOL MODEL

19

together, the figures presented in Table 2 suggest little benefit to students from any particular
schooling form.

[INSERT TABLE 2 ABOUT HERE]

Cross
-
level Interaction Effects


However, despite the suggestions of policymakers promoting the K
-
8 schooling form,
models estimating average grade span effects across different types of students and schools may
miss some of the effects of the school type. Rather, we might be missing a
story of how these
impacts vary across different types of students. Because recent previous research has linked K
-
8
models to achievement in urban districts, which disproportionally serve disadvantaged
populations of students, the next set of models direc
tly test whether the K
-
8 model benefits
precisely these types of students. Theory suggests that these models are more apt to provide a
“zone of comfort” (Simmons & Blyth, 1987) that both buffers early adolescents against the
developmental changes they are

experiencing and provides access to more stable, personable
relations. Those from disadvantaged backgrounds would likely benefit most from these settings,
as these environments may offset the adverse effects of their
socio
-
demographic and academic
backgr
ounds. Tables 3 and 4 report the results of the models that directly test whether these
students benefit from attending K
-
8 schools, or any other model, for that matter.


There is, however, mixed evidence in support of this hypothesis. For example, Tab
le 3,
Model 5, reports on the cross
-
level interactions between variables indicating single parent and
grade span. Thus, the coefficients associated with each interaction term represent a unique effect
of single parent for each of the four grade span indic
ator variables. None of these interactions
are significant predictors of math scores, and only the interaction between single parent and the
grades 7
-
8 model is significantly associated with reading scores, albeit negatively (ß =
-
0.04,
t
=
MIDDLE SCHOOL MODEL

20

-
2.11,
p

= 0.0
35). A similar picture emerges in Model 6, which examines the interaction
between native English
-
speaking students and grade span. Here, one would expect significant
positive relationships between the interaction terms and both outcomes, with native Engl
ish
language speakers having an advantage across the different models. This advantage is only
present for native English
-
speaking students in the K
-
8 model (ß = 0.07,
t
= 2.17,
p

= 0.030),
suggesting that those for whom English is not one’s native languag
e would be associated with
decreases in math scores in K
-
8 schools. If anything, this result casts doubt on the notion that
K
-
8 schools provide a more appropriate fit for disadvantaged students, specifically those that are
non
-
native speakers of English
.

[INSERT TABLE 3 ABOUT HERE]


Model 7, however, provides a nice contrast. This model includes interaction terms
between the disability and grade span indicator variables. In addition to reporting a small, but
significant relationship between students
in the grades 7
-
8 model and reading scores (ß = 0.02,
t
= 2.12,
p

= 0.034), it also shows a strong and significant relationship for the interaction between
disability and the K
-
8 model (ß = 0.06,
t
= 2.36,
p

= 0.018). This is the strongest evidence in
s
upport of the K
-
8 model for disadvantaged students, specifically for those students who have
been identified as having a disability. But, this relationship only applies to reading scores, as
there is no significant relationship between this interaction te
rm and math scores.


The final model in Table 4 reports the results for the interaction terms between minority
status and grade spans. If it were advantageous for minority students to attend one grade span
versus another, one would expect a significan
t negative relationship between the non
-
minority
and grade span interaction terms and the two dependent cognitive outcomes. For all the four
interaction terms and both dependent variables, the relationships are weak and statistically non
-
MIDDLE SCHOOL MODEL

21

significant: none

of the grade spans provide an advantage for
Hispanic, African
-
American,
Native
-
American and multi
-
racial students (those coded as 0). Not reported in this table are the
results for the interaction between SES and the grade span indicator variables. Howe
ver, similar
to the results reported in Model 8, none of these interaction coefficients were significant.


In sum, the results reported in Models 5
-
8 provide minimal evidence that any one grade
span model best serves students with disadvantaged socio
-
demog
raphic and academic
backgrounds. Therefore, there is little support for the second hypothesis. The only noteworthy
result was the small, but significant relationship for those students with disabilities in the K
-
8
model. What do these results have to s
ay about the presumed benefits of the K
-
8 model versus
the others?

