MANY THE MILES TO SCHOOL: THE ROLE OF RESIDENTIAL AND SCHOOL 

LOCATION IN CHOOSING AND GOING TO SCHOOL 

 
By 
 

Danielle Michelle Edwards 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 

 

2021

A DISSERTATION 

Submitted to  

Michigan State University 

in partial fulfillment of the requirements 

for the degree of 

Education Policy—Doctor of Philosophy 

ABSTRACT 

 

 

 
By 
 

Danielle Michelle Edwards 

MANY THE MILES TO SCHOOL: THE ROLE OF RESIDENTIAL AND SCHOOL 

LOCATION IN CHOOSING AND GOING TO SCHOOL 

 

Over the last 30 years, school choice policies have weakened the link between home 

residence and school assignment by allowing students to attend schools other than their 

neighborhood school. In theory, school choice policies can raise student achievement by 

increasing access to effective schools and by creating competitive pressure for schools to 

improve their academic quality. However, geographic factors may act as barriers to participating 

in school choice policies and constrain access to effective schools. To date, little attention has 

been paid to how geography shapes participation in and effectiveness of school choice policies. 

 

In this dissertation, comprised of three papers, I provide some of the first evidence 

concerning the roles of distance, residential mobility, school district boundaries, and access to 

transportation in participation in formal school choice programs and access to effective schools. 

Also, I estimate the impacts of school transportation—a policy that can mitigate the negative 

effects of these geographic factors on student outcomes. I examine these relationships in 

Michigan where students have been able to participate in inter-district and charter school choice 

for over 25 years. I use student-level enrollment, achievement, and address records for Michigan 

public school students over seven years to describe geographic inequities in participation in 

choice use and access to effective schools as well as to estimate the effects of the school bus on 

student attendance and achievement.  

In my first paper, I estimate a set of hazard models to determine the relationships between 

residential mobility, commute time to school, and exit from school choice programs. I find that

 the majority of exits from school choice programs correspond to a residential move. 

Furthermore, the probability that a student exits charter school and inter-district choice programs 

increases as the time spent commuting to school past their assigned school increases. These 

findings establish that participation in school choice policies can be determined by where schools 

are located in relation to students’ residences. 

Even where school choice participation is widespread, geographic factors may still 

constrain access to effective schools. In my second paper, I investigate whether students living in 

Detroit attend the highest quality schools in their choice sets, as determined by levels of and 

contributions to achievement, using a set of discrete choice models. I find that students are more 

likely to attend the higher quality schools in their choice sets when their choice sets are restricted 

to schools located within Detroit, implying that access to effective schools is constrained by 

geographic factors.  

In addition to influencing access to effective schools, geographic factors can also affect 

student outcomes. In my final paper, I exploit the walking distance cutoffs that determine 

transportation eligibility to provide some of the first causal evidence of the effects of school 

transportation on student attendance and achievement using a regression discontinuity design. I 

find that transportation eligibility increases attendance rates and decreases the probability of 

being chronically absent especially for disadvantaged students. However, my results provide no 

evidence that school transportation affects achievement. 

Taken together, the findings of this dissertation provide substantial evidence that where 

students live in relation to where they go to school affects their educational opportunities and 

outcomes. I also show that public policy has the potential to mitigate the negative effects of these 

relationships. 

My shotgun rider on my drive through graduate school.

To David. 

 

iv 

ACKNOWLEDGEMENTS 

 
 
 

 

The road to my degree has been filled with many roadblocks, bypasses, detours, and pit 

stops. Along my route, I have met many intelligent and caring people that have assisted me in 

my research, influenced my thinking, and helped me overcome some of my greatest challenges. 

My successes would not be possible without these students, faculty members, friends, and family 

members. I am ever grateful for your support and am honored that you chose to spend your time 

and effort investing in me.  

 

First, I would like to acknowledge those who provided funding and data access for and 

helped collect the data used in this dissertation. Papers 1 and 3 in this dissertation are supported 

by the Institute of Education Sciences, U.S. Department of Education, through Grant 

R305C100025 to The Administrators of the Tulane Educational Fund. The opinions expressed 

are those of the author and do not represent views of the Institute or the U.S. Department of 

Education. Additionally, this dissertation used data structured and maintained by the MERI-

Michigan Education Data Center (MEDC). MEDC data is modified for analysis purposes using 

rules governed by MEDC and are not identical to those data collected and maintained by the 

Michigan Department of Education (MDE) and/or Michigan’s Center for Educational 

Performance and Information (CEPI). Results, information and opinions solely represent the 

analysis, information, and opinions of the author and are not endorsed by, or reflect the views or 

positions of, grantors, MDE, CEPI or any employee thereof. All errors are my own. Additionally, 

I would like to thank Olivia Czajka, Alex Johann, and Mallory Weiner for assistance in 

collecting district transportation policies and data from the Detroit Parents’ Guide to Schools.  

 

v 

 

Research support not only comes from those directly involved, but from the staff who 

ensure we meet all of the requirements and have all the materials we need to do our work. Thank 

you Joni Burns for helping me navigate the paperwork associated with each of my program 

requirements during my first four years of graduate school. I did not realize how much easier you 

made my life until you had moved onto bigger and better things. Michelle Huhn not only helped 

me plan events, schedule meetings, and find the resources I needed; she was a sympathetic ear 

during some of my hardest days of graduate school. Michelle, you are invaluable to the faculty, 

staff, and students you serve. I am ever grateful for your support.  

Multiple peers have improved my work through collaboration, feedback, and countless 

conversations. I would like to thank Amy Auletto, Jason Burns, Tara Kilbride, Ijun Lai, and 

Jesse Wood for the hours they have spent staring at whiteboards with me, listening to my ideas, 

and critiquing them. From calculating commute times to defining what makes a school a school 

to perfecting the modeling of a stacked differences-in-differences design, you all have pushed me 

to think deeply about how to conduct rigorous quantitative research. In particular, I would like to 

thank Kaitlin Anderson. She not only provided me guidance on coding and some of the first 

feedback on my work, but she has shaped how I approach data analysis problems in my research. 

To this day, I think “What would Kaitlin suggest I do?” when I am unsure how to handle data 

issues. Sandy Frost Waldron has given me an immeasurable amount of professional and personal 

support integral to my success. Thank you Sandy for providing honest feedback on my 

professional decisions, dropping everything when I needed to talk, and even driving through the 

middle of the night to get us to AERA. 

 

Each of my committee members has contributed to my development as a researcher and 

scholar. Amita Chudgar’s excellent teaching gave me a strong base of knowledge concerning the 

 

vi 

enduring issues of and methods used in the economics of education of literature. She ignited my 

interest in understanding how where students live and where they go to school affect their access 

to effective schools and teachers. Scott Imberman built on the foundation of knowledge Amita 

laid by exposing me to cutting edge papers and methods in the economics discipline. His honest 

feedback on my research has ensured that it answers important questions in the most rigorous 

way possible. Furthermore, his support and promotion of my research and career has opened 

many doors for me in the economics of education world. Thank you, Amita and Scott. 

 

Amy Ellen Schwartz was one of the first scholars outside of MSU to recognize my talent 

and go out of her way to nurture my development. The excitement she expressed towards my 

research while we were in Barcelona encouraged me to submit my work to top-tier journals. 

Furthermore, her mentorship and own research have kept my line of work concerning school 

transportation moving forward. Amy, I hope to continue to work with you to understand how 

education, transportation, and housing policies affect students’ lives.  

 

Josh Cowen has served as my navigator as I traveled down the long and winding roads of 

graduate school. Every time I reached a fork in the road, Josh gave me unbiased, professional 

advice and empowered me to make informed decisions about my career. When I hit a roadblock, 

he got out of the car and helped me clear my path to success. Most importantly, he made me 

“practice on the concrete” by giving me the opportunity to work on high stakes projects with 

meaningful implications not only for my career but for real-world education policy. He trusted 

me to drive his work on school transportation with the Urban Institute, provide accurate 

information to policymakers when his research practice partnership was establishing itself, and 

take the lead on multiple grant-funded research projects. Josh, you serve as one of the best 

examples of what success in this field looks like. Learning from you and working with you has 

 

vii 

been my most rewarding professional relationship to date. I look forward with great excitement 

to what we will both accomplish as we transition from student and advisor to colleagues and 

friends. 

 

I would not have been able to start on the path to a Ph.D. without the unconditional love 

and support of my parents, Mandee and Dan Sanderson, and my brother, Adam. The time and 

treasure they invested in me has allowed me to live a life that is better than anything that I could 

have imagined. The countless hours spent traveling to softball tournaments, quizzing me on the 

capitals of all the countries and states, and arguing with me strengthened me intellectually and 

emotionally. I hope to provide as many opportunities to future generations as they have for me. 

Thank you. 

 

 The one person that this dissertation would not have been possible without is my loving 

husband, David Edwards. His strong, consistent, and supportive presence has allowed me to 

flourish not only in my professional life but in my personal life. He has sacrificed his time and 

money to ensure that all my basic needs are met so I can do this important work. David has 

listened to countless presentations and research ideas—he probably can explain a regression 

discontinuity design better than many people with a doctorate. He has provided the 

encouragement and emotional support necessary for me to overcome some of the greatest 

challenges I faced in graduate school. Not only has he been my biggest cheerleader, but he is my 

most influential critic, challenging my thoughts and actions when he knows that I can do better. 

David, you are the best example of what it means to love and serve someone unconditionally. I 

hope I can support you and your endeavors at least half as much as you have mine. 

 

viii 

TABLE OF CONTENTS 

 
 
 

LIST OF TABLES ......................................................................................................................... xi 
 
LIST OF FIGURES ..................................................................................................................... xiii 
 
INTRODUCTION ........................................................................................................................... 1 
REFERENCES ................................................................................................................................ 7 
 
PAPER 1 THE FARTHER YOU GO, THE CLOSER YOU GET: THE ROLES OF 
RESIDENTIAL MOBILITY AND DISTANCE IN PARTICIPATION IN PUBLIC SCHOOL 
CHOICE ........................................................................................................................................ 10 
Introduction ............................................................................................................................... 10 
Background: Determinants of Participation in Public School Choice Programs ...................... 15 
Who Chooses? ....................................................................................................................... 15 
Who Leaves? ......................................................................................................................... 16 
Residential Mobility .............................................................................................................. 18 
Residential Mobility and Participation in School Choice ................................................. 19 
Distance ................................................................................................................................. 19 
Prior Evidence Concerning Commutes to School ............................................................. 22 
Context: Michigan, A Mature School Choice Market ............................................................... 22 
Data ............................................................................................................................................ 25 
RQ 1: How often do students participating in formal school choice policies change residences? 
Are residential moves associated with participation in school choice programs? .................... 28 
RQ 2: How far do students travel to school? How does this differ for students participating in 
formal school choice programs? ................................................................................................ 33 
RQ3: What are the roles of residential mobility and commute time to school in continued 
participation in formal school choice policies? ......................................................................... 39 
Results ................................................................................................................................... 44 
Discussion .................................................................................................................................. 49 
APPENDIX ................................................................................................................................... 54 
REFERENCES .............................................................................................................................. 57 
 
PAPER 2 JUST OUT OF REACH? UNRESTRAINED SUPPLY, CONSTRAINED DEMAND, 
AND ACCESS TO EFFECTIVE SCHOOLS IN AND AROUND DETROIT ............................ 65 
Introduction ............................................................................................................................... 65 
Motivation: Supply of, Demand for, and Access to Effective Schools in Choice-Rich Areas . 69 
Supply .................................................................................................................................... 69 
Demand .................................................................................................................................. 70 
Parental Preferences .......................................................................................................... 71 
Barriers to Access .............................................................................................................. 73 
Accurate and Transparent Information .............................................................................. 74 
Background on School Choice in Michigan and Detroit ........................................................... 75 
Data ............................................................................................................................................ 78 

 

ix 

Measures of Sector and Location .......................................................................................... 79 
Measures of Poverty and Access to Transportation .............................................................. 80 
Academic Quality Measures .................................................................................................. 82 
School Characteristics ........................................................................................................... 85 
Methods ..................................................................................................................................... 86 
RQ 1: What are the roles of poverty, access to transportation, and student’s assigned school 
in use of school choice? ......................................................................................................... 86 
RQ2: Do Detroit students attend the more effective schools and/or the schools with the 
highest levels of achievement within their choice sets? ........................................................ 88 
Results ....................................................................................................................................... 92 
RQ 1: What are the roles of poverty, access to transportation, and student’s assigned school 
in use of school choice? ......................................................................................................... 92 
RQ2: Do Detroit students attend the more effective schools and/or the schools with the 
highest levels of achievement within their choice sets? ........................................................ 95 
Heterogeneity by Location and Sector .............................................................................. 96 
Discussion ................................................................................................................................ 105 
REFERENCES ............................................................................................................................ 107 
 
PAPER 3 ANOTHER ONE RIDES THE BUS: THE IMPACT OF SCHOOL 
TRANSPORTATION ON STUDENT OUTCOMES IN MICHIGAN ...................................... 114 
Introduction ............................................................................................................................. 114 
Background: The Policy Relevance of Student Attendance .................................................... 116 
How Can School Transportation Affect Attendance and Achievement? ................................ 118 
Prior Research on School Transportation and Student Outcomes ........................................... 120 
School Transportation in Michigan ......................................................................................... 121 
Data .......................................................................................................................................... 122 
District Transportation Policies ........................................................................................... 122 
Analytic Sample .................................................................................................................. 126 
Treatment and Forcing Variables ........................................................................................ 127 
Attendance and Achievement Measures ............................................................................. 128 
Method ..................................................................................................................................... 130 
Validity of Design ............................................................................................................... 132 
Results ..................................................................................................................................... 137 
Specification Checks ........................................................................................................... 143 
Falsification Test ................................................................................................................. 146 
Discussion ................................................................................................................................ 147 
APPENDIX ................................................................................................................................. 152 
REFERENCES ............................................................................................................................ 158 

 

 

 

 

x 

LIST OF TABLES 

 
 
 

Table 1. 2017-18 Student Characteristics by Choice Use ............................................................. 27 

Table 2. Residential and Choice Mobility Behaviors by Initial Choice ........................................ 29 

Table 3. 2016-17 Student Characteristics by Residential Mobility between 2016-17 and 2017-18
 ....................................................................................................................................................... 31 

Table 4. Residential Behaviors by Initial Choice for Students who Leave their Initial Choice .... 32 

Table 5. 2017-18 Average Commute Times and Distances to Attended and Nearest School by 
Grade ............................................................................................................................................. 35 

Table 6. Additional Distance Traveled Past Nearest School by Grade, Locale, and Choice ........ 39 

Table 7. 2012-13 Kindergarten Student and School Characteristics by Mobility and Initial 
Choice ............................................................................................................................................ 41 

Table 8. Predicting Exit from Initial Choice from Kindergarten to 5th Grade: Full Sample ......... 46 

Table 9. Predicting Exit from Initial Choice from Kindergarten to 5th Grade: Same Residence 
Sample ........................................................................................................................................... 50 

Table 10. Student and Neighborhood Characteristics by Sector of School Attended ................... 82 

Table 11. Student Weighted Assigned School Characteristics by Sector of School the Student 
Attends ........................................................................................................................................... 86 

Table 12. Sector of Attendance by Grade ..................................................................................... 89 

Table 13. School Characteristics by Sector and Grade ................................................................. 90 

Table 14. Estimated Marginal Effects of Student and Assigned School Characteristics on School 
Type Attended ............................................................................................................................... 94 

Table 15. Conditional Logit Predictions of School Attendance for School Characteristics by 
Grade ............................................................................................................................................. 96 

Table 16. Percent of Students Who Have Ever Attended Another Sector by Sector and Location 
between 2012-13 and 2017-18 ...................................................................................................... 98 

Table 17. Percent of Students Attending a Different Sector Last Year by Sector and Mobility 
Type ............................................................................................................................................. 100 

Table 18. Conditional Logit Predictions of School Attendance for School Characteristics by 
Grade and Location ..................................................................................................................... 102 

 

xi 

Table 19. District Transportation Policy Provisions ................................................................... 124 

Table 20. District Characteristics of Sample, 2017-18. ............................................................... 125 

Table 21. Student Characteristics of Analytic Sample ................................................................ 130 

Table 22. Estimated Coefficients of Transportation Eligibility for Balance Tests ..................... 134 

Table 23. Estimated Effects of Transportation Eligibility ........................................................... 141 

Table 24. Heterogeneous Effects of Transportation Eligibility by Economically Disadvantaged 
Status ........................................................................................................................................... 142 

Table 25. Estimated Effects of Transportation Eligibility, Donut Regressions .......................... 145 

Table 26. Placebo Effects of Transportation Eligibility .............................................................. 148 

Table 27. Estimated Coefficients of Transportation Eligibility for Balance Tests, All Bandwidths 
and Splines ................................................................................................................................... 153 

Table 28. Estimated Coefficients of Transportation Eligibility, All Bandwidths and Splines .... 154 

Table 29. Heterogeneous Effects by Economically Disadvantaged Status, All Bandwidths and 
Splines ......................................................................................................................................... 155 

Table 30. Estimated Effects of Transportation Eligibility, Non-Parametric Approach .............. 156 

 

 

 

 

xii 

LIST OF FIGURES 

 
 
 

Figure 1. 2017-18 Participation in School Choice by District of Residence ................................. 24 

Figure 2. Average Commute Time to Attended School by Grade, Locale, and Choice ............... 36 

Figure 3. Commute Time to Nearest School By Grade, Locale, and Choice ................................ 37 

Figure 4. Kaplan-Meier Survival Probabilities for All Students (Survival=remain in initial 
choice). Kindergarten to 5th Grade ................................................................................................ 42 

Figure 5. Kaplan-Meier Survival Probabilities by Residential Mobility (Survival=remain in 
initial choice). Kindergarten to 5th Grade ...................................................................................... 42 

Figure 6. Kaplan-Meier Survival Probabilities by Attended School Driving Time Quartile 
(Survival=remain in initial choice). Kindergarten to 5th Grade ..................................................... 43 

Figure 7. Kaplan-Meier Survival Probabilities by Attended School Driving Time Quartile 
(Survival=remain in initial choice). Kindergarten to 5th Grade for Same Residence Sample ....... 55 

Figure 8. Kaplan-Meier Survival Probabilities Additional Driving Time Quartile 
(Survival=remain in initial choice). Kindergarten to 5th Grade ..................................................... 56 

Figure 9. Distribution of Observations ........................................................................................ 135 

Figure 10. Unadjusted Average Attendance Rate and Proportion Chronically Absent by Distance 
from Threshold ............................................................................................................................ 138 

Figure 11. Unadjusted Average Standardized Math and ELA Test Scores by Distance from 
Threshold ..................................................................................................................................... 139 

Figure 12. Distribution of Observations without Mobile Home Park Census Block .................. 157 

 

 

xiii 

INTRODUCTION 

 

Where students live and where they attend school affect their long-term outcomes. Living 

in neighborhoods with higher incomes and levels of educational attainment increases enrollment 

in college preparatory courses in high school, college attendance, and adult earnings in addition 

to decreasing the likelihood of single parenthood (Chetty, Hendren, & Katz, 2016; Chetty & 

Hendren, 2018; Rosenbaum, 1995). One possible mechanism that can explain neighborhood 

differences in outcomes is the strength of the schools located within neighborhoods. In fact, 

Chetty, Hendren, Kline, and Saez (2014) find that areas that contain schools with higher test 

scores and graduation rates have higher levels of economic mobility. Traditionally, students are 

assigned to schools by their residence making it difficult to separate the contributions of 

neighborhoods and schools to student outcomes. However, recent evidence leveraging policies 

that allow students to attend schools outside of their neighborhoods confirms that both 

neighborhoods and schools contribute to student achievement. Nonetheless, the effect of schools 

on achievement is larger (Carlson & Cowen, 2015).  

 

Over the last 30 years, many cities and states have implemented school choice policies 

that break the traditional link between residence and school assignment by allowing students to 

attend schools other than their residentially assigned school. In theory, school choice policies can 

increase access to desirable schools for families unable to afford to live near these schools. They 

can also increase student achievement if families prefer academic effectiveness, possess accurate 

information on school quality, and have the ability to physically access effective schools. 

Evidence concerning the effectiveness of school choice policies is mixed. Research examining 

the effects of charter schools on achievement in large cities like Boston and New Orleans finds 

that they positively impact the test scores of historically disadvantaged students (Abdulkadiroglu, 

 

1 

Angrist, Dynarski, Kane, & Pathak, 2011; Harris & Larsen, 2016). In contrast, recent studies of 

statewide voucher programs show that they have negative effects on student achievement (Figlio 

& Karbownik, 2016; Mills & Wolf, 2018; Waddington & Berends, 2018). Little evidence exists 

concerning the effects of inter-district choice with a handful of studies showing that those who 

consistently use these policies have higher test scores (Carlson & Lavertu, 2017; Carlson, 

Lavery, & Hughes, 2018).  

 

Although some evidence indicates that school choice policies are effective in raising 

achievement, relationships between residential and school location may constrain widespread 

participation in school choice programs, its ability to increase access the effective schools as well 

as its effectiveness improving student outcomes. Families may not take advantage of school 

choice programs if they value proximity to home when choosing schools and there are few 

alternatives to their neighborhood option around them. Even if families have strong preferences 

for school effectiveness, burdensome enrollment policies, lack of sufficient transportation, and 

residential mobility may constrain students’ abilities to attend preferred schools. Furthermore, 

traveling long distances to school may have deleterious effects on student outcomes independent 

of choice of school especially if there is no school provided transportation.  

 

In this dissertation, comprised of three papers, I examine how geographic factors, 

including distance from home to school, residential mobility, district boundaries, and access to 

transportation promote or restrict participation in school choice programs and constrain access to 

effective schools. Furthermore, I evaluate a policy that could mitigate the possible negative 

effects of these factors on student attendance and achievement: school provided transportation. I 

investigate the relationships between where students live, where they go to school, and their 

outcomes in Michigan where public school choice programs are prevalent and well-established 

 

2 

in all regions of the state. Michigan students can attend their assigned school, other schools in 

their resident district, schools in other traditional public school districts, and charter schools. 

Almost a quarter of Michigan public school students participate in formal school choice 

programs. Additionally, charter school and inter-district choice have existed in Michigan for over 

twenty-five years, making it a mature and stabilized education market. Furthermore, there exists 

a substantial amount of variation in the supply of charter schools, the rules that govern inter-

district choice, and the provision of transportation across the state, providing an ideal setting to 

understand how geographic factors interact with policy in a choice-rich environment.  

 

In each of my three papers, I primarily use student-level enrollment, achievement, and 

address records for all Michigan public school students from the 2012-13 to the 2018-19 school 

years provided by the Michigan Department of Education (MDE) and Center for Educational 

Performance and Information (CEPI). I employ a variety of quantitative and quasi-experimental 

methods including survival analysis, discrete choice models, and a regression discontinuity 

design to describe the roles of residential and school location in participation in school choice 

programs and estimate the causal effects of school transportation on student attendance and 

achievement. I note that each of these papers was prepared as separate articles for publication in 

peer-reviewed journals. Therefore, each paper has its own literature review, description of the 

context, and explanation of the data even though many of these aspects are shared across all 

three papers.  

 

In Paper 1, I describe the roles of residential mobility and distance from home and school 

in participation in and exit from public school choice programs. First, I provide some of the first 

evidence concerning differences in residential mobility between students who participate in 

school choice and those who do not. Second, I estimate the commute times from home to school 

 

3 

for all Michigan students and examine differences in commute times for students who use charter 

school and inter-district choice. These estimated commute times are some of the first 

descriptions of how long it takes students to travel to school outside of large cities. Finally, I 

estimate a set of hazard models to investigate the roles of residential mobility and commute time 

in exit from public school choice programs. I find that most exits from school choice programs 

are accompanied by residential moves. Furthermore, students have a higher probability of 

leaving inter-district and charter school choice as the additional commute time to the attended 

school past the assigned school increases. Taken together, these findings imply that residential 

stability and proximity to home play large roles in families’ abilities to consistently participate in 

school choice programs.  

 

Geographic factors may not only influence families’ decisions to participate in or exit 

from school choice programs. Also, they could constrain access to effective schools even where 

school choice participation is widespread. Prior research shows that families in choice-rich cities 

value proximity to home just as much if not more than academic quality when choosing schools 

(Denice & Gross, 2016; Glazerman & Dotter, 2017; Harris & Larsen, 2019; Hastings, Kane, & 

Staiger, 2005; Lincove, Cowen, & Imbrogno, 2018). However, this preference for proximity may 

represent an inability to attend schools farther from home, especially for disadvantaged students. 

In Paper 2, I test whether distance to school and district boundaries constrain access to schools 

with higher levels of and contributions to achievement for students living in Detroit where one in 

five students attend a school located outside of the city, almost one half attend a charter school, 

and less than a quarter of students attend their assigned school. First, I show that students living 

in neighborhoods with lower incomes and rates of car ownership are less likely to participate in 

public school choice or have access to the highest quality schools using a set of multinomial 

 

4 

logistic regressions. Then, I examine whether students in Detroit attend schools with higher 

levels of and contributions to achievement using discrete choice models. I find that families’ 

preferences for academic quality are stronger when their choice sets are restricted to schools 

within Detroit. These findings imply that geographic factors likely constrain access to effective 

schools even where school choice use is prevalent. 

 

In addition to shaping participation in school choice and access to effective schools in 

choice-rich environments, the relationship between residential and school location may affect 

student attendance and achievement. In fact, prior research shows that longer commute times 

have a negative relationship with attendance but little association with achievement (Stein & 

Grigg, 2019; Blagg, Rosenboom, & Chingos, 2018). One intervention that can mitigate the 

effects of distance to school on student outcomes is school transportation. By removing the 

burden of transporting students to and from school from families, the school bus can make it 

easier for students to get to school, thus increasing their attendance and possibly their 

achievement. In Paper 3, I provide some of the first causal evidence of the effects of the school 

bus on student attendance and achievement using a regression discontinuity design that exploits 

the walking distance cutoffs that determine transportation eligibility. I find that transportation 

eligibility increases attendance rates and reduces the probability of being chronically absent. 

These effects are largest for economically disadvantaged students—the students who are least 

likely to have access to private forms of transportation and most likely to be chronically absent 

(Urban Institute Student Transportation Working Group, 2018; Balfanz & Byrnes, 2012). My 

results provide little evidence that transportation eligibility affects achievement. These findings 

imply that increasing the amount of transportation provided may be an appropriate intervention 

 

5 

to reduce chronic absenteeism. However, widespread provision of school transportation may not 

be the most cost-effective or efficient way to allocate this expensive resource.  

 

Taken together, this body of work shows that school location in relation to residential 

location is a determinant of participation in school choice programs, access to effective schools, 

and student outcomes. Also, I show that its influence can be moderated by school transportation 

policies. Given my findings, future school choice research should account for geographic factors 

when describing school choice use, determining its effectiveness, and studying the consequences 

of choice policies. Additionally, more research is needed to understand how elements of school 

choice policies, in particular enrollment rules and the provision of transportation, promote or 

restrict access to effective schools, especially those located farther from home. Finally, the next 

generation of school choice research should consider residential decisions as a form of school 

choice, study the effects of these decisions on participation in and the effectiveness of school 

choice programs, and understand how local housing and public transportation policies interact 

with school choice policies.  

 

 

 

 

6 

REFERENCES 

 

7 

REFERENCES 

 
 
 

Abdulkadiroğlu, A., Angrist, J., Dynarski, S., Kane, T., & Pathak, P. (2011). Accountability and 
flexibility in public schools: Evidence from Boston's charters and pilots. The Quarterly 
Journal of Economics, 126(2), 699-748. doi: 10.1093/qje/qjr047. 

 
Balfanz, R., & Byrnes, V. (2012). The importance of being in school: A report on absenteeism in 

the nation's public schools. The Education Digest, 78(2), 4. 

 
Blagg, K., Rosenboom, V., & Chingos, M. (2018). The Extra Mile: Time to School and Student 
Outcomes in Washington, D.C. Washington, DC: Urban Institute. Retrieved on July 30, 
2020 from: https://www.urban.org/research/publication/extra-mile-time-school-and-
student-outcomes-washington-dc  

 
Carlson, D., & Cowen, J. (2015). Student neighborhoods, schools, and test score growth: 

Evidence from Milwaukee, Wisconsin. Sociology of Education, 88(1), 38-55. 

 
Carlson, D., & Lavertu, S. (2017). Interdistrict Open Enrollment in Ohio. Columbus, Ohio: 

Thomas B. Fordham Institute. Retrieved on February 24, 2020 from: 
https://fordhaminstitute.org/sites/default/files/publication/pdfs/FORDHAM%20Open%20
Enrollment%20Report_Online%20final_0.pdf  

 
Carlson, D., Lavery, L., & Hughes, T. (2018). Should I Stay or Should I Go? Open Enrollment 
Decisions and Student Achievement Trajectories. Social Science Quarterly, 99(3), 1089-
1104. 

 
Chetty, R., Hendren, N., & Katz, L. (2016). The effects of exposure to better neighborhoods on 

children: New evidence from the Moving to Opportunity experiment. American 
Economic Review, 106(4), 855-902. 

 
Chetty, R., Hendren, N., Kline, P., & Saez, E. (2014). Where is the land of opportunity? The 
geography of intergenerational mobility in the United States. The Quarterly Journal of 
Economics, 129(4), 1553-1623. 

 
Chetty, R., & Hendren, N. (2018). The impacts of neighborhoods on intergenerational mobility 

II: County-level estimates. The Quarterly Journal of Economics, 133(3), 1163-1228. 

 
Denice, P., & Gross, B. (2016). Choice, preferences, and constraints: Evidence from public 

school applications in Denver. Sociology of Education, 89(4), 300-320. 

 
Figlio, D., & Karbownik, K. (2016). Evaluation of Ohio’s edchoice scholarship program: 

Selection, competition, and performance effects. Thomas B. Fordham Institute. Retrieved 
February 24, 2020, from https://files.eric.ed.gov/fulltext/ED575666.pdf  

 

 

8 

Glazerman, S., & Dotter, D. (2017). Market signals: evidence on the determinants and 

consequences of school choice from a citywide lottery. Educational Evaluation and 
Policy Analysis, 39(4), 593-619. 

 
Harris, D. N., & Larsen, M. (2019). The identification of schooling preferences: Methods and 

evidence from post-Katrina New Orleans. Technical Report, Education Research 
Alliance for New Orleans. Retrieved on July 10, 2020 from: 
https://educationresearchalliancenola.org/files/publications/Harris-Larsen-How-Parents-
Choose-2019-07-05.pdf 

 
Harris, D. N., & Larsen, M. (2016). The effects of the New Orleans post-Katrina school reforms 
on student academic outcomes. Technical Report, Education Research Alliance for New 
Orleans. Retrieved on February 24, 2020 from: 
https://dspace.lafayette.edu/bitstream/handle/10385/2096/Larsen-
EducationResearchAllianceforNewOrleansTechnicalReport-2016.pdf?sequence=1  

 
Hastings, J. S., Kane, T. J., & Staiger, D. O. (2005). Parental preferences and school 

competition: Evidence from a public school choice program (No. w11805). National 
Bureau of Economic Research. 

 
Lincove, J. A., Cowen, J. M., & Imbrogno, J. P. (2018). What's in Your Portfolio? How Parents 

Rank Traditional Public, Private, and Charter Schools in Post-Katrina New Orleans’ 
Citywide System of School Choice. Education Finance and Policy, 13(2), 194-226. 

 
Mills, J. N., & Wolf, P. J. (2017). Vouchers in the bayou: The effects of the Louisiana 

Scholarship Program on student achievement after 2 years. Educational Evaluation and 
Policy Analysis, 39(3), 464-484. 

 
Rosenbaum, J. E. (1995). Changing the geography of opportunity by expanding residential 

choice: Lessons from the Gautreaux program. Housing Policy Debate, 6(1), 231-269. 

 
Stein, M. L., & Grigg, J. A. (2019). Missing bus, missing school: Establishing the relationship 

between public transit use and student absenteeism. American Educational Research 
Journal, 56(5), 1834-1860. 

 
Urban Institute Student Transportation Working Group. (2018). The road to school: How far 

students travel to school in Denver, Detroit, New Orleans, New York City, and 
Washington, DC. Washington, DC: Urban Institute. Retrieved on February 24, 2020 
from:https://greaterdc.urban.org/sites/default/files/publication/97151/the_road_to_school
_7.pdf  

 
Waddington, R. J., & Berends, M. (2018). Impact of the Indiana Choice Scholarship Program: 

Achievement effects for students in upper elementary and middle school. Journal of 
Policy Analysis and Management, 37(4), 783-808. 

 

9 

PAPER 1 THE FARTHER YOU GO, THE CLOSER YOU GET: THE ROLES OF 

RESIDENTIAL MOBILITY AND DISTANCE IN PARTICIPATION IN PUBLIC 

SCHOOL CHOICE  

Introduction 

Traditionally, the particular public school that students attend is determined by the 

particular place in which they live. Over the past thirty years, this inextricable link has been 

weakened by public school choice polices that allow students to attend schools according to their 

needs, values, preferences, and goals—including schools outside of residential assignment zones 

(Levin, 2015). Forty-three states have at least one charter school, and all but three states have 

policies that govern inter-district and intra-district choice, permitting students to attend schools 

in other traditional public school (TPS) districts and other schools in their district of residence 

respectively (David & Hesla, 2018; Wixom & Kelly, 2018). During the 2016-17 school year, 

over twenty percent of U.S. public school students attended a school other than their residentially 

assigned school (Wang, Rathburn, & Musu, 2019).  

Proponents of school choice argue that these policies create more equitable access to 

effective schools for families who are unable to afford homes in the most desirable districts and 

catchment zones (Levin, 2015). Also, they contend that school choice creates competitive 

pressures on existing schools to meet families’ needs and improve productivity since schools 

must attract students to maintain enrollment (Chubb & Moe, 1990; Friedman, 1962). Therefore, 

school choice policies have the potential to improve student outcomes if they allow students to 

attend more effective schools and improve the quality of the supply of schools. Although public 

school choice policies are well established and widespread, the evidence concerning their 

effectiveness to raise student achievement for all students who participate in them as well as 

 

10 

those who remain in the surrounding TPSs is mixed. On average, disadvantaged students in 

urban areas experience increases in achievement when they attend charter schools and participate 

in inter-district choice (Abdulkadiroglu, Angrist, Dynarski, Kane, & Pathak, 2011; Carlson & 

Lavertu, 2017; Carlson, Lavery, & Hughes, 2018; Harris & Larsen, 2016). Additionally, the 

majority of evidence concerning competitive effects of charter schools suggests that they do not 

negatively impact achievement for students in TPS districts (Booker, Gilpatric, Gronberg, & 

Jansen, 2008; Cordes, 2017; Imberman, 2011; Sass, 2006; Winters, 2012; Zimmer & Buddin, 

2009; Zimmer, Gill, Booker, Lavertu, Sass, & Witte, 2009).1  

Although there exists evidence of positive effects of charter schools and inter-district 

choice, especially for disadvantaged students, participation in public school choice may be 

constrained by residential and school location. In order for school choice policies to increase 

achievement, families must prefer academic effectiveness over other school features and be able 

to physically access multiple schooling options in addition to having accurate information 

concerning school quality (Chubb & Moe, 1990; Glazerman & Dotter, 2017). Although parents 

have strong preferences for academic quality, they value proximity from home just as much, if 

not more than achievement (Denice & Gross, 2016; Glazerman & Dotter, 2017; Harris & Larsen, 

2019; Hastings, Kane, & Staiger, 2005; Lincove, Cowen, & Imbrogno, 2018). Families may also 

express their preferences for academic quality through residential decisions or may be unable to 

do so due to residential instability. Furthermore, many students cannot physically access schools 

outside of their neighborhood. Low income families may not have access to sufficient 

 

1Additionally, there exists an extensive literature on the effects of private school voucher programs on student 
outcomes. In Michigan, the context of this study, there are no voucher programs since the use of public funding for 
private schools is prohibited by the Michigan State Constitution. Therefore, I focus my review of the literature on 
evaluations of public school choice policies other than where studies of private school choice are particularly 
instructive.  

 

11 

transportation to send their students to schools far from home (Urban Institute Student 

Transportation Working Group, 2018), and rural students may not have more than one accessible 

school in a reasonable distance from home (Blagg & Chingos, 2017; Catt & Shaw, 2019). 

Although distance and residential location could be determinants in participating in 

public school choice, in addition to influencing its effectiveness, little research has directly 

examined where students live in relation to where they go to school or its role in participation in 

school choice programs especially outside of large, choice-rich cities or over time. To my 

knowledge, little evidence exists concerning the role of residential mobility in participation in 

public school choice. A handful of studies describe how far students travel to school and its 

associations with attending higher quality schools and student outcomes in Baltimore, Denver, 

Detroit, New Orleans, New York, and Washington, D.C. (Blagg, Rosenboom, & Chingos, 2018; 

Cordes & Schwartz, 2018; Cowen, Edwards, Sattin-Bajaj, & Cosby, 2018; Denice & Gross, 

2018; Stein & Grigg, 2019; Urban Institute Student Transportation Working Group, 2018). 

Studies of parental preferences account for distance as one of many school characteristics parents 

may value (Denice & Gross, 2016; Glazerman & Dotter, 2017; Harris & Larsen, 2019; Hastings, 

Kane, & Staiger, 2005; Lincove, Cowen, & Imbrogno, 2018). Additionally, evidence exists 

concerning the lack of school choice options for students outside of choice-rich cities (Blagg & 

Chingos, 2017; Catt & Shaw, 2019).  

This paper describes the relationships between where students live, where they go to 

school, and participation in inter-district and charter school choice in Michigan over six years 

using a rich panel of student level enrollment, achievement, and address data. Specifically, I ask: 

 

12 

1.  How often do students participating in formal school choice policies change 

residences? Are residential moves associated with participation in school choice 

programs? 

2.  How far do students travel to school? How does this differ for students participating 

in formal school choice programs? 

3.  What are the roles of residential mobility and commute time to school in continued 

participation in formal school choice policies? 

To answer these questions, I first describe the residential mobility patterns of students 

who participate in formal school choice policies. This is a critical piece largely unaddressed in 

the student mobility and school choice literatures because families can informally choose schools 

through residential decisions and could exit school choice due to residential instability unrelated 

to preferences for effective schools. Next, I estimate commute times for all Michigan students to 

their attended school and nearest school and examine differences for those who use inter-district 

and charter school choice. Finally, I explore the roles of residential mobility and commute time 

in continued participation in school choice using a set of hazard models.  

I find that students using public school choice travel farther to school and are more likely 

to change residences. In addition, students who change residences have a higher probability of 

leaving school choice programs. In fact, residential moves accompany over half of exits from 

school choice programs in Michigan. Also, I find that the commute time relative to students’ 

assigned schools may play a role in the use of and duration in choice. Students are more likely to 

use school choice when the nearest school in the district they live in is farther from home, and 

many use it to attend schools closer to home. Additionally, students have a higher probability of 

leaving inter-district and charter school choice when the additional commute time past their 

 

13 

nearest school increases. Taken together, these findings imply that residential mobility and 

distance likely play a role in families’ school choice decisions, especially in their decision to 

remain in formal school choice programs. Future policies and research concerning school choice 

should account for the roles of residential and school location in influencing the effectiveness of 

school choice programs to improve access to effective schools and student outcomes.  

This paper extends the current literature concerning participation in school choice 

programs in multiple ways. First, it provides some of the first evidence concerning the roles 

residential mobility and distance in who participates in and, in particular, who leaves inter-

district and charter school choice programs. This paper is one also of the first to describe 

commute times for public school students outside of choice-rich cities, which have been the 

focus of nearly all school transportation studies to date. Finally, this is one of the only studies 

that examines the factors that predict exit from both inter-district and charter school choice 

within the same context.  

This paper proceeds as follows: First, I discuss the prior literature concerning who 

participates in and exits formal school choice policies. Second, I examine the literature 

concerning residential mobility, commute to school, and their relationships with school choice 

participation. Third, I provide context concerning school choice policies in Michigan, the setting 

of my study. Fourth, I describe my rich panel of student-level data used in my analyses. Next, I 

explain the methods and results of each of my research questions. Finally, I discuss the 

implications of my findings for policy and future research.  

 

 

 

14 

Background: Determinants of Participation in Public School Choice Programs 

Who Chooses? 

