Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
The role of the timing of school changes and school quality in the impact of student mobility: evidence from Clark County, Nevada
(USC Thesis Other)
The role of the timing of school changes and school quality in the impact of student mobility: evidence from Clark County, Nevada
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
1
THE ROLE OF THE TIMING OF SCHOOL CHANGES AND SCHOOL QUALITY IN THE
IMPACT OF STUDENT MOBILITY:
EVIDENCE FROM CLARK COUNTY, NEVADA
A DISSERTATION
PRESENTED TO THE FACULTY OF THE GRADUATE SCHOOL
OF THE UNIVERSITY OF SOUTHERN CALIFORNIA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR DEGREE OF
DOCTOR OF PHILOSOPHY (URBAN EDUCATION POLICY)
BY
RICHARD OSBOURNE WELSH
AUGUST 2015
2
The Role of the Timing of School Changes and School Quality in the Impact of Student
Mobility: Evidence from Clark County, Nevada
Richard Osbourne Welsh, Ph.D.
University of Southern California 2015
Student mobility, or the movement of students from one school to another, is a
widespread phenomenon in the U.S. (Mehana & Reynolds, 2004; Reynolds, Chen, & Herbers,
2009; United States Government Accountability Office, 2010). Frequent student mobility is most
common and likely to be damaging among low-income and minority students (Hanushek, Kain,
& Rivkin, 2004; Institute of Medicine and National Research Council, 2010; Reynolds et al.,
2009; Schwartz, Stiefel, & Chalico, 2009; Xu, Hannaway, & D’Souza, 2009). Nevertheless,
educational policymakers and researchers often make divergent assumptions regarding the
impact of changing schools. There is still much disagreement on whether student mobility is a
helpful or harmful phenomenon and there is no consensus about the effects of student mobility
(Grigg, 2012; Mehana & Reynolds, 2004). Student mobility has the potential to improve student
achievement in urban school districts. However, a better sense of the school changes that help or
hurt achievement in a complex landscape is needed to fulfill this potential. In essence, student
mobility is an important educational policy and equity issue given the ongoing demographic
shifts within urban school districts.
School quality and the timing of school changes are crucial factors in determining the net
impact of changing schools on student achievement. Students may switch schools at any time
between or during the academic year. The timing of school changes may play an important role
in the extent of the disruptive effect of changing schools and whether students transfer to higher
3
quality schools. However, the majority of the extant literature has focused on student mobility
between school years and little attention has been paid to differentiating the timing of school
changes. In sum, the when (timing) and where (school quality) of school changes appears to be
pivotal in determining whether changing schools have a beneficial or harmful effect on student
achievement.
Using student-level data from Clark County School District (CCSD) from 2007-08
through to 2012-13, this dissertation provides a comprehensive analysis of the prevalence and
impact of student mobility across the timing of school changes and directs much needed attention
to student mobility, an area of study in which there has been little research in CCSD.
1
CCSD
offers a fascinating social and geographic locale in which to study student mobility in a large
urban district. CCSD is the fifth largest and fastest growing urban district in the U.S. CCSD is
also the largest of Nevada’s 17 districts and has 70 percent of the state’s public school students.
Clark County is one of sixteen counties in Nevada and consists of five cities (Las Vegas, North
Las Vegas, Boulder City, Henderson, and Mesquite) and a number of surrounding smaller
jurisdictions. Though Clark County was once relatively small and homogenous, it is now a large
and diverse county with over two million people. CCSD has experienced a substantial increase in
Hispanic students and English Language Learners in the past two decades. CCSD has a
traditional governance structure with a locally elected school board operating most public
schools and school assignment is based on attendance zones.
2
It can be argued that CCSD is
1
According to a three-month “Educational and Operational Efficiency Study” that was privately funded by the
UCLA Dream Foundation and conducted by Gibson Consulting Group, an Austin-based educational research
company, CCSD has an in-district student mobility rate of over 30 percent.
2
CCSD has limited open enrollment in three forms: magnet schools, charter schools, and an open enrollment policy.
There are 25 magnet schools in the district (less than 10 percent of the total number of public schools in CCSD).
Students in CCSD may apply to magnet schools regardless of their residence and seats in magnet schools are
assigned based on a lottery. There are less charter schools than magnet schools in CCSD. Charter schools are open
to all students. There is also an open enrollment policy that provides choice for families that are not satisfied with
their neighborhood schools. Schools with available seats are publicized and parents may apply for a space in these
4
representative of the typical large urban district in the US in many regards including educational
governance structure, ethnic and racial heterogeneity and demographic shifts in recent decades.
In three interrelated empirical essays, this dissertation examines the role of the timing of
school changes and school quality in student mobility patterns and effects within an urban school
district. The first paper analyzes the exit and destination patterns of mobile students by students’
race/ethnicity, prior achievement, and school quality to determine whether differential mobility
patterns exist. The second paper estimates the impact of student mobility on the achievement of
mobile students using two different quasi-experimental methods (student fixed effects and
propensity score matching). The third paper investigates the relationship between student
mobility, segregation and achievement gaps over time within an urban school district using
measures of segregation such as dissimilarity and isolation indices. Together these three papers
provide a detailed overview of the prevalence and impact of student mobility across the timing of
school changes within a school district. This dissertation is one of the first studies to
comprehensively differentiate and classify the various timing and types of student mobility (e.g.,
moves between versus during the year, structural versus non-structural moves) across a range of
grades (K-12) in a large urban district. It is also one of the first studies to use two identification
strategies in estimating mobility effects. Overall, this dissertation provides a better understanding
of student mobility especially the role of the timing of school changes in mobility patterns and
effects. The results illuminate policy implications as well as directions for future research.
The findings of first paper suggest that there are differential mobility patterns in CCSD as
the likelihood and timing of a school change is determined by student characteristics and origin
school quality plays an important role in determining the quality of schools that students transfer
schools even though they do not reside in the attendance zone. Open enrollment is on a first-come, first-serve basis
and transportation is not provided.
5
to. The results from second paper imply that the impact of student mobility varies significantly
with the timing of school changes: switching schools between academic years is typically
associated with low insignificant costs of moving and positive math achievement gains whereas
the opposite is true for changing schools during the school year. The results from the third paper
imply that student mobility (especially during the year mobility) may be related to and influence
disadvantages within a school district such as segregation.
This dissertation takes an important step toward a better understanding of the variations
in the patterns and effects of student mobility across timing of school changes and student
characteristics within an urban school district. This new knowledge can help shape policies that
more effectively target and support the mobile student subgroups in most need. Four critical
scholarly and policy contributions emerge from this dissertation. First, differentiating the timing
of school changes is a necessary starting point for considering the patterns and effects of student
mobility. When all moves are combined, important variation between students, patterns and
effects are left unidentified and critical nuances in the student mobility landscape are ignored.
These variations may have crucial ramifications for the internal and external validity of student
mobility studies. Second, this dissertation highlights the multi-layered heterogeneity of the
mobile student population. Not all mobile students are the same and student mobility effects vary
across student characteristics. There are even further variations within student characteristics by
the timing of school changes. For instance, Black mid-year movers have statistically significant
differences with their counterparts that switched schools in the summer. Differentiating the
timing of school changes helps in disentangling the heterogeneity of the mobile student
population, however, variations across student characteristics compel researchers and
policymakers to not adopt a one-size-fits-all approach to student mobility.
6
Third, student mobility that occurs during the year matters. Ironically, the most
disadvantaged and low-achieving mobile student subgroup is also seemingly overlooked by
researchers and policymakers. This dissertation illustrates that even in the event of positive
changes in school quality, these students experience substantial costs of moving. This
dissertation is one of the first studies to identify and provide insights on the mobility patterns and
effects of students who changed schools both between and during the school year in the same
academic year, or "ultra-movers." These students are indicative of the heterogeneous nature of
student mobility. Fourth, student mobility provides a multi-purpose tool to learn more about and
influence disadvantage in a traditional urban school district. Much of the attention on student
mobility as a policy lever is centered on school choice policies. However, given the relationship
between the timing of school changes and segregation, policymakers and researchers may use
student mobility patterns to provide insights on the underlying mechanisms of desegregation and
achievement gaps within schools and districts.
7
for
Rose-Patricia, Lloyd, Kimberley and Solomon
with much love
8
ACKNOWLEDGEMENTS
My name is the only listed author, however, this dissertation and its completion is forever
indebted to the intellectual, moral and emotional support of a village. First, the journey began
when a charismatic auto mechanic approached an exceptional young teacher in Kingston,
Jamaica circa the early 1980s. My parents have been there from day one. It is fitting to start by
acknowledging Rose-Patricia Boland and Lloyd Welsh, my greatest teachers. I hope I have put
all your hard-work and sacrifices to good use. I became a father to a phenomenal human being
whilst completing this dissertation. Solomon Lloyd Welsh, my firstborn, you have inspired me in
ways that bring tears to my eyes. Sol is the product of a union that has sustained and carried me
through the lowest of lows and the highest of highs. Kimberley, my better half, has always been
there for me and words do little justice to describe her indelible impact. For all my family, too
many to name individually, I am grateful for your understanding and your unrelenting support as
I completed this dissertation.
I could not have imagined a more insightful and supportive dissertation committee:
Dominic J. Brewer, Katharine O. Strunk and Gary D. Painter. I greatly appreciate all the input,
encouragement, and constructive criticism. Most of all, I am thankful for your confidence in my
capacity to develop as a scholar. Dom, my committee chair, deserves more thanks than words
can express. I remember the first time we spoke as if it were yesterday and I will never forget
your unshakeable faith and confidence in my abilities to deliver at the highest-level. I could not
have asked for a more bespoke advisor. Gary provided invaluable perspectives and his feedback
consistently pushed me to higher heights. Katharine’s incisive and illuminating advice elevated
my work to new heights.
9
In addition to my dissertation committee, I must also acknowledge the support at the
Rossier School of Education that was essential to completing this dissertation. Rochelle Hardison
is a dear friend who I consider family. Whether it was ensuring that I got squeezed into Dom’s
hectic schedule or simply lending a listening ear for my frustrations, I will forever treasure our
conversations and your words of guidance played a major role in where I am today. To all the
Rossier folks, past and present, who ensured all deadlines and administrative issues were squared
away, I thank you. In particular, Dianne Morris, Aba Cassell, Laura Romero and Katie Moulton
deserve special thanks. I am also grateful to my peers at USC, especially my cohort. My first
year as a Ph.D. student was memorable because of the people I grew with. They have influenced
the content of this dissertation in some shape or form. I am also especially indebted to the DJB
fraternity – Andrew, Matt, Michelle, Tenice and Tien. I appreciate all the whiteboard sessions
and kind words that helped to make this dissertation a reality.
It would be remiss not to acknowledge my prior colleagues and advisors on the Farm
who guided me to the Rossier School of Education. David Boyer and Pete Klenow are two
notables among the many persons who supported my journey from an undergraduate at Stanford
University to a Ph.D. graduate at the University of Southern California.
Finally, there is a place and a unique set of people that constitute the core of who I am.
Jamaica, Jamaicans and all my friends back home kept me grounded with my eye on the prize.
When I was battered and bruised, I went home for renewal and returned reenergized for the task
at hand. I will never forget where I come from and ‘nuh weh nuh betta than yard.’
My heartfelt appreciation and gratitude are owed to many people, including some not
mentioned by name, who played a part in this monumental accomplishment. Thank you.
10
TABLE OF CONTENTS
Dedication………………………………………………………………………………… 7
Acknowledgements……………………………………………………………………….. 8
Table of Contents…………………………………………………………………………..10
Chapter One: Student Mobility, the Timing of School Changes and School Quality…….11
Chapter Two: Disentangling the Timing and Impact of Student Mobility………………..68
Chapter Three: Student Mobility, Segregation and Achievement Gaps…………………..122
Conclusion………………………………………………………………………………….157
References ………………………………………………………………………………….161
Appendix…………………………………………………………………………………. ..169
11
CHAPTER ONE
STUDENT MOBILITY, THE TIMING OF SCHOOL CHANGES AND SCHOOL QUALITY
Abstract
Although low-income and minority students disproportionately change schools, student mobility
can also be used to facilitate access to higher quality schools. The timing of school changes may
play an important role in student mobility patterns and effects, however, most studies only focus
on student mobility that occurs between school years. This paper compares and contrasts the
patterns of student mobility across the timing of non-structural school changes (when students
change schools on their own accord such as switching elementary schools) in Clark County,
Nevada. The results highlight the importance of differentiating student mobility by the timing of
school changes given the variation in student mobility rates and statistically significant
differences among mobile students across the timing of school changes. The findings also
demonstrate the prevalence of mid-year moves within a school district and draw attention to
discipline-related mobility (school-initiated student mobility resulting in the placement of
students in alternative schools) that largely occur during the school year. The results also indicate
that students in low quality schools are more likely to switch to other low quality schools
whereas students in high quality schools are more likely to transfer to other high quality schools,
regardless of students’ prior achievement or the timing of school changes. This evidence of
differential mobility patterns raises equity concerns regarding access to high quality schools,
especially for students enrolled in low quality schools. Policy implications and areas for future
research are discussed.
12
STUDENT MOBILITY, THE TIMING OF SCHOOL CHANGES AND SCHOOL QUALITY
Student mobility, or the movement of students across schools, is a widespread
phenomenon in the U.S. A 2010 Government Accountability Office study found that only five
percent of students in K-8 did not change schools by the eighth grade (United States Government
Accountability Office, 2010). Student mobility rates are higher in urban school districts relative
to suburban and rural districts (United States General Accounting Office, 1994). Moreover,
frequent student mobility is most common and likely to be damaging among particular
subgroups that tend to concentrate in urban school districts, namely low-income and minority
students (Hanushek et al., 2004; Reynolds et al., 2009; Schwartz et al., 2009; Xu et al., 2009).
Although student mobility is generally associated with a decline in student achievement (Institute
of Medicine and National Research Council, 2010; Mehana & Reynolds, 2004; Reynolds et al.,
2009), some scholars posit that moving to a higher quality school may offset the disruptions
accompanied with changing schools and improve student achievement (de la Torre & Gwynne,
2009; Engberg, Gill, Zamarro, & Zimmer, 2012; Hanushek et al., 2004; Temple & Reynolds,
1999). Thus, student mobility presents both a challenge and a policy opportunity for urban
school districts. On the one hand, low-income and minority students disproportionately change
schools. On the other hand, student mobility can be used to facilitate access to higher quality
schools in order to improve the achievement of these students.
In the extant literature on student mobility in the U.S, relatively few studies focus
exclusively on intra-district mobility (Cullen, Jacob, & Levitt, 2005; Fantuzzo, LeBoeuf, Chen,
Rouse, & Culhane, 2012; Grigg, 2012; Heinlein & Shinn, 2000; Kerbow, 1996; Parke &
Kanyongo, 2012; Schwartz, Stiefel, & Chalico, 2009; Temple & Reynolds, 1999; Welsh, Duque,
13
& McEachin, forthcoming).
3
Most student mobility studies use national data sets (Burkam, Lee,
& Dwyer, 2009; Gasper, DeLuca, & Estacion, 2010, 2012; Pribesh & Downey, 1999; Rumberger
& Larson, 1998) or document intra- and inter-district mobility in a state (Fong, Bae, & Huang,
2010; Hanushek et al., 2004; Mao, Whitsett, & Mellor, 1997; Xu et al., 2009). In addition, intra-
district mobility studies are generally limited to school choice environments (Cullen et al., 2005;
Heinlein & Shinn, 2000; Kerbow, 1996; Schwartz et al., 2009; Temple & Reynolds, 1999; Welsh
et al., forthcoming). Given that patterns of mobility may be dependent on the educational
governance context, there is a need for research that speaks to within-district mobility patterns in
a “traditional” district.
4
The timing of school changes (for e.g., students may switch schools in the summer or
during the school year) may play an important role in the disruption associated with student
mobility and whether students switch to better schools. However, most mobility studies have
been limited to school changes that occur between academic years. It is important to investigate
whether mobility patterns vary systematically by student and school characteristics as well as
whether these patterns vary across the timing of school changes. If within a school district,
advantaged, white and higher-achieving students tend to enroll in or are more likely to switch to
higher quality schools while disadvantaged, minority and lower-achieving students tend to enroll
in or are more likely to switch to lower quality schools, then there exist differential mobility
patterns that raise important equity concerns as they may lead to increasing student segmentation
within a school district by students’ characteristics, achievement or school quality.
3
Students may switch schools within a school district or between school districts. Intra-district mobility occurs
when the origin schools (the school that students are currently enrolled in) and the destination schools (the school
that students transfer to) are in the same school district. Inter-district mobility occurs when the origin and destination
schools are in different school districts.
4
I define ‘traditional’ as a school district with limited open enrollment policies, attendance zones and most schools
operated by a locally elected school board. Although school choice policies have increased in prominence in recent
decades, the majority of school districts in the U.S are ‘traditional.’
14
This paper evaluates the patterns of student mobility across the timing of school changes
in the Clark County School District (CCSD). Using student-level data from 2007-08 through to
2012-13, I compare the exit patterns and examine the destination schools of different groups of
mobile students categorized by the timing of school changes in CCSD. These descriptive
analyses of mobility patterns in CCSD serve as a prelude to causal modelling and provide a
better understanding of the general relationships between the who (characteristics of mobile
students), when (timing of school changes) and where (destination school quality or the quality
of the schools that students switch to) of student mobility.
This paper is one of the first studies to
compare and contrast patterns of intra-district student mobility across the timing of school
changes. Specifically, I ask the following research questions:
a) Does the likelihood of making a non-structural school change differ by students’ and
schools’ characteristics?
b) Do students’ prior achievement and the quality of their current school predict whether
students move to higher quality schools?
c) How do mobility patterns vary with the timing of non-structural school changes?
This paper demonstrates the importance of differentiating student mobility by the timing
of school changes. Non-structural mobility rates vary by the timing of school changes and not all
mobile students are the same. Nine percent of students made non-structural school changes
between school years, five percent switched schools during the school year and three percent
changed schools both between and during the school year in the same academic year. The results
indicate that low-income, minority and low-achieving students in low quality schools are more
15
likely to change schools during the school year whereas high-achieving, advantaged and Asian
students in high quality schools are more likely to switch schools in the summer. More
importantly, regardless of their prior achievement or the timing of school changes, students in
low quality schools are more likely to switch to another low quality school and less likely to
switch to a high quality school, whereas students in high quality schools are more likely to
transfer to high quality schools and less likely to switch to low quality schools. These differential
student mobility patterns in CCSD may lead to increased student segmentation based on student
characteristics and school quality. The rest of the paper proceeds as follows. I first provide a
brief rationale for investigating student mobility patterns across the timing of school changes
within a school district. Next, I describe data and methods before presenting results. I conclude
with a discussion of policy implications and areas for future research.
Intra-District Student Mobility, the Timing of School Changes and School Quality
Intra-district student mobility is important for three main reasons. First, the majority of
student mobility occurs within the same school district as opposed to switching to schools in a
different school district (Hanushek et al., 2004; Kerbow, 1996; Pribesh & Downey, 1999; Xu et
al., 2009). Second, intra-district mobility is generally limited to poor and minority students who
tend to switch schools frequently within an urban school district (Alexander, Entwisle, &
Dauber, 1996; Hanushek et al., 2004; Mao et al., 1997; Xu et al., 2009). Alexander et al. (1996)
found that lower income students transferred within the school district more often while rich,
white students were more likely to move across districts (Alexander et al., 1996). Hanushek et
al. (2004) highlighted that African American and Hispanic students were at least twice as likely
to switch schools within a district than White students and attributed some of the difference to
16
the concentration of minority students in large urban districts (Hanushek et al., 2004). Third,
intra-district student mobility, especially for frequent movers, is typically not linked to
improvements in school quality (Hanushek et al., 2004; Xu et al., 2009).
Student mobility is driven by a confluence of social and economic factors including
residential mobility, family circumstances, or the preference to attend a higher quality school or a
school that better suits a student’s needs (Kerbow, Azcoitia, & Buell, 2003; Kerbow, 1996;
Pribesh & Downey, 1999; Rumberger, Larson, Ream, & Palardy, 1999; Rumberger & Larson,
1998; Rumberger, 2003; Swanson & Schneider, 1999). Although students may change schools
for many different reasons, the majority of student mobility overlaps with residential mobility
(Institute of Medicine and National Research Council, 2010; Reynolds et al., 2009; Rumberger,
2003). Historically, this is largely due to the presence of attendance zones that link school
assignment to a student’s residence. In urban areas and densely populated cities, residential
mobility is even more likely to result in student mobility (Temple & Reynolds, 1999). However,
not all school changes are caused by residential mobility and about 40 percent of student
mobility is due to school-related factors (Kerbow, 1996; Rumberger et al., 1999). Typically
administrative data provide little information about the exact reasons why students change
schools (Grigg, 2012; Hanushek et al., 2004; Institute of Medicine and National Research
Council, 2010; Xu et al., 2009). A substantial proportion of intra-district student mobility is
generally associated with negative reasons such as job loss or family disruption (“reactive”)
rather than transferring to a higher quality or a better fit school (“strategic”) (Alexander et al.,
1996; Hanushek et al., 2004; Rumberger et al., 1999; Xu et al., 2009).
There is no consensus on the causal impact of student mobility on student achievement
(Hanushek et al., 2004; Mehana & Reynolds, 2004). This is partly because the effect of changing
17
schools on academic achievement depends on many factors including but not limited to the type
and timing of the student mobility, the motivating factors driving the decision to change schools
and the quality of schools students transfer to (Institute of Medicine and National Research
Council, 2010; Mehana & Reynolds, 2004; Reynolds et al., 2009). In addition to research design,
a key barrier to credible causal estimates in the mobility literature is distinguishing between the
array of types and timing of student mobility.
5
In order to gain a better understanding of how
non-structural student mobility affects students’ academic outcomes, it is necessary to
extensively document the relationships between the who (mobile students), the when (the timing
of school changes) and the where (the quality of school students transfer to) of student mobility.
The Timing of School Changes
Students can change schools at different points throughout the course of a given school
year including between academic years or within the school year. However, little attention has
been paid to differentiating student mobility by the timing of school changes (Grigg, 2012;
Reynolds et al., 2009; Schwartz et al., 2009). The majority of studies in the extant literature have
examined student mobility that occurs in the summer and most studies on intra-district mobility
classify student mobility based on the number of moves over a given period of time. In the over
150 mobility studies published since 1984, few studies have documented rates of student
mobility that occur during the school year (Alexander et al., 1996; Burkam et al., 2009; Engec,
2006; Fong et al., 2010; Grigg, 2012; Hanushek et al., 2004; Kerbow, 1996; Mao et al., 1997;
Parke & Kanyongo, 2012; Schwartz et al., 2009). Nevertheless, these studies generally found
5
There are two main types of student mobility: 1) structural moves that occur after the completion of a terminal
grade (e.g., elementary to middle school transitions) and 2) non-structural moves that occur when students change
schools of their own volition (e.g., switching elementary schools). Though I include brief comparisons with
structural moves, I am primarily interested in non-structural mobility in this paper. Non-structural movers are the
students that move to a new school on their own accord and the student subgroup that mobility policies in
“traditional” school districts may target and influence.
18
within-year student mobility rates ranging from 7 to 16 percent. African American and low-
income students were considerably more likely to change schools during the school year
(Hanushek et al., 2004; Schwartz et al., 2009). Special education and English Language Learner
students were less likely to switch schools within the academic year (Hanushek et al., 2004; Mao
et al., 1997; Schwartz et al., 2009).
Student mobility during the school year may be more disruptive and affect a more at-risk
and vulnerable population than moves between academic years (Alexander et al., 1996; Burkam
et al., 2009; Grigg, 2012; Hanushek et al., 2004; Schwartz et al., 2009). During the year mobility
may also be influenced by schools’ policies and practices. Under the No Child Left Behind Act
(NCLB), students that change schools during the school year do not count towards schools’
accountability, thus schools have an incentive to avoid accountability for lower-achieving
students through activities that may also increase student mobility (Weckstein, 2003).
6
Additionally, under the NCLB, families in low-performing public schools were given school
choice. Yet, there is limited evidence on whether parents take advantage of the choice provided
when their children attend schools labelled as “in need of improvement” (Howell, 2006). Hence,
examining student mobility during the academic year may also provide useful insights on
whether schools are ‘pushing out’ low-achieving students and whether families are exercising
school choice under NCLB in ‘traditional’ districts.
Lumping together all non-structural movers irrespective of the timing of school changes
is problematic. In particular, combining all non-structural moves may mask important variations
among mobile students as well as mobility patterns and effects. The patterns and impact of
6
Under NCLB, the population of students included in the calculation of a schools’ accountability status are students
who have been continuously enrolled in the same school for a full academic year. In Nevada, a student is considered
continuously enrolled if that student was enrolled in the particular school on or before the fourth Friday of the school
year. In Nevada, the same accountability rules that are applicable to regular schools are also applied to “alternative
schools.”
19
student mobility may depend on the timing of school changes. There may also be statistically
significant differences among mobile students by the timing of school changes. Hence,
differentiating student mobility across the timing of school changes may help address some of
the internal validity concerns of student mobility studies. It is plausible that without such
distinctions, the estimate derived in most mobility studies may be averaging the impact of
changing schools across various timing of school changes which raises questions about the
internal and external validity of the findings (for e.g., estimates of the effect of between-year
mobility may be contaminated by the presence of unidentified students who were also within-
year movers and if movers differ by the timing of school changes, then insights from between-
year movers may not be readily applicable to mid-year movers). Thus, detailed categorization of
the type and timing of school changes is an important step toward crafting more appropriate
empirical strategies to obtain credible causal estimates of the impact of student mobility on
student achievement.
School Quality
Student mobility may increase student achievement if students move from lower quality
to higher quality schools, and any disruptive effects of the move are offset by gains from
improved school quality (de la Torre & Gwynne, 2009; Engberg et al., 2012; Hanushek et al.,
2004; Rumberger et al., 1999; Swanson & Schneider, 1999; Temple & Reynolds, 1999).
However, the majority of the extant literature suggests that most student mobility decisions may
be driven by factors other than school quality (e.g., changes in residence). More importantly,
whether or not a student moves to a higher quality school can be predicted by student
characteristics (Cullen et al., 2005; Hanushek et al., 2004; Rumberger et al., 1999; Schwartz et
al., 2009; Xu et al., 2009). Students who choose to move to a higher quality school are often
20
higher-achieving, less likely to live in poverty, and more likely to be White (Cullen et al., 2005;
Hanushek et al., 2004; Schwartz et al., 2009; Xu et al., 2009).
7
Student mobility often occurs in particular clusters of schools with similar racial, ethnic,
income and achievement characteristics (Kerbow, 1996). Furthermore, the achievement levels of
the schools that students exit are a strong predictor of the schools they transfer to (Kerbow, 1996;
Mao, Whitsett, & Mellor, 1997; Welsh, Duque & McEachin, forthcoming). Cullen et al. (2005)
found scant evidence that students systematically take advantage of open enrollment to attend
higher quality schools. In addition, Xu et al. (2009) found that non-structural moves within
choice districts were not associated with positive changes in math and reading. Overall, there is
little evidence of systematic non-structural student mobility from a lower to a higher quality
school within a school district.
Table 1A in the appendix summarizes 21 peer-reviewed empirical K-12 intra-district
mobility studies published between 1994 and 2014. The majority of intra-district mobility studies
investigate student mobility in school districts with school choice policies, with a number of
studies focusing on Chicago (Cullen et al., 2005; Heinlein & Shinn, 2000; Herbers, Reynolds, &
Chen, 2013; Kerbow, 1996; Ou & Reynolds, 2008; Schwartz et al., 2009; Temple & Reynolds,
1999). Although some studies have examined student mobility prompted by school closures (de
la Torre & Gwynne, 2009; Engberg et al., 2012; Kirshner, Gaertner, & Pozzoboni, 2010), the
predominant focus of intra-district mobility studies has been estimating the effect of mobility on
student achievement (Alexander et al., 1996; Grigg, 2012; Mantzicopoulos & Knutson, 2000;
Parke & Kanyongo, 2012; Scherrer, 2013; D. Wright, 1999). In the extant literature, few studies
7
School quality is typically measured in two ways: either a school’s achievement level or growth. In this paper, I
primarily classify schools based on students’ achievement levels. I provide greater detail on the indicators of school
quality in the methods section.
21
examine the exit and destination patterns of mobile students within a traditional urban school
district.
In this paper, I examine the exit and destination patterns of mobile students by students’
race/ethnicity, prior achievement and school quality across the timing of school changes. A
comprehensive analysis of mobility patterns offers useful insights for estimating the impact of
student mobility and a better understanding of possible school-initiated student mobility during
the academic year. Differential mobility patterns within non-choice school districts may also
imply an increase in student segmentation over time and raise important equity concerns in urban
school districts. In the next section, I describe the data and the empirical approach employed to
investigate mobility patterns in CCSD.
Data and Methods
Data
I use a six year panel of student-level data for all students in the CCSD from 2007-08
through to 2012-13.
8
The data contains students’ demographic characteristics and annual test
scores from the Nevada Proficiency Examination Program. Demographic data includes
indicators for students’ gender, race/ethnicity (Black, Hispanic, Asian, White), free and reduced
priced lunch (FRPL), English Language Learner (ELL) and special education statuses. Students
are tested in reading and math in grades 3-8 and take the High School Proficiency Exam (HSPE)
in grade 10. I standardize test scores for students in grades 3 through 10 by grade and year,
8
As of 2012-13, there are 357 schools in CCSD (217 elementary schools, 59 middle schools, 49 high schools, 24
alternative schools, and 8 special schools). I have data on some of the alternative schools operated by CCSD
including: 5 behavior, 3 continuation schools, 4 juvenile detention centers and 6 adult education schools. I also have
data on all of the 25 CCSD-run magnet schools and career and technical academies that provide parents with school
choice and 9 Prime 6 schools (The Prime 6 Initiative was first adopted in 1994 in an effort to provide support for 9
schools located in West Las Vegas that serve a predominantly low-income African American and Hispanic
population). I only have data for the charter schools in CCSD for the 2012-13 school year.
22
relative to the school mean, as well as relative to the district mean.
9
I complement the student-
level data with school-level data on school locations and facilities including zip code and the
year the school was built in addition to publicly available accountability data.
10
Detailed longitudinal data that tracks the dates and sequence of school changes allows for
in-depth classification of the timing of student mobility across a range of grades (K-12). Unique
student and school identifiers in the data link students to schools in each year and across multiple
school years. I assume that all school changes between school years in grade 6 and 9 are
transitions from elementary to middle and middle to high schools respectively, with the
exception of students enrolled in combination schools, of which there are relatively few.
11
The
reasons for enrollment and withdrawal, as reported by schools, such as whether the student
previously attended a school in another state or the student left a school to transfer to private or
charter schools in the district are also available. Table 2A in the appendix provides details of the
definitions of all the data fields used in this study.
I use two main analytical samples. For the descriptive analysis, I use yearly cross-
sections of all K-12 students. My primary unit of analysis is the student-year yielding 1,946,446
9
I use the annual means and standard deviations on the test scores for each grade and year to standardize student
achievement. Standardizing the outcomes makes it possible to compare students' test scores over time, as well as
across grades and provides insight into how a student’s math and reading test scores are compared to students in the
same grade in the school and in the district. If negative relative to the school mean, this indicates that the student in
the school received a math or reading test score below the school average.
10
I obtained Adequate Yearly Progress (AYP) designation data from 2007-08 through to 2011-12 from the Nevada
Department of Education (ayp.nv.gov) and nevadareportcard.com. Schools are designated by the Nevada
Department of Education as: exemplary, continuing exemplary, exemplary turnaround, high achieving, adequate, on
watch list and in need of improvement. In compliance with NCLB, Nevada AYP classifications are made annually
based on the percentage of students tested, the percentage of students tested who score at or above the proficient
level on annual statewide tests, and school attendance or graduation rates.
11
I exclude students who were transported out of the district (students who live in Clark County but attend schools
in another district) and remove all within-school Americans with Disabilities Act (ADA) changes. There are a
number of students who attended two schools at the same time (typically a high school and a technical school). For
these concurrent enrollments, I only use mobility associated with comprehensive high schools and ignore the other
schools. Xu et al. (2009) and Burkam et al. (2009) highlight the importance of including grade retention variables in
mobility studies. For students who were held back in the same grade in the same school (grade retention) or were
accelerated a grade in the same school, I identify those specific moves as possible controls in the analysis. However,
there was limited grade retention or skipping grades in CCSD over the period of study with less than two percent of
students and the majority of changes occurring in grades 9-12.
23
student-years with 548,523 unique students.
12
For the empirical analysis, I use a sample of
students that have been continuously enrolled in a CCSD school for at least two consecutive
academic years (in other words, students need at least two observations to be included and
students with only one observation were dropped from the sample). This sample includes
1,826,170 student-years with 428,247 unique students.
CCSD is a large, diverse urban school district with average annual enrollment of over
300,000 students. Over the period of study, CCSD experienced ongoing demographic shifts
including an increasing proportion of Hispanic and low-income students and a decreasing
proportion of non-minority students. On average, roughly 42 percent of students are Hispanic, 33
percent are White, 13 percent are African American and 8 percent are Asian. The proportion of
African American students slightly decreased over the period of study from 14 to 13 percent
while the proportion of Hispanic students increased over time from 40 percent in 2007-08 to 43
percent in 2012-13. There was a marked decline in the proportion of White students from 36
percent in 2007-08 to 29 percent in 2012-13. Similarly, the proportion of Asian students also
decreased but to a lesser extent. The proportion of special education status students rose from 10
to 12 percent while the proportion of ELL students mildly fluctuated but declined overall from
20 percent in 2007-08 to 16 percent in 2012-13. The proportion of FRPL students increased over
time from 44 percent in 2007-08 to 50 percent in 2012-13.
Methods
The empirical analysis consists of two parts: an exit analysis and a destination analysis.
First, I examine the probability of a student making a non-structural move based on the
characteristics of the student and the school they exit. Second, I investigate the role of the quality
12
I include students in untested grades (K-2, 9, 11, 12) that do not have achievement data. If this sample is further
restricted to students with achievement data, there are 985,805 student-years with 205,123 unique students.
24
of origin schools and student characteristics in predicting the quality of the destination schools of
non-structural movers across the different timing of school changes. Similar to previous mobility
studies, I focus on and present results for mathematics achievement as math is predominantly
learned in school rather than the home (especially starting in the elementary years) and mobility
effects may be more detectable using math as opposed to reading (Hanushek et al., 2004;
Raudenbush, Jean, & Art, 2011; Rumberger et al., 1999; Xu et al., 2009). I categorize non-
structural movers by the timing of school changes: between-year switcher or a student who made
a non-structural move between school years; within-year switcher or a student who switched
schools at least once during the school year and; “ultra-mover” or a student who changed schools
both between and during the school year in the same academic year.