[INSERT TABLE 4 ABOUT HERE]

Discussion

Taken together, the results of this study call into question the policy focus on changing
the schooling form in which middle grades students are edu
cated. Using a nationally
representative dataset, we have estimated an array of models to detect differences in student
performance. However, we have found very few differences in student performance by the type
of middle grades school they attend.


While

finding little evidence in support of either hypothesis, this research contributes to a
broader conversation on middle grades schooling and adolescent development in two important
ways. First, it moves research attention away from detecting average effec
ts across all early
adolescents and focuses on the effects for early adolescents with certain characteristics, in this
case those with disadvantaged socio
-
demographic or academic characteristics who are also the
target of most school
-
level reforms. On thi
s point, there was one notable finding: students
MIDDLE SCHOOL MODEL

22

identified as having a disability benefitted from the K
-
8 grade span. This is an example of how
the fit between these students and the K
-
8 environment might be more appropriate given the
needs of this speci
fic subpopulation of students. Findings such as this should encourage
researchers to move beyond average effects and identify, more specifically, how certain school
-
level reforms differentially impact different types of adolescents.

In light of this fin
ding, more research is needed to pinpoint the mechanisms through
which K
-
8 models might benefit students with disabilities. One set of possible
explanations
focuses on how K
-
8 models influence adolescents’ and teachers’ social relations. For example,
thes
e explanations draw attention to how schools’ grade spans influence teachers’ ability to align
their instruction and curriculum with other teachers’ expectations. Nonetheless, investigations of
these mechanisms are needed to determine whether and to what d
egree these malleable factors
mediate the relationship between middle schools’ grade span and students’ outcomes.


The second contribution of this work corroborates and extends previous discussions
(Eccles & Roeser, 1999, 2009) on the ecological components

of school systems and their
influence on adolescent development. Eccles and Roeser (1999, 2009) parse the complex
hierarchy of school systems into levels, specifically the classroom level and the school/district
level. At the classroom level, several nuan
ced components, such as instruction, emotional
support, management and motivational climate, teacher beliefs, efficacy and expectations among
other characteristics have proximal predictive power on student outcomes. At the school/district
level, attributes

such as overall school climate, school size, curricular differentiation, and middle
grade span/transitions are thought to be largely predictive as well. However, in specifically
examining middle grade span while considering the developmental needs of ado
lescents, this
study suggests that decade long efforts made to change the grade span of schools as a means to
MIDDLE SCHOOL MODEL

23

better meet the developmental needs of early adolescents at the school/district level are likely
better spent at the classroom level through quali
ty teacher
-
student interactions and differentiated
instruction.

This work re
-
focuses researchers’ attention on the classroom
-
level “best practices” that
were originally thought to be the main advantage of middle schools. Given that the classroom is
the pr
imary arena through which students engage subject matter, this is the level where early
adolescents need to have the most appropriate developmental fit. The school and its grade span
may be too large of an aggregate category to detect any meaningful effec
ts. Rather, the
developmental needs of early adolescents, particularly those with disadvantaged socio
-
demographic or academic characteristics, need to be addressed at the level of the classroom. It is
there where factors such as maintaining an intense fo
cus on academic achievement, proactive
intervention, and teacher competencies matter most for adolescents. Incidentally, these factors,
as well as a small number of others, are those that are associated with the highest performing
middle schools, regardle
ss of their grade span (Williams, Kirst, & Haertel, 2010). This requires
researchers to move beyond school
-
level factors and towards classroom level practices that can
be adapted to suit the varied developmental needs of early adolescents.