 

Understanding who participates in school choice gives some insight into which types of 

students may have preferences for and access to school choice programs. The earliest studies that 

describe participation in inter-district choice use district-level data to explore student flows 

between districts. They show that students from districts with high achieving students and high 

income families are more likely to leave their districts to attend even more advantaged districts 

(Carlson, Lavery, & Witte, 2011; Holme & Richards, 2009; Reback, 2008; Welsch, Statz, & 

Skidmore, 2010). More recent work describing participation in inter-district choice uses student-

level data. A study of Colorado students finds that economically advantaged students living in 

high achieving districts with fewer economically disadvantaged students have a higher 

probability of participating in inter-district choice (Lavery & Carlson, 2015). Similarly, students 

that are higher achieving and more advantaged than their peers living in the same district are 

more likely to leave their district of residence for higher achieving, smaller, and more 

advantaged districts in Ohio (Carlson & Lavertu, 2017). Findings concerning inter-district choice 

students in Michigan are more nuanced. A higher percentage of students who begin in inter-

district choice in kindergarten are White, economically advantaged, and live farther away from 

their assigned school (Edwards, 2021). In particular, kindergarteners in Detroit who leave the 

city to attend school are more likely to be White or Asian and live closer to Detroit’s borders 

(Lenhoff, Singer, Pogodzinski, & Cook, 2020). In contrast, students who begin using inter-

district choice after attending a school in their district of residence are more likely to be low 

achieving, economically disadvantaged, and underrepresented minorities (Cowen, Creed, & 

Keesler, 2015).  

 

15 

 

In general, studies of the demographics of charter school students in California, Texas, 

North Carolina, and Michigan find that a high proportion of charter school students are Black, 

economically disadvantaged, and low achieving (Bifulco & Ladd, 2006; Booker, Zimmer, & 

Buddin, 2005; Edwards & Cowen, 2019; Ni, 2012). This is most likely because the majority of 

charter schools are located in urban areas which have more historically disadvantaged students. 

As of the 2017-18 school year, 56 percent of charter schools were in cities while only a quarter 

of traditional public schools were located in urban areas (U.S. Department of Education, 

National Center for Education Statistics, 2020). Studies that compare the characteristics of 

students who attend charter schools and those who do not within urban districts find that the 

comparatively more advantaged students who are assigned to schools or districts with higher 

levels of disadvantaged students are more likely to attend charter schools (Bifulco, Ladd, & 

Ross, 2009; Ni, 2012). In sum, the literature shows that, for the most part, students participating 

in inter-district and charter school choice seem to be more advantaged than their peers.  

Who Leaves? 

 

If there is high attrition from inter-district choice and charter schools, the hypothetical 

benefits of school choice policies may not be realized. The negative effects of student mobility 

may outweigh any gains in achievement students experience from participating in school choice 

(Hanushek, Kain, & Rivkin, 2004). Furthermore, neighborhood schools may not respond to 

competitive pressures to improve if they know or perceive that students will eventually return to 

their schools (Creed, 2016). In fact, empirical evidence concerning inter-district choice 

participation and student achievement shows that students who begin in inter-district choice and 

do not return to their home districts experience increases in achievement, but those who return to 

their districts of residence experience small declines (Carlson & Lavertu, 2017; Carlson, Lavery, 

 

16 

& Hughes, 2018). Students who begin using inter-district choice after attending a school in their 

district of residence have no change in achievement on average (Cowen & Creed, 2017).  

 

 A handful of studies examine how many and which students leave school choice 

programs. Overall, disadvantaged students have a higher probability of exiting school choice. In 

a study of private school choice, Cowen, Fleming, Witte, and Wolf (2012) find that Black 

students, students with lower achievement, and students attending schools with a higher 

proportion of voucher students were more likely to exit Milwaukee’s private school voucher 

program and return to public schools. Two papers describe the characteristics of students who 

leave inter-district choice. In Colorado, four out of five students continued to enroll in a TPS 

district other than the one they lived in. Students were more likely to exit if they were low 

income and lived in low-poverty districts (Lavery & Carlson, 2015). In contrast, over 60 percent 

of students who began in inter-district choice in kindergarten had exited by 5th grade in 

Michigan. Black, economically disadvantaged, and lower achieving students, especially those 

who were attending schools with low achieving and high risk students, had a higher probability 

of exiting (Cowen, Creed, & Keesler, 2015). I directly build on the Michigan study by 

incorporating commute times and residential mobility into the model and estimating these 

models for students who begin in charter schools as well.  

 

To my knowledge, few studies directly examine the characteristics of students who exit 

charter schools. Taken together, they find that non-White, low achieving, and low income 

students are more likely to exit charter schools (Finch, Lapsley, & Baker-Boudissa, 2009; Ni, 

2012). Much of the work concerning who exits charter schools focuses on whether charter 

schools “push out” difficult to educate students. Students with disabilities and English Learners 

are no more likely to exit charter schools than their more advantaged counterparts or similar 

 

17 

students in surrounding traditional public schools (Winters, 2014; Winters, 2015). The research 

concerning whether students with low levels of achievement are more likely to exit charter 

schools is mixed. Evidence form New York City and Denver shows that lower performing 

students are more likely to exit charter schools while a study of a large Midwestern school 

district finds no statistically significant difference in achievement levels of those who leave 

charter schools (Winters, Clayton, & Carpenter, 2017; Zimmer & Guarino, 2013). However, 

these studies find that there is no difference in exit rates between charter schools and their 

surrounding traditional public schools (Nichols-Barrer, Gleason, Gill, & Tuttle, 2016; Winters, 

Clayton, & Carpenter, 2017; Zimmer & Guarino, 2013). Thus, differences in attrition between 

charter school students likely reflect the characteristics of mobile students within districts.  

Residential Mobility 

One factor that could influence participation in school choice programs is residential 

mobility. Most of the work that directly examines the effects of residential mobility focuses on 

its impact on student performance. The findings of these studies are nuanced, but they show that 

residential mobility is negatively associated with achievement especially when accompanied by a 

change in schools. Two studies using nationally representative survey data find that students that 

move schools and residences experience decreases in achievement on average but moving earlier 

during high school is associated with increases in test scores (Pribesh & Downey, 1999; 

Swanson & Schneider, 1999). In an urban district in Tennessee, Voight, Shinn, & Nation (2012) 

find that residential mobility has a negative relationship with achievement in elementary and 

middle school. Similarly, students in New York City experience decreases in test scores after 

they move neighborhoods, but these relationships are attenuated when the residential move is 

accompanied with a move to a higher quality school (Cordes, Schwartz, Stiefel, & Zabel, 2016). 

 

18 

One study estimates the causal impacts of residential and school mobility. It finds that changing 

residences has a negative effect on achievement unless the student did not change schools and 

only moved a short distance (Cordes, Schwartz, & Stiefel, 2019). 

Residential Mobility and Participation in School Choice 

 

To my knowledge, little evidence exists concerning the role of residential mobility in 

participation in formal school choice policies. However, it is likely that residential moves induce 

entry in and exit from formal school choice programs. Families may make different choices 

about schools after a residential move since families prefer schools closer to home, report having 

difficulty transporting their students to school, and are more likely to use school choice when 

their default option is farther from their residence (Denice & Gross, 2016; Edwards, 2021; 

Glazerman & Dotter, 2017; Harris & Larsen, 2019; Hastings, Kane, & Staiger, 2005; Jochim, 

DeArmond, Gross, & Lake, 2014; Lenhoff, Singer, Stokes, & Mahowald, 2021; Lincove, 

Cowen, & Imbrogno, 2018). Furthermore, residential decisions can be used to informally choose 

a school since a student’s assigned school is determined by residence in the majority of U.S. 

school districts. To test my hypothesis that many families enter into or exit from formal school 

choice programs at the time of a residential move, I investigate whether or not changes in 

residence between school years are associated with changes in use of formal school choice 

policies. This provides some of the first evidence of the relationship between school choice use 

and residential mobility. 

Distance 

In addition to residential mobility, distance to school may also promote or restrict 

participation in school choice. First, distance to school choice options can serve as a barrier to 

using formal school choice policies since it is likely that families must travel outside of the their 

 

19 

neighborhoods in order to participate in school choice. Studies of parent preferences in choice-

rich cities like Denver, New Orleans, and Washington, D.C. show that families rank schools 

closer to home higher on enrollment applications indicating that they may not be willing or able 

to leave their neighborhoods to attend more effective schools (Denice & Gross, 2016; Glazerman 

& Dotter, 2017; Harris & Larsen, 2019; Hastings, Kane, & Staiger, 2005; Lincove, Cowen, & 

Imbrogno, 2018). These revealed preferences are likely to be function of families’ abilities (or 

inabilities) to transport their children to schools farther from home. Only six states require 

districts to transport students using inter-district choice and 17 states mandate that charter school 

students are provided transportation (McShane & Shaw, 2020). In choice-rich cities, most 

parents report that they drive their children to school and that a lack of reliable transportation is a 

barrier to sending their children to the desired school and getting them to school regularly 

(Jochim et al., 2014; Lenhoff et al., 2021). Furthermore, families living in high-poverty 

neighborhoods, those who are possibly the most likely to benefit from school choice, are less 

likely to have access to a car, making it difficult to attend schools outside of their neighborhoods 

(Urban Institute Student Transportation Working Group, 2018).  

Distance may be a larger barrier to entry for students in population sparse areas since 

there are longer distances between schools and towns creating fewer schooling options proximal 

to home. If inter-district and charter school choice was made universal, about a quarter of rural 

elementary school students would have access to an additional schooling option within five miles 

from home. In contrast, almost two-thirds of urban students would have increased access to 

schools through universal public school choice (Blagg & Chingos, 2017). Furthermore, Catt and 

Shaw (2019) show that six percent of rural elementary school students and about a quarter of 

 

20 

rural high school students in Indiana do not have a magnet, charter, or private school within a 

thirty minute commute from home. 

Although distance from home to school choice options is likely to be a barrier to 

participating in school choice, long commutes to the assigned school could induce students to 

use school choice by lowering the opportunity costs of participation. For example, it is likely that 

the additional distance past the assigned school to attend a school of choice is shorter for students 

farther away from their assigned school. Some students may even be able to attend schools closer 

to home through school choice. Therefore, the additional cost in terms of distance of 

participating in school choice is likely lower for students who live farther away from their 

assigned school. In fact, the few studies that predict participation in public school choice as a 

function of distance to assigned school find that students who live farther from their nearest or 

assigned school are more likely to use inter-district and charter school choice (Bifulco, Ladd, & 

Ross, 2009; Edwards, 2021; Singer, 2020). In particular, Edwards (2021) finds that commute 

time to the nearest school plays a larger role in the decision to participate in inter-district choice 

for rural students with one in five rural students using inter-district choice to attend a school 

closer to home.  

To test my hypothesis that distance to school choice options inhibits participation in 

choice but long commutes to the assigned school could induce students to use school choice, I 

compare commute times to the attended school and the nearest school for students who 

participate in school choice and those who do not. I also calculate the percent of students using 

school choice to attend a school closer to home.  

 

21 

Prior Evidence Concerning Commutes to School 

Recent evidence concerning commute times in from choice-rich cities shows that Black 

and economically advantaged students travel farther to school, students who travel farther to 

attend school attend higher quality schools as measured by test scores, types of programs offered, 

and resources, and students who have longer commute times are more likely to have higher rates 

of absenteeism (Blagg, Rosenboom, & Chingos, 2018; Cordes & Schwartz, 2018; Cowen et al., 

2018; Denice & Gross, 2018; Stein & Grigg, 2019; Urban Institute Student Transportation 

Working Group, 2018). In particular, Stein, Burdick-Will, & Grigg (2020) predict exit from 

Baltimore high schools, which operate under an open enrollment system instead of residential 

assignment, as a function of difficulty of commute. They find that students living farther away 

from their school in 9th grade are more likely to transfer and attend schools closer to home. As 

for school choice participation, students who attend charter schools in elementary school have 

longer commute times on average, but high school students attending traditional public schools 

travel as far if not farther to school than charter school students (Urban Institute Student 

Transportation Working Group, 2018). I add to these studies by providing some of the first 

estimates of commute times to school outside of cities and use them to predict exit from inter-

district and charter school choice over time. 

Context: Michigan, A Mature School Choice Market 

 

For over twenty years, Michigan students have been able to attend charter schools, other 

schools in their district of residence, and schools in other traditional public school districts in 

addition to their assigned school. In 1994, the Michigan State Legislature enacted Part 6A of the 

Revised School Code which allows community colleges, public universities, intermediate 

schools districts (ISDs), and TPS districts to authorize charter schools (Michigan Department of 

 

22 

Education, 2017). In contrast with the majority of charter school laws, public post-secondary 

institutions may authorize charter schools located anywhere in the state without oversight from 

local governments. Therefore, there is no one body that controls where schools are located, when 

they open, or when they close. Over 80 percent of Michigan charter schools are authorized by 

universities or community colleges. Furthermore, each charter school has its own application 

process. Although Michigan law prohibits charter schools from practicing selective enrollment 

policies and stipulates that they must hold a lottery to determine admission if they are 

oversubscribed, filling out multiple applications without a guarantee of enrollment may be 

prohibitive for entry into the charter sector (Michigan Department of Education, 2017). As of the 

2017-18 school year, about one in ten Michigan public school students attended one of its 368 

charter schools.  

 

Since 1996, Michigan TPS districts have been able to enroll students from surrounding 

districts and ISDs. Under Acts 105 and 105c of the Michigan Revised School Code, also known 

as Michigan’s Schools of Choice Program, Michigan districts may accept students from districts 

in their own ISD or students living in districts in contiguous ISDs respectively. 97 percent of 

Michigan TPS districts have participated in either 105 or 105c in the last decade with over 80 

percent participating in both (Edwards & Cowen, 2020). Districts that decide to participate in 

Schools of Choice determine how many students they accept, which grades, programs, and 

schools non-resident students can enroll in, the timeframe they accept applications, and whether 

or not they offer transportation to non-residents. Under Michigan law, districts cannot select 

which non-resident students enroll in their district with few exceptions. Districts may refuse 

enrollment to students who have been suspended or expelled. Additionally, they do not have to 

accept students from districts outside their ISD with an Individualized Education Program (IEP) 

 

23 

if they do not have an agreement with the students’ district of residence. Outside of these 

exceptions, oversubscribed districts must hold a lottery. Districts may also enter into local 

cooperative agreements with other districts to enroll their students. Unlike Schools of Choice, 

selective enrollment practices are allowed by these local cooperative agreements (Michigan 

Department of Education, 2013). During the 2017-18 school year, about 13 percent of Michigan 

public school students used inter-district choice.  

Figure 1. 2017-18 Participation in School Choice by District of Residence 

Figure 1A: Inter-District Choice 

 

      Figure 1B: Charter School Choice 

Note. The denominator of the proportions in Panels A and B include all residents regardless if they attend a 
traditional public school or charter school.  
 

Due to its prevalence, longevity, and lack of regulation, Michigan’s school choice system 

  

 

 

is an ideal setting to study the role of location in school choice participation. First, Michigan has 

a relatively high proportion of Michigan public school students participating in inter-district and 

charter school choice with substantial use of school choice outside of urban areas. Figures 1A 

and 1B display participation in inter-district and charter school choice by district of residence. A 

higher percentage of rural students use inter-district choice than students living in other locales. 

Although over sixty percent of charter school students live in the Metro Detroit region, charter 

 

24 

schools do exist in many rural districts throughout the state. In addition to its widespread use, the 

maturity of Michigan’s school choice programs allows me to examine participation in a 

stabilized and developed schooling market. Finally, the absence of regulations on the supply of 

schools in Michigan provides conditions closest to the free market ideal of school choice. 

Michigan’s charter school laws are considered some of the least regulated (Candal, 2018; 

Ziebarth, 2019). In theory, this allows for unfettered access to effective schools and the 

opportunity for families’ schooling decisions to regulate the market, truly testing whether public 

school choice policies can increase access to effective schools and improve school productivity.  

Data 

My main sources of data are student-level enrollment and achievement records from the 

Michigan Department of Education (MDE) and Center for Educational Performance and 

Information (CEPI). These data include student demographic information (e.g., race and 

ethnicity, gender, disability status, English Learner status, and economically disadvantaged 

status2), student test scores on state standardized achievement exams (either the Michigan 

Educational Assessment Program, MEAP, or the Michigan Student Test of Educational Progress, 

M-STEP), and student addresses geocoded at the census block level for all Michigan public 

school students from 2012-13 to 2017-18. Additionally, I use a school-level data made publicly 

available by MDE, CEPI, and Michigan’s Department of Technology, Management, and Budget 

that includes each district and school’s sector, address, and educational settings as well as district 

boundaries.  

To create my full analytic sample, I begin with 8,808,831 student-year observations 

between the 2012-13 and 2017-18 school years of students attending a traditional public school 

2 In Michigan, students are considered economically disadvantaged if they qualify for free or reduced lunch, receive 
food (SNAP) or cash assistance (TANF), or they are homeless, migrant, or in foster care. 

 

 

25 

or a charter school offering a general education setting. First, I exclude less than two percent of 

observations for the following reasons: 67,352 observations of students attending schools in 

juvenile detention centers, boarding schools, virtual schools, and other residential schools since 

these students do not commute to school; 64,680 observations of students who either attend a 

school that changes districts or live in a census block where the district of residence changes 

during the panel since these changes could induce them to technically use school choice without 

switching schools or residences; 12,931 observations for students who do not have a school in 

their district of residence that offers his or her grade and therefore must participate in school 

choice by definition; 3,270 observations that I am unable to calculate the distance from the 

student’s residence to either their attended or nearest school; 2,675 student-year observations of 

students that are reported in a grade higher than the terminal grade of their attended school. Next, 

I remove 2.4% of observations where a student was held back or skipped a grade since an 

abnormal grade progression could induce a change in schools for different reasons that most 

students in the sample. Finally, I exclude of observations for homeless students (2.1% of my 

sample) since they do not have a stable residence by definition. My final analytic sample consists 

of 8,331,445 student-year observations representing almost 2.2 million unique students. I draw 

from this sample for each of my subsequent analyses. 

The main focus of my paper are the relationships between use of choice, residence, and 

school attended. I study two forms of public school choice: charter school and inter-district 

choice. I consider all other public school students, those attending a school in the district that 

they live in, their resident district, as not participating in a formal public school choice policy.3 I 

determine whether a student attends a TPS or charter school using information about schools 

3 I note that attending a school in the district of residence does not mean that families are not actively choosing a 
school. It is likely that they chose their residence so they could send their children to their desired school. 

 

 

26 

made publicly available by MDE and CEPI. I determine a student’s district of residence using 

the coordinates of the population weighted centroid of their resident census block and district 

boundaries. I consider a student to be a non-resident, one using inter-district choice, if they are 

attending a TPS in a district other than the one they live in.  

Table 1. 2017-18 Student Characteristics by Choice Use 
Non-

Full 

Sample 
1,356,085 
100% 
49% 
67% 
17% 
8% 
4% 
5% 
51% 
13% 
7% 
27% 
44% 
12% 
17% 
0.03 
0.02 

Resident 
1,064,568 
78% 
49% 
71% 
13% 
8% 
4% 
4% 
47% 
13% 
7% 
23% 
46% 
13% 
18% 
0.07 
0.06 

Resident  Charter 
169,793 
121,724 
9% 
13% 
50% 
50% 
30% 
69% 
17% 
53% 
10% 
8% 
4% 
2% 
4% 
5% 
52% 
76% 
12% 
13% 
12% 
4% 
62% 
27% 
37% 
28% 
4% 
13% 
5% 
23% 
-0.31 
0.00 
0.02 
-0.26 

 
Number of Students 
Percent of Sample 
Female 
White 
Black 
Hispanic 
Asian 
Other Race 
Econ. Dis. 
SWD 
EL 
City 
Suburb 
Town 
Rural 
Avg. Std. Math Score 
Avg. Std. ELA Score 
Note. Non-residents are defined as students who attend a traditional public school that is not in their district of 
residence. Econ. Dis., EL, and SWD stand for Economically Disadvantaged, English Learner and Student with 
Disability respectively. Locale is determined by the National Center of Education Statistics locale code of the 
student’s district of residence. ELA is an abbreviation for English Language Arts. Math and ELA test scores are 
from the Michigan Student Test of Educational Progress (M-STEP) and are standardized within grade, subject, and 
year at the state level. Test scores are available only for students who were in grades 3 through 8 in 2017-18  
with valid test scores.  

In Table 1, I present the summary statistics for students in the most recent school year of 

my sample, 2017-18, as well as differences between students attending charter schools, students 

using inter-district choice, and students attending a school in their district of residence. Overall, a 

higher percentage of students using formal school choice policies come from disadvantaged 

backgrounds and have lower average achievement compared to resident students. Differences are 

smaller between resident and non-resident students. Additionally, a higher percentage of non-

 

27 

resident students live in rural areas compared to resident students. In contrast, the majority of 

charter school students live in cities and are Black and economically disadvantaged.  

RQ 1: How often do students participating in formal school choice policies change 

residences? Are residential moves associated with participation in school choice programs? 

I hypothesize that many families enter into or exit from formal school choice programs at 

the time of a residential move. To test this hypothesis, I first describe the role of residential 

mobility in school choice decisions. Students in my analysis are considered to be residentially 

mobile if they live in a different census block than the previous year regardless of whether the 

new residence is within the boundaries of the same district.4 Moving residences, even when it 

does not change their district of residence, could change the student’s assigned school as well as 

the distance to other schooling options. I determine that students leave their initial choice at time 

t if they no longer using the form of choice they used at time t-1. For example, if a student leaves 

a non-resident school to attend a charter school, they would be leaving their initial choice. In 

contrast, if a student switches from one charter school to another charter school, they would not 

be leaving their initial choice. However, if a non-resident student moves into the district they are 

attending school in and continue to attend school there after they move, they would leave choice 

although they did not switch schools. 

 

4 5.2% of student-year observations have multiple addresses within the same school year. To deduplicate the 
addresses, I first drop observations with residences in a district that does not match the reported district of residence 
from MDE when the student has an observation that does match it since that is likely to be the address at the time 
the school data was collected. Next, I drop excess observations that have commute times that are over an hour to 
their attended school when the student has one that is closer because it is likely that they lived at the closer address 
when attending the reported school. Third, I drop observations where a student does not live in the district they 
attend but has an observation with an address in the district they attend. Finally, I drop the remaining duplicated 
observations at random since I have no indication of when or how long they lived at a given residence within a 
school year. 
 

 

28 

To answer my first research question, I first show the differences in residential mobility 

rates and exit from students’ initial choices by participation in formal school choice policies. 

Then, I examine how the characteristics of residentially mobile students differ from students who 

do not switch residences. Finally, I describe the relationships between school attended, moving 

residences and exit from the students’ initial choice. I accomplish this using two different 

samples drawn from my full analytic sample. First, I examine residential mobility and 

participation in school choice between two school years for all students. Specifically, this sample 

includes all students in my main sample that attend a Michigan public school in both 2016-17 

and 2017-18. While this sample permits us to describe residential mobility across all grades for 

the majority of students in Michigan, it does not allow us to examine residential mobility over 

time or account for a student’s initial choice. Thus, I also explore residential mobility and choice 

use for students who begin kindergarten in the first year of my panel and follow them through 5th 

grade, the last year of my panel.  

Table 2. Residential and Choice Mobility Behaviors by Initial Choice  

Mobility between 2016-17 and 

Mobility between Kindergarten 

 

N 

2017-18  
Changes 
Residences 

N 

and 5th Grade  

Changes 
Residences 

43% 
49% 
56% 

12% 
15% 
19% 

 66,493  
 7,977  
 8,546  

 953,714  
 143,791  
 111,238  

 
Resident 
Non-Resident 
Charter 
Note. Non-residents are defined as students who attend a traditional public school that is not in their district of 
residence. Columns 1-3 include students who are in my full sample during the 2016-17 and 2017-18 school years 
and a student is considered to change residences if they live in a different census block in 2017-18 than they did in 
2016-17. Columns 4-6 contain students who are in kindergarten in the initial year of the sample, 2012-13 and are in 
my sample all years of the panel. A student is considered to change residences if they live in a different census block 
than they did in kindergarten at any time before 6th grade. Rows represent choice in the initial year of the sample. I 
consider students who are no longer using the same choice policy to attend school to exit their initial choice. 
Students who use the same choice but change schools or district are not considered to exit their initial choices.  
 
 

In Table 2, I display the percent of students who are residentially mobile and the percent 

who exit choice for both samples by their initial choice. Across all forms of choice, less than 20 

 

29 

Exits 
Initial 
Choice 
3% 
12% 
17% 

Exits 
Initial 
Choice 
15% 
44% 
43% 

percent of students were residentially mobile between 2016-17 and 2017-18, but almost half of 

kindergarten students moved residences at least once by 5th grade. In both samples, a slightly 

higher percentage of students who attended a non-resident or charter school were residentially 

mobile. Compared to the percent of resident students who move residences, percent of resident 

students that no longer attend a school in their district of residence is much smaller. However, 

the percentages of students who change residences and exit their initial choice are similar for 

students using inter-district and charter school choice. This implies that most exits from formal 

school choice programs may be accompanied by residential moves.  

 

Next, I compare the characteristics of residentially mobile and immobile students by their 

initial choice. Table 3 displays these average differences in student demographic characteristics 

for those who are residentially mobile and those who are not between 2016-17 and 2017-18.5 

Across all initial choices, a higher percentage of residentially mobile students are Black, 

economically disadvantaged and live in cities compared to other students using the same form of 

choice. Residentially mobile students also have average test scores prior to moving residences 

that are 0.2 to 0.3 standard deviations lower than their immobile counterparts. Additionally, 

residentially mobile students are much more likely to exit their initial choice. For example, only 

four percent of students who live in the same residence between 2016-17 and 2017-18 and use 

inter-district choice in 2016-17 no longer use it during the next school year while almost 60 

percent of students who move residences exit inter-district choice. However, a smaller 

percentage of resident students who change residences exit their initial choice compared to 

students using formal school choice policies. Taken together, the evidence from Tables 2 and 3 

show that not only are students participating in school choice are more residentially mobile and 

5 Differences are similar for the kindergarten sample. However, kindergarten students do not have test scores for 
most of the panel, especially at the time of initial choice. Results are available by request. 

 

 

30 

that residentially mobile students are more likely to exit their initial choice, but residentially 

mobile choosers have a higher rate of exit from their initial choice than residentially mobile 

students attending a school in their district of residence. 

Table 3. 2016-17 Student Characteristics by Residential Mobility between 2016-17 and 2017-18 

Resident 

Non-Resident 

Same 

Charter 

Same 

 

 

Same 

Residence 
843,583 
1% 
49% 
73% 
12% 
7% 
4% 
4% 
41% 
12% 
7% 
22% 
48% 
13% 
17% 
0.12 
0.11 

Changes 
Residence 
20,873 
59% 
51% 
65% 
20% 
8% 
1% 
6% 
64% 
14% 
3% 
31% 
37% 
13% 
19% 
-0.19 
-0.15 

Changes 
Residence 
110,131 
19% 
49% 
60% 
23% 
9% 
2% 
6% 
68% 
15% 
8% 
31% 
40% 
14% 
15% 
-0.27 
-0.26 

Residence 
122,918 
4% 
50% 
71% 
15% 
8% 
2% 
5% 
45% 
13% 
4% 
27% 
37% 
13% 
23% 
0.04 
0.05 

Residence 
89,932 
11% 
50% 
32% 
50% 
10% 
4% 
4% 
70% 
12% 
13% 
61% 
30% 
4% 
5% 
-0.28 
-0.22 

Number of Students 
Pct. Exit Initial Choice 
Female 
White 
Black 
Hispanic 
Asian 
Other Race 
Econ. Disadvantaged 
Student with Disability 
English Learner 
City 
Suburb 
Town 
Rural 
Std. Math Score 
Std. ELA Score. 
Note. Non-residents are defined as students who attend a traditional public school that is not in their district of 
residence. Columns represent choice in the initial year of the sample, 2016-17. A student is considered to change 
residences if they live in a different census block in 2017-18 than they did in 2016-17. Students who use the same 
choice but change schools or district are not considered to exit their initial choices. Locale is determined by the 
National Center of Education Statistics locale code of the student’s district of residence. ELA is an abbreviation for 
English Language Arts. Math and ELA test scores are from the Michigan Student Test of Educational Progress (M-
STEP) and are standardized within grade, subject, and year at the state level. Test scores are available only for 
students who were in grades 3 through 8 in 2016-17 with valid test scores. 
 

Changes 
Residence 
21,306 
40% 
50% 
21% 
66% 
7% 
2% 
4% 
84% 
12% 
9% 
69% 
24% 
3% 
3% 
-0.54 
-0.49 

Now that I established that students participating in school choice are more likely to be 

residentially mobile and exit their initial choice when residentially mobile, I examine the 

percentage of exits from the initial choice that are accompanied by a residential move. Table 4 

describes residential mobility patterns for those who leave their initial choice. Almost two-thirds 

of students who no longer attend a school in resident district move residences. A similar 

 

31 

percentage of students who leave inter-district choice are residentially mobile. About half of 

charter school students who exit the charter sector move residences. Because students who move 

residences can attend the same school or district but enter or exit inter-district choice, I also 

calculate the percent of students who move residences and exit choice separately for those who 

attend the same school district between years and those who do not. Higher percentages of 

students who exit their resident district or inter-district choice and move residences during the 

2016-17 school year attend a school in the same district the next school year than change 

districts. Furthermore, one in five students who uses inter-district choice in kindergarten 

eventually moves into the district that they were using inter-district choice to attend. Overall, the 

findings of Table 4 show that the majority of students who leave their initial choice also move 

residences with many who leave their initial choice attending the same school. Taken together, 

the findings of Tables 2, 3, and 4 suggest that many exits from formal school choice programs 

may be accompanied by a residential move. 

Table 4. Residential Behaviors by Initial Choice for Students who Leave their Initial Choice  
Mobility between Kindergarten 

Mobility between 2016-17 and 

and 5th Grade 

Non-

Resident 

 

Resident 

2017-18 
Non-

Resident 
17,446 
29% 
29% 

9,927 
36% 
27% 

32,055 
36% 
26% 

Charter 
3,662 
49% 
51% 

Charter  Resident 
18,783 
54% 
46% 

 
Number of Students 
Same Residence 
Changes Residence, New 
District 
Changes Residence, Same 
N/A 
District 
Note. Non-residents are defined as students who attend a traditional public school that is not in their district of 
residence. Columns 1-3 include students who are in my full sample during the 2016-17 and 2017-18 school years 
and a student is considered to change residences if they live in a different census block in 2017-18 than they did in 
2016-17. Columns 4-6 contain students who are in kindergarten in the initial year of the sample, 2012-13, and are in 
my sample all years of the panel. A student is considered to change residences if they live in a different census block 
than they did in kindergarten at any time before 6th grade. I consider students who are no longer using the same 
choice policy to attend school to exit their initial choice. Students who use the same choice but change schools or 
district are not considered to exit their initial choices. Changes Residence, New District refers to students who move, 
but attend a school in a different district than they did the previous year. 
 

38% 

3,480 
25% 
28% 

47% 

42% 

N/A 

37% 

 

32 

RQ 2: How far do students travel to school? How does this differ for students participating 

in formal school choice programs? 

The relationship between where families live and their decisions to use and continue 

participating in formal school choice policies may not only be shaped by residential mobility but 

by the distance between schooling options and their residences. Thus, I also examine differences 

in commute times and distances from home to school between students participating in school 

choice and those who do not. I calculate commute times (in minutes) and commute distances (in 

miles) by car using Here Application Program Interface (API) from the population weighted 

centroid of the student’s home census block to their attended school as well as the nearest TPS in 

their district of residence offering their grade, my proxy for assigned school.6 Estimating the 

commute time to students’ nearest schools allows me to determine how much farther students are 

traveling to use school choice options and test my hypothesis that students who live farther from 

their assigned school are more likely to use school choice. I use the nearest school instead of the 

assigned school because I do not have data concerning student’s assigned schools or catchment 

zones for all districts in Michigan over time. I determine the student’s nearest school by 

 

6 Although I do not have exact addresses for students, I believe that using the population weighted centroid of the 
student’s resident census block provides reliable estimates for the following reasons. First, over half of U.S. census 
blocks are smaller than a tenth of a square mile, implying that my calculations should be within 528 feet of the 
actual address on average (Federal Communications Commission, 2015). To investigate errors associated with 
addresses coarsened to the block level and disparities in these errors between urban and rural locales in my sample, I 
calculated the geodetic distance from the population-weighted centroid of each student’s census block to the 
population-weighted centroid of the nearest census block within the same school district. This distance should 
provide an estimate of the possible size of the measurement error since the population weighted centroid of a 
student’s census block should be closer to their home than the population weighted centroid of the next census 
block. I find that the median distance between census blocks for students in my full sample is less than a tenth of a 
mile. Because census blocks are larger in area in rural locales (Federal Communications Commission, 2015), I also 
examine distance between census blocks in rural districts. I find that the median distance is about three tenths of a 
mile with less than one percent of rural students living in a census block that has a distance to the center of the next 
census block over one mile. Because these distances are fairly small, I conclude that using the population weighted 
centroids of the student’s resident census block to estimate commute times should provide fair estimates in cities as 
well as in rural areas. To account for the differences in the size of the census blocks between rural and urban areas 
and the larger errors in actual address, I focus my comparisons of travel times between students who use school 
choice and those who do not within locales. When the centroids do not fall on a road, I use the nearest road to the 
centroid to calculate drive time.  

 

33 

calculating the geodetic distance to each school offering his or her grade in their district of 

residence. I consider the school with the shortest distance to be the nearest school.  

I calculate commute time and distance by car using the fastest route assuming normal 

traffic when the student leaves home at 8am on a weekday to best estimate travel conditions 

during the morning commute to school. Since commutes cannot be calculated for past dates, I 

estimate commutes times on a weekday at the end April between 2019 and 2021. Although there 

have been changes to traffic patterns in the years between the beginning of my panel, September 

2012, and when I calculated the commute times, I do not believe that these changes are large 

enough to bias my estimates. Furthermore, these estimates do not account for extreme weather 

conditions found at other times in the school year in Michigan, allowing them to be comparable 

with other states. I consider my calculated commute times to be estimates of students’ actual 

commute since I do not know the student’s exact address, the mode of transportation students 

used to get to school, the exact time they leave or arrive to school, accurate weather conditions, 

or whether there was a significant change in traffic patterns during my panel. 

Since the size of, number of, and distance between schools varies across grades, I focus 

on analyzing commute times and distances for students in kindergarten, 6th grade, and 9th grade, 

the grades students most commonly change schools, separately. First, I present average 

commutes times and distances to each student’s attended and nearest school for all Michigan 

public school students by grade for the most recent school year, 2017-18, in Table 5. I choose to 

focus on the most recent school year in this analysis to more accurately reflect commute times 

and distances estimated. On average, students travel 8 to 11 minutes (3 to 5 miles) to school. 

Students in 9th grade travel farther to school than elementary school students. This is likely a 

function of the number and size of high schools compared to elementary schools. Additionally, 

 

34 

students live about 6 to 9 minutes away (2 to 4 miles) from their nearest school offering their 

grade in their district of residence.  

Table 5. 2017-18 Average Commute Times and Distances to Attended and Nearest School by 
Grade 

100,279 
8.6 
3.7 
6.1 
2.4 

108,231 
9.5 
4.2 
7.5 
3.1 

  Kindergarten  6th Grade  9th Grade 
109,155 
10.7 
4.6 
8.8 
3.5 

Number of Students 
Min. to Attended School 
Miles to Attended School 
Min. to Nearest School 
Miles to Nearest School 
Note. Commute times and distances are calculated using HERE API assuming normal traffic at 8 am on a weekday 
in April between 2019 and 2021. I determine the student’s nearest school, my  proxy for assigned school, by 
calculating the geodetic distance to each school offering his or her grade in their district of residence. I consider the 
school with the shortest distance to be the nearest school. 
 

Reporting state-level averages may mask differences due to access to multiple schooling 

and choice options. For example, rural students may have to travel farther to school than students 

living in cities regardless of whether or not they attend their resident school since schools are 

more spread out in rural areas due to population sparsity. Therefore, I compare commute times to 

attended school and nearest school within locales to determine how commutes to school differ 

for students participating in formal school choice programs in Figures 2 and 3. Furthermore, I 

focus on commute times in my discussion of my results since I contend that time more accurately 

reflects how individuals experience their commute since they account for traffic patterns, the 

availability of express ways, and the number of stoplights that may make commutes of the same 

distance take longer.  

In Figure 2, I display commute times to the attended school by grade, locale, and 

participation in school choice programs. Across all grades and locales, students participating in 

formal school choice programs travel farther to school on average. In kindergarten, students 

using inter-district choice travel twice as far to school than students attending a school in their  

district of residence with slightly smaller differences in 6th and 9th grade. In kindergarten and 6th  

 

35 

Figure 2. Average Commute Time to Attended School by Grade, Locale, and Choice 
Figure 2A: Kindergarten 

)
s
e
t
u
n
i
M

i

(
 
e
m
T
 
e
t
u
m
m
o
C

24
22
20
18
16
14
12
10
8
6
4
2
0

14.4

t
n
e
d
i
s
e
R
-
n
o
N
City

6.0

t
n
e
d
i
s
e
R

10.3

6.3

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

Figure 2B: 6th Grade 

16.8

13.5

18.0

16.4

14.3

11.6

7.8

9.1

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

t
n
e
d
i
s
e
R
-
n
o
N
Suburb

t
n
e
d
i
s
e
R
-
n
o
N
Town

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

-
n
o
N
Rural

 

)
s
e
t
u
n
i
M

i

(
 
e
m
T
 
e
t
u
m
m
o
C

24
22
20
18
16
14
12
10
8
6
4
2
0

14.5

t
n
e
d
i
s
e
R
-
n
o
N
City

7.5

t
n
e
d
i
s
e
R

14.3

12.2

16.7

13.3

9.0

9.4

11.2

8.0

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

t
n
e
d
i
s
e
R
-
n
o
N
Suburb

t
n
e
d
i
s
e
R
-
n
o
N
Town

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

Figure 2C: 9th Grade 

)
s
e
t
u
n
i
M

i

(
 
e
m
T
 
e
t
u
m
m
o
C

24
22
20
18
16
14
12
10
8
6
4
2
0

9.8

t
n
e
d
i
s
e
R

15.9

15.8

15.9

17.6

17.0

12.7

9.1

9.7

9.9

t
n
e
d
i
s
e
R
-
n
o
N
City

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

t
n
e
d
i
s
e
R
-
n
o
N
Suburb

t
n
e
d
i
s
e
R
-
n
o
N
Town

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

-
n
o
N
Rural

 

18.3

17.5

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

-
n
o
N
Rural

 

22.1

19.4

 

 

 

Note. Non-residents are defined as students who attend a traditional public school that is not in their district of 
residence. Locale is determined by the National Center of Education Statistics locale code of the student’s district of 
residence. Commute times and distances are calculated using HERE API assuming normal traffic at 8 am on a 
weekday in April between 2019 and 2021. 
 

 

36 

Figure 3. Commute Time to Nearest School By Grade, Locale, and Choice 
Figure 3A: Kindergarten 

11.1

9.8

9.2

8.7

8.1

7.2

5.0

4.7

4.3

6.0

5.5

5.1

)
s
e
t
u
n
i
M

i

(
 
e
m
T
 
e
t
u
m
m
o
C

14

12

10

8

6

4

2

0

t
n
e
d
i
s
e
R

t
n
e
d
i
s
e
R
-
n
o
N
City

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

t
n
e
d
i
s
e
R
-
n
o
N
Suburb

t
n
e
d
i
s
e
R
-
n
o
N
Town

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

-
n
o
N
Rural

 

Figure 3B: 6th Grade 

10.8

8.8

9.4

9.3

11.7

10.7

7.3

7.6

7.1

5.9

6.2

5.1

)
s
e
t
u
n
i
M

i

(
 
e
m
T
 
e
t
u
m
m
o
C

14

12

10

8

6

4

2

0

t
n
e
d
i
s
e
R

t
n
e
d
i
s
e
R
-
n
o
N
City

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

t
n
e
d
i
s
e
R
-
n
o
N
Suburb

Figure 3C: 9th Grade 

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

t
n
e
d
i
s
e
R
-
n
o
N
Town

11.3

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

-
n
o
N
Rural

 

11.9

11.6

)
s
e
t
u
n
i
M

i

(
 
e
m
T
 
e
t
u
m
m
o
C

14

12

10

8

6

4

2

0

7.5

7.6

6.9

8.7

8.6

8.2

9.5

9.8

9.8

t
n
e
d
i
s
e
R

t
n
e
d
i
s
e
R
-
n
o
N
City

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

t
n
e
d
i
s
e
R
-
n
o
N
Suburb

t
n
e
d
i
s
e
R
-
n
o
N
Town

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

r
e
t
r
a
h
C

t
n
e
d
i
s
e
R

-
n
o
N
Rural

 

 

 

 

Note. Non-residents are defined as students who attend a traditional public school that is not in their district of 
residence. Locale is determined by the National Center of Education Statistics locale code of the student’s district of 
residence. Commute times and distances are calculated using HERE API assuming normal traffic at 8 am on a 
weekday in April between 2019 and 2021. I determine the student’s nearest school, my proxy for assigned school, 
by calculating the geodetic distance to each school offering his or her grade in their district of residence. I consider 
the school with the shortest distance to be the nearest school. 