Exit Analysis. In order to examine the relationship between exiting patterns and student
and school characteristics, I use the following linear probability model:
𝑌 is 𝑡 = 𝛽 0
+ 𝛽 1
Achieve
ist-1
+ 𝛽 2
SchQuality
st
+ 𝛽 3
SchQuality
st
*Achieve
ist-1
+ 𝛽 4
X
is 𝑡
+ 𝛽 5
Z
s 𝑡
+ π
t
+ γ
g
+ 𝑒 i 𝑡
(1)
where 𝑌 ist
is a dichotomous outcome variable that is equal to one if student i in school s at
time t made a non-structural school change. I estimate the probability separately for the
aforementioned three categories of mobile students as well as all non-structural moves
combined. Achieve
ist-1
is lagged student achievement (relative to the school) and controls for the
academic history of students before enrolling at a new school. This variable assesses whether
students’ prior academic performance predicts future mobility, specifically the marginal effect of
a one unit (a standard deviation) increase in students’ z score on the probability of changing
25
schools. SchQuality
st
is measured by the percentage of students in a school scoring proficient or
above on state accountability tests.
13
This captures the marginal effect of school quality in year t
on the probability of making a non-structural school change in that year.
14
𝛽 3
indicates whether
the coefficient on student achievement varies by the quality of school they exited. In particular,
𝛽 3
is the impact of a one percent increase in the percentage of proficient students in a school on
the probability of changing schools for a mean student in the school. X
ist
is a vector of student-
level characteristics including gender, the aforementioned racial categories, FRPL, ELL and
special education statuses. Z
s 𝑡 is a vector of school-level characteristics including the percentage
of Black, Hispanic, Asian, FRPL, ELL, Special Education, and male students in the school. In all
models I utilize grade (γ
g
) and year (π
t
) fixed effects to control for unobservable differences
across time and between grades and use robust standard errors clustered at the school level.
Finally, I include school fixed-effects to remove any time-invariant differences among schools.
The school fixed-effect specification examines whether students from different points across the
achievement distribution within a given school are more or less likely to make a non-structural
move.
13
In the typical accountability system, schools are predominantly evaluated on the achievement level of their
students (Polikoff, McEachin, Wrabel & Duque, 2014). Even though parents may use information from multiple
sources when choosing schools, school accountability reports generated by state education agencies remain the only
formal source of information on school quality available to parents (Garcia, 2011). The school quality information
available to parents in Nevada is likely to originate from schools’ accountability reports that are accessible online.
These reports have two measures of school quality: the proportion of students in each accountability category and
the Nevada School Performance Framework (NSPF) ratings that started in 2011-12. Given that measures of
achievement levels are more readily available to parents and students, it is reasonable to assume, as evidenced by the
school capitalization literature (Black & Machin, 2010), that parents readily respond to schools’ achievement levels
and use these measures when making decisions about what school to attend rather than schools’ value-add. In this
paper, I model student mobility as closely as possible to how families choose a school. Hence, I focus on schools’
achievement levels and not growth, as families likely do not have access to the latter. I use the percentage of
students in a school classified as proficient or higher as the indicator of school quality.
14
For between-year non-structural moves, I use SchQuality
st-1, hence the quality of the schools that students exited
may be correlated with their prior achievement. However, the correlation between school quality in the previous
year and prior achievement is small (0.05) thus I interpret the school quality coefficient in a similar fashion to the
other timing of school changes.
26
Destination Analysis. In order to evaluate whether students are switching to higher
quality schools, I utilize a multinomial framework as follows to predict the quality of the
destination school for non-structural movers:
𝑃𝑟 𝑜𝑏 ( 𝐷𝑒𝑠𝑡 𝑖𝑛𝑎𝑡 𝑖𝑜𝑛 𝑖𝑡
= 𝑗 )
=
exp ( β 0 + β 1 𝐴𝑐 ℎ 𝑖𝑒 𝑣 𝑒 𝑖 ( 𝑡 − 1 ) 𝑗 + β 2 𝑂 𝑟 𝑖𝑔𝑖 𝑛 𝑄 𝑢 𝑎 𝑙 𝑖𝑙 𝑡 𝑦 𝑖 ( 𝑡 − 1 ) 𝑗 + β 3𝐴𝑐 ℎ 𝑖𝑒 𝑣 𝑒 ∗ 𝑂 𝑟 𝑖𝑔𝑖 𝑛 𝑄 𝑢 𝑎 𝑙 𝑖𝑡𝑦 𝑖 ( 𝑡 − 1 ) 𝑗 + β 4 X
𝑖𝑡𝑗 )
∑ ( β 0 + β 1 𝐴𝑐 ℎ 𝑖𝑒 𝑣 𝑒 𝑖 ( 𝑡 − 1 ) 𝑗 + β 2 𝑂 𝑟 𝑖𝑔𝑖 𝑛 𝑄 𝑢 𝑎 𝑙 𝑖𝑡𝑦 𝑖 ( 𝑡 − 1 ) 𝑗 + β 3𝐴𝑐 ℎ 𝑖𝑒 𝑣 𝑒 ∗ 𝑂 𝑟 𝑖𝑔𝑖 𝑛 𝑄 𝑢 𝑎 𝑙 𝑖𝑡𝑦 𝑖 ( 𝑡 − 1 ) 𝑗 + β 4 X
𝑖𝑡𝑗 )
𝐽 𝑗 = 1
Where Destination
it
represents school quality and is the categorical outcome variable for
a student i in time t, with j=1 for schools in the bottom third of the district achievement
distribution in the prior year, =2 for schools in the middle third of the distribution (base
category), and =3 for schools in the top third of achievement distribution. Achieve indicates
whether the student was in the bottom, middle, or top third of the district’s achievement
distribution for their grade in the previous year, OriginQuality represents the prior year school
quality of the non-structural movers’ previous school, calculated in the same manner as the
Destination variable, and Achieve*OriginQuality is an interaction between the student’s prior
year achievement and the origin school’s prior school quality. I also include the full set of
student level controls, X, and an indicator for year.
Results
I present a descriptive overview of student mobility in CCSD before the results of the
empirical analysis. Table 1 shows the mobility rates by the type and timing of school changes.
Over the period of study, roughly 31 percent of students changed schools each year, regardless of
the type or timing of school change. There was little year-to-year variation in total student
27
mobility rates. Total student mobility increased slightly from 32 percent in 2008-09 to 33 percent
in 2009-10 before declining to 30 percent in 2012-13. However, mobility rates vary by the type
of school change. Approximately 14 percent of students made structural moves and 17 percent of
students made non-structural school changes each year.
15
Non-structural mobility rates also vary
by the timing of school changes. On average, about nine percent of students made non-structural
moves between school years, five percent of students changed schools at least once during the
school year and roughly three percent of students changed schools both between and during the
school year in the same academic year (“ultra-movers”).
16
[Insert Table 1 around here]
Table 2 illustrates that non-structural mobility rates vary by students’ income, ethnicity,
educational status and gender. Low-income, Black, Hispanic, special education, ELL and male
students had above average mobility rates. For instance, roughly 27 percent of Black students
made a non-structural move annually, more than twice the rate of White and Asian students.
Hispanic students were the second most mobile racial/ethnic category with about 18 percent of
students making a non-structural school change annually. Approximately 22 percent of FRPL
students made a non-structural move each year relative to 13 percent of students who were not
free and reduced price lunch recipients. The black/white gap in mobility rates (14 percent) is
15
In addition, of the 548,523 unique students, 22 percent made one non-structural move, 8 percent made two moves
and less than 5 percent of students changed schools three or more times between 2007-08 and 2012-13.
16
On average, roughly seven percent of students left the district within the academic year and roughly three percent
of students transferred into the CCSD annually during the school year. Exiting in high school, about 10 percent on
average, was twice as high as exiting during elementary and middle schools (grades K-8), when roughly five percent
of students left the district within the academic year. Most of the leavers from CCSD also exited the state. Nearly
half of all the students that exited the district during the school year transferred to a school in another state. Only two
percent of leavers remained within-state and transferred to school in Nevada outside the CCSD. About seven percent
switched to a school in another country and less than five percent switched to a private school in the district. About
six percent of students had an incomplete record of transfer (for e.g., a record of withdrawing from a school but no
destination school record). In addition, students who entered and exited the district during the school year were also
below average in their schools with entrants having higher achievement levels than within-year leavers.
28
larger than the income gap (9 percent). ELL and special education students also had high
mobility rates with about a fifth of each subgroup making non-structural school changes annually
over the six year span. There was also a slight gender gap in mobility rates with males having a
higher rate of mobility than females.
[Insert Table 2 around here]
Table 2 also demonstrates further variation in non-structural mobility rates by students’
income, ethnicity, educational status and gender across the timing of school changes. Overall,
African American, low-income, Hispanic, male, special education and ELL status students
switched schools at substantially higher rates within the academic year than Asian, White and
non-FRPL students. When one considers all mobility that occurred during the school year (mid-
year and ultra-movers), the majority of non-structural moves for these students occur mid-year as
opposed to between school years. The opposite is true for advantaged, non-minority and female
students when tend to change schools predominantly between school years. For instance, roughly
12 percent of low-income students switched schools during the school year relative to about 6
percent of non-FRPL students. Similarly, about 15 percent of Black students changed schools
during the year compared to about 5 percent for White and Asian students.
Table 3 illustrates that not all mobile students are the same. There were statistically
significant differences among movers by type and timing of student mobility. Non-structural
movers had statistically significant differences from students who made structural moves across
all demographic and achievement characteristics. On average a higher proportion of low-income,
male, special education, ELL and minority students made non-structural moves relative to
structural moves. Non-structural movers also had below average achievement relative to their
schools and the district whereas structural movers had above average achievement levels.
29
[Insert Table 3 around here]
Non-structural movers differ by the timing of school changes. Students who switched
schools between school years differed from students who switched schools during the school
year across all observable characteristics. On average, students who changed schools during the
school year were a higher proportion of African American, Hispanic, male, ELL, special
education and FRPL students than their counterparts who made non-structural moves between
school years. Conversely, lower percentages of Asian and White students switched schools
during the year compared to between-year school changes.
Moreover, within-year switchers had
lower achievement levels relative to their school and district than students who changed schools
between academic years (-0.36SD vs. -0.12SD in math relative to school).
17
There is another group of non-structural movers that has been largely overlooked in the
mobility literature: students who change schools both between and during the school year in the
same academic year. I have termed these students “ultra-movers”. These students appear to be a
subset of more disadvantaged within-year movers. However, “ultra-movers” differed from
within-year non-structural movers across all observable characteristics except gender and math
proficiency. On average, “ultra-movers” had a higher proportion of African American and low-
income students relative to within-year movers. These students also had lower achievement
levels relative to their schools and the district compared to within year movers. In fact, “ultra-
movers” were the lowest achieving group of mobile students (-39SD in math relative to school).
Reasons for Student Mobility
17
Students who switched schools in the fall semester (August to December) had statistically significant differences
with students who changed schools in the Spring semester (December to June) across all achievement and
demographic characteristics except special education status, math achievement relative to school, and whether
students were performing below average in math in their schools. On average, a greater proportion of African
American, White, Male and FRPL students switched schools during the spring semester, whereas a larger
percentage of Hispanic, Asian, ELL and special education students changed schools in the fall semester.
30
Students may change schools at different times for varying reasons, however, data on the
motivating factors for student mobility is rarely available.
Notwithstanding, I create three
indicators of the reasons for changing schools including: residential change, AYP-designation
related move and school-initiated school changes due to disciplinary reasons.
18
Regardless of the
timing of non-structural student mobility, 68 percent of school changes were to a school in a
different zip code. This is consistent with prior findings in the student mobility literature that
roughly two-thirds of all moves are due to residential changes. However, the proportion of
students moving to a school in a new zip code varies by the timing of student mobility: 83
percent of students who changed schools between academic years switched to a school in a
different zip code compared to 48 percent for mid-year movers. This implies that most school
changes during the summer involve residential changes and possibly greater distance between
the origin and destination schools whereas only about half of during the year mobility involves
residential changes and mid-year moves tend to occur within a cluster of nearby schools (schools
within the same zip code) that do not require moving large distances.
There are also variations in transfer rates by schools’ accountability status and the timing
of school changes. About 53 percent of school changes between school years were from schools
classified as “in need of improvement” compared to 60 percent for mid-year moves and 66
percent for “ultra-moves.” The results suggest that students in higher quality schools may be
more likely to transfer between school years whereas school changes during the school year may
be concentrated in lower quality schools, especially for ultra-movers. The results imply that
18
In the absence of students’ addresses or zip codes, I use schools’ zip codes to identify student mobility that is
concurrent with changes in residences. I classify moves to a school in a different zip code as a proxy for student
mobility caused by residential mobility. Another reason for changing schools may be AYP-related school choice.
Under the NCLB, Title I schools that do not make AYP for two consecutive years are classified as “In Need of
Improvement” and required to provide school choice options. Students can transfer to a school making AYP, thus I
identify students who transfer from a school designated as “In Need of Improvement.” Discipline-related mobility is
classified as all school changes to and from behavior or continuation schools or juvenile detention centers. The data
on school discipline is as reported by the schools.
31
families may be utilizing school choice under NCLB to exit low-achieving schools during the
school year.
School discipline. School discipline provides an important yet overlooked example of
school policies and practices that may induce student mobility. When a student in CCSD
commits “any act defined as dangerous or antisocial behavior,” there are three main possible
outcomes: suspension of student (if more than 10 days, Board approval is required); placement in
behavior or continuation schools and; referral to the Clark County Juvenile Services.
19
School
officials have considerable leeway in their interpretation of the code of conduct to discipline
students. Placing students in behavior or continuation schools is largely left up to the discretion
of school personnel (the principal) and is listed as an additional consequence for all offenses in
the code of conduct. Even though school administrators must submit recommendations for
expulsion to the Education Services Division (ESD), the gross majority of recommendations are
approved by the ESD: in the 2009-10 school year, less than one percent of recommendations for
expulsion (22 of 4,544) were rejected by the ESD (Gibson Consulting Group, 2011).
The results suggest that discipline-related mobility accounts for a non-trivial amount of
student mobility that occurs within a school district. Over the period of study, about 6000
students were placed in behavior and continuation schools or juvenile correctional facilities
annually. Overall, six percent of all student mobility (structural and non-structural) and about
19
CCSD offers two main alternative education programs for students with disciplinary problems: five behavior
schools (grades 6-12) and three continuation schools (two serve grades 9-12; the other serves grades 6-8). Students
in behavior schools are either recommended for expulsion or placed directly in the behavior school by their
principals without a recommendation for expulsion.
Behavior schools are intended to be brief intervention programs
lasting about four to nine weeks for students who committed disciplinary infractions at regular schools. Students are
expected to return to a regular school following completion of the behavior school program that consists of required
academic courses plus assistance and support to improve social and behavioral skills. Continuation schools are
limited to students expelled from their home school via a Board of Trustee ratified expulsion. After successfully
completing continuation schools, students with “limited” expulsions may return to a regular school under trial
enrollment conditions. In addition, students who commit infractions that not only violate school rules but also
federal, state and local laws are typically referred to Juvenile Detention Centers in CCSD.
32
nine percent of all non-structural school changes were discipline-related. More importantly, the
results suggest that these school-initiated school changes largely occur during the school year.
Roughly two percent of school changes that occurred between school years were discipline-
related. Conversely, 14 percent of mid-year moves and 20 percent of “ultra-moves” were
discipline-related. In addition, multiple school changes during the school year were largely
discipline-related. For instance, of the students who changed schools three times during the
school year, 34 percent made discipline-related school changes as opposed to less than 5 percent
for students who made one mid-year move. The majority of disciple-related mobility (about
three-fourths) was due to placement in behavior schools as opposed to placement in continuation
schools or referral to juvenile services. The overwhelming majority of discipline-related mobility
occurred in grades 8 through 10 and was highest in grades 8 and 9.
In addition, there were a few differences in the timing of school changes between the
disciplinary options. The majority of mobility to and from behavior schools occurred in grades 7
through 10 and was highest in grade 8. More than 95 percent of placement in behavior schools
occurred during the year and about a third of these students were ultra-movers. The majority of
continuation school placement occurred in grades 8 through 10 and was highest in grade 9.
About four-fifth of these placements occurred during the year and nearly half of these students
were ultra-movers. Students were generally sent to correctional facilities in high schools with
grade 9 having the highest rate. However, unlike placement in behavior and continuation
schools, nearly half of mobility to and from correctional facilities occurred between school years.
The results imply that mobile students affected by discipline-related school changes are
disproportionately Black, Hispanic, male, low income, low-achieving and receiving special
education services. On average 28 percent of discipline-related movers were Black (relative to a
33
district average of 13 percent), 46 percent were Hispanic (relative to a district average of 41
percent), 74 percent were male (relative to a district average of 51 percent), 16 percent were
receiving special education services (relative to a district average of 11 percent), and 71 percent
were FRPL recipients (relative to a district average of 49 percent). These students were also the
lowest achieving students in CCSD – only 22 percent were proficient in math and 85 percent
were below the district’s average in math. Discipline-related mobile students’ average
achievement was more than half of a standard deviation below their school and the district
average.
There were statistically significant differences between discipline-related and non-
structural movers. When all non-structural moves are lumped together, discipline-related movers
were, on average, a higher proportion of Black, male, special education and low-income students
with much lower achievement. Discipline-related mobile students differed from non-structural
movers across all demographic and achievement characteristics. The differences between non-
structural movers and discipline-related movers remained statistically significant when the
timing of school changes were differentiated. Moreover, there were interesting statistically
significant differences among students placed in behavior and continuation schools as well as
students placed in juvenile detention. For instance, continuation students and students placed in
juvenile detention differed across most demographic characteristics but not achievement,
behavior school students differed from students placed in juvenile detention, and behavior school
students differed from continuation school students . In sum, discipline-related mobility is an
important type of student mobility because it occurs mainly during the school year and may be
largely attributed to schools’ personnel and practices. The results imply discipline-related movers
differ from other types of movers regardless of the timing of school changes but there may also
34
be interesting heterogeneity within the discipline-related mobile student population worthy of
further empirical study.
Student Mobility from the Schools’ Perspective.
In order to better understand the variation in non-structural mobility across the timing of
school changes in CCSD, I examine student mobility at the school-level. Specifically, I focus on
the percent of students leaving each school or the average school turnover. Without
distinguishing the timing of school changes, the average non-structural exit rates of schools in
CCSD over the period of study was roughly 21 percent. However, there was a notable variation
in non-structural exit rates among schools. Non-structural exit rates vary by schools’ income,
racial/ethnicity and achievement characteristics. The results indicate that schools with higher
mobility rates also have a higher proportion of low-income and minority students than schools
with lower mobility rates. There is also a negative relationship between non-structural exit rates
and school quality: schools with higher mobility rates typically have progressively lower school
quality.
Moreover, the results indicate further considerable variations in non-structural exit rates
by school characteristics across the timing of school changes. Table 4 summarizes the average
annual between-year and within -year exit rates by school characteristics. Non-structural exit
rates also vary by level of schooling across the timing of school changes. In elementary and high
schools, the between-year rate was higher than the within-year exit rate (especially for high
schools where the between-year rate was more than twice that of the within-year rate).The
within-year exit rate in middle schools was slightly higher than mid-year exit rates in high
schools. Interestingly, in middle schools, the between-year and within-year exit rates were
almost equal. This suggests that mid-year moves are especially relevant in middle schools.
35
[Insert Table 4 around here]
Disciplinary schools including behavior and continuation schools as well as schools in
the Clark County Juvenile Justice System had some of the highest non-structural mobility rates
that were largely driven by during the year mobility. In these schools, roughly 32 percent of
students exited during the school year relative to 13 percent between school years. This within-
year churn stands in stark contrast to magnet schools that had 9 percent of students exiting
between school years and only 3 percent exiting during the school year. The substantially higher
within year exit rate of disciplinary schools confirms that discipline-related mobility largely
occurs during the school year as opposed to between school years. On average, disciplinary
schools also had lower quality than regular schools.
In addition, I also examined discipline-related mobility by school characteristics. The
average discipline-related exit rate in CCSD was roughly two percent. The discipline-related
mobility rate was slightly higher in middle schools (1.93 percent) than high schools (1.87
percent). Magnet schools had average discipline-related exit rate ranging from zero to four
percent with an average over the period of study of just over one percent. Charter schools had a
much lower rate – less than half of a percent (only one year of data). Combination schools had
the highest discipline-related mobility with 28 percent of students exiting due to school
discipline. Finally, special education schools had an above average rate of roughly four percent.
School quality is negatively correlated with discipline-related mobility (-0.3). Lower
quality schools typically have higher discipline-related mobility rates. The results indicate that
schools in the bottom third of the district’s achievement distribution have much higher
discipline-related mobility rates (4 percent) than average (0.007) and high achieving school
(0.004). Furthermore, the demographic characteristics most correlated with schools’ discipline-
36
related mobility rate are the percent of black students (0.23) and percent of male students (0.58)
in a school. Schools’ discipline-related mobility rate was also strongly correlated with mid-year
exit rate (0.56 for mid-year and 0.76 for ultra-mover).
There are also interesting differences in non-structural exit rates by schools’ income and
demographic composition across the timing of school changes. Table 4 also shows that, on
average, schools with a low proportion of FRPL students had higher between-year exit rates and
lower within-year rates. Moreover, the difference between between-year and within-year exit
rates decreased as the concentration of FRPL students in a school increased. The results imply
that within-year mobility is a relatively larger part of total non-structural student mobility for
schools with a large proportion of low-income students - within-year exit rates are comparable to
between-year exit rates for these schools. Similarly, the results indicate that the within-year exit
rates of schools increased with the proportion of minority students and the between-year and
within-year exit rates were comparable for schools with a high proportion of minority students.
The same is also true for school quality: within-year exit rates increased as the percent of
proficient students in a school decreased and schools labelled as “in need of improvement” had
the highest within-year exit rate and schools classified as “high-achieving and above” had the
lowest within-year exit rate.
Overall, the results imply that mid-year mobility largely affects lower quality schools
with large proportions of low-income and minority students. It is worthy to note the considerable
prevalence of during the year mobility in the lowest achieving schools in the district (the bottom
two deciles in the district, schools with less than 45 percent of proficient students). These schools
had within-year exit rates that were higher than between-year exit rates and the average within-
37
year exit rate of schools in the bottom decile of the district’s achievement distribution was more
than three times that of schools in the top decile.
Table 5 shows the exit school quality by destination school quality across the timing of
school changes. Overall, the results indicate that the majority of school changes are to schools of
equivalent quality. Students exiting higher quality schools (schools with greater than 55 percent
of proficient students) generally transferred to a higher quality school between school years
relative to during the academic year moves. However, students in schools with less than 55
percent of proficient students (bottom 4 deciles) transferred to higher quality schools during the
year relative to school changes between school years. For instance, students in schools in the
bottom decile that switched during the year transferred to schools almost in the second decile
(1.91). This provides suggestive evidence that students in some of the lowest-achieving schools
who switched schools during the year may be “inching up.” These students are transferring to
schools of higher quality relative to their previous schools even though their new school is still
low-achieving within the district.
[Insert Table 5 around here]
Table 6 demonstrates the exit school by destination school demographic composition.
With the exception of students in schools with greater than 90 percent non-white students, mid-
year movers switched to schools with a greater proportion of minority students relative to school
changes between academic years. Overall, the results imply that students in low quality schools
with a high percentage of minority students may be moving to marginally higher quality schools
with similar demographic composition during the school year. Students in higher quality schools
with low concentrations of minority students transfer to better schools in the summer. Students in
these schools that moved during the year typically switched to lower quality schools with higher
38
concentrations of minority students. In the following sections, I present and discuss results from
the empirical analysis examining the likelihood of exiting schools and the quality of schools that
students transfer to across the timing of school changes.
[Insert Table 6 around here]
Exit Analysis
Table 7 illustrates the linear probability of a non-structural move by the timing of school
changes. Column (1) presents linear probabilities accounting only for students’ characteristics.
Column (2) accounts for students’ and schools’ characteristics. Column (3) includes both
students’ and schools’ characteristics as well as an interaction between school quality and
student achievement. Column (4), the preferred specification, includes all the aforementioned
controls and school fixed effects.
[Insert Table 7 around here]
When all non-structural moves were lumped together, a one standard deviation increase
in prior achievement was associated with a six percentage points decrease in the probability of
changing schools. This implies that higher achieving students in schools are less likely to make a
non-structural move. Black, male and FRPL students were more likely to switch schools whereas
Asian, special education and ELL students were less likely to make a non-structural move. The
results also indicate that students in higher quality schools were less likely to switch schools. A
10 percent increase in the percentage of proficient students in a school was associated with a
roughly ten percentage points decrease in the probability of changing schools. The interaction
between student achievement and school quality is significant indicating that the role of prior
achievement varies by the quality of schools students exit.
39
A more complex narrative of exit patterns emerges when student mobility is
differentiated by the timing of school changes. When one considers between year non-structural
moves, a one unit increase in a student’s z score was associated with a roughly two percentage
points increase in the likelihood of changing schools. In other words, higher achieving students
are more likely to change schools in the summer. Black, Hispanic, male, ELL and low-income
students were less likely to switch schools between school years whereas Asian students were
more likely to change schools between academic years. The results indicate that a 10 percent
increase in the percentage of proficient students in a school was associated with a roughly five
percentage points increase in the probability of changing schools between school years. Stated
differently, students in higher quality schools are more likely to change schools in the summer.
The interaction between prior achievement and school quality was insignificant across all
specifications. This suggests that the effect of prior achievement on the likelihood of changing
schools between school years does not vary with the quality of schools that mobile students exit.
For mid-year movers, the results suggest that lower achieving students were more likely
to change school during the school year. However, prior achievement plays a lesser role in mid-
year moves when school characteristics are accounted for.
20
Black, Hispanic, male, ELL and
low-income students were more likely to change schools during the academic year. The results
also suggest that students in higher quality schools were less likely to change schools during the
school year. A 10 percent increase in school proficiency was associated with about a six
20
As Column (1) shows, when only students’ characteristics are accounted for, a one unit increase in a student’s z
scores was associated with a two percentage points decrease in the likelihood of changing schools during the school
year. As Columns 3 and 4 show, the coefficients on prior student achievement for within-year moves were
dramatically smaller and lost statistical significance upon the inclusion of student/school interactions and school
fixed effects respectively. The results suggest that, among the population of students within a district, the higher
achieving students are more likely to switch schools between school years whereas lower achieving students are
more likely to change schools during the school year. However, among the population of students within individual
schools, higher achieving students are more likely to change schools in the summer but are not more or less likely
than lower-achieving students to switch schools during the academic year.
40
percentage points decrease in the probability of changing schools during the school year. The
interaction between student achievement and school quality was a significant predictor of within
year non-structural moves. In other words, the effect of prior achievement on the likelihood of
changing schools during the school year varies with the quality of schools mobile students exit.
Ultra-movers are an interesting group of mid-year movers that illustrate the profound
heterogeneity of the mobile student population within an urban district. The results indicate that
a one unit increase in student achievement was associated with a two percentage points increase
in the likelihood of changing both between and during the academic year. This implies that
higher achieving students are more likely to be ultra-movers. Black and low-income students
were more likely to be ultra-movers whereas Asian and Hispanic were less likely to be ultra-
movers. The results also suggest that students in higher quality schools are less likely to be ultra-
movers. Thus, ultra-movers appear to be largely higher achieving students in low quality schools.
It is plausible that ultra-movers may reflect high achieving low-income and minority students
from highly motivated families with strong school quality preferences seeking to escape lower
quality schools. The interaction between achievement and school quality is significant indicating
that the effect of prior achievement on the likelihood of making ultra-moves varies with the
origin school quality of mobile students.
In order to examine whether exit patterns are driven in part by school discipline, I re-run
the analysis using a sample that excludes discipline-related mobility. There are a few noteworthy
differences in the results. When discipline-related school changes are excluded, male students
were no longer more likely to change schools during the school year. This suggests that school
discipline plays a significant role in mid-year mobility for male students and partly contribute to
differential exit patterns, especially the gender gap in during the year school changes. In
41
addition, I predict the probability of a student making a discipline-related move based on both
student and school characteristics. The results indicate that higher achieving students were less
likely to be discipline-related mobile students. A one standard deviation increase in prior
achievement was associated with a one percentage point decrease in the likelihood of changing
schools due to school discipline. Black, male and low-income students were more likely and
Asian students were less likely to be discipline-related movers. The results imply that time-
invariant school characteristics may partly influence the likelihood of a discipline-related
mobility as the coefficients on school characteristics lost significance upon the inclusion of
school fixed effects. The interaction between student achievement and school quality was
positive and significant. This suggests that as school quality increases, mean students in a school
are more likely to be discipline-related movers.
In sum, the results suggest that non-minority, advantaged students with higher prior
achievement in higher quality schools tend to change schools in the summer whereas lower
achieving low-income, minority and male students in lower quality schools typically switch
schools during the academic year. Students receiving special education services were more likely
to switch schools in the summer and less likely to change schools during the year,
21
whereas the
opposite is true for ELL students. Lower-achieving students may change schools during the
school year for several possible reasons including family circumstances, the search for better
schools or school-initiated mobility such as “pushing out.”
Interestingly, the results imply that there may be variations in the role of school
characteristics in exiting patterns by the timing of school changes. Time-varying unobserved
21
When school fixed effects are included the coefficients become insignificant. This may be because special
education students are non-randomly distributed between schools and these students are more likely to cluster. Thus
there may be some unobserved school-level selection and there is evidence to suggest this may be the case. There
are ten schools in CCSD with over 50 percent of their student population receiving special education services and
special education students appear to move within this cluster of schools.
42
school characteristics (for e.g., variations in school-level policies or changes in the principal or
composition of teachers) appear to play a considerable role in between year moves whereas time-
invariant unobserved school characteristics (for e.g., reputation and school discipline practices)
may play an influential role in during the year mobility.
22
More importantly, the interaction of
student achievement and school quality is not significant for between year school changes but is
significant for mid-year and ultra-moves. This suggests the effect of achievement on the
likelihood of changing schools between school years does not vary with school quality but the
opposite is true for mid-year and ultra-moves. This suggests that during the year mobility is
determined more by the interaction of student and school characteristics than solely prior
achievement.
23
Overall, the results in the exit analysis demonstrate that lumping all non-structural moves
masks the crucial differences in mobility patterns across the timing of school changes. However,
examining exit patterns only provides a part of the story of changing schools. The evaluation of
destination patterns in the next section further illuminates the relationships between mobility
patterns, the timing of school changes and school quality.
Destination Analysis
22
When school fixed effects were included, the coefficient on school quality varies for between and during the year
moves. For mid-year and ultra-movers, the coefficient on school quality becomes insignificant when school fixed
effects are included suggesting that time-invariant school characteristics explain a significant portion of the effect of
school attributes on the likelihood of changing schools. Conversely, the coefficient for between school years
increased and gained statistical significance upon the inclusion of school fixed effects.
This implies that for between-
year nonstructural moves, changes in school quality still explain the probability of changing schools even after
accounting for time-invariant unobserved differences between schools. Although there is a fair amount of variation
in school quality between schools, there is little variation in quality within schools. Only 22 of 2,077 school-years
had no yearly school quality change, however, the average change was less than one percent. Thus most schools had
fluctuations in school quality from year to year but these changes were small.
23
The interaction for mid-year and ultra-movers is negative. When the sum of the coefficients and the interaction
term is estimated, the results imply that between-year movers are more likely to be higher achieving students in high
quality schools; mid-year movers are more likely to be higher achieving students in low quality schools or lower
achieving students in high quality schools; and ultra-movers are more likely to be higher achieving students in low
quality schools.
43
Table 8 presents the change in log odds of moving to a low and high achieving school
relative to moving to a school in the middle third of the district’s achievement distribution for all
non-structural moves combined. The results imply that origin school quality plays a significant
role in the quality of schools that students transfer to. For instance, students in low achieving
schools were more likely to transfer to another low achieving school and less likely to transfer to
a high achieving school. Moreover, regardless of prior achievement, students in high achieving
schools were more likely to switch to other high achieving schools.
[Insert Table 8 around here]
Student characteristics also predict the quality of schools students transfer to. Black and
Hispanic students were more likely to switch to low achieving schools and less likely to transfer
to high achieving schools. Asian students were less likely to switch to low achieving schools.
Male students were more likely to switch to low achieving schools and low income students
were less likely to transfer to high achieving schools. Overall, the results imply there is sorting
based on both student and school characteristics. Low-income, minority students typically
change to low achieving rather than high achieving schools and high achieving students
generally switch to high rather than low achieving schools. Students in high achieving schools,
regardless of whether they are low achieving or high achieving, tend to switch to other high
achieving schools.
Because it is possible to still have a multiplicative interaction with non-linear regression
models (Buis, 2010), I calculate the marginal effects of these interactions to investigate the
potential for a multiplicative interaction between student and school achievement and student
mobility. The marginal effects are the differences in predicted probabilities of attending a given
quality school between students in the bottom and top third of the achievement distribution
44
relative to students in the middle third, by school of origin. Table 9 presents the marginal effects
for all non-structural moves combined. The results confirm that there are differential mobility
patterns by origin school quality regardless of the timing of student mobility or students’ prior
achievement. For instance, students in a low-achieving school, whether the student is low,
average or high-achieving, were more likely to transfer to a low-achieving school and less likely
to switch to a high-achieving school, relative to their counterparts in average-achieving schools.
Conversely, students in high-achieving schools, regardless of their prior achievement, were more
likely to transfer to another high-achieving school and less likely to transfer to low-achieving
schools.
[Insert Table 9 around here]
The destination patterns are similar across the different timing of non-structural school
changes.
24
The results indicate that there are differential mobility patterns across low-and high
achieving schools. There is evidence of sorting based not only on students’ demographics and
prior achievement, but also by school quality in CCSD. The results imply that origin school
quality may be a more influential factor in determining destination school quality than prior
student achievement and the timing of school changes. Given that most high-achieving students
are enrolled in high-achieving schools, and low-achieving students tend to cluster in lower
quality schools, these systematic differences in student mobility may lead to student
segmentation within CCSD over time.
On the one hand, differential mobility patterns may be indicative of several interrelated
factors related to the demand for school quality. Indeed, parents in school choice environments
reported difficulties in choosing schools and locating a good fit for their children (Jochim,
24
For brevity’s sake, I only present results for all non-structural moves. Results by the timing of school changes are
available upon request.