While sh
ifting attention away from large
-
scale structural factors such as grade span, it is
also important to consider just how much an effect, on average, one can expect from such
changes. While Rockoff and Lockwood (2010) report sizable positive effects of K
-
8 s
chools,
Byrnes and Ruby’s (2007) estimates are much more modest. While the results reported here
generally found no such advantage, any differences between students had little to do with
schools. Rather, much of the difference can be attributed to studen
t
-
level factors. Specifically,
Model 1, which includes no explanatory variables at either level, shows that about 85 percent of
MIDDLE SCHOOL MODEL

24

the variation in math and 83 percent in reading can be attributed to the student
-
level. This
finding supports Byrnes and Ruby’
s (2007) call for a more reasonable expectation as to what
effects school systems and broad school
-
based reforms can have on student achievement. If a
majority of variation in achievement can be attributed to students themselves, current methods
that tie
school performance to rewards and consequences may be misguided. One reason why
Model 4, the final model that tests the first hypothesis, explains about 70 percent of the between
-
student variation in math scores is that the model includes covariates perta
ining to students’
socio
-
demographic characteristics, SES for example. These are factors that schools and their
administrators cannot address at the school
-
level.

While recognizing the contributions of this study, caution is warranted when interpreting
the modest results. Three limitations are noteworthy. First, the models employed in this study
do not include measures of these critical classroom
-
level practices.
During the ECLS
-
K spring
-
eighth grade data collection, one teacher
-
level background and three child
-
level subject matter
(i.e., English, mathematics, and science) questionnaires were used to collect data from the
sampled children’s teachers. The third sec
tion on these questionnaires included questions about
the instructional practices in the classroom, such as specific instructional activities and curricular
focus, and assigned books and textbooks. In this last section, the items specified activities and
p
ractices that were relevant to the subject domain. Future analyses will seek to incorporate
measures related to these classroom practices in specific subject domains to better account for
their influence on early adolescents’ achievement and whether these

influences vary by schools’
grade span.

Second, this study focused exclusively on cognitive outcomes, specifically achievement
in math and reading. Historically, much of the research on middle schools and early adolescent
MIDDLE SCHOOL MODEL

25

development has focused on no
n
-
cognitive outcomes such as self
-
perceptions, school
attachment, and achievement motivation, among others. Therefore, it is important to note that
the lack of relationship between grade span and the achievement measures does not extend these
non
-
cogniti
ve outcomes. For example, the benefits of K
-
8 schools reported by Authors (2006)
centered on non
-
academic outcomes such as self
-
esteem and whether one was the target of a
threat. These outcomes may also be equally important from a policy perspective. He
re, too,
future analyses using the
ECLS
-
K should consider the relationship between grade span and these
important outcomes.

Finally,
s
tudents are not randomly assigned to schools in the ECLS
-
K, and so these data
have the same potential selection bias as
all other observational studies. To limit the magnitude
of this bias, this study employs the standard strategy of using control variables that have been
associated with students’ academic achievement in previous research. As with all analyses
based on ob
servational data (and even for some studies based on randomized experimental data),
caution must be exercised in interpreting the few estimated significant effects as causal; it is
through the accumulation of similar estimates from studies with varying dat
a and alternative
methodologies that a conclusion that estimated effects are indeed causal becomes substantiated.

These limitations notwithstanding, there are three immediate implications to be derived
from this research. First, the results should give pause to reformers who are considering whole
-
scale changes to the ways in which grade spans are organized for early
adolescents. Reforms
such as these are very costly, and their effects are not uniformly beneficial

nor may they be
beneficial in the aggregate. At best, large districts that can provide a different number of grade
span configurations for early adolescent
s should do so in a deliberate manner that contributes to
MIDDLE SCHOOL MODEL

26

a more consistent evidentiary base. The results reported here question whether districts should
move beyond these relatively small pilot efforts and adopt the K
-
8 model on a larger scale.