 

37 

grade, non-resident students travel farther to school than charter school students on average. 

However, charter school students and students using inter-district choice travel similar distances 

on average in 9th grade. In Figure 3, I compare commute times to the nearest school between 

resident, non-resident, and charter school students. Average differences in time to nearest school 

between students who participate in school choice and those who do not are much smaller than 

differences in commute time to attended school. In town and rural districts, students who use 

inter-district choice live farther away from their nearest school than those who attend a school in 

their resident district. This implies that the opportunity cost of using inter-district choice may be 

lower for rural students whose nearest school, their likely default option, is farther away from 

their residence.  

Since I show that there are differences in commute times to students’ schools as well as 

nearest schools by choice use, I also explore how much longer students using formal choice 

policies spend commuting to school than they would if they attended their nearest school in 

Table 6. The additional minutes or miles a student travels are the differences between the 

attended school commute and the nearest school commute (attended school commute - nearest 

school commute). Students with negative additional commute times attend schools closer to 

home than their nearest school. Non-resident students travel an additional 6 to 10 minutes past 

their nearest school to their attended school on average. City students have the highest additional 

drive time. The average additional distance traveled by charter school students ranges from 3 to 9 

minutes depending on grade and locale. Some students who use formal school choice policies 

actually attend a school closer to home than their nearest school. In Table 6, I also report the 

percent of students attending schools closer to home than their nearest school. Approximately 

one in five non-resident students living in rural areas and towns attend a school closer to home. 

 

38 

Similarly, over twenty percent of students attending a charter school attend a school closer to 

home. Taken together, the results I present in Figure 3 and Table 6 show that many students use 

school choice policies to attend schools closer to home or when their assigned school is far from 

home. This implies that proximity to home relative to their nearest or assigned school is a likely 

determinant in participation in formal school choice policies. 

 

 

Non-Resident 

Charter 

 

 

 

 

 

 

 

6.7 
3.6 
18% 

 

8.0 
5.6 
19% 

 

8.4 
4.6 
12% 

 

6.9 
5.0 
27% 

 

4.9 
3.3 
32% 

 

5.9 
2.9 
22% 

 

7.5 
4.9 
19% 

City  Suburb  Town  Rural  City  Suburb  Town  Rural 
 

Table 6. Additional Distance Traveled Past Nearest School by Grade, Locale, and Choice 
 
 
Kindergarten 
9.6 
Additional Minutes 
5.3 
Additional Miles 
Pct. Attending Closer to Home   8% 
 
6th Grade 
9.0 
Additional Minutes 
5.2 
Additional Miles 
Pct. Attending Closer to Home  11% 
9th Grade 
8.3 
Additional Minutes 
4.9 
Additional Miles 
Pct. Attending Closer to Home  14% 
Note. Non-residents are defined as students who attend a traditional public school that is not in their district of 
residence. Locale is determined by the National Center of Education Statistics locale code of the student’s district of 
residence. Additional Minutes and Miles are the difference between the commute to the attended school and the 
student’s nearest school. Commute times and distances are calculated using HERE API assuming normal traffic at 8 
am on a weekday in April between 2019 and 2021. I determine the student’s nearest school, my proxy for assigned 
school, by calculating the geodetic distance to each school offering his or her grade in their district of residence. I 
consider the school with the shortest distance to be the nearest school. I determine the percent of students attending 
school closer to home using additional commute time in minutes. Negative additional minutes means that the student 
attends a school closer to home than the nearest school in their district of residence.  
 

3.3 
2.1 
35% 

 

6.7 
4.7 
29% 

 

6.7 
3.8 
17% 

 

6.2 
4.5 
25% 

 

5.9 
3.1 
22% 

 

5.1 
3.2 
25% 

 

6.8 
4.5 
19% 

7.0 
4.6 
21% 

 

8.0 
5.3 
20% 

4.6 
2.8 
28% 

6.6 
5.1 
28% 

8.6 
6.5 
27% 

7.2 
4.3 
18% 

6.4 
4.9 
26% 

RQ3: What are the roles of residential mobility and commute time to school in continued 

participation in formal school choice policies? 

Since I find differences in residential mobility and distance to school between students 

who participate in formal school choice programs and those who do not, I formally examine 

whether residential mobility and distance traveled to school are associated with mobility out of 

inter-district or charter school choice. I accomplish this by estimating a set of hazard models on 

 

39 

the students in my full analytic sample who were in kindergarten in 2012-13, participated in 

either charter school or inter-district choice in kindergarten, and were in my full sample all six 

years of the panel.7 I focus on kindergarten, so I know with certainty what the student’s initial 

choice was. I exclude students who leave Michigan public schools during the panel since I 

cannot account for their residential mobility at the time they leave the panel. Hazard models 

allow me to account for the relationship between the passage of time and mobility out of schools 

and have been used in prior studies of mobility and exit out of school choice (Cowen, Creed, & 

Keesler, 2015; Cowen, Fleming, Witte, & Wolf, 2012; Finch, Lapsley, & Baker-Boudissa, 2009; 

Lavery & Carlson, 2015). I estimate the hazard of leaving inter-district choice and charter 

schools separately. I also estimate my models on the sample restricted to students who never 

move residences since students who move residences are likely to have different choice sets 

before and after moving and the distance between home and school during the previous school 

year may be less relevant after changing residences.  

Table 7 displays the average student characteristics of the non-resident and charter school 

students in my sample and compares students who leave their initial choice to those who remain 

in it through 5th grade. The majority of students who leave either form of choice move residences 

at some point between kindergarten and 5th grade. Those who switch out of inter-district choice 

travel farther past their nearest school than students that remain. Additionally, a higher 

percentage of students who leave inter-district choice are economically disadvantaged.  

 

 

 

7 In my main specifications, I also exclude 2,300 students who attend a school that does not have test scores in at 
least one year of the panel because a school’s average achievement level is a significant predictor of mobility. Many 
of these schools without test scores only offer grades K-2. Results are similar with the full sample and are available 
by request. 

 

40 

Full Sample 

Non-Resident 

Charter 
Stays 

Same Residence Sample 

Non-Resident 

Charter 
Stays 

Switch 

Full 

Stays 

Switch 

Full 

Switch 

Full 

Stays 

Switch 

Full 

3,599 
22% 
14.4 
7.6 
6.8 
50% 
80% 
9% 
5% 
2% 
4% 
38% 
10% 
3% 
21% 
39% 
12% 
28% 

6,268 
48% 
14.9 
7.2 
7.6 
49% 
78% 
11% 
6% 
2% 
5% 
43% 
10% 
3% 
22% 
41% 
12% 
25% 

2,669 
84% 
15.5 
6.7 
8.7 
49% 
75% 
13% 
6% 
1% 
5% 
51% 
11% 
3% 
24% 
43% 
11% 
22% 

Table 7. 2012-13 Kindergarten Student and School Characteristics by Mobility and Initial 
Choice 
 
 
 
Student Characteristics 
Number of Students 
Switches Residences 
Min. to Attended Sch. 
Min. to Nearest Sch. 
Additional Minutes 
Female 
White 
Black 
Hispanic 
Asian 
Other Race 
Econ. Dis. 
Student with Disability 
English Learner 
City 
Suburb 
Town 
Rural 
Note. The same residence sample only includes students who live at the same residence all years of the panel. Non-
residents are defined as students who attend a traditional public school that is not in their district of residence. A 
student is considered to switch residences if they live in two different census blocks during the panel. Commute 
times are calculated using HERE API assuming normal traffic at 8 am on a weekday in April between 2019 and 
2021. I determine the student’s nearest school, my proxy for assigned school, by calculating the geodetic distance to 
each school offering his or her grade in their resident district. I consider the school with the shortest distance to be 
the nearest school. Additional Minutes are the difference between the commute to the attended school and the 
student’s nearest school. Econ. Dis. stands for Economically Disadvantaged. Locale is determined by the National 
Center of Education Statistics locale code of the student’s resident district. ELA is an abbreviation for English 
Language Arts. Math and ELA test scores are from the Michigan Student Test of Educational Progress (M-STEP) 
and are standardized within grade, subject, and year at the state level. 
 

7,955 
56% 
11.0 
5.2 
5.9 
50% 
33% 
52% 
7% 
4% 
5% 
73% 
7% 
10% 
61% 
29% 
5% 
5% 

4,640 
44% 
11.1 
5.4 
5.7 
51% 
32% 
52% 
8% 
4% 
4% 
71% 
8% 
11% 
62% 
28% 
5% 
5% 

3,315 
72% 
11.0 
4.9 
6.1 
50% 
33% 
53% 
5% 
4% 
5% 
75% 
6% 
7% 
60% 
30% 
4% 
5% 

3,234 
N/A 
14.7 
7.8 
6.9 
49% 
81% 
8% 
5% 
2% 
4% 
35% 
10% 
4% 
19% 
38% 
12% 
31% 

2,813 
N/A 
14.2 
8.0 
6.2 
49% 
82% 
7% 
5% 
2% 
4% 
33% 
10% 
3% 
19% 
38% 
12% 
31% 

3,524 
N/A 
10.9 
5.5 
5.4 
50% 
41% 
41% 
8% 
5% 
5% 
62% 
8% 
13% 
55% 
32% 
6% 
7% 

2,604 
N/A 
10.7 
5.6 
5.1 
50% 
39% 
43% 
9% 
5% 
4% 
63% 
8% 
15% 
58% 
29% 
6% 
7% 

421 
N/A 
17.9 
6.8 
11.1 
49% 
76% 
13% 
7% 
2% 
3% 
44% 
16% 
5% 
19% 
42% 
10% 
29% 

920 
N/A 
11.4 
5.2 
6.3 
50% 
45% 
37% 
6% 
7% 
6% 
60% 
7% 
10% 
47% 
38% 
6% 
8% 

First, I estimate the unconditional hazard of leaving inter-district or charter school choice 

and graph the probability of staying in the initial form of choice as Kaplan-Meier survival curves 

in Figures 4A and 4B respectively. For both non-resident and charter school students, the 

probability of remaining in their initial choice decreases over time. Next, I plot the Kaplan-Meier 

curves separately for students who change residence and those who do not in Figure 5. These 

figures show that residentially mobile students have an increased hazard of exiting their initial 

choice. In fact, Figure 5A shows that few students who live at the same residence between  

 

41 

Figure 4. Kaplan-Meier Survival Probabilities for All Students (Survival=remain in initial 
choice). Kindergarten to 5th Grade 

           Figure 4B: Charter School Students 

Figure 4A: Non-Resident Students 

 

 

  

 
Note. Non-residents are defined as students who attend a traditional public school that is not in their resident district. 
The unit of Analysis Time is years starting with the 2012-13 school year and ending with the 2017-18 school year. 
Survival probability is the probability of remaining in the same choice between each year. 
 
Figure 5. Kaplan-Meier Survival Probabilities by Residential Mobility (Survival=remain in 
initial choice). Kindergarten to 5th Grade 
          Figure 5A:  Non-Resident Students 

Figure 5B: Charter School Students 

             

 

 
Note. Non-residents are defined as students who attend a traditional public school that is not in their resident district. 
The unit of Analysis Time is years starting with the 2012-13 school year and ending with the 2017-18 school year. 
Survival probability is the probability of remaining in the same choice between each year. I consider a student to 
change residences if they do not live in the same census block all years of the panel. 
 
kindergarten and 5th grade leave inter-district choice; exit is driven by residentially mobile 

 

students. Figures 6A and 6B examine the unconditional hazards of leaving inter-district choice or 

charter school choice by quartile of total drive time to school in kindergarten for the full sample.8 

Non-resident students who travel farther to school have an increased hazard of exiting their 

initial choice while there are small differences by commute time for charter school students. The 

8 Results for the same residence sample can be found in Figure 7 in the Appendix.  

 

 

42 

differences in hazards by quartiles of additional minutes traveled to school past their nearest 

school are slightly larger as seen in Figure 8 in the Appendix.  

Figure 6. Kaplan-Meier Survival Probabilities by Attended School Driving Time Quartile 
(Survival=remain in initial choice). Kindergarten to 5th Grade 
            Figure 6A: Non-Resident Students 

 Figure 6B: Charter School Students 

             

 

 

 
Note. Non-residents are defined as students who attend a traditional public school that is not in their resident district. 
The unit of Analysis Time is years starting with the 2012-13 school year and ending with the 2017-18 school year. 
Survival probability is the probability of remaining in the same choice between each year. Quartiles are determined 
using the commute time to attended school in kindergarten. Commute times and distances are calculated using 
HERE API assuming normal traffic at 8 am on a weekday in April between 2019 and 2021.  
 
 

Next, I condition the student’s probability of leaving their initial choice by residential 

mobility, distance traveled, and student and school characteristics. Formally, I estimate Equation 

1 separately for students who begin in inter-district choice or charter schools:  

ℎ!(#)=&(#)exp	(+,-./,0#123,4!"5#+7.-#809,!,"%#5&+:;',(%)+<=',(%)) 
Where ℎ!(#) is the hazard that student i fails to continue in their initial choice at time t. The 
baseline hazard function, &(#) is assumed to have a Weibull distribution, as is common in most 
applications of survival analysis (Manton, Singer, & Woodbury, 1992).9 	+,-./,0#123,4!" 
7.-#809,!,"%# is either the overall travel time to the student’s attended school, the commute time 

equals one when a student lives in a different census block at time t than he or she did at time t-1. 

(1) 

 

 

9 I choose a Weibull distribution for the hazard function since the hazard increases over time at a decreasing rate as 
suggested visually in the Kaplan-Meier curves and by the shape parameter. I also estimate Cox proportional hazard 
models. The results are similar and available by request. 

 

43 

to their nearest TPS in their district of residence, or the additional travel time to the attended 

school past the nearest school. I estimate specifications of my main model with linear and 

quadratic functional forms of each distance measure as well as combinations of overall distance 

with either nearest or additional distance. ;',(%) is a vector of student characteristics including 
statuses, and locale of the student’s district of residence. =',(%) includes the following school 

gender, race, economically disadvantaged, English Learner, and student with disability (SWD) 

characteristics: total enrollment, the percent of female, Black, Hispanic, Asian, Other Race, 

economically disadvantaged, and English Learner students, the percent of students with 

disabilities, and the average standardized math score on state exams.10 I also include the average 

drive time to school in the vector of school characteristics to help account for differences in exit 

rates for schools that are located far from residential areas in general. Because average drive time 

could be related to the number of students using choice, I also include the percent of non-resident 

students attending the school in =',(%)	in the inter-district choice models. Standard errors are 
Table 8 Panel A displays the estimates of my coefficients of interest, +,-./,0#123,4!", 
and the variants of 7.-#809,!,"%#, on the full sample of non-resident students. Coefficients are 

clustered at the district of residence.  

Results 

displayed as log odds coefficients. Across all of my specifications, moving residences is a large 

and significant predictor of leaving inter-district choice holding constant other risk factors 

including economic disadvantaged status, race, and school achievement levels and 

demographics. In Column 1, I show that having a minute increase in commute time is associated 

10 I also estimate Equation 1 with standardized ELA score instead of math score. Results are similar and available by 
request. 

 

 

44 

with a small increase in the probability of exiting inter-district choice. In contrast, students who 

live farther away from their nearest school in their district of residence are less likely to leave 

inter-district choice as shown in Column 3. This implies that students who live farther from their 

assigned school may have a lower opportunity cost to continue participating in inter-district 

choice. Because of the positive relationship between leaving inter-district choice and actual 

commute time and the negative relationship between exit and travel time to the nearest school, I 

also examine the relationship between additional distance traveled to the attended school past 

their nearest school for non-resident students in Column 5. For every additional minute traveled 

to the attended school past the nearest school, there is an increase in the hazard of exiting inter-

district choice.  

To further examine these relationships, I include both the actual commute time and either 

the commute time to the nearest school or the additional minutes traveled in the model. When 

both commute times to the attended and nearest school are included in the model as linear terms, 

their relationships with exit are the same as the models when only one of the distance measures 

was included. Interestingly, they are similar in magnitude in opposite directions, indicating that if 

distance to both the attended and nearest school increases, there would be little change in the 

probability of exiting. To test this, I estimate a model with commute time to the attended school 

and the additional commute time measure. The coefficient on commute time to the attended 

school represents the change in the probability of exiting if both the time it takes to get to the 

nearest and attended school increased. The coefficient on additional minutes represents the 

change in the probability of exiting if the time it takes to get to the attended school increases, 

holding the distance to the nearest school constant. Results of this model are displayed in 

Column 9. There is no significant relationship between overall drive time and exiting  

 

45 

Table 8. Predicting Exit from Initial Choice from Kindergarten to 5th Grade: Full Sample 
Panel A: Non-Resident Students 
(1) 
  
Resident Mover 

(6) 

(2) 

(4) 

(5) 

(7) 

2.974*** 
(0.054) 

2.959*** 
(0.054) 

2.958*** 
(0.054) 

 
 
 
 

2.958*** 
(0.054) 
0.011*** 
(0.003) 
-0.012*** 
(0.004) 

 
 
 
 
 
 
 
 

0.012*** 
(0.004) 

 
 
 
 

-0.000 
(0.000) 
-3.437*** 
(0.384) 

(8) 

2.959*** 
(0.054) 
0.011* 
(0.006) 
-0.010 
(0.009) 

 
 

0.000 
(0.000) 
-0.000 
(0.000) 

 
 

(9) 

2.958*** 
(0.054) 
-0.001 
(0.005) 

 
 

0.012*** 
(0.004) 

 
 
 
 
 
 

(10) 

2.958*** 
(0.054) 
-0.004 
(0.009) 

 
 

0.014** 
(0.006) 
0.000 
(0.000) 

 
 

-0.000 
(0.000) 
-3.404*** 
(0.392) 

-3.430*** 
(0.385) 

-3.439*** 
(0.389) 

-3.430*** 
(0.385) 

Min. to Attended School 

 

Min. to Nearest School 

 

Additional Min. 

 

Min. to Attended School Sq. 

 

Min. to Nearest School Sq. 

 

Additional Min. Sq. 

 

Constant 

 

 

 
 

 

2.964*** 
(0.054) 
0.011*** 
(0.003) 

2.964*** 
(0.054) 
0.011* 
(0.006) 

(3) 

2.973*** 
(0.054) 

 
 

 
 

-0.011*** 
(0.004) 

 
 
 
 
 
 
 
 

-0.011 
(0.008) 

 
 
 
 

-0.000 
(0.000) 

 
 

 
 
 
 

-0.000 
(0.000) 

 
 
 
 

 
 
 
 

0.0111*** 
(0.002) 

 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 

-3.471*** 
(0.384) 

-3.473*** 
(0.388) 

-3.328*** 
(0.383) 

-3.330*** 
(0.384) 

-3.437*** 
(0.384) 

 

46 

Table 8 (cont’d) 
Panel B: Charter School Students 
  
Resident Mover 

(1) 

Min. to Attended School 

 

Min. to Nearest School 

 

Additional Min. 

 

Min. to Attended School Sq. 

 

Min. to Nearest School Sq. 

 

Additional Min. Sq. 

 

Constant 

 

 

1.613*** 
(0.087) 
0.004 
(0.004) 

 
 
 
 
 
 
 
 
 
 

(2) 

1.613*** 
(0.087) 
0.002 
(0.005) 

 
 
 
 

0.000 
(0.000) 

 
 
 
 

(3) 

1.613*** 
(0.085) 

 
 

(4) 

1.612*** 
(0.085) 

 
 

-0.019*** 
(0.007) 

 
 
 
 
 
 
 
 

-0.007 
(0.013) 

 
 
 
 

-0.001 
(0.001) 

 
 

(5) 

1.609*** 
(0.087) 

 
 
 
 

0.007** 
(0.003) 

 
 
 
 
 
 

(6) 

1.609*** 
(0.087) 

 
 
 
 

0.011*** 
(0.004) 

 
 
 
 

-0.000* 
(0.000) 
-3.772*** 
(0.381) 

(7) 

1.610*** 
(0.087) 
0.004 
(0.004) 
-0.020*** 
(0.007) 

 
 
 
 
 
 
 
 

(8) 

1.609*** 
(0.086) 
0.003 
(0.005) 
-0.008 
(0.013) 

 
 

0.000 
(0.000) 
-0.001 
(0.001) 

 
 

(9) 

1.610*** 
(0.087) 
-0.016** 
(0.007) 

 
 

0.020*** 
(0.007) 

 
 
 
 
 
 

-3.702*** 
(0.388) 

-3.747*** 
(0.388) 

-3.702*** 
(0.388) 

(10) 

1.611*** 
(0.086) 
-0.025*** 
(0.009) 

 
 

0.022*** 
(0.007) 
0.001*** 
(0.000) 

 
 

-0.001*** 
(0.000) 
-3.669*** 
(0.387) 

-3.798*** 
(0.381) 

-3.792*** 
(0.381) 

-3.711*** 
(0.388) 

-3.764*** 
(0.387) 

-3.759*** 
(0.384) 

Note. Total Observations Panel A: 24,397. Panel B: 32,315. Estimated coefficients are display as log odd coefficients. Standard errors clustered at the resident 
district level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Student’s gender, race, economically disadvantaged, English Learner, and disability statuses, the 
locale of the student’s resident district and the following school characteristics total enrollment, the percent of female, Black, Hispanic, Asian, Other Race, 
economically disadvantaged, and English Learner students, the percent of students with disabilities, the average standardized math score on state exams, and the 
average drive time at time t-1 are included as covariates. In Panel A I also include the percent of non-resident students at time t-1 as a covariate. A student is 
considered be resident mover if they live in a different census block at time t than they did at t-1. Commute times are calculated using HERE API assuming 
normal traffic at 8 am on a weekday in April between 2019 and 2021. I determine the student’s nearest school, my proxy for assigned school, by calculating the 
geodetic distance to each school offering his or her grade in their resident district. I consider the school with the shortest distance to be the nearest school. 
Additional Minutes are the difference between the commute to the attended school and the student’s nearest school. 

 

47 

inter-district choice, but an increase in the additional travel time is associated with an increased 

hazard of leaving, indicating that relative distance past the default option, not total time traveled, 

influences decisions to remain in inter-district choice. The inclusion of the quadratic terms in the 

models in Columns 2, 4, 6, 8, and 10 does not meaningfully change the relationship between 

commute time and exit for any of my commute time measures suggesting that the relationship is 

linear.  

Table 8 Panel B presents the analogous models for the sample of charter school students. 

Similar to non-resident students, charter school students have a higher probability of exiting the 

charter sector if they move residences between time t and t-1. While students who have a longer 

commute to their nearest school are less likely to exit charter schools, there is no significant 

relationship between commute time to the attended school and exit. There is a small but positive 

relationship with exiting choice and additional time traveled to the attended school past the 

nearest school. When both overall commute time to the attended school and the additional time 

to school are included in the model, an increase in distance to the attended school past the nearest 

school is associated with an increased hazard of leaving while an increase in both the distance to 

the nearest and attended schools is associated with a decreased hazard of leaving. This implies 

that charter school students are more likely to leave the charter sector the farther their attended 

school is relative to their nearest school, but less likely to exit if both their nearest school and 

attended school are farther from home.  

The commute time from home to school at time t-1 may not be relevant to school choice 

decisions at time t if students no longer live at the same residence. Therefore, I also estimate a 

version of Equation 1 on the sample of students who live at the same address between 

kindergarten and 5th grade without the residential mobility indicator. Results of this specification 

 

48 

are presented in Table 9. For non-resident students, the direction of the relationships between all 

the commute time measures and exiting are in the same direction as the models estimated on the 

full sample, but their magnitudes are larger. Additionally, there is some evidence that some of 

these relationships are quadratic in nature. Still, when both overall and additional commute time 

are in the model, only additional commute time is statistically significant. Similarly, the 

relationships between commute times and the probability of exit are larger in magnitude and 

statistically significant for charter school students who do not change residences. In the model 

with both overall and additional commute time, only additional commute time is statistically 

significant. These results indicate that the commute time to school relative to the default option 

may be more important when making school choice decisions than overall distance especially for 

families whose residence is stable over time.  

Discussion 

In this paper, I provide some of the first evidence of the differences in residential 

mobility and commute times for students who use inter-district and charter school choice and 

their roles in continued participation in formal school choice programs. My findings show that 

both residential mobility and commute time are likely determinants of participation in school 

choice. The majority of students who leave their initial choice also move residences, and many 

who leave their initial choice actually attend the same school but move into or out of the district 

where they are attending school. As for distance, I find that many families use inter-district or 

charter school choice to attend schools closer to home or when their assigned school is farther 

from home. Similarly, students are more likely to exit their initial choice when the commute time 

to their attended school relative to their nearest school is longer. Taken together, these results  

 

 

49 

Table 9. Predicting Exit from Initial Choice from Kindergarten to 5th Grade: Same Residence Sample 
Panel A: Non-Resident Students 
(1) 
  
Min. to Attended School 

(2) 

(7) 

(8) 

(3) 
 
 

(4) 
 
 

-0.062*** 
(0.013) 

-0.067** 
(0.033) 

 
 
 
 
 
 
 
 

-3.358*** 
(1.208) 

 
 
 
 

0.000 
(0.001) 

 
 

-3.342*** 
(1.197) 

0.043*** 
(0.006) 

0.103*** 
(0.016) 

 
 
 
 

-0.001*** 
(0.000) 

 
 
 
 

-4.800*** 
(1.205) 

 
 
 
 
 
 
 
 
 
 
-4.242*** 
(1.221) 

 

 

Min. to Nearest School 

 

Additional Min.  

 

Min. to Attended School Sq. 

 

Min. to Nearest School Sq. 

 

Additional Min. Sq. 

Constant 

 

 

 
 

 

0.0427*** 
(0.007) 
-0.059*** 
(0.013) 

0.104*** 
(0.016) 
-0.051 
(0.038) 

 
 
 
 
 
 
 
 

-3.961*** 
(1.214) 

 
 

-0.001*** 
(0.000) 
-0.000 
(0.002) 

 
 

-4.583*** 
(1.192) 

(9) 

-0.016 
(0.014) 

(10) 
0.044 
(0.033) 

 
 

0.059*** 
(0.013) 

 
 
 
 
 
 

-3.961*** 
(1.214) 

 
 

0.060*** 
(0.019) 
-0.001 
(0.001) 

 
 

0.000 
(0.001) 
-4.560*** 
(1.240) 

(5) 
 
 
 
 

(6) 
 
 
 
 

0.044*** 
(0.006) 

0.081*** 
(0.010) 

 
 
 
 
 
 

-4.055*** 
(1.217) 

 
 
 
 

-0.001*** 
(0.000) 
-4.262*** 
(1.201) 

50 

Table 9 (cont’d) 
Panel B: Charter School Students 
  
Min. to Attended School 

(1) 

 

Min. to Nearest School 

 

Additional Min. 

 

Min. to Attended School Sq. 

 

Min. to Nearest School Sq. 

 

Additional Min. Sq. 

 

Constant 

 

0.018*** 
(0.006) 

 
 
 
 
 
 
 
 
 
 
-4.996*** 
(0.755) 

(2) 

0.023* 
(0.013) 

 
 
 
 

-0.000 
(0.000) 

 
 
 
 

 
 
 
 
 
 
 
 

 
 
 
 

-0.002 
(0.001) 

 
 

 
 
 
 
 
 
 
 

 
 
 
 
 
 

 
 
 
 

-0.001* 
(0.000) 
-4.950*** 
(0.759) 

(3) 
 
 

-0.030*** 
(0.011) 

(4) 
 
 

0.010 
(0.027) 

(5) 
 
 
 
 

0.022*** 
(0.005) 

(6) 
 
 
 
 

0.033*** 
(0.007) 

(7) 

0.019*** 
(0.006) 
-0.032*** 
(0.010) 

(8) 
0.021 
(0.013) 
0.005 
(0.027) 

 
 

-0.000 
(0.000) 
-0.002 
(0.001) 

 
 

(9) 

-0.012 
(0.012) 

(10) 
-0.023 
(0.018) 

 
 

0.032*** 
(0.010) 

 
 

0.038*** 
(0.011) 
0.001* 
(0.000) 

 
 

-0.001*** 
(0.000) 
-4.838*** 
(0.772) 

 
 
 
 
 
 

-5.019*** 
(0.762) 

-4.864*** 
(0.750) 

-5.070*** 
(0.743) 

-4.938*** 
(0.753) 

-4.898*** 
(0.754) 

-5.078*** 
(0.744) 

-4.898*** 
(0.754) 

Note. Total Observations Panel A: 15,043. Panel B: 15,768. Estimated coefficients are display as log odd coefficients. Standard errors clustered at the resident 
district level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Student’s gender, race, economically disadvantaged, English Learner, and disability statuses, the 
locale of the student’s resident district and the following school characteristics total enrollment, the percent of female, Black, Hispanic, Asian, Other Race, 
economically disadvantaged, and English Learner students, the percent of students with disabilities, the average standardized math score on state exams, and the 
average drive time at time t-1 are included as covariates. In Panel A I also include the percent of non-resident students at time t-1 as a covariate. Commute times 
are calculated using HERE API assuming normal traffic at 8 am on a weekday in April between 2019 and 2021. I determine the student’s nearest school, my 
proxy for assigned school, by calculating the geodetic distance to each school offering his or her grade in their resident district. I consider the school with the 
shortest distance to be the nearest school. Additional Minutes are the difference between the commute to the attended school and the student’s nearest school.

 

51 

imply that families not only consider overall distance but focus on distance to school relative to 

the distance to the default option when making school choice decisions. 

 

Because of the importance of location in families’ school choice decisions, it is unlikely 

that the existence of school choice policies alone will increase equitable access to effective 

schools or force schools to compete for students, even if families prefer high quality schools. 

Families are likely weighing the proximity of the school against the increases in quality their 

child may experience if he or she attended a school farther away from home. In particular, they 

may be considering the distance relative to other schooling options just as much if not more than 

overall distance to school, indicating that they are accounting for other choices when making 

their decisions.  

One possible policy that could mitigate the role of residential location in participation in 

school choice policies is the provision of transportation. School choice theorists claim that 

increased access and competition would be limited without an expanded school transportation 

system (Levin, 2015; Chubb & Moe, 1990). If parents are not personally responsible for 

transporting their children to and from school every day, they may be more willing and able to 

send them to schools farther from home. In fact, recent evidence from New York City finds that 

being eligible for the school bus mitigates the negative effects of distance on choice of school 

(Trajkovski, Zabel, & Schwartz, 2021). Furthermore, families may be able to continue to 

participate in choice programs after a residential move if there is guaranteed transportation. 

However, few states mandate that schools provide transportation for students participating in 

school choice (McShane & Shaw, 2020). More empirical evidence is needed concerning the 

effects of transportation on student outcomes and its cost-effectiveness to determine its 

feasibility as a policy solution. 

 

52 

My results also show that it is imperative that future work describing use of school choice 

policies or their effectiveness in increasing student outcomes should account for residential 

mobility and distance. For example, research that describes access to school choice options (i.e., 

Catt & Shaw, 2019; Blagg, & Chingos, 2017) should consider relative distance to school in 

addition to overall distance to various school choice options. Additionally, work examining 

differential exit from choice options should account for different residential mobility patterns 

between sectors. Finally, the literature concerning the effectiveness of school choice programs 

must incorporate the role of location and specifically residential mobility in their evaluations. A 

theory of school choice that does not account for physical access to schooling options, families’ 

preferences for proximity, and their residential choices will not accurately reflect the factors that 

influence families’ decisions.  

 

 

 

53 

APPENDIX

 

54 

Figure 7. Kaplan-Meier Survival Probabilities by Attended School Driving Time Quartile 
(Survival=remain in initial choice). Kindergarten to 5th Grade for Same Residence Sample  
         Figure 7A: Non-Resident Students           

         Figure 7B: Charter School Students 

    

 
Note. Students in the sample live at the same residence all analysis years. Non-residents are defined as students who 
attend a traditional public school that is not in their resident district. The unit of Analysis Time is years starting with 
the 2012-13 school year and ending with the 2017-18 school year. Survival probability is the probability of 
remaining in the same choice between each year. Quartiles are determined using the commute time to attended 
school in kindergarten. Commute times and distances are calculated using HERE API assuming normal traffic at 8 
am on a weekday in April between 2019 and 2021. 
 
 

 

 

55 

Figure 8. Kaplan-Meier Survival Probabilities Additional Driving Time Quartile 
(Survival=remain in initial choice). Kindergarten to 5th Grade 
Figure 8A: Full Sample Non-Resident Students  Figure 8B: Full Sample Charter Sch. Students 

 

 

 

 

Figure 8C: Same Res. Non-Resident Students    Figure 8D: Same Res. Charter School Students 

 

 

 
Note. Students in the same residence sample live at the same residence all analysis years. Non-residents are defined 
as students who attend a traditional public school that is not in their resident district. The unit of Analysis Time is 
years starting with the 2012-13 school year and ending with the 2017-18 school year. Survival probability is the 
probability of remaining in the same choice between each year. Quartiles are determined using the additional 
commute time to attended school past the nearest school in kindergarten. I determine the student’s nearest school, 
my proxy for assigned school, by calculating the geodetic distance to each school offering his or her grade in their 
resident district. I consider the school with the shortest distance to be the nearest school. Commute times and 
distances are calculated using HERE API assuming normal traffic at 8 am on a weekday in April between 2019 and 
2021. 
 
 
 

 

 

56 

REFERENCES

 

57 

REFERENCES 

 
 
 

 

 

58 

Abdulkadiroğlu, A., Angrist, J. D., Dynarski, S. M., Kane, T., & Pathak, P. (2011). 

Accountability and flexibility in public schools: Evidence from Boston's charters and 
pilots. The Quarterly Journal of Economics, 126(2), 699-748. doi: 10.1093/qje/qjr047. 

 
Bifulco, R., & Ladd, H. (2006). The impacts of charter schools on student achievement: 

Evidence from North Carolina. Education Finance and Policy, 1(1), 50-90. 

 
Bifulco, R., Ladd, H., & Ross, S. (2009). The effects of public school choice on those left 

behind: Evidence from Durham, North Carolina. Peabody Journal of Education, 84(2), 
130-149. 

 
Blagg, K., & Chingos, M. (2017). Who Could Benefit from School Choice: Mapping Access to 

Public and Private Schools. Washington, DC: Brookings Institution. Retrieved on 
February 24, 2020 from: https://www.brookings.edu/wp-
content/uploads/2017/03/es_20170330_chingos_evidence_speaks.pdf  

 
Blagg, K., Rosenboom, V., & Chingos, M. (2018). The Extra Mile: Time to School and Student 
Outcomes in Washington, D.C. Washington, DC: Urban Institute. Retrieved on July 30, 
2020 from: https://www.urban.org/research/publication/extra-mile-time-school-and-
student-outcomes-washington-dc  

 
Booker, T., Gilpatric, S., Gronberg, T., & Jansen, D. (2008). The Effect of Charter Schools on 

traditional Public School Students in Texas: Are Children Who Stay Behind Left Behind?  

 
Booker, K., Zimmer, R., & Buddin, R. (2005). The effects of charter schools on school peer 

composition. 

 
Candal, C. (Ed.). (2018). National charter school law rankings and scorecard-2018: The 
essential guide for policymakers and advocates. The Center for Education Reform. 
Retrieved on December 17, 2019 from: https://edreform.com/wp-
content/uploads/2019/01/CER_National-Charter-School-Law-Rankings-and-Scorecard-
2018_screen_1-30-19.pdf 

 
Carlson, D., & Lavertu, S. (2017). Interdistrict Open Enrollment in Ohio. Columbus, Ohio: 

Thomas B. Fordham Institute. Retrieved on February 24, 2020 from: 
https://fordhaminstitute.org/sites/default/files/publication/pdfs/FORDHAM%20Open%20
Enrollment%20Report_Online%20final_0.pdf  

 
Carlson, D., Lavery, L., & Hughes, T. (2018). Should I Stay or Should I Go? Open Enrollment 
Decisions and Student Achievement Trajectories. Social Science Quarterly, 99(3), 1089-
1104. 

Carlson, D., Lavery, L., & Witte, J. (2011). The determinants of inter-district open enrollment 

flows: Evidence from two states. Educational Evaluation and Policy Analysis, 33(1), 76-
94. doi: 10.3102/0162373710388643. 

 
Catt, A., & Shaw, M. (2019). Rural Communities Lacking Multi-Sector Schooling Options: 

Indiana’s School Choice Deserts. Journal of School Choice, 13(4), 576-598. doi: 
10.1080/15582159.2019.1691850.  

 
Chubb, J., & Moe, T. (1990). Politics, markets, and America's schools. Brookings Institution 

Press. 

 
Cordes, S. (2017). In pursuit of the common good: The spillover effects of charter schools on 

public school students in New York City. Education Finance and Policy, 1-49. 

 
Cordes, S.. & Schwartz, A. (2018). Does Pupil Transportation Close the School Quality Gap? 

Evidence from New York City. Research Report. Updated. Urban Institute. Retrieved on 
July 30, 2020 from: https://www.urban.org/research/publication/does-pupil-
transportation-close-school-quality-gap  

 
Cordes, S., Schwartz, A., & Stiefel, L. (2019). The effect of residential mobility on student 

performance: Evidence from New York City. American Educational Research 
Journal, 56(4), 1380-1411. 

 
Cordes, S., Schwartz, A., Stiefel, L., & Zabel, J. (2016). Is neighbourhood destiny? Exploring the 

link between neighbourhood mobility and student outcomes. Urban Studies, 53(2), 400-
417. 

 
Cowen, J., & Creed, B. (2017). Public school choice and student achievement: Evidence from 

Michigan’s inter-district open enrollment system. AERA Open, 3(3). doi: 
10.1177/2332858417731555.  

 
Cowen, J., Creed, B., & Keesler, V. (2015). Dynamic Participation in Inter-District Open 

 

Enrollment: Evidence from Michigan 2005-2013. Working Paper# 49. Education Policy 
Center at Michigan State University. Retrieved on February 24, 2020 from: 
https://files.eric.ed.gov/fulltext/ED574662.pdf  

Cowen, J., Edwards, D., Sattin-Bajaj, C., & Cosby, M. (2018). Motor City Miles: Student Travel 

to Schools in and Around Detroit. Washington, DC: Urban Institute. Retrieved on July 
30, 2020 from: https://www.urban.org/research/publication/motor-city-miles-student-
travel-schools-and-around-detroit/view/full_report  

 
Cowen, J., Fleming, D., Witte, J., & Wolf, P. (2012). Going public: Who leaves a large, 

longstanding, and widely available urban voucher program?. American Educational 
Research Journal, 49(2), 231-256. 

 

 

59 

Creed, B. (2016). The systemic effects of school choice induced competition: Defining 

competition and evaluating its effects on the outcomes of all students. Michigan State 
University. 

David, R., & Hesla, K. (2018). Estimated public charter school enrollment, 2017-2018. National 

Alliance for Public Charter Schools. Retrieved on May 29, 2020 from: 
https://www.publiccharters.org/sites/default/files/documents/2018-
03/FINAL%20Estimated%20Public%20Charter%20School%20Enrollment%2C%20201
7-18.pdf  

 
Denice, P., & Gross, B. (2018). Going the Distance: Understand the Benefits of a Long 

Commute to School. Washington, DC: Urban Institute. Retrieved on July 30, 3030 from: 
https://www.urban.org/research/publication/going-distance-understand-benefits-long-
commute-school  

 

Denice, P., & Gross, B. (2016). Choice, preferences, and constraints: Evidence from public 

school applications in Denver. Sociology of Education, 89(4), 300-320. 

 
Edwards, D. (in press). Over the River and through the Woods: Describing Who Uses School 

Choice in Rural Michigan. Journal of School Choice. 

 
Edwards, D. & Cowen, J. (2020). All Decisions are Local: How District Rules Can Promote or 

Restrict School Choices. 

 
Edwards, D. & Cowen, J. (2019) Who Chooses? Charter and Non-Resident Enrollment in 

Michigan. East Lansing, MI: Education Policy Innovation Collaborative. Retrieved on 
February 24, 2020 from: https://epicedpolicy.org/wp-
content/uploads/2019/06/EPIC_Choice_Policy_Brief_June2019.pdf  

 
Federal Communications Commission. (2015). More about census blocks. FCC Form 477. 

Retrieved on September 28, 2020 from: 
https://transition.fcc.gov/form477/Geo/more_about_census_blocks.pdf  

 
Finch, H., Lapsley, D., & Baker-Boudissa, M. (2009). A survival analysis of student mobility 

and retention in Indiana charter schools. education policy analysis archives, 17, 18. 