45
Dearmond, Gross, & Lake, 2014). Parents may consider other non-academic attributes such as
distance and transportation options, sports and extracurricular programs, curricular focus, and
historical “reputation” (Hastings, Kane & Staiger, 2008). Thus, parents still seek a more optimal
match but this may not always include sending their children to a higher quality school. Another
plausible explanation of differential mobility patterns is that families with higher-achieving
students are considering and prioritizing school quality when switching schools and will likely
attend the higher quality schools in CCSD. Differential mobility patterns may also be partly
attributed to disparities in information about school quality across families.
On the other hand, supply-related factors such as school-initiated sorting in a
“traditional” district provide an alternate interpretation of the findings. This may occur in several
ways such as “pushing out” low-achieving students or “cream skimming” high-achieving
students. For instance, some schools may be “pushing out” students, especially during the school
year. There is evidence of an interesting relationship between gender and the timing of non-
structural school changes with male students being more likely to switch schools during the
school year due to schools’ discipline policies and practices. It is also conceivable that
differential mobility patterns may be the product of a mixture of both undesirable school-
initiated student mobility as well as demand asymmetries among families.
Specification Checks and Limitations
The first set of specification checks investigates whether the results are sensitive to the
measures of student achievement and school quality. I re-run the exit analyses using different
indicators of student achievement and school quality. For student achievement, I use students z
score relative to district instead of school, whether or not the student is proficient and the
students’ place in the school achievement distribution where Low
ist
and High
ist
represent the
46
main effects of being in the bottom and top third, respectively, of a school’s achievement
distribution, by grade and year. For school quality, I use schools’ z score. Overall, the results are
qualitatively similar regardless of the measures of student achievement and school quality used.
For the destination analysis, instead of using categorical independent variables, I re-run the
models using dummy and continuous variables. In particular, I use students’ z score and whether
the student was proficient in math. For school quality, I used schools’ z score and the percentage
of proficient students. The results are qualitatively similar.
In addition, I use an alternative strategy to account for unobserved changes over time at
the school level. Instead of using school quality in the exit analysis, I use changes in school
quality (SQ
st
–SQ
st-1
), which examine the trajectory of the schools. The interpretation of the
school quality indicator changes when I make school quality marginal. This analysis tackles a
different but related research question of whether the trajectory or growth patterns of schools are
playing a role in the likelihood of changing schools. The results are similar: students in schools
that the percent proficient of students was increasing (higher performing schools) were more
likely to exit during the summer and less likely to change schools during the school year.
The second set of specification checks examines whether the results are sensitive to the
open enrollment options in the CCSD. Student mobility patterns may be driven by mobility to
and from magnet and charter schools. I re-run the exit and destination analyses excluding student
mobility associated with these schools. The results did not change when charter schools were
excluded. However, there were interesting changes in exit patterns for within-year movers when
magnet schools were excluded. Specifically, the school quality coefficient lost statistical
significance across all specifications. This suggests that the role of school quality in mid-year
patterns may be largely attributed to magnet schools in CCSD and that differential exit patterns
47
for during the year mobility are partly driven by magnet schools in CCSD. In addition, I use a
more restricted sample of students that were continuously enrolled for the length of the panel
excluding entrants, leavers and students who left and returned to the district. The results are
qualitatively similar.
The third set of specification checks examine whether the results vary for elementary,
middle, and high schools. I re-run the exit and destination analyses separately by the level of
schooling. The results are qualitatively similar except for the role of school characteristics in the
timing of exit patterns. For elementary schools, the coefficient on school quality is significant
only for between year moves, in middle schools the coefficient is significant only for mid-year
moves and the coefficient is insignificant across the timing of schools changes in high schools.
This implies that school characteristics play an influential role in between school year moves in
elementary schools and mid-year moves in middle schools.
Student achievement may be both a cause and a consequence of student mobility. The
persistence of mobility effects is a key factor in unravelling this complex relationship. The
majority of the literature has viewed student achievement as a consequence rather than a cause of
mobility. However, while the effect of mobility on achievement may decay over time (Hanushek
et al., 2004; Kain & O’Brien, 1998; Kerbow, 1996), the effect of achievement on the likelihood
of changing schools may be more lasting. Further disentangling this relationship is a direction for
future research.
Limitations. The study has two main limitations. First, I cannot definitively disentangle
school-driven sorting from student-driven sorting. There is no way to determine whether a
student chose to leave a school or was pushed out by a school. The same holds for the patterns of
school entry as there is no way to disentangle “cream-skimming” from demand-side market
48
failures (e.g., information asymmetry). Although I find evidence that origin school quality
predicts destination school quality, I cannot differentiate whether students attending low-
achieving schools are making informed, but different, decisions than their high-achieving peers
or if there are structural barriers or school-initiated sorting mechanisms that impede them from
attending high-achieving schools. In other words, without data on the exact reasons for switching
schools, it is difficult to definitively know whether the sorting is driven by the demand side (e.g.,
differential parental preferences) or the supply side (e.g., “cream-skimming” or “pushing out”).
Second, even though additional valuable insight can be garnered by examining the geographic
location of origin and destination schools given the importance of location in school choice
decisions, I currently do not have the data to conduct a geographic analysis.
Discussion
This paper examines patterns of student mobility in a large “traditional” urban school
district. Three critical lessons emerge from the multitude of findings. First, the importance of
differentiating student mobility by the timing of school changes cannot be overemphasized. The
results illustrate that granular definitions of timing of school changes expose important nuances
and differences in mobile students and mobility patterns that are masked when all non-structural
moves are combined. Second, mid-year mobility matters most. The most disadvantaged and low-
achieving subpopulation of mobile students tends to move mid-year and school-initiated changes
appear most likely to occur during the school year. Indeed, the results highlight an important yet
overlooked relationship between mid-year school changes and school discipline. Third, student
mobility may lead to stratification over time within a ‘traditional’ urban school district. The
findings of this paper provide strong evidence that within an urban school district, advantaged,
49
non-minority and higher-achieving students tend to enroll in or are more likely to switch to
higher quality schools while disadvantaged, minority and lower-achieving students tend to enroll
in or are more likely to switch to lower quality schools.
The results of this study demonstrate the importance of differentiating student mobility
by the timing of school changes. Non-structural mobility rates vary by the timing of school
changes and not all mobile students are the same. Non-structural mobility rates also vary by
students’ ethnicity, gender and socioeconomic status: the mobility rates of economically
disadvantaged students, minority, ELL students and students receiving special education services
are higher relative to their advantaged, White and Asian counterparts and peers not receiving
services. There are also systematic demographic and achievement differences between movers
by the timing of student mobility. Furthermore, the role of school characteristics on the
likelihood of making non-structural moves also varies by the timing of school changes. Thus, it
is understandable why when all non-structural moves are combined, the results and the
consequent inferences differ from student mobility differentiated by the timing of school
changes.
It is important to differentiate student mobility by the timing of school changes because
lumping all non-structural moves or focusing solely on between year moves simplifies the
complex web of school changes that occurs within school districts and masks the profound
heterogeneity of the mobile student population. The results suggest that prior studies that did not
distinguish non-structural mobility by the timing of school changes may have underestimated the
true prevalence of changing schools given the prevalence of mid-year mobility. Furthermore, the
statistically significant differences among mobile students by the timing of school changes make
differentiating the timing of non-structural student mobility an important exercise, especially for
50
investigating the impact of changing schools on student achievement. The extant literature may
also be treating different mobile students as one homogenous or monolithic group. Mobility
studies that focus solely on between year moves not only ignore a significant proportion of the
mobile student population within a school district but also may conflate between year movers
with other categories of mobile students such as ultra-movers. In sum, failure to differentiate the
timing of school changes will likely lead to biased results when examining student mobility
patterns and effects.
This paper illustrates that mid-year mobility matters most for many reasons. During the
year mobility is prevalent, disproportionately affects Black, male, low-income and low-achieving
students and may be driven in part by school discipline policies. About half of all non-structural
moves occurred during the school year (mid-year or ultra-mover). Low-income, minority and
low-achieving students are more likely to change schools during the school year. Male students
are also more likely to switch schools during the school year when discipline-related moves are
included. The results provide suggestive evidence that school discipline may be affecting within-
year mobility patterns. When faced with increasing accountability pressures, schools may focus
on low-performing students (Rouse, Hannaway, Goldhaber, & Figlio, 2013). School discipline
policies may facilitate the “gaming” of accountability incentives and result in schools initiating
student mobility during the school year. Moreover, school discipline policies and practices may
also affect mid-year student mobility in ways that are not school-initiated. Disciplinary reasons
may also prompt non-structural school changes as interactions with school personnel or
disagreements with school discipline policies and practices may spur parents to change schools
during the school year. Hence, school discipline policies may partly explain the differential
mobility patterns in a variety of ways.
51
In essence, the extant literature has largely ignored arguably the most important mobile
student subgroup. Within year non-structural movers are a lower-achieving and more
disadvantaged subset of non-structural movers. There may also be further heterogeneity in the
mid-year mobile student population that has been overlooked by researchers and policymakers.
This paper is the first study to identify ultra-movers – or students who changed schools between
and during the school year in the same academic year.
The findings of this paper imply that student mobility may lead to stratification within a
“traditional” district over time. Whether driven by demand or supply factors, the results raise
important equity questions about the growing disparity in opportunities and the educational
experiences of students in low quality schools in “traditional” urban school districts. Prior
research found that mobile students in school choice districts typically switched to schools of
similar quality and that the quality of schools students transfer from is a significant predictor of
the quality of schools students transfer to (Kerbow, 1996, Welsh et al. forthcoming). This study
illustrates that origin school quality is a significant predictor of destination school quality
regardless of the timing of school changes and students’ prior achievement in a traditional urban
district.
Policy Implications
The results highlight a few policy implications. First, the findings suggest that a one-size-
fit-all approach to student mobility policy is inappropriate and provide strong justification for
identifying, monitoring and supporting mid-year and ultra-movers. The considerable variation in
mobility rates across the timing of school changes reinforces the importance of extensive data
collection on student mobility in urban school districts. Detailed data on the timing of school
changes makes it possible to identify students who move during the school year as well as
52
schools that have a relatively high within-year exit rate. The results suggest that discipline-
related mobility is more likely to occur during the school year and since most mobility studies do
not differentiate within-year moves, possible undesirable school practices may also be
overlooked. The characteristics of mid-year movers coupled with the relationship between mid-
year mobility and school discipline compels policymakers to place greater attention on school
changes that occur during the academic year. In addition, school districts may also use student
mobility as a measure of disadvantage in order to target resources and support. The timing of
non-structural school changes may serve as an indicator for the nature and magnitude of support
districts provide to mobile students and schools.
Second, the prevalence of mid-year student mobility and the variation in mobility rates by
the timing of school changes within an urban school district may also have implications for
curriculum. As policymakers nationwide and in urban districts mull decisions for curriculum,
student mobility should be a pertinent consideration given the prevalence of mid-year moves.
Highly standardized curriculum may prove beneficial in many senses but may also have other
costs. It is worth considering whether variations in curriculum by the timing of school changes
may help better accommodate mobile students and ease the strain of student mobility on teachers
and schools.
Directions for Future Research
The results also highlight some areas for future research. First, it is critical to better
understand possible drivers of differential mobility patterns such as why parents choose schools
for varying reasons. However, it is difficult to capture the nuances of student mobility using only
administrative data. A qualitative complementary study would be useful to provide more on the
ground, fine-grained possible explanations of student mobility patterns. It would be especially
53
helpful to learn more about the reasons for mobility as well as what school attributes families of
mobile students value when making enrollment decisions. Additional valuable insights can also
be garnered by examining the geographic location of origin and destination schools given the
importance of location in student mobility decisions.
Second, it would be useful to delve deeper into the relationship between student mobility
and school discipline in urban districts. The results raise questions about school discipline
policies and the discretion of school officials in placing students in alternative schools for
disciplinary reasons. Even though a considerable portion of student mobility may be driven by
out of school factors, discipline related-mobility is an important yet overlooked school factor.
There is a need for research on the functioning and effectiveness of zero-tolerance school
discipline policies in order to learn more about the relationship between student mobility and
school discipline. It can be argued that discipline-related mobility should be considered another
distinct type of mobility. Discipline-related school changes do not reflect school choice such as
other non-structural school changes. However, discipline-related movers resemble non-structural
movers more than they do structural movers. Mixed methods may help disentangle student-
initiated from school-initiated moves that are related to school discipline and offer insights on the
complex relationship between student mobility and school discipline.
Third, further research is needed on the role origin schools play in the enrollment
decisions of mobile students. Regardless of the timing of school changes or students’ prior
achievement, origin school quality appears to be the dominant factor in determining the quality
of schools mobile students transfer to. There may be key differences in how schools approach
mobile students that may also contribute to these results. Teachers and counselors may also
affect destination school quality in ways worthy of further research. There may also be peer
54
networks that influence the choice of new schools for mobile students. Given the importance of
origin schools in mobility patterns, it is crucial to get a better sense of how these schools
influence the decision to exit and the choice of destination schools of mobile students.
Conclusion
In sum, this paper provides empirical evidence on the relationship between the timing of
school changes and the exit and destination patterns of mobile students. It bolsters our
understanding of the heterogeneous nature of non-structural movers and the role that student and
school characteristics play in the likelihood of changing schools and the quality of schools
mobile students transfer to across the timing of school changes. These descriptive analyses of
students in CCSD also offer guiding insights for estimating the impact of student mobility on
student achievement within a school district.
55
Chapter One Tables
Table 1
Student mobility rates by the type and timing of school changes: 2008-09 – 2012-13
2008-09 2009-10 2010-11 2011-12 2012-13
Mover 0.32 0.33 0.31 0.30 0.30
(0.46) (0.47) (0.46) (0.46) (0.45)
Type of student mobility
Structural mover 0.14 0.14 0.14 0.14 0.14
(0.35) (0.35) (0.35) (0.35) (0.34)
Non-structural mover 0.18 0.19 0.17 0.16 0.16
(0.38) (0.39) (0.38) (0.37) (0.37)
Timing of non-structural mobility
Between-year 0.09 0.10 0.09 0.08 0.08
(0.28) (0.30) (0.28) (0.27) (0.27)
Within-year 0.06 0.06 0.05 0.05 0.05
(0.24) (0.24) (0.23) (0.23) (0.23)
Between and within-year 0.03 0.03 0.03 0.03 0.03
(0.17) (0.18) (0.17) (0.17) (0.16)
N 325712 322683 321678 320725 329663
Note. N includes leavers, entrants and continuously enrolled students. Standard deviations in parentheses.
56
Table 2
Average annual non-structural mobility rates, by the timing of student mobility and students’ characteristics: 2008-09-2012-13
FRPL Non-
FRPL
Black Hispanic Asian White Special
Ed
Non-
SpecialEd
ELL Non-
ELL
Male Female All
All non-
structural school
changes (%)
21.7 13.0 27.4 18.1 13.0 13.0 21.3 16.8 20.0 16.7 18.0 16.5 17.2
By Timing
Between-year
(%)
10.1 7.3 12.0 8.8 7.6 7.4 10.5 8.4 9.5 8.4 8.8 8.6 8.6
Within-year (%) 7.4 3.9 9.6 6.3 3.7 3.7 6.8 5.6 7.6 5.4 6.0 5.3 5.7
Between and
within-year (%)
4.2 1.7 5.8 3.1 1.6 1.8 4.0 2.8 2.9 2.9 3.2 2.6 2.9
Observations 793824 816542 215297 675880 125769 513984 176467 1407301 253993 1340544 826991 783394 1620639
57
Table 3
Demographic and achievement characteristics of mobile students by the type and timing of student mobility: 2008-09 – 2012-13
Type of Mobility Timing of Non-structural mobility
District
Avg.
Structural Non-
structural
Diff
(NS-S)
BY
switcher
WY
switcher
Diff (WY-
BY)
Ultra-
mover
Diff (UM-
WY)
Black 0.13 0.13 0.21 -0.09 0.18 0.23 0.06 0.27 0.03
(0.34) (0.33) (0.41) (0.001) (0.38) (0.42) (0.002) (0.44) (0.003)
Hispanic 0.41 0.42 0.44 -0.02 0.43 0.45 0.02 0.44 -0.01
(0.49) (0.50) (0.50) (0.002) (0.50) (0.50) (0.003) (0.50) (0.004)
Asian 0.08 0.08 0.06 0.02 0.07 0.05 -0.02 0.04 -0.01
(0.27) (0.27) (0.24) (0.001) (0.25) (0.22) (0.001) (0.20) (0.002)
White 0.32 0.32 0.24 0.07 0.27 0.22 -0.06 0.20 -0.02
(0.47) (0.47) (0.43) (0.002) (0.44) (0.41) (0.002) (0.40) (0.003)
Male 0.51 0.51 0.53 -0.02 0.52 0.55 0.04 0.56 0.007*
(0.50) (0.50) (0.50) (0.002) (0.50) (0.50) (0.003) (0.50) (0.004)
Special Education 0.11 0.11 0.13 -0.02 0.14 0.13 -0.01 0.15 0.02
(0.31) (0.31) (0.34) (0.001) (0.33) (0.34) (0.002) (0.35) (0.003)
English Language Learner 0.16 0.10 0.19 -0.09 0.18 0.21 0.03 0.16 -0.02
(0.37) (0.29) (0.39) (0.001) (0.38) (0.41) (0.002) (0.36) (0.003)
FRPL 0.49 0.51 0.61 -0.09 0.57 0.64 0.06 0.70 0.08
(0.50) (0.50) (0.49) (0.002) (0.49) (0.48) (0.003) (0.46) (0.003)
Math Achievement (z score relative to
school)
0.00 0.02 -0.25 0.26 -0.12 -0.36 -0.23 -0.39 -0.03
(1.00) (0.99) (1.02) (0.03) (0.99) (1.02) (0.005) (1.02) (0.008)
Math Achievement (z score relative to
district)
0.00 0.02 -0.33 0.34 -0.15 -0.48 -0.33 -0.55 -0.07
(1.00) (0.99) (1.00) (0.004) (0.97) (0.98) (0.005) (0.97) (0.007)
Below Average in Math (relative to
school)
0.48 0.47 0.58 -0.11 0.54 0.62 0.09 0.64 0.02
(0.50) (0.50) (0.49) (0.002) (0.50) (0.48) (0.003) (0.48) (0.004)
Below Average in Math (relative to
district)
0.48 0.48 0.62 -0.14 0.55 0.68 0.13 0.71 0.03
(0.50) (0.50) (0.49) (0.002) (0.50) (0.47) (0.003) (0.46) (0.003)
Math proficiency 0.61 0.64 0.49 0.15 0.57 0.41 -0.17 0.42 -0.005*
(0.49) (0.48) (0.50) (0.002) (0.50) (0.49) (0.003) (0.49) (0.004)
58
N 1946446 229557 289,146 140439 120796 46903
Note. BY= Between school years, WY= Within the school year .Ultra-mover refers to student that changed schools both between and within the school year in
the same academic year. Standard deviations in parentheses. *- differences are not statistically significant
59
Table 4
School Characteristics by Exit rates, by the timing of school changes: 2008-09 – 2012-13
School Characteristics Between-yr exit rate (%) Within-yr exit rate (%)
Percent FRPL (deciles)
1 (0-19.3%) 15.9 4.3
2(19.4-26.2%) 9.4 4.2
3 (26.3-33.3%) 10.1 5.3
4 (33.4-42.2%) 10.3 5.9
5 (42.3-50.9%) 10.4 7.7
6 (51-58.2%) 10.3 7.6
7 (58.3 -66.8%) 9.4 8.8
8 (66.9 -76.1%) 9.5 8.8
9 (76.2-85.2%) 10.7 9.4
10 (85.3 -100%) 10.5 8.8
Percent Minority (deciles)
1(0-31.9%) 8.5 2.9
2 (32-40.3%) 9.2 4.5
3 (40.4 – 48.8%) 11.2 5.2
4 (48.9 -56.5%) 9.7 6.8
5 (56.6-64.6%) 11.6 8.0
6 (64.7-73%) 12.4 8.2
7 (73.1-79.7%) 10.1 7.6
8 (79.8-85.2%) 11.6 9.0
9 (85.3 -91.3%) 11.2 10.9
10 (91.4-100%) 10.6 8.4
Percent Proficient (deciles)
1 (0 -30.8%) 13.6 15.6
2 (30.9 – 45%) 9.0 9.5
3 (45.1-52.5%) 8.8 7.0
4 (52.6- 57.9%) 9.4 7.4
5 (58 -62.3%) 10.2 7.2
6 (62.4-67%) 9.9 6.4
7 (67.1-71.1%) 9.9 6.1
8 (71.1 – 74.8%) 10.3 5.5
9 (74.9 -80.6%) 11.4 4.7
60
10 (80.7 - 100%) 9.1 3.6
61
Table 5
Exit school quality by destination school quality for non-structural movers by the timing of school changes
Average Destination Decile
Percent Proficient
range of decile
Exit School Decile All moves
combined
Between
year
Within
year
Between
and Within
0- 32.6% 1 1.79 1.48 1.91 1.84
32.7%-43.4% 2 2.88 2.89 2.93 2.77
43.5%-50.5% 3 3.67 3.71 3.73 3.47
50.6%-55.7% 4 4.26 4.29 4.32 4.03
55.8%-61.1% 5 4.93 5.08 4.88 4.62
61.2%-65.6% 6 5.82 6.03 5.69 5.49
65.7%-70.2% 7 6.42 6.68 6.25 6.06
70.3%-74.1% 8 6.86 7.07 6.76 6.27
74.2%-79.6% 9 7.7 8.0 7.42 6.93
79.7%-100% 10 8.62 8.82 8.37 7.84
62
Table 6
Exit school by destination school demographic composition for non-structural movers by the timing of school changes
Average Destination Decile
Percent nonwhite
range of decile
Exit School Decile All moves
combined
Between
year
Within
year
Between
and Within
0- 35.5% 1 1.86 1.65 2.09 2.3
35.6%-41.7% 2 2.83 2.53 3.04 3.3
41.8%-49.4% 3 3.48 3.23 3.70 3.81
49.5%-57.5% 4 4.38 4.23 4.47 4.62
57.6%-64.8% 5 5.02 4.9 5.11 5.15
64.9%-73.3% 6 6.01 5.9 6.12 6.0
73.4%-79.1% 7 6.84 6.77 6.88 6.9
79.2%-84.2% 8 7.51 7.47 7.58 7.43
84.3%-90.4% 9 8.14 8.15 8.18 8.02
90.5%-100% 10 9.0 9.09 8.98 8.86
63
Table 7
Estimation of linear probability of non-structural move by timing of school changes: 2007-08 -2012-13
(1) (2) (3) (4)
All moves combined (N=773,180)
Prior student achievement -0.03*** -0.03*** -0.07*** -0.06***
(0.002) (0.002) (0.005) (0.004)
Percent proficient in school -0.29*** -0.29*** -0.10**
(0.07) (0.07) (0.03)
Prior achievement * Percent proficient 0.08*** 0.05***
(0.01) (0.006)
Constant 0.75*** 0.56*** 0.56*** 0.87***
(0.009) (0.16) (0.15) (0.09)
Between-year moves (N=119,866)
Prior student achievement 0.03*** 0.03*** 0.02*** 0.02**
(0.003) (0.002) (0.008) (0.006)
Percent proficient in school 0.04 0.04 0.05**
(0.02) (0.02) (0.02)
Prior achievement * Percent proficient 0.02 0.01
(0.01) (0.01)
Constant 1.23*** 1.61*** 1.16*** 0.76***
(0.01) (0.08) (0.08) (0.12)
Within-year moves (N=118,894)
Prior student achievement -0.03*** -0.02*** 0.0003 -0.001
(0.002) (0.002) (0.004) (0.004)
Percent proficient in school -0.15* -0.16* -0.06
(0.08) (0.08) (0.04)
Prior achievement * Percent proficient -0.04*** -0.03***
(0.008) (0.008)
Constant -0.14*** -0.17 -0.17 0.13
(0.01) (0.14) (0.14) (0.13)
Between & Within year moves (N=118,894)
Prior student achievement -0.008*** -0.007*** 0.01*** 0.01***
(0.001) (0.001) (0.004) (0.004)
Percent proficient in school -0.05* -0.06* -0.02
(0.02) (0.03) (0.02)
64
Prior achievement * Percent proficient -0.04*** -0.04***
(0.006) (0.006)
Constant -0.09*** -0.19*** -0.18*** 0.14
(0.006) (0.04) (0.04) (0.10)
Student Level Controls Yes Yes Yes Yes
School Level Controls No Yes Yes Yes
Student/School interactions No No Yes Yes
School Fixed Effects No No No Yes
*** p<0.001, ** p<0.01, * p<0.05
Note. I use cluster-robust standard errors at the school-level to account for within school correlation of the student-level error terms, as well as school-level serial
correlation. Student-level controls include indicators for race/ethnicity (Black, Hispanic, Asian), gender, special education, ELL and FRPL status. School-level
controls include: Percent Black, Percent Hispanic, Percent Asian, Percent Male, Percent special education, Percent ELL and Percent FRPL.
65
Table 8
Multinomial Logit Regression Predicting Destination School Achievement for All non-structural moves combined (N=127,713)
Destination School Quality
Low Achieving High Achieving
Low Achieving Student -0.25*** -0.04
(0.03) (0.04)
High Achieving Student -0.43*** 0.13**
(0.06) (0.05)
Low Achieving Origin School 1.75*** -0.48***
(0.13) (0.09)
High Achieving Origin School -0.23 1.57***
(0.10) (0.10)
Low Student*Low School 0.11* -0.03
(0.05) (0.06)
Low Student*High School 0.06 0.15**
(0.06) (0.05)
High Student*Low School 0.08 -0.06
(0.08) (0.09)
High Student*High School -0.01 0.22**
(0.07) (0.07)
Black 0.28*** -0.28***
(0.05) (0.04)
Hispanic 0.37*** -0.26***
(0.04) (0.04)
Asian -0.10* 0.05
(0.06) (0.05)
Male 0.06*** -0.03
(0.02) (0.02)
Special Education 0.06* 0.08**
(0.03) (0.03)
ELL -0.43*** -0.21***
(0.04) (0.05)
FRPL -0.04 -0.27***
(0.04) (0.04)
66
Constant -0.78*** -1.04***
(0.11) (0.12)
Pseudo R-squared 0.22 0.22
Model chi-square 2304.64 2304.64
*** p<0.001, ** p<0.01, * p<0.05
Note: Robust standard errors in parentheses. I use cluster-robust standard errors at the school-level to account for within school correlation of the student-level
error terms, as well as school-level serial correlation.
67
Table 9
Marginal effects of origin school quality on destination school quality by student achievement (All moves, N=127,713)
Destination School
Student Achievement/Origin School Quality Low Achieving
School
Average
Achieving
School
High
Achieving
School
Low Achieving School vs Middle Achieving School
Low Achieving Student 0.44*** -0.27*** -0.17***
(0.03) (0.03) (0.02)
Middle Achieving Student 0.46*** -0.29*** -0.17***
(0.03) (0.03) (0.01)
High Achieving Student 0.47*** -0.28*** -0.19***
(0.03) (0.03) (0.02)
High Achieving School vs Middle Achieving School
Low Achieving Student -0.18*** -0.23*** 0.41***
(0.03) (0.03) (0.03)
Middle Achieving Student -0.17*** -0.28*** 0.45***
(0.02) (0.03) (0.03)
High Achieving Student -0.16*** -0.29*** 0.46***
(0.02) (0.03) (0.03)
Note. Standard errors in parentheses; *** p<0.001, ** p<0.01, * p<0.05. The reference group for each student is a student in the same third of achievement in a
school in the middle third of achievement. For example, in the first row, a low achieving student in a low achieving school has a significantly higher probability
of moving to a low achieving school than a low achieving student in an average achieving school.
68
CHAPTER TWO
DISENTANGLING THE TIMING AND IMPACT OF STUDENT MOBILITY
Abstract
Although student mobility is a widespread phenomenon across urban school districts in the U.S.,
the majority of the research on the effects of changing schools focuses mainly on school changes
that occur between school years and is correlational due to methodological challenges. Further,
few studies have disentangled the net impact of changing schools into effects associated with
changes in school quality from those linked to the disruption of changing schools. This paper
estimates the impact of student mobility on the achievement of mobile students across the timing
of school changes using student fixed effects and propensity score matching. The results indicate
that the impact of student mobility varies with the timing of school changes: student mobility
between school years was associated with small, insignificant transactions costs of moving
whereas moves during the school year were associated with large, significant transition costs of
moving and a -0.08SD decline in math achievement growth. There is little evidence of positive
systematic changes in school quality regardless of the timing of student mobility. The disruptive
effect of student mobility appears to be the driving factor of the overall impact of student
mobility. The results also imply that even though the reasons for mobility may likely bias
estimates from quasi-experimental methods, there may be still an independent adverse effect
(especially for mid-year moves) associated with changing schools. Policy implications and
directions for future research are also discussed.
69
DISENTANGLING THE TIMING AND IMPACT OF STUDENT MOBILITY
Student mobility is prevalent nationwide and especially widespread in urban school
districts. Most K-12 students make at least one non-structural move in their educational career.
There is a longstanding concern among families, educators, researchers and policymakers about
whether changing schools inhibits or improves student achievement. Despite a robust literature,
there is no consensus on the effects of student mobility on student achievement. On the one hand,
the majority of the extant literature suggests that changing schools is associated with lower test
scores (Institute of Medicine and National Research Council, 2010; Mehana & Reynolds, 2004;
Reynolds et al., 2009; United States Government Accountability Office, 2010). On the other
hand, several scholars posit that switching to higher quality schools may improve student
achievement (de la Torre & Gwynne, 2009; Engberg et al., 2012; Hanushek et al., 2004;
Rumberger et al., 1999; Temple & Reynolds, 1999).
The divergent assumptions of educational policymakers and researchers regarding the
impact of student mobility are reflected in education policy. Some policies discourage student
mobility whereas others encourage changing schools. Several researchers have suggested ways
to reduce student mobility such as allowing students who moved out of the catchment area to
continue attending the same school, increasing parent engagement and improving families’ sense
of belonging in the school community, and providing parents with information about the
importance of school stability and attendance zones (Kerbow, 1996; Kerbow, Azcoitia, & Buell,
2003; Rumberger & Larson, 1998; Rumberger, 2003; Temple & Reynolds, 1999). Over the past
decade, there has been a proliferation of school choice policies that rely on non-structural
mobility to drive innovation and competition among schools, raise school quality, and improve
student outcomes (KE Bulkley, Henig, & Levin, 2010; Levin, Daschbach, & Perry, 2010). It is
70
important and timely to gain a better understanding of the impact of student mobility on student
achievement within urban school districts given the growing importance of student mobility in
education policy and its potential for improving educational outcomes.
The non-random nature of student mobility makes it difficult to attribute a causal link
between changing schools and student achievement. Consequently, the lion’s share of research
on the impact of student mobility is correlational in nature. The impact of student mobility on
student achievement depends on multiple factors including: the frequency of student moves, the
timing of school changes (between vs. during the school year), the nature of the school change
(between vs. within school district), grade, a student’s previous history of mobility, the reason
for student mobility as well as a student’s personal and family characteristics, and the quality of
schools students transfer to (Institute of Medicine and National Research Council, 2010; Mehana
& Reynolds, 2004; Reynolds et al., 2009; Rumberger, 2002; United States Government
Accountability Office, 2010). Student mobility studies vary in the identification strategy and data
used to estimate the effect of mobility. Reynolds et al. (2009) found that the specification of
covariates had the biggest influence on effect sizes and there is substantial variation among
published and unpublished studies in controlling for pre-mobility achievement and
characteristics (Reynolds et al., 2009).
Even though the quality of student mobility studies estimating the impact of changing
schools has increased over time (Institute of Medicine and National Research Council, 2010),
estimates of the impact of changing schools in the extant literature remain susceptible to bias
from the reasons for mobility and omitted contributing factors. The majority of the estimates of
the impact of student mobility in the extant literature typically include the possible effect of the
circumstances that precipitated the change in schools (e.g., job loss) on student achievement.
71
Unless all the influences in the decision to move that also affect achievement are completely
accounted for, the estimate of mobility effects will be bundled with family circumstances and
economic shocks that may have prompted the change of schools (Hanushek et al., 2004). For
instance, changing schools may have no effect, and the link between student mobility and
educational outcomes may be a vestige of economic hardships, residential mobility, and
transitions to parents’ new employment etc. that prompted families to change schools. Thus,
negative effects associated with student mobility may contain both the impact of changing
schools and the unobserved factors that led to the school change.
Estimates of the impact of changing schools may also be biased by the lack of granular
differentiation of the type and timing of school changes. Few studies differentiate structural
moves from non-structural moves. The majority of the extant student mobility literature does not
differentiate non-structural moves by the timing of school changes. In addition, the net impact of
student mobility on student achievement includes the disruptive costs of changing schools and
the effects on student achievement attributable to changes in school quality (Tiebout effect).
Only one previous study has separated the impact of the changes in school quality on student
achievement from the disruptive effects of student mobility (Hanushek et al., 2004). The net
impact of student mobility as well as it components (disruptive and Tiebout effects) may also
vary with the timing of school changes. Prior research has also highlighted the importance of
examining conditional effects or how mobility effects may vary with students’ characteristics
(Institute of Medicine and National Research Council, 2010). Overall, there is a need for
research on the effects of intra-district student mobility that simultaneously accounts for the
heterogeneity of mobile students as well as the type and the timing of school changes,
72
disentangles the disruptive versus Tiebout effects, and offers insights on how biased estimates of
mobility effects may be to the reasons prompting the school change.
In this paper, I use student-level data from Clark County, Nevada to estimate the impact
of changing schools on the achievement of mobile students within an urban school district. I pay
particular attention to the role that the timing of non-structural school changes play in
determining the net effect of student mobility. This study disentangles the net impact of mobility
into changes in school quality and disruption effects across the timing of school changes. I use
two quasi-experimental methods with different identification strategies to estimate mobility
effects. I complement the primary student fixed effects models with propensity score matching in
order to use a different counterfactual that may shed light on the extent of the bias attributed to
reasons for mobility. This paper is one of the first studies to apply multiple methods to estimate
the impact of student mobility and contributes to a better understanding of the role of the timing
of school changes in mobility effects. In particular, I ask the following research questions:
a.) What is the impact of student mobility on mobile students’ achievement?
b.) How do the disruptive effects and the changes in school quality (Tiebout effect)
associated with student mobility vary with the timing of school changes?
The results indicate that the effect of student mobility varies with the timing of school
changes: student mobility between school years is associated with a positive impact on math
achievement growth whereas moves during the school year are associated with a negative effect
on math achievement growth. Furthermore, the transactions costs of moving vary by students’
race/ethnicity and income as well as the timing of school changes. Overall, the results paint a
73
nuanced mobility effects landscape and demonstrate compelling heterogeneity in the mobile
student population. In what follows, I provide a brief empirical and conceptual overview of the
effects of student mobility on mobile students’ achievement. Next, I describe the data and
methods employed in this study before presenting the results. I conclude with a discussion of
policy implications and areas for future research.