Second,

these results should further remind reformers that the K
-
8 model, or other grade
span model serving early adolescents,

is not a one size fits all solution. School
-
level reform must
be carefully constructed in each locale to reflect both individual capaci
ty and needs. Because of
the doubly difficult challenges that disadvantaged early adolescents must confront, creating the
right fit for these students at a developmentally tumultuous time requires flexibility that few
districts can afford to provide. Whi
le certainly not suggesting that all districts convert their
middle school from one model from another, it may be that districts that can afford to provide a
range of different configurations would be well positioned to offer these options in a way that
be
st match students’ characteristics. This would require the research community to identify how
the effects of the configurations vary across different types of students. This study has attempted
to do just that. This study also points to the fact that th
ere is much to be learned about how and
why the K
-
8 model, as reported by these data, is advantageous for students with disabilities.

Third, this study also highlights the importance of employing a comparative longitudinal
framework on nationally represe
ntative data. This is one of a small number of studies to compare
students in different middle school models that also makes use of previous data on students’
achievement. Moreover, the results

or, the lack thereof

have a level of external validity that
f
ew studies can match. This is clearly a strength of the ELCS study design. While a number of
other studies have employed a comparative longitudinal framework, few have done so using
nationally representative data. Large
-
scale complex survey data, despi
te its observational nature
and inability to infer causality, provide a unique opportunity to examine a wide range of student
-

and school
-
level characteristics.

MIDDLE SCHOOL MODEL

27

References


Alspaugh, J. W. (2001). Achievement loss associated with the transition to middl
e school and high
school.
Journal of Educational Research, 92
(1), 20
-
25.

Anderman, E. M., & Maehr, M. L. (1994). Motivation and schooling in the middle grades.
Review of
Educational Research, 64
(2), 287
-
309.

Barber, B. K., & Olsen, J. A. (2004). Assessin
g the transitions to middle and high school.
Journal of
Adolescent Research, 19
(1), 3
-
30.

Blyth, D. A., Simmons, R. G., & Bush, D. (1978). The transition into early adolescence: A longitudinal
comparison of youth in two educational contexts.
Sociology of
Education, 51
(3), 149
-
162.

Byrnes, V., & Ruby, A. (2007). Comparing achievement between K
-
8 and middle schools: A large
-
scale
empirical study.
American Journal of Education, 114
(1), 101
-
135.

Cook, P. J., MacCoun, R., Muschkin, C., & Vigdor, J. (2008). Th
e negative impacts of starting middle
school in sixth grade.
Journal of Policy Analysis and Mangement, 27
, 104
-
121.

Cuban, L. (1992). What happens to reforms that last? The case of the junior high school.
American
Educational Research Journal, 29
(2), 227
-
251.

Deci, E. L., & Ryan, R. M. (1994). Promoting self
-
determined education.
Scandinavian Journal of
Educational Research, 38
(1), 3
-
14.

Eccles, J., Lord, S., & Midgely, C. (1991). What are we doing to early adolescents? The impact of
educational contexts

on early adolescents.
American Journal of Education, 99
(4), 521
-
542.

Eccles, J. S., & Roeser, R. (1999). School and community influences on human development. In M.
Bornstein & M. Lamb (Eds.),
Developmental Psychology: An advanced textbook

(4th ed., pp.
503
-
554). Mahwah, NJ: Lawrence Erlbaum. .

Eccles, J. S., & Roeser, R. W. (2009). Schools, Academic Motivation, and Stage
-
Environment Fit. In R.
MIDDLE SCHOOL MODEL

28

M. Lerner & L. Steinberg (Eds.),
Handbook of Adolescent Psychology, 3rd Edition
. Hoboken,
NJ: Wiley.

Eccles, J.
S., Wigfield, A., Midgley, C., & Reuman, D. (1993). Negative effects of traditional middle
schools on students' motivation.
The Elementary School Journal, 93
, 553
-
574.

French, S. E., Seidman, E., Allen, L., &

Aber, J. L. (2006). The development of ethnic identity during
adolescence.
Developmental Psychology, 42
(1), 1
-
10.

Friedel, J. M., Cortina, K. S., Turner, J. C., & Midgley, C. (2010). Changes in efficacy beliefs in
mathematics across the transition to mid
dle school: Examining the effects of perceived teacher
and parent goal emphases.
Journal of Educational Psychology, 102
(1), 102
-
114.