 
Friedman, M. (1962). Capitalism and freedom. Chicago, IL: University of Chicago Press. 
 
Glazerman, S., & Dotter, D. (2017). Market signals: evidence on the determinants and 

consequences of school choice from a citywide lottery. Educational Evaluation and 
Policy Analysis, 39(4), 593-619. 

 
Hanushek, E., Kain, J., & Rivkin, S. (2004). Disruption versus Tiebout improvement: The costs 

and benefits of switching schools. Journal of public Economics, 88(9-10), 1721-1746. 

 
 

 

60 

Harris, D., & Larsen, M. (2019). The identification of schooling preferences: Methods and 

evidence from post-Katrina New Orleans. Technical Report, Education Research 
Alliance for New Orleans. Retrieved on July 10, 2020 from: 
https://educationresearchalliancenola.org/files/publications/Harris-Larsen-How-Parents-
Choose-2019-07-05.pdf  

 
Harris, D., & Larsen, M. (2016). The effects of the New Orleans post-Katrina school reforms on 

student academic outcomes. Technical Report, Education Research Alliance for New 
Orleans. Retrieved on February 24, 2020 from: 
https://dspace.lafayette.edu/bitstream/handle/10385/2096/Larsen-
EducationResearchAllianceforNewOrleansTechnicalReport-2016.pdf?sequence=1  

 
Hastings, J., Kane, T., & Staiger, D. (2005). Parental preferences and school competition: 
Evidence from a public school choice program (No. w11805). National Bureau of 
Economic Research. 

 
Holme, J.,  & Richards, M. (2009). School choice and stratification in a regional context: 

Examining the role of inter-district choice. Peabody Journal of Education, 84(2), 150-
171. doi: 10.1080/01619560902810120. 

 
Imberman, S. (2011). The effect of charter schools on achievement and behavior of public school 

students. Journal of Public Economics, 95(7-8), 850-863. 

 
Jochim, A., DeArmond, M., Gross, B., & Lake, R. (2014). How parents experience public school 

choice. Center for Reinventing Public Education. Retrieved on January 19, 2020 from: 
https://www.crpe.org/publications/how-parents-experience-public-school-choice  

 
Lavery, L., & Carlson, D. (2015). Dynamic participation in inter-district open 

enrollment. Educational Policy, 29(5), 746-779. doi: 10.1177/0895904813518103. 

 
Lenhoff, S., Singer, J., Pogodzinski, B., & Cook, W. (2020). Exiting Detroit for school: 

inequitable choice sets and school quality. Journal of Education Policy, 1-23. 

 
Lenhoff, S., Singer, J., Stokes, K., & Mahowald, J. (2021). Why do Detroit students miss school? 

Implications for returning to school after COVID-19. Detroit Education Research 
Partnership, Wayne State University. https://education.wayne.edu/detroit-education-
research-partnership  

 
Levin, H. (2015). Issues in education privatization. In H.F. Ladd & M.E. Goertz (Eds.), 
Handbook of research in education finance and policy (pp. 411-422). Routledge. 

 
Lincove, J., Cowen, J., & Imbrogno, J. (2018). What's in Your Portfolio? How Parents Rank 

Traditional Public, Private, and Charter Schools in Post-Katrina New Orleans’ Citywide 
System of School Choice. Education Finance and Policy, 13(2), 194-226. 

 

 

61 

Manton, K., Singer, B., & Woodbury, M. (1992). Some issues in the quantitative characterization 

of heterogeneous populations. In J. Trussell, R. Hankison, & J. Tilton (Eds.), 
Demographic Applications of Event History Analysis (pp. 8–37). Oxford, England: 
Clarendon Press. 

 
McShane, M. & Shaw, M. (2020). Transporting school choice students: A primer on states’ 

transportation policies related to private, charter, and open enrollment students. 
EdChoice. Retrieved on May 29, 2020 from: https://www.edchoice.org/wp-
content/uploads/2020/03/Transporting-School-Choice-Students-by-Michael-Q-McShane-
and-Michael-Shaw.pdf  

 
Michigan Department of Education. (2013). Michigan State School Aid Act [MCL 388.1705; 

388.1705c; Sections 105, 105c]. Retrieved on December 17, 2019 from: 
http://mi.gov/documents/mde/choice1_279579_7.pdf  

 
Michigan Department of Education. (2017). Charter Schools- Questions and Answers. Retrieved 

on December 17, 2019 from: 
https://www.michigan.gov/documents/PSAQA_54517_7.pdf    

Ni, Y. (2012). The sorting effect of charter schools on student composition in traditional public 

schools. Educational Policy, 26(2), 215-242. 

 
Nichols-Barrer, I., Gleason, P., Gill, B., & Tuttle, C. (2016). Student selection, attrition, and 

replacement in KIPP middle schools. Educational Evaluation and Policy Analysis, 38(1), 
5-20. 

 
Pribesh, S., & Downey, D. (1999). Why are residential and school moves associated with poor 

 

school performance?. Demography, 36(4), 521-534. 

 
Reback, R. (2008). Demand (and supply) in an inter-district public school choice 

program. Economics of Education Review, 27(4), 402-416. doi: 
10.1016/j.econedrev.2008.01.001. 

 
Sass, T. (2006). Charter schools and student achievement in Florida. Education Finance and 

Policy, 1(1), 91–122.  

 
Singer, J. (2020). Student stratification among a combination of school choice policies in 

Detroit. Journal of School Choice, 14(1), 122-153. 

 
Stein, M., Burdick-Will, J. & Grigg, J. (2020). A choice too far: Transit difficulty and early high 

school transfer. Educational Researcher. Online first.  

 
Stein, M., & Grigg, J. (2019). Missing bus, missing school: Establishing the relationship between 

public transit use and student absenteeism. American Educational Research 
Journal, 56(5), 1834-1860. 

 

 

62 

Swanson, C., & Schneider, B. (1999). Students on the move: Residential and educational 

mobility in America's schools. Sociology of Education, 54-67. 

 
Trajkovski, S., Zabel, J., & Schwartz, A. (2021). Do school buses make school choice 

work?. Regional Science and Urban Economics, 86, 103607. 

 
Urban Institute Student Transportation Working Group. (2018). The road to school: How far 

students travel to school in Denver, Detroit, New Orleans, New York City, and 
Washington, DC. Washington, DC: Urban Institute. Retrieved on February 24, 2020 
from:https://greaterdc.urban.org/sites/default/files/publication/97151/the_road_to_school
_7.pdf.  

 
U.S. Department of Education, National Center for Education Statistics. (2020). Characteristics 

of elementary and secondary schools. Retrieved on May 29, 2020 from: 
https://nces.ed.gov/programs/coe/indicator_cla.asp  

 
Voight, A., Shinn, M., & Nation, M. (2012). The longitudinal effects of residential mobility on 
the academic achievement of urban elementary and middle school students. Educational 
Researcher, 41(9), 385-392. 

 
Wang, K., Rathburn, A., & Musu, L. (2019). School Choice in the United States: 2019. NCES 

2019-106. National Center for Education Statistics. Retrieved on February 24, 2020 
from: https://nces.ed.gov/pubs2019/2019106.pdf  

 
Welsch, D., Statz, B., & Skidmore, M. (2010). An examination of inter-district public school 

transfers in Wisconsin. Economics of Education Review, 29(1), 126-137. 
doi:10.1016/j.econedrev.2009.07.008. 

 
Winters, M., Clayton, G., & Carpenter II, D. (2017). Are low-performing students more likely to 
exit charter schools? Evidence from New York City and Denver, Colorado. Economics of 
Education Review, 56, 110-117. 

 
Winters, M. (2015). Understanding the gap in special education enrollments between charter and 

traditional public schools: Evidence from Denver, Colorado. Educational 
Researcher, 44(4), 228-236. 

 
Winters, M. (2014). Why the Gap? English Language Learners and New York City Charter 

Schools. Civic Report No. 93. Manhattan Institute for Policy Research. Retrieved on July 
30 202 from: https://files.eric.ed.gov/fulltext/ED564925.pdf  

 
Winters, M. (2012). Measuring the effect of charter schools on public school student 

achievement in an urban environment: Evidence from New York City. Economics of 
Education review, 31(2), 293-301. 

 

 

63 

Wixom, M. & Keily, T. (2018). 50-State Comparison: Open Enrollment Policies (2018 Update) 

Education Commission of the States. Retrieved on December 17, 2019 from: 
https://www.ecs.org/open-enrollment-policies/  

Ziebarth, T. (2019). Measuring up to the model: A ranking of state public charter school 

laws. National alliance for public charter schools. Retrieved on July 30, 2020 from: 
https://www.publiccharters.org/ranking-state-public-charter-school-laws-2019  

 
Zimmer, R. & Buddin, R. (2009). Is Charter School Competition in California Improving the 
Performance of Traditional Public Schools? Public Administration Review. 69(5): 831- 
845.  

 
Zimmer, R., Gill, B., Booker, K., Lavertu, S., Sass, T., & Witte, J. (2009). Charter Schools in 

Eight States: Effects on achievement, attainment, integration, and competition. MG-869. 
Pittsburgh: RAND.  

 
Zimmer, R., & Guarino, C. M. (2013). Is there empirical evidence that charter schools “push 

out” low-performing students?. Educational Evaluation and Policy Analysis, 35(4), 461-
480.  

 

 
 

 

 

64 

PAPER 2 JUST OUT OF REACH? UNRESTRAINED SUPPLY, CONSTRAINED 

DEMAND, AND ACCESS TO EFFECTIVE SCHOOLS IN AND AROUND DETROIT 

Introduction 

Over the last three decades, school choice policies have increased the number and 

diversity of public schooling options available to students beyond a residential school 

assignment. Currently, thirty-four states have mandatory open enrollment policies, and forty-four 

states have charter school laws (Wixom & Keily, 2018; Wixom, 2018). About forty percent of 

families recognize that they have a choice in public schools, and one in four students in the 

United States enroll in a school other than the one they were assigned (National Center for 

Education Statistics, 2018). In cities, the number of students choosing to attend a non-

neighborhood school is closer to fifty percent (Jochim, DeArmond, Gross, & Lake, 2014).  

Some proponents of school choice argue that it can increase the supply of effective 

schools in addition to access to these schools. In theory, allowing students to attend schools 

outside of their neighborhood or even their local district breaks the monopoly that individual 

schools and districts have on students and thus creates competition for students. This encourages 

the establishment of many schools with a variety of curriculum, offerings, and management 

strategies to meet the needs and preferences of students and their families. Under such a system, 

schools may change their offerings to meet families’ preferences in order to attract students. If 

they fail to attract students, they will shut down. Therefore, competition for students would result 

in the creation of a more effective supply of schools as long as parents prefer academic 

performance or school’s effectiveness in increasing achievement, have accurate and transparent 

information concerning school quality, and the ability to overcome barriers associated with 

attending school outside of their neighborhood (Chubb & Moe, 1990; Friedman, 1962).  

 

65 

Critics of this theory have argued that competition, by itself, cannot regulate the quality 

of the supply of schools. Since schooling is compulsory, schools will draw students regardless of 

the school’s quality if there is no other options or at least no more effective options. Furthermore, 

there is little incentive for effective or high performing schools to expand their capacity (Harris, 

2017). Therefore, more recent school choice scholars have proposed systems of managed 

competition or portfolio models where schools are held accountable for their academic 

performance by districts or authorizers in addition to families (Harris, 2017; Hill, 2006). These 

proposals also call for cooperation and coordination across schools within an education market in 

order to regulate not only the quality but the quantity of schools.  

Regardless of whether school choice policies increase the supply of effective schools, 

these policies will only increase access to more effective options if there is demand for effective 

schools. To accomplish this, families must prefer academic quality over other school 

characteristics, effective schools must be accessible, and there must exist accurate and 

transparent information concerning school quality (Harris, 2017; Levin, 2015). Prior work on 

parent preferences in choice-rich environments finds that families do value academic 

achievement. However, there are other school characteristics they value similarly. The most 

prevalent is proximity to home—even in a market where many schools are in principle available 

(Denice & Gross, 2016; Glazerman & Dotter, 2017; Harris & Larsen, 2019; Hastings, Kane, & 

Staiger, 2005; Lincove, Cowen, & Imbrogno, 2018). Although much of this literature frames 

proximity as a preference, distance is likely a constraint on the quantity of schools families’ have 

the ability to attend, particularly for the most disadvantaged students. Families living in high-

poverty neighborhoods are less likely to have access to a car (Urban Institute Student 

Transportation Working Group, 2018). Furthermore, most parents in choice-rich cities report that 

 

66 

they drive their students to school with about one third reporting transportation as a barrier to 

their preferred schools (Jochim et al., 2014).  

In this paper, I test whether distance and district boundaries are constraints to accessing 

high quality and effective schools, as measured by academic performance, for students living in 

Detroit, Michigan. Detroit provides an ideal context to test whether the supply of effective 

schools is constrained by geographic barriers for a few reasons. First, Detroit students have many 

schooling options that are geographically dispersed. They can choose from schools within their 

own district (intra-district choice), schools in surrounding districts (inter-district choice), and 

charter schools. Additionally, there is little accountability, coordination, transparency, or 

accessibility in the Detroit education market. Michigan’s charter school law receives high marks 

from pro-school choice organizations for its lack of regulation on and use of non-district 

authorizers (Candal, 2018; Ziebarth, 2019). Since universities and colleges can authorize charter 

schools throughout the state with little oversight from the state or districts and there are no caps 

on brick and mortar open enrollment charter schools (Michigan Department of Education, 2017), 

little coordinated oversight of the quantity or quality of schools exists in and around Detroit. 

Also, there exists no centralized enrollment system or universal transportation policies for 

Detroit schools unlike other cities with similar levels of choice including New Orleans. Since 

Detroit has an unrestrained supply of schools with little regulation on their accessibility, it is an 

ideal case to test whether geography constrains families’ ability to attend preferred schools in 

absence of universal enrollment or transportation policies. Furthermore, this permits the test of 

the role of the residentially assigned school in school choice decisions as well. In sum, I test 

whether the mere existence of school choice policies are enough to provide meaningful choice 

between schools. 

 

67 

To do so, I first describe the role of geography, in terms of district boundaries and 

distance, in the use of formal school choice policies. I focus on whether students who are 

disadvantaged in terms of poverty or access to transportation are more or less likely to use school 

choice. Furthermore, recognizing that in a high-choice environment the decision to enroll in 

one’s default neighborhood school is itself a “school choice,” I provide some of the first 

evidence for the ways in which a student’s residentially assigned school may factor into family 

decision-making. I accomplish this using a set of multinomial logistic regressions that predict 

sector and location of school attended as a function of student and assigned school 

characteristics. Then, I test whether Detroit students attend more effective schools, as measured 

by their contribution to student’s test scores, or schools with higher levels of achievement within 

their choice sets using discrete choice models. More specifically, I ask: 

1.  What are the roles of poverty, access to transportation, and student’s assigned school 

in use of school choice? 

2.  Do Detroit students attend the more effective schools and/or the schools with the 

highest levels of achievement within their choice sets?  

 I provide evidence that families’ choice sets are likely constrained to within city 

boundaries or within their TPS district. Specifically, disadvantaged students, particularly those 

living in neighborhoods where families have lower incomes and do not have access to a car are 

less likely to participate in public school choice, especially the options located outside of city 

limits. Furthermore, I show that families’ preferences for academic quality and school 

effectiveness are stronger when choice sets are restricted to schools within Detroit or DPSCD 

and in earlier grades. I conclude by recommending that increasing the accessibility of choice 

 

68 

options through transportation and enrollment policies may increase access to effective schools 

in a choice-rich system.  

This study adds to the literature concerning preferences for schools in choice-rich cities 

in multiple ways. First, it is one of the few studies to frame distance and district boundaries as 

barriers to attending effective schools instead of a preference, especially for disadvantaged 

families. Furthermore, I test whether these constraints limit access to academic quality. Also, I 

describe choice in a deregulated environment that lacks a centralized enrollment system, retains 

little local control of the supply of schools, and allows students to attend schools outside of 

district boundaries. In particular, this paper is one of the first to expand students’ choice sets to 

include choices available through inter-district choice—a prevalent but understudied form of 

choice—where other major forms of choice within multiple public school systems is possible.  

Motivation: Supply of, Demand for, and Access to Effective Schools in Choice-Rich Areas 

Supply  

 

Although school choice advocates argue that school choice can increase the effectiveness 

of the supply of schools, little empirical evidence exists that competition, by itself, increases 

school quality, even when there are preferences for school effectiveness. First, studies of school 

leaders’ perceptions of and responses to competition for students in Milwaukee and New 

Orleans, two choice-rich cities, find that schools are more likely to increase the marketing of 

their schools than improve achievement or instruction in response to competitive pressures 

(Jabbar, 2015; Loeb, Valant, & Kasman, 2011). Furthermore, studies that measure the extent to 

which competitive pressures from charter and private schools increase achievement of students 

remaining in traditional public schools find some small positive effects (Booker, Gilpatric, 

Gronberg, & Jansen, 2008; Cordes, 2017; Figlio & Hart, 2014; Sass, 2006; Winters, 2012) with 

 

69 

some studies finding no effect (Zimmer & Buddin, 2009; Zimmer, Gill, Booker, Lavertu, Sass, & 

Witte, 2009), and one finding some negative effects (Imberman, 2011).  

 

A set of studies of school choice in New Orleans also illustrate that even when there is 

demand for academic quality, competition, by itself, may not increase the supply of effective 

schools. In New Orleans, research shows that families have preferences for schools with high 

levels of academic achievement as measured by school performance ratings (Harris & Larsen, 

2019; Lincove, Cowen, & Imbrogno, 2018). Although families listed high performing schools as 

their preferred option on the common application, less that 40 percent of students attend their 

first choice schools. Most attend lower quality schools than their first choice due to the limited 

supply of high performing schools in New Orleans (Lincove, Valant, & Cowen, 2018). 

Furthermore, a recent study finds that most of the improvements in the effectiveness, as 

measured by value-added measures, of the supply of schools in New Orleans over time can be 

attributed to the takeover of low performing schools and the opening of higher quality schools 

rather than the development of existing schools (Harris, Liu, Gerry, & Arce-Trigatti, 2019). Most 

of these takeovers and openings were a result of the performance-based charter authorizing 

process instead of competition by itself (Bross, Harris, & Liu, 2016).  

Demand  

Taken together, there is inconsistent evidence that shows school choice policies create a 

more effective supply of schools without external performance accountability. However, these 

policies can still increase access to effective schools if three conditions are met: families must 

have strong preferences for academic performance, the ability to overcome barriers associated 

with attending a school other than the assigned school, and accurate information concerning 

 

70 

school quality (Chubb & Moe, 1990; Harris, 2017; Levin, 2015). I discuss the empirical evidence 

concerning each of these assumptions in turn. 

Parental Preferences 

When surveyed, parents state that the most important single factor when choosing a 

school is academic quality. This stated preference is especially strong for low income families 

(Bell, 2009; Jochim et al., 2014; Schneider & Buckley, 2002; Schneider, Teske, Marshall, & 

Roch, 1998). Nonetheless, self-reported preferences are subject to social desirability bias. For 

example, when surveyed, Washington, D.C. parents reported that they prioritize academic 

quality over other factors in their school search. However, an analysis of their internet searches 

shows that parents search schools’ student demographics and location more often than their test 

scores (Schneider & Buckley, 2002).  

Studies that use parents’ rank-ordered (i.e. revealed) preferences on enrollment 

applications show that parents value higher levels of academic achievement, as measured by the 

schools’ average test scores, accountability ratings, and peer quality (Abdulkadiroglu, Pathak, 

Schellenberg, & Walters, 2020; Beuermann, Jackson, Navarro-Sola, & Pardo, 2019; Denice & 

Gross, 2016; Glazerman & Dotter, 2017; Glazerman, 1998; Harris & Larsen, 2019; Hastings, 

Kane, & Staiger, 2005; Lincove, Cowen, & Imbrogno, 2018). Results concerning preferences for 

effective schools, as measured by their contribution to student achievement as growth percentiles 

or value-added measures, are mixed. In New York City and Washington, D.C. there is little 

evidence that families prefer schools with larger contributions to student achievement 

(Abdulkadiroglu et al. 2020; Glazerman & Dotter, 2017). In contrast, studies of New Orleans and 

Trinidad and Tobago find that parents value schools with higher valued added measures on high 

stakes tests in addition to overall achievement (Beuermann et al., 2019; Harris & Larsen, 2019). 

 

71 

In this study, I use state-calculated accountability ratings and author-constructed school level 

value-added measures to test whether parents prefer academic quality when choosing schools. 

Although parents have strong preferences for academic achievement, they value other 

qualities as well as achievement. Studies of Washington, D.C. and Minneapolis show that 

families prefer schools where their child is not a racial minority (Glazerman & Dotter, 2017; 

Glazerman, 1998). Specifically, Denice and Gross (2016) find that White families in Denver 

have preferences for schools with few racial minority students. In New Orleans, families have a 

strong preference for extracurricular activities and childcare (Harris & Larsen, 2019). 

Additionally, some evidence exists that families prefer district-run schools as well as private 

schools over charter schools in Denver and New Orleans (Denice & Gross, 2016; Lincove, 

Cowen, & Imbrogno, 2018). 

One of the most valued school characteristics is proximity from home (Abdulkadiroglu et 

al. 2020; Beuermann et al., 2019; Denice & Gross, 2016; Glazerman & Dotter, 2017; Glazerman, 

1998; Harris & Larsen, 2019; Hastings, Kane, & Staiger, 2005; Lincove, Cowen, & Imbrogno, 

2018; Singer, 2020). Families prefer schools that are closer to home, have shorter commutes, and 

are accessible by public transportation. To my knowledge, only one other study examines how 

student choice sets are constrained by geography. Denice and Gross (2016) find that when choice 

sets are restricted to a 2 mile radius around a student’s home, all families, regardless of race, 

have strong preferences for high performing schools, implying that distance is a possible 

constraint to access to effective schools. I add to this literature by providing some of the first 

empirical evidence that families’ choices—especially those of income-disadvantaged families—

are meaningfully constrained by distance and geography even with multiple options legally 

available. Additionally, I examine the role of the student’s assigned school in their use of school 

 

72 

choice. Then, I use this information to restrict students’ choice sets to estimate their underlying 

preferences for academic quality. 

Preferences for school characteristics also vary across race, income, and academic ability. 

Racial minority, lower income, and lower achieving students have an especially strong 

preference for proximity that outweighs their preference for academic achievement (Denice & 

Gross, 2016; Harris & Larsen, 2019; Hastings, Kane, & Staiger, 2005). For example, Hastings, 

Kane, and Staiger (2005) find that as their child’s test score and their income increase, families 

are more willing to travel farther to access higher quality schools. Differential preferences 

between advantaged and disadvantaged students could create a two-tiered system where high 

income and high achieving students attend high performing schools and while low income and 

low achieving students are stuck in their neighborhood schools (Hastings, Kane, & Staiger, 

2005).  

Barriers to Access  

 

The difference in stated and revealed preferences, particularly for disadvantaged students, 

may be driven by constraints on their ability to overcome barriers associated with attending 

school outside of their neighborhoods. First, the revealed preference for schools closer to home 

could in fact be due to a lack of access to transportation. Families may be unable, but not 

unwilling to attend higher quality schools farther from home. Few states require schools to 

provide transportation to students using choice policies to attend them (McShane & Shaw, 2020). 

Thus, it tends to be the sole responsibility of the parent to get their child to and from school 

every day. This may be particularly difficult for low income families who are less likely to have 

access to a car or other forms of direct transportation (Urban Institute Student Transportation 

Working Group, 2018). In choice-rich cities, over twenty percent of parents report that 

 

73 

transportation is a barrier in sending their child to the school of their choice (Jochim et al., 2014). 

Furthermore, differences in preferences for academic achievement by race disappear when 

choice sets are restricted to schools close to home (Denice & Gross, 2016). Therefore, it is likely 

that students attend schools near their residence out of necessity rather than solely preference. 

 

Another likely set of barriers to access to effective schools is enrollment systems, 

practices, and policies. Parents, especially disadvantaged parents, cite application deadlines, the 

number of applications, difficulty with paperwork, and confusion with eligibility requirements as 

problems when choosing schools (Jochim et al., 2014). Furthermore, unregulated enrollment 

systems could create unequal access by providing opportunities for schools to select the best 

students or push out lower achieving students. One solution that can ease the burden on parents 

and create more equitable opportunities is unified enrollment systems where parents only fill out 

one application and a third party determines enrollment through a random lottery. At least eight 

cities have instituted a unified enrollment system over the last decade (Ekmekci & Yenmez, 

2019). Parents in Denver report that the common application made the enrollment process more 

manageable (Gross, DeArmond, & Denice, 2015). Furthermore, causal evidence exists showing 

that the adoption of the unified enrollment system increased the percent of minority students and 

English Learners participating in charter school choice in Denver (Winters, 2015).  

Accurate and Transparent Information 

 

Families also need accurate information on school quality in order to choose effective 

schools. Few families have accurate information on the school’s contribution to learning or 

achievement and instead rely on the social connections and the demographics of the student body 

to determine school quality (Schneider & Buckley, 2002; Schneider et al., 1998). About a quarter 

of families reported that they were unable to find an adequate amount of information to make the 

 

74 

best school choice decision for their child (Jochim et al., 2014). However, information 

interventions that send school-level achievement information to families have increased the 

number of students attending effective schools (Hastings & Weinstein, 2008; Valant, 2014). 

Although many unified enrollment systems also increase the amount of publicly available 

information concerning school quality and offerings, parents in cities with these systems report 

wanting more tailored and rich information about schools (Gross, DeArmond, & Denice, 2015).  

Background on School Choice in Michigan and Detroit 

The amount of choice, and, in particular, the lack of centralized planning concerning the 

quantity and quality of schools serving Detroit students has been indicated as a contributor to the 

well-documented financial decline and eventual bankruptcy of Detroit Public Schools (Strauss 

July 15 2016). Detroit has an option demand choice system where students are assigned to a 

school within the newly formed Detroit Public Schools Community District (DPSCD) based on 

their residence but can option to attend other schools if the receiving school has space available 

similar to Denver and Washington, D.C. (Bell, 2009). For over twenty years, Detroit students 

have been able to attend charter schools located inside and outside of Detroit, schools in other 

districts, and other traditional public schools in Detroit instead of their assigned school. The 

Michigan legislature passed Act 105 of the Revised School Code in 1996 enacting the Schools of 

Choice Program which oversees inter-district choice in Michigan (Michigan Department of 

Education, 2013). Under Acts 105 and 105c, Michigan school districts have the option to accept 

students from other districts in their intermediate school district and surrounding intermediate 

school districts (ISDs). Districts that choose to receive students through Schools of Choice 

decide the number of students they enroll, the grades, programs, and buildings non-resident 

students can enroll in, and the timeframe they accept applications. Thus, sending districts, 

 

75 

including Detroit, are not guaranteed to enroll a majority of their resident population. If a district 

chooses to not participate in Schools of Choice, they still are able to make agreements with 

specific districts to accept their students or they can accept non-resident students on a case by 

case basis. Districts are not required to provide transportation to non-resident students, but they 

can if they choose to do so (Michigan Department of Education, 2013). For example, less than 

twenty percent of Metro Detroit districts offer any transportation to non-resident students with 

only two of almost seventy districts reporting that they will cross district boundaries to provide 

transit. During the 2017-18 school year, Detroit students attended 60 different traditional public 

school districts within the three surrounding intermediate school districts. 

Additionally, Michigan districts can choose to allow students to attend schools other than 

the one they are assigned to within the district through intra-district choice. Within DPSCD, 

students are able to choose from their assigned school, other neighborhood schools, and magnet 

schools. There were 24 magnet schools in DPSCD with 14 having competitive application or 

examination requirements during of the 2018-19 school year. DPSCD only guarantees 

transportation to K-8 students who attend their assigned school and live at least three quarters of 

a mile away from their school. All DPSCD high school students receive bus passes (Urban 

Institute Student Transportation Working Group, 2018).  

Furthermore, Detroit has one of the highest rates of charter school attendance of cities in 

the United States at almost fifty percent. Only New Orleans and Washington, D.C. have a higher 

percentage of students attending charter schools (Hesla, White, & Gerstenfeld, 2019). In 1993, 

Michigan adopted Part 6A of the Revised School Code which allowed districts, intermediate 

school districts, community colleges, and universities to authorize charter schools. (Michigan 

Department of Education, 2017). In contrast with the majority of charter school laws in the 

 

76 

United States, authorizers in Michigan may approve applications for charter schools located 

anywhere in the state without oversight from state or local governments (Wixom, 2018). Thus, 

there is no one body overseeing where schools open, how many schools open, when schools 

open or close, and what types of schools exist in Detroit. In contrast, New Orleans only allows 

the Orleans Parish School Board and the state to authorize charter schools (Wixom, 2018). While 

charter schools in Michigan cannot practice selective admissions policies, enrollment may be a 

barrier since each charter school oversees its own application process leaving families to apply to 

each school individually with no guarantees of enrollment in contrast to Denver, Washington, 

D.C., and New Orleans that have centralized enrollment systems. Furthermore, charter schools 

are not required to provide transportation to its students, unlike New Orleans and Washington, 

D.C. (Urban Institute Student Transportation Working Group, 2018). During the 2018-19 school 

year, there were 380 charter schools operating in Michigan, with almost one hundred located 

within the boundaries of DPSCD. 

Compared to other choice-rich cities including New Orleans, Denver, and Washington, 

D.C., Detroit’s system of school choice remains one of the least regulated due to its lack of 

planning concerning school openings and closings, centralized enrollment systems, and universal 

transportation policies. When the Michigan legislature was crafting the bailout of Detroit Public 

Schools in 2016, the creation of a non-partisan entity, the Detroit Education Commission, was 

proposed to oversee the opening and closing of schools, serve as an accountability mechanism, 

manage a centralized enrollment system, and coordinate transportation needs across all 

traditional public schools and charter schools in Detroit (Coalition for the Future of Detroit 

Schoolchildren, 2015). While the state approved the return of an elected school board and the 

creation of a new debt-free school district, the Detroit Public Schools Community District, no 

 

77 

city-wide oversight, enrollment system, or transportation policy was put in place (Gray June 9 

2016). 

During his 2018 State of the City address, Mayor Duggan announced the Get On and 

Learn bus loop (GOAL Line) as one of the first efforts of citywide coordination between DPSCD 

and charter schools. His objective was to decrease the number of Detroit residents attending 

schools outside of the city since over twenty percent were leaving the city every day to attend 

school (Levin March 21 2019). Currently, fourteen schools in northwest Detroit, half of which 

are charter schools, participate in the bus loop where students can get on at any of the schools to 

travel to any other school on the loop (Community Education Commission, 2019). In addition to 

transportation to and from school, GOAL Line provides transportation to an after school 

program. An analysis of ridership and parent surveys finds that GOAL Line is mostly used for its 

after school program rather than transportation (Edwards, Anderson, & Mohr, 2019). 

Data 

The main sources of data for this paper are student-level records from 2012-13 to 2017-

18 provided by the Michigan Department of Education (MDE) and the Center for Educational 

Performance and Information (CEPI). These records include student addresses geocoded at the 

census block level, student demographic information (e.g., race and ethnicity, gender, disability 

status, English learner status, and economic disadvantaged status11), and student test scores on 

the Michigan Student Test for Educational Progress (M-STEP) for all Michigan public school 

students. I generate the school-level variables used in my analysis from the student-level data 

and a variety of publicly available datasets made available by MDE and CEPI. These school-

 

11In Michigan, students are considered economically disadvantaged if they qualify for free or reduced lunch, receive 
food (SNAP) or cash assistance (TANF), or they are homeless, migrant, or in foster care. 
 

 

78 

level datasets contain information about all schools in Michigan including the school’s sector, 

address, their accountability rating on Michigan’s Overall School Index System, and graduation 

rates. Additionally, I create a set of variables describing the programs schools offer using the 

2018-19 Detroit Parents’ Guide to Schools which includes application requirements, uniform 

requirements, transportation, before/after school care, and top activities for all public schools 

located within the boundaries of DPSCD. Finally, I use 2017-18 catchment zone boundaries 

provided by DPSCD to determine students’ residentially assigned schools.12 

Measures of Sector and Location 

 

To answer my research questions, I estimate the relationships between student and school 

characteristics and the use of choice for students living in Detroit. As discussed above, Detroit 

students can attend their assigned school, other DPSCD schools, traditional public schools (TPS) 

in other districts, and charter schools located inside and outside of city limits. Therefore, I define 

use of choice as a categorical variable with five mutually exclusive outcomes determined by the 

sector and location of the school attended: the student’s assigned school, another DPSCD school, 

a charter school located inside of Detroit, a TPS in another district, and a charter school located 

outside of Detroit. First, I determine the students’ assigned schools using the coordinates of the 

population weighted centroids of their residences and the 2017-18 DPSCD catchment zones. I 

 

12 I begin with 112,756 students living within the boundaries of DPSCD during the 2017-18 school year. I use 
census block to school district conversions curated by the Missouri Census Data Center to determine which students 
live in Detroit. I exclude 28 students living in 10 census blocks that are split between catchment zones. Next, I 
restrict the sample to students attending a traditional public school or charter school offering general education in the 
three intermediate school districts serving the majority of Detroit students, Wayne Regional Education Service 
Agency (RESA), Oakland Schools, and Macomb Intermediate School District. I exclude 4,372 (3.9%) students 
attending virtual schools, boarding schools, schools in juvenile detention centers, and other residential schools 
because they do not commute to school on a daily basis. 18,101 (16.7%) students in my sample attend multiple 
schools during the 2017-18 school year. For these students, I use the observation associated with the school that 
determined by MDE and CEPI to be the student’s most primary enrollment. For the 5,567 (5.1%) students who have 
multiple residences while attending their most primary school, I choose a residence to use at random since I am 
unable to discern which residence was their primary residence throughout the school year. Finally, I exclude 271 
observations of students living in census tracts without reported median income. My full sample includes 108,085 
students. 

 

79 

consider all other DPSCD schools, regardless if they are neighborhood or magnet schools, as 

Other DPSCD schools for that particular student. I determine whether or not a charter school and 

is located inside or outside of Detroit using the school’s coordinates and the 2016-17 Michigan 

school district boundaries from the Michigan Department of Technology, Management, and 

Budget, the most recent year available at the time of analysis. 

I calculate driving times (in minutes) for each student from his or her census block or 

tract to his or her assigned school for all students in the sample. For distances within a 2-mile 

radius (as the “crow flies”), drive times are calculated from the center of each students’ home 

census block to the center of each school’s census block. For distances more than 2 miles, drive 

times are calculated from the center from each students’ home census tract to the center of the 

school’s census tract. I estimate driving times using the Google Distance Matrix application 

programming interface (API). Because Google API does not calculate drive times from the past, 

drive times are calculated using predicted travel conditions for weekdays between September 

2017 and April 2020 assuming usual traffic at 8 am. Drive times are calculated for weekdays 

during the months of the school year where little to no snow would be expected (September 

through November and March through May).  

Measures of Poverty and Access to Transportation 

 

My first research question examines choice use for students who are likely to face 

difficulty attending schools farther away from home. Two groups of students that are likely to 

have difficulty overcoming barriers associated with using school choice are impoverished 

students and students without access to transportation. In addition to the measure of student 

economic disadvantage provided by MDE and CEPI, I use the median income of the student’s 

resident census tract from the 2017 American Community Survey (ACS) as a measure of 

 

80 

poverty. My first measure of transportation access, which is also from the ACS, is the percent of 

residents that do not own a car in the student’s resident census tract. Since Detroit has an 

inefficient public transportation system compared to other choice rich cities and the majority of 

Detroit parents report driving their children to school, car ownership is likely necessary to 

transport students to schools outside of their neighborhoods (Jochim et al., 2014; Urban Institute 

Student Transportation Working Group, 2018).  

Another factor in choice use may be whether the school itself provides transportation. 

The only school a Detroit student may be guaranteed transportation to is their assigned school. 

Students attending their assigned school and living more than 0.75 miles walking distance from 

that school may ride the school bus in elementary and middle school. To account for the role of 

school provided transportation in decisions to participate in school choice, I create an indicator 

of transportation eligibility. To determine whether a student is eligible for transportation to their 

assigned school, I calculate the walking distance from the centroid of the student’s resident 

census block to the school using Here Technologies API assuming average traffic. I consider 

students who are in kindergarten through eighth grade who live more than 0.75 miles from their 

assigned school as eligible for transportation to their assigned school. 

In Table 10, I provide summary statistics for my full sample as well as differences by 

sector of school attended. The vast majority of the students in my sample are Black and 

economically disadvantaged. Students who attend their assigned school have shorter drive times 

to their assigned school on average than the students attending schools in other sectors. 

Additionally, students attending schools outside of Detroit live in census tracts with slightly 

higher rates of car ownership and higher median incomes. 

 
 

 

81 

Table 10. Student and Neighborhood Characteristics by Sector of School Attended 

Attends 
Assigned 
School 

22,838 
90% 
$28,410  
4.04 
23% 
32% 
48% 
81% 
3% 
2% 
14% 
1% 
14% 
14% 

Attends 
Other 
DPSCD 
School 
27,629 
86% 
$29,549  
5.31 
23% 
37% 
49% 
84% 
2% 
1% 
13% 
1% 
18% 
12% 

Attends 
Outside 

Full 

Sample 
108,085 
90% 
$29,103  
4.94 
23% 
40% 
49% 
82% 
5% 
1% 
11% 
1% 
14% 
12% 

Attends 
Outside 
Charter 
 
N 
14,560 
Economically Disadvantaged 
92% 
Avg. Census Tract Median Inc. 
$29,801  
Drive Min. to Assigned Sch. 
4.58 
Avg. Pct. No Car in Tract 
20% 
Transportation Eligible 
47% 
Female 
50% 
Black 
79% 
White 
12% 
Hispanic 
3% 
Asian 
4% 
Other Race 
2% 
Student with Disability  
11% 
English Learner 
13% 
Note. DPSCD is Detroit Public Schools Community District. TPS is an abbreviation for traditional public school. A 
student is considered Economically Disadvantaged if he or she receives free or reduced lunch, his or her family 
receives food (SNAP) or cash (TANF) assistance, or is in foster care, is homeless or migrant. Drives times from the 
population weighted centroid of the student’s home census block to their assigned school were calculated using 
Google Distance Matrix API assuming usual traffic at 8am on a weekday. Students are considered transportation 
eligible if they are in grades K-8 and live more than 0.75 miles walking distance from their assigned school. 
 
Academic Quality Measures 

Attends 
Detroit 
Charter 
34,029 
93% 
$28,592  
5.16 
24% 
50% 
50% 
82% 
5% 
1% 
12% 
1% 
12% 
12% 

TPS 
9,029 
87% 
$30,291  
5.88 
21% 
28% 
49% 
84% 
6% 
2% 
5% 
3% 
9% 
5% 

The notion of “school quality” is itself a fairly subjective construct, and whether and to 

what extent schools are held accountable by federal, state, or local jurisdictions for some metric 

of performance is an enduring debate among education policymakers. This is a particularly 

important consideration in a study of school choice, which itself is a policy alternative 

predicated—as I describe above—at least in part on the notion that different families value 

different aspects of a school’s contribution to their children’s success. On the other hand, that 

school ratings on different dimensions are a relevant part of the current policy environment, and 

school characteristics—regardless of whether schools are in control of them—can and do 

constitute an important part of parental decision making, especially in high-choice environments 

(Lincove, Cowen, & Imbrogno 2018). 

 

82 

For these reasons, rather than choose a single metric of academic quality, I focus on three 

different metrics that capture different aspects of a school’s academic performance. The 

measures are the school’s state-generated academic accountability rating, the school’s 

contribution to student test scores as measures by the school’s valued added measure, and the 

school’s graduation rate during the 2016-17 school year. I choose to use the year prior to the 

sample because this is the information parents would have had to use to make their school choice 

decisions. In particular, I calculate academic quality as follows:  

1.  School’s rating on the Michigan School Index System. Index values range from 0 to 

100. This is a composite measure made up of six components: student growth on state 

assessments, student proficiency on state assessments, school quality or student 

success, graduation rate, English-learner progress, and assessment participation. This 

measure is highly correlated with the school’s average performance on state 

assessments. I use the school’s rating from MDE as a school quality measure because 

it is made publicly available on a “Parent Dashboard” maintained by the state. I refer 

to this as a school’s accountability rating or ranking because the index is the 

accountability system for Michigan schools under the Every Student Succeeds Act 

(ESSA). 