The Impact of Student Mobility: An Empirical and Conceptual Overview
In the past decades, numerous studies have attempted to estimate the impact of student
mobility.
25
Early student mobility studies typically used cross sectional data, compared mobile
students to stable students, and did not always include controls for the prior achievement and
demographic characteristics of students (Hanushek et al., 2004; Mehana & Reynolds, 2004;
Reynolds et al., 2009). Studies that did not control for the demographic and prior achievement of
students typically found that student mobility negatively impacts student achievement
(Rumberger, 2002; United States General Accounting Office, 1994). However, research has
demonstrated that inclusion of controls for prior achievement and for students’ background
characteristics reduces the statistical significance and size of student mobility effects (Alexander
et al., 1996; Heinlein & Shinn, 2000; Pribesh & Downey, 1999; Reynolds et al., 2009; Temple &
Reynolds, 1999; Wright, 1999). Indeed, pre-existing differences in achievement account for
25
Overall, within-district student mobility is associated with more negative student outcomes, such as lower
achievement scores, than across-district student moves (Hanushek et al., 2004; Institute of Medicine and National
Research Council, 2010; Mehana & Reynolds, 2004; D. Wright, 1999; Xu et al., 2009). Xu and colleagues (2009)
found that whereas within district moves hurt students, cross-district moves benefitted or had no effect on students’
mathematics performance (Xu et al., 2009). Student mobility to a different district within the same region
significantly increased achievement but student mobility within a district (especially for frequent movers) and
moves to a new district in a new region were not linked to improvements in school quality (Hanushek et al., 2004).
This is not surprising given that a substantial proportion of intra-district student mobility is reactive rather than
strategic (Alexander et al., 1996; Hanushek et al., 2004; Institute of Medicine and National Research Council, 2010;
Rumberger et al., 1999).
74
about half of the variation in achievement between mobile and non-mobile students (Alexander
et al., 1996; Pribesh & Downey, 1999).
In their meta-analysis of 26 studies dated between 1975 and 1994 that examined the
effect of student mobility on reading and math achievement in K-6, Mehana and Reynolds
(2004) found that changing schools has a negative effect on reading (-0.25 of a standard
deviation) and math (-0.22SD) achievement (Mehana & Reynolds, 2004). However, only five of
the studies included were published in peer-reviewed journals, few of the studies controlled for
pre-mobility achievement and student mobility included both structural and non-structural
moves. In another meta-analysis of 16 studies from 1990 through 2008 with a measure of pre-
mobility achievement and a focus on non-structural student mobility, Reynolds et al. (2009)
found that student mobility was associated with a -0.07SD and -0.08SD in reading and math
achievement for each additional move (Reynolds et al., 2009). These estimates of the impact of
changing schools remained negative and significant but were much smaller compared to those
found by Mehana and Reynolds (2004). Reynolds et al. (2009) also illustrated that larger effect
sizes may be due to frequent mobility: the effect sizes for frequent student mobility were -
0.21SD and -0.23SD respectively (Reynolds et al., 2009).
More importantly, students who make non-structural school changes may differ from
non-mobile students in unobserved ways that are not accounted for by including background
characteristics and prior achievement. Thus, the inclusion of a host of school and student
characteristics does not assure the identification of the casual effect of student mobility due to
the non-randomness of student moves (Hanushek et al., 2004). Studies in the past decade have
provided more convincing estimates of the impact of changing schools by using longitudinal data
that tracks individual students over time, and including student fixed effects in regression models
75
(Grigg, 2012; Hanushek et al., 2004; Schwartz et al., 2009; Xu et al., 2009). The inclusion of
student fixed effects compares movers to themselves rather than non-movers as opposed to
earlier studies that compared mobile students to non-mobile students. Comparing the before and
after academic performance of the same mobile student is considered a better methodological
approach than comparing movers and non-movers in the identification of mobility effects
(Hanushek et al., 2004). Student fixed effects also control for time-invariant student
characteristics or unobserved attributes that do not change over time such as motivation and
ability. Unsurprisingly, estimates of the negative effect of changing schools have decreased even
when student fixed effects are employed. For example, Hanushek et al. (2004) demonstrated that
the estimate of non-structural mobility declined from -0.17SD when comparing movers to non-
movers to -0.014SD when one controls for pre-mobility achievement in a value-added
specification.
Disentangling Mobility Effects
Figure 1 illustrates a simplified conceptual framework of how changing schools may
affect mobile students’ achievement. The conceptualization of the impact of student mobility on
mobile students’ achievement is complex. Students attend a given school nested within a
neighborhood in a larger school district. Observable and unobservable student, school and
neighborhood characteristics may directly influence students’ achievement. The reasons why
students switch schools may also be related to these observed and unobserved student, school
and neighborhood characteristics. More importantly, the school-related or non-school-related
circumstances that prompted a student to change schools may also directly impact student
achievement. Moving from one school to another in a given school year may be due to various
reasons including but not limited to any combination of the following: residential change,
76
negative family circumstances (e.g., divorce or job loss), positive family circumstances (e.g., job
promotion), school quality preferences and other school-related factors. In essence, a
constellation of concurrent and possibly competing effects determine the net impact of student
mobility on mobile students’ achievement including: the motivating reason for student mobility,
the adjustment costs of a new school and/or neighborhood environment and the effect of moving
to a higher quality school.
[Insert Figure 1 around here ]
The costs of student mobility. Transactions costs of student mobility refer to the
disruption after a transfer to a new school. Student mobility introduces discontinuity in learning
environments and disrupts stability (Institute of Medicine and National Research Council, 2010;
Reynolds et al., 2009; Temple & Reynolds, 1999). Several theories of child development
emphasize the importance of regularity and stability in early learning environments
(Bronfenbrenner, 1979, 1989, 1994; Cole, Cole, & Lightfoot, 2005; M. Wright & Masten, 2005).
The developmental perspective posits that disruptions in a child’s social, emotional and cognitive
development may harm students’ educational outcomes (Institute of Medicine and National
Research Council, 2010; Temple & Reynolds, 1999).
26
Regardless of the motivating reason, type
or timing of student mobility, changing schools is an ecological transition. Ecological transitions
are changes in the environments, roles and expectations of students (Bronfenbrenner, 1979).
26
For example, Bronfenbrenner’s ecological model of human development has two main premises. First, human
development occurs through increasingly complex and reciprocal interactions between organisms and their
immediate environment (persons, objects and symbols). Regular interactions over extended periods of time is best
for effective development; these interactions, or proximal processes include individual or group play, reading,
athletic activities and learning new skills (Bronfenbrenner, 1994). Second, the nature (form, power, content and
direction) of the proximal processes impacting development is a joint function of the characteristics of the
developing human and the environment in which these interactions occur as well as the developmental outcome of
interest. Even though proximal processes may have a greater effect on development than environments, for
outcomes measuring developmental competence such as academic achievement, proximal processes are more
impactful in “more advantaged and stable environments throughout the life course” (p. 38, Bronfenbrenner, 1997).
77
Student mobility is a major disruption in the school organization that students are
exposed to. When students switch schools, the school (physical) environment as well as the
instruction of students including the curriculum, textbooks, teachers and teaching style may
differ dramatically across the origin and destination schools (Institute of Medicine and National
Research Council, 2010; Kerbow, 1996; Mehana & Reynolds, 2004; Reynolds et al., 2009;
Temple & Reynolds, 1999). Research has found that there are many difficulties associated with
switching schools such as acclimatizing to curriculum changes and school environment as well
as struggling with subject matter (Alexander et al., 1996; Kerbow, 1996; Lash & Kirkpatrick,
1990; Mehana & Reynolds, 2004; Nelson, Simoni, & Adelman, 1996; Rumberger et al., 1999).
When students move, there is a dislocation in their social environment. Student mobility disrupts
the network of social relationships students have with peers and adults in schools and
neighborhoods that constitute social capital (Coleman, 1988). Changing schools disrupts and
weakens peer relationships, lowers social ties and school attachments as well as engagement in
school and community, and increases the risk of underachievement (Astone & Mclanahan, 1994;
Coleman, 1988; Pribesh & Downey, 1999; South, Haynie, & Bose, 2007; Swanson & Schneider,
1999; Teachman, Paasch, & Carver, 1996). Research has also found that mobile students
experience challenges in making new friends and acclimating socially to a new school
environment (Rumberger et al., 1999).
The nature and extent of adjustment costs related to student mobility may also vary with
the timing of school changes. Student mobility during the school year may be more disruptive
than moves between academic years (Alexander et al., 1996; Burkam et al., 2009; Grigg, 2012;
Hanushek et al., 2004; Schwartz et al., 2009). This may be due to several likely reasons such as
higher adjustment costs of navigating two or more different schooling environments over the
78
course of a given school year; limited available options for destination schools, hence schooling
decisions for mid-year movers may be driven more by convenience and availability than school
quality preferences; and, exigent motivating reasons for student mobility such as an unexpected
job loss. Prior research has found that moves during the school year have a negative impact on
student achievement (Burkam et al., 2009; Engec, 2006; Grigg, 2012; Hanushek et al., 2004;
Schwartz et al., 2009). Some studies have also found that moves during the academic year do not
have significantly greater adverse effect than school changes between academic years (Grigg,
2012; Hanushek et al., 2004).
The benefits of student mobility. “Tiebout” improvements occur when students switch
to higher quality schools and any disruptive effects of changing schools are offset by gains from
improved school quality (de la Torre & Gwynne, 2009; Engberg et al., 2012; Hanushek et al.,
2004; Rumberger et al., 1999; Swanson & Schneider, 1999; Temple & Reynolds, 1999). This
“school quality” or “Tiebout” effect may represent improvements in the level of resources,
quality of teachers and peers or other attributes of the quality of a school that positively impact
student achievement (Hanushek et al., 2004). There is little evidence of systematic non-structural
student mobility from a lower to a higher quality school within a school district (Cullen et al.,
2005; Hanushek et al., 2004). Kerbow (1996) reported that students moved to a higher quality
school less than half the time, even when school quality is the motivating reason for changing
schools (Kerbow, 1996).
Changing schools is a disruptive event that will likely have at least short-term
transactions costs for mobile students. Student mobility may adversely affect mobile students’
achievement as they adjust to new schools and/or new neighborhoods. Consequently, the
majority of the literature on student mobility has conceptualized switching schools as a harmful,
79
negative event (Burkam et al., 2009). However, there is a considerable degree of “push” and
“pull” in how student mobility may affect student achievement. In their seminal study that
separated the impact of the changes in school quality from the disruptive effects of student
mobility, Hanushek et al. (2004) found significant differences in the average change in school
quality and transactions costs of changing schools across the timing of school changes and
destination of school change (Hanushek et al., 2004). The elements of these transactions costs
measured by the empirical model are school assimilation or schools’ policies and practices to
assimilate new entrants, and disruption or the costs of moving independent of school quality.
Thus, school-related and non-school-related transition costs are lumped together in the
estimation of transactions costs. It is important to highlight that these transactions costs may also
include the changes in family circumstances associated with a school change (Hanushek et al.,
2004).
In this paper, I examine how the impact of intra-district student mobility varies with the
timing of school changes. Specifically, I decompose the net effect of mobility into changes in
school quality and transactions costs as well as examine conditional effects. I complement the
student fixed effects model with propensity score matching to learn more about the nature and
extent of the bias attributed to reasons for mobility. Less than a handful of student mobility
studies have used propensity score matching as an empirical strategy (Carolan, 2013; de la Torre
& Gwynne, 2009; Gasper et al., 2012) and no previous study has used propensity score matching
or propensity score matching in tandem with student fixed effects to estimate the effect of non-
structural student mobility on students’ achievement growth. In the next section, I describe the
data and empirical strategies employed in this study.
80
Data and Methods
Data
I use a six year panel of student-level data for all students in the CCSD from 2007-08
through to 2012-13.
27
The data contains students’ demographic characteristics and annual test
scores from the Nevada Proficiency Examination Program. Demographic data includes
indicators for students’ gender, race/ethnicity (Black, Hispanic, Asian, White), free and reduced
priced lunch (FRPL), English Language Learner (ELL) and special education statuses. Students
are tested in reading and math in grades 3-8 and take the High School Proficiency Exam (HSPE)
in grade 10. I standardize test scores for students in grades 3 through 10 by grade and year,
relative to the school mean, as well as relative to the district mean.
28
I complement the student-
level data with school-level data on schools’ location and facilities including zip code and the
year the school was built in addition to publicly available accountability data.
29
Detailed longitudinal data that tracks the dates and sequence of school changes allows for
in-depth classification of the timing of student mobility across a range of grades (K-12). Unique
student and school identifiers in the data link students to schools in each year and across multiple
27
As of 2012-13, there are 357 schools in CCSD (217 elementary schools, 59 middle schools, 49 high schools, 24
alternative schools, and 8 special schools). I have data on some of the alternative schools operated by CCSD
including: 5 behavior, 3 continuation schools, 4 juvenile detention centers and 6 adult education schools. I also have
data on all of the 25 CCSD-run magnet schools and career and technical academies that provide parents with school
choice and 9 Prime 6 schools (The Prime 6 Initiative was first adopted in 1994 in an effort to provide support for 9
schools located in West Las Vegas that serve a predominantly low-income African American and Hispanic
population). I only have data for the charter schools in CCSD for the 2012-13 school year.
28
I use the annual means and standard deviations on the test scores for each grade and year to standardize student
achievement. Standardizing the outcomes makes it possible to compare students' test scores over time, as well as
across grades and provides insight into how a student’s math and reading test scores are compared to students in the
same grade in the school and in the district. If negative relative to the school mean, this indicates that the student in
the school received a math or reading test score below the school average.
29
I obtain Adequate Yearly Progress (AYP) designation data from 2007-08 through to 2011-12 from the Nevada
Department of Education (ayp.nv.gov) and nevadareportcard.com. Schools are designated by the Nevada
Department of Education as: exemplary, continuing exemplary, exemplary turnaround, high achieving, adequate, on
watch list and in need of improvement. In compliance with NCLB, Nevada AYP classifications are made annually
based on the percentage of students tested, the percentage of students tested who score at or above the proficient
level on annual statewide tests, and school attendance or graduation rates.
81
school years. I assume that all school changes between school years in grades 6 and 9 are
transitions from elementary to middle and middle to high schools respectively, with the
exception of students enrolled in combination schools, of which there are relatively few.
30
The
reasons for enrollment and withdrawal, as reported by schools, such as whether the student
previously attended a school in another state or the student left a school to transfer to private or
charter schools in the district are also available.
I use a sample of students that have been continuously enrolled in a CCSD school for at
least two consecutive academic years (in other words, students need at least two observations to
be included and students with only one observation were dropped from the sample). This sample
includes 1,826,170 student-years with 428,247 unique students. I organize the data into two
cohorts across K-8 including: a cohort of kindergarteners in 2007-08 tracked through grade 5 in
2012-13; and a cohort of grade 3 students in 2007-08 who are followed through to grade 8 in
2012-13. Although the majority of the students in both cohorts have been enrolled for the
duration of the panel, the cohorts include entrants, leavers and students who left and returned to
the district. There are 148,010 student-years in the K-5 cohort (30,973 unique students), 149,059
student-years in the 3-8 cohort (30,692 unique students). The flow of students in and out of both
cohorts was similar. The rate of students leaving (4 percent) was slightly higher than the
proportion of students entering the district during the school year (3 percent).
Methods
30
I exclude students who were transported out of the district (students who live in Clark County but attend schools
in another district) and remove all within-school Americans with Disabilities Act (ADA) changes. There are a
number of students who attended two schools at the same time (typically a high school and a technical school). For
these concurrent enrollments, I only use mobility associated with comprehensive high schools and ignore the other
schools. Xu et al. (2009) and Burkam et al. (2009) highlight the importance of including grade retention variables in
mobility studies. For students who were held back in the same grade in the same school (grade retention) or were
accelerated a grade in the same school, I identify those specific moves as possible controls in the analysis. However,
there is limited grade retention or skipping grades in CCSD over the period of study with less than two percent of
students and the majority of changes occurring in grades 9-12.
82
There are four main empirical challenges in estimating the causal impact of student
mobility. First, mobile students are different from non-mobile students in observed (for e.g.,
demographics) and unobserved (for e.g., motivation) ways. These differences in student
characteristics may influence the decision to change schools as well as academic achievement
thus, naïve OLS estimates of mobility effects will likely be upwardly biased due to selection
bias. Decisions about changing schools may also be correlated with observable as well as
unobservable family and student characteristics that also influence student achievement. Second,
there is considerable heterogeneity among mobile students by the type and timing of student
mobility. Structural movers differ from non-structural movers and between-year non-structural
movers differ from during the year non-structural movers. Third, mobility impacts students in
different ways. In other words, mobility effects may vary by the student’s race/ethnicity, FRPL,
ELL and special education statuses. Finally, the circumstances that resulted in students changing
schools, whether positive or negative (for e.g., job loss or gain, marriage or divorce) may also
directly influence student achievement. It is plausible that the reasons students switch schools
such as family disruptions may also impact student achievement even if the student did not
change schools.
The empirical analysis consists of two parts. First, I use a student fixed effects model as
the primary method to estimate the impact of student mobility on mobile students’ achievement.
In addition to estimating the impact of changing schools, I also use the student fixed effects
model to disentangle the net effect of student mobility into the effect associated with changes in
school quality and the transition costs of moving or effect associated with disruption. The fixed
effects strategy largely addresses three of the four empirical challenges. First, the fixed effects
strategy overcomes the problem of selection bias by excluding non-mobile students from the
83
estimation of mobility effects. In order to mitigate the non-randomness of student mobility, the
impact of student mobility is identified by comparing mobile students to themselves before and
after the school change. Thus, the effects of mobility are not based on movers versus non-movers
comparisons and thus not susceptible to selection bias associated with the pre-existing
differences between mobile and non-mobile students. Student fixed effects also controls for
observable and unobservable student characteristics that do not change over time that may be
associated with the type of mobility and academic achievement. However, unobservable time
varying student characteristics are not accounted for and may still bias mobility effects. The
second challenge was addressed by separating the structural from non-structural moves and
differentiating non-structural moves by the timing of school changes. The third challenge was
addressed by estimating conditional effects, or mobility effects by students’ race/ethnicity and
income. The fourth challenge was left unaddressed and estimates from student fixed effects
model may be biased by the circumstances prompting student mobility that may have a direct
impact on student achievement.
I use propensity score matching (PSM) as a secondary strategy to estimate the effect of
changing schools on mobile students’ achievement. Another way to conceptualize the effect of
student mobility is to consider the counterfactual question: if mobile students had stayed in the
same school, what would their achievement be? In an ideal experiment, similar students would
be randomly assigned to the treatment of changing schools thus the estimation of the impact of
changing schools would include non-mobile students. It is also plausible that estimates
comparing mobile to non-mobile students may differ from estimates of comparing mobile
students to themselves. PSM addresses all of the four empirical challenges but in incomplete
ways. First, PSM overcomes selection bias by comparing mobile students to a control group of
84
non-mobile students with similar propensity to move but who did not change schools. This
replicates a mobility experiment by including non-mobile students in the estimation of mobility
effects. However, PSM only controls for selection based on observable student and school
characteristics. PSM does not account for the biases based on unobservable student
characteristics that may impact both student mobility and student achievement. The second
challenge is addressed in a similar fashion to fixed effects and the third challenge is addressed by
restrictive matching conditions that match students within cohort, gender, race and income. The
fourth challenge is addressed in some ways using PSM. If the effect of the reasons for mobility is
strongly correlated with students’ observable characteristics, then the matching process may
account for some of the bias of the reason of mobility related to students’ observable
characteristics. However, estimates from PSM are susceptible to bias if the reasons for mobility
are related to unobservable student characteristics.
Overall, each method addresses selection bias and the role of unobservable characteristics
in the estimation of mobility effects using different identification strategies. Student fixed effects
address selection bias and the role of unobservable student characteristics by comparing mobile
students to themselves but the estimates are susceptible to remaining biases from the impact of
the reasons for mobility and time varying unobservable student characteristics. PSM
approximates a mobility experiment but estimates may also have some remaining bias from
selection due to unobservable characteristics. Comparing the estimates of both methods may
provide useful insights on the biases that are accounted for as well as remaining bias in mobility
effects. Specifically, PSM may offer insights on the extent to which fixed effects estimates are
biased by the reason for mobility that are related to students’ observable characteristics and
85
whether the comparison used in the estimation of mobility effects (mobile student to themselves
vs. mobile-non-mobile students) leads to different estimates.
Similar to previous mobility studies, I focus on and present results for mathematics
achievement as math is predominantly learned in school rather than the home (especially starting
in the elementary years) and mobility effects may be more detectable using math as opposed to
reading (Hanushek et al., 2004; Raudenbush et al., 2011; Rumberger et al., 1999; Xu et al.,
2009). I categorize non-structural movers by the timing of school changes: between-year
switcher or a student who made a non-structural move between school years; within-year
switcher or a student who switched schools at least once during the school year and; “ultra-
mover” or a student who changed schools both between and during the school year in the same
academic year. I also focus on elementary and middle schools (grades 3-8) because students do
not take annual standardized tests in high schools and multiple test observations are needed for
models examining students’ yearly achievement growth.
Student fixed effects. In order to estimate the impact of changing schools on mobile
students’ achievement, I start with the following generalized fixed effect model:
𝑌 𝑖𝑡
= 𝑀 𝑖𝑡
𝛼 + 𝑿 𝒊𝒕
𝜷 + 𝛿 𝑖 + 𝜖 𝑖𝑡
( 1 )
Where the outcome, 𝑌 𝑖𝑡
is a student’s math achievement growth (Y
it
– Y
it-1
), standardized
relative to district. 𝑀 𝑖𝑡
is an indicator variable of student mobility indicating whether student i
changed schools in year t. I run separate models for the aforementioned categories of
nonstructural movers. 𝑿 𝒊𝒕
is a vector of time varying student characteristics (FRPL, ELL and
special education status). 𝛿 𝑖 is a student fixed effect which accounts for time-invariant
86
unobservable characteristics. 𝛼 is the coefficient of interest and can be interpreted as the impact
of changing schools that is identified by changes over time in the achievement gains of mobile
students. In this model, 𝛼 includes the changes in school quality (Tiebout effect) and the
transition costs of moving.
In order to disentangle the changes in school quality from the transition costs of moving,
I estimate the following model:
𝑌 𝑖𝑡
= 𝑀 𝑖 𝑡 𝛼 + 𝑚 𝑖𝑡
𝛼 ′
+ 𝑿 𝒊𝒕
𝜷 + 𝛿 𝑖 + 𝜖 𝑖𝑡
( 2 )
Similar to equation (1), the outcome 𝑌 𝑖𝑡
is the change in annual math learning or achievement
growth for student i in year t and 𝑀 𝑖𝑡
indicates whether a student changed schools. The
additional term 𝑚 𝑖𝑡
indicates student mobility prior to the summer before the current school
year. 𝑚 𝑖𝑡
=1 if 𝑠 𝑡 = 𝑠 𝑡 − 1
and either 𝑠 𝑡 is not equal to 𝑠 𝑡 − 2
or 𝑚 𝑖𝑡 − 1
; and otherwise 𝑚 𝑖𝑡
=0.
When 𝑀 𝑖𝑡
=1,𝑚 𝑖𝑡
= 0 indicating a student who moved but remained in the same school for at
least a year following the move. 𝛼 ′ is the average difference in learning between two academic
years – the year after a student changes school and the year before the move. Hanushek et al.
(2004) argues that 𝛼 ′ is determined by differences in school quality (Hanushek et al., 2004).
31
𝛼
can be considered the gross temporary effect of student mobility on the rate of learning and 𝛼 ′
can be viewed as the long term or steady state change in the rate of learning following a school
31
Mobility effects are identified through the change in achievement gains over time for mobile students or by
comparing the academic performance before and after a school change for the same student. The identification
strategy is to compare mobile students to themselves using multiple test score observations to separate the changes
in school quality. Thus, the model requires students to have at least three observations within the panel. The
identification of the changes in school quality is based on three main assumptions: a) assimilation and disruption
costs are incurred in the first school year or the year of the move, after which mobile students are treated as
incumbent students; b) on average, students do not experience temporary losses in the year prior to the school
change; c) students recover to pre-school change achievement growth rates in the years following the year of the
school change (Hanushek et al., 2004). I provide evidence testing these assumptions in the results.
87
change. In this model, 𝛼 − 𝛼 ′
is the transitory disruption effect of student mobility. These
disruptions include both school and non-school related factors in the adjustment to student
mobility such as assimilation and getting accustomed to a new school environment that is
assumed to occur in the first school year.
Propensity score matching. I use PSM to create a control group and then estimate the
differences in achievement between mobile students and students in the comparison groups.
Although PSM cannot account for selection on unobservable characteristics, it can mitigate the
influence of unobservable characteristics by restrictions on matches and if observed
characteristics are correlated and capture many of the salient unobserved characteristics. In order
to mitigate selection on unobservable characteristics, I impose a few restrictive conditions on the
matching techniques. I limit the pool of eligible comparison non-mobile students to students in
the same cohort, gender, race/gender, grade and year that did not change schools over the period
of study. I also match within district as opposed to within school.
32
Propensity score matching
provides “treatment on the treated estimates,” or the impact of changing schools on mobile
students. I conduct separate analyses by the timing of school changes.
32
In addition to the impact of changing schools on movers, student mobility may also affect non-mobile students.
These mobility externalities or spillover effects arise as high student turnover introduces discontinuities in the
learning environment of all students (both mobile and non-mobile) and may disrupt teaching and curriculum
development (Alexander et al., 1996; Hanushek et al., 2004; Institute of Medicine and National Research Council,
2010; Kerbow, 1996; Raudenbush et al., 2011; Reynolds et al., 2009; Rumberger et al., 1999; Rumberger, 2003).
Thus, non-mobile students in schools with high student turnover may also be affected by student mobility.
Hanushek and colleagues (2004) presented evidence of a negative externality of student mobility or considerable
costs borne by non-mobile students and highlighted that the effect of student mobility on non-mobile students is
greater for low-income and minority students who generally enroll in schools with high mobility rates (Hanushek et
al., 2004). Raudenbush et al. (2011) found similar results and effect size with respect to school-level mobility
(about a -0.03SD) as well as evidence of grade-level and cumulative effects associated with the within year in-
migration rate (Raudenbush et al., 2011). In addition, South and colleagues (2007) found that non-mobile students
in high mobility schools have higher dropout rates, lower levels of school attachment and lower academic
performance (South et al., 2007). Matching within schools raises concerns about the lack of independence across
treated and control units, a central underlying assumption of propensity score matching. Without data at the
classroom level, one cannot distinguish whether mobile and non-mobile students at the same school in the same
grade were classmates.
88
First, I create the propensity score or estimate the likelihood of making a non-structural
move within the school district (the treatment) based upon observable characteristics in the year
prior to student mobility using the following conditional logit model:
𝑌 𝑖𝑡
= 𝛽 + 𝑿 𝒊𝒕
𝜷 + 𝜖 𝑖𝑡
( 3 )
Where 𝑌 𝑖𝑡
is a dichotomous indicator of whether student i switched schools in time t. I
run separate models for the different categories of nonstructural movers. 𝑿 𝒊𝒕
is a vector of student
characteristics (demographic and achievement). I use a set of rich student background variables
to estimate the propensity score of changing schools including: lagged math test scores (relative
to the district); demographic variables (race/ethnicity, gender, FRPL status, special education and
ELL status); interactions of test scores and all demographic variables; interactions of gender and
race/ethnicity; interactions of special education status and race/ethnicity; interactions of FRPL
status and race/ethnicity; and interactions of ELL and race/ethnicity. For the K-5 cohort, I
exclude students who changed schools in grade 3 or prior untested grades (K-2). Thus, I compare
a mobile student that changed schools only once in elementary school in either grade 4 or grade
5 to a similar student who did not switch elementary schools. For the 3-8 cohort, I exclude
students who made non-structural changes in grade 3. Thus, I compare a mobile student that
changed schools only once in elementary and middle schools (assuming students who did not
change in tested elementary grades also did not switch schools in K-2) in grades 4 through 8 with
a student that did not make a non-structural move in either elementary or middle school. Thus,
89
prior mobility history and the frequency of mobility are taken into account when forming the
matched comparison groups across both cohorts.
33
Next, I match a mobile student with a student with similar baseline characteristics who
did not change schools (non-mobile student) but had a similar propensity to change schools. I use
nearest neighbor matching (one-to-one) with replacement and a bandwidth of 0.001.
34
I then
examine the balance of the treatment and comparison groups by testing for significant
differences in baseline test scores and demographic characteristics between the two groups in
their baseline test scores and demographic variables. In the results, I provide evidence of the
balance of matched samples. In specification analyses detailed later in this paper, I use alternate
matching techniques to test the robustness of PSM.
After creating the treatment and control groups, I conduct an ordinary least squares
(OLS) regression to improve precision and account for any lingering differences between the
matched comparison group. I use the following model to estimate the effect of changing schools:
∆ 𝑌 𝑖𝑠 𝑡 = 𝑿 𝒊 𝒔 𝒕
𝜷 + 𝑴 𝒊 𝒔 𝒕
𝜷 + 𝒁 𝒔𝒕
𝜷 + 𝜖 𝑖𝑡
Where ∆ 𝑌 𝑖𝑠 𝑡 is the difference in math test score before and after changing schools of
student i in school s at time t ( 𝑌 𝑖𝑠 𝑡 − 𝑌 𝑖𝑠 𝑡 − 1
). 𝑋 𝑖𝑠 𝑡
is a vector of student characteristics including
33
In the extant literature, little attention has been paid to the possible effects of student mobility in untested grades
on the estimates of changing schools in elementary and middle schools. Most mobility studies focus on grades in
elementary and middle schools given that students are tested annually and achievement can be tracked over time.
Typically grade 3 is the first tested grade and often the first included year. However, school changes in early
untested grades may have small lingering effects on achievement (Burkam et al., 2009). Therefore, accounting for
mobility in these untested grades is important in estimating the impact of mobility on mobile students’ educational
outcomes. I account for changing schools in untested grades by excluding these students from the PSM sample and
including indicator for mobility in these grades in specification checks of student fixed effects models.
34
Given the systematic differences between mobile and non-mobile students, it may be hard to find good close
matches. It is reasonable that the population of non-mobile students with similar propensity to move may be limited,
thus matching with replacement may provide closer matches. Matching with replacement reduces the distance
between the treatments and controls but may affect efficiency. I match without replacement in specification checks.
90
gender, educational status (FRPL, ELL and special education) and race/ethnicity. 𝑀 𝑖𝑠 𝑡
is an
indicator of whether student i in school t switched to a new school in time t. In additional
specifications, I include 𝒁 𝒔𝒕
a vector of school-level characteristics (origin school) such as the
percentage of proficient, Black, Hispanic, Asian, FRPL, ELL and special education students. I
use robust standard errors clustered at the school-level.
Results
Before describing the empirical results, I provide a brief overview of the rates of student
mobility and the characteristics of mobile students across the timing of school changes in the
empirical sample. Table 1 summarizes student mobility among the grades in the K-5 and 3-8
cohorts. On average, 14 percent of students in the K-5 cohort made non-structural moves
between school years, 7 percent made mid-year moves and 3 percent switched schools both
between and during the school year in the same academic year (ultra-movers). Interestingly, total
non-structural mobility (regardless of the timing of school changes) was higher in untested
grades (K-2) than in the tested grades (3-5).
35
Of the K-5 cohort, 38 percent did not make a non-
structural move over the period of study, 31 percent made one non-structural move, 17 percent
made two non-structural moves and about 14 percent made three or more non-structural moves
35
Student mobility in untested grades (K-2) in elementary schools was prevalent in CCSD. About half of all non-
structural moves in elementary schools occurred in untested grades (K-2). In fact, the grades with the highest total
nonstructural student mobility (regardless of the timing of school changes) were grades one, two and three
respectively. During the year mobility (mid-year moves as well as ultra-moves) decreased as grades increased thus
mid-year moves were more prevalent in untested grades relative to grades 3-5. Similarly, the rate of between year
mobility was highest in grade 1. Overall, in the K-5 cohort, student mobility in untested grades is higher than that of
tested grades across the timing of school changes. There was also a notable trend of repeat movers among untested
grades: about half of the nonstructural movers that switched schools in kindergarten also switched schools in either
grade 1 or 2 and roughly a third of non-structural movers in grade 1 also switched schools in grade 2. Students who
changed schools in untested grades also changed schools in tested grades. For instance, about 39 percent of
kindergarteners that changed schools also switched schools in grade 3, 34 percent switched in grade 4 and 30
percent switched in grade 5. Movers in grades 1 and 2 demonstrated similar trends of switching in tested grades.
Overall, the majority of students that switched schools in tested grades (57 percent) previously changed schools in
untested grades.
91
between 2007-08 and 2012-13. In the 3-8 cohort, on average, 11 percent of students changed
schools between school years, 7 percent changed within the school year and 2 percent changed
schools both between and during the school year. Of the 3-8 cohort, 51 percent did not make a
non-structural move over the period of study, 28 percent made one nonstructural move, 13
percent moved twice and roughly 8 percent of students moved three or more times.
[Insert Table 1 around here]
Table 2 presents demographic and achievement characteristics of mobile students by the
timing of school changes for both cohorts. Non-structural movers vary by the timing of school
changes. For both cohorts, between year movers were disproportionately Black and low-income
students performing below their school and the district averages. Ultra-movers appear to be a
more disadvantaged subset of within-year movers and had the lowest achievement among all
categories of mobile students. These disproportionately low-income and Black students were
also some of the lowest achieving students performing between a third and half a standard
deviation below average in their schools and district.
[Insert Table 2 around here]
Mobile students also differ by the timing of school changes. The results indicate that
students who changed schools between school years in the K-5 cohort were significantly
different from mid-year movers across all demographic and achievement characteristics except
gender and special education status. In the 3-8 cohort, between and mid-year movers differed
across all characteristics except special education status. There were notable variations in the
differences between ultra-movers and mid-year movers across cohorts. In the K-5 cohort, mid-
year and ultra-movers had a handful of statistically significant differences (only Asian, low-
income, math achievement relative to district and math proficiency) whereas in the 3-8 cohort
92
mid- year and ultra-movers had statistically significant differences across all characteristics
except Hispanic, White and gender. In the next section, I present estimates of the impact of
student mobility across the timing of school changes using a student fixed effects model.