Giordano, P. C. (2003). Relationships in adolescence.
Annual Review of Sociology, 29
, 257
-
281.

Goldin, C. (1999). A brief

history of education in the United States
NBER Historical Paper

(Vol. 119).
Cambridge, MA.

Grills
-
Taquechel, A. E., Norton, P., & Ollendick, T. H. (2010). A longitudinal examination of factors
predicting anxiety during the transition to middle school.
Anx
iety, Stress & Coping: An
International Journal, 23
(5), 493
-
513.

Hanushek, E. A., Kain, J. F., & Rivkin, S. G. (2004). Disruption versus tiebout improvement: The costs
and benefits of switching schools.
Journal of Public Economics, 88
, 1721
-
1746.

Hough, D. L. (2005). The rise of the "Elemiddle" schools: Not every K
-
8 school truly applies best
middle
-
level practices and deserves the designation.
School Administrator, 62
(3), 10
-
13.

Juvonen, J., Le, V.
-
N., Kaganoff, T., Augustine, C. H., & Constant,
L. (2004). Focus on the Wonder
Years: Challenges Facing the American Middle School. Santa Monica, CA: RAND Corporation.

Mac Iver, M. A., & Mac Iver, D. J. (2006). Which bets paid off? Early findings on the impact of private
management and K
-
8 conversion re
forms on the achievement of Philadelphia students.
Review of
MIDDLE SCHOOL MODEL

29

Policy Research, 23
(5), 1077
-
1093.

Maehr, M. L., & Midgley, C. (1996).
Transforming school cultures
. Boulder, CO: Westview Press.

Najarian, M., Pollack, J. M., Sorongon, A. G., & Hausken, E. G.
(2009).
Early Childhood Longitudinal
Study, Kindergarten Class of 1998
-
99 (ECLS
-
K) Psychometric Report for the Eighth Grade
.
Washington, D.C.: National Center for Education Statistics.

Poncelet, P., & Metis Associates (2004). Restructuring schools in Clev
eland for the social, emotional,
and intellectual development of early adolescents.
Journal of Education for Students Placed at
Risk, 9
(2), 91
-
96.

Reardon, S. F., Cheadle, J. E., & Robinson, J. P. (2009). The effect of Catholic schooling on math and
readi
ng development in Kindergarten through fifth grade.
Journal of Research on Educational
Effectiveness, 2
(1), 45
-
87.

Rockoff, J. E., & Lockwood, B. (2010). Stuck in the middle: Impacts of grade configurations in public
schools.
Journal of Public Economics,
94
(11
-
12), 1051
-
1061.

Rubin, D. B. (1987).
Multiple imputation for nonresponse in surveys
. New York: John Wiley & Sons,
Inc.

Rudolph, K. D., Lambert, S. F., Clark, A. G., & Kurlakowsky, K. D. (2001). Negotiating the transition
to middle school: the role o
f self
-
regulatory processes.
Child Development, 72
, 929
-
946.

Rumberger, R. W. (1995). Dropping out of middle school: A multilevel analysis of students and schools.
American Educational Research Journal, 32
(3), 583
-
625.

Seidman, E., Aber, J. L., Allen, L.
, & French, S. E. (1994). The impact of school transitions in early
adolescence on the self
-
system and social context of poor urban youth.
Child Development,
65
(2), 507
-
522.

Seidman, E., Aber, J. L., Allen, L., & French, S. E. (1996). The impact of the tr
ansition to high school
MIDDLE SCHOOL MODEL

30

on the self
-
system and perceived social context of poor urban youth.
American Journal of
Community Psychology, 24
(4), 489
-
515.

Simmons, R. G., & Blyth, D. A. (1987).
Moving into adolescence: The Impact of pubertal change and
school

context.

New York: De Gruyter.

Sirin, S. R. (2005). Socioeconomic status and academic achievement: A meta
-
analytic review of
research.
Review of Educational Research, 75
(3), 417
-
453.