2.  School Value-added Measure. To measure school’s contribution to student learning, I 

calculate the school’s value-added measures, the school’s contribution to students’ 

math and English Language Arts (ELA) test scores, for the 2016-17 school year for 

students in grades K-8. Although value-added measures may be better measures of 

school quality, they may not be related to family’s preferences for high quality 

schools since Detroit families do not directly observe them. To construct school-level 

 

83 

value-added measures, I follow the procedure described by Koedel, Mihaly, and 

Rockoff (2015).13 I choose this procedure since it shrinks the estimated error variance 

which is preferred when using value-added estimates in secondary analysis since the 

errors could attenuate the results (Koedel, Mihaly, & Rockoff, 2015). I report the 

results of the models using the school’s value-added to math test scores. Results using 

the ELA measures are similar and available by request. 

3.  Graduation Rate. Value-added measures are unavailable for most high schools since 

students are only administered a state standardized exam once between grades 9 and 

12 in Michigan. Therefore, I use four year graduation rates as my secondary academic 

quality measure for students in grades 9-12. In addition, graduate rates have become a 

typical alternative to test scores as outcome measures in school choice evaluations 

(Angrist, Cohodes, Dynarski, Pathak, & Walters, 2016; Wolf, Kisida, Gutmann, 

Puma, Eissa, & Rizzo, 2013).  

 
 

 

13 To calculate a school’s value-added measures for the 2015-16, 2016-17, and 2017-18 school years, I use student 
test scores from the M-STEP for the 2014-15 through the 2017-18 school years. I first estimate the following 

equation: !"#$%&'!"=)#+)$&+#ℎ!"%$+)&(&+#ℎ!"%$)&+)'(&+#ℎ!"%$)'+)('/+!"%$+))('/+"%$)&+
)*('/+!"%$)'+0+,1+2-,3+4!" where !"#$%&'!" is either student i’s math or ELA test score at time t year 
standardized within subject, grade, and year at the state level. &+#ℎ!"%$ and '/+!"%$ are student i’s math and ELA 
test scores respectively for the prior school year. 0+, is a vector of student characteristics including race, gender, and 
economically disadvantaged, disability, and English Learner statuses. 2-, is a school by year fixed effect. 35 is 
using the following equation: 6."/07 =+."6."7+(1−+.")6̅ where 6."/07  is the school by year shrunken value-added 
estimate, 6."7 is the unshrunken, recovered estimate, 6̅ is the school’s average valued-added over time, and +."=
12!12!34"#5  . ;<& is the estimated variance of the school’s valued added measures, and =."7  estimated error variance for the 
value-added for school s in year t (Koedel, Mihaly, & Rockoff, 2015). I use 6."/07,	the shrunken value added estimate 

recovered as the school value-added measure for year t. Next, I shrink these measures towards a Bayesian prior 

for school s for the 2016-17 school year, in my models. Since I use school by year valued added measures, this 
allows a school’s effectiveness to vary by year. Therefore, a school’s value added to a student’s test score in 2015-
16 should not be reflected in the school’s 2016-17 value added estimate.  
 

 

84 

School Characteristics  

To examine other preferences specifically in the Detroit area where this study is focused, 

I use information on school offerings found in the 2018-19 Detroit Parents’ Guide to Schools 

(DPGS), one of the first sources of centralized information on school in Detroit that includes 

facts on charter schools in addition to DPSCD schools. I use a dataset created from the DPGS 

that contains information on application requirements, school hours, transportation, before and 

after school care, uniforms, security, and top activities for all public schools located within 

Detroit. Additionally, prior research has shown that families have preferences for student 

demographics. To measure the concentration of students by demographic characteristics, I  

construct the percent of students who are female, Black, White, Asian, Hispanic, Other Race, and  

economically disadvantaged as well as the percent of students with disabilities and who are 

English learners at each school using student-level demographic data. 

Table 11 displays the average school characteristics for the student’s assigned school by 

sector of school attended. Students choosing to leave the city have default options with a higher 

percentage of Black students on average. Students choosing to attend other DPSCD schools and 

schools outside of Detroit have assigned schools with lower than average accountability ratings.  

 

 

 

85 

Full 

Attends 
Outside 

Table 11. Student Weighted Assigned School Characteristics by Sector of School the Student 
Attends 

Attends 
Assigned 
School 

Attends 
Outside 
Average Assigned School 
Charter 
Characteristics 
N 
14,560 
Avg. Pct. Female 
52% 
Avg. Pct. Black 
92% 
Avg. Pct. White  
2% 
Avg. Pct. Hispanic 
3% 
Avg. Pct. Asian 
2% 
Avg. Pct. Other Race 
1% 
Avg. Pct. Economically Disadvantaged 
87% 
Avg. Pct. English Learner 
5% 
Avg. Pct. Students with Disabilities 
17% 
Avg. Total Enrollment 
557 
Avg. Accountability Rating 
27.93 
Avg. Math Value-Added  
-0.14 
Avg. Graduation Rate 
73% 
Pct. with After School Care  
16% 
Pct. Have Sport as Top Activity 
37% 
Note. Assigned school characteristics are weighted by the number of students in the location and/or sector. DPSCD 
is Detroit Public Schools Community District. TPS is an abbreviation for traditional public school. A student is 
considered Economically Disadvantaged if he or she receives free or reduced lunch, his or her family receives food 
(SNAP) or cash (TANF) assistance, or is in foster care, is homeless or migrant. Average math value-added is the 
average for students in grades K-8 and average graduation rate is the average in grades 9-12.  
 

Attends 
Detroit 
Charter 
34,029 
53% 
86% 
2% 
10% 
1% 
1% 
85% 
9% 
18% 
646 
29.63 
-0.13 
70% 
21% 
42% 

Sample 
108,085 
53% 
87% 
2% 
9% 
1% 
1% 
85% 
9% 
18% 
629 
28.79 
-0.13 
70% 
17% 
39% 

TPS 
9,029 
53% 
91% 
1% 
6% 
1% 
1% 
84% 
5% 
20% 
588 
27.75 
-0.13 
67% 
11% 
29% 

22,838 
52% 
82% 
3% 
13% 
1% 
1% 
85% 
12% 
17% 
695 
29.80 
-0.13 
74% 
17% 
43% 

Attends 
Other 
DPSCD 
School 
27,629 
53% 
87% 
2% 
10% 
0% 
1% 
85% 
9% 
19% 
606 
27.72 
-0.14 
69% 
16% 
37% 

Methods 

RQ 1: What are the roles of poverty, access to transportation, and student’s assigned 

school in use of school choice? 

To examine the differences in school choice take-up among disadvantaged students, I 

estimate a multinomial logistic regression since my dependent variable is a categorical variable. 

This model is described in Equation 2:  

ln(!6(#)
!6(%))=&%+&&()*+,-.'+&(log	(234-5+-+)*23)')+6789:;<=*+>+
&,?)@.B*C5D')+&-EF.(G-H-IG3'.+	J*K+L/+M0+N'  
Where ?'(2) is the probability that student i attends a DPSCD school other than his or her 

   (2) 

  

 

 

assigned school (an Other DPSCD school, a Detroit Charter school, an Outside TPS school or 

 

86 

indicator that equals one when student i is considered economically disadvantaged. 

the percent of residents who do not own a car in student i’s resident census tract c respectively. 

Outside Charter school) relative to attending his or her assigned school, ?'(0). My variables of 
interest measure the student’s exposure to poverty and access to transportation. ()*+,-.' is an 
log	(234-5+-+)*23)') and ?)@.B*C5D') are the natural logarithm of the median income of and 
6789:;<=*+	is a vector of indicators for having a drive time between their residence and their 
fifteen minutes. Finally, EF.(G-H-IG3'. is a binary variable than equals one if the student is 
J* is a vector of student characteristics including gender, race/ethnicity, and student with 
disability (SWD), and English Learner (EL) statuses. L/ and M0 are grade and assigned school 

assigned school either between five and ten minutes, ten to fifteen minutes, or greater than 

eligible to ride a school bus to his or her assigned school.  

fixed effects respectively. Including assigned school fixed effects allows me to estimate 

associations between choice use and student characteristics using comparisons between students 

with the same default option and a similar choice set. However, I also estimate the model 

represented in Equation 2 using zip code fixed effects, P1, instead of assigned school fixed effects 

as another measure of residential location for a specification check. Results of this model are 

similar and available by request. Furthermore, I use zip code fixed effects in the models that 

estimate the relationships between sector attended and assigned school characteristics instead of 

assigned school fixed effects since they would be perfectly collinear with the assigned school 

characteristics. This model is represented by Equation 3: 

ln(!6(#)
!6(%))=&%+&&()*+,-.'+&(log	(234-5+-+)*23)')+6789:;<=*+>+
&,?)@.B*C5D')+&-EF.(G-H-IG3'.+&2R)5432-)SF5G-@T.+J*K+U+V+L/+P1+N'      (3) 

 

87 

Where R)5432-)SF5G-@T.	is either accountability rating or the school’s 2016-17 math value-

added measure for elementary and middle school students, and graduation rate for high school 

students. I estimate the relationships between choice use and academic quality separately for the 

accountability rating and the other measures since accountability ratings include an achievement 

growth measure thus, accountability rating and value-added are highly correlated. U+ is a vector 

of characteristics of assigned school j which includes the percentages of female, White, Asian, 

Hispanic, Other Race, economically disadvantaged, and English Learner students, the natural 

logarithm of total enrollment, the percent of students with disabilities, and indicators for whether 

or not the school offers after school care and has a sport as its top activity. 

RQ2: Do Detroit students attend the more effective schools and/or the schools with the 

highest levels of achievement within their choice sets?  

To examine preferences for school sector, proximity from home, academic quality and 

student demographics, I estimate the relationship between the school characteristics outlined 

above and the probability that a student attends each school available to them. I accomplish this 

using a conditional logit model under the framework for alternative specific discrete choice 

models created by McFadden (1974), which stresses individuals making decisions between a 

competing set of discrete choices and is also a common approach in the school and college 

choice literature (e.g., Carlson, Cowen & Fleming 2013; Harris & Larsen, 2019; Long, 2004). 

Thus, I must construct a choice set for each student in the sample. I begin with the sample of 

students created for my first research question and restrict it to students in kindergarten, 6th 

grade, and 9th grade, the grades that students typically choose new schools. This sample includes 

8,970 kindergarteners, 8,628 6th graders, and 8,719 9th graders.  

 

 

 

88 

Table 12. Sector of Attendance by Grade 
 

 

Kindergarten  Grade 6  Grade 9 
8,719 
17% 
32% 
27% 
15% 
10% 

8,970 
24% 
24% 
33% 
4% 
14% 

8,628 
20% 
23% 
35% 
6% 
16% 

Attends Assigned School 
Attends Other DPSCD School 
Attends Detroit Charter 
Attends Outside TPS 
Attends Outside Charter 
Note. DPSCD is Detroit Public Schools Community District. TPS is an abbreviation for traditional public school. 
 

Table 12 displays the percent of Detroit students attending each sector by grade. 

Regardless of the grade, less than a quarter of students attend their assigned school, 

approximately forty percent attend a charter school, and about one in five students leave the city 

to attend school. In kindergarten, a higher percentage of students attend their assigned school 

compared to the other grades while more 9th graders attend outside TPS schools and other 

DPSCD schools.  

In Table 13, I report the characteristics of the schools in Detroit students’ choice sets by 

sector separately for each analysis grade. A student’s choice set is constructed by creating all 

pairwise combinations of students and the sample of schools offering his or her grade. For each 

student, the sample of schools in his or her choice set includes all traditional public schools and 

charter schools in the three intermediate school districts surrounding Detroit, Wayne Regional 

Education Service Authority, Oakland Public Schools, and Macomb Intermediate School District 

that offer general education, and serve at least one Detroit student, making it plausibly accessible 

to Detroit students. Detroit students attend over 600 different schools located within the 

boundaries of 69 different traditional public school districts in the tri-county area. There are 361, 

278, and 169 schools within the choice sets of kindergarten, 6th grade and 9th grade students 

respectively. The average demographic characteristics for schools located within Detroit reflect 

the population of students living in Detroit. Across all grades, schools located outside of the city, 

 

89 

55 
51% 
94% 
92% 
443 
47.23 
-0.02 
47% 
62% 
64% 
47% 
89% 
 

54 
51% 
92% 
93% 
464 
44.60 
-0.02 
35% 
48% 
69% 
48% 
81% 
 

172 
52% 
40% 
56% 
425 
70.04 
0.01 
N/A 
N/A 
N/A 
N/A 
N/A 
 

105 
52% 
38% 
53% 
652 
62.21 
0.01 
N/A 
N/A 
N/A 
N/A 
N/A 
 

DPSCD  Detroit Charter  Outside TPS  Outside Charter 
63 
50% 
64% 
73% 
553 
58.40 
-0.02 
N/A 
N/A 
N/A 
N/A 
N/A 
 

Table 13. School Characteristics by Sector and Grade 
Kindergarten 
Number of Schools 
Avg. Pct. Female 
Avg. Pct. Underrepresented Minority 
Avg. Pct. Economically Disadvantaged 
Avg. Enrollment 
Avg. Accountability Rating 
Avg. Math Value-Added 
Pct. with Before School Care  
Pct. with After School Care  
Pct. Sport Top Activity 
Pct. Do Not Offer Transportation 
Pct. Uniform Required 
Grade 6 
Number of Schools 
Avg. Pct. Female 
Avg. Pct. Underrepresented Minority 
Avg. Pct. Economically Disadvantaged 
Avg. Enrollment 
Avg. Accountability Rating 
Avg. Math Value-Added 
Pct. with Before School Care  
Pct. with After School Care  
Pct. Sport Top Activity 
Pct. Do Not Offer Transportation 
Pct. Uniform Required 
Grade 9 
Number of Schools 
Avg. Pct. Female 
Avg. Pct. Underrepresented Minority 
Avg. Pct. Economically Disadvantaged 
Avg. Enrollment 
Avg. Accountability Rating 
Avg. Graduation Rate 
Pct. with Before School Care  
Pct. with After School Care  
Pct. Sport Top Activity 
Pct. Do Not Offer Transportation 
Pct. Uniform Required 
Note. DPSCD is Detroit Public Schools Community District. TPS is an abbreviation for traditional public school. 
Information concerning Before School Care, After School Care, Transportation, and Uniform Requirements is only 
available for schools located within the city of Detroit since it comes from the Detroit Parents’ Guide to Schools. 15 
new schools are not included in the average school demographics and school quality ratings since they do not have 
them for the 2016-17 school year. In addition, 27 elementary and middle schools do not have math value-added 
measures since they do not have test score in both 2015-16 and 2016-17. 9 schools that were open in 2016-17 do not 
have a graduation rate.

56 
50% 
64% 
72% 
563 
57.14 
-0.03 
N/A 
N/A 
N/A 
N/A 
N/A 
 

24 
49% 
58% 
73% 
562 
51.44 
78% 
N/A 
N/A 
N/A 
N/A 
N/A 

71 
51% 
96% 
86% 
494 
32.78 
-0.11 
3% 
24% 
39% 
1% 
97% 
 

63 
52% 
97% 
85% 
500 
31.35 
-0.12 
2% 
21% 
41% 
0% 
97% 
 

28 
52% 
96% 
80% 
635 
36.99 
65% 
0% 
0% 
21% 
0% 
93% 

27 
50% 
89% 
91% 
445 
32.33 
66% 
11% 
15% 
48% 
56% 
56% 

90 
51% 
41% 
50% 
1076 
65.59 
83% 
N/A 
N/A 
N/A 
N/A 
N/A 

 

90 

particularly outside TPS schools, have higher accountability ratings. However, fewer Detroit 

students attend these schools. Interestingly, Detroit charter schools have higher accountability 

ratings and value-added measures than DPSCD schools on average in kindergarten and 6th grade, 

but in 9th grade, DPSCD schools have higher average accountability ratings and similar 

graduation rates than Detroit charter schools. This change in academic quality may explain the 

differences in attendance by sector within Detroit between 9th grade and the earlier grades seen in 

Table 12 if families have preferences for academic achievement. 

To examine the attributes of the schools students attend, I estimate a discrete choice 

model represented by Equations 4 and 5 separately for students in kindergarten, 6th, and 9th 

grades: 

Pr(Y'3=Z)= 47689
∑ 47689
8

 

 

 

 

 

(4) 

 

(5) 

Where: 

indicating the sector and location of the school with the reference group being student i’s 

['3&=!"#$%&!'+6789:;<=*6>+U6V+&&SF5G-@T3+\'+N'3 

in which Y'3 equals one for student i if they attend school k. ['3 is the predicted utility student i 
receives from attending school k based on school k’s characteristics. !"#$%&! is a vector of 
binary variables, ]@ℎ3D,?_C,3,,3@D*-@Cℎ5D@3D3,]F@.-43Cℎ5D@3D3, and ]F@.-43`?_3 
assigned school. ]@ℎ3D,?_C,3, equals one if school k is a DPSCD school but not the student’s 
assigned school. Similarly, ,3@D*-@Cℎ5D@3D3,	]F@.-43Cℎ5D@3D3, and ]F@.-43`?_3 equal one 
if the school is in the sector and location delineated by that variable. 6789:;<=*6 is a vector of 
greater than 40 minutes. Similar to model represented in Equation 2, SF5G-@T3 is one of the 

the student’s choice set either between 10 and 20 minutes, 20 to 30 minutes, 20 to 40 minutes or 

indicators for having a drive time (in minutes) from student i’s home census block to school k in 

 

91 

following academic quality measures: accountability rating, school-level 2016-17 math value-

added in kindergarten and 6th grade, and in 9th grade, graduation rate. Again, I estimate the 

relationships for the quality measures separately due to the high correlation between 

accountability rating and value-added. U6 is a vector of characteristics of school k which includes 
include student fixed effects, \', which accounts for all of the student i’s personal characteristics 

percent of female students, underrepresented minority students (non-White, non-Asian), and of 

economically disadvantaged students, and the natural logarithm of total enrollment.14 Also, I 

allowing me to make comparisons between schools within student i’s choice set. 

Results 

RQ 1: What are the roles of poverty, access to transportation, and student’s assigned 

school in use of school choice? 

Table 14 contains the estimated marginal effects of the models represented by Equations 

2 and 3. Columns 1 through 5 present the results of the models represented by Equation 2. The 

results in Column 1 show that students living in census tracts with high rates of residents without 

a car and lower median incomes are more likely to attend their assigned school holding all else 

constant. Additionally, students who live farther from their assigned school and those who are 

eligible for the school bus to it have a lower probability of attending their assigned school. Since 

bus eligibility is a function of walking distance to the assigned school, this finding may be driven 

by distance rather than the opportunity to use school provided transportation. As for other 

 

14 @: also includes two missing data indicators. A'BC$ℎ%%/; equals one if the school is in its first year of operation 

in 2017-18 meaning that it would not have any school characteristics or quality measures from the prior school year. 
15 schools in the sample are new schools. Additionally, 27 elementary and middle schools do not have math value-
added measures, and 9 high schools do not have graduation rates available for the 2016-17 school year. For schools 
with missing data, I recode the missing variable for this school to 0 and include indicators for missing value-added 
and graduation rate data. I also drop one school from the choice sets and the 8 students that attend it because it is 
missing the accountability rating. 

 

92 

sectors, students who have low rates of car ownership in their resident census tract are less likely 

to attend schools located outside of Detroit city limits. Furthermore, economically disadvantaged 

students are more likely to attend charter schools but less likely to attend other DPSCD schools. 

Overall, these results show that exposure to poverty, access to private transportation, and 

distance to assigned school play a role in whether or not a student uses school choice, especially 

to attend schools located outside of the city.  

Columns 6 through 15 of Table 14 contain the estimated marginal effects of the model 

represented by Equation 3 which includes assigned school characteristics as explanatory 

variables. The probability that a student attends his or her assigned school increases if it has a 

smaller concentration of economically disadvantaged students, a higher accountability rating, 

and does not offer after school care. However, there is no significant relationship between 

assigned school’s value-added and choice use. Thus, families may value student demographics 

and school performance levels, a finding consistent with prior literature. While this may suggest 

that Detroit families have preferences for these school attributes, it is possible that they are 

correlated with other unobserved features that families use to determine which school to send 

their child to.  

In contrast, the estimated marginal effects for many of the assigned school characteristics 

for attending another DPSCD school have the opposite signs than the predicted effects for 

attending the assigned school. These findings may imply that families substitute other DPSCD 

schools for their assigned school when the assigned school has fewer desirable characteristics. 

Furthermore, the estimated marginal effects of assigned school characteristics for attending 

schools outside of Detroit are small and statistically insignificant in most cases. This suggests 

that families that choose schools outside of Detroit may not even consider their assigned school 

 

93 

Table 14. Estimated Marginal Effects of Student and Assigned School Characteristics on School Type Attended 

 

 
 

Econ. Dis. 
 
Log. Med. Income 
 
5-10 Min. Drive Time 
 
10-15 Min. Drive Time 
 
15-20 Min. Drive Time 
 
Pct. No Car Ownership 
 
Transportation Eligible 
 
Sch. Pct. Econ. Dis. 
 
Accountability Rating 
 
Math Value-added 
 
Graduation Rate 
 
Offers Afterschool Care 
 
Sport Top Activity 
 
Assigned Sch. Fixed Effects 
Zip Code Fixed Effects 
Grade Fixed Effects 
Observations 

(1) 

Assigned 
School 

(2) 

Other 
DPSCD 

(3) 

Detroit 
Charter 

(4) 

Outside 

TPS 

(5) 

Outside 
Charter 

(6) 

Assigned 
School 

(7) 

Other 
DPSCD 

(8) 

Detroit 
Charter 

(9) 

Outside 

TPS 

(10) 

Outside 
Charter 

(11) 

Assigned 
School 

(12) 
Other 
DPSCD 

  

-0.010 
(0.012) 
-0.037* 
(0.017) 
-0.083* 
(0.008) 
-0.180* 
(0.016) 
-0.235 
(0.022) 
0.001* 
(0.000) 
-0.091* 
(0.009) 

 
 

 
 
 
 
 
 
 

 
X 

X 

 

 

 

  

-0.098* 
(0.006) 
0.036 
(0.020) 
0.035* 
(0.006) 
0.093* 
(0.013) 
0.134* 
(0.013) 
0.001* 
(0.000) 
0.028* 
(0.008) 

  

0.104* 
(0.016) 
0.039* 
(0.018) 
0.037* 
(0.009) 
0.044* 
(0.016) 
0.240* 
(0.096) 
0.001* 
(0.000) 
0.044* 
(0.010) 

  

-0.010 
(0.006) 
-0.004 
(0.006) 
0.000 
(0.003) 
0.019* 
(0.007) 
0.030 
(0.028) 
-0.001* 
(0.000) 
0.010* 
(0.005) 

  

0.014* 
(0.007) 
-0.034 
(0.018) 
0.011 
(0.007) 
0.024 
(0.019) 
-0.168* 
(0.058) 
-0.003* 
(0.001) 
0.009 
(0.006) 

 
 
 
 
 
 
 
 
 
 
 
 
X 
 
X 

 
 
 
 
 
 
 
 
 
 
 
 
X 
 
X 

 
 
 
 
 
 
 
 
 
 
 
 
X 
 
X 

 
 
 
 
 
 
 
 
 
 
 
 
X 
 
X 

  

-0.012 
(0.013) 
-0.045* 
(0.016) 
-0.083* 
(0.009) 
-0.152* 
(0.019) 
-0.244* 
(0.025) 
0.001* 
(0.000) 
-0.092* 
(0.008) 
-0.287* 
(0.146) 
0.001* 
(0.001) 

 
 
 
 

-0.051* 
(0.016) 
0.024 
(0.015) 

 
X 
X 

  

-0.098* 
(0.006) 
0.039* 
(0.019) 
0.033* 
(0.008) 
0.054* 
(0.016) 
0.151* 
(0.023) 
0.001 
(0.001) 
0.025* 
(0.008) 
0.367 
(0.224) 
-0.003* 
(0.001) 

 
 
 
 

0.049* 
(0.020) 
-0.016 
(0.017) 

 
X 
X 

  

0.105* 
(0.016) 
0.030 
(0.015) 
0.049* 
(0.010) 
0.047* 
(0.015) 
0.184* 
(0.054) 
0.001 
(0.001) 
0.044* 
(0.009) 
-0.115 
(0.159) 
0.001 
(0.000) 

 
 
 
 

0.003 
(0.017) 
-0.024 
(0.013) 

 
X 
X 

  

-0.009 
(0.006) 
-0.009 
(0.006) 
-0.002 
(0.003) 
0.007 
(0.008) 
0.031* 
(0.015) 
-0.000* 
(0.000) 
0.016* 
(0.004) 
-0.008 
(0.052) 
0.000 
(0.000) 

 
 
 
 

-0.001 
(0.007) 
-0.004 
(0.005) 

 
X 
X 

  

0.014* 
(0.007) 
-0.016 
(0.012) 
0.004 
(0.007) 
0.044* 
(0.011) 
-0.122* 
(0.035) 
-0.002* 
(0.000) 
0.008 
(0.005) 
0.044 
(0.089) 
0.000 
(0.000) 

 
 
 
 

-0.000 
(0.008) 
0.020* 
(0.007) 

 
X 
X 

  

-0.012 
(0.013) 
-0.044* 
(0.016) 
-0.082* 
(0.009) 
-0.155* 
(0.019) 
-0.254* 
(0.029) 
0.001** 
(0.000) 
-0.092* 
(0.008) 
-0.254 
(0.145) 

 
 

0.026 
(0.053) 
0.002 
(0.002) 
-0.046* 
(0.017) 
0.022 
(0.015) 

 
X 
X 

  

-0.099* 
(0.006) 
0.038* 
(0.019) 
0.031* 
(0.007) 
0.057* 
(0.017) 
0.174* 
(0.030) 
0.001 
(0.001) 
0.024* 
(0.008) 
0.333 
(0.216) 

 
 

-0.041 
(0.046) 
-0.002* 
(0.001) 
0.041 
(0.021) 
-0.013 
(0.017) 

 
X 
X 

(13) 

Detroit 
Charter 

  

0.105* 
(0.016) 
0.030* 
(0.015) 
0.049* 
(0.010) 
0.049* 
(0.015) 
0.182* 
(0.057) 
0.001 
(0.001) 
0.044* 
(0.009) 
-0.111 
(0.157) 

 
 

0.016 
(0.046) 
-0.000 
(0.00) 
0.006 
(0.019) 
-0.024 
(0.013) 

 
X 
X 

108,085 

(14) 

Outside 

TPS 

  

-0.009 
(0.006) 
-0.009 
(0.006) 
-0.002 
(0.003) 
0.008 
(0.008) 
0.033* 
(0.015) 
-0.000* 
(0.000) 
0.015* 
(0.004) 
-0.018 
(0.054) 

 
 

0.016 
(0.015) 
-0.000 
(0.000) 
-0.003 
(0.008) 
-0.004 
(0.005) 

 
X 
X 

(15) 

Outside 
Charter 

  

0.014* 
(0.007) 
-0.015 
(0.012) 
0.004 
(0.007) 
0.040* 
(0.011) 
-0.136* 
(0.036) 
-0.002* 
(0.000) 
0.008 
(0.005) 
0.050 
(0.087) 

 
 

-0.017 
(0.022) 
0.001* 
(0.000) 
0.002 
(0.008) 
0.019* 
(0.007) 

 
X 
X 

108,085 

108,085 

108,085 

108,085 

108,085 
Note. Standard errors in parentheses. * p<0.05. Standard errors are clustered at the level of the geographic fixed effect in the model. Models include student race, 
gender, English Learner, and disability statuses and school’s percent female, White, Asian, Hispanic, Other Race, English Learner, students with disabilities and 
the natural logarithm of enrollment as covariates. Drive times from the population weighted centroid of the student’s home census block to their assigned school 
were calculated using Google Distance Matrix API assuming usual traffic at 8am on a weekday. DPSCD is Detroit Public Schools Community District. TPS is an 
abbreviation for traditional public school. A student is considered Econ. Dis. (Economically Disadvantaged) if he or she receives free or reduced lunch, his or her 
family receives food (SNAP) or cash (TANF) assistance, or is in foster care, is homeless or migrant. Accountability Rating is the school’s score on Michigan’s 
School Index System.  

108,085 

108,085 

108,085 

108,085 

108,085 

108,085 

108,085 

108,085 

108,085 

 

94 

as an option. Taken together, these results provide some suggestive evidence that different 

Detroit families have different and restricted choice sets. Some families may only consider 

DPSCD schools or schools within Detroit whereas others only look outside of city limits for a 

school for their child. 

RQ2: Do Detroit students attend the more effective schools and/or the schools with the 

highest levels of achievement within their choice sets?  

Table 15 presents the results of the discrete choice model described in Equations 4 and 5 

for all students in the sample. As previously described, the models were restricted by grade to 

ensure than all students in the model had the same choice set. Estimated coefficients are reported 

as log odds coefficients. All the estimated coefficients for the sector indicators are negative and 

significant in the main models. This indicates that on average, students are more likely to attend 

their assigned schools than charter schools, schools outside of Detroit, and other DPSCD 

schools, even when distance is held constant. The coefficient estimates for all commute time 

indicators are negative and significant across all specifications indicating a negative relationship 

between increased distance and attendance. Thus, families are less likely to choose schools 

farther from home. Also, I find that students have a higher probability of attending schools with 

higher rates of economically disadvantaged students in the majority of the models. As for 

academic quality, kindergarten and 6th grade students have a higher probability of attending 

schools with higher accountability ratings and higher contributions to student achievement as 

measured by math value-added. As for 9th grade, there is no significant relationship between 

accountability rating or graduation rate.  

 

 

95 

Table 15. Conditional Logit Predictions of School Attendance for School Characteristics by 
Grade 
 

K 

6 

9 

 

(4) 
-2.199* 
(0.036) 
-1.647* 
(0.036) 
-2.195* 
(0.070) 
-1.290* 
(0.041) 
-2.265* 
(0.026) 
-4.039* 
(0.057) 
-5.419* 
(0.168) 
-5.605* 
(0.356) 
0.167 
(0.101) 
 
 
0.164* 
(0.073) 
 
 
2,398,584 

(1) 
-2.445* 
(0.033) 
-2.022* 
(0.035) 
-3.200* 
(0.071) 
-1.829* 
(0.040) 
-2.442* 
(0.027) 
-4.052* 
(0.056) 
-6.026* 
(0.238) 
-5.622* 
(0.409) 
0.420* 
(0.112) 
0.009* 
(0.001) 
 
 
 
 
3,238,170 

(2) 
-2.454* 
(0.033) 
-1.936* 
(0.033) 
-3.223* 
(0.068) 
-1.790* 
(0.040) 
-2.437* 
(0.027) 
-4.055* 
(0.056) 
-6.013* 
(0.238) 
-5.551* 
(0.409) 
0.118 
(0.113) 
 
 
0.401* 
(0.067) 
 
 
3,238,170 

(3) 
-2.220* 
(0.036) 
-1.755* 
(0.038) 
-2.260* 
(0.070) 
-1.375* 
(0.042) 
-2.273* 
(0.026) 
-4.051* 
(0.057) 
-5.429* 
(0.168) 
-5.609* 
(0.355) 
0.395* 
(0.104) 
0.007* 
(0.001) 
 
 
 
 
2,398,584 

(5) 
-1.210* 
(0.039) 
-1.235* 
(0.042) 
-1.421* 
(0.054) 
-0.986* 
(0.054) 
-1.596* 
(0.027) 
-3.029* 
(0.047) 
-4.782* 
(0.155) 
-5.404* 
(0.502) 
0.974* 
(0.122) 
-0.001 
(0.001) 
 
 
 
 
1,473,511 

(6) 
-1.184* 
(0.038) 
-1.125* 
(0.041) 
-1.425* 
(0.053) 
-0.961* 
(0.051) 
-1.609* 
(0.027) 
-3.040* 
(0.047) 
-4.776* 
(0.155) 
-5.425* 
(0.503) 
0.983* 
(0.108) 
 
 
 
 
-0.001 
(0.001) 
1,473,511 

Other DPSCD School 
 
Detroit Charter 
 
Outside TPS 
 
Outside Charter 
 
10-20 Min. Drive Time 
 
20-30 Min. Drive Time 
 
30-40 Min. Drive Time 
 
40-60 Min. Drive Time 
 
Pct. Econ. Dis. 
 
Accountability Rating 
 
Math Value-Added 
 
Graduation Rate 
 
Observations 
Note. Heteroskedasticity robust standard errors in parentheses. * p<0.05. Estimated relationships are reported as log 
odds coefficients. Models include school percent female enrollment, percent underrepresented minorities, natural log 
of enrollment, indicators for being a new school in 2017-18, having missing math value-added or graduation rate 
data. DPSCD is Detroit Public Schools Community District. TPS is an abbreviation for traditional public school. 
Drives times from the population weighted centroid of the student’s home census block to each school in the 
student’s choice set were calculated using Google Distance Matrix API assuming usual traffic at 8am on a weekday. 
A student is considered Econ. Dis. (Economically Disadvantaged) if he or she receives free or reduced lunch, his or 
her family receives food (SNAP) or cash (TANF) assistance, or is in foster care, is homeless or migrant. 
 
Heterogeneity by Location and Sector 

Although there exist positive relationships between my measures of academic quality for 

Michigan schools and the probability of attending a school for kindergarten and 6th grade 

students, these relationships may differ if the student’s choice set is geographically restricted. 

 

96 

Indeed, my findings from the first research question suggest that families’ choice sets may be 

restricted by location and/or sector before considering individual school qualities especially for 

students with little access to transportation or impoverished students. Specifically, the results 

indicate that some families are choosing between DPSCD schools, within city limits, or solely 

outside of the city. To investigate this further, I examine mobility between sectors over time for 

Detroit students.  

Table 16 presents the percent of students who have ever attended another sector between 

2012-13 and 2017-18 by sector attended during the year of analysis. Only nine percent of 

students attending a school inside of Detroit during the 2017-18 school year have ever attended a 

public school located outside of city limits while living in Detroit. More specifically, about two 

percent of students who did not attend an outside TPS school in the year of analysis ever 

attended one. In contrast, about four out of ten students attending an outside Detroit school have 

attended a Detroit school in the past. Additionally, about a quarter of DPSCD students have 

attended a non-DPSCD school and a similar amount of non-DPSCD students have attended a 

DPSCD school. 

 

 

97 

Table 16. Percent of Students Who Have Ever Attended Another Sector by Sector and Location between 2012-13 and 2017-18 

Ever 
Assigned 
School 
100% 
29% 
16% 
16% 
11% 
13% 
43% 
61% 
15% 

Ever 
Other 
DPSCD 
School 
28% 
100% 
15% 
17% 
10% 
13% 
47% 
68% 
14% 

School in 2017-18 
Assigned School 
Other DPSCD School 
Detroit Charter 
Outside TPS 
Outside Charter 
Outside School 
Inside School 
DPSCD School 
Non-DPSCD School 
Note. Sample includes the 108,085 Detroit residents attending public school in 2017-18. Whether a student has ever attended a school in a sector is determined 
using student enrollment data from the 2012-13 school year through the 2017-18 school year. DPSCD is Detroit Public Schools Community District. TPS is an 
abbreviation for traditional public school.

N 
22,808 
27,659 
34,029 
9,029 
14,560 
23,589 
84,496 
50,467 
57,618 

Ever Attends 
Non-DPSCD 
24% 
29% 
100% 
100% 
100% 
100% 
56% 
27% 
100% 

Ever 
Outside 
Charter 
5% 
7% 
9% 
18% 
100% 
69% 
7% 
6% 
33% 

Ever 
Outside 
Detroit 
7% 
8% 
10% 
100% 
100% 
100% 
9% 
8% 
47% 

Ever 
Detroit 
Charter 
19% 
23% 
100% 
28% 
21% 
24% 
53% 
21% 
69% 

Ever 
Outside 
TPS 
2% 
2% 
2% 
100% 
2% 
40% 
2% 
2% 
17% 

Ever 
Inside 
Detroit 
100% 
100% 
100% 
48% 
34% 
39% 
100% 
100% 
75% 

Ever 
Attends 
DPSCD 
100% 
100% 
27% 
28% 
18% 
22% 
71% 
100% 
25% 

 

98 

Although examining the mobility of all students over time may be informative, this 

analysis may understate the amount of mobility that happens between sectors since it includes 

students who did not make active choices in some years. To more accurately explore the 

movement between sectors of students, I examine the mobility between sectors of students who 

changed schools between the 2016-17 and 2017-18 school years in Table 17. I consider the 

mobility patterns of structural movers, students who were in the last or terminal grade of the 

school they attended in 2016-17, and non-structural movers separately. Almost 90 percent of 

structural and non-structural movers who attended a Detroit school in 2017-18 also attended one 

the previous year. However, only a third of non-structural movers attending a school outside of 

the city during the year of analysis attended one the year before while three quarters of structural 

movers remained in an outside school between years. About two-thirds of structural and non-

structural movers attending a DPSCD school in 2017-18 were in a DPSCD school in 2016-17. 

Interestingly, 88 percent of structural movers attending a non-DPSCD school attended a non-

DPSCD school in both years.

 

99 

TPS 

Outside 

School in 2016-17 

Table 17. Percent of Students Attending a Different Sector Last Year by Sector and Mobility Type 
Panel A: Non-Structural Movers 
 
School in 2017-18 
Assigned School  
Other DPSCD School 
Detroit Charter 
Outside TPS 
Outside Charter 
Outside School 
Inside School 
DPSCD School 
Non-DPSCD School 
Panel B: Structural Movers 
 

Outside 
Charter  Outside 
11% 
10% 
16% 
34% 
34% 
34% 
13% 
11% 
23% 

Detroit 
Charter 
30% 
26% 
37% 
33% 
33% 
33% 
31% 
28% 
35% 

Assigned 
27% 
29% 
23% 
13% 
18% 
16% 
26% 
28% 
20% 

Other 
DPSCD 
32% 
35% 
24% 
20% 
16% 
17% 
30% 
34% 
21% 

4% 
3% 
4% 
17% 
5% 
10% 
4% 
4% 
7% 

7% 
7% 
12% 
17% 
28% 
24% 
9% 
7% 
17% 

Inside  DPSCD 
58% 
64% 
47% 
33% 
34% 
33% 
56% 
61% 
42% 

89% 
90% 
84% 
66% 
66% 
66% 
87% 
89% 
77% 

School in 2016-17 

Non-
DPSCD 

42% 
36% 
53% 
67% 
66% 
67% 
44% 
39% 
58% 

TPS 

Outside 

School in 2017-18 
Assigned School  
Other DPSCD School 
Detroit Charter 
Outside TPS 
Outside Charter 
Outside School 
Inside School 
DPSCD School 
Non-DPSCD School 
Note. Main sample includes Detroit residents attending public school in 2017-18 and in 2016-17. Structural Movers are defined as students who were in the 
terminal grade of their school in 2016-17 meaning that their school did not offer the grade, they would be in in 2017-18. Non-structural movers are all other 
students who switched schools between the two school years. DPSCD is Detroit Public Schools Community District. TPS is an abbreviation for traditional public 
school.  

Inside  DPSCD 
73% 
64% 
13% 
10% 
9% 
10% 
45% 
67% 
12% 

Assigned 
33% 
19% 
6% 
4% 
4% 
4% 
17% 
25% 
5% 

Outside 
Charter  Outside 
9% 
10% 
15% 
75% 
73% 
74% 
12% 
9% 
41% 

Detroit 
Charter 
18% 
26% 
72% 
14% 
19% 
17% 
43% 
23% 
47% 

Other 
DPSCD 
40% 
45% 
7% 
6% 
4% 
5% 
28% 
43% 
6% 

1% 
1% 
1% 
59% 
1% 
30% 
1% 
1% 
14% 

8% 
9% 
14% 
16% 
71% 
44% 
11% 
8% 
27% 

27% 
36% 
87% 
90% 
91% 
90% 
55% 
33% 
88% 

91% 
90% 
85% 
25% 
27% 
26% 
88% 
91% 
59% 

Non-
DPSCD 

 

100 

When the evidence from Tables 14, 16, and 17 is combined, it seems likely that some 

families have choice sets restricted to schools within city limits or within DPSCD. Furthermore, 

the differences in who attends outside schools and the lack of relationship between attending 

outside schools and assigned school characteristics suggest that those attending outside schools 

may choose to attend schools outside of the city before considering specific schools. In 

accordance with these findings, I estimate the model represented in Equations 4 and 5 separately 

by grade and for students attending: inside Detroit, outside Detroit, a DPSCD school, or a non-

DPSCD school. Thus, the schools in the students’ choice set are restricted to schools in that 

location/sector for each analysis. The results of these analysis are presented in Table 18. Panel A 

displays estimates for the Inside and DPSCD samples and Panel B shows the results for the 

outside and non-DPSCD samples. Estimates should be interpreted in comparison to other schools 

within that location or sector.  