Student Fixed Effects Model
Table 3 shows estimates of the impact of changing schools on mobile students’
achievement in mathematics for both cohorts. Column (1) provides cross-sectional estimates
based on achievement level in year t. This estimation of the impact of student mobility on
student achievement levels in the same academic year that the student switched schools is
prevalent in many earlier mobility studies that do not include student fixed effects and compared
movers to non-movers. Column (2) provides a simple annual gain model with student fixed
effects that does not distinguish the immediate costs of moving from subsequent changes in
school quality. More recent mobility studies using panel data adopt a similar approach and
provide estimates of the impact of changing schools on students’ achievement gains. These
estimates include possible confounding influences that prompt student mobility such as family
shocks and incorporates both the changes in school quality and the transition costs of moving.
Column (3) disentangles the long-term school quality effect from the short–term disruptive
impact of switching schools by considering the time pattern of achievement gains related to
school changes. 𝛼 can be considered as the gross temporary effect of changing schools on the
rate of learning. 𝛼 ′ is the average “Tiebout effect” or the steady state change in the rate of
learning following a move that can be viewed as the long run impact of moving attributed to the
change in school quality that movers experience. 𝛼 - 𝛼 ′ represents the transitory disruption
costs or the sum disruptive effects including the loss of school specific capital, assimilation
93
efforts, other costs of disruption and changes in family circumstances that may accompany a
school change.
[Insert Table 3 around here]
When all non-structural school changes are combined (regardless of the timing of student
mobility), on average, movers had math achievement levels that were significantly lower than
the achievement of non-movers (-0.26SD in the K-5 cohort and -0.32SD in the 3-8 cohort).
However, movers differ from non-movers in statistically significant ways, thus these estimates
cannot be interpreted as the causal impact of changing schools. In the value-added specification
in Column (2), the estimates of the impact of changing schools on achievement growth was
markedly lower and lost statistical significance. As Column (3) illustrates, the average “Tiebout
effect” was also insignificant for both cohorts.
36
In addition, the average transaction costs were
also insignificant. The results are qualitatively similar to those found by Hanushek et al. (2004).
When one compares movers to themselves over time and all nonstructural moves were lumped
together, the results imply student mobility in elementary and middle schools may lead to a small
negative decline on math achievement gains. Both studies found no evidence of changes in
school quality when all within-district non-structural moves were combined. However, Hanushek
et al. (2004) found that within district switchers experienced a significant -0.24SD transaction
cost of moving whereas this study finds much lower and insignificant transaction costs when all
moves are lumped together.
Table 3 illustrates that the overall impact of student mobility, as well as the effects
attributable to transactions costs and changes in school quality, varies with the timing of school
changes. As Column (2) illustrates, for the K-5 cohort, between-year school changes were
36
The school quality decomposition (estimates in Column (3) for Cohort K-5) is based solely on grade 4 school
changes or a student who did not change schools in grade 3 or 5.
94
associated with a 0.05SD increase in mobile students’ math achievement growth (for the 3-8
cohort, the coefficient was positive, 0.01SD, but insignificant). The average transactions costs
and the changes in school quality for between school year moves were small and insignificant for
both cohorts. The results suggest that student mobility in the summer may be especially
beneficial for mobile students in elementary schools relative to higher grades. The findings also
imply that the achievement gains experienced by between year movers in elementary schools are
largely due to low transitions costs of moving rather than systematic positive changes in school
quality.
Column (2) also demonstrates that switching schools during the academic year was
associated with a -0.08SD decline in math achievement growth for students in both cohorts. The
average transaction costs for during the year moves were large and statistically significant for
both cohorts (-0.10SD for the K-5 cohort and -0.06SD for the 3-8 cohort). There was no
evidence of significant changes in school quality for mid-year movers in both cohorts.
Unsurprisingly, the results imply that student mobility during the year requires greater
adjustments that may adversely affect achievement compared to between year school changes.
“Ultra-movers” or students who switched schools both between and during the school
year in the same academic year also experienced a statistically significant decline in math
achievement gains in both cohorts (-0.08SD and -0.07SD for the K-5 and 3-8 cohorts
respectively). Interestingly, the average transaction costs were large but significant for “ultra-
movers” in K-5 cohorts. In the 3-8 cohort, transactions costs were -0.08SD. There is also little
evidence to suggest that ultra-movers were systematically switching to higher quality schools in
both cohorts.
95
In sum, the results imply that the impact of student mobility depends on the timing of
school changes.
37
The components of the net impact of changing schools, disruptive and Tiebout
effects, also vary across the timing of school changes. The results suggest that mobility effects
are largely driven by the transactions cost of moving. However, the estimates of the transitory
disruption also include changes in family circumstances that may be associated with changing
schools. Thus, the higher and more significant average transactions costs of mid-year and ultra-
movers relative to between-year movers may also reflect the possibility that students moving
during the school year may experience a higher rate of negative shocks such as job loss or
divorce that are correlated with changing schools. Therefore, the reasons for mobility may be a
driving factor in the variation in the transition cost of moving and thus mobility effects across the
timing of school changes.
The results imply between year moves may be likely related to positive circumstances for
reasons for mobility given the small insignificant transaction costs of moving. In this case, the
reasons for mobility may upwardly bias estimates as the family disruptions, especially in earlier
grades, may create positive changes in a students’ learning environment that benefit student
achievement even though there were no significant changes in school quality. To be sure, there
may also be a positive independent effect associated with changing schools in the summer that
allows students to assimilate more effectively and reduce the transition costs of student mobility.
Conversely, the significant large negative transactions costs of moving for during the year
mobility imply that the reasons for mobility for mid-year moves may downwardly bias mobility
37
I also examine whether mobility effects are significantly different across the timing of school changes by
comparing the estimates of between-year movers, within-year movers and ultra-movers. The results indicate that for
both cohorts, moves during the academic year (mid-year as well as ultra-moves) had a significantly greater adverse
effect than school changes between academic years. This is the opposite of prior research that found that between-
year moves were no worse than mid-year moves (Grigg, 2012; Hanushek et al., 2004). However, ultra-moves did
not have a significantly more adverse effect than mid-year school changes.
96
effects. It is plausible that the negative circumstances associated with changing school during the
school year such as job loss may be captured in the disruptive effects. There may also be a
negative independent effect associated with changing schools during the school year that
increases the transition costs of student mobility.
Even though there are interesting differences in the impact of student mobility on
students’ achievement gains across timing of school changes, there may also be further
differences by student characteristics. In the next section, I examine the conditional effects of
student mobility and whether there are further differences in transactions costs and changes in
school quality by students’ race/ethnicity, income and educational status across the timing of
school changes.
Conditional effects. In order to investigate how mobility effects vary across mobile
students’ characteristics, I compute separate estimates by race/ethnicity, income, gender and
educational status across the timing of school changes. Tables 4 present the estimates of effect of
student mobility for low income, special education and ELL students in the K-5 and 3-8 cohorts
across the timing of school changes. Table 4 indicates mobility effects vary by the timing of
school changes for low-income students. Mid-year FRPL recipients had a -0.08SD decline in
math achievement growth. Low-income mid-year movers also had considerable transactions
costs (-0.07SD for both cohorts). Table 4 also demonstrates that special education students in
both cohorts had mostly insignificant mobility effects with the exception of ultra-movers in the
3-8 cohort. Special education ultra-movers had a -0.09SD decline in achievement gains and -
0.12SD transactions costs. Mid-year ELL movers in the 3-8 cohort had a -0.08SD decline in
math achievement growth. These students had significant negative transition costs of moving (-
0.11SD). ELL ultra-movers in the 3-8 cohorts had a similar decline in math achievement gains
97
and significant transaction costs (-0.10SD). Regardless of the timing of school changes, there
was little evidence of changes in school quality for low-income, ELL and special education
students.
[Insert Table 4 around here]
Tables 5 present the estimates of effect of student mobility for Black, Hispanic and
Male students in the K-5 and 3-8 cohorts across the timing of school changes. Table 5 illustrates
that the impact of student mobility for Black students, especially those in elementary schools,
vary with the timing of school changes. There is suggestive evidence that Black students who
switched schools in the summer had positive achievement gains and low transactions costs of
moving across both cohorts, however, the coefficients are insignificant. Conversely, mid-year
black movers in the K-5 cohort experienced a -0.08SD decline in math achievement growth and
considerable negative transactions cost (-0.10SD). Similarly, black ultra-movers in the 3-8 cohort
had a substantial decline in achievement gains (-0.10SD) and transactions cost (-0.16SD).
[Insert Table 5 around here]
Table 5 also demonstrates that the impact of student mobility on Hispanic students also
varies with the timing of school changes. Hispanic students in the 3-8 cohort had a -0.06SD
decline for during the year moves. Hispanic mid-year movers in the 3-8 cohort also had
significant negative transactions costs (-0.08SD). Hispanic ultra-movers in both cohorts had
significant negative achievement gains (-0.09SD in K-5 cohort, -0.04SD in the 3-8 cohort) and
substantial transactions cost (-0.12SD in the K-5 cohort and -0.07SD in the 3-8 cohort). The
results suggest that the impact of mobility for male students also depends on the timing of school
changes, especially for school changes in elementary schools. For male students in both cohorts,
mid-year moves were associated with negative achievement gains (-0.05SD in the K-5 cohort
98
and -0.06SD in the 3-8 cohort). Male mid-year movers also had high transactions, -0.06SD and -
0.08SD for the K-5 and 3-8 cohort respectively. Similarly, male ultra-movers in the 3-8 cohort
had negative achievement gains (-0.06SD) and transactions costs (-0.10SD).
Overall, the results imply that student mobility is most harmful for students who switch
schools during the academic year. Mid-year mobility is most impactful for low-income,
Hispanic, male and ELL students (and Black students in elementary grades) who experienced
substantial losses in achievement growth. These students also had some of the highest
transactions costs of moving. In addition, ultra-movers experienced negative achievement gains
regardless of income or educational status. The results confirm the heterogeneity among the
mobile student population and demonstrate considerable variation in the transactions costs and
changes in school quality across the timing of school changes. Similar to the findings of
Hanushek and colleagues (2004), transactions costs vary by students’ race/ethnicity and income
with low-income and minority students experiencing larger average transitions costs of moving.
Transactions costs also vary across the timing of school changes for students with similar
characteristics. Student mobility effects in this section were identified by comparing mobile
students to themselves over time and using the time pattern of achievement gains. In the next
section, I present estimates of the impact of changing schools using propensity score matching
(PSM) that compares mobile students to similar non-mobile students.
Propensity Score Matching
Before presenting estimates of the impact of student mobility on mobile students’ math
achievement, I briefly describe the balance of matched samples for both cohorts across the
timing of school changes. Table 6 provides a sense of whether PSM ‘works’ by testing how
‘good’ the matched samples are across the timing of school changes or how well matching has
99
eliminated or reduced the association between student mobility and the covariates in the K-5 and
3-8 cohorts respectively. The table summarizes two checks for evaluating the balance of matched
samples: testing the average difference in covariates between the matched samples and
examining standardized bias, or the difference in the means of the treatment and control group
for each covariate in standard deviation units. There should be no statistically significant
differences in the observable characteristics between students in the matched samples. It is also
suggested that the overall balance or average standardized bias across all covariates should be
less than five percent after matching (Rosenbaum & Rubin, 1985). The balance of individual
covariates for matched samples across the timing of school changes in both cohorts is available
upon request.
[Insert Table 6 around here]
In the final restricted K-5 cohort sample, there were 15,434 unique students: 12,755 did
not switch elementary schools and 2,679 changed schools once in either grade 4 or 5. In the final
restricted 3-8 cohort sample, there were 23,728 unique students: 15,314 did not make a non-
structural school change in either elementary or middle schools and 8,414 made a non-structural
move in middle school in grades 4 through 8. Table 6 shows that few of the covariates had
statistically significant differences suggesting a balance between the treatment and control group
in the matched samples across the timing of school changes for both cohorts. Few covariates had
standardized bias values over five percent in the matched samples across both cohorts. The
average standardized bias ranged between 1.4 and 2.6 in the K-5 cohort and between 0.6 and 0.9
for the 3-8 cohort. All are below the recommended maximum value. When one compares the
change in the standardized bias between the raw and matched samples across the timing of
school changes in both cohorts, the results suggest that PSM reduced a considerable proportion
100
of bias. In essence, evaluating the balance of the matched samples provides evidence that the
models have been correctly specified and can be considered reasonably valid.
Table 7 illustrates the estimates of the impact of student mobility on students’ math
achievement levels and growth using propensity score matching. Column (1) presents the effect
of making a non-structural move on math achievement levels, whereas Column (2) shows the
effect of changing schools on math achievement growth. In the K-5 cohort, all the coefficients
were insignificant but the magnitude and direction of the coefficients are suggestive. Changing
schools was associated with a negative effect across the timing of school changes but the
coefficients were relatively larger for school changes that occurred during the school year. Table
7 also illustrates that overall there were few differences in results across the timing of school
changes between the two cohorts. In the 3-8 cohort, when all moves are combined, student
mobility was associated with negative effect on math achievement. Between year moves in the 3-
8 cohort was associated with a -0.06SD decline in math achievement growth. During the year
mobility was associated with a larger negative effect on math achievement growth (-0.11SD).
Ultra-movers had large, negative but insignificant effects on achievement across both cohorts.
[Insert Table 7 around here]
Similar to the estimates from the student fixed effects model, the results from PSM also
imply that the impact of student mobility on achievement varies with the timing of school
changes. The results suggest that mid-year moves typically have a large, significant adverse
effect on math achievement whereas between year moves result in smaller, insignificant effects
on achievement. Ultra-movers had insignificant effects implying that selection on observables
account for a considerable portion of estimated mobility effects for this category of movers. On
average, the estimates from both methods are qualitatively similar. This suggests that estimates
101
from students fixed effects models may be capturing an independent effect of changing schools
or selection on unobservables, especially for mid-year mobility. Given that estimates from the
student fixed effects model account for the time-invariant observable characteristics of mobile
students, it is reasonable that changing schools during the school year has a significant negative
impact on math achievement growth independent of the circumstances prompting the school
change.
Specification Checks and Limitations
Fixed effects. The first set of specification checks investigates the assumptions under
which the changes in school quality can be disentangled from the disruptions accompanying
student mobility. The identification strategy that attributes the change in achievement growth
patterns to changes in school quality rests on three assumptions (Hanushek et al., 2004). First,
there should be no temporary loss in achievement the year prior to the move. Second, without
any changes in school quality, students generally recover to pre-move achievement growth rates
in the years succeeding a school change. Third, disruptions accompanying the move that affect
achievement persist for only the year of the move. In order to test the first assumption, I include
a dummy variable that identifies changes in achievement growth in the year prior to a move. To
examine the validity of the second and third assumption, I separate 𝑚 𝑖 𝑡 into an indicator for a
move one year prior and a move two years prior.
Table 8 presents the average changes in achievement gains in the year prior to the move,
the year of the move, the year following the move and two years following the move for the 3-8
cohort (I did not run for the K-5 cohort as there were too few observations). The coefficient for
achievement growth differences on the year prior to the move was insignificant across the timing
of school changes. Although there was some variation in achievement gains in the years
102
following the move, the coefficients were insignificant. However, there was suggestive evidence
of a pre-move dip for mid-year and ultra- movers. This implies estimates of the changes in
school quality would be biased upward for during the year mobility. Interestingly, Hanushek et
al. (2004) also found evidence of a significant pre-move dip for movers within a district. It is
important to note that these specification checks cannot account for constant shocks
accompanying a move that persist over time and may confound the estimated changes in school
quality. Overall, the results suggest that changes in average school quality may be a reasonable
explanation for the differences in the time patterns of achievement gains.
[Insert Table 8 around here]
The second set of specification checks examine whether mobility effects may be sensitive
to the inclusion of certain subpopulations in the empirical sample such as discipline-related
mobility and students capitalizing on open enrollment options in the district.
In particular,
mobility to and from magnet and charter schools or discipline-related mobility may influence
mobility effects. I re-run the analysis separately without magnet and charter schools to test
whether the results are shaped by open enrollment options. In both instances, the results were
qualitatively similar. Next, I exclude discipline-related mobility and the results remained largely
the same except for mid-year mobility in the 3-8 cohort. In particular, the effect of changing
schools during the academic year on math achievement growth declined from -0.08SD to -
0.05SD when discipline-related mobility was excluded from the sample. The effect for ultra-
movers was also reduced from -0.08SD to -0.06SD. In both cases, the transactions costs of
moving were lower but still significant. This suggests that discipline-related mobility may be an
important driver of the adverse effects of mid-year mobility in urban school districts and raises
questions on the effectiveness of school discipline policies in improving student achievement.
103
Finally, I restrict the sample to continuously enrolled students that have been in a CCSD school
for the duration of the panel and exclude entrants, leavers and students who left and returned to
the district. The results are qualitatively similar.
Propensity score matching. The first set of specification checks use alternate matching
techniques as the estimates of the effect of student mobility may be sensitive to the matching
procedure used. Tables 9 and 10 provide various estimates from the different matching methods
for the K-5 and 3-8 cohort respectively. I employ various techniques including nearest neighbor
with replacement and stratification matching. Overall, the results are qualitatively similar across
the various matching techniques across both cohorts. Indeed, as the table shows, controlling for
student characteristics significantly reduces the estimates and in some cases the statistical
significance. However, these regressions do not account for possible confounding variables that
may affect student mobility and students’ achievement levels. In addition, I also compare the
achievement differences between the treatment and comparison groups without covariate
adjustments. The results are qualitatively similar.
[Insert Table 9 and 10 around here]
Limitations. This paper cannot make causal claims as mobile students were not
randomly assigned to schools. In addition to omitted variable bias, the reasons for mobility
remain the concerning source of bias. Similar to prior research on student mobility, this analysis
is limited by data availability and there is no information on the circumstances that precipitated
the change in schools (e.g., job loss). I do not have information to control for the process of
assigning students to schools and I cannot account for the possible sorting into neighborhoods.
Thus, there are remaining biases linked to the reasons for mobility and unobservable
characteristics that are not accounted for in the estimates of each method. While each method
104
addresses selection bias, no method fully accounts for a time varying shock that prompted
student mobility that may also directly influence student achievement. This may be captured in
the fixed effects estimates as part of the reason for mobility (transaction cost) and it may be
linked to unobservable characteristics that are not captured by PSM.
A popular critique of using propensity score matching is that unobservable characteristics
that are not captured may bias the results. In this case, movers may vary from non-mobile
students in ways that are measured as well as ways that are unmeasured. However, I argue that
the sample restrictions coupled with matching on students’ prior test scores and other covariates
may reasonably capture students’ unobservable characteristics. Moreover, PSM in conjunction
with OLS have been previously used by recent studies in the charter school literature to estimate
the impact of schools and have replicated experimental estimates in some instances (Bifulco,
2010; Cook, Shadish, & Wong, 2008; Fortson, 2012; Furgeson et al., 2012; Tuttle, Gill, &
Gleason, 2013; Zimmer, Gill, Attridge, Obenauf, & Lansing, n.d.). Notwithstanding, although
multiple specifications of exhaustive as well as parsimonious models are used to predict
propensity scores, estimates from PSM may be biased by omitted or unmeasured covariates that
are highly correlated with student mobility and student achievement.
Another limitation with the PSM analysis in this study is the limited power due to sample
restrictions and conditions for comparison groups. In the K-5 cohort, by excluding mobile
students who switched schools in K-3 (untested grades and the first year with achievement data
for PSM), the size of the treatment group was considerably reduced. Moreover, given the sample
restrictions, it is likely that treatment groups largely consist of students in CCSD with relatively
lower propensities to move. This may partly explain the lack of significance in several PSM
estimates. It may also limit the external validity of the PSM estimates given it captures only a
105
small slice of the notably heterogeneous mobile student population. Nonetheless, even though
the coefficients on the impact of student mobility are mostly statistically insignificant, the
direction and magnitude provide some insights.
Discussion
In this paper, I estimate the impact of intra-district student mobility on the math
achievement of mobile students and disentangle the net effect of changing schools across the
timing of school changes. The results highlight the complexities involved in estimating the
impact of student mobility in an urban school district. Two key takeaways may be distilled from
the findings. First, the impact of student mobility varies with the timing of school changes.
However, the relationship between the timing of school changes and transactions costs of
moving as well as the differences in mobile students also suggest that the variations in mobility
effects across the timing of school changes may reflect differences in the reasons for mobility.
The results also imply that even though the reasons for mobility may likely bias estimates from
quasi-experimental methods, there may still be an independent adverse effect (especially for
mid-year moves) associated with changing schools that vary with the timing of school changes.
Second, the disruptive effect of student mobility appears to be the driving factor of mobility
effects. It is plausible that minimizing transactions costs coupled with equivalent transfers in
school quality may also result in positive achievement gains from changing schools.
The results imply that the impact of student mobility depends on the timing of school
changes. Between year moves have a positive impact on math and little insignificant costs of
moving. During the year mobility has a negative impact on math and large significant costs of
moving. However, the timing of school changes appears to also be driven in part by the reasons
106
for mobility. It is difficult to distill the independent effect of the timing of school changes from
the impact of timing attributable to the reasons for mobility. The results suggest that estimates
may be a mixture of both an independent effect of changing schools and possible effects
attributed to the reasons for mobility. In addition, this study also illustrates the importance of
differentiating the timing of school changes when estimating the mobility effects. Overall, when
all moves are combined, mobility effects tend to be negative and significant. Differentiating
student mobility by the timing of school changes illuminates important variations and paints a
more nuanced picture of mobility effects. Moreover, mobility studies that lump all moves
together mask crucial variation in transactions costs among mobile students by the timing of
school changes.
The results also suggest that transactions costs may be an equally important determinant
of the net effect of student mobility as changes in school quality. Mobile students typically have
no systematic positive changes in school quality. The results suggest that some school changes
may lead to an improvement in achievement not because of changes in school quality but rather
because of low and insignificant costs of moving coupled with equivalent school quality. Thus,
minimizing transactions costs associated with changing schools may lead to positive mobility
effects provided students switch to schools of similar quality.
Policy Implications
The results highlight a few critical policy implications. First, states and districts should
focus on reducing the transactions costs of student mobility especially for low-achieving mobile
students that switch schools during the academic year. Given the prevalence of student mobility
within an urban school district, policymakers should explore creative ways to help reduce the
possible school-related element of disruption associated with changing schools. Policymakers
107
should develop and refine “what works” strategies to ease the transition of mid-year movers.
These may include facilitating the creation of teacher materials for mobile students or
implementing a peer buddy system within schools and the district. Assisting schools and teachers
that experience high mobility rates to prepare for the turnover in students may result in
mitigating and reducing the disruptive effects associated with student mobility. Providing
additional financial and curriculum support and resources for schools with higher mid-year
mobility rates may also help mitigate the transactions costs of student mobility. Overall,
establishing a clear and effective framework for acclimating mobile students within urban school
districts will be even more important as student mobility rates (especially mid-year) persist and
increase over time.
Second, greater coordination of education policy and other social and economic policy is
needed to comprehensively address the effects of changing schools. The results imply that the
reasons for mobility may play a significant role in the timing of school changes and mobility
effects. Transactions costs may also reflect factors outside the schools’ purview such as housing
or employment stability that influence both student mobility and student achievement. However,
given the link between residential mobility and student mobility as well as the role of social
relationships in the costs and benefits of moving, it compels policymakers to also focus on
relational connections and social support. In essence, student mobility represents the intersection
of schooling and society. Thus, policymakers should recognize that student mobility is a broader
social issue and craft a commensurate response. This may include more detailed employment and
housing information for mobile students that enables tracking of changes in the lives of mid-year
movers. Districts may also consider assigning low-achieving, low-income and minority mobile
students to case workers that can navigate and coordinate social services across multiple
108
agencies to help reduce the harmful effects that may be attributable to the reasons for mobility.
In sum, there is a need for policies that are systematic and that recognize the collective
circumstances that shape the individual achievement of mobile students.
Third, the results raise questions about policies that encourage student mobility such as
school choice policies. Similar to previous studies, this paper finds little evidence of students
systematically switching to higher quality schools. Indeed, it is probable that changing schools
will be accompanied by costs and adjustments. However, the likelihood of transferring to a
higher quality school is low and is often an unrealized possibility for most mobile students. Thus,
the costs are almost sure but the benefits of changing schools generally only accrue to relatively
few students who switch to better schools. Although the results highlight the importance of
minimizing transactions costs, switching to better schools is not mutually exclusive. The results
compel policymakers to address the challenges families face when choosing schools and the
likely disruption associated with changing schools as policies that encourage student mobility are
given greater consideration.
Directions for Future Research
There are also a few areas in which future research may make important contributions.
First, a better understanding of why students change schools and the timing of school changes is
needed. The results suggest that the last rung on the causal ladder is accounting for family shocks
that result in student mobility. This may be explored by linking survey data on the reasons for
moving with achievement data for mobile students. Additionally, mixed methods and qualitative
studies of student mobility are necessary to unpack the complicated mobility landscape. Mobility
studies may also incorporate neighborhood characteristics, school location and distance between
school and home to provide additional insights on the reasons for mobility.
109
Second, examining the specific nature of disruption costs may also provide valuable
insights on the impact of changing schools. Disruption associated with student mobility may
involve several possible elements such as loss of instructional time when switching schools or
acclimation to a new school environment. More research is needed to delve into this black box
and gain a more granular understanding of what aspects of changing schools generate the
greatest transition costs. It would also be useful to better understand how transition and
disruption costs are affected by the quality of the origin as well as the destination school. For
instance, some schools may have certain resources or provide a type of environment that reduces
the transition costs of moving. There may also be important variations in schools’ policies and
practices within an urban district to acclimatize incoming students as well as manage exiting
students that also warrant greater attention. Finally, valuable insights can be garnered by
exploring empirical strategies that further disentangle the transactions costs into school related
disruptions, non- school-related disruptions and disruptions related to the circumstances
prompting student mobility.
Finally, a better understanding of the impact of student mobility on non-cognitive
educational outcomes across the timing of school changes is also needed. Non-cognitive
outcomes such as students’ attitude and behavior may also play equally important roles as
cognitive outcomes such as test scores in families’ school choice. Most student mobility studies
measure school quality using schools’ achievement levels. Nevertheless, the conceptualization of
school quality based solely on achievement scores may be inadequate as parents may define
school quality in complicated ways and test scores may fail to capture a range of other attributes
mobile families may find appealing in a school. In addition, future research may also examine
the effect of student mobility on postsecondary and labor market outcomes such as attending and
110
completing college or the likelihood of being arrested. Hence, more robust measures of school
quality and examining a wider range of possible educational outcomes that student mobility may
affect can help to generate new knowledge about how families choose schools and how changing
schools impact students’ overall learning and developmental progress.
Conclusion
This study contributes to a stronger consensus on the impact of student mobility across
the timing of school changes. The results illustrate that the estimation of mobility effects is
complex and demonstrate compelling heterogeneity in the mobile student population within a
large urban school district. The reasons for mobility and the many contributing factors to student
achievement remain challenges that may introduce bias into estimates of changing schools
derived from quasi-experimental methods. Nevertheless, the estimates derived in this paper takes
a considerable step by differentiating whether school changes occurred during the academic year
or between school years and using two different identification strategies. It offers useful insights
on the biases accounted for and the remaining biases that may affect mobility estimates.
111
Chapter Two Tables and Figures
Figure 1.A model of the key relationships among student mobility and mobile students’ educational outcomes.
School
District
Neighborhood
School
Student
Mobile
Students’
Educational
Outcomes
Student
Mobility
School-related and
non-school-related
circumstances
Adjusting to a new school:
Social Capital (Peer effects at school
level)
Ecological Model of Human
Development (School, Classroom and
Teacher Effects from different
curriculum, teachers and expectations)
Adjusting to a new neighborhood:
Neighborhood Effects and
Social Capital (Peer effects at
neighborhood level)
School Quality (Tiebout effect)
112
Table 1
Mobility rates by the timing of school changes and grades across cohorts, 2007-08 – 2012-13
Cohort K-5 (N=30,973) Cohort 3-8 (N=18235)
BY switcher WY switcher Ultra-mover BY switcher WY switcher Ultra-mover
Grade K N/A 0.11 N/A
(0.31)
Grade 1 0.16 0.07 0.03
(0.37) (0.26) (0.18)
Grade 2 0.15 0.07 0.03
(0.35) (0.25) (0.17)
Grade 3 0.16 0.06 0.03 N/A 0.10 N/A
(0.36) (0.24) (0.16) (0.30)
Grade 4 0.12 0.06 0.02 0.13 0.07 0.02
(0.32) (0.24) (0.14) (0.34) (0.26) (0.15)
Grade 5 0.11 0.05 0.02 0.13 0.06 0.03
(0.31) (0.22) (0.14) (0.34) (0.24) (0.16)
Grade 6 N/A 0.08 N/A
(0.27)
Grade 7 0.09 0.07 0.02
(0.28) (0.25) (0.15)
Grade 8 0.08 0.06 0.02
(0.27) (0.24) (0.14)
Note. BY= Between school years, WY= Within the school year .Ultra-mover refers to student that changed schools both between and within the school year in
the same academic year. Standard deviations in parentheses.
113
Table 2
Demographic and achievement characteristics of mobile students by the timing of school changes, 2007-08 – 2012-13
Cohort K-5 Cohort 3-8
BY switcher WY switcher Ultra-mover Cohort Avg. BY switcher WY switcher Ultra-mover Cohort
Avg.
Black 0.17 0.22 0.27 0.13 0.19 0.23 0.26 0.13
(0.37) (0.41) (0.44) (0.33) (0.39) (0.42) (0.44) (0.34)
Hispanic 0.44 0.49 0.46 0.44 0.43 0.47 0.45 0.43
(0.50) (0.50) (0.50) (0.50) (0.49) (0.50) (0.50) (0.49)
Asian 0.07 0.06 0.05 0.08 0.08 0.05 0.04 0.08
(0.25) (0.24) (0.21) (0.27) (0.27) (0.23) (0.21) (0.27)
White 0.28 0.20 0.18 0.31 0.27 0.21 0.20 0.32
(0.45) (0.40) (0.38) (0.46) (0.44) (0.40) (0.40) (0.47)
Male 0.52 0.52 0.51 0.51 0.52 0.56 0.56 0.51
(0.50) (0.50) (0.50) (0.50) (0.50) (0.50) (0.50) (0.50)
Special Education 0.14 0.12 0.14 0.11 0.13 0.14 0.16 0.11
(0.34) (0.32) (0.35) (0.32) (0.34) (0.34) (0.37) (0.32)
English Language Learner 0.24 0.29 0.25 0.25 0.13 0.20 0.17 0.15
(0.42) (0.46) (0.44) (0.44) (0.34) (0.40) (0.38) (0.35)
FRPL 0.61 0.72 0.78 0.56 0.62 0.73 0.76 0.54
(0.47) (0.45) (0.41) (0.50) (0.48) (0.44) (0.42) (0.50)
Math Achievement
(school)
-0.10 -0.34 -0.38 0.007 -0.14 -0.36 -0.44 0.003
(1.00) (1.02) (1.02) (0.99) (0.99) (1.03) (1.03) (1.00)
Math Achievement
(district)
-0.12 -0.39 -0.50 0.006 -0.17 -0.48 -0.60 0.002
(0.97) (0.99) (0.99) (1.00) (0.98) (0.99) (1.02) (1.00)
Math proficiency 0.66 0.55 0.51 0.72 0.56 0.43 0.39 0.62
(0.47) (0.50) (0.50) (0.46) (0.50) (0.50) (0.49) (0.49)
N 17,311 10,401 3,209 148,010 10,651 11,101 2,323 149,059
Note. BY= Between school years, WY= Within the school year .Ultra-mover refers to student that changed schools both between and within the school year in
the same academic year. Math achievement is the z score of students relative to their school or the district. For the K-5 cohort, achievement statistics is based
on tested grades in the cohort (3-5) (N=51,715). Standard deviations in parentheses. **- p <0.001, * - p<0.05.
114
Table 3
Estimated effects of nonstructural student mobility on mathematics achievement by timing of school changes: Student fixed effects
model
Cohort K-5 Cohort 3-8
(1) (2) (3) (1) (2) (3)
All non-structural moves
combined
𝛼 -0.26*** -0.005 0.02 -0.32*** -0.003 -0.01
(0.01) (0.02) (0.05) (0.02) (0.01) (0.02)
𝛼 ′ 0.03 -0.02
(0.06) (0.01)
Between School Years 𝛼 -0.13*** 0.05** 0.04 -0.14*** 0.01 0.02
(0.01) (0.02) (0.01) (0.02) (0.01) (0.02)
𝛼 ′ -0.02 0.005
(0.05) (0.02)
During the School year 𝛼 -0.35*** -0.08*** -0.01 -0.39*** -0.08*** -0.09***
(0.02) (0.02) (0.06) (0.02) (0.01) (0.02)
𝛼 ′ 0.09 -0.03
(0.07) (0.005)
Between & During the School
year
𝛼 -0.40*** -0.08* -0.02 -0.46*** -0.07** -0.05
(0.03) (0.04) (0.09) (0.04) (0.02) (0.03)
𝛼 ′ 0.10 0.03
(0.11) (0.02)
Note: Robust standard errors in parentheses; *** p<0.001, ** p<0.01, * p<0.05. All specifications include indicators for free and reduced price lunch eligibility,
ELL and special education status. Standard errors are clustered by schools. Column 2 and 3 includes student fixed effects.
115
Table 4
Estimated effects of nonstructural student mobility on mathematics achievement by timing of move and students’ income and
educational status
Cohort K-5 Cohort 3-8
FRPL Special
Ed
ELL FRPL Special
Ed
ELL
(1) (2) (1) (2) (1) (2) (1) (2) (1) (2) (1) (2)
Between
School Years
𝛼 0.03 0.03 0.05 0.04 0.09 0.07 0.02 0.02 0.009 0.008 0.05 0.05
(0.01) (0.01) (0.03) (0.03) (0.03) (0.10) (0.01) (0.01) (0.03) (0.03) (0.03) (0.03)
𝛼 ′ -0.01 -0.03 -0.04 -0.001 -0.003 -0.007
(0.008) (0.02) (0.01) (0.007) (0.02) (0.03)
During the
School year
𝛼 -0.06** -0.06** -0.06 -0.06 -0.05 -0.05 -0.06*** -0.06*** -0.002 -0.002 -0.08** -0.08**
(0.02) (0.02) (0.04) (0.04) (0.03) (0.03) (0.01) (0.03) (0.03) (0.03) (0.03) (0.03)
𝛼 ′ 0.009 0.03 -0.007 0.01* 0.001 0.03
(0.008) (0.02) (0.02) (0.006) (0.06) (0.07)
Between &
During the
School year
𝛼 -0.04 -0.03 0.06 0.06 -0.06 -0.06 -0.05** -0.04** -0.09* -0.09* -0.08* -0.08*
(0.02) (0.10) (0.07) (0.07) (0.04) (0.04) (0.02) (0.02)
(0.04) (0.04) (0.04) (0.04)
𝛼 ′ 0.03 -0.0009 0.04 0.04 0.03 0.02
(0.01) (0.03) (0.03) (0.01) (0.02) (0.02)
Note: Robust standard errors in parentheses; *** p<0.001, ** p<0.01, * p<0.05. Standard errors are clustered by schools. All specifications include student fixed
effects.