Theriot, M. T., Craun, S. W., & Dupper, D. R. (2010). Multilevel evalua
tion of factors predicting school
exclusion among middle and high school students.
Children and Youth Services Review, 32
(1),
13
-
19.

Tourangeau, K., Nord, C., Lê, T., Sorongon, A. G., & Najarian, M. (2009). Early Childhood
Longitudinal Study, Kindergarten

Class of 1998

99 (ECLS
-
K), combined user’s manual for the
ECLS
-
K eighth
-
grade and K

8 full sample data files and electronic codebooks. Washington, DC:
National Center for Education Statistics, Institute of Education Sciences.

Way, N., Reddy, R., & Rhodes,

J. (2007). Students’ perceptions of school climate during the middle
school years: Associations with trajectories of psychological and behavioral adjustment.
American Journal of Community Psychology, 40
(3/4), 194
-
213.

Weiss, C. C., & Baker
-
Smith, E. C. (
2010). Eighth grade school form and resilience in the transition to
high school: A comparison of middle schools and K
-
8 schools.
Journal of Research on
Adolescence, 20
(4), 825
-
839.

Weiss, C. C., & Kipnes, L. (2006). Reexamining middle school effects: A co
mparison of middle grades
students in middle schools and

K
-
8 schools.
American Journal of Education, 112
(2), 239
-
272.

Williams, T., Kirst, M., &

Haertel, E. (2010). Gaining Ground in the Middle Grades: Why Some
Schools Do Better: A Large
-
Scale Study of Middle Grades Practices and Student Outcomes
MIDDLE SCHOOL MODEL

31

Middle Grades Study
. Mountain View, CA: EdSource.

MIDDLE SCHOOL MODEL

32


Tables


Table 1

Weighted Descriptive Statistics
Organized by Middle Level Grade Span (
N

= 5,585)



K
-
8

% or Mean

(SD)

6
-
8

% or Mean

(SD)

7
-
8

% or Mean

(SD)

7
-
12

% or Mean

(SD)

K
-
12

% or Mean

(SD)

Overall

% or Mean

(SD)

Student
-
level







N

(%)

8.55

59.07

19.08

8.75

4.54



Math 8
th

Grade

1.251

(0.407)

1.348

(0.453)

1.451

(0.401)

1.393

(0.404)

1.233

(0.508)

1.384

(0.447)


Reading 8
th

Grade

1.010

(0.351)

1.216

(0.369)

1.259

(0.360)

1.291

(0.370)

1.147

(0.456)

1.245

(0.379)


Math 5
th

Grade

1.147

(0.440)

1.310

(0.483)

1.416

(0.422)

1.324

(0.450)

1.191

(0.537)

1.344

(0.469)


Reading 5
th

Grade

1.107

(0.340)

1.230

(0.341)

1.264

(0.313)

1.280

(0.347)

1.184

(0.407)

1.252

(0.346)


SES

-
0.449

(0.691)

-
0.238

(0.748)

-
0.097

(0.804)

-
0.252

(0.761)

-
0.485

(0.773)

-
0.199

(0.777)


Native English


77.41

79.64

80.76

91.99

87.43

83.60


Male


57.45

52.88

52.05

56.02

54.63

52.30


Disability

20.69

17.31

15.42

14.30

21.39

17.14

MIDDLE SCHOOL MODEL

33



Single Parent


30.56

25.94

25.46

31.88

30.99

24.63


Non
-
m
inority


55.36

56.90

58.52

56.08

59.91

61.98

School
-
level








Free Lunch

0.481

(1.114)

0.224

(1.160)

0.017

(0.934)

0.085

(1.130)

0.411

(1.337)

0.175

(1.108)