After restricting the choice sets to schools located inside of Detroit and DPSCD, students 

still have a lower probability of attending schools farther from home. Additionally, students are 

less likely to attend Other DPSCD and Detroit charter schools compared to their assigned 

schools. In contrast with the full sample results, students attend schools with a lower percent of 

economically disadvantaged students when the sample is restricted to all schools inside Detroit. 

These contrasting results are explained by the strong positive relationship between the percent of 

economically disadvantaged students and attendance when the choice set is restricted to schools 

outside of Detroit as seen in Table 18 Panel B. Taken together, these results imply that students 

inside Detroit choose schools with fewer economically disadvantaged students but those who 

attend schools outside are more likely to attend the more impoverished schools within their 

choice set. 

 

101 

Table 18. Conditional Logit Predictions of School Attendance for School Characteristics by Grade and Location 
Panel A: Choice Sets Restricted to Schools Located Inside or Outside of DPSCD Borders 
 

DPSCD 

Inside 

 

 
Other DPSCD Sch. 
 
Detroit Charter 
 
10-20 Min. Drive  
 
20-30 Min. Drive  
 
30-40 Min. Drive  
 
Pct. Econ. Dis. 
 
Acct. Rating 
 
Math Value-Added 
 
Graduation Rate 
 
Before Care 

 
After Care 

 
No Transport. 

 
Uniform  

 
Sport Top Activity 

 
Observations 

K 

6 

9 

(1) 

-2.417* 
(0.034) 
-1.964* 
(0.047) 
-2.583* 
(0.032) 
-4.166* 
(0.086) 
-6.246* 
(1.000) 
-0.645* 
(0.188) 
0.008* 
(0.001) 

 
 
 
 
-0.060 
(0.044) 
0.077* 
(0.036) 
0.295* 
(0.040) 
0.399* 
(0.071) 
-0.188* 
(0.029) 
922,824 

(2) 

-2.418* 
(0.034) 
-1.910* 
(0.046) 
-2.576* 
(0.032) 
-4.163* 
(0.086) 
-6.239* 
(1.000) 
-0.866* 
(0.188) 

 
 
0.346* 
(0.071) 

 
 
0.050 
(0.044) 
0.046 
(0.035) 
0.256* 
(0.040) 
0.349* 
(0.071) 
-0.169* 
(0.030) 
922,824 

(3) 

-2.212* 
(0.037) 
-1.635* 
(0.047) 
-2.399* 
(0.031) 
-4.267* 
(0.098) 
-6.178* 
(1.001) 
-1.220* 
(0.179) 
0.007* 
(0.001) 

 
 
 
 
-0.004 
(0.047) 
-0.045 
(0.037) 
0.117* 
(0.040) 
0.354* 
(0.070) 
-0.067* 
(0.029) 
790,452 

(4) 

(5) 

-2.189* 
(0.037) 
-1.552* 
(0.047) 
-2.394* 
(0.031) 
-4.261* 
(0.098) 
-6.193* 
(1.001) 
-1.504* 
(0.179) 

 
 
0.098 
(0.079) 

 
 
0.059 
(0.046) 
-0.038 
(0.037) 
0.142* 
(0.041) 
0.340* 
(0.072) 
-0.062* 
(0.029) 
790,452 

-1.406* 
(0.043) 
-1.059* 
(0.061) 
-1.529* 
(0.032) 
-2.790* 
(0.067) 
-3.387* 
(0.585) 
-0.793* 
(0.208) 
0.002* 
(0.001) 

 
 
 
 
-1.090* 
(0.171) 
0.248 
(0.142) 
-0.064 
(0.049) 
-0.047 
(0.047) 
-0.117* 
(0.029) 
361,185 

(6) 

-1.341* 
(0.041) 
-0.746* 
(0.060) 
-1.545* 
(0.032) 
-2.828* 
(0.066) 
-3.415* 
(0.585) 
-1.459* 
(0.195) 

 
 
 
 
-0.003* 
(0.001) 
-0.846* 
(0.181) 
0.251 
(0.146) 
-0.260* 
(0.050) 
0.090 
(0.051) 
-0.171* 
(0.029) 
361,185 

 

 

6 

9 

(10) 
-2.213* 
(0.040) 

 
 
-2.482* 
(0.050) 
-4.131* 
(0.136) 
-5.073* 
(1.002) 
-0.344 
(0.331) 

 
 
0.206 
(0.154) 

 
 
0.419* 
(0.136) 
0.051 
(0.060) 

 
 
0.334* 
(0.169) 
-0.129* 
(0.042) 
236,628 

(11) 
-1.588* 
(0.049) 

 
 
-1.203* 
(0.046) 
-2.490* 
(0.089) 
-15.881* 
(0.098) 
1.719* 
(0.316) 
0.008* 
(0.001) 

 
 
 
 
 
 
 
 
 
 
-0.480* 
(0.104) 
-0.496* 
(0.043) 
119,084 

(12) 
-1.526* 
(0.048) 

 
 
-1.273* 
(0.046) 
-2.521* 
(0.090) 
-15.871* 
(0.101) 
0.282 
(0.297) 

 
 
 
 
-0.000 
(0.001) 

 
 
 
 
 
 
-0.084 
(0.111) 
-0.508* 
(0.042) 
119,084 

(9) 

-2.256* 
(0.041) 

 
 
-2.467* 
(0.050) 
-4.114* 
(0.137) 
-5.019* 
(1.002) 
-0.029 
(0.308) 
0.013* 
(0.002) 

 
 
 
 
-0.086 
(0.151) 
-0.005 
(0.060) 

 
 
0.352* 
(0.167) 
-0.151* 
(0.041) 
236,628 

K 

(7) 

(8) 

-2.486* 
(0.037) 

 
 
-2.541* 
(0.050) 
-3.946* 
(0.118) 
-17.292* 
(0.047) 
-1.059* 
(0.294) 
0.010* 
(0.001) 

 
 
 
 
-0.106 
(0.155) 
-0.115* 
(0.056) 
0.392 
(0.249) 
0.356* 
(0.155) 
-0.156* 
(0.037) 
308,850 

-2.453* 
(0.036) 

 
 
-2.544* 
(0.049) 
-3.956* 
(0.118) 
-18.295* 
(0.047) 
-0.961* 
(0.304) 

 
 
0.698* 
(0.125) 

 
 
0.347* 
(0.144) 
-0.081 
(0.054) 
0.176 
(0.250) 
0.394* 
(0.158) 
-0.153* 
(0.037) 
308,850 

102 

Table 18. (cont’d) 
Panel B: Choice Sets Restricted DPSCD Schools or No DPSCD Schools 
 

Outside 

 

K 

6 

9 

K 

No DPSCD 

6 

9 

(1) 

(2) 

(3) 

(4) 

(5) 

(6) 

(12) 
0.166* 
(0.057) 
0.335* 
(0.045) 
-1.925* 
(0.035) 
-3.270* 
(0.054) 
-4.821* 
(0.155) 
-5.280* 
(0.503) 
1.882* 
(0.138) 

 
 
0.592* 
(0.078) 
-1.803* 
(0.055) 
-3.416* 
(0.077) 
-4.767* 
(0.171) 
-4.579* 
(0.354) 
1.905* 
(0.176) 
-0.000 
(0.003) 

 
 
1.199* 
(0.069) 
-1.922* 
(0.058) 
-3.523* 
(0.081) 
-5.443* 
(0.246) 
-4.730* 
(0.406) 
2.333* 
(0.206) 
0.017* 
(0.003) 

 
 
0.058 
(0.069) 
-1.651* 
(0.063) 
-3.030* 
(0.073) 
-4.649* 
(0.163) 
-4.781* 
(0.504) 
2.059* 
(0.237) 
-0.022* 
(0.002) 

(8) 
-1.321* 
(0.072) 
0.067 
(0.041) 
-2.380* 
(0.032) 
-4.017* 
(0.063) 
-5.869* 
(0.238) 
-5.349* 
(0.408) 
0.611* 
(0.144) 

 
 
0.068 
(0.100) 

 
 
1.260* 
(0.072) 
-1.949* 
(0.059) 
-3.565* 
(0.080) 
-5.421* 
(0.246) 
-4.645* 
(0.405) 
2.073* 
(0.216) 

 
 
1.272* 
(0.258) 

 
 
0.582* 
(0.078) 
-1.803* 
(0.055) 
-3.413* 
(0.076) 
-4.755* 
(0.170) 
-4.540* 
(0.354) 
1.872* 
(0.161) 

 
 
-0.030 
(0.222) 

(9) 
-0.437* 
(0.075) 
0.269* 
(0.039) 
-2.169* 
(0.031) 
-3.934* 
(0.063) 
-5.256* 
(0.170) 
-5.349* 
(0.356) 
0.638* 
(0.123) 
0.001 
(0.001) 

(7) 
-1.146* 
(0.075) 
0.135* 
(0.041) 
-2.380* 
(0.032) 
-4.005* 
(0.064) 
-5.883* 
(0.238) 
-5.440* 
(0.408) 
0.911* 
(0.141) 
0.008* 
(0.001) 

 
Outside TPS 
 
Outside Charter 
 
10-20 Min. Drive  
 
20-30 Min. Drive  
 
30-40 Min. Drive  
 
40-60 Min. Drive  
 
Pct. Econ. Dis. 
 
Acct. Rating 
 
Math Value-Added 
 
Graduation Rate 
 
Observations 
Note. Heteroskedasticity robust standard errors in parentheses. *p<0.05. Estimated relationships are reported as log odds coefficients. Models include school 
percent female enrollment, percent underrepresented minorities, natural log of enrollment, indicators for being a new school in 2017-18, having missing math 
value-added or graduation rate data. School offering variables only included in models that only include students who attend schools within the boundaries of 
DPSCD. DPSCD is Detroit Public Schools Community District. TPS is an abbreviation for traditional public school. A student is considered Econ. Dis. 
(Economically Disadvantaged) if he or she receives free or reduced lunch, his or her family receives food (SNAP) or cash (TANF) assistance, or is in foster care, 
is homeless or migrant.

(10) 
-0.471* 
(0.075) 
0.248* 
(0.040) 
-2.166* 
(0.031) 
-3.923* 
(0.063) 
-5.224* 
(0.169) 
-5.285* 
(0.355) 
0.554* 
(0.120) 

 
 
-0.161 
(0.098) 

(11) 
0.180* 
(0.057) 
0.429* 
(0.045) 
-1.906* 
(0.035) 
-3.283* 
(0.054) 
-4.899* 
(0.155) 
-5.380* 
(0.503) 
1.322* 
(0.154) 
-0.008* 
(0.001) 

 
 
-0.021 
(0.067) 
-1.648* 
(0.063) 
-3.020* 
(0.072) 
-4.553* 
(0.163) 
-4.691* 
(0.502) 
3.063* 
(0.191) 

 
 
 
 
0.001 
(0.001) 
119,084 

 
 
 
 
0.001 
(0.001) 
629,706 

 
 
 
 
308,850 

 
 
 
 
236,628 

 
 
308,850 

 
 
 
 
119,084 

 
 
236,628 

 
 
 
 
1,339,800 

 
 
1.339,800 

 
 
 
 
1,047,480 

 
 
 
 
629,706 

 
 
1,047,480 

 

103 

By restricting the choice set to schools inside Detroit or just to DPSCD schools, I can 

also include school characteristics from the Detroit Parents’ Guide to Schools as seen in Table 

18 Panel A. Across grades, student have a lower probability of attending schools that list a sport 

as a top activity. In the earlier grades, students have a higher probability of attending schools that 

require uniforms and when all Detroit schools are included in the choice set, they are more likely 

to attend schools that do not offer transportation. Since these characteristics are likely correlated 

to other unobserved characteristics of schools that families used to determine where they send 

their child, I do not interpret these as preferences. 

As for measures of academic quality, students seem to have a higher probability of 

attending a school as its accountability rating increases when the choice set is restricted to 

schools inside of Detroit or DPSCD across all grades. The relationships are larger when the 

choice set is restricted to DPSCD schools, suggesting that there may be a larger preference for 

high accountability ratings after accounting for geographic and administrative barriers. 

Additionally, there is a positive relationship between school value-added and attendance for 

kindergarteners. The results for the choice sets for students who attend outside or non-DPSCD 

schools vary by grade. In kindergarten, there is a positive relationship between accountability 

rating, value-added, and attendance for outside Detroit and non-DPSCD choice sets. However, 

there is no significant relationship between either of the academic quality measures and 

attendance for 6th grade students in either of these choice sets. A negative relationship exists 

between accountability rating and attendance for 9th grade students in both outside and non-

DPSCD choice sets. In kindergarten, there is a positive relationship with school valued added 

and attendance. Therefore, these findings suggest that the possible preferences for academic 

 

104 

quality and student demographics may be constrained for students looking to attend school 

outside of Detroit or DPSCD especially in the later grades.  

Discussion 

 

In this paper, I show that students’ choice sets are likely geographically constrained, 

especially for impoverished students and students with little access to private transportation in 

Detroit, a choice-rich city and region without regulations that promote accessibility or 

transparency. While I am unable to directly test whether families do not prefer high quality 

schools since I do not have any record of parent’s set of preferred schools, I provide some strong 

suggestive evidence that families have some preference for higher performing schools when 

families are choosing between schools located within Detroit and especially within DPSCD. 

Additionally, I offer some indication that families have limited access to the relatively higher 

quality schools located outside of the city, constraining their ability to attend schools with their 

desired characteristics. This contributes to the parental preference literature by framing distance 

and administrative barriers as constraints to families’ preferences instead of as a preference itself 

and expanding choice sets to include inter-district choice options.  

The findings of this study provide a compelling empirical example of the idea that strong 

preferences for effective schools can exist throughout an education market and, yet students may 

also have differential access to high quality schools due to where they live and what their 

historical demographic and current economic circumstances are especially when there is little 

regulation on where schools locate and what enrollment rules they govern by exists. Thus, it is 

unlikely that parents and families can regulate the school supply through market transactions 

even when they have strong preferences for academic quality. This implies that choice-rich 

cities, regardless of whether they offer intra-district, charter school, or inter-district choice, may 

 

105 

need to provide oversight of the schooling market to ensure an adequate supply of effective 

schools. This oversight likely includes centralized planning of the quantity and location of 

schools and performance accountability as suggested by Harris (2017).  

Furthermore, choice-rich cities may need to ensure that effective schools are accessible to 

all students. Two policies that could promote or restrict access to effective schools are 

transportation policies and enrollment policies. Increasing school transportation has the potential 

to remove the burden from parents of transporting their children to school, making it easier to 

attend schools located farther from home. However, few states require to provide transportation 

to students attending charter schools or participating in inter-district choice (McShane & Shaw, 

2020). Furthermore, enrollment policies have the potential to promote or restrict access to 

effective schools. Multiple applications with varying deadlines likely restrict access to effective 

schools for the most vulnerable populations, possibly explaining why Detroit students have 

difficulty accessing the highest quality schools, especially in the later grades. A centralized 

enrollment application could increase access to effective schools through choice since causal 

evidence exists showing that they are effective in increasing enrollment in charter schools for 

disadvantaged students (Winters, 2015). To date, however, only a handful of cities have 

implemented these policies. Extensions of this work and others in this literature should consider 

in particular the role that enrollment rules play in promoting or restricting access to schools even 

in a system where choice is widely available. 

 

106 

REFERENCES

 

107 

REFERENCES 

 

Abdulkadiroğlu, A., Pathak, P., Schellenberg, J., & Walters, C. (2020). Do parents value school 

effectiveness?. American Economic Review, 110(5), 1502-39. 

 
Angrist, J., Cohodes, S., Dynarski, S., Pathak, P., & Walters, C. (2016). Stand and deliver: 

Effects of Boston’s charter high schools on college preparation, entry, and 
choice. Journal of Labor Economics, 34(2), 275-318. 

 
Bell, C. (2009). All choices created equal? The role of choice sets in the selection of 

schools. Peabody Journal of Education, 84(2), 191-208. 

 
Beuermann, D., Jackson, C., Navarro-Sola, L., & Pardo, F. (2019). What is a good school, and 
can parents tell? Evidence on the multidimensionality of school output (No. w25342). 
National Bureau of Economic Research. 

 
Booker, T., Gilpatric, S., Gronberg, T., & Jansen, D. (2008). The Effect of Charter Schools on 

traditional Public School Students in Texas: Are Children Who Stay Behind Left Behind?  

 
Bross, W., Harris, D., & Liu, L. (2016). The effects of performance-based closure and charter 
takeover on student performance. Technical Report, Education Research Alliance for 
New Orleans. Retrieved on July 10, 2020 from: 
https://educationresearchalliancenola.org/files/publications/Bross-Harris-Liu-The-
Effects-of-Performance-Based-School-Closure-and-Charter-Takeover-on-Student-
Performance.pdf 

 
Candal, C. (Ed.). (2018). National charter school law rankings and scorecard-2018: The 
essential guide for policymakers and advocates. The Center for Education Reform. 
Retrieved on December 17, 2019 from: https://edreform.com/wp-
content/uploads/2019/01/CER_National-Charter-School-Law-Rankings-and-Scorecard-
2018_screen_1-30-19.pdf 

 
Carlson, D., Cowen, J., & Fleming, D. (2013). Life after vouchers: What happens to students 

who leave private schools for the traditional public sector?. Educational Evaluation and 
Policy Analysis, 35(2), 179-199. 

 
Chubb, J., & Moe, T. (1990). Politics, markets, and America's schools. Brookings Institution 

Press. 

 
Coalition for the Future of Detroit Schoolchildren. (2015). The Choice is Ours: Road to 

Excellence for Troubled Michigan Begins in Detroit. Retrieved on December 17, 2019 
from: http://detroiteducationcoalition.org/wp-
content/uploads/2018/09/CFDS_ChoiceIsOurs.pdf 

 

 

108 

Community Education Commission. (2019). Participating GOAL Line Schools. Retrieved on 

December 17, 2019 from: https://www.goaldetroit.org/goal-line/#schools  

 
Cordes, S. (2017). In pursuit of the common good: The spillover effects of charter schools on 

public school students in New York City. Education Finance and Policy, 1-49. 

 
Denice, P., & Gross, B. (2016). Choice, preferences, and constraints: Evidence from public 

school applications in Denver. Sociology of Education, 89(4), 300-320. 

 
Edwards, D., Anderson, K., & Mohr, E. (2019). Transportation, Choice, and Parental Priorities: 

An Analysis of Detroit’s GOAL Line Program. Education Policy Innovation 
Collaborative. East Lansing, MI. Retrieved on January 28, 2021 from: 
https://epicedpolicy.org/wp-content/uploads/2020/05/GOALLine_Ridership.pdf  

 
Ekmekci, M., & Yenmez, M. (2019). Common enrollment in school choice. Theoretical 

Economics, 14(4), 1237-1270. 

 
Figlio, D., & Hart, C. (2014). Competitive effects of means-tested school vouchers. American 

Economic Journal: Applied Economics, 6(1), 133-56. 

 
Friedman, M. (1962). Capitalism and freedom. Chicago, IL: University of Chicago Press. 
 
Glazerman, S. (1998). School Quality and Social Stratification: The Determinants and 

Consequences of Parental School Choice. Presented at the American Educational 
Research Association Annual Meeting.  

 
Glazerman, S., & Dotter, D. (2017). Market signals: evidence on the determinants and 

consequences of school choice from a citywide lottery. Educational Evaluation and 
Policy Analysis, 39(4), 593-619. 

 
Gray, K. (2016, June 9). Legislature OKs $617M Detroit Public Schools Rescue Plan. Detroit 

Free Press. Retrieved on December 17, 2019 from: 
https://www.freep.com/story/news/politics/2016/06/09/dps-package-wins-gop-support-
headed-gov-snyder/85630880/  

 
Gross, B., DeArmond, M., & Denice, P. (2015). Common Enrollment, Parents, and School 

Choice: Early Evidence from Denver and New Orleans. Making School Choice Work 
Series. Center on Reinventing Public Education. 
https://files.eric.ed.gov/fulltext/ED556474.pdf  

 
Harris, D., & Larsen, M. (2019). The identification of schooling preferences: Methods and 

evidence from post-Katrina New Orleans. Technical Report, Education Research 
Alliance for New Orleans. Retrieved on July 10, 2020 from: 
https://educationresearchalliancenola.org/files/publications/Harris-Larsen-How-Parents-
Choose-2019-07-05.pdf  

 

 

109 

Harris, D., Liu, L., Gerry, A., & Arce-Trigatti, P. (2019). From revolution to evolution: Market 
dynamic in school value-added and marketed program offering under the post-Katrina 
school reforms in New Orleans. Technical Report, Education Research Alliance for New 
Orleans. Retrieved on July 10, 2020 from: 
https://educationresearchalliancenola.org/files/publications/2019-08-20-Harris-et-al-
From-Revolution-to-Evolution.pdf  

 
Harris, D. (2017). Why managed competition is better than a free market for schooling. 

Brookings Institution. Retrieved on July 10, 2020 from: https://www.brookings.edu/wp-
content/uploads/2017/03/harris-paper.pdf  

Hastings, J., Kane, T., & Staiger, D. (2005). Parental preferences and school competition: 
Evidence from a public school choice program (No. w11805). National Bureau of 
Economic Research. 

 
Hastings, J., & Weinstein, J. (2008). Information, school choice, and academic achievement: 

Evidence from two experiments. The Quarterly journal of economics, 123(4), 1373-1414.  

 
Hesla, K., White J., & Gerstenfeld, A.. (2019). "A Growing Movement: America's Largest 
Public Charter School Communities." National Alliance for Public Charter Schools. 
Retrieved on July 10, 2020 on: https://files.eric.ed.gov/fulltext/ED595146.pdf 

Hill, P. (2006). Put Learning First: A Portfolio Approach to Public Schools. PPI Policy 

Report. Progressive Policy Institute. Retrieved on July 10, 2020 from: 
https://files.eric.ed.gov/fulltext/ED491223.pdf 

 
Imberman, S. (2011). The effect of charter schools on achievement and behavior of public school 

students. Journal of Public Economics, 95(7-8), 850-863. 

 
Jabbar, H. (2015). “Every kid is money” market-like competition and school leader strategies in 

New Orleans. Educational Evaluation and Policy Analysis, 37(4), 638-659. 

 
Jochim, A., DeArmond, M., Gross, B., & Lake, R. (2014). How parents experience public school 

choice. Center for Reinventing Public Education. Retrieved on January 19, 2020 from: 
https://www.crpe.org/publications/how-parents-experience-public-school-choice  

 
Koedel, C., Mihaly, K., & Rockoff, J. (2015). Value-added modeling: A review. Economics of 

Education Review, 47, 180-195. 

 
Levin, H. (2015). Issues in education privatization. In H.F. Ladd & M.E. Goertz (Eds.), 
Handbook of research in education finance and policy (pp. 411-422). Routledge. 

Levin, K. (2016, March 21). Mayor Duggan launched a bus route to draw students back to 
Detroit schools. Is it working? Chalkbeat. Retrieved on December 17, 2019 from: 
https://chalkbeat.org/posts/detroit/2019/03/21/goal-line-bus-route-childcare/  

 

 

110 

 

 

 

Lincove, J., Cowen, J., & Imbrogno, J. (2018). What's in Your Portfolio? How Parents Rank 

Traditional Public, Private, and Charter Schools in Post-Katrina New Orleans’ Citywide 
System of School Choice. Education Finance and Policy, 13(2), 194-226. 

 
Lincove, J., Valant, J., & Cowen, J.  (2018). You can't always get what you want: Capacity 

constraints in a choice-based school system. Economics of Education Review, 67, 94-109. 

 
Loeb, S., Valant, J., & Kasman, M. (2011). Increasing choice in the market for schools: Recent 
reforms and their effects on student achievement. National Tax Journal, 64(1), 141-164. 

 
Long, B. (2004). How have college decisions changed over time? An application of the 
conditional logistic choice model. Journal of econometrics, 121(1-2), 271-296. 

 
McFadden, D.  (1974). Conditional logit analysis of qualitative choice behavior. In Frontiers in 

econometrics, edited by Paul Zarembka, pp. 105–142. New York: Academic Press. 

 
McShane, M. & Shaw, M. (2020). Transporting school choice students: A primer on states’ 

transportation policies related to private, charter, and open enrollment students. 
EdChoice. Retrieved on May 29, 2020 from: https://www.edchoice.org/wp-
content/uploads/2020/03/Transporting-School-Choice-Students-by-Michael-Q-McShane-
and-Michael-Shaw.pdf  

 
Michigan Department of Education. (2013). Michigan State School Aid Act [MCL 388.1705; 

388.1705c; Sections 105, 105c]. Retrieved on December 17, 2019 from: 
http://mi.gov/documents/mde/choice1_279579_7.pdf  

 
Michigan Department of Education. (2017). Charter Schools- Questions and Answers. Retrieved 

on December 17, 2019 from: 
https://www.michigan.gov/documents/PSAQA_54517_7.pdf   

 
National Center for Education Statistics. (2018). Digest of Education Statistics, 2016 (NCES 

2017-094), Table 206.40. Retrieved on July 31, 2020 from: 
https://files.eric.ed.gov/fulltext/ED580954.pdf  

 
Sass, T. (2006). Charter schools and student achievement in Florida. Education Finance and 

Policy, 1(1), 91–122.  

 
Schneider, M., & Buckley, J. (2002). What do parents want from schools? Evidence from the 

Internet. Educational evaluation and policy analysis, 24(2), 133-144. 

 
Schneider, M., Teske, P., Marshall, M., & Roch, C. (1998). Shopping for schools: In the land of 

the blind, the one-eyed parent may be enough. American Journal of Political Science, 
769-793. 

 
Singer, J. (2020). Student stratification among a combination of school choice policies in 

Detroit. Journal of School Choice, 14(1), 122-153. 

 

111 

 

 
Strauss, V. (2016, July 15). How charter schools in Michigan have hurt traditional public 

schools, new research finds. The Washington Post. Retrieved on December 17, 2019 
from: https://www.washingtonpost.com/news/answer-sheet/wp/2016/07/15/how-charter-
schools-in-michigan-have-hurt-traditional-public-schools-new-research-finds/  

 
Urban Institute Student Transportation Working Group. (2018). The road to school: How far 

students travel to school in Denver, Detroit, New Orleans, New York City, and 
Washington, DC. Washington, DC: Urban Institute. Retrieved on February 24, 2020 
from:https://greaterdc.urban.org/sites/default/files/publication/97151/the_road_to_school
_7.pdf.  

Valant, J. (2014). Better Data, Better Decisions: Informing School Choosers to Improve 

Education Markets. American Enterprise Institute for Public Policy Research. Retrieved 
on July 31, 2020 from: https://files.eric.ed.gov/fulltext/ED555547.pdf  

 
Winters, M. (2015). Narrowing The Charter Enrollment Gap: Denver's Common-Enrollment 
System. Manhattan Institute. Retrieved on July 10, 2020 from: https://www.manhattan-
institute.org/html/narrowing-charter-enrollment-gap-8235.html  

 
Winters, M. (2012). Measuring the effect of charter schools on public school student 

achievement in an urban environment: Evidence from New York City. Economics of 
Education review, 31(2), 293-301. 

 
Wixom, M.  & Keily, T. (2018). 50-State Comparison: Open Enrollment Policies (2018 Update) 

Education Commission of the States. Retrieved on December 17, 2019 from: 
https://www.ecs.org/open-enrollment-policies/  

 
Wixom, M. (2018). 50-State Comparison: Charter School Policies. Education Commission of the 

States. Retrieved on December 17, 2019 from: https://www.ecs.org/charter-school-
policies/  

 
Wolf, P., Kisida, B., Gutmann, B., Puma, M., Eissa, N., & Rizzo, L. (2013). School Vouchers 

and Student Outcomes: Experimental Evidence from Washington, DC. Journal of Policy 
Analysis and Management, 32(2), 246-270. 

 
Ziebarth, T. (2019). Measuring up to the model: A ranking of state public charter school 

laws. National alliance for public charter schools. Retrieved on July 30, 2020 from: 
https://www.publiccharters.org/ranking-state-public-charter-school-laws-2019  

 
Zimmer, R. & Buddin, R. (2009). Is Charter School Competition in California Improving the 
Performance of Traditional Public Schools? Public Administration Review. 69(5): 831- 
845.  

 

 

112 

Zimmer, R., Gill, B., Booker, K., Lavertu, S., Sass, T., & Witte, J. (2009). Charter Schools in 

Eight States: Effects on achievement, attainment, integration, and competition. MG-869. 
Pittsburgh: RAND.  

 

 

 

113 

PAPER 3 ANOTHER ONE RIDES THE BUS: THE IMPACT OF SCHOOL 

TRANSPORTATION ON STUDENT OUTCOMES IN MICHIGAN 

Introduction 

Since 1869, some public school districts have provided students transportation to school 

to ensure consistent attendance and allow districts to consolidate (McDonald & Howlett, 2007). 

Today, over 25 million children, about half of U.S. public school students, ride a school bus to 

and from school. During the 2015-16 school year, school districts spent over $24 billion on 

student transportation nationally, about $1,000 per student transported—constituting 

approximately eight percent of average per pupil expenditures (National Center for Education 

Statistics, 2019). To reduce these high costs of the school bus, districts have cut school bus 

routes, provided public transit passes, contracted with rideshare companies, halted bus service 

altogether, and in some states, charged parents for transportation services in recent years (Bergal 

June 16 2015; Cornwall May 1 2018). 

Decreasing the availability of district-provided transportation may be detrimental to 

student outcomes. School transportation can positively affect school attendance, an increasingly 

important outcome, by providing a reliable, consistent, and safe mode of transportation. This 

removes the logistical and financial burdens of school transit from parents, making it easier for 

students to get to school regularly, increasing how often they attend school. These burdens may 

be especially prohibitive for low-income families since they are more likely to live in 

neighborhoods that are unsafe to walk through, have little access to car or any form of direct 

transportation, and be chronically absent, commonly defined as missing more than ten percent of 

possible school days (Balfanz & Byrnes, 2012; Sampson, Raudenbush, & Earls, 1997; Urban 

Institute Student Transportation Working Group, 2018). In addition to increasing attendance, 

 

114 

school transportation may raise student achievement since school attendance has a positive effect 

on student achievement and is positively associated with on-time graduation and socio-emotional 

outcomes (Aucejo & Romano, 2016; Kirksey, 2019; Gershenson, Jacknowitz, & Brannegan, 

2017; Gottfried, 2010, 2011, 2014b Kirksey, 2019).  

Although districts have funded school transportation for over 150 years, little research 

has existed concerning its role in increasing student outcomes until recently. Descriptive 

evidence shows that riding the school bus, the most prevalent form of school transportation, is 

positively associated with attendance (Cordes, Leardo, Rick & Schwartz, 2019; Gottfried, 2017). 

However, these differences in attendance between students who ride the bus and those who do 

not can be explained by student and school characteristics, implying that students who use the 

bus do so to attend better schools (Cordes et al., 2019). To my knowledge, no evidence exists 

concerning the direct effects of school bus transportation on student achievement.  

In this paper, I provide some of the first causal evidence concerning the effects of school 

transportation on student attendance and achievement. I ask:  

1.  What is the effect of school transportation on student attendance? 

2.  What is the effect of school transportation on student achievement? 

3.  How do these effects differ for economically disadvantaged students?  

To answer these questions, I estimate the effects of transportation on students’ attendance rates, 

an indicator for being chronically absent, and standardized math and English Language Arts 

(ELA) test scores using a rich panel of statewide, student-level enrollment records, achievement, 

and address data as well as a unique dataset of local transportation provisions I collected from 

the 50 largest districts in Michigan. To estimate transportation effects, I exploit the walking 

distance cutoffs that determine student eligibility for the school bus using a regression 

 

115 

discontinuity design. Because my treatment is school bus eligibility, my estimates are the intent-

to-treat (ITT) effects of riding the school bus, the policy-relevant estimate of school 

transportation since policymakers can only control who is eligible for transportation and cannot 

force students to ride the bus. 

 

I find that transportation eligibility increases attendance rates for economically 

disadvantaged students by two-thirds of a percentage point, the equivalent of approximately one 

day in a 180 day school year. Similarly, transportation eligibility decreases the probability of 

being chronically absent for economically disadvantaged students by two to four percentage 

points. I find little evidence that transportation eligibility affects attendance for students who are 

not economically disadvantaged. Finally, I detect no significant effects of the school bus on 

student achievement. These findings imply that school transportation is likely most important for 

students who would not attend school as frequently without a safe and reliable mode of 

transportation. Thus, districts should consider implementing a means-tested transportation 

program or using it as an intervention to reduce chronic absenteeism to more effectively allocate 

this expensive resource.  

 

This paper proceeds as follows. First, I discuss how school transportation can affect 

attendance and achievement and the prior research concerning school transportation. Second, I 

outline Michigan’s statewide transportation policies, the context of my study. Third, I describe 

my data, its sources, and my sample. Next, I explain my research design and how I test its 

validity. Finally, I present my results and discuss their implications for policy.  

Background: The Policy Relevance of Student Attendance 

Over the last decade, there has been an increased focus on reducing student absenteeism 

to improve student outcomes. It is well-established that school attendance is strongly associated 

 

116 

with student achievement and educational attainment with some causal evidence suggesting that 

absences decrease student test scores (Aucejo & Romano, 2016; Gottfried, 2009, 2010, 2011, 

2019; Gershenson, et al, 2017; Kirksey, 2019). In particular, chronic absenteeism, commonly 

defined as missing more than ten percent of days in a school year, is particularly detrimental to 

school performance. Students who are chronically absent have lower levels of academic 

achievement, are less eager to learn, and are less likely to graduate on time (Allensworth & 

Easton, 2007; Chang & Romero, 2008; Gottfried, 2014b). Chronic absenteeism not only affects 

those who miss school, but their classmates. An increase in the percent of classmates who are 

chronically absent is associated with lower test scores (Gottfried, 2019). 

In addition to student outcomes, chronic absenteeism can negatively impact district 

budgets and school performance on state accountability systems. In seven states, including 

California and Texas, the amount of funding a district receives is tied to its average daily 

attendance (Baker, 2014). In Michigan, the setting of this study, school funding is determined by 

the number of students in attendance on two count days during the school year. Furthermore, 

Michigan districts only receive funds for the instructional days they have over 75 percent of 

students enrolled in attendance (MI Sec 388.1701). Therefore, districts risk losing funding when 

they have high chronic absenteeism rates. Additionally, schools likely face state intervention if 

they have high chronic absenteeism rates in most states. Under the Every Student Succeeds Act 

(ESSA), states are required to report chronic absenteeism rates on state report cards. 

Furthermore, over 70 percent of states chose to incorporate it into their state accountability 

system, including Michigan (Bauer, Liu, Schanzenbach, & Shambaugh, 2018).  

Although schools are held accountable for chronic absenteeism in addition to student 

achievement and attainment, family characteristics and environmental factors largely influence 

 

117 

these student outcomes, including attendance (Lenhoff & Pogodzinski, 2018). Prior research 

finds a strong positive relationship between poverty, absences, and chronic absenteeism (Balfanz 

& Byrnes, 2012; Dougherty, 2018; Gottfried, 2014a; Morrissey, Hutinson, & Winsler, 2014; 

Ready, 2010). Access to health care, housing instability, neighborhood crime, car ownership, and 

access to public transit may influence impoverished students’ abilities to attend school regularly 

(Baker, 2014; Lenhoff & Pogodzinski, 2018). For example, low-income students are more likely 

to miss school because of chronic health conditions (Bauer et al., 2018; Meng, Babey, & 

Wolstein, 2012). Furthermore, district-level chronic absenteeism rates are higher in cities with 

high rates of asthma, violent crime, cold weather, and residential vacancy (Singer, Cook, 

Lenhoff, & Pogodzinski, 2019). 

A particularly salient factor that may negatively impact attendance is distance to school. 

Less than ten percent of students who live more than a mile away from their school walk or bike 

to school (Federal Highway Administration, 2019). Therefore, students who live farther from 

school likely rely on a car, bus, or train to transport them to school. Because families who live in 

high poverty neighborhoods are less likely to own a car, distance likely inhibits low-income 

students from attending school unless they have access to public or school transportation (Urban 

Institute Student Transportation Working Group, 2018). In fact, research on commute times and 

absenteeism in Washington, D.C. and Baltimore, two cities with high percentages of low-income 

students, finds that students with longer commutes have higher rates of absenteeism (Blagg, 

Rosenboom, & Chingos, 2018; Stein & Grigg, 2019).  

How Can School Transportation Affect Attendance and Achievement? 

One school resource that could mitigate the effects of distance to school and other 

familial and environmental factors that negatively impact school attendance is publicly provided 

 

118 

school transportation. School buses can reduce the difficulty of traveling to and from school. 

First, they remove the logistical and financial burdens of transporting students to school from 

parents. Without school or public transportation, parents must have access to transportation, have 

adequate time to take their child to and from school, and be able to afford any additional 

expenses that the commute incurs. These burdens may be especially prohibitive for low-income 

families since they are less likely to have reliable access to transportation and adequate time or 

money to transport their children to school. Second, the school bus provides a daily routine. 

Research shows that having routines reduces stress in students (Wolin & Bennett, 1984). If the 

school bus provides families a routine by having a reliable and consistent mode of transportation, 

it could reduce stress, increasing positive attitudes towards going to school, leading to higher 

attendance (Gottfried, 2017). Finally, school buses offer a safe way to get to school. If students 

have to walk through dangerous neighborhoods to get to school or brave extreme weather 

conditions, they may not go to school. Thus, by providing a reliable and safe way of getting to 

school, school buses could make it easier for students to get to school on a regular basis, 

increasing their attendance.  

In addition to raising attendance rates, school transportation can affect student 

achievement. The most direct way that school transportation can change achievement is by 

increasing attendance. Since I hypothesize that school transportation raises attendance rates and 

it is well established that regular attendance increases test scores, it follows that school 

transportation could have a positive effect on achievement as well. However, riding the school 

bus could have negative effects on achievement even if it increases attendance. Riding the school 

bus to school likely takes longer than riding in a private car due to the multiple stops and 

circuitous routes. Thus, long bus rides could harm student achievement by reducing the time 

 

119 

students have to do homework, participate in extra-curricular activities, or sleep. To my 

knowledge, only one descriptive study of the relationship between commute times and 

achievement exists. Blagg, Rosenboom, & Chingos (2018) find no association between test 

scores and commute time. Additionally, riding the school bus could decrease test scores if 

bullying, fighting, and other undesirable social behaviors take place on the bus.  

Prior Research on School Transportation and Student Outcomes 

Prior research concerning district-provided transportation finds a positive association 

between transportation and attendance. In Baltimore, Burdick-Will, Stein, & Grigg (2019) show 

that incidences of violent crime decrease attendance for 9th graders who have to walk through the 

neighborhoods where the crimes occur. However, there is no change in attendance when students 

use a district-provided bus pass to ride the public bus through the same neighborhoods. This 

implies that by providing a safe mode of transportation, districts may mitigate some of the 

negative effects of traveling to school. Using a nationally representative sample of 

kindergarteners, Gottfried (2017) finds that riding the school bus is associated with increased 

attendance. Similarly, school bus riders in New York City have higher attendance rates and are 

less likely to be chronically absent. However, differences in attendance between students who 

ride the bus and those who do not are explained by student and school characteristics, implying 

that riding the bus may allow students to attend better schools rather than directly increase 

attendance (Cordes et al., 2019). To my knowledge, there is no evidence concerning the 

relationship between school bus transportation and student achievement.  

Although these studies attempt to account for student, school, and district characteristics 

through covariates and fixed effects, it is likely that there are unobserved characteristics in the 

error term that are correlated with riding the school bus biasing the results. I add to these studies 

 

120 

by providing some of the first causal evidence of the direct effects of the school bus on 

attendance and achievement using a regression discontinuity design that compares the outcomes 

of students who live on opposite sides of the walking distance cutoff that determines whether the 

district provides transportation. Plausibly, the only difference between these two groups is that 

one is transportation eligible, and one is not, allowing me to estimate the impact of school 

transportation on student outcomes. Furthermore, my design allows me to compare outcomes of 

students who attend the same school, allowing me to account for choice of school. 

School Transportation in Michigan 

According to state law, Michigan school districts are not required to provide 

transportation to general education students, but districts may choose to do so at their discretion 

(MI Sec 380.1321). Although most decisions concerning the provision of school transportation 

are made at the local level, there exist state laws that regulate how transportation is provided and 

to whom if districts offer it. First, the decision to provide transportation must be made at the 

elementary, middle, or high school level. For example, if a district offers transportation to one 

elementary school, they must offer it to all elementary schools. However, they would not have to 

provide it to high school students. Second, districts that provide transportation must offer it to 

resident students who attend the public school “which they are eligible to be admitted” and live 

more than 1.5 miles from that school (MI Sec 380.1321). It is up to interpretation whether this 

means that districts only have to provide transportation to students who attend their assigned 

school or all schools in the resident district that the student is eligible to attend. However, many 

districts state that they only provide transportation to the assigned school. Additionally, some 

districts offer transportation to students who live closer to their school than the 1.5 mile state 

mandated cutoff. Finally, Michigan districts cannot charge resident students for transportation 

 

121 

(MI Sec 380.1321). Although there are some state funds for transportation expenses, the majority 

of school transportation costs are covered by district’s operational budgets (MI Sec 388.1674). 