116
Table 5
Estimated effects of nonstructural student mobility on mathematics achievement by timing of move and students’ gender and
race/ethnicity
Cohort K-5 Cohort 3-8
Black Hispanic Male Black Hispanic Male
(1) (2) (1) (2) (1) (2) (1) (2) (1) (2) (1) (2)
Between School
Years
𝛼 0.08** 0.08** 0.03 0.03 0.06 0.05 0.03 0.03 0.03* 0.03* 0.01 0.01
(0.03) (0.03) (0.02) (0.02) (0.02) (0.06) (0.02) (0.02) (0.02) (0.02) (0.01) (0.01)
𝛼 ′ 0.005 -0.02 -0.001 -0.01 -0.001 -0.004
(0.02) (0.008) (0.009) (0.01) (0.009) (0.007)
During the School
year
𝛼 -0.08** -0.08* -0.03 -0.04 -0.05* -0.05* -0.03 -0.03 -0.07*** -0.06*** -0.06*** -0.06***
(0.03) (0.03) (0.02) (0.09) (0.02) (0.02) (0.02) (0.03) (0.01) (0.01) (0.01) (0.01)
𝛼 ′ 0.02 0.0004 0.007 0.004 0.02 0.02
(0.02) (0.009) (0.01) (0.01) (0.006) (0.007)
Between & During
the School year
𝛼 0.003 0.003 -0.09** -0.09** -0.04 -0.04 -0.11*** -0.10** -0.04* -0.04* -0.06*** -0.06***
(0.05) (0.05) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.02) (0.02) (0.02) (0.02)
𝛼 ′ -0.001 0.03 0.02 0.06 0.03 0.04
(0.02) (0.02) (0.02) (0.02) (0.01) (0.01)
Note: Robust standard errors in parentheses; *** p<0.001, ** p<0.01, * p<0.05. Standard errors are clustered by schools. All specifications include student fixed
effects.
117
Table 6
Summary of matched samples across the timing of school changes
Cohort K-5 Cohort 3-8
Average
Standardized
Bias
% of balance
covariates
Average
Standardized Bias
% of balance
covariates
All moves combined 1.4 100% 0.6 100%
Between year move 1.4 100% 0.7 100%
During the year move 2.6 100% 0.7 100%
Ultra mover 2.4 100% 0.9 100%
Note. There are 36 covariates in all the specifications.
118
Table 7
Estimated effects of non-structural student mobility on mathematics achievement by timing of school changes: Propensity score
matching
Cohort K-5 Cohort 3-8
(1) (2) (1) (2)
All moves combined -0.06 -0.05 -0.09* -0.08**
(0.05) (0.03) (0.04) (0.03)
Between year move -0.01 0.02 -0.08 -0.06*
(0.06) (0.04) (0.04) (0.03)
During the year move -0.06 -0.03 -0.10* -0.11***
(0.07) (0.05) (0.04) (0.03)
Ultra mover -0.13 -0.13 -0.09 -0.09
(0.12) (0.09) (0.06) (0.05)
Note. Robust standard errors in parentheses; *** p<0.001, ** p<0.01, * p<0.05.
119
Table 8
Estimates of the change in achievement gains prior to and following a move by the timing of school changes
All moves
combined
BY WY Ultra-moves
Year prior to move 0.002 -0.03 0.02 0.03
(0.02) (0.03) (0.02) (0.05)
Year of move -0.007 -0.03 -0.05* -0.04
(0.03) (0.04) (0.02) (0.05)
One year following move -0.04 0.04 -0.008 -0.06
(0.03) (0.04) (0.03) (0.04)
Two years following move -0.02 -0.05 -0.009 0.01
(0.02) (0.03) (0.03) (0.03)
Note. *** p<0.001, ** p<0.01, * p<0.05; BY= Between school years, WY= Within the school year .Ultra-mover refers to student that changed schools both
between and within the school year in the same academic year.
120
Table 9
Estimated effects of nonstructural student mobility on mathematics achievement by timing of school changes using several propensity
score matching methods: K-5 cohort
All moves combined
Between-Year Within-Year Ultra-mover
Estimation Method (1) (2) (1) (2) (1) (2) (1) (2)
Nearest Neighbor matching -0.03 -0.04* -0.01 -0.02 -0.09* -0.08** -0.14 -0.13
(0.02) (0.02) (0.03) (0.02) (0.05) (0.03) (0.08) (0.07)
Nearest neighbor without replacement -0.04 -0.03 -0.04 0.01 -0.06 -0.05 -0.14 -0.13
(0.03) (0.02) (0.03) (0.02) (0.06) (0.04) (0.11) (0.08)
Stratification Matching -0.04* -0.04*** -0.008 -0.007 -0.11*** -0.08*** -0.18*** -0.13
(0.02) (0.02) (0.02) (0.02) (0.04) (0.03) (0.07) (0.06)
Main strategy without covariate
adjustment (nearest neighbor with
replacement and a caliper of 0.001)
-0.06 -0.05 -0.02 0.06 -0.05 -0.04 -0.12 -0.11
(0.05) (0.03) (0.05) (0.03) (0.07) (0.05) (0.12) (0.08)
Note. *** p<0.001, ** p<0.01, * p<0.05; Two calipers were used: a caliper of .2 of the standard deviation of the logit propensity score and a caliper of 0.001. The
results are qualitatively similar and only results using the caliper of 0.001 are presented. Bootstrap standard errors are reported where available.
121
Table 10
Estimated effects of nonstructural student mobility on mathematics achievement by timing of school changes using several propensity
score matching methods: 3-8 cohort
All moves combined
Between-Year Within-Year Ultra-mover
Estimation Method (1) (2) (1) (2) (1) (2) (1) (2)
Nearest Neighbor matching -0.07*** -0.07** -0.02 -0.03* -0.12*** -0.12*** -0.13*** -0.12***
(0.01) (0.009) (0.02) (0.01) (0.02) (0.01) (0.05) (0.04)
Nearest neighbor without replacement -0.08*** -0.07** -0.05* -0.04* -0.14*** -0.13*** -0.13* -0.12**
(0.02) (0.01) (0.02) (0.01) (0.03) (0.02) (0.06) (0.05)
Stratification Matching -0.08*** -0.07*** -0.03* -0.03** -0.12*** -0.12*** -0.15*** -0.12***
(0.009) (0.007) (0.01) (0.01) (0.02) (0.01) (0.04) (0.03)
Main strategy without covariate
adjustment (nearest neighbor with
replacement and a caliper of 0.001)
-0.08** -0.07*** -0.05 -0.03 -0.12** -0.12** -0.16* -0.14**
(0.03) (0.02) (0.04) (0.02) (0.03) (0.02) (0.07) (0.05)
Note. *** p<0.001, ** p<0.01, * p<0.05; Two calipers were used: a caliper of .2 of the standard deviation of the logit propensity score and a caliper of 0.001. The
results are qualitatively similar and only results using the caliper of 0.001 are presented. Bootstrap standard errors are reported where available.
122
CHAPTER THREE
STUDENT MOBILITY, SEGREGATION AND ACHIEVEMENT GAPS
Abstract
The prevalence of student mobility and school segregation are two important issues facing urban
school districts that are often addressed in separate conversations. Nevertheless, the relationship
between both phenomena may have important equity implications given evidence of differential
mobility patterns that may lead to a more segmented school system. This paper provides a
descriptive analysis of racial, income and achievement segregation in Clark County, Nevada and
examines the relationship between student mobility and school segregation within a large urban
school district. The results indicate that racial segregation was low to moderate as unevenness
increased and isolation decreased over the period of study. Hispanic students were the most
segregated racial/ethnic group in the school district. Income segregation was also moderate but
income segregation decreased over time. Similarly, racial achievement gaps increased and the
income achievement gap declined over time. The results suggest that more segregated schools
typically have smaller within-school achievement gaps, a lower proportion of proficient students
and higher non-structural mobility rates (especially during the year mobility) than less segregated
schools. This implies there is clustering of low-achieving students from different ethnicities in
segregated schools, which may not be beneficial for student achievement in urban school
districts. The results also indicate that students in intensely segregated minority schools transfer
to less segregated schools, regardless of the timing of school changes. Overall, the findings
provide suggestive evidence that student mobility may contribute to desegregation in both
positive and negative ways. Policy implications and directions for future research are discussed.
123
STUDENT MOBILITY, SEGREGATION AND ACHIEVEMENT GAPS
Student mobility is pervasive across the U.S and is especially prevalent in urban school
districts (Institute of Medicine and National Research Council, 2010; United States Government
Accountability Office, 2010). The majority of the extant literature on student mobility has
examined changing schools from the students’ perspective and focus primarily on how student
mobility affects student achievement (Institute of Medicine and National Research Council,
2010; Reynolds et al., 2009). However, student mobility may have other consequences within an
urban school district outside of mobile students’ achievement. Few studies consider student
mobility from the perspective of schools and districts. In light of evidence of differential
mobility patterns that imply changing schools may lead to unintended consequences over time,
such as increasing segmentation of student populations by students’ backgrounds, achievement,
or school quality (Kerbow, 1996; Welsh, Duque & Mceachin, forthcoming), it is important to
examine how student segmentation has changed over time by race/ethnicity and achievement
levels in urban school districts. The lack of a formal definition of segmentation makes it difficult
for one to determine how differentiated an educational system has to be to label it as
“segmented.” Nevertheless, a better understanding of the relationship between student mobility,
school segregation and achievement gaps over time in urban districts may offer valuable insights
about the equity implications of student mobility.
The relationship between segregation and student mobility is an important but overlooked
issue in urban districts. Student mobility and segregation share several interesting commonalities
that make examining the relationship between the two phenomena worthwhile. Student mobility
is widespread and segregation is rising. Schools have been re-segregated in the decades since the
124
1954 landmark Brown v. Board of Education case ruling (Frankenberg, Lee, & Orfield, 2003;
Orfield & Yun, 1999; Orfield, 2001). Both phenomena concern equality of educational
opportunity. Segregation in schools and urban districts is a prominent educational equity issue
(Orfield, 1983). The Coleman report published in 1966 highlighted the prevalence of school
segregation in the US and its adverse effects on the equality of educational opportunity and
students’ educational outcomes (Coleman, Campbell, & Hobson, 1966). Student mobility and
segregation is especially concerning in urban school districts in light of ongoing demographic
shifts. Changing demographics have had adverse effects on desegregation, especially the influx
of minority students in urban school districts (Bifulco & Ladd, 2007; Frankenberg et al., 2003;
Orfield & Lee, 2007). Moreover, court decisions in recent decades have made it more
challenging for districts to maintain integrated schools (Condron, Tope, Steidl, Freeman, &
Colleges, 2013; Orfield & Lee, 2007). Low-income and minority students are disproportionately
affected by both student mobility and segregation. The persistence and rise of segregation is
particularly concerning given evidence of the positive influence of desegregation on educational
and labor market outcomes of minority as well as non-minority students (Johnson, 2011;
Kurlaender & Yun, 2001; Wells & Crain, 1994). More importantly, student mobility and racial
income and segregation may also partly explain the achievement gap (Bifulco & Ladd, 2007;
Hanushek et al., 2004).
Notwithstanding, the relationship between segregation and student mobility is complex.
Both may be a cause and consequence of each other. The sorting of students between schools
over time may play a critical role in determining how segregation within schools and a school
district changes over time. It is also plausible that the demographic composition of schools may
be a role in families’ decision to switch schools. Racial and income groups may be segregated in
125
a myriad of ways and students may change schools at different times and with varying
frequencies. Both student mobility and segregation may be largely influenced by out-of-school
factors and represent the intersection of society and schooling. Nevertheless, school quality plays
a major role in determining the net impact of both processes on educational outcomes. More
importantly, student mobility and segregation can be manipulated and influenced through
education policy. However, both represent policy challenges and there is much to learn about
how these phenomena interact and how both affect education equity and students’ outcomes in
urban school districts.
Surprisingly, student mobility and school segregation are often addressed in different
silos. The majority of the extant literature tends to focus on segregation in school choice contexts
or states with districts with court-ordered desegregation plans (Bifulco & Ladd, 2007; Condron
et al., 2013; Johnson, 2011). It is important to learn more about the relationship between student
mobility, school segregation and test score gaps such as whether mobile students are transferring
to more segregated schools and/or making racially segregating moves. The relationship between
student mobility and school segregation may also vary with the timing of school changes. There
is a need for research that explores the relationship between intra-district student mobility and
segregation across the timing of school changes within ‘traditional’ urban school districts over
time.
This paper examines the relationship between student mobility, segregation and
achievement gaps within a school district over time. Using student-level data from 2007-08
through to the 2012-13 school year, this study employs multiple measures of segregation
including dissimilarity and isolation indices to provide a descriptive analysis of racial, income
and achievement segregation. The analysis moves beyond the black-white comparisons and
126
includes several racial and income groups to reflect the multiethnic nature of an urban school
district. This is one of the first studies to examine the relationship between intra-district student
mobility, segregation and achievement gaps. Specifically, I ask the following research
questions:
a.) To what extent have racial, income and achievement segregation, and achievement
gaps changed over time?
b.) How does the relationship between segregation and student mobility vary across the
timing of school changes?
c.) Are mobile students switching to more segregated schools?
The results indicate that racial segregation was moderate but increasing in CCSD.
Overall, racial achievement gaps increased as unevenness increased and isolation decreased over
the period of study. Hispanic students were the most segregated racial/ethnic group in the school
district. Income segregation was also moderate, however, income segregation and the income
achievement gap declined over time. Achievement segregation was high and increased over time.
The results suggest that more segregated schools typically have smaller within-school
achievement gaps, a lower proportion of proficient students and higher non-structural mobility
rates (especially during the year mobility) than less segregated schools. There is also evidence
suggesting that student mobility may contribute positively to desegregation as students in
intensely segregated minority schools transfer to less segregated schools, regardless of the timing
of school changes. The rest of the paper proceeds as follows. I first provide a brief overview of
the literature on segregation, achievement gaps and student mobility. Following this, I describe
127
the data and methodological approach employed in this study. Next, I present results and
conclude with a discussion of policy implications and directions for future research.
Segregation, Achievement Gaps and Student Mobility
Segregation refers to the physical separation of different racial and ethnic groups. Even
though the Brown v. Board decision resulted in the desegregation of schools in the 1970s, there
has been persistent segregation at substantial levels across the U.S. (Frankenberg et al., 2003;
Orfield & Yun, 1999; Orfield, 1983). Racial segregation across schools within an urban school
district is significantly higher than racial segregation within schools (Conger, 2005). Although a
handful of studies have investigated the impact of racial segregation within schools (Conger,
2005; Mickelson, 2001), the vast majority of research literature on segregation has examined the
extent of segregation or whether students of different racial, ethnic and income profiles attend
different schools (Logan, Minca, & Adar, 2012). Most of the research has focused on racial
segregation at the district-level (between-schools) use school-level data and focus on the Black-
White dichotomy even though Asian and Hispanic students account for an increasing part in the
racial composition of the U.S student population (Frankenberg et al., 2003; Orfield & Lee,
2007). Researchers also conceptualize and measure segregation in a myriad of ways.
Nevertheless, the most oft used indicator is a measure of the proportion of minority students in a
school which may not accurately capture segregation between groups within a district (Condron
et al., 2013). There is also a growing body of research evaluating the effect of racial segregation
on student and school performance (Bifulco & Ladd, 2007; Logan et al., 2012).
School segregation separates children and stratifies the type of school they attend, leaving
minority children in inferior schools (Orfield & Yun, 1999). Black and Hispanic students are
128
concentrated in high poverty schools compared to White and Asian students (Orfield & Lee,
2005). Orfield and Lee (2005) also found that Black or Hispanic students are more likely to
attend urban schools (Orfield & Lee, 2005). Although White students are the most racially
isolated racial/ethnic group, segregation is rising for African American and Latino students
(Frankenberg et al., 2003). Overall, numerous studies indicate that racial composition has direct
and independent effects on student performance (Logan et al., 2012).
One of the main reasons why segregation is widely regarded as a critical issue in
education policy is its presumed relationship with the achievement gap. However, the
relationship between segregation and achievement gaps is understudied (Condron et al., 2013).
An achievement gap is defined as the differences or disparities in students’ educational outcomes
such as test scores between different groups of students. The achievement gap between black and
white students is an important component of black/white economic inequality (Condron et al.,
2013; Jencks & Phillips, 2011). Over the past half century, minority students (Black and
Hispanic) have performed considerably better than their peers in the past and this has contributed
to a narrowing of the achievement gap with White students whose achievement has remained
fairly constant over time (Berends, 2014). Unlike racial achievement gaps, the socioeconomic or
income achievement gap has remained sizable and has been growing over time (Berends, 2014;
Reardon, 2011). Condron and colleagues (2013) also found that black-white dissimilarity and
black student isolation partly explain black-white achievement gaps (Condron et al., 2013).
Overall, the majority of studies have found that desegregation is helpful for students of
all races. Coleman and colleagues (1966) found a negative association between the concentration
of poverty within a school and student performance that has been confirmed by several studies in
recent decades (Coleman et al., 1966). Racial isolation of minorities in majority-minority schools
129
concentrations are associated with lower academic achievement and inferior educational
opportunities (Coleman et al., 1966; Logan et al., 2012) Johnson (2011) found desegregation’s
impact on racial equality to be deep, wide, and long-lasting (Johnson, 2011). Black Americans
who attended schools integrated by court order were more likely to graduate, go on to college,
and earn a degree than black Americans who attended segregated schools (Johnson, 2011).
Desegregation also had a positive impact on labor market and other lifestyle outcomes (Johnson,
2011). Using panel data from Texas, Hanushek and Rivkin (2009) found that a higher percentage
of black schoolmates reduced achievement for African Americans and segregation levels
explained a notable portion of the racial achievement gap (Hanushek & Rivkin, 2009).
There are several possible ways that desegregation may impact students and schools.
Presumably, the central impact of desegregation comes from the peers of students’ or the peer
effect. Simply put, it is advantageous to attend a school where students are more successful
(Hanushek, Kain, Markman, & Rivkin, 2003). Peer effects are not the only consideration as
school context and characteristics may also be crucial factors. Segregated schools typically are
unequally resourced (for e.g., class sizes, school facilities, teacher quality and curriculum), thus
attending such schools may adversely affect achievement, especially for low-income and
minority students (Condron et al., 2013). Johnson (2011) posits improvement in access to school
resources as one of the mechanisms through which desegregation benefits students (Johnson,
2011).
However, the majority of the extant literature on student mobility suggests that the reason
for changing schools may be driven more by out-of-school factors such as residential changes
rather than preferences for school quality. It is also plausible that parents navigate tradeoffs such
as higher-achieving peers versus well-matched peers (Duflo, 2011). Moreover, student mobility
130
may lead to stratification within a school district as students of different achievement levels and
racial and income groups are increasingly unevenly distributed within a district and have less
interactions with each other.
This study explores the ways in which student mobility and segregation overlap to affect
students, schools and districts. I also compare the patterns of student mobility, segregation and
achievement gaps at the district, school and student-level. This descriptive analysis of the
relationships between segregation, student mobility and achievement gaps is the first step toward
simultaneously investigating multiple explanations of achievement gap in order to learn more
about the causes of educational disparities. This paper provides a view into the key but
understudied relationship between student mobility and segregation. In the next section, I
describe the data and methodological approach employed in this study.
Data and Methods
Data
I use a six year panel of student-level data for all students in the CCSD from 2007-08
through to 2012-13.
38
The data contains students’ demographic characteristics and annual test
scores from the Nevada Proficiency Examination Program. Demographic data includes
indicators for students’ gender, race/ethnicity (Black, Hispanic, Asian, White), free and reduced
priced lunch (FRPL), English Language Learner (ELL) and special education statuses. Students
are tested in reading and math in grades 3-8 and take the High School Proficiency Exam (HSPE)
38
As of 2012-13, there are 357 schools in CCSD (217 elementary schools, 59 middle schools, 49 high schools, 24
alternative schools, and 8 special schools). I have data on some of the alternative schools operated by CCSD
including: 5 behavior, 3 continuation schools, 4 juvenile detention centers and 6 adult education schools. I also have
data on all of the 25 CCSD-run magnet schools and career and technical academies that provide parents with school
choice and 9 Prime 6 schools (The Prime 6 Initiative was first adopted in 1994 in an effort to provide support for 9
schools located in West Las Vegas that serve a predominantly low-income African American and Hispanic
population). I only have data for the charter schools in CCSD for the 2012-13 school year.
131
in grade 10. I standardize test scores for students in grades 3 through 10 by grade and year,
relative to the school mean, as well as relative to the district mean.
39
I complement the student-
level data with school-level data on school locations and facilities including zip code and the
year the school was built in addition to publicly available accountability data.
40
Detailed longitudinal data that tracks the dates and sequence of school changes allows for
in-depth classification of the timing of student mobility across a range of grades (K-12). Unique
student and school identifiers in the data link students to schools in each year and across multiple
school years. I assume that all school changes between school years in grade 6 and 9 are
transitions from elementary to middle and middle to high schools respectively, with the
exception of students enrolled in combination schools, of which there are relatively few.
41
The
reasons for enrollment and withdrawal, as reported by schools, such as whether the student
previously attended a school in another state or the student left a school to transfer to private or
charter schools in the district are also available.
39
I use the annual means and standard deviations on the test scores for each grade and year to standardize student
achievement. Standardizing the outcomes makes it possible to compare students' test scores over time, as well as
across grades and provides insight into how a student’s math and reading test scores are compared to students in the
same grade in the school and in the district. If negative relative to the school mean, this indicates that the student in
the school received a math or reading test score below the school average.
40
I obtained Adequate Yearly Progress (AYP) designation data from 2007-08 through to 2011-12 from the Nevada
Department of Education (ayp.nv.gov) and nevadareportcard.com. Schools are designated by the Nevada
Department of Education as: exemplary, continuing exemplary, exemplary turnaround, high achieving, adequate, on
watch list and in need of improvement. In compliance with NCLB, Nevada AYP classifications are made annually
based on the percentage of students tested, the percentage of students tested who score at or above the proficient
level on annual statewide tests, and school attendance or graduation rates.
41
I exclude students who were transported out of the district (students who live in Clark County but attend schools
in another district) and remove all within-school Americans with Disabilities Act (ADA) changes. There are a
number of students who attended two schools at the same time (typically a high school and a technical school). For
these concurrent enrollments, I only use mobility associated with comprehensive high schools and ignore the other
schools. Xu et al. (2009) and Burkam et al. (2009) highlight the importance of including grade retention variables in
mobility studies. For students who were held back in the same grade in the same school (grade retention) or were
accelerated a grade in the same school, I identify those specific moves as possible controls in the analysis. However,
there was limited grade retention or skipping grades in CCSD over the period of study with less than two percent of
students and the majority of changes occurring in grades 9-12.
132
I use two main analytical samples. For the descriptive analysis, I use yearly cross-sections of all
K-12 students. My primary unit of analysis is the student-year yielding 1,946,446 student-years
with 548,523 unique students.
42
Methods
Similar to previous mobility studies, I focus on and present results for mathematics
achievement as math is predominantly learned in school rather than the home (especially starting
in the elementary years) and mobility effects may be more detectable using math as opposed to
reading (Hanushek et al., 2004; Raudenbush et al., 2011; Rumberger et al., 1999; Xu et al.,
2009). I categorize non-structural movers by the timing of school changes: between-year
switcher or a student who made a non-structural move between school years; within-year
switcher or a student who switched schools at least once during the school year and; “ultra-
mover” or a student who changed schools both between and during the school year in the same
academic year.
In order to examine student mobility at the school-level, I focus on the percent of students
leaving each school or the average school turnover across the timing of non-structural school
changes. In addition, I measure school quality by the percentage of proficient students and
include other school-level characteristics including the percentage of Black, Hispanic, Asian,
FRPL, ELL, Special Education, and male students in the school. Finally, I identify racially
segregating school transfers, or a non-structural school transfer in which a student switched to a
school that is greater than 60 percent nonwhite and more than 10 percentage points more
nonwhite than their previous school (I run separate analyses for Black and Hispanic instead of
percentage nonwhite).
42
I include students in untested grades (K-2, 9, 11, 12) that do not have achievement data. If this sample is further
restricted to students with achievement data, there are 985,805 student-years with 205,123 unique students.
133
Segregation
I use dissimilarity and isolation indices to evaluate segregation between schools in CCSD
by race/ethnicity, income and achievement over time. Each index measures different dimensions
of segregation: the dissimilarity index captures unevenness or the distribution of racial groups
and the isolation index captures the extent of isolation or potential contact between racial groups
(Massey & Denton, 1988). Thus, I characterize segregation along two of the five dimensions
classified by Massey and Denton (1988): unevenness and exposure. The combination of these
multiple measures of segregation provides a detailed picture of the distribution of students in
schools within the district as well as inter-racial contact. I calculate the dissimilarity and isolation
indices for multiple combinations of four racial categories (Black, White, Asian and Hispanic),
one income category (FRPL students) and two achievement categories (whether the student was
below math in the district or proficient in math).
The dissimilarity index measures what percentage of the racial group’s population would
need to change schools for the racial groups to be evenly distributed within the school district.
Generally, a dissimilarity index below .3 is low segregation, between .3 and .6 is moderate
segregation and above .6 is high segregation (Massey & Denton, 1988). I calculate the
dissimilarity index using the following formula:
DI
d 𝑡 = ½ ∑ | (a 𝑠 𝑡
/ A
dt
) – (b 𝑠 𝑡
/ B
d 𝑡 ) | (1)
Where DI
d 𝑡 is the dissimilarity index of district d at time t. a 𝑠 t
is the number of “a”
students in school s at time t and A
d 𝑡 is the number of “a” students in all schools in district d at
134
time t. b 𝑠 𝑡 is the number of “b” students in school s at time t. B
d
is the number of “b” students in
all schools in district d at time t.
Racial groups being disproportionately located in particular schools does not equate to
little or no contact between the groups. It is also useful to know whether students in a certain
racial group have a high degree of possible interaction with other racial groups. The isolation
index measures the isolation between two racial groups or the exposure (degree of contact) of
one group to another. I calculate the isolation index using the following formula:
II
d 𝑡 = ∑ (a
𝑠 𝑡
/ A
d 𝑡 ) – (a 𝑠 𝑡
/ C
s 𝑡 ) (2)
Where II
d 𝑡 is Isolation index of district d at time t. a
𝑠 𝑡
is the number of “a” students in
school s at time t and A
d 𝑡 is the number of “a” students in all schools in district d at time t. C
s 𝑡 is
the number of ALL students in school s at time t.
In order to categorize the concentration of students in schools, I create school-level racial,
income and achievement segregation indicators to complement the segregation indices including:
a) predominantly minority – greater than 50 percent of students in a school is non-White; b)
intensely segregated minority schools – 90-100 percent of student body is minority; c) intensely
segregated white schools – 90-100 percent of student body is White; d) racially unbalanced
school (black) – the percent of Black students in a school is 20 percentage points higher or lower
than the percent of Black students in the district; e) racially unbalanced school (Hispanic) – the
percent of Hispanic students in a school is 20 percentage points higher or lower than the percent
of Hispanic students in the district; f) high poverty schools – greater than 50 percent of students
in a school are free and reduced price lunch recipients; g) extreme poverty schools – 90-100
135
percent of student body are FRPL recipients; h) predominantly low achieving – greater than 50
percent of students in school are achieving below district average; i) intensely low achieving –
90-100 percent of student body is achieving below the district average and; j) intensely high
achieving schools – 90-100 percent of student body is performing above the district average and;
k) low achieving schools – less than 10 percent of the student body is performing above the
district average.
Achievement Gaps
I also calculate district-level and school-level achievement gaps over the period of study.
For example, to compute the white-black district-level achievement gap, I subtract the
standardized mean math achievement of black students from that of white students. I also
calculate yearly within-school achievement gap, or the achievement gap between racial and
income groups within a school. For instance, a within school achievement gap between blacks
and whites for year t is calculated by subtracting the standardized achievement of black students
within a school from the mean achievement of white students in that school for year t.
Results
Table 1 provides a summary of the demographic characteristics of CCSD. CCSD is a
large, diverse urban school district with average annual enrollment of more than 300,000
students. On average, roughly 42 percent of students are Hispanic, 33 percent are White, 13
percent are African American and 8 percent are Asian. The proportion of African American
students slightly decreased over the period of study from 14 to 13 percent while the proportion of
Hispanic students increased over time from 40 percent in 2007-08 to 43 percent in 2012-13.
There was a marked decline in the proportion of White students from 36 percent in 2007-08 to
136
29 percent in 2012-13. Similarly, the proportion of Asian students also decreased but to a lesser
extent. The proportion of special education status students rose from 10 to 12 percent while the
proportion of ELL students mildly fluctuated but declined overall from 20 percent in 2007-08 to
16 percent in 2012-13. The proportion of FRPL students increased over time from 44 percent in
2007-08 to 50 percent in 2012-13.
Table 2 shows segregation among schools in CCSD from 2007-08 through to 2012-13
using the dissimilarity index. The results indicate that although overall racial segregation in
CCSD was moderate, unevenness in the distribution of students by race/ethnicity in the district
increased over the period of study. On average, the dissimilarity index was lowest for Black-
Others (0.249) and highest for Hispanic-White (0.482). Black-White, Black-Asian and Black-
Hispanic segregation increased over the period of study. Similarly, Hispanic-White and
Hispanic-Asian segregation increased over time and were higher compared to those of Black
students. The results imply that Hispanic students are more unevenly distributed than Black
students. Indeed, Hispanic students were the most highly unevenly distributed racial group. In
sum, the results suggest that most racial groups are moderately evenly distributed between
schools in CCSD but segregation across all racial categories is increasing.
[Insert Table 2 around here]
Table 2 also illustrates that unlike racial segregation, income segregation decreased over
the period of study from 0.446 to 0.350. In other words, free and reduced priced lunch recipients
are somewhat evenly distributed between schools and the distribution of students by income has
improved over time in CCSD. Students achieving below the district average in math and students
proficient in math both had moderate segregation over the period of study (0.302 and 0.350
respectively). However, the distribution of proficient students between schools grew slightly
137
more uneven over time whereas the distribution of below average students fluctuated but did not
increase over the period of study. The results suggest that the segregation of high achieving
students is increasing in CCSD. In particular, these students are more unevenly distributed
relative to low-achieving students and the gap appears to be widening over time.
Table 3 indicates that overall isolation among racial groups was also low to moderate and
decreased over the period of study. Black and Asian students had low isolation whereas Hispanic
and White students had moderate isolation. Hispanic students were the most isolated racial group
(0.530) and Asian students (0.126) were the least isolated racial group. Furthermore, isolation
decreased over time for all racial groups except Hispanic students. Overall, the results imply that
there is a considerable degree of potential contact between racial student subgroups in CCSD that
has increased.
43
[Insert Table 3 around here]
Table 3 also shows that free and reduced price lunch recipients were more isolated (.600)
than any of the racial groups. This implies that income isolation is more prevalent and
concerning than racial isolation. However, isolation by income slightly decreased over time.
Isolation by achievement level was also high and increasing in the district. Students below
average had an average isolation index of 0.525 relative to 0.651 for proficient students. The
results suggest that both low and high achieving students are increasingly having contact with
students of similar achievement levels. The findings also imply that isolation by achievement
levels may be even higher and more important than racial and income segregation. Indeed,
proficient students are the most isolated of all racial, income and achievement groups.
43
I also compute the dissimilarity and isolation indices for ELL and special education students. ELL students were
moderately unevenly distributed (.465) relative to special education students (.121) who were somewhat evenly
distributed across schools. Both ELL and special education students had low isolation (.330 and .124 respectively).
138
The results suggest that there is increasing stratification within the district as racial and
achievement segregation rose over time.
44
There was an increase in racial segregation among all
racial groups but income segregation decreased over the period of study. African-American and
Asian students were fairly evenly distributed and had a high degree of possible interaction with
other racial groups. Hispanic students were the most segregated racial group on both the
dissimilarity and isolation indices. FRPL students were somewhat evenly distributed (.399) but
were also fairly high isolated (.6). Overall, segregation between schools was moderate in CCSD
and unevenness increased while isolation decreased. The rise in achievement segregation is
particularly noteworthy. Low achieving students are somewhat evenly distributed but highly
isolated. High achieving students were more unevenly distributed and isolated than their lower
achieving peers.
Table 4 presents district-level achievement gaps and the average of within-school
achievement gaps over the period of study. The results indicate sizable achievement gaps
between racial groups that increased over time. On overage, White students in CCSD performed
0.76 of a standard deviation higher than Black students and the achievement gap between White
and Black students increased from 0.72SD in 2007-08 to 0.82SD in 2012-13. The achievement
gap between White and Hispanic students was smaller than the White-Black achievement gap
(0.52SD). The White-Hispanic achievement gap also slightly increased over time. The Asian-
Black achievement gap was the largest in CCSD with Asian students performing nearly a
44
The first set of specification checks examines the sensitivity of the results to open enrollment options in the school
district. Prior research has found that charter schools may increase racial segregation (Bifulco & Ladd, 2007). I
estimate the dissimilarity and isolation indices while excluding open enrollment options – magnet and charter
schools – in the CCSD. First, I excluded the 13 charter schools in 2012-13 and the results changed. In particular,
racial and income segregation was lower across all groups when charter schools were excluded. Achievement
segregation also decreased. The results were similar when magnet schools were separately excluded. The results
suggest that open enrollment options may increase racial, income and achievement segregation (both unevenness
and isolation) within school districts. It also implies that magnet and charter schools may also partly explain the rise
in racial and income segregation between school years in urban school districts. Nevertheless, the segregation levels
in CCSD remained moderate and increased over time with or without open enrollment options.
139
standard deviation above Black students. Similar to the White-Black achievement gap, the
Asian-Black achievement gap also increased over the period of study. The Asian-White
(0.19SD) and the Hispanic-Black (0.29SD) comparisons were the smallest gaps in test scores in
CCSD. However, these smaller achievement gaps were also increasing over time. In addition to
disparities in achievement between racial groups, there was also a notable achievement gap
between income groups. Non-FRPL recipients outperformed FRPL students by about half of a
standard deviation. However, the income achievement gap decreased over time. Within-school
achievement gaps demonstrated a similar pattern to district-level achievement gaps.