High Minority


49.60


40.02

41.71

32.92

33.71

35.06


Cohort Size >180

13.40

78.83

91.36

49.81

5.98

58.80


Total Enrollment








<150

2.78

0.59

0.01

1.99

2.87

2.28


150
-
299

10.56

2.93

3.20

17.60

20.02

8.35


300
-
499

39.97

17.51

19.71

17.66

13.59

20.27


500
-
749

28.89

24.43

34.54

5.26

36.12

27.46


>750

17.81

54.41

42.32

57.49

27.40

41.50


Note.
SD

reporte
d only for continuous variables. Percentages reported for indicator variables. Descriptive statistics based on the
pooled estimates derived from 20 imputed data sets
.
MIDDLE SCHOOL MODEL

34

Table 2: Multilevel Model Predicting Students’ Eighth Grade Ach
ievement from Schools’ Middle Level Grade Span (
N

= 5,585)



Model 1

Model 2

Model 3

Model 4


Math

Reading

Math

Reading

Math

Reading

Math

Reading

Intercept

1.460
(0.009)***

1.312

(0.008)***

0.435

(0.016)***

0.378

(0.017)***

0.427

(0.019)***

0.396

(0.021)***

0.427

(0.022)***

0.413

(0.023)***

Fixed Effects










Student
-
level
a










Math 5
th

Grade



0.765
(0.008)***


0.765

(0.008)***


0.764

(0.008)***



Reading 5
th

Grade





0.763

(0.010)***


0.761

(0.010)***


0.761

(0.010)***


Disability



-
0.037

(0.009)***

-
0.055

(0.009)***

-
0.038

(0.009)***

-
0.056

(0.009)***

-
0.038

(0.009)***

-
0.056

(0.009)***


Native English



-
0.018

(0.009)***

-
0.015

(0.009)

-
0.021

(0.010)*

-
0.020

(0.009)*

-
0.021

(0.010)*

-
0.019

(0.009)*


Single Parent



-
0.021

(0.008)**

-
0.009

(0.008)

-
0.020

(0.008)*

-
0.009

(0.008)

-
0.021

(0.008)*

0.009

(0.008)



Non
-
minority



-
0.020

(0.008)**

-
0.050

(0.008)***

-
0.015

(0.008)

-
0.037

(0.008)***

-
0.015

(0.008)

-
0.377

(0.008)***


School
-
level
b










% Free lunch






-
0.012

(0.004)**

-
0.009

(0.005)

-
0.012

(0.005)*

-
0.008

(0.005)


K
-
8








-
0.004

(0.014)

-
0.007

(0.014)

MIDDLE SCHOOL MODEL

35


7
-
8








0.010

(0.010)

0.017

(0.010)


7
-
12








0.024

(0.015)

-
0.011

(0.014)


K
-
12








-
0.007

(0.022)

-
0.043

(0.022)

Random Effects










SD
(
between
-
schools
)

0.172

0.155


0.060

0.057

0.060

0.056

0.060

0.056


SD

(
between
-
students
)

0.401

0.341

0.220

0.217

0.220

0.217

0.220

0.216



Notes.
*
p
< .05
**
p

< .01
***
p

<

.001 (two

tailed tests)

Values in parentheses are standard errors.

a

Additional student
-
level covariates include SES and an indicator for male.

b

Additional student
-
level covariates include indicators for high minority, cohort size > 180, and total
school size. Grades 6
-
8 schools
are the reference category.



MIDDLE SCHOOL MODEL

36


Table 3: Multilevel Model Predicting Students’ Eighth Grade Achievement from Cross
-
level
Interactions between Single Parent and Grade Span and Native English Speakers and Grade Span
(
N

= 5,585)



Model 5

Model 6


Math

Reading

Math

Reading

Intercept

0.426

(0.022)***

0.411

(0.023)***

0.423

(0.023)***

0.411

(0.024)***

Fixed Effects
a






K
-
8

-
0.001

(0.015)

-
0.008

(0.015)

-
0.063

(0.030)*

-
0.026

(0.030)


7
-
8

0.014

(0.011)

0.026

(0.011)*

0.041

(0.021)*

0.025

(0.021)