However, districts can charge students using inter-district choice for transportation if they choose 

to provide it to them. I also note that the above regulations do not apply to charter schools since 

they do not have resident students (Michigan Department of Education, 2017).  

Data 

To estimate the effects of school transportation on student attendance and achievement, I 

primarily use student-level records from the Michigan Department of Education (MDE) and the 

Center for Educational Performance and Information (CEPI), and transportation policies 

collected from the 50 largest school districts in Michigan. The student-level records include 

enrollment and demographic information (e.g. race and ethnicity, gender, disability status, 

English Learner status, and economically disadvantaged status), number of days the student 

attended school, the number of days the student could have attended the school, test scores on 

state standardized exams (either the Michigan Educational Assessment Program, MEAP, or the 

Michigan Student Test of Educational Progress, M-STEP), and student addresses geocoded at the 

census block level for all Michigan public school students from the 2012-13 school year to the 

2018-19 school year. I also use school-level records made publicly available by MDE and CEPI 

that include the school’s address, educational settings, and grades offered.  

District Transportation Policies 

 

As described above, school transportation in Michigan is at the discretion of local 

districts. There is no centralized state-administered data resource on individual district policies, 

so local nuances must be collected directly at the district level. For this paper, I collected district 

transportation policies for the 50 largest traditional public school districts in Michigan from 

 

122 

district websites and bylaws during Fall 2019. The largest districts were determined by 

enrollment during the 2017-18 school year, the most recent year of student-level data available at 

the time of collection. These 50 districts account for nine percent of districts in Michigan but 

contain one-third of the total public school student population in the state. I coded the policies 

for the date it was last changed, eligibility requirements, the modes of transportation the district 

offers, and any restrictions on the provision of transportation. In particular, I collected 

information concerning the walking distance cutoffs that determined eligibility for school-

provided transportation. I use these cutoffs as a form of exogeneous variation to estimate the 

causal effects of school transportation.  

In Table 19, I examine variation in transportation policy provisions across districts in my 

sample. I consider differences by schooling level because Michigan state law requires that 

decisions regarding the provision of transportation be made at the elementary, middle, and high 

school levels. All but one of the districts in my sample offer transportation (Grosse Pointe) and 

all but two districts in my sample only offer yellow bus transportation (Ann Arbor and Detroit to 

high school students). For these reasons, I focus on the districts that offer school bus 

transportation in my analysis. Therefore, the results of my analysis can be interpreted as the 

effects of the school bus on student outcomes. I also examine the prevalence of two types of 

restrictions on school bus eligibility in Table 19: attendance at assigned school and walking 

distance cutoffs. 22 districts in my sample explicitly stipulate that students are only offered 

district-provided transportation if they attend their assigned school. However, the absence of the 

assigned school provision does not mean that the other districts in the sample provide 

transportation to all schools in the district. They may rely on the language in the state law 

concerning “eligible to be admitted” to only offer transportation to the assigned school.  

 

123 

Table 19. District Transportation Policy Provisions 
  
Total Districts 
Offers transportation 
Must Attend Assigned School 
Mode of Transportation 
Yellow Bus 
City Bus 
Walking Distance Cutoff 
.25 Miles 
.5 Miles 
.75 Miles 
1 Miles 
1.5 Miles 
No Cutoff 
Note. Sample includes the 50 largest traditional public school districts in Michigan in terms of enrollment during the 
2017-18 school year.  
 

Elementary  Middle  High 
50 
49 
22 
 
47 
2 
 
0 
2 
1 
5 
26 
15 

50 
49 
22 
 
49 
0 
 
2 
1 
1 
8 
23 
14 

50 
49 
22 
 
49 
0 
 
1 
5 
2 
17 
10 
14 

Thirty-four districts in my sample explicitly state the walking distance cutoff that 

determine transportation eligibility. In elementary school, the majority of districts have walking 

distances of one or 1.5 miles. In middle and high school most districts have walking distances of 

1.5 miles, the maximum distance set by the state. Districts who choose a cutoff closer than the 

state mandated cutoff may do so to minimize or maximize the number of transportation eligible 

students. For example, the percent of students eligible for transportation in districts in my sample 

that have a 0.75 mile walking distance cutoff would increase by 25 percentage points if they 

changed the cutoff to 0.5 miles from school, likely increasing transportation costs. Thus, I focus 

on the effects of school transportation in districts with a 1.5 mile cutoff in my analysis because it 

is less likely to have been manipulated by districts which could bias my results. 

 
 
 
 
 

 

124 

Table 20. District Characteristics of Sample, 2017-18. 

  

State  Sample  1.5 Mile Cutoff 

50 

10,856 

23 

12,541 

63 
35% 
57% 
9% 
51% 
70% 
10% 
9% 
6% 
5% 
37% 
12% 
11% 
94.42 
14% 
0.23 
0.26 

65 
32% 
62% 
6% 
51% 
67% 
14% 
8% 
6% 
5% 
39% 
13% 
8% 
93.94 
17% 
0.20 
0.21 

537 
2,520 
108 
6% 
27% 
67% 
52% 
79% 
8% 
7% 
2% 
5% 
55% 
14% 
4% 
93.05 
19% 
-0.02 
-0.02 

N Districts 
Avg. Total Enrollment 
Avg. Sq. Miles 
City 
Suburb 
Rural 
Avg. Pct. Female 
Avg. Pct. White 
Avg. Pct. Black 
Avg. Pct. Hispanic 
Avg. Pct. Asian 
Avg. Pct. Other Race 
Avg. Pct. Econ. Dis. 
Avg. Pct. SWDs 
Avg. Pct. ELs 
Avg. Attendance Rate 
Avg. Pct. Chronic. Abs. 
Avg. Std. Reading Score 
Avg. Std. Math Score 
Note. Unweighted district characteristics created using student level data. Econ. Dis., SWD, and EL are 
abbreviations for economically disadvantaged, student with disabilities, and English Learner respectively. In 
Michigan, students are considered economically disadvantaged if they qualify for free or reduced lunch, receive 
food (SNAP) or cash assistance (TANF), or they are homeless, migrant, or in foster care. Sample includes the 50 
largest traditional public school districts in Michigan in terms of enrollment during the 2017-18 school year. 1.5  
mile cutoff includes all districts that have a 1.5 mile walking distance cutoff for at least one grade between grades K-
8. Attendance rates are from public report made available by the Center for Education Performance and Information. 
3 districts with less than 10 students do not have attendance rates available. 24 districts have missing chronic 
absenteeism rates. One district does not have test scores. 
 

In Table 20, I compare the average characteristics of districts in the collected and analytic 

samples to all districts in the state during the 2017-18 school year. Sampled districts are more 

densely populated since they serve more students but are smaller in land area than the average 

Michigan district. A higher percentage of sampled districts are located in cities and suburbs. 

Also, they have a higher percentage of non-White students, a lower percentage of economically 

disadvantaged students, and higher average achievement.  

 

125 

Analytic Sample 

To construct my sample, I begin with 428,174 student-year observations of students in 

grades K-8 who have a walking distance cutoff of 1.5 miles and attend their nearest school 

offering their grade in their resident district, my proxy for assigned school, between the 2012-13 

and 2018-19 school years.15 It is likely that students are only guaranteed transportation to their 

assigned school since 22 of the 50 of the district transportation policies I collected explicitly 

stated that students were only guaranteed transportation to their assigned school. Also, I cannot 

guarantee that districts that do not have this provision in their transportation policy do not limit 

transportation eligibility to a student’s assigned school since districts may rely on the language in 

the state law or use their discretion to do so. Even if a district offers transportation to all schools 

in the district, they must offer it to their assigned school as well. Therefore, by restricting the 

sample to students who likely attend their assigned school, I am ensuring all students in my 

sample are transportation eligible if they live more than 1.5 miles from school.  

I exclude 2,333 (0.5%) student-year observations of homeless students and 44,760 

(10.5%) student-year observations of students with disabilities from my sample because they 

may receive school transportation regardless of distance between home and school. Under the 

McKinney-Vento Act, districts are required to transport homeless students to their school of 

origin (U.S. Department of Education, 2018). Students with disabilities are guaranteed 

transportation if their Individualized Education Program (IEP) Team deems it as a necessary 

service (U.S. Department of Education, 2009). Although the number of students with disabilities 

 

15 To determine a student’s nearest school, I first determine which district the student lives in using the population 
weighted centroid of their resident census block and district boundary shape files from the Michigan Department of 
Technology, Management, and Budget. Then, I calculate the geodetic (“as the crow flies”) distance to from their 
census block to each school in their resident district that offers general education and the student’s grade excluding 
virtual schools, boarding schools, and other residential schools. I use the school with the shortest distance as their 
nearest school. 
 

 

126 

whose IEPs include the provision of transportation may vary by district, existing evidence shows 

that a large percentage of students with disabilities receive transportation services regardless of 

whether it is specified in their IEP. Thus, I drop all students with disabilities from my sample.16 

My final analytic sample includes 380,909 student-year observations.17  

Treatment and Forcing Variables 

I use a strict regression discontinuity design that leverages the walking distance cutoffs to 

estimate the effects of transportation eligibility on student attendance and achievement outcomes. 

This design assumes that, local to the cutoff, the average student on either side of the cutoff is 

identical, with one exception: one side is eligible for transportation and the other is not. Thus, 

any estimated differences in outcomes can be attributed to transportation eligibility as long as 

families do not choose their residence based on transportation eligibility and there is nothing 

other than the outcomes that change discontinuously at the cutoff. I consider my estimates to be 

the intent to treat (ITT) effects of riding the school bus because I use transportation eligibility 

rather than bus ridership to determine treatment since I do not have data concerning which 

students actually ride the bus on a daily basis. I likely underestimate the effects of riding the 

school bus on student outcomes because I consider possible non-compliers—those eligible for 

transportation but do not ride the bus—as treated. However, these ITT effects can be considered 

a more policy-relevant parameter than the average treatment effect on the treated because 

policymakers cannot force students to use a bus. Instead, they can change who is eligible for 

 

16 My data include special education services provided. However, few students have transportation reported as a 
necessary service. According to an interview with a Detroit Public Schools Community District administrator, about 
sixty percent of special education students received door-to-door transportation to any school in the district during 
the 2017-18 school year (Sattin-Bajaj, 2018). Although DPSCD transported over 3,300 students with disabilities, 
significantly fewer students had transportation reported as a necessary service in the administrative data during the 
2017-18 school year. Therefore, I have no indication in the data concerning which students with disabilities receive 
transportation and choose to drop them all to ensure they do not bias my results.  
17 Also, I exclude 172 observations of students who do not have reported attendance variables.  

 

127 

transportation to either encourage or restrict school bus use. Therefore, the true impact of 

changes to transportation policies is likely to include compliers and non-compliers making the 

effect of transportation eligibility, the ITT effect, the most policy-relevant estimate.  

Students are considered transportation eligible if they live more than 1.5 miles walking 

distance from school. Walking distance, my forcing variable, is calculated from the population 

weighted centroid of the student’s home census block to the exact address of their attended 

school using Here Application Programming Interface (API), a similar tool to Google Maps, 

using the quickest route assuming average traffic. Because I do not have any data concerning 

who districts consider transportation eligible, I assume that walking distance perfectly predicts 

treatment. This assumption likely biases my results towards zero. If there is an effect of the bus 

on student outcomes, the inclusion of the outcomes of students considered treated who are not 

transportation eligible should weaken my estimates. Similarly, the presence of outcomes of 

students considered not treated who actually are eligible for the school bus in the sample should 

also reduce the estimated effect of transportation eligibility.  

Attendance and Achievement Measures 

I estimate the effects of school bus eligibility on student attendance and achievement. I 

use two measures of attendance: the student’s annual attendance rate and an indicator for being 

chronically absent. I calculate the attendance rate using the rules set out by CEPI. I divide the 

number of days the student attended the school by the number of days they could have possibly 

attended.18 Since I theorize that the school bus increases attendance by lowering the financial and 

 

18 Prior to the 2017-18 school year, a student was considered in attendance if they attended any part of the school 
day. Starting in the 2017-18 school year, a student had to attend at least 50 percent of the school day to be 
considered in attendance. Although I contend that this change should not affect my results because my estimates are 
created by comparing attendance outcomes from the same year due to the inclusion of grade-by-school-by year fixed 
effects, I estimate my models separately for the years before and after the change in the attendance definition as a 
robustness check. Results are similar in direction and magnitude for each sample and are available by request. 

 

128 

time costs associated with transporting students to and from school, I hypothesize that 

transportation eligibility may increase the likelihood that a student attends school on a regular 

basis rather than marginally increasing attendance. Therefore, I also use an indicator for being 

chronically absent as an outcome. I consider a student to be chronically absent if their attendance 

rate is less than 90 percent, which is MDE’s definition of chronic absence (Center for 

Educational Performance and Information, 2020). I measure achievement using test scores on the 

state standardized exam, the MEAP or the M-STEP, for students in grades 3 through 8. 

Specifically, I use math and ELA test scores standardized within grade, subject, and year.19  

Table 21 examines differences in characteristics and outcomes between transportation 

eligible and ineligible students in my full and analytic samples. For my analysis, I restrict my 

sample to students who live within 0.4 miles of the cutoff, my preferred bandwidth for 

estimating causal effects. Just under half of students in my sample are transportation eligible. A 

lower percentage of English Learners, White, and economically disadvantaged students are 

transportation eligible in the full sample. However, there are few differences between 

transportation eligible and ineligible students who live closer to the walking distance cutoff. As 

for attendance rates, students who are transportation eligible have slightly higher attendance rates 

than students who do not. These differences are similar for students who live within 0.4 miles of 

the cutoff. Additionally, transportation eligible students have higher test scores, but this 

difference is much smaller within my preferred bandwidth. 

 

 

 

19 The MEAP was administered during the first two school years of my panel, 2012-13 and 2013-14, while the M-
STEP was administered from 2014-15 to 2018-19. Because I standardize the test scores within grade, subject, and 
year and use school-by-grade-by-year fixed effects in my models, estimates are created by comparing test scores 
from the same year. Therefore, changes in the test administered should not affect my results. 
 

 

129 

Table 21. Student Characteristics of Analytic Sample 

1.5 Mile Cutoff 

Full Sample 

0.4 Mile Bandwidth 

 
 
 

0.77 
0% 
50% 
80% 
6% 
7% 
4% 
3% 
47% 
25% 
7% 
95.70 
0.27 
0.21 

51,347 
1.30 
0% 
50% 
75% 
8% 
7% 
6% 
4% 
33% 
11% 
8% 
95.57 
0.35 
0.30 

93,281 
1.47 
45% 
50% 
75% 
8% 
6% 
7% 
4% 
32% 
11% 
8% 
95.68 
0.36 
0.30 

380,909 
1.73 
44% 
51% 
78% 
7% 
6% 
6% 
3% 
37% 
17% 
7% 
95.79 
0.36 
0.29 

212,737  168,172 
2.95 
100% 
51% 
75% 
8% 
5% 
8% 
4% 
26% 
7% 
7% 
95.90 
0.46 
0.38 

Full Sample  Not Eligible  Eligible  Full Sample  Not Eligible  Eligible 
41,934 
1.69 
100% 
51% 
74% 
8% 
6% 
8% 
4% 
31% 
11% 
7% 
95.83 
0.37 
0.31 

N 
Avg. Walk Distance 
Pct. Transport Elig. 
Pct. Female 
Pct. White 
Pct. Black  
Pct. Hispanic 
Pct. Asian 
Pct. Other Race 
Pct. Econ. Dis. 
Pct. English Learner 
Pct. Chronic Absent 
Avg Attendance Rate 
Avg. Std. Math Score 
Avg. Std. ELA Score 
Note. Sample includes all student by year observations in analytic sample. 0.4 Mile bandwidth includes all students 
who have a walking distance from home to school that is between 1.1 miles and 1.9 miles. Walking distances are 
calculated from the population weighted centroid of the student’s resident census block to their school assuming 
average traffic using Here API. Students are transportation eligible (Transport Elig.) if their walking distance to 
school is greater than 1.5 miles . Econ. Dis. is an abbreviation for economically disadvantaged. In Michigan, 
students are considered economically disadvantaged if they qualify for free or reduced lunch, receive food (SNAP) 
or cash assistance (TANF), or they are homeless, migrant, or in foster care. Students are considered chronically 
absent if they miss more than ten percent of possible days in the school year. Math and English Language Arts 
(ELA) test scores on the Michigan Educational Assessment Program, MEAP, or the Michigan Student Test of 
Educational Progress, M-STEP are standardized within grade, subject and year. Because students take these tests in 
grades 3-8, 285,025 and 283,579 student-year observations in my sample have math or ELA test scores respectively.  
 

Method 

 

I estimate the effects of school bus eligibility on student attendance and achievement 

using a strict regression discontinuity design. Specifically, I exploit the walking distance cutoffs 

that determine eligibility for district transportation. I estimate: 

!!"#$=#%+#&%&'(')&*!#$++,-'./012*!#$3+4'(5+6"#$+7!"#$  

Where !!"#$ is one of the following four outcomes for student i in grade g who attends nearest 

                   (6) 

school j at time t: attendance rate, an indicator that equals one if student i is chronically absent, 

 

130 

equals one if student i is eligible to receive transportation to school j. Students in my sample are 

standardized math test score, or standardized ELA test score. %&'(')&*!#$, my treatment indicator 
school bus eligible if they live more than 1.5 miles from school. +,-'./012*!#$3 is a flexible 
%&'(')&*!#$ in my preferred models. 4'( contains student characteristics including race, gender, 

function of the walking distance from student i’s home to their school j in year t, my forcing 

variable. I use a linear term of my forcing variable and its interaction with my treatment, 

economically disadvantaged status, and English Learner indicators. In the models where math or 

ELA test score is the outcome, I include a lagged test score in my vector of student 

characteristics to account for prior achievement.  

Transportation eligibility, school characteristics, and neighborhood characteristics are 

likely correlated and associated with my outcomes. Therefore, my estimates of the effects of 

school bus eligibility would be biased if I do not account for students’ schools and 

neighborhoods. In fact, Cordes et al. (2019) find that most differences in attendance between 

students who ride the bus and those who do not in New York City are due to differences in the 

schools they attend. This implies that students who ride the bus do so to attend better schools, 

biasing the direct effect of the school bus on student attendance. To ensure that where students 

choose to attend school does not bias my results, I include, 6"#$, a grade-by-school-by-year fixed 

effect in my preferred models. This ensures that my estimates are created by only comparing 

students who attend the same school and the same grade during the same school year. Because I 

restrict my sample to students who attend their nearest school, my proxy for assigned school, 

6"#$ not only holds constant school characteristics but neighborhood characteristics as well, 

accounting for choice of school and home. Furthermore, the grade-by-school-by year fixed 

effects account for grade and year specific trends. I cluster by standard errors by school. 

 

131 

One of the conditions of my research design is that I must limit my sample to 

observations local to the cutoff to produce causal estimates of the effects of transportation 

eligibility. Thus, I estimate Equation 6 on the sample of students who live within 0.4 miles of the 

walking distance cutoff, my preferred bandwidth. My choice of bandwidth is informed by the 

optimal bandwidth procedures proposed by Calonico, Cattaneo, & Titiunik (2014). Additionally, 

I hypothesize that district-provided transportation has larger effects for impoverished students 

since low-income students are less likely to have access to direct forms of transportation and 

they are more likely to live in unsafe neighborhoods (Urban Institute Student Transportation 

Working Group, 2018; Sampson, Raudenbush, & Earls, 1997). Therefore, they are more likely to 

rely on schools to provide a reliable and safe way to get to school. To test this hypothesis, I also 

estimate the model represented by Equation 6 on samples restricted to either economically 

advantaged or disadvantaged students, my poverty indicator. In Michigan, students are 

considered economically disadvantaged if they qualify for free or reduced lunch, receive food 

(SNAP) or cash assistance (TANF), or they are homeless, migrant, or in foster care. 

Validity of Design 

 

In order for my research design to estimate the causal effect of school bus eligibility, two 

assumptions must hold. First, I must assume that nothing other than treatment and outcomes 

change discontinuously at the cutoff. If other factors change at the cutoff, then my estimate of the 

effect of transportation eligibility may be biased since my treatment indicator would be 

correlated with unobserved characteristics in the error term. To provide evidence that this 

assumption likely holds, I examine whether or not my treatment, %&'(')&*!#$ predicts pre-

determined observable student characteristics using balance tests. Specifically, I estimate:  

8!"#$=#%+#&%&'(')&*!#$++,-'./012*!#$3+6"#$+7!"#$ 

 

 

(7) 

 

132 

where 8!"#$ is an indicator for one of the following student characteristics for student i in grade g 

attending nearest school j at time t: female, White, Black, Hispanic, Asian, Other Race, 

economically disadvantaged, or English Learner. I estimate the model represented in Equation 7 

on the sample of students who live within 0.4 miles of the walking distance cutoff using either a 

linear term of -'./012*!#$ and its interaction with my treatment, %&'(')&*!#$, or a linear term, a 
standard errors for %&'(')&*!#$ for each of the outcomes I predict in my balance tests. I find no 

quadratic term, and their interactions with treatment. Table 22 displays the coefficients and 

significant differences in any of my observable student characteristics at the walking distance 

cutoff, providing some confidence that no other characteristics change discontinuously at the 

cutoff other than the treatment and the outcomes.20 

Second, I must assume that families do not manipulate themselves into treatment, 

meaning that they do not choose their residences based on school bus eligibility. If families do 

manipulate themselves into treatment, it is likely that there are unobserved characteristics 

correlated with treatment biasing my estimates of transportation eligibility. Although school bus 

eligibility rules are publicly available, they tend to be buried within district bylaws. Furthermore, 

families may be more focused on which school their children would be assigned to attend rather 

than whether or not they would be eligible for the school bus when choosing a home. To provide 

evidence that this assumption holds, I first visually check for discontinuities using histograms of 

the frequency of observations around the cutoff. In Figure 9, I present the histograms of the 

frequency of observations in my sample as a function of distance from the cutoff. In Figure 9A, I 

use a bin size of 0.01 miles and find that there is a large number of observations between 1.5 and  

 

 
 

20 I estimate versions of Equation 2 using various bandwidths. Results are similar and can be found in Table 27. Less 
than five percent of estimates have p-values less than 0.05 providing confidence that my research design is internally 
valid.  

 

133 

Table 22. Estimated Coefficients of Transportation Eligibility for Balance Tests 
  
OUTCOMES 
  
Female 

(1) 

  

Coefficient (SE)  Coefficient (SE) 

(2) 

  

0.012 
(0.016) 
0.035 
(0.024) 
-0.017 
(0.014) 
-0.016 
(0.010) 
0.004 
(0.015) 
-0.006 
(0.008) 
-0.026 
(0.023) 
-0.003 
(0.014) 

 

-0.002 
(0.011) 
0.017 
(0.017) 
-0.003 
(0.010) 
-0.008 
(0.008) 
0.001 
(0.008) 
-0.008 
(0.005) 
0.00571 
(0.015) 
-0.002 
(0.011) 

White 

Black 

Hispanic 

Asian 

Other Race 

Econ. Dis. 

EL 

 

 

 

 

 

 

 

 
 

 

 

93,281 

93,281 

Observations 
WALKING DISTANCE FUNCTIONAL FORM 
Linear Term 
Linear Term interacted with Transportation Eligibility Indicator 
Quadratic Term 
Quadratic Term interacted with Transportation Eligibility Indicator 
Note. Standard errors in parentheses. *p<0.1, ** p<0.05, ***p<0.01. Models include school-by-grade-by-year fixed 
effects. Standard errors are clustered at the school level. The sample includes student-year observations living within 
0.4 miles of the transportation eligibility cutoff, my preferred bandwidth. ED and EL are abbreviations for 
economically disadvantaged and English Learner respectively. In Michigan, students are considered economically 
disadvantaged if they qualify for free or reduced lunch, receive food (SNAP) or cash assistance (TANF), or they are 
homeless, migrant, or in foster care. 
 
1.51 miles from the cutoff. Taken by itself, this could imply that families are choosing homes 

X 
X 
X 
X 

X 
X 

  

right over the cutoff in order to receive treatment. However, there is little difference in the 

number of observations on either side of the cutoff when I compare the number of observations 

that are 0.1 miles from the cutoff in Figure 9B. When combined, the histograms show that there 

is some evidence of bunching within 53 feet of the cutoff but little evidence of bunching at 530 

feet of the cutoff.  

 

 

134 

Figure 9. Distribution of Observations 
 
Figure 9A: 0.01 Mile Bin Size  

 

 

 

 

 

Figure 9B: 0.1 Mile Bin Size   

 

 

Note. Sample includes observations of students in the analytic sample living between zero and five miles from their 
attended school. 
 

Next, I formally test for bunching around the cutoff using the statistical test proposed by 

 

McCrary (2008). If there is statistically significant evidence of a discontinuity in the density of 

observations at the cutoff, then it is likely that families manipulate themselves into treatment. I 

perform the McCrary test using multiple bandwidths and bin sizes. In addition to the choice of 

bandwidth, the choice of bin size is important in a McCrary test since it regresses the number of 

observations within a bin as a function of each bin’s midpoint to detect discontinuities in density 

at the cutoff (McCrary, 2008). At my preferred bandwidth, 0.4 miles, and the optimal bin size, 

 

135 

0.004 miles, I find a statistically significant and positive discontinuity in the density of 

observations. However, if I use a bin size of 0.1 miles, I do not detect a statistically significant 

discontinuity in the density of observations at the cutoff. Similar to the findings of the visual 

examinations of the histograms, the results of the McCrary test provide evidence of bunching 

within a hundred feet of the cutoff but not within five hundred feet of the cutoff.  

Given these findings, I argue that it is unlikely that this bunching is evidence of 

manipulation since it is unlikely that families who choose their residence for school bus 

eligibility can control whether their home is within a hundredth or a tenth of the mile from the 

cutoff. Rather, this discontinuity is a likely result of an idiosyncrasy of the distribution of 

residences in the data. To investigate whether or not the bunching is an artifact of normal 

residential patterns, I first examine the characteristics of students who live within 0.01 miles of 

the cutoff and the census blocks they live in. Although five students live in a census block in a 

given year on average in my sample, I find that there is one census block that is one thousandth 

of a mile from the walking distance cutoff and is home to over 130 students a year. This census 

block contains a mobile home park with 430 multi-family homes all with the same address 

(MHVillage Inc., 2021). Because the McCrary test uses local polynomial regressions that weigh 

observations closer to the cutoff more heavily, this census block likely causes the discontinuity 

in the density of observations. To explore this, I estimate the McCrary test without the mobile 

home park census block and find no statistically significant evidence of bunching. Furthermore, 

there is little visual evidence of bunching in the histograms of the frequency of observations in 

the sample with the mobile home park census block is removed as seen in Figure 12 in the 

Appendix. Therefore, this one census block which exhibits a predictable residential pattern likely 

drives the bunching.  

 

136 

Although the evidence presented above provides confidence that my regression 

discontinuity design is internally valid, I cannot fully rule out the possibility of manipulation. To 

address this concern, I estimate the model represented by Equation 6 using a donut regression 

discontinuity approach as a specification check. Specifically, I exclude observations extremely 

close to the cutoff, including the mobile home park census block, where there could be possible 

manipulation. Dropping observations at data heaps produces unbiased estimates of the treatment 

effect in regression discontinuity designs (Barreca, Lindo, & Waddell, 2015). 

Results 

 

Before I present the results of my main models, I display graphs of the unadjusted 

average attendance and achievement outcomes by distance from school in Figures 10 and 11 to 

visually examine discontinuities in the outcomes at the transportation eligibility cutoff. In 

Figures 10A and 10B, I detect no visual change in attendance rate in the full sample or the 

sample of economically advantaged students. However, a different pattern emerges in the 

attendance rates of economically disadvantaged students in Figure 10C. For economically 

disadvantaged students who are not eligible to ride the school bus, there is a negative relationship 

between distance and attendance rate. However, this relationship does not exist for transportation 

eligible students. Furthermore, there is visual evidence of a discontinuity in attendance rates at 

the cutoff. Taken together, the evidence from Figure 10C implies that the school bus not only 

mitigates the negative effects of distance on attendance for economically disadvantaged students 

but improves their attendance as well.  

 

 

137 

Figure 10. Unadjusted Average Attendance Rate and Proportion Chronically Absent by Distance from Threshold  
Figure 10A: Attendance Rate Full Sample Figure 10B: Attendance Rate Not Econ. Dis. Figure 10C: Attendance Rate Econ. Dis. 

 
Figure 10D: Chronic. Absent Full Sample Figure 10E: Chronic. Absent Not Econ. Dis.     Figure 10F: Chronic. Absent Econ. Dis. 

 

 

 

  

Note. Each dot represents the average outcome for all observations within a 0.05 mile bin width. Students are considered chronically absent if they miss more 
than ten percent of possible days in the school year. Walking distances are calculated from the population weighted centroid of the student’s resident census 
block to their school assuming average traffic using Here API. The eligibility cutoff is 1.5 miles. Econ. Dis. is an abbreviation for economically disadvantaged. 
In Michigan, students are considered economically disadvantaged if they qualify for free or reduced lunch, receive food (SNAP) or cash assistance (TANF), or 
they are homeless, migrant, or in foster care. 
 
 
 
 

 

 

 

138 

Figure 11. Unadjusted Average Standardized Math and ELA Test Scores by Distance from Threshold 
        Figure 11A: Math Score Full Sample  Figure 11B: Math Score Not Econ. Dis.       Figure 11C: Math Score Econ. Dis. 

Figure 11D: ELA Score Full Sample       Figure 11E: ELA Score Not Econ. Dis. 

  

 

 
 Figure 11F: ELA Score Econ. Dis. 

  
 
Note. Each dot represents the average outcome for all observations within a 0.05 mile bin width. Math and English Language Arts (ELA) test scores on the 
Michigan Educational Assessment Program, MEAP, or the Michigan Student Test of Educational Progress, M-STEP are standardized within grade, subject and 
year. Walking distances are calculated from the population weighted centroid of the student’s resident census block to their school assuming average traffic using 
Here API. The eligibility cutoff is 1.5 mile. Econ. Dis. is an abbreviation for economically disadvantaged. In Michigan, students are considered economically 
disadvantaged if they qualify for free or reduced lunch, receive food (SNAP) or cash assistance (TANF), or they are homeless, migrant, or in foster care. 

  

 

 

139 

An analogous pattern emerges in Figures 10D, 10E, and 10F which displays the 

unadjusted proportion of chronically absent students by distance from the threshold for the full 

sample and the samples restricted to economically advantaged or disadvantaged students. There 

is little difference in the proportion of students who are chronically absent between 

transportation eligible and ineligible students in the full sample or the sample of economically 

advantaged students. However, there is evidence that proportion of chronically absent students  

increases as distance to school increases for economically disadvantaged students who are not 

school bus eligible. Furthermore, there is evidence that there is a discontinuity in the proportion 

of students who are chronically absent at the cutoff implying that the school bus reduces the 

probability of being chronically absent. Figure 11 graphs the relationships between distance to 

school and standardized math and ELA test scores. There is no evidence of large slope changes 

or discontinuities for any of my samples, implying that transportation eligibility may have no 

effect on achievement.  

 

Although the evidence provided in Figure 10 shows that economically disadvantaged 

students who are transportation eligible have higher attendance rates and are less likely to be 

chronically absent, it may be biased by unaccounted for student or school characteristics. In 

Table 23, I present the results of the model represented by Equation 6 using a 0.4 mile 

bandwidth, providing regression adjusted, causal estimates of the effects of transportation 

eligibility for the full sample. Columns 1 and 2 display the effects of school bus eligibility on 

attendance rates and the indicator for being chronically absent respectively. I detect a small 

positive effect of transportation eligibility on attendance that is statistically significant at a 90 

percent confidence level. Specifically, I find that transportation eligibility increases attendance 

rates by approximately 0.2 percentage points. This is equivalent to almost a half a day increase in 

 

140 

attendance in a 180 day school year. Similarly, I find that school bus eligibility decreases the 

probability of being chronically absent by 1.5 percentage points. In Columns 3 and 4, I present 

the results for math and ELA test scores. I do not detect any statistically significant effect of 

transportation eligibility on student achievement.  

Table 23. Estimated Effects of Transportation Eligibility  
  

(1) 

(2) 

(3) 

(4) 

 

  
Transportation Eligibility 

 

Distance 

 

Distance*Transport. Eligibility  

 

Constant 

Attendance Rate Chronically Absent Math Score ELA Score 

  

0.197* 
(0.119) 
-0.581 
(0.367) 
0.444 
(0.492) 
96.03*** 
(0.088) 

  

-0.015*** 
(0.006) 
0.044** 
(0.018) 
-0.0268 
(0.0228) 
0.0571*** 
(0.004) 

  

  

0.009 
-0.004 
(0.011) 
(0.010) 
-0.026 
-0.008 
(0.035) 
(0.025) 
-0.010 
0.0340 
(0.0360) 
(0.047) 
0.129***  0.113*** 
(0.007) 
(0.008) 

 
 

 

 

 

 

52,448 
0.666 

93,281 
0.020 

93,281 
0.017 

52,228 
0.578 

Observations 
Adj R Squared 
Note. Standard errors in parentheses. *p<0.1, ** p<0.05, ***p<0.01. Models include student race, gender, 
economically disadvantaged status, and English Learner covariates and school-by-grade-by-year fixed effects. 
Standard errors are clustered at the school level. The sample includes student-year observations living within 0.4 
miles of the transportation eligibility cutoff, my preferred bandwidth. Students are considered to be transportation 
eligible if students live farther than 1.5 miles walking distance from school. Walking distances are calculated from 
the population weighted centroid of the student’s resident census block to their school’s address assuming average 
traffic using Here API. Students are considered chronically absent if they miss more than ten percent of possible 
days in the school year. Math and English Language Arts (ELA) test scores on the Michigan Educational 
Assessment Program, MEAP, or the Michigan Student Test of Educational Progress, M-STEP are standardized 
within grade, subject and year. 
 

To examine whether or not the effects of transportation eligibility are larger for 

economically disadvantaged students, I estimate the model represented by Equation 6 on samples  

restricted to either economically advantaged or disadvantaged students. The results of these 

specifications are displayed in Table 24. I find little evidence that transportation eligibility  

affects the attendance rate or the probability of being chronically absent for economically 

advantaged students. Instead, I find that the positive effects of school bus eligibility found in the

 

141 

Table 24. Heterogeneous Effects of Transportation Eligibility by Economically Disadvantaged Status 
  

(4) 

(5) 

(3) 

(6) 

(7) 

(8) 

Chronically Absent 
Not Econ. 

Math Score 

Not Econ. 

ELA Score 

Not Econ. 

 

 

  
Transportation Eligibility 

 

Distance 

 

Distance*Transport. Eligibility 

 

Constant 

 
 

(2) 
(1) 
Attendance Rate 

Not Econ. 

Dis. 
  

-0.011 
(0.105) 
-0.146 
(0.328) 
0.268 
(0.466) 
96.36*** 
(0.069) 

 

Econ. Dis. 

  

0.661** 
(0.285) 
-1.582** 
(0.752) 
0.895 
(1.123) 
93.75*** 
(0.209) 

 

Dis. 
  

-0.003 
(0.005) 
0.026* 
(0.014) 
-0.026 
(0.020) 
0.043*** 
(0.003) 

 

Econ. Dis. 

  

-0.040*** 
(0.014) 
0.096** 
(0.044) 
-0.060 
(0.058) 
0.172*** 
(0.010) 

 

Dis. 
  

-0.010 
(0.012) 
-0.011 
(0.030) 
0.062 
(0.043) 
0.146*** 
(0.008) 

 

Econ. Dis. 

  

0.006 
(0.017) 
0.010 
(0.048) 
-0.045 
(0.069) 
0.004 
(0.013) 

 

16,692 
0.615 

Dis. 
  

0.004 
(0.013) 
-0.044 
(0.039) 
0.051 
(0.056) 
0.126*** 
(0.009) 

 

Econ. Dis. 

  

0.013 
(0.022) 
0.036 
(0.066) 
-0.170** 
(0.084) 
-0.012 
(0.014) 

 

63,299 
0.003 

29,982 
0.011 

63,299 
0.003 

Observations 
Adj R Squared 
Note. Standard errors in parentheses. *p<0.1, ** p<0.05, ***p<0.01. Models include student race, gender, and English Learner covariates and school-by-grade-
by-year fixed effects. Standard errors are clustered at the school level. The sample includes student-year observations living within 0.4 miles of the transportation 
eligibility cutoff, my preferred bandwidth. Samples are restricted to either economically disadvantaged or advantaged students. Students are considered to be 
transportation eligible if students live farther than 1.5 miles walking distance from school. Walking distances are calculated from the population weighted 
centroid of the student’s resident census block to their school’s address assuming average traffic using Here API. Students are considered chronically absent if 
they miss more than ten percent of possible days in the school year. Math and English Language Arts (ELA) test scores on the Michigan Educational Assessment 
Program, MEAP, or the Michigan Student Test of Educational Progress, M-STEP are standardized within grade, subject and year. Econ. Dis. is an abbreviation 
for economically disadvantaged. In Michigan, students are considered economically disadvantaged if they qualify for free or reduced lunch, receive food (SNAP) 
or cash assistance (TANF), or they are homeless, migrant, or in foster care.

35,690 
0.543 

35,756 
0.656 

16,538 
0.570 

29,982 
0.007 

 

142 

full sample are driven by the effects for economically disadvantaged students. Specifically, I find 

that transportation eligibility increases attendance rates for economically disadvantaged students 

by two-thirds of a percentage point. This is equivalent to about one day in a 180 day school year. 

Moreover, school bus eligibility decreases the probability of being chronically absent for 

economically disadvantaged students by four percentage points. This large effect on chronic 

absenteeism for economically disadvantaged students confirms my hypothesis that school  

transportation has the greatest impact on students who do not have a reliable, consistent, and safe 

way to get to school without it. Finally, Table 24 shows little evidence that school bus eligibility 

has a significant effect on achievement for economically advantaged or disadvantaged students. 

Specification Checks 

In order to ensure that my results are not sensitive to my choices in functional form, 

bandwidth, estimator, and sample, I perform the following specification checks. First, I estimate 

Equation 1 using various bandwidths and a quadratic polynomial term of distance to school, my 

forcing variable. Results of these specifications are similar and can be found in Tables 28 and 29 

in the Appendix. Second, I use a nonparametric estimator, the optimal bandwidth calculated 

using the method proposed by Calonico, Cattaneo, & Titiunik (2014), and the robust bias-

corrected inference procedures detailed by Calonico, Cattaneo, & Farrell (2020) with district, 

grade, and year fixed effects to estimate the effects of school bus eligibility on my attendance 

outcomes.21 Specifically, I estimate local linear polynomial regressions with a first order 

 

21 At the time of this writing, the rdrobust command that implements the procedures discussed by Calonico, 
Cattaneo, & Titiunik (2014) and Calonico, Cattaneo, & Farrell (2020) does not allow for the use of fixed effects. 
Therefore, I must create binary indicators for each school-by-grade-by-year fixed effect. However, I have over 
20,000 school-by-grade-by-year combinations in my data and Stata does not allow for that many variables. 
Therefore, I use the district, grade, and year fixed effects for the non-parametric models due to computational 
limitations at this time. Furthermore, the rdrobust command was unable to calculate standard errors in the models 
where standardized test score was an outcome. Therefore, I only display the results of the attendance outcomes in 
Table 30 in the Appendix.  

 

143 

polynomial function to construct the estimates and a second order polynomial function to 

construct the bias correction with triangular kernel functions. Results of the models using the 

nonparametric estimator are displayed in Table 30 in the Appendix. The estimates of 

transportation eligibility from the nonparametric model are similar in direction and slightly larger 

in magnitude than the models using a parametric estimator and school-by-grade-by-year fixed 

effects. 