45
[Insert Table 4 around here]
In sum, the results suggest that there was increasing racial and achievement segregation
as well as widening achievement gaps that accompanied considerable non-structural mobility
rates over the period of study. Racial achievement gaps increased while the dissimilarity index
(unevenness) increased and the isolation index (isolation) decreased. Conversely, income
achievement gaps and income segregation decreased over time. Similar to other districts
nationwide, racial segregation is increasing, however, unlike the national trend, CCSD is
experiencing declining isolation and increasing unevenness. The results are also similar to prior
research that found that as black-white dissimilarity increased, racial achievement gaps also
increased (Condron et al., 2013).
Segregation, Achievement Gaps and the Timing of School Changes
Table 5 presents the average within-school achievement gaps and school turnover by the
timing of school changes across different classifications of school-level segregation. Over the
45
I also calculate achievement gaps between non-ELL and ELL students as well as between non-special education
and special education students. Both achievement gaps are large. On average, non-ELL students performed .8 of a
standard deviation higher than ELL students and non-special education students performed nearly one standard
deviation above special education students at the district-level.
140
period of study, 67 percent of schools were predominantly nonwhite, 13 percent were intensely
segregated (minority), 6 percent were racially unbalanced (Black) and 47 percent were racially
unbalanced (Hispanic). The proportion of predominantly nonwhite schools was similar across
elementary, middle and high schools. However, the percentage of intensely segregated schools
was higher in elementary schools (16 percent) relative to middle (10 percent) and high (6
percent) schools. Similarly, 5 percent of elementary schools were racially unbalanced (Black)
schools compared to 1 percent of middle schools and 4 percent of high schools. The majority of
magnet and discipline alternative schools (behavior and continuation schools) were
predominantly nonwhite (90 percent for magnet schools and 95 percent for discipline schools).
However, 11 percent of magnet schools were intensely segregated minority schools.
[Insert Table 5 around here]
White-Black, White- Hispanic and the Non-FRPL-FRPL within-school achievement gaps
were lower in predominantly nonwhite schools relative to schools that were not predominantly
nonwhite. Conversely, the Asian-Hispanic gap was higher in predominantly nonwhite schools.
Similarly, the White-Hispanic achievement gap in intensely segregated minority and racially
unbalanced (Black) schools was less than half that of non-intensely segregated and racially
balanced (Black) schools. Interestingly, FRPL and non-FRPL students in intensely segregated
schools had similar test scores and thus a negligible achievement gap. In addition, the Asian-
Black and the Asian-Hispanic achievement gaps remained constant irrespective of the
segregation levels in schools whereas the White-Black and White-Hispanic achievement gaps
were lower in more segregated schools. Overall, the results imply that racial and income
achievement gaps are lower in more segregated schools.
141
Non-structural mobility rates were higher in predominantly nonwhite schools especially
for during the year school changes (9 percent versus 5 percent). Interestingly, predominantly
nonwhite schools also had during the year mobility rates that were equal to mobility rates
between school years. When one also considers the higher rates of ultra-movers in these schools,
the results imply that students in predominantly nonwhite schools switched schools just as or
more often during the school year than between academic years. For both intensely segregated
and racially unbalanced (Black) schools, during the year nonstructural mobility rates were higher
than between year mobility rates. Racially unbalanced (Black) schools had the highest within
year non-structural mobility rates. Indeed, 18 percent of students in these schools exited during
the school year and 9 percent changed schools both between and during the school year. The
results suggest that more segregated schools typically have a higher non-structural mobility rate
and mid-year school changes are especially prevalent in these schools.
Table 5 also illustrates that more segregated schools typically have a lower proportion of
proficient students. For instance, in predominantly nonwhite schools 56 percent of students were
proficient relative to 72 percent in schools that were not predominantly nonwhite. The negative
relationship between segregation and school quality is confirmed by the findings on
accountability status. A higher proportion of segregated schools were classified as a school ‘in
need of improvement’ compared to less segregated schools. For example, 71 percent of intensely
segregated minority schools were classified as ‘in need of improvement’ relative to 45 percent
for non-intensely segregated schools. The results also imply that older and larger schools are also
more segregated in the district. Less segregated schools are generally nearly ten years newer than
more segregated schools.
142
Overall the results suggest that more segregated schools typically have smaller
achievement gaps, a lower proportion of proficient students and higher non-structural mobility
rates (especially during the year mobility) than less segregated schools. There appears to be a
positive relationship between school segregation and within year non-structural moves: schools
with higher levels of segregation also have greater proportions of students exiting during the
school year. The results also imply that the achievement gap is smaller in more segregated
schools because of the presence of similar low-achieving students regardless of race/ethnicity,
whereas larger achievement gaps in less segregated schools suggest minority students tend to be
low-achieving and non-minority students are higher-achieving resulting in larger achievement
gaps. From the district’s perspective, this is not a beneficial trend given than prior research
demonstrates that high-achieving peers improve the student achievement of all students in a
school.
Student Mobility and Segregation
In the final set of analyses, I examine the segregation levels of the schools that non-
structural movers transfer to and the possible differences between movers from similarly
segregated schools who switched to schools with different segregation levels. These results
provide a better sense of whether mobile students are switching to more or less segregated
schools and whether there are important variations across the timing of school changes. Overall,
regardless of the timing of school changes, mobile students in predominantly nonwhite schools
switched to similar schools. For instance, 92 percent of between year nonstructural movers from
predominantly nonwhite schools transferred to another predominantly nonwhite school (93
percent of mid-year movers made similar moves). Interestingly, a notable proportion of students
in schools that are not predominantly nonwhite switched to predominantly nonwhite schools. For
143
example, about 17 and 29 percent of non-structural movers from non-predominantly nonwhite
schools transferred to predominantly nonwhite schools between and during the school year
respectively (31 percent of ultra-movers made similar moves). The results imply that during the
year mobility by students in less segregated schools may be more likely to result in switching to
more segregated schools.
Interestingly, regardless of the timing of school changes, a considerable proportion of
nonstructural movers in intensely segregated minority schools switched to non-intensely
segregated schools (about 62 percent transferred to another intensely segregated school). About
38 percent of between year movers, 41 percent of mid-year movers and 45 percent of ultra-
movers in intensely segregated schools switched to less segregated schools. It is interesting to
note that a higher proportion of within year nonstructural movers in intensely segregated schools
transferred to less segregated schools compared to between year movers. Conversely, the
overwhelming majority of students in non-intensely segregated schools switched to another non-
intensely segregated school regardless of the timing of school changes (96 percent of between
year movers, 94 percent of mid-year movers and 94 percent of ultra-movers). The results suggest
that some mobile students in the most segregated schools are transferring to less segregated
schools (whether by choice or by chance) especially during the academic year. This implies that
student mobility may positively contribute to desegregation in urban school districts as
considerable proportions of mobile students in intensely segregated schools switch to non-
intensely segregated schools.
I further examined the differences between nonstructural movers in intensely segregated
schools that transferred to another intensely segregated school and those that switched to a less
segregated school. Table 6 presents the differences in non-structural movers (regardless of the
144
timing of school changes) who exited intensely segregated schools. Overall the results indicate
statistically significant differences between the two groups of mobile students. Students who
switched to less segregated schools were, on average, a higher proportion of Black and White
students and a slightly lower proportion of FRPL students relative to their counterparts who
switched to other intensely segregated schools. A greater proportion of Hispanic mobile students
in intensely segregated schools switched to other intensely segregated schools than to less
segregated schools. Interestingly, students who switched to less segregated schools had lower
achievement than their counterparts (-.60SD versus .53SD).
[Insert Table 6 around here]
There are also further statistically significant differences among students who switched
from intensely segregated schools to less segregated schools by the timing of school changes. In
particular, between and during the year movers had significant differences across all student and
school characteristics except Asian, White, Special Education, and the percent of nonwhite,
FRPL, ELL, special education students and ultra-movers in a school. A greater proportion of
Black, Asian, special education, ELL and male students switched during the school year whereas
a greater proportion of FRPL and Hispanic students switched between school years. During the
year movers had lower achievement than between year movers who transferred to less
segregated schools (-.68SD versus .39SD). Between year movers also exited from higher quality
schools (52 percent vs 46 percent of proficient students in schools) with a lower proportion of
Black students and a higher proportion of Hispanic students than mid-year movers.
Moreover, there are also statistically significant differences between mid-year movers
and ultra-movers who switched from intensely segregated to less segregated schools. Ultra-
movers and mid-year movers had statistically significant differences across all student and exit
145
school characteristics except Asian, White, FRPL, math proficiency and percent of nonwhite
students in schools. Ultra-movers were a more disadvantaged subset of mid-year movers.
Specifically, a greater proportion of Black and male students were ultra-movers relative to mid-
year movers. Ultra-movers also had lower achievement relative to mid-year movers (-.79SD
versus -.68SD). Ultra-movers also exited slightly lower quality schools relative to mid-year
movers who switched from intensely segregated to less segregated schools. The results highlight
the heterogeneity in the mobile student population within an urban school district with different
students in different schools with comparable segregation levels making similar school changes
at different times.
Similar to mobile students in intensely segregated minority schools, a substantial
proportion of non-structural movers exiting racially unbalanced (Black) schools switched to
racially balanced (Black) schools regardless of the timing of school changes. More than half of
between year movers (53 percent) from racially unbalanced schools switched to racially balanced
schools. The proportion of mid-year and ultra-movers making similar school changes was even
higher (54 percent of mid-year movers and 65 percent of ultra-movers made similar moves).
Conversely, mobile students in racially balanced schools typically transferred to another racially
balanced school irrespective of the timing of school changes. The results suggest student
mobility may result in desegregation as the majority of non-structural movers in racially
unbalanced schools switch to racially balanced schools.
Interestingly, the results for racially unbalanced (Hispanic) schools paint a different
picture of how mobility may affect segregation within an urban school district. The results for
racially unbalanced (Hispanic) schools differ from those of intensely segregated minority and
racially unbalanced (Black) schools. In particular, most nonstructural movers in racially
146
unbalanced (Hispanic) schools transferred to another racially unbalanced school (78 percent).
The trend was similar for during the year mobility with 75 percent of mid-year movers and 72
percent of ultra-movers in racially unbalanced schools switching to other unbalanced schools.
Moreover, a notable proportion of students in racially balanced schools transferred to racially
unbalanced (Hispanic) schools: 16 percent of between year movers, 20 percent for mid-year
movers and ultra-movers. These results suggest that in some instances student mobility may
result in greater segregation as the overwhelming majority of mobile students in racially
unbalanced (Hispanic) schools transfer to another unbalanced school, and a notable proportion of
non-structural movers in racially balanced schools switch to racially unbalanced schools.
In essence, the results suggest that student mobility may contribute to desegregation in
positive and negative ways. Overall, mobile students typically switched to schools with similar
segregation levels. However, there is evidence that many students in the most segregated schools
in CCSD switch to less segregated schools, especially during the school year. There was also
little evidence of racially segregating transfers in CCSD regardless of the timing of school
changes. The results also illustrate the multi-layered heterogeneity among nonstructural movers
across the timing of school changes. Students who exit segregated schools for less segregated
schools typically differed from students who exit segregated schools for similar schools.
Moreover, there were further differences between the students who transferred to less segregated
schools across the timing of school changes.
Discussion
This paper examines the relationships between student mobility, segregation and
achievement gaps within an urban school district. Two key takeaways emerge from the findings.
147
First, there is evidence of stratification in the CCSD. Differential mobility patterns are
accompanied by an increase in racial and achievement segregation. The results imply that
achievement segregation is just as or even more important than racial and income segregation in
urban school districts. Second, student mobility may influence segregation in positive and
negative ways. The results suggest that some non-structural moves may add to segregation
whereas others may boost desegregation. In essence, there is a reasonable possibility student
mobility may have a net positive effect on desegregation.
The paper suggests that segregation remains a pertinent issue in school districts. The
results imply that differential mobility patterns may increase racial and achievement segregation
but may reduce income segregation. Although it is difficult to estimate how quickly stratification
is increasing, examining the relationship between student mobility and school segregation
provides a way to see how stratification is urban school districts in changing.
This study also illustrates that student mobility may affect segregation in various ways.
There are flows and patterns across the timing of schools changes as well as the segregation
levels of origin and destination schools that create a matrix of numerous complex interactions.
Student mobility patterns across the timing of school changes may influence the net segregation
within a school district. The results from this paper suggest that intra-district student mobility in
urban school districts may aid desegregation as students exit intensely segregated schools for less
segregated schools.
Policy Implications
A few policy implications emerge from this study. First, desegregation ought to be a high
priority of policymakers. This paper adds to a growing number of studies that have found that
segregation is rising in urban districts. Segregation still remains an issue worthy of
148
policymakers’ attention. Second, policymakers should consider using student mobility as a
multi-purpose policy tool. Changing schools may be used as a measure of disadvantage for
students as well as a measure of effectiveness for schools. Policymakers should also explore
possible constitutional ways to use student mobility as a desegregation tool. In the wake of the
2007 Supreme Court decisions on desegregation that deemed the majority of voluntary
desegregation programs by school districts unconstitutional, there is need to consider feasible
options within the law to attain integrated schools (Orfield & Lee, 2007). Student mobility is a
possible policy lever to affect desegregation that warrants further consideration.
Directions for Future Research
The findings also provide some directions for future research. First, a better
understanding of families’ demand for less segregated schools among mobile students and how
this may vary with the timing of school changes is needed. A complementary qualitative study
may provide better insights on whether non-structural moves that aid desegregation are by choice
or by chance and determine whether parents are actively seeking less segregated schools. For
instance, the relationship between student mobility and segregation may be a byproduct of
parents’ demand for other school characteristics. Additionally, a better understanding of how
segregation levels of origin schools affect mobility decisions will help unpack the multi-
dimensional heterogeneity in families’ preferences that may influence the relationship between
student mobility and segregation. Second, studies with classroom-level data that allow for
estimation of within-school segregation may provide a stronger link between segregation and
student mobility at the school level. These investigations will further illuminate how segregation
and student mobility affects schools. Future work should delve deeper into unlocking how the
149
relationship between the two can be used to improve students’ outcomes in urban school
districts.
Conclusion
This study is exploratory and finds critical connections between the patterns of two
important phenomena in urban school districts, student mobility and school segregation. The
findings suggest that differential mobility patterns in urban school district may result in increased
racial and achievement segregation. This paper also reinforces the need for greater attention on
relationship student mobility and segregation given the significant equity implications.
150
Chapter Three Tables
Table 1
Student Demographic Statistics of CCSD, K-12, 2007-08 -2012-13
2007-08 2008-09 2009-10 2010-11 2011-12 2012-13
African-American 0.14 0.14 0.15 0.13 0.12 0.13
(0.35) (0.35) (0.35) (0.33) (0.33) (0.34)
Hispanic 0.40 0.40 0.41 0.42 0.43 0.43
(0.49) (0.49) (0.49) (0.49) (0.50) (0.50)
White 0.36 0.35 0.34 0.31 0.30 0.29
(0.48) (0.48) (0.47) (0.46) (0.46) (0.46)
Asian 0.09 0.10 0.10 0.07 0.07 0.07
(0.29) (0.30) (0.29) (0.25) (0.25) (0.25)
Pacific N/A N/A N/A 0.01 0.01 0.01
(0.12) (0.12) (0.12)
Multirace N/A N/A N/A 0.05 0.06 0.06
(0.22) (0.24) (0.24)
Male 0.51 0.51 0.51 0.51 0.51 0.51
(0.50) (0.50) (0.50) (0.50) (0.50) (0.50)
Special Education 0.10 0.10 0.10 0.10 0.09 0.12
(0.30) (0.29) (0.30) (0.30) (0.29) (0.32)
English Language Learner 0.20 0.17 0.16 0.16 0.17 0.16
(0.40) (0.37) (0.37) (0.36) (0.37) (0.37)
Free and Reduced Lunch 0.44 0.44 0.43 0.50 0.53 0.50
(0.50) (0.50) (0.50) (0.50) (0.50) (0.50)
N 325807 325712 322683 321678 320725 329663
Note. N includes leavers, entrants and current enrolled students. Standard deviations in parentheses. Pacific and Multirace ethnic/race classification by CCSD
started in 2010-11.
151
Table 2
School Segregation in CCSD, K-12, Dissimilarity Index, 2007-08 -2012-13
2007-08 2008-09 2009-10 2010-11 2011-12 2012-13 Average
Black-White 0.362 0.363 0.365 0.394 0.408 0.420 0.386
Black-Hispanic 0.300 0.298 0.303 0.303 0.301 0.310 0.302
Black-Asian 0.343 0.351 0.353 0.405 0.418 0.432 0.385
Black-Others 0.237 0.231 0.232 0.256 0.261 0.272 0.249
Hispanic-White 0.474 0.480 0.484 0.480 0.483 0.489 0.482
Hispanic-Asian 0.433 0.437 0.441 0.447 0.455 0.462 0.446
Hispanic-Others 0.394 0.397 0.403 0.393 0.392 0.398 0.396
Asian-White 0.267 0.274 0.282 0.323 0.334 0.338 0.304
Asian-Others 0.284 0.292 0.296 0.337 0.353 0.363 0.321
FRPL-Others 0.446 0.410 0.375 0.390 0.429 0.350 0.399
Below Average-Others 0.310 0.306 0.283 0.298 0.303 0.309 0.302
Proficient-Others 0.342 0.353 0.337 0.352 0.362 0.352 0.350
N 335 342 344 350 350 362 2083
152
Table 3
School Segregation in CCSD, K-12, Isolation Index, 2007-08 -2012-13
2007-08 2008-09 2009-10 2010-11 2011-12 2012-13 Average
Black-Others 0.196 0.194 0.193 0.178 0.177 0.178 0.186
White-Others 0.474 0.469 0.465 0.429 0.421 0.408 0.444
Hispanic-Others 0.517 0.522 0.528 0.535 0.541 0.533 0.530
Asian-Others 0.136 0.140 0.144 0.112 0.112 0.112 0.126
FRPL-Others 0.608 0.576 0.542 0.607 0.680 0.585 0.600
Below Average-Others 0.516 0.519 0.531 0.525 0.525 0.532 0.525
Proficient-Others 0.614 0.628 0.648 0.683 0.712 0.623 0.651
N 335 342 344 350 350 362 2083
153
Table 4
Achievement Gaps of CCSD, K-12, 2007-08 -2012-13
2007-08 2008-09 2009-10 2010-11 2011-12 2012-13 Average
District-Level
White -Black 0.716 0.726 0.726 0.792 0.793 0.820 0.763
White- Hispanic 0.528 0.520 0.521 0.501 0.503 0.534 0.518
Asian- Black 0.839 0.849 0.870 1.024 1.052 1.074 0.951
Asian- Hispanic 0.651 0.642 0.665 0.733 0.761 0.786 0.706
Asian-White 0.122 0.122 0.144 0.231 0.258 0.254 0.189
Hispanic-Black 0.188 0.207 0.205 0.291 0.291 0.291 0.288
Non-FRPL-FRPL 0.533 0.507 0.474 0.500 0.579 0.493 0.514
Average Within School
White -Black 0.55 0.55 0.56 0.59 0.58 0.59
(0.23) (0.23) (0.22) (0.27) (0.28) (0.28)
White- Hispanic 0.34 0.34 0.35 0.31 0.30 0.31
(0.22) (0.22) (0.24) (0.25) (0.27) (0.25)
Asian- Black 0.69 0.68 0.69 0.81 0.85 0.82
(0.29) (0.30) (0.33) (0.37) (0.36) (0.37)
Asian- Hispanic 0.46 0.46 0.48 0.52 0.56 0.55
(0.28) (0.28) (0.31) (0.34) (0.36) (0.34)
Asian-White 0.12 0.13 0.13 0.21 0.26 0.23
(0.27) (0.27) (0.28) (0.33) (0.35) (0.34)
Hispanic-Black 0.21 0.22 0.22 0.28 0.28 0.28
(0.21) (0.23) (0.21) (0.25) (0.25) (0.27)
Non-FRPL-FRPL 0.31 0.33 0.30 0.31 0.36 0.25
(0.30) (0.21) (0.21) (0.21) (0.32) (0.25)
N 335 342 344 350 350 362
Note. Standard deviations in parentheses.
154
Table 5
School Segregation, Within-School Achievement Gaps and Schools’ exit rates: 2008-09 – 2012-13
Predominantly Nonwhite Intensely Segregated
(Minority)
Racially Unbalanced
(Black)
Racially Unbalanced
(Hispanic)
0 1 0 1 0 1 0 1
White-Black 0.61 0.53 0.57 0.47 0.56 0.40 0.53 0.59
(0.29) (0.28) (0.27) (0.36) (0.28) (0.41) (0.26) (0.30)
White- Hispanic 0.33 0.27 0.32 0.12 0.30 0.14 0.31 0.26
(0.18) (0.26) (0.22) (0.31) (0.23) (0.36) (0.21) (0.27)
Asian- Black 0.74 0.74 0.74 0.77 0.74 0.76 0.71 0.78
(0.39) (0.42) (0.38) (0.61) (0.40) (0.71) (0.33) (0.49)
Asian- Hispanic 0.46 0.48 0.49 0.42 0.48 0.50 0.50 0.45
(0.35) (0.39) (0.35) (0.54) (0.36) (0.68) (0.31) (0.45)
Non-FRPL-FRPL 0.43 0.24 0.34 0.00 0.30 0.18 0.33 0.25
(0.21) (0.28) (0.22) (0.40) (0.27) (0.32) (0.19) (0.34)
BY exit rate (%) 0.07 0.09 0.08 0.08 0.08 0.10 0.08 0.08
(0.08) (0.08) (0.08) (0.07) (0.08) (0.07) (0.08) (0.08)
WY exit rate (%) 0.05 0.09 0.07 0.10 0.07 0.18 0.08 0.07
(0.05) (0.08) (0.08) (0.07) (0.07) (0.19) (0.10) (0.04)
UM exit rate (%) 0.01 0.03 0.03 0.03 0.03 0.09 0.03 0.02
(0.03) (0.05) (0.05) (0.05) (0.04) (0.13) (0.06) (0.03)
School Quality 0.72 0.56 0.63 0.50 0.62 0.38 0.59 0.64
(0.14) (0.17) (0.17) (0.15) (0.17) (0.22) (0.17) (0.17)
Percent of INOI
schools
0.25 0.60 0.45 0.71 0.49 0.39 0.51 0.45
(0.43) (0.49) (0.50) (0.45) (0.50) (0.49) (0.50) (0.50)
Enrollment 897.38 956.56 952.82 831.03 965.00 456.85 958.57 912.23
(698.62) (621.75) (669.96) (472.69) (640.83) (595.49) (659.83) (635.42)
Year School was
Built
1996 1987 1991 1983 1990 1981 1991 1987
(11.79) (17.94) (16.07) (19.44) (16.43) (24.43) (14.86) (18.48)
N 691 1,392 1,811 272 1,966 117 1,110 973
155
Table 6
Differences in Non-structural Movers who exited from intensely segregated schools
Students that switched to IS
school from another IS school
Students that switched to non-
IS school from IS school
Difference
African-American 0.20 0.28 -0.08***
(0.40) (0.45) (0.004)
Hispanic 0.73 0.61 0.12***
(0.45) (0.49) (0.005)
Asian 0.02 0.02 -0.005**
(0.13) (0.15) (0.001)
White 0.05 0.08 -0.03***
(0.21) (0.27) (0.003)
Male 0.53 0.55 -0.02***
(0.50) (0.50) (0.005)
Special Education 0.12 0.14 -0.02***
(0.32) (0.34) (0.004)
English Language Learner 0.39 0.26 0.13***
(0.49) (0.44) (0.005)
Free and Reduced Lunch 0.85 0.84 0.01***
(0.36) (0.37) (0.04)
Student z score -0.53 -0.60 0.07***
(0.97) (0.98) (0.01)
Math proficiency 0.42 0.39 0.02**
(0.49) (0.49) (0.007)
School Quality 0.49 0.48 0.01***
(0.13) (0.15) (0.002)
Percentage Black 0.16 0.18 -0.02***
(0.14) (0.14) (0.001)
Percentage Hispanic 0.75 0.72 0.03***
(0.15) (0.14) (0.002)
Percentage White 0.05 0.06 -0.01***
(0.02) (0.02) (0.0002)
Percentage Asian 0.02 0.03 -0.01***
(0.02) (0.02) (0.0001)
Percentage Male 0.52 0.53 -0.01***
156
(0.03) (0.06) (0.0004)
Percentage Special Education 0.10 0.11 -0.01***
(0.03) (0.03) (0.0004)
Percentage ELL 0.43 0.38 0.05***
(0.18) (0.17) (0.002)
Percentage FRPL 0.83 0.82 0.01***
(0.13) (0.12) (0.001)
Percentage Nonwhite 0.94 0.92 0.02***
(0.02) (0.02) (0.0002)
Percentage Btw-year 0.10 0.09 0.01***
(0.12) (0.08) (0.001)
Percentage Within-year 0.09 0.11 -0.02***
(0.05) (0.10) (0.0007)
Percentage Umover 0.04 0.05 -0.01***
(0.04) (0.07) (0.0006)
N 21970 15160
157
CONCLUSION
In three interrelated empirical papers, this dissertation examines student mobility patterns
and effects across the timing of school changes in Clark County, Nevada. The first paper
describes the characteristics of mobile students, when they move and the quality of schools that
mobile students transfer to across the timing of school changes. The second paper disentangles
the transition costs of moving from changes in school quality associated with student mobility
across the timing of school changes. The third paper investigates the relationship between
student mobility, segregation and achievement gaps in an urban district over time.
The results indicate that the likelihood of making a non-structural school change and the
timing of the school change differ by students’ and schools’ characteristics. More importantly,
regardless of a student’s prior achievement or the timing of school changes, origin school quality
is predictive of destination school quality. Overall, the results imply that school changes between
school years are associated with positive gains in math achievement and little transactions costs
whereas school changes during the year are associated with negative gains in math achievement
and significant transactions costs. In addition, the results indicate that transactions costs vary by
the timing of school changes as well as student characteristics. The results suggest that more
segregated schools typically have smaller achievement gaps, a lower proportion of proficient
students and higher non-structural mobility rates (especially during the year mobility) than less
segregated schools. There is also evidence suggesting that student mobility may contribute
positively to desegregation as students in intensely segregated minority schools transfer to less
segregated schools, regardless of the timing of school changes.
158
This dissertation builds on and extends the student mobility literature in four key areas:
intra-district mobility, the timing of school changes, the effects of student mobility, and the
relationship between segregation and student mobility. In particular, this dissertation is one of
the first studies to investigate: (1) the variation in student mobility patterns and effects across the
timing of school changes within an urban district; (2) the transition costs of moving and changes
in school quality for mobile students across the timing of school changes within a school district;
(3) the effects of student mobility in elementary versus middle school grades using multiple
cohorts; (4) the conditional effects of intra-district student mobility across the timing of school
changes, or the heterogeneous nature of the impact of changing schools on different student
subgroups by the timing of school changes; (5) the impact of nonstructural mobility on math
achievement levels and growth using propensity score matching; (6) the effects of student
mobility using two different identification strategies; (7) the role of student mobility in untested
grades in the estimation of student mobility effects; (8) the characteristics and mobility effects of
ultra-movers or students who changed schools both between and during the school year in the
same academic year and ; (9) the relationship between student mobility, segregation and
achievement gaps at the district-, school-, and student-level.
This dissertation reinforces the importance of differentiating student mobility by the
timing of school changes. There are several instances where the results when one considers all
moves combined vary substantially from the results when school changes were categorized by
the timing of moves. In short, lumping school changes masks important variation in the patterns
and effects of student mobility. Given evidence of differential mobility patterns, the timing of
school changes may also provide useful but limited insights on the nature of the stratification.
159
The results of this dissertation also draw attention to the role of the reasons for student
mobility in mobility effects. The plausible impact of causes of mobility on the effects of mobility
remains a strong barrier to causal estimates. Overall, pre-existing differences between mobile
and non-mobile studies explain a considerable portion of the relationship between student
mobility and student achievement. Moreover, the timing of student mobility may reflect
underlying economic and social circumstances such as job loss or promotion and influence the
choice of schools. The systematic differences in student and (to a lesser extent) school
characteristics that predict the timing of school changes may reflect varying reasons for mobility.
This implies that there may be systematic differences in the reasons for mobility by student
characteristics and school characteristics. The relationship between transactions costs and the
timing of school changes also suggests that the reasons for mobility may be a prominent
consideration in student mobility effects.
One of the most important scholarly and policy implications emanating from this
dissertation is the importance of paying greater attention to moves that occur during the year.
About half of non-structural student mobility within CCSD occurred during the school year.
Mobility, particularly mid-year, is particularly prevalent among low-income, minority and low-
achieving students. During the year mobility may also create issues for schools. Of all the timing
of school changes, researchers and policymakers should focus intensely on mid-year movers.
Indeed, policies and interventions that enhance the positive aspects and reduce the disruptive
elements of during the year student mobility have the potential to improve student achievement
and school quality within urban school districts. These programs may include a) more effective
and targeted support (for e.g., providing more parent resources for students who move frequently
during the school year); b) school transfer rules (for e.g, setting transfer deadlines and conditions
160
for the timing of school changes) and; c) strategies for capitalizing on mobility (for e.g.,
providing rewards to low-income minority families that move to higher quality schools).
Future research on student mobility must embrace qualitative methods to push the
literature forward. Interviews, participant observations, surveys and focus groups etc. will
contribute more nuanced perspectives about how students and families experience student
mobility that are beyond the scope of quantitative studies. Mixed methods studies that provide a
granular description of student mobility as well as broader quantitative trends will result in a
better understanding of the prevalence and impact of student mobility in urban school districts.
This research design may uncover new insights about the relationships between the reasons for
mobility, the timing of school changes and student achievement by connecting the abstract to the
real life experiences of mobile students and families.
In sum, student mobility highlights disadvantages in new ways and provides an insightful
angle on the challenges facing urban schools and districts. Furthermore, the similarities in
mobility rates, patterns and effects between choice and non-choice based school districts suggest
that student mobility may be an important consideration in the equity and improvement in
education regardless of differences in educational governance context. Student mobility is also a
good segue to learn more about segmentation and may be a useful tool for policymakers in urban
districts to track the extent and effects of sorting between schools. This dissertation makes
notable contributions to the student mobility literature, however, there remains much room for
improving the knowledge base of the relationship between student achievement, school quality
and student mobility.
161
References
Alexander, K. L., Entwisle, D. R., & Dauber, S. L. (1996). Children in motion: School transfers
and elementary school performance. Journal of Educational Research, 90(1), 3–12.
http://doi.org/10.1080/00220671.1996.9944438
Astone, N. M., & Mclanahan, S. S. (1994). Family structure, residential mobility, and school
dropout: A research note. Demography, 31(4), 575–584. http://doi.org/10.2307/2061791
Berends, M. (2014). Achievement gap. In D. J. Brewer & L. Picus (Eds.), Encyclopedia of
education economics & finance (pp. 18–23). Thousand Oaks, CA: SAGE Publications, Inc.
http://doi.org/10.4135/9781483346595.n12
Bifulco, R. (2010). Can Propensity Score Analysis Replicate Estimates Based on Random
Assignment in Evaluations of School Choice? A Within-Study Comparison. Center for
Policy Research Working Paper. Retrieved from
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1805890
Bifulco, R., & Ladd, H. (2007). School choice, racial segregation, and test score gaps: Evidence
from North Carolina’s charter school program. Journal of Policy Analysis and
Management, 26(1), 31. Retrieved from
http://onlinelibrary.wiley.com/doi/10.1002/pam.20226/full
Bronfenbrenner, U. (1979). The ecology of human development: Experiments by design and
nature. Cambridge, MA: Harvard University Press.
Bronfenbrenner, U. (1989). Ecological systems theory. In R. Vasta (Ed.), Six theories of child
development: Revised formulations and current issues (vol. 6, pp. 187–250). Greenwich,
CT: Jai Press.
Bronfenbrenner, U. (1994). Ecological models of human development. In International
encyclopedia of education. Oxford, United Kingdom: Elsevier.
Bulkley, K., Henig, J., & Levin, H. (2010). Politics, governance, and the new portfolio models
for urban school reform: Between public and private. Cambridge, MA: Harvard Education
Press.
Burkam, D., Lee, V., & Dwyer, J. (2009). School mobility in the early elementary grades:
Frequency and impact from nationally-representative data. Paper presented at the
workshop on the impact of mobility and change on the lives of young children, schools, and
neighborhoods, June 29-30. Washington, DC: National Academies.
Carolan, B. V. (2013). School transitions and students’ achievement in the fifth grade. The
Journal of Educational Research, 106(5), 372–383.
http://doi.org/10.1080/00220671.2012.736432
162
Cole, M., Cole, S., & Lightfoot, C. (2005). The development of children. London, UK:
Macmillian.
Coleman, J. (1988). Social capital in the creation of human capital. American Journal of
Sociology, 94, S194–S120. http://doi.org/10.1086/228943
Coleman, J., Campbell, E., & Hobson, C. (1966). Equality of educational opportunity. US
Government Priinting, 1066–5684. Retrieved from
https://scholar.google.com/scholar?hl=en&q=the+coleman+report&btnG=&as_sdt=1%2C5
&as_sdtp=#3
Condron, D. J., Tope, D., Steidl, C. R., Freeman, K. J., & Colleges, W. S. (2013). Racial
segregation and the Black / White achievement gap, 1992 To 2009. The Sociological
Quarterly, 54, 130–157.
Conger, D. (2005). Within-school segregation in an urban school district. Educational
Evaluation and Policy Analysis. Retrieved from
http://epa.sagepub.com/content/27/3/225.short
Cook, T. D., Shadish, W. R., & Wong, V. C. (2008). Three conditions under which experiments
and observational studies produce comparable causal estimates: New findings from within-
study comparisons. Journal of Policy Analysis and Management, 27(4), 724–750.
http://doi.org/10.1002/pam.20375
Cullen, J. B., Jacob, B. A., & Levitt, S. D. (2005). The impact of school choice on student
outcomes: An analysis of the Chicago Public Schools. Journal of Public Economics, 89(5-
6), 729–760. http://doi.org/10.1016/j.jpubeco.2004.05.001
De la Torre, M., & Gwynne, J. (2009). When schools close: Effects on displaced students in
Chicago public schools. Chicago, IL.
Engberg, J., Gill, B., Zamarro, G., & Zimmer, R. (2012). Closing schools in a shrinking district:
Do student outcomes depend on which schools are closed? Journal of Urban Economics,
71(2), 189–203. http://doi.org/10.1016/j.jue.2011.10.001
Engec, N. (2006). Relationship between mobility and student performance and behavior. The
Journal of Educational Research, 99(3), 167–178. http://doi.org/10.3200/JOER.99.3.167-
178
Fantuzzo, J. W., LeBoeuf, W. A., Chen, C.-C., Rouse, H. L., & Culhane, D. P. (2012). The
unique and combined effects of homelessness and school mobility on the educational
outcomes of young children. Educational Researcher, 41(9), 393–402.
http://doi.org/10.3102/0013189X12468210
163
Fong, A., Bae, S., & Huang, M. (2010). Patterns of student mobility among English Language
Learner students in Arizona public schools (Issues & Answers Report, REL 2010-No. 093).