7
-
12

0.027

(0.016)

-
0.009

(0.015)

0.029

(0.038)

0.027

(0.037)


K
-
12

-
0.007

(0.024)

-
0.034

(0.023)

0.046

(0.057)

-
0.042

(0.056)

Cross
-
level Interactions







Single Parent x
K
-
8

-
0.012

(0.027)

0.004

(0.025)




Single Parent x 7
-
8

-
0.019

(0.021)

-
0.043

(0.020)*




Single Parent x 7
-
12

-
0.012

(0.030)

-
0.010

(0.028)




Single Parent x
K
-
12

0.003

(0.041)

-
0.039

(0.037)




Native English x
K
-
8



0.070

(0.032)*

0.022

(0.032)


Native English x 7
-
8



-
0.038

(0.023)

-
0.009

(0.023)


Native English x 7
-
12



-
0.005

(0.039)

-
0.042

(0.039)


Native English x
K
-
12



-
0.056

(0.058)

-
0.001

(0.056)

MIDDLE SCHOOL MODEL

37

Random Effects






SD
(
between
-
schools
)

0.060

0.056

0.059

0.056


SD

(
between
-
students
)

0.220

0.216

0.220

0.216


Notes.
*
p
< .05
**
p

< .01
***
p

< .001 (two

tailed tests)

Values in parentheses are standard
errors.

a

Additional student
-
level covariates include SES and indicators for male, disability, native
English
-
speaker, single parent, and non
-
minority. Add
itional school
-
level covariates include
indicators for high minority, cohort size > 180, and total school size. Grades 6
-
8 schools are the
reference category

MIDDLE SCHOOL MODEL

38

Table 4: Multilevel Model Predicting Students’ Eighth Grade Achievement from Cross
-
level
Interact
ions between Disability and Grade Span and Non
-
minority and Grade Span (
N

= 5,585)



Model 7

Model 8


Math

Reading

Math

Reading

Intercept

0.427

(0.022)***

0.412

(0.023)***

0.428

(0.022)***

0.412

(0.023)***

Fixed Effects
a






K
-
8

-
0.004

(0.015)

-
0.018

(0.014)

-
0.005

(0.016)

0.000

(0.016)


7
-
8

0.010

(0.011)

0.022

(0.010)*

0.007

(0.012)

0.009

(0.012)


7
-
12

0.029

(0.016)

-
0.004

(0.015)

0.030

(0.016)

-
0.010

(0.015)


K
-
12

-
0.004

(0.023)

-
0.037

(0.022)

-
0.009

(0.024)

-
0.033

(0.023)

Cross
-
level Interactions







Disability x
K
-
8

0.002

(0.028)

0.064

(0.027)*




Disability x 7
-
8

-
0.000

(0.026)

-
0.034

(0.024)




Disability x 7
-
12

-
0.037

(0.038)

-
0.046

(0.036)




Disability x
K
-
12

-
0.020

(0.042)

0.028

(0.040)




Non
-
m
inority

x
K
-
8



0.004

(0.024)

-
0.025

(0.024)


Non
-
m
inority

x 7
-
8



0.007

(0.018)

0.023

(0.018)


Non
-
m
inority

x 7
-
12



-
0.023

(0.029)

0.000

(0.029)


Non
-
m
inority

x
K
-
12



0.010

(0.040)

-
0.037

(0.038)

Random Effects






SD
(
between
-
schools
)

0.060

0.056

0.060

0.056

MIDDLE SCHOOL MODEL

39


SD

(
between
-
students
)

0.220

0.216

0.220

0.216


Notes.
*
p
< .05
**
p

< .01
***
p

< .001 (two

tailed tests)

Values in parentheses are standard
errors.

a

Additional student
-
level covariates include SES and indicators for male, disability, native
English
-
speaker, single parent, and non
-
minority. Additional school
-
level covariates include
indicators for high minority, cohort size > 180, and total school siz
e. Grades 6
-
8 schools are the
reference category