Finally, I estimate the model represented by Equation 1 on samples that exclude 

observations within 0.01, 0.05, and 0.1 miles of the cutoff. Although I argue that the 

discontinuity in the density of observations that I detect at the cutoff is driven by an idiosyncrasy 

of residential patterns, I use this donut regression discontinuity approach to account for possible 

manipulation extremely close the cutoff as a robustness check. I present the results of the donut 

regressions in Table 25. First, I find that the positive effect of school bus eligibility on attendance 

rates in the full sample is not robust to the exclusion of observations close to the cutoff. Second, I 

show that the estimated transportation eligibility effects on attendance rates for economically 

disadvantaged students are similar in direction and magnitude but lose statistical significance 

when the sample is reduced. Finally, I find that the estimated effects of transportation eligibility 

on chronic absenteeism for the full sample and the sample of economically disadvantaged 

students are similar and statistically significant when observations within 0.01 or 0.05 miles from 

the cutoff are excluded. However, they are no longer statistically significant when all 

observations within 0.1 miles from the cutoff are excluded. Because this sample exclusion 

reduces the sample by almost one third, my inability to detect significant effects may be due to a 

loss of power. Taken together, the results of these specification checks show that, at the very 

least, my finding that transportation eligibility reduces the probability of being chronically absent 

 

144 

Table 25. Estimated Effects of Transportation Eligibility, Donut Regressions 
  
(7) 

(5) 

(2) 

(1) 

(3) 

(4) 

(6) 

(8) 

(9) 

(10) 

(11) 

(12) 

 

  
0.01 Mile Exclusion 
Transport. Eligibility 

 

Observations 
0.05 Mile Exclusion 
Transport. Eligibility 

 

Observations 
0.1 Mile Exclusion 
Transport. Eligibility 

 

Full 

Sample 

  

0.135 
(0.126) 
90,042 

  

0.146 
(0.165) 
80,665 
  

Attendance Rate 

Not Econ. 

Dis. 

  

-0.052 
(0.107) 
61,292 

  

-0.070 
(0.137) 
54,702 

  

  

Econ. 
Dis. 

  

0.562* 
(0.292) 
28,750 

  

0.451 
(0.362) 
25,963 

Chronically Absent 

Full 

Sample 

Not Econ. 

Dis. 

Econ. 
Dis. 

  

-0.011* 
(0.006) 
90,042 

  

-0.014* 
(0.008) 
80,665 
  

  

-0.002 
(0.005) 
61,292 

  

-0.003 
(0.006) 
54,702 

  

  

-0.033** 
(0.013) 
28,750 

  

-0.034* 
(0.018) 
25,963 
  

Full 

Sample 

  

-0.007 
(0.010) 
50,841 

  

-0.002 
(0.012) 
45,423 
  

Math Score 
Not Econ. 

Dis. 

  

-0.013 
(0.012) 
34,741 

  

-0.019 
(0.013) 
31,028 

  

0.016 
(0.200) 
68,223 

-0.178 
(0.182) 
46,371 

0.258 
(0.446) 
21,852 

-0.008 
(0.011) 
68,223 

0.002 
(0.010) 
46,371 

-0.028 
(0.025) 
21,852 

0.004 
(0.017) 
38,256 

-0.012 
(0.020) 
26,221 

Econ. 
Dis. 

  

0.002 
(0.017) 
16,100 

  

0.026 
(0.023) 
14,395 
  
0.034 
(0.033) 
12,035 

Full 

Sample 

  

0.009 
(0.012) 
50,629 

  

0.002 
(0.015) 
45,232 

ELA Score 
Not Econ. 

Dis. 

  

0.005 
(0.013) 
34,680 

  

-0.007 
(0.017) 
30,973 

  

  

-0.014 
(0.022) 
38,096 

-0.005 
(0.025) 
26,169 

Econ. 
Dis. 

  

0.011 
(0.022) 
15,949 

  

0.018 
(0.026) 
14,259 
  
-0.052 
(0.036) 
11,927 

Observations 
Note. Standard errors in parentheses. *p<0.1, ** p<0.05, ***p<0.01. Models include walking distance to school, its interaction with transportation eligibility, 
student race, gender, economically disadvantaged status, and English Learner covariates and school-by-grade-by-year fixed effects. Standard errors are clustered 
at the school level. The sample includes student-year observations living within 0.4 miles of the transportation eligibility cutoff, my preferred bandwidth. 
Observations that are within either 0.01, 0.05, or 0.1 miles are excluded. Students are considered to be transportation eligible if students live farther than 1.5 
miles walking distance from school. Walking distances are calculated from the population weighted centroid of the student’s resident census block to their 
school’s address assuming average traffic using Here API. Students are considered chronically absent if they miss more than ten percent of possible days in the 
school year. Math and English Language Arts (ELA) test scores on the Michigan Educational Assessment Program, MEAP, or the Michigan Student Test of 
Educational Progress, M-STEP are standardized within grade, subject and year. Econ. Dis. is an abbreviation for economically disadvantaged. In Michigan, 
students are considered economically disadvantaged if they qualify for free or reduced lunch, receive food (SNAP) or cash assistance (TANF), or they are 
homeless, migrant, or in foster care

 

145 

for economically disadvantaged students is likely not a result of my choices in functional form, 

bandwidth, estimator, and sample. 

Falsification Test 

 

Although my results are robust to many of my specification choices, there still may exist 

unobserved characteristics that are correlated with living 1.5 miles away from the schools in my 

sample, biasing my estimates. For example, if multi-family housing is located closer to schools 

while single family homes are more likely to be located past the walking distance cutoff, then my 

estimates could be biased by the effects of living in single family homes on my outcomes. To 

provide evidence that it is unlikely that living in the neighborhoods that are 1.5 miles away from 

the schools in my sample is correlated with unobserved factors that determine either attendance 

or achievement, I estimate the model represented by Equation 1 on the sample of elementary 

school students who attend their nearest school and live in a district in my sample that has a 

walking distance cutoff of 1.5 miles for middle schools but not for elementary schools. This 

sample provides an ideal test to falsify my results for two reasons. First, students in the placebo 

sample are not affected by the 1.5 mile cutoff that determines transportation eligibility in my 

main sample. Therefore, my results are likely biased if I detect a significant effect of 

transportation eligibility when I estimate my model on this sample. Second, the students in my 

placebo sample are very similar to those in my main sample because they live in the same 

neighborhoods and could even be in the same families as students in my main sample. Therefore, 

if there are family or neighborhood characteristics that are biasing my estimates of transportation 

eligibility, they should affect my placebo sample as well because they live in the same 

neighborhoods as my main sample.  

 

146 

To ensure that students in my placebo sample live in the same neighborhoods as those in 

my main sample, I use the walking distance from home to the nearest middle school in their 

resident district as my forcing variable in my falsification test. Therefore, I consider students in 

my placebo sample as treated if they live 1.5 miles or more from the middle school that they 

would attend, the same schools that students attend in my main sample.22 I present the results of 

my falsification tests for each of my four outcomes for the full sample and the samples restricted 

to either economically advantaged or disadvantaged students in Table 26. For all outcomes and 

samples, I do not detect a statistically significant effect of transportation eligibility. Furthermore, 

most of the estimates are in the opposite direction of the main results. Thus, the results of this 

falsification test provide evidence it is likely that no unobserved characteristics associated with 

living in the neighborhoods in my sample are correlated with my treatment and predict my 

outcomes, giving more confidence that my models are estimating the causal effect of 

transportation eligibility. 

 Discussion 

In this paper, I provide some of the first causal evidence of school transportation effects 

on student attendance and achievement. I find that being eligible for school bus transportation—

the intent-to-treat (ITT) estimate of school bus use—decreases the probability of being 

chronically absent especially for economically disadvantaged students by two to four percentage 

points. Combined with prior research, my results show that the school bus can mitigate the 

 

22 I also estimate my falsification test using distance from home to the attended elementary school as the forcing 
variable. The results of this falsification test are similar to the results of the falsification test presented in Table 8 and 
are available by request. Because the elementary schools may be located in different neighborhoods than middle 
schools, students in the placebo sample considered treated using the distance to the attended school may be less 
likely to live in the same neighborhoods as the students in the main sample than the students in the placebo sample 
that are considered treated using the distance to the nearest middle school. Therefore, I prefer distance to middle 
school as the forcing variable determining treatment in my falsification test. 

 

147 

Table 26. Placebo Effects of Transportation Eligibility 
  
(4) 

(1) 

(2) 

(3) 

(5) 

(6) 

(7) 

(8) 

(9) 

(10) 

(11) 

(12) 

Full 

Sample 

  

-0.118 

(0.116) 
0.248 

(0.422) 
-0.094 

Attendance Rate 

Not Econ. 

Dis. 

  

-0.075 

(0.132) 
-0.340 

(0.408) 
0.990 

Econ. 
Dis. 

  

-0.168 

(0.221) 
1.111 

(0.751) 
-1.769 

Chronically Absent 

Full 

Sample 

Not Econ. 

Dis. 

  

0.001 

(0.007) 
-0.000 

(0.022) 
0.012 

  

-0.006 

(0.006) 
0.020 

(0.016) 
0.005 

Econ. 
Dis. 

  

0.013 

(0.012) 
-0.020 

(0.046) 
-0.002 

Full 

Sample 

  

-0.001 

(0.019) 
0.028 

(0.057) 
-0.071 

Math Score 
Not Econ. 

Dis. 

  

-0.007 

(0.023) 
0.016 

(0.077) 
-0.015 

Econ. 
Dis. 

  

0.002 

(0.034) 
0.089 

(0.087) 
-0.195* 

Full 

Sample 

  

-0.018 

(0.018) 
0.023 

(0.072) 
-0.035 

ELA Score 
Not Econ. 

Dis. 

  

-0.012 

(0.025) 
0.004 

(0.090) 
-0.042 

Econ. 
Dis. 

  

-0.024 

(0.033) 
0.093 

(0.111) 
-0.084 

(0.724) 
96.24*** 
(0.090) 

(0.754) 
96.42*** 
(0.096) 

(1.208) 
94.49*** 
(0.175) 

(0.035) 
0.048*** 
(0.005) 

(0.032) 
0.035*** 
(0.004) 

(0.073) 
0.138*** 
(0.009) 

(0.067) 
0.197*** 
(0.0144) 

(0.104) 
0.238*** 
(0.018) 

 

 

  
Transportation Eligibility 

 

Distance 

 

Distance*Transport. Elig. 

 

Constant 

(0.087) 
0.131*** 
(0.018) 

(0.117) 
0.167*** 
(0.023) 

(0.146) 
-0.029 
(0.026) 

(0.099) 
0.021 
(0.023) 

 

 

 

 

 

 

 

0.008 

0.012 

0.011 

0.015 

0.007 

56,186 

34,521 

56,186 

34,521 

21,665 

 
 
Observations 
Adj R Squared 
0.547 
Note. Standard errors in parentheses. *p<0.1, ** p<0.05, ***p<0.01. Models include student race, gender, economically disadvantaged status, and English 
Learner covariates and school-by-grade-by-year fixed effects. Standard errors are clustered at the school level. The placebo sample includes student-year 
observations of elementary school students who attend their nearest school in their resident district, live in districts with a walking distance cutoff of 1.5 miles for 
middle school but not elementary school, and live within 0.4 miles of the transportation eligibility cutoff, my preferred bandwidth. Students are considered to be 
transportation eligible if students live farther than 1.5 miles walking distance from their nearest middle school. Walking distances are calculated from the 
population weighted centroid of the student’s resident census block to their nearest middle school’s address assuming average traffic using Here API. Students 
are considered chronically absent if they miss more than ten percent of possible days in the school year. Math and English Language Arts (ELA) test scores on 
the Michigan Educational Assessment Program, MEAP, or the Michigan Student Test of Educational Progress, M-STEP are standardized within grade, subject 
and year. Econ. Dis. is an abbreviation for economically disadvantaged. In Michigan, students are considered economically disadvantaged if they qualify for free 
or reduced lunch, receive food (SNAP) or cash assistance (TANF), or they are homeless, migrant, or in foster care. 

15,066 

9,589 

0.594 

5,477 

0.583 

9,573 

0.528 

15,033 

21,665 

0.006 

 

5,460 

0.611 

0.553 

 

 

 

 

 

148 

negative effects of distance on attendance for low-income students. In particular, the large effects 

of transportation eligibility on chronic absenteeism for vulnerable students provide some 

compelling evidence that school-provided transportation most likely increases attendance for 

students who would regularly miss school without the school bus, in accordance with my 

hypothesis. However, my results show little evidence that district-provided transportation has an 

effect on student achievement. Although it is well-established that attendance is positively 

associated with achievement, the possible positive benefits of increased attendance on 

achievement due to the use of the school bus could be outweighed by the negative effects of long 

commutes or increased bullying, fighting, or participation in other undesirable social behaviors 

on the school bus.  

For districts focused on increasing attendance rates, my findings may imply that at least 

maintaining existing bus routes, if not increasing the amount of transportation provided, could 

help curb chronic absenteeism. However, widespread provision of school transportation may not 

be the most cost-effective or efficient way to allocate this expensive resource since it has little 

effect on attendance for economically advantaged students or on achievement. Instead, districts 

concerned with reducing chronic absenteeism could implement means-tested school bus 

programs that provide transportation for free to all low-income students, similar to the National 

School Lunch Program. Additionally, districts could use school transportation as an intervention 

for chronically absent students. If it is identified that getting to school is a barrier to regular 

attendance for a particular student, schools could choose to provide that student transportation 

regardless of where they live.  

Although I provide evidence that transportation eligibility reduces chronic absenteeism 

for low-income students, more evidence is needed in order to make any strong policy 

 

149 

recommendations concerning transportation. First, my study generalizes largely only to districts 

located in cities and suburbs, not rural communities, since the districts in my sample have fewer 

economically disadvantaged students and mostly are suburban. In particular, I only estimate the 

effects of transportation eligibility for students living in a district with a 1.5 mile walking 

distance cutoff, local to the cutoff. The school bus likely has a larger effect in rural districts 

where half of students live at least ten minutes away from the nearest school by car (Edwards, in 

press). Furthermore, increasing access to school bus transportation may not raise attendance for 

students living within a half mile of their school since the burden of walking to school is much 

smaller for them than students living over a mile away from school. 

Second, research concerning the effects of the school bus on other outcomes, the 

mechanisms that could explain variation in school bus effects, and the effectiveness of other 

modes of transit would help provide policymakers with a more complete picture of the 

effectiveness of district transportation when making decisions concerning its provision. For 

example, riding the bus could increase disciplinary incidents and lower other socioemotional 

outcomes which could outweigh its positive effects on attendance. Furthermore, the effects of the 

school bus on achievement as well as other outcomes may vary by time spent on the bus. Riding 

the school bus may make a child attend school more often, but the longer he or she spends on the 

bus, the more difficulty he or she may have concentrating at school, negatively affecting 

achievement and negating any of the attendance benefits of riding the bus on achievement. 

Additionally, other modes of transit, like public transportation or ridesharing services, could be 

more effective in getting students to school regularly. Finally, cost-effectiveness studies of 

school transportation are needed to determine whether the benefits of increasing or changing the 

provision of transportation would offset the high costs associated with the school bus. In 

 

150 

conclusion, this paper provides a starting point for future work on the effects of school 

transportation, which I have shown to have meaningful effects on student attendance and chronic 

absence—an increasingly salient metric for education policy and decision-making.  

 
 

 

 
 

 

151 

APPENDIX

 

152 

Table 27. Estimated Coefficients of Transportation Eligibility for Balance Tests, All Bandwidths 
and Splines 
  
OUTCOMES 
Linear Spline 
0.5 Mile Bandwidth 

(3) 
Black  Hispanic  Asian  Other Race  Econ. Dis. 

Female  White 

(8) 
EL 
  

(7) 

(4) 

  

(5) 

  

(6) 

  

(1) 

  

(2) 

  

  

-0.007 
(0.004) 
-0.008 
(0.005) 
-0.003 
(0.006) 
-0.007 
(0.007) 
-0.013 
(0.008) 
-0.013 
(0.009) 

-0.006 
(0.007) 
-0.006 
(0.008) 
-0.014 
(0.009) 
-0.017 
(0.011) 
-0.019 
(0.0123) 
-0.037** 
(0.015) 

  

-0.003 
(0.013) 
0.006 
(0.015) 
-0.020 
(0.017) 
-0.022 
(0.019) 
-0.033 
(0.020) 
-0.007 
(0.025) 

0.003 
(0.021) 
-0.026 
(0.023) 
-0.013 
(0.023) 
-0.022 
(0.025) 
0.0113 
(0.0301) 
0.023 
(0.038) 

0.000 
(0.009) 
-0.002 
(0.011) 
-0.003 
(0.012) 
-0.001 
(0.013) 
-0.003 
(0.013) 
0.009 
(0.015) 
  
-0.004 
(0.014) 
-0.003 
(0.014) 
0.001 
(0.015) 
-0.005 
(0.016) 
0.005 
(0.020) 
0.018 
(0.023) 

  

  

 

0.4 Mile Bandwidth 

 

0.3 Mile Bandwidth 

 

0.25 Mile Bandwidth 

 

0.2 Mile Bandwidth 

 

0.15 Mile Bandwidth 
  
Quadratic Spline 
0.5 Mile Bandwidth 

 

0.4 Mile Bandwidth 

 

0.3 Mile Bandwidth 

 

0.25 Mile Bandwidth 

 

0.2 Mile Bandwidth 

 

0.004 
(0.0091) 
-0.002 
(0.011) 
0.010 
(0.011) 
0.005 
(0.013) 
-0.004 
(0.016) 
-0.010 
(0.021) 
  

0.007 
(0.014) 
0.012 
(0.016) 
-0.012 
(0.020) 
-0.009 
(0.022) 
-0.011 
(0.0255) 
-0.003 
(0.034) 

0.023 
(0.014) 
0.017 
(0.017) 
0.023 
(0.020) 
0.031 
(0.021) 
0.039 
(0.024) 
0.030 
(0.029) 
  

0.019 
(0.022) 
0.035 
(0.024) 
0.037 
(0.026) 
0.057** 
(0.029) 
0.038 
(0.0325) 
0.056 
(0.038) 

-0.011 
(0.008) 
-0.003 
(0.010) 
-0.008 
(0.011) 
-0.012 
(0.012) 
-0.014 
(0.013) 
-0.021 
(0.015) 
  
-0.005 
(0.012) 
-0.017 
(0.014) 
-0.014 
(0.015) 
-0.020 
(0.016) 
-0.018 
(0.0188) 
-0.014 
(0.023) 

-0.007 
(0.007) 
-0.008 
(0.008) 
-0.019** 
(0.010) 
-0.017* 
(0.009) 
-0.013 
(0.011) 
0.002 
(0.015) 
  

-0.010 
(0.009) 
-0.016 
(0.010) 
-0.001 
(0.012) 
-0.000 
(0.014) 
0.009 
(0.0159) 
0.000 
(0.016) 

0.002 
(0.007) 
0.001 
(0.008) 
0.007 
(0.012) 
0.005 
(0.013) 
0.001 
(0.014) 
0.002 
(0.015) 
  

0.001 
(0.013) 
0.004 
(0.015) 
-0.008 
(0.015) 
-0.019 
(0.017) 
-0.011 
(0.0167) 
-0.006 
(0.019) 

0.15 Mile Bandwidth 
  
Note. Standard errors in parentheses. *p<0.1, ** p<0.05, ***p<0.01. Models include walking distance to school, its 
interaction with transportation eligibility, and school-by-grade-by-year fixed effects. Standard errors are clustered at 
the school level. In the models with a quadratic spline, I also include a quadratic term of walking distance and its 
interaction with transportation eligibility. Students are considered to be transportation eligible if students live farther 
than 1.5 miles walking distance from school. Walking distances are calculated from the population weighted 
centroid of the student’s resident census block to their school’s address assuming average traffic using Here API. 
Econ. Dis. and EL are abbreviations for economically disadvantaged and English Learner respectively. In Michigan, 
students are considered economically disadvantaged if they qualify for free or reduced lunch, receive food (SNAP) 
or cash assistance (TANF), or they are homeless, migrant, or in foster care. 
 
 
 

 

 

153 

Table 28. Estimated Coefficients of Transportation Eligibility, All Bandwidths and Splines 
  

(1) 

(2) 

(3) 

(4) 

  
Linear Spline 
0.5 Mile Bandwidth 

 

0.4 Mile Bandwidth 

 

0.3 Mile Bandwidth 

 

0.25 Mile Bandwidth 

 

0.2 Mile Bandwidth 

 

0.15 Mile Bandwidth 
  
Quadratic Spline 
0.5 Mile Bandwidth 

 

0.4 Mile Bandwidth 

 

0.3 Mile Bandwidth 

 

0.25 Mile Bandwidth 

 

0.2 Mile Bandwidth 

 

0.299* 
(0.160) 
0.362** 
(0.184) 
0.358 
(0.221) 
0.461* 
(0.252) 
0.312 
(0.269) 
0.528* 
(0.308) 

-0.016** 
(0.008) 
-0.017* 
(0.010) 
-0.015 
(0.012) 
-0.021 
(0.013) 
-0.017 
(0.014) 
-0.025 
(0.017) 

Attendance Rate  Chronically Absent  Math Score  ELA Score 

  

0.175 
(0.108) 
0.197* 
(0.119) 
0.306** 
(0.142) 
0.289* 
(0.159) 
0.363* 
(0.202) 
0.354 
(0.233) 

  

  

-0.014** 
(0.005) 
-0.015*** 
(0.006) 
-0.015** 
(0.007) 
-0.012 
(0.008) 
-0.014 
(0.010) 
-0.016 
(0.012) 

  

  

-0.005 
(0.009) 
-0.004 
(0.010) 
0.002 
(0.011) 
-0.001 
(0.012) 
-0.004 
(0.014) 
-0.008 
(0.018) 

  

  

0.005 
(0.009) 
0.009 
(0.011) 
0.017 
(0.013) 
0.021 
(0.014) 
0.024 
(0.015) 
0.031 
(0.020) 

  

-0.002 
(0.012) 
-0.005 
(0.014) 
-0.013 
(0.016) 
-0.009 
(0.019) 
-0.019 
(0.023) 
-0.026 
(0.028) 

0.016 
(0.015) 
0.018 
(0.016) 
0.015 
(0.017) 
0.020 
(0.021) 
0.012 
(0.023) 
-0.013 
(0.027) 

0.15 Mile Bandwidth 
  
Note. Standard errors in parentheses. *p<0.1, ** p<0.05, ***p<0.01. Models include walking distance to school, its 
interaction with transportation eligibility, student race, gender, economically disadvantaged status, and English 
Learner covariates and school-by-grade-by-year fixed effects. Standard errors are clustered at the school level. In the 
models with a quadratic spline, I also include a quadratic term of walking distance and its interaction with 
transportation eligibility. Students are considered to be transportation eligible if students live farther than 1.5 miles 
walking distance from school. Walking distances are calculated from the population weighted centroid of the 
student’s resident census block to their school’s address assuming average traffic using Here API. Students are 
considered chronically absent if they miss more than ten percent of possible days in the school year. Math and 
English Language Arts (ELA) test scores on the Michigan Educational Assessment Program, MEAP, or the 
Michigan Student Test of Educational Progress, M-STEP are standardized within grade, subject and year.  

 

154 

Table 29. Heterogeneous Effects by Economically Disadvantaged Status, All Bandwidths and Splines 
  

(3) 

(5) 

(7) 

(6) 

(1) 
(2) 
Attendance Rate 

(4) 
Chronically Absent 

Math Score 

(8) 

ELA Score 

Not Econ. Dis.  Econ. Dis.  Not Econ. Dis.  Econ. Dis.  Not Econ. Dis.  Econ. Dis.  Not Econ. Dis.  Econ. Dis. 

 

  
Linear Spline 
0.5 Mile Bandwidth 

 

0.4 Mile Bandwidth 

 

0.3 Mile Bandwidth 

 

0.25 Mile Bandwidth 

 

0.2 Mile Bandwidth 

 

0.15 Mile Bandwidth 
  
Quadratic Spline 
0.5 Mile Bandwidth 

 

0.4 Mile Bandwidth 

 

0.3 Mile Bandwidth 

 

0.25 Mile Bandwidth 

 

0.2 Mile Bandwidth 

 

  

-0.018 
(0.091) 
-0.011 
(0.105) 
0.075 
(0.121) 
0.077 
(0.133) 
0.217 
(0.152) 
0.274 
(0.173) 

0.074 
(0.141) 
0.162 
(0.153) 
0.142 
(0.172) 
0.258 
(0.187) 
0.165 
(0.218) 
0.421* 
(0.251) 

  

0.576** 
(0.259) 
0.661** 
(0.285) 
0.805** 
(0.378) 
0.764* 
(0.408) 
0.891 
(0.563) 
0.797 
(0.686) 

0.880** 
(0.364) 
0.852* 
(0.449) 
0.963* 
(0.581) 
1.133 
(0.695) 
0.939 
(0.722) 
1.048 
(0.900) 

  

-0.003 
(0.004) 
-0.003 
(0.005) 
-0.004 
(0.006) 
-0.001 
(0.007) 
-0.007 
(0.008) 
-0.009 
(0.009) 

-0.004 
(0.007) 
-0.006 
(0.008) 
-0.003 
(0.009) 
-0.010 
(0.009) 
-0.007 
(0.010) 
-0.016 
(0.012) 

  

-0.036*** 
(0.012) 
-0.040*** 
(0.014) 
-0.037** 
(0.018) 
-0.034 
(0.021) 
-0.035 
(0.028) 
-0.039 
(0.034) 

  
-0.043** 
(0.018) 
-0.040 
(0.024) 
-0.043 
(0.032) 
-0.050 
(0.036) 
-0.049 
(0.041) 
-0.053 
(0.054) 

  

-0.007 
(0.011) 
-0.010 
(0.012) 
-0.006 
(0.014) 
-0.012 
(0.014) 
-0.006 
(0.016) 
0.004 
(0.022) 

-0.010 
(0.015) 
-0.007 
(0.015) 
-0.011 
(0.018) 
0.006 
(0.022) 
0.003 
(0.026) 
-0.011 
(0.029) 

  

-0.004 
(0.016) 
0.006 
(0.017) 
0.020 
(0.020) 
0.020 
(0.023) 
0.006 
(0.026) 
-0.021 
(0.034) 

0.011 
(0.021) 
-0.003 
(0.024) 
-0.018 
(0.031) 
-0.028 
(0.034) 
-0.047 
(0.047) 
-0.030 
(0.068) 

  

0.005 
(0.011) 
0.004 
(0.013) 
0.010 
(0.015) 
0.016 
(0.015) 
0.016 
(0.018) 
0.026 
(0.025) 

0.010 
(0.016) 
0.008 
(0.017) 
0.009 
(0.020) 
0.012 
(0.025) 
0.001 
(0.027) 
-0.033 
(0.032) 

  

0.003 
(0.018) 
0.013 
(0.022) 
0.039 
(0.027) 
0.031 
(0.031) 
0.037 
(0.035) 
0.037 
(0.040) 

0.029 
(0.031) 
0.042 
(0.036) 
0.036 
(0.036) 
0.045 
(0.040) 
0.029 
(0.046) 
-0.003 
(0.060) 

  

  

  

  

  

  

  

0.15 Mile Bandwidth 
  
Note. Standard errors in parentheses. *p<0.1, ** p<0.05, ***p<0.01. Models include student race, gender, and English Learner covariates and school-by-grade-
by-year fixed effects. Standard errors are clustered at the school level. In the models with a quadratic spline, I also include a quadratic term of walking distance 
and its interaction with transportation eligibility. Students are considered to be transportation eligible if students live farther than 1.5 miles walking distance from 
school. Walking distances are calculated from the population weighted centroid of the student’s resident census block to their school’s address assuming average 
traffic using Here API. Students are considered chronically absent if they miss more than ten percent of possible days in the school year. Math and English 
Language Arts (ELA) test scores on the Michigan Educational Assessment Program, MEAP, or the Michigan Student Test of Educational Progress, M-STEP are 
standardized within grade, subject and year. Econ. Dis. is an abbreviation for economically disadvantaged. In Michigan, students are considered economically 
disadvantaged if they qualify for free or reduced lunch, receive food (SNAP) or cash assistance (TANF), or they are homeless, migrant, or in foster care. 

 

155 

Table 30. Estimated Effects of Transportation Eligibility, Non-Parametric Approach 
  

(1) 

(2) 

(4) 

(5) 

(3) 

(6) 

Attendance Rate 

Chronically Absent 

 
  
Conventional Estimate 

Robust and Bias-Corrected 
Estimate 

 

 

 

Full 

Sample 

  

0.413*** 
(0.157) 
0.408** 
(0.164) 

 

Not 
Econ. 
Dis. 
  

0.068 
(0.109) 
0.076 
(0.127) 

 

Econ. Dis. 

Full 

Sample 

0.992*** 
(0.360) 
1.002*** 
(0.361) 

-0.019*** 
(0.006) 
-0.021*** 
(0.007) 

  

 

  

 

Not 
Econ. 
Dis. 
  

-0.003 
(0.005) 
-0.004 
(0.006) 

 

Econ. Dis. 

-0.048*** 
(0.013) 
-0.052*** 
(0.015) 

  

 

380,909 
112,374 
0.486 
0.764 

142,661 
36,155 
0.483 
0.719 

380,909 
95,782 
0.411 
0.655 

238,248 
89,341 
0.572 
0.913 

238,248 
65,789 
0.415 
0.683 

N 
142,661 
Effective N 
39,871 
Bandwidth for Estimate 
0.530 
Bandwidth for Bias Correction 
0.864 
Note. Standard errors in parentheses. *p<0.1, ** p<0.05, ***p<0.01. Models include student race, gender, 
economically disadvantaged status, and English Learner covariates and district, grade, and year fixed effects. 
Standard errors are clustered at the school level. Bandwidths are in miles. Optimal bandwidths and bias-corrected 
and robust estimates are calculated using the procedures proposed by Calonico, Cattaneo, and Titiunik (2014) and 
Calonico, Cattaneo, and Farrell (2020). Students are considered to be transportation eligible if students live farther 
than 1.5 miles walking distance from school. Walking distances are calculated from the population weighted 
centroid of the student’s resident census block to their school’s address assuming average traffic using Here API. 
Students are considered chronically absent if they miss more than ten percent of possible days in the school year. 
Econ. Dis. is an abbreviation for economically disadvantaged. In Michigan, students are considered economically 
disadvantaged if they qualify for free or reduced lunch, receive food (SNAP) or cash assistance (TANF), or they are 
homeless, migrant, or in foster care. 
 
 

 

156 

Figure 12. Distribution of Observations without Mobile Home Park Census Block 
Figure 12A: 0.01 Mile Bin Size 

 

 

 

 

 

Figure 12B: 0.1 Mile Bin Size 

  

 

 

 

Note. Sample includes observations of students in the analytic sample living between zero and five miles from their 
attended school excluding the census block containing Independence Woods Mobile Home Park.  
 
 

 

 

 

157 

REFERENCES 

 

158 

Allensworth, E., & Easton, J. (2007). What matters for staying on-track and graduating in 

Chicago public high schools. Chicago, IL: University of Chicago Consortium on Chicago 
School Research. 

 
Aucejo, E., & Romano, T. (2016). Assessing the effect of school days and absences on test score 

performance. Economics of Education Review, 55, 70-87. 

 
Baker, B. (2014). Not Making the Grade: How Financial Penalties for School Absences Hurt 
Districts Serving Low-Income, Chronically Ill Kids. A Guide for State Policymakers. 
ChangeLab Solutions. Retrieved on Jan. 26, 2021 from: 
https://changelabsolutions.org/sites/default/files/School-
Financing_StatePolicymakers_FINAL_09302014.pdf  

 
Balfanz, R., & Byrnes, V. (2012). The importance of being in school: A report on absenteeism in 

the nation's public schools. The Education Digest, 78(2), 4. 

 
Barreca, A., Lindo, J., & Waddell, G. (2016). Heaping‐induced bias in regression‐discontinuity 

designs. Economic inquiry, 54(1), 268-293. 

 
Bauer, L., Liu, P., Schanzenbach, D., & Shambaugh, J. (2018). Reducing chronic absenteeism 

under the every student succeeds act. The Hamilton Project, Brookings Institute. 
Retrieved on Sept. 28, 2020 from: https://www.attendanceworks.org/wp-
content/uploads/2018/04/Hamilton_project_-
reducing_chronic_absenteeism_under_the_every_student_succeeds_act.pdf  

 
Bergal, J. (2015, June 16). School districts are billing parents for bus rides. Stateline, an 

initiative of Pew Charitable Trusts. Retrieved on Sept. 28, 2020 from: 
https://www.pewtrusts.org/en/research-and-analysis/blogs/stateline/2015/6/16/school-
districts-are-billing-parents-for-bus-rides  

 
Blagg, K., Rosenboom, V., & Chingos, M. (2018). The Extra Mile: Time to School and Student 
Outcomes in Washington, D.C. Washington, DC: Urban Institute. Retrieved on Jul. 30, 
2020 from: https://www.urban.org/research/publication/extra-mile-time-schooland-
student-outcomes-washington-dc  

 
Burdick-Will, J., Stein, M., & Grigg, J. (2019). Danger on the way to school: exposure to violent 

crime, public transportation, and absenteeism. Sociological Science, 6, 118-142. 

 
Calonico, S., Cattaneo, M., & Farrell, M. (2020). Optimal bandwidth choice for robust bias-

corrected inference in regression discontinuity designs. The Econometrics Journal, 23(2), 
192-210. 

REFERENCES 

 
 
 

 

 

159 

Calonico, S., Cattaneo, M., & Titiunik, R. (2014). Robust nonparametric confidence intervals for 

regression‐discontinuity designs. Econometrica, 82(6), 2295-2326. 

 
Center for Educational Performance and Information. (2020). MI School Data Student Counts: 

Attendance. Retrieved on Sept. 28, 2020 from: 
https://www.mischooldata.org/DistrictSchoolProfiles2/StudentInformation/StudentCount
s/Attendance2.aspx  

 
Chang, H., & Romero, M. (2008). Present, Engaged, and Accounted for: The Critical Importance 

of Addressing Chronic Absence in the Early Grades. Report. National Center for 
Children in Poverty. 

 
Cordes, S., Leardo, M., Rick, C., & Schwartz, A. (2019). Can school buses drive down (chronic) 

absenteeism. Absent from school: Understanding and addressing student absenteeism, 
121-136. 

 
Cornwall, G. (2018, May 1). How lack of access to transportation segregates schools. Forbes. 

Retrieved on Sept. 28, 2020 from: 
https://www.forbes.com/sites/gailcornwall/2018/05/01/why-tech-is-prepping-to-overhaul-
school-transportation/#6ffe5512588a  

 
Dougherty, S. (2018). How Measurement and Modeling of Attendance Matter to Assessing 

Dimensions of Inequality. Journal of Education for Students Placed at Risk 
(JESPAR), 23(1-2), 9-23. 

 
Edwards, D. (in press). Over the river and through the woods: The role of distance in 

participation in rural school choice. Journal of School Choice. 

 
Federal Highway Administration. (2019). Children’s travel to school: 2017 national household 

travel survey. U.S. Department of Transportation. Retrieved on Sept. 28, 2020 from: 
https://nhts.ornl.gov/assets/FHWA_NHTS_%20Brief_Traveltoschool_032519.pdf 

 
Gershenson, S., Jacknowitz, A., & Brannegan, A. (2017). Are student absences worth the worry 

in US primary schools?. Education Finance and Policy, 12(2), 137-165. 

 
Gottfried, M. (2019). Chronic absenteeism in the classroom context: Effects on 

achievement. Urban Education, 54(1), 3-34. 

 
Gottfried, M. (2017). Linking getting to school with going to school. Educational Evaluation 

and Policy Analysis, 39(4), 571-592. 

 
Gottfried, M. (2014a). Can neighbor attributes predict school absences?. Urban 

Education, 49(2), 216-250. 

 

 

160 

Gottfried, M. (2014b). Chronic absenteeism and its effects on students’ academic and 

socioemotional outcomes. Journal of Education for Students Placed at Risk 
(JESPAR), 19(2), 53-75. 

 
Gottfried, M. (2011). The detrimental effects of missing school: Evidence from urban 

siblings. American Journal of Education, 117(2), 147-182. 

 
Gottfried, M. (2010). Evaluating the relationship between student attendance and achievement in 

urban elementary and middle schools: An instrumental variables approach. American 
Educational Research Journal, 47(2), 434-465. 

 
Gottfried, M. (2009). Excused versus unexcused: How student absences in elementary school 
affect academic achievement. Educational Evaluation and Policy Analysis, 31(4), 392-
415. 

 
Kirksey, J. (2019). Academic harms of missing high school and the accuracy of current policy 

thresholds: Analysis of preregistered administrative data from a California school 
district. AERA Open, 5(3). 

 
Lenhoff, S., & Pogodzinski, B. (2018). School organizational effectiveness and chronic 

absenteeism: Implications for accountability. Journal of Education for Students Placed at 
Risk (JESPAR), 23(1-2), 153-169. 

 
McCrary, J. (2008). Manipulation of the running variable in the regression discontinuity design: 

A density test. Journal of econometrics, 142(2), 698-714. 

 
McDonald, N., & Howlett, M. (2007). Funding for pupil transportation: framework for 

analysis. Transportation research record, 2009(1), 98-103. 

 
Meng, Y., Babey, S., & Wolstein, J. (2012). Asthma-related school absenteeism and school 
concentration of low-income students in California. Preventing chronic disease, 9. 

 
MHVillage Inc. (2021). Independence Woods. Retrieved on Jan. 26, 2021 from: 

https://www.mhvillage.com/parks/591  

 
Michigan Department of Education. (2017). Charter Schools- Questions and Answers. Retrieved 

on Jan. 26, 2021 from: https://www.michigan.gov/documents/PSAQA_54517_7.pdf   

 
MI Sec. 380.1321. Transportation for pupils; requirements; payment. Retrieved on Sept. 28, 

2020 from: http://legislature.mi.gov/doc.aspx?mcl-380-1321  

 
MI Sec. 388.1674. School bus driver safety instruction; cost of instruction and driver 

compensation; nonspecial education auxiliary services transportation; inspection costs. 
Retrieved on Feb. 12, 2021 from: 
http://www.legislature.mi.gov/(S(4e0vpo0wmue3nsobxqyhzzwu))/mileg.aspx?page=get
Object&objectName=mcl-388-1674 

 

161 

 
MI Sec. 388.1701.Eligibility to receive state aid. Retrieved on Feb. 12, 2021 from: 

http://www.legislature.mi.gov/(S(xse2efdmzydgfz0ybkfzlgiz))/mileg.aspx?page=getobjec
t&objectname=mcl-388-1701&query=on&highlight=attendance#5 

  
Morrissey, T., Hutchison, L., & Winsler, A. (2014). Family income, school attendance, and 
academic achievement in elementary school. Developmental psychology, 50(3), 741. 

 
National Center for Education Statistics. (2019). Digest of Education Statistics, 2018 (NCES 

2020-009), Table 236.90. Retrieved on Sept. 28, 2020 from: 
https://nces.ed.gov/programs/digest/d18/tables/dt18_236.90.asp  

 
Ready, D. (2010). Socioeconomic disadvantage, school attendance, and early cognitive 

development: The differential effects of school exposure. Sociology of Education, 83(4), 
271-286. 

 
Sampson, R., Raudenbush, S., & Earls, F. (1997). Neighborhoods and violent crime: A 

multilevel study of collective efficacy. Science, 277(5328), 918-924. 

 
Sattin-Bajaj, C. (2018). It’s Hard to Separate Choice from Transportation. Washington, DC: 

Urban Institute. Retrieved on Oct. 27, 2020 from: 
https://www.urban.org/sites/default/files/publication/99246/school_transportation_policy
_in_practice.pdf  

 
Singer, J., Cook, W., Lenhoff, S., & Pogodzinski, B. (2019). Detroit’s uniquely challenging 

context for student attendance. Detroit: Detroit Education Research Partnership, Wayne 
State University. Retrieved on Sept. 28, 2020 from: 
https://education.wayne.edu/detroit_ed_research/uniquely_challenging_context_report.pd
f  

 
Stein, M., & Grigg, J. (2019). Missing bus, missing school: Establishing the relationship between 

public transit use and student absenteeism. American Educational Research 
Journal, 56(5), 1834-1860. 

 
Urban Institute Student Transportation Working Group. (2018). The road to school: How far 

students travel to school in Denver, Detroit, New Orleans, New York City, and 
Washington, DC. Washington, DC: Urban Institute. Retrieved on Feb. 24, 2020 from: 
https://greaterdc.urban.org/sites/default/files/publication/97151/the_road_to_school_7.pd
f  

 
U.S. Department of Education (2018). Education for homeless children and youth program non-

regulatory guidance. Retrieved on Sept. 28, 2020 from: https://nche.ed.gov/mckinney-
vento/  

 

 

162 

U.S. Department of Education (2009). Questions and answers on serving children with 

disabilities eligible for transportation. Retrieved on Sept. 28, 2020 from: 
https://sites.ed.gov/idea/files/OMB_08-0101_Transportation-11-4-09_FINAL.pdf  

 
 Wolin, S., & Bennett, L. (1984). Family rituals. Family process, 23(3), 401-420. 
 

 

163