Regional Educational Laboratory West. Washington, DC.
Fortson, K. (2012). Using an experimental evaluation of charter schools to test whether
nonexperimental comparison group methods can replicate experimental impact estimates.
… Evaluation and …. Retrieved from http://mathematica-
mpr.com/publications/pdfs/education/expereval_charterschools.pdf
Frankenberg, E., Lee, C., & Orfield, G. (2003). A multiracial society with segregated schools:
Are we losing the dream? Retrieved from https://escholarship.org/uc/item/3rh7w18g.pdf
Furgeson, J., Gill, B., Haimson, J., Killewald, A., Mc-, Moira, C., … Bing-ru, T. (2012). Charter-
school management organizations: Diverse strategies and diverse student impacts.
Mathematica Policy …. Retrieved from http://eric.ed.gov/?id=ED528536
Garcia, D. R. (2011). The Achilles’ Heel of school choice policies: The obstacles to reporting
school accountability results to parents. Journal of School Choice, 5(1), 66–84.
http://doi.org/10.1080/15582159.2011.548249
Gasper, J., DeLuca, S., & Estacion, A. (2010). Coming and going: Explaining the effects of
residential and school mobility on adolescent delinquency. Social Science Research, 39(3),
459–476. http://doi.org/10.1016/j.ssresearch.2009.08.009
Gasper, J., DeLuca, S., & Estacion, A. (2012). Switching schools: Revisiting the relationship
between school mobility and high school dropout. American Educational Research Journal,
49(3), 487–519. http://doi.org/10.3102/0002831211415250
Grigg, J. (2012). School enrollment changes and student achievement growth: A case study in
educational disruption and continuity. Sociology of Education, 85(4), 388–404.
http://doi.org/10.1177/0038040712441374
Hanushek, E. A., Kain, J. F., Markman, J. M., & Rivkin, S. G. (2003). Does peer ability affect
student achievement? Journal of Applied Econometrics, 18(5), 527–544.
http://doi.org/10.1002/jae.741
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(9-10), 1721–
1746. http://doi.org/10.1016/S0047-2727(03)00063-X
Hanushek, E. A., & Rivkin, S. G. (2009). Harming the best: How schools affect the black-white
achievement gap. Journal of Policy Analysis and Management, 28(3), 366–393.
http://doi.org/10.1002/pam.20437
164
Heinlein, L. M., & Shinn, M. (2000). School mobility and student achievement in an urban
setting. Psychology in the Schools, 37(4), 349–357. http://doi.org/10.1002/1520-
6807(200007)37:4<359::AID-PITS6>3.0.CO;2-1
Herbers, J. E., Reynolds, A. J., & Chen, C.-C. (2013). School mobility and developmental
outcomes in young adulthood. Development and Psychopathology, 25(2), 501–15.
http://doi.org/10.1017/S0954579412001204
Howell, W. (2006). Switching schools? A closer look at parents’ Initial Interest in and
knowledge about the choice provisions of No Child Left Behind. Peabody Journal of
Education. Retrieved from
http://www.tandfonline.com/doi/abs/10.1207/S15327930pje8101_7
Institute of Medicine and National Research Council. (2010). Student mobility: Exploring the
impact of frequent moves on achievement (Summary of a workshop). (A. (Rapporteur)
Beatty, Ed.). Washington, DC: The National Academies Press.
Jencks, C., & Phillips, M. (2011). The Black-White test score gap. Retrieved from
https://books.google.com/books?hl=en&lr=&id=Ywb7r1oOxJYC&oi=fnd&pg=PA1&dq=J
encks+and+Phillips,+1998&ots=xpt7VYXhkq&sig=9imG8k4Q9lIbGHF_GpNUg2yN5UA
Jochim, A., Dearmond, M., Gross, B., & Lake, R. (2014). How Parents Experience Public
School Choice.
Johnson, R. C. (2011). Long-run Impacts of School Desegregation & School Quality on Adult
Attainments. Retrieved from http://www.nber.org/papers/w16664
Kain, J., & O’Brien, D. (1998). A Longitudinal Assessment of Reading Achievement: Evidence
from the Harvard/UTD Texas Schools Project. University of Texas at Dallas, Dallas, TX.
Retrieved from http://www.utdallas.edu/research/tsp-
erc/pdf/wp_kain_1999_longitudinal_assessment_reading.pdf.pdf
Kerbow, D. (1996). Patterns of urban student mobility and local school reform. Journal of
Education for Students Placed at Risk, 1(2), 147–169.
http://doi.org/10.1207/s15327671espr0102_5
Kerbow, D., Azcoitia, C., & Buell, B. (2003). Student mobility and local school improvement in
Chicago. The Journal of Negro Education, 72(1), 158–164. http://doi.org/10.2307/3211299
Kirshner, B., Gaertner, M., & Pozzoboni, K. (2010). Tracing Transitions: The Effect of High
School Closure on Displaced Students. Educational Evaluation and Policy Analysis, 32(3),
407–429. http://doi.org/10.3102/0162373710376823
Kurlaender, M., & Yun, J. (2001). Is Diversity a Compelling Educational Interest? Evidence
from Louisville. Retrieved from http://eric.ed.gov/?id=ED456196
165
Lash, A., & Kirkpatrick, S. L. (1990). A classroom perspective on student mobility. The
Elementary School Journal, 91(2), 176–191. http://doi.org/10.1086/461645
Levin, H., Daschbach, J., & Perry, A. (2010). A diverse education provider. In K. Bulkley, J.
Henig, & H. Levin (Eds.), Politics, governance and the new portfolio models for urban
school reform: Between public and private (pp. 323–340). Cambridge, MA: Harvard
Education Press.
Logan, J., Minca, E., & Adar, S. (2012). The Geography of Inequality Why Separate Means
Unequal in American Public Schools. Sociology of Education. Retrieved from
http://soe.sagepub.com/content/85/3/287.short
Mantzicopoulos, P., & Knutson, D. (2000). Head Start children: School mobility and
achievement in the early grades. The Journal of Educational Research, 93(5), 305–311.
http://doi.org/10.1080/00220670009598722
Mao, M., Whitsett, M., & Mellor, L. (1997). Student mobility, academic performance and school
accountability. ERS Spectrum, 16(1), 3–15.
Massey, D., & Denton, N. (1988). The dimensions of residential segregation. Social Forces,
67(2), 281–315. http://doi.org/10.2307/2579183
Mehana, M., & Reynolds, A. J. (2004). School mobility and achievement: A meta-analysis.
Children and Youth Services Review, 26(1), 93–119.
http://doi.org/10.1016/j.childyouth.2003.11.004
Mickelson, R. (2001). Subverting Swann: First-and second-generation segregation in the
Charlotte-Mecklenburg schools. American Educational Research Journal. Retrieved from
http://aer.sagepub.com/content/38/2/215.short
Nelson, P., Simoni, J., & Adelman, H. (1996). Mobility and school functioning in the early
grades. The Journal of Educational Research, 89(6), 365–369.
http://doi.org/10.1080/00220671.1996.9941340
Orfield, G. (1983). Public school desegregation in the United States, 1968-1980. Retrieved from
http://escholarship.org/uc/item/85w788b9.pdf
Orfield, G. (2001). Schools more separate: Consequences of a decade of resegregation. Retrieved
from http://eric.ed.gov/?id=ED459217
Orfield, G., & Lee, C. (2005). Why segregation matters: Poverty and educational inequality.
Retrieved from http://escholarship.org/uc/item/4xr8z4wb.pdf
Orfield, G., & Lee, C. (2007). Historic Reversals, Accelerating Resegregation, and the Need for
New Integration Strategies. Civil Rights Project/Proyecto Derechos Civiles. Retrieved from
http://eric.ed.gov/?id=ED500611
166
Orfield, G., & Yun, J. (1999). Resegregation in American schools. Retrieved from
http://escholarship.org/uc/item/6d01084d.pdf
Ou, S., & Reynolds, A. J. (2008). Predictors of educational attainment in the Chicago
Longitudinal Study. School Psychology Quarterly, 23(2), 199–229.
http://doi.org/10.1037/1045-3830.23.2.199
Parke, C. S., & Kanyongo, G. Y. (2012). Student attendance, mobility, and mathematics
achievement in an urban school district. The Journal of Educational Research, 105(3), 161–
175. http://doi.org/10.1080/00220671.2010.547231
Pribesh, S., & Downey, D. B. (1999). Why are residential and school moves associated with poor
school performance? Demography, 36(4), 521–534. http://doi.org/10.2307/2648088
Raudenbush, S. W., Jean, M., & Art, M. (2011). Year-by-year and cumulative impacts of
attending a high-mobility elementary school on children’s mathematics achievement in
Chicago, 1995 to 2005. In G. J. Duncan & R. Murnane (Eds.), Whither opportunity? Rising
inequality, schools, and children’s life chances (pp. 359–376). New York, NY: Russell
Sage Foundation.
Reardon, S. F. (2011). The widening academic achievement gap between the rich and the poor:
New evidence and possible explanations. In G. J. Duncan & R. Murnane (Eds.), Whither
opportunity? Rising inequality, schools, and children’s life chances (pp. 91–115). New
York, NY: Russell Sage Foundation.
Reynolds, A. J., Chen, C., & Herbers, J. (2009). School mobility and educational success: A
research synthesis and evidence on prevention. Paper presented at the workshop on the
impact of mobility and change on the lives of young children, schools, and neighborhoods,
June 29-30. Washington, DC: The National Academies.
Rosenbaum, P., & Rubin, D. (1985). Constructing a control group using multivariate matched
sampling methods that incorporate the propensity score. The American Statistician.
Retrieved from http://amstat.tandfonline.com/doi/abs/10.1080/00031305.1985.10479383
Rouse, C. E., Hannaway, J., Goldhaber, D., & Figlio, D. (2013). Feeling the Florida Heat? How
Low-Performing Schools Respond to Voucher and Accountability Pressure. American
Economic Journal: Economic Policy, 5(2), 251–81. Retrieved from
http://ideas.repec.org/a/aea/aejpol/v5y2013i2p251-81.html
Rumberger, R. W. (2002). Student mobility and academic achievement. ERIC Clearinghouse on
Elementary and Early Childhood Education.
Rumberger, R. W. (2003). The causes and consequences of student mobility. Journal of Negro
Education, 72(1), 6–21. http://doi.org/10.2307/3211287
167
Rumberger, R. W., & Larson, K. A. (1998). Student mobility and the increased risk of high
school dropout. American Journal of Education, 107(1), 1–35.
http://doi.org/10.1086/444201
Rumberger, R. W., Larson, K. A., Ream, R. K., & Palardy, G. J. (1999). The educational
consequences of mobility for California students and schools. Berkeley, CA: Policy
Analysis for California Education.
Scherrer, J. (2013). The negative effects of student mobility: Mobility as a predictor, Mobility as
a mediator. International Journal of Education Policy and Leadership, 8(1). Retrieved from
http://journals.sfu.ca/ijepl/index.php/ijepl/article/view/400
Schwartz, A., Stiefel, L., & Chalico, L. (2009). The multiple dimensions of student mobility and
implications for academic performance: Evidence from New York City elementary and
middle school students. Available at SSRN 1508603. New York, NY.
South, S. J., Haynie, D. L., & Bose, S. (2007). Student mobility and school dropout. Social
Science Research, 36(1), 68–94. http://doi.org/10.1016/j.ssresearch.2005.10.001
Swanson, C. B., & Schneider, B. (1999). Students on the move: Residential and educational
mobility in America’s Schools. Sociology of Education, 72(1), 54–67.
http://doi.org/10.2307/2673186
Teachman, J., Paasch, K., & Carver, K. (1996). Social capital and dropping out of school early.
Journal of Marriage and the Family, 58(3), 773–783. http://doi.org/10.2307/353735
Temple, J. A., & Reynolds, A. J. (1999). School mobility and achievement: Longitudinal
findings from an urban cohort. Journal of Social Psychology, 37(4), 355–377.
http://doi.org/10.1016/S0022-4405(99)00026-6
Tuttle, C., Gill, B., & Gleason, P. (2013). KIPP Middle Schools: Impacts on Achievement and
Other Outcomes. Final Report. … Policy Research, Inc. Retrieved from
http://eric.ed.gov/?id=ED540912
United States General Accounting Office. (1994). Elementary school children: Many change
schools frequently, harming their education. Washington, DC.
United States Government Accountability Office. (2010). Many challenges arise in educating
students who change schools frequently. Washington, DC.
Weckstein, P. (2003). Accountability and student mobility under Title I of the No Child Left
Behind act. Journal of Negro Education, 117–125. http://doi.org/10.2307/3211295
Wells, A., & Crain, R. (1994). Perpetuation theory and the long-term effects of school
desegregation. Review of Educational Research. Retrieved from
http://rer.sagepub.com/content/64/4/531.short
168
Welsh, R., Duque, M., & McEachin, A. (Forthcoming ). School choice, student mobility and
school quality. Education Finance and Policy.
Wright, D. (1999). Student mobility: A negligible and confounded influence on student
achievement. The Journal of Educational Research, 92(6), 347–353.
http://doi.org/10.1080/00220679909597618
Wright, M., & Masten, A. (2005). Resilience processes in development. In S. Goldstein & R.
Brooks (Eds.), Handbook of resilience in children (pp. 17–37). New York, NY: Springer
US. http://doi.org/10.1007/0-306-48572-9_2
Xu, Z., Hannaway, J., & D’Souza, S. (2009). Student transience in North Carolina: The effect of
mobility on student outcomes using longitudinal data. Washington, DC.
Zimmer, R., Gill, B., Attridge, J., Obenauf, K., & Lansing, E. (n.d.). Charter school authorizers
and student achievement, 59–85.
169
Appendix
Table A1
Summary of empirical K-12 intra-district student mobility studies: 1994-2014
Study (APA citation)
Location, Data and
Grade Span
Measure of Student
Mobility Outcomes and Methods Results
Alexander, K.L., Entwisle,
D.R., & Dauber, S.L. (1996).
Children in motion: school
transfers
and elementary school
performance. The Journal of
Educational Research, 90(1),
3–12.
• Baltimore,
Maryland
• Longitudinal data:
Beginning School
Study - A five year
panel (1982-1987)
following a sample of
767 first graders in 20
public schools
• Grades 1-5
Non-structural
mobility (Within-
year (within the same
city); Between years
(within the same city
and out of city))
• Number of moves
over 5 year period
(total, one move vs.
none, two or more
moves vs. none)
• Academic and non-
academic outcomes: report
card grades, math and
reading test scores, risk of
retention and receipt of
special education services
• OLS regression models
(test scores), ordered
logistic regression (marks
and special education
services) and a logistic
regression (retention status)
with controls for students'
academic and demographic
characteristics
Advantaged students transferred
outside the school district while
disadvantaged students changed
schools within the school district.
Students who moved had lower
test scores (approximately one-
tenth of a standard deviation),
lower marks, a higher risk of
retention and received more
elaborate special education
services. Most results (except
math achievement) lost statistical
significance after the inclusion of
prior achievement and
background characteristics as
controls. Multiple moves were
significant for test scores.
170
Been, V., Ellen, I. G.,
Schwartz, A. E., Stiefel, L.,
& Weinstein, M. (2011).
Does losing your home mean
losing your school?: Effects
of foreclosures on the school
mobility of children.
Regional Science and Urban
Economics, 41(4), 407-414.
• New York City
• Longitudinal data
• Grades 1-8
Non-structural
mobility
• Whether student
remained in the same
school across school
years
• Student mobility
• Regression analysis with
controls for race, poverty,
gender, grade and origin
school
Students living in properties with
foreclosure notices were more
likely to switch schools within
the district and less likely to exit
the district. There is little
evidence of differences across
racial groups. Children living in
buildings that entered foreclosure
as well as mobile students
(changed schools without
foreclosure) tend to move to
lower-performing schools.
Cullen, J. B., Jacob, B. A., &
Levitt, S. D. (2005). The
impact of school choice on
student outcomes: an analysis
of the Chicago Public
Schools. Journal of Public
Economics, 89(5-6), 729–760
• Chicago, Illinois•
Longitudinal data -
Administrative data
from the Chicago
Public Schools 1993-
1995 (194 lotteries at
19 high schools)•
Grades 9-12
Non-structural
mobility • Whether a
student opted out of
their neighborhood
school in high school
•Academic outcomes: test
scores• OLS regression
models and Instrumental
variables using distance to
school
Students who opt out of their
assigned neighborhood school
are more likely to graduate than
peers who remained in the same
school. However, students who
opted out are systematically
different from those who did not.
Career academies have a positive
impact on student achievement.
de La Torre, M., & Gwynne,
J. (2009). When schools
close: Effects on displaced
students in Chicago public
schools. Chicago, IL:
University of Chicago
Consortium on Chicago
School Research.
• Chicago, Illinois
• Longitudinal data:
administrative data on
public schools
between 2001 and
2006
• Grades K-12
Structural mobility
• Whether or not a
student attended a
school that closed
• Academic and non-
academic outcomes: math
and reading test scores,
attendance to summer
schools, retention, and
referral to special education
services
• Propensity score matching
that matched closing
schools to similar schools,
HLM to analyze student
achievement, logistic model
for non-academic outcomes
Most students leaving closing
schools transfer to similar low-
achieving schools. Displaced
students are more likely to
change schools a second time
after leaving for a school closure.
Students who transferred to a
higher-achieving school had
larger gains in reading and math
relative to students who
transferred to lower-achieving
schools.
171
Dunn, M., Kadane, J., &
Garrow, J. (2003).
Comparing harm done by
mobility and class absence:
Missing students and missing
data. Journal of Educational
and Behavioral Statistics,
28(3), 269–88.
• Pittsburgh,
Pennsylvania
• Longitudinal data:
administrative data on
public school students
between 1998 and
2000
• Grades 9-11
Non-structural
mobility
• Number of times a
student changed
schools in period of
study (No moves in
three year span vs. at
least one move)
•Academic outcomes: test
scores
• OLS regression models
Mobility has a negative
correlation with student
achievement
Engberg, J., Gill, B.,
Zamarro, G., & Zimmer, R.
(2012). Closing schools in a
shrinking district: Do student
outcomes depend on which
schools are closed? Journal of
Urban Economics, 71(2),
189–203.
• Anonymous urban
district
• Longitudinal data:
student-level data
from 2005-06 through
to 2008-09
• Grades K-8
Structural mobility
• school closures
• Academic outcomes: math
and reading test scores
• Instrumental variables
with controls for student
characteristics
School closures can have
negative effects on test scores
and attendance, but these effects
are lessened or offset when
students change to higher-
performing (value-added)
schools. No evidence of adverse
effects on students in receiving
schools.
Fleming, C. B., Harachi, T.
W., Catalano, R. F.,
Haggerty, K. P., & Abbott, R.
D. (2001). Assessing the
effects of a school-based
intervention on unscheduled
school transfers during
elementary school.
Evaluation Review 25(6),
655-679.
• Seattle,
Washington•
Longitudinal data:
Five year panel of 10
public elementary
schools that
participated in the
study starting in
1992-1993• Grades
K-6
Non-structural
mobility • Whether
the student made a
school change within
five years •Student Mobility • HLM
Students in intervention schools
were less likely to change
schools in the first 5 years of
project.
172
Grigg, J. (2012). School
Enrollment Changes and
Student Achievement
Growth: A case study in
educational disruption and
continuity. Sociology of
Education, 85(4), 388-404.
• Nashville,
Tennessee
• Longitudinal data:
Administrative data
on Metropolitan
Nashville Public
Schools (1998-2003)
• Grades 3-8
Non-structural and
structural mobility
(Between year
structural moves;
Between year non-
structural moves;
During the year non-
structural moves;
During the year-
compulsory
(expulsions))
• Indicator of annual
mobility
• Academic outcomes: Test
score growth in reading and
mathematics
• A student fixed effect
model with controls for
student characteristics,
grade-by-year fixed effects
All four types of mobility have a
significant negative effect on
annual gains in math and
reading. Estimates of during the
year mobility are not
significantly different from
between year estimates.
Gruman, D., Harachi,T.,
Abbott,R., Catalano,R., &
Fleming,C. (2008).
Longitudinal effects of
student mobility on three
dimensions of elementary
school engagement. Child
Development, 79(6), 1833-
1852.
• The Pacific
Northwest
• Longitudinal data:
A sample of about
1000 2nd-5th grade
students in 10
elementary schools
• Grades 2-5
Non-structural
mobility
• Number of school
changes (No school
change vs. one or
more school changes)
• Academic and non-
academic outcomes:
classroom participation,
positive attitude toward
school, teacher ratings of
student's academic
performance
• HLM with covariates for
student characteristics
School changes predicted
decreases in academic
performance and classroom
participation.
Heinlein, L. M., & Shinn, M.
(2000). School mobility and
student achievement in an
urban
setting. Psychology in the
Schools, 37(4), 349-357.
• New York City,
New York
• Longitudinal data:
A sample of nearly
800 sixth graders in a
school district in New
York City
• Grades K-6
Non-structural
mobility
•Total number of
moves before third
grade
• Total number of
moves from fourth
through sixth grade
• Academic Outcomes:
Sixth grade reading and
math achievement
• OLS regression model
controlling for third-grade
achievement, SES and
gender; Logistic regression
analyses
No significant association
between mobility and student
achievement in sixth grade after
controlling for demographic
variables and prior achievement.
Mobility prior to third grade may
be a predictor of sixth-grade
achievement than later mobility
173
Herbers, J. E., Reynolds, A.
J., & Chen, C.-C. (2013).
School mobility and
developmental outcomes in
young adulthood.
Development and
Psychopathology, 25(2),
501–15.
• Chicago, Illinois•
Longitudinal data:
CLS data for 25 years
• Grades K-12
Residential and non-
structural mobility •
Number of years
students changed
schools in K-12, K-4,
4-8 and 8-12, •
Number of years
students changed
residences between
K-12, • Two or more
moves, three or more
moves and four or
more moves during
K-12
Academic and non-
academic outcomes: highest
grade completed, on time
graduation, occupational
prestige, depression
symptoms and criminal
arrests HLM with covariates
for student characteristics,
prior achievement and
behavioral history, Linear
regressions and binary
logistic regressions
Students with more school
changes between K-12 are less
likely to complete high school on
time and more likely to be
arrested as adults. More negative
outcomes are associated with
moves later in school career,
especially between 4th and 8th
grades. The number of school
moves predicted outcomes even
after controlling for contextual
factors such as residential
mobility and family poverty.
Kerbow, D. (1996). Patterns
of student mobility and local
school reform. Journal of
Education
for Students Placed at Risk,
1(2), 147-69.
• Chicago, Illinois
• Longitudinal data:
over 2,500 Chicago
public elementary
students
• Grades K-6
Non-structural
student mobility
• Number of school
changes over a two-
year period
• Academic Outcomes:
math and reading test scores
• Mixed methods: HLM;
Surveys with sxth graders
The pace of the curriculum was
substantially slower in schools
with highly mobile student
populations relative to schools
with stable student population.
Students who moved four or
more times were one full school
year behind their stable
counterparts. Mobility typically
occurs in clusters of schools with
similar characteristics and the
achievement levels of origin
schools strongly predicts
destination schools.
174
Kirshner, B., Gaertner, M., &
Pozzoboni, K. (2010).
Tracing Transitions The
Effect of High School
Closure on Displaced
Students. Educational
evaluation and policy
analysis, 32(3), 407-429.
• Urban District in
Western U.S
• Longitudinal data
• Grades 6-10
Structural mobility
• Whether or not a
students switched
schools due to school
closure
• Academic outcomes: test
scores, dropout rates and
graduation rates
• Mixed methods:
quantitative analyses for
academic performance and
open-ended surveys for
students' experiences with
closure and transitioning to
new schools
HLM
Test scores changed significantly
after a school closure
announcement. Post-closure
years were associated with an
increased likelihood of dropping
out. Mobile students experience
several challenges after school
closure such as forming new
relationships and adjusting to
different academic norms and
expectations.
Mantzicopoulos, P. &
Knutson, D (2000). Head
start children: School
mobility and achievement in
the early grades. The Journal
of Educational Research,
93(5), 305-311.
• Midwestern district
• Longitudinal data: a
sample of 90 children
who participated in
Head Start and
attended public
school • Grades K-2
Non-structural
mobility • Number of
school changes over
three years
• Academic outcomes: test
scores (second grade
academic achievement) •
HLM, Logistic regression
Frequent school changes were
associated with lower
achievement levels.
Nelson, P. S., Simoni, J. M.,
& Adelman, H. S. (1996).
Mobility and school
functioning in the
early grades. Journal of
Educational Research, 89(6),
365-369.
• Large urban district
• Longitudinal data:
sample of 2500
kindergarten and first
graders from 24
elementary schools
tracked for three
years
• Grades K-4
Non-structural
mobility
• Whether a student
changed schools in a
school year
• Total number of
moves over a three-
year period
• Academic and non-
academic outcomes: reading
and math grades, work
habits and cooperation
Student mobility is related to
poor behavior ratings and hailing
from single-parent families.
Schools with the highest rates of
mobility also have greater
proportions of minority and low-
income students.
Ou, S & Reynolds, A. (2008).
Predictors of educational
attainment in the Chicago
Longitudinal
Study. School Psychology
Quarterly, 23(2), 199-229.
• Chicago, Illinois
• Longitudinal data:
Chicago Longitudinal
Study (over 1000,
low income African
American students)
• Preschool through
Age 20
Non-structural
mobility
• Number of moves (
one, two and three or
more)
• Academic outcomes: high
school dropout
• OLS regression model
with controls for prior
achievement,
socioeconomic status,
parent education and other
family characteristics,
logistic regression
Student mobility is a strong
predictor of educational
attainment. Enrollment in a
magnet high school is associated
with higher educational
attainment
175
Parke, C. S., & Kanyongo, G.
Y. (2012). Student
attendance, mobility, and
mathematics
achievement in an urban
school district. Journal of
Educational Research,
105(3), 161-175.
• Large northeastern
urban school district
• Cross-sectional
data: 2004-05 school
year for all 80 schools
in the district
• Grades 1-12
Non-structural
student mobility
(within year)
• Whether student
changed schools in a
school year
• Academic outcomes: Math
and reading test scores
• Chi-square analyses,
ANOVA
Student mobility has a negative
relationship with math
achievement. There is no
evidence of differential effects
across ethnicities.
Scherrer, J. (2013). The
Negative Effects of Student
Mobility: Mobility as a
Predictor, Mobility as a
Mediator. International
Journal ofEducation Policy &
Leadership 8 (1).
• Nationwide & a
medium-sized urban
district in the
Northeast•
Longitudinal data:
ECLS-K & data on
21 elementary
schools in northeast
school district •
Grades K-8
Non-structural
mobility • Whether a
student changed
schools sometime
between third and
fifth grade
• Academic outcomes: fifth
grade reading achievement •
HLM and OLS regression
models
Student mobility is a significant
predictor of 5th grade reading
achievement and mobile students
are predicted to score lower than
stable counterparts. Mobility has
a negative relationship with
school reading achievement.
Schwartz, A.E., Stiefel, L., &
Chalico, L., (2009). The
multiple dimensions of
student mobility
and implications for
academic performance:
evidence from New York
City elementary and middle
school students. A condition
report for the New York
education finance research
consortium. Institute for
Education and Social Policy,
New York, NY.
• New York City,
New York
• Longitudinal data:
1995-96 to 2007-08
on students in NYC
public schools
(sample of eighth
graders who have
potentially been in the
district continuously
since 3rd grade)
• Grades 1-8
Non-structural
(within year and
between school
years; cumulative
mobility) and
structural (attending
the last grade offered
by a school or if 80%
or more of a
student’s cohort also
changed schools)
mobility
• Whether student
changed schools in
an academic year
• Academic outcomes: 8th
grade reading
• Regression model with
controls for lagged reading
score from 3rd grade,
student demographics, types
of mobility and student
fixed effects
Student mobility has a
significant negative effect on 8th
grade reading scores - every
additional move results in a
0.054 SD decline in reading
scores; changing schools three
times by 8th grade lowers
reading scores by 0.162 SD; 1
during the year move has a
significant -0.022SD effect and 3
moves during the year over the
period has a -0.088SD
176
Temple, J & Reynolds, A.
(1999). School mobility and
achievement: Longitudinal
findings
from an urban cohort.
Journal of School
Psychology, 37(4), 355-377.
• Chicago, Illinois
• Longitudinal data: a
sample of African
American students in
the CLS
• Grades K-7
Non-structural
mobility
• Number of times a
student changed
school between
kindergarten and
seventh grade
• Whether the student
changed schools four
or more times
• Academic outcomes:
reading and mathematics
achievement
• Ordered probit regression,
OLS regression models with
controls for student
characteristics
Students who moved frequently
between kindergarten and
seventh grade performed about a
grade level behind their stable
peers (mobility reduced test
scores by roughly 0.10SD). One
half of this difference can be
attributed to the lower
achievement of movers before
switching schools.
Wright, D (1999) Student
Mobility: A Negligible and
Confounded Influence on
Student Achievement. The
Journal of Educational
Research, 92(6), 347-353.
• Midwest Urban
Center• Cross
sectional data: 3rd
and 4th graders in 33
elementary schools •
Grades 3-4
Non-structural
mobility (within-
year)
• Academic outcomes:
reading and mathematics
test scores • ANOVAs
Student mobility is confounded
by family income and ethnicity.
177
Table A2
Description of data fields used in study
Variable Definition
Student-Level
Current Math Achievement A student’s test score (reading and math) in a
school year (standardized relative to school and
district)
Prior Math Achievement A student’s test score in the previous school year
(only available for grades 4-10)
Below average student (school) =1 if student’s test score is below their school’s
average test score
Below average student (district) =1 if student’s test score is below the district’s
average test score
Race/Ethnicity Black, Hispanic, Asian, White
Gender =1 if student is male and 0 if student is female
Free and Reduced Price Lunch (FRPL) status =1 if student qualifies for free and reduced lunch
and 0 otherwise
English Language Learner (ELL) status =1 for ELL status, 0 otherwise
Special Education status =1 for Special education status and 0 otherwise
Repeater (grade repetition) = 1 if student repeated grades in the school year, 0
otherwise
Skipped grade =1 if student had grade promotion in the school
year, 0 otherwise
Movers A student who changed schools at least once
between OR during the school year
Structural movers A student who changed schools due to the structure
of the education system. There are two main
structural moves: the transition from elementary to
middle school (grade 5 to 6) and the transition from
middle to high school (grade 8 to 9). It is assumed
that all structural moves occur between school
years
Between-year non-structural movers A student who made a non-structural move
between school years
During the year switcher A student who switched schools at least once
178
during the school year and remained in the district
for the entire school year. The student was enrolled
in the district at the end of the school year
Ultra-movers A student who changed schools at least once
between AND during the school year
Reason for mobility = three indicators including: residential change
(using school zip codes), behavior-related moves,
and AYP-designation related move
Reenrollment lag time = the number of days between when a student
leaves their old school and when they enroll in a
new school
School-Level
Student demographics Percentage of Black, Asian, White, Hispanic,
Nonwhite, FRPL, ELL, and Special Education in a
school
School quality Percentage of proficient students in a school
School enrollment = the natural log of schools’ total enrollment
School mobility rate (average school turnover) = percent of nonstructural mobile students
(calculated with all moves combined and separately
by the timing of school changes)
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Student mobility in policy and poverty context: two essays from Washington
PDF
Superintendents and Latino student achievement: promising practices that superintendents use to influence the instruction and increase the achievement of Latino students in urban school districts
PDF
The teacher evaluation: key components that positively impact teacher effectiveness and student achievement
PDF
The role of the superintendent in raising student achievement: a superintendent effecting change through the implementation of selected strategies
PDF
How urban school superintendents effectively use data-driven decision making to improve student achievement
PDF
The impact of leadership on student achievement in high poverty schools
PDF
Equity and access: the under-identification of African American students in gifted programs
PDF
Secondary school reform and improved math achievement: a case study of site efforts at Mission Valley High School
PDF
More than sanctions: California's use of intensive technical assistance in a high stakes accountability context to close achievement gaps
PDF
Allocation of educational resources to improve student learning: case studies of California schools
PDF
Household mobility and neighborhood impacts
PDF
Systemic change and the system leader: a case study of superintendent action to improve student achievement in a large urban school district
PDF
Educational resource allocation at the middle school level: a case study of six middle schools in one California district
PDF
Perceived impact of Masters in Governance training on student achievement and governance
PDF
The superintendent and reform: a case study of action by the system leader to improve student achievement in a large urban school district
PDF
Urban schools that have narrowed the achievement gap: middle school math achievement in an urban setting
PDF
The impact of programs, practices, and strategies on student academic performance: a case study
PDF
Are there hidden costs associated with teacher layoffs? The impact of job insecurity on teacher effectiveness in the Los Angeles Unified School District
PDF
Perceived factors for narrowing the achievement gap for minority students in urban schools: cultural norms, practices and programs
PDF
Narrowing the achievement gap: Factors that support English learner and Hispanic student academic achievement in an urban intermediate school
Asset Metadata
Creator
Welsh, Richard Osbourne
(author)
Core Title
The role of the timing of school changes and school quality in the impact of student mobility: evidence from Clark County, Nevada
School
Rossier School of Education
Degree
Doctor of Philosophy
Degree Program
Urban Education Policy
Publication Date
07/31/2016
Defense Date
06/23/2015
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
achievement gaps,changing schools,Clark County,impact of student mobility,mobility effects,mobility rate,Nevada,OAI-PMH Harvest,school accountability,school changes,school mobility,school quality,school segregation,school transitions,student achievement,student mobility,student segmentation,Tiebout improvement,timing of student mobility
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Brewer, Dominic J. (
committee chair
), Painter, Gary Dean (
committee member
), Strunk, Katharine O. (
committee member
)
Creator Email
richarow@usc.edu,rowelsh@uga.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-624039
Unique identifier
UC11303974
Identifier
etd-WelshRicha-3785.pdf (filename),usctheses-c3-624039 (legacy record id)
Legacy Identifier
etd-WelshRicha-3785.pdf
Dmrecord
624039
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Welsh, Richard Osbourne
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
achievement gaps
changing schools
impact of student mobility
mobility effects
mobility rate
school accountability
school changes
school mobility
school quality
school segregation
school transitions
student achievement
student mobility
student segmentation
Tiebout improvement
timing of student mobility