University of Southern California Dissertations and Theses

Essays on urban and real estate economics

(USC Thesis Other)

Essays on urban and real estate economics

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Essays on Urban and Real Estate Economics
by
Eunjee Kwon
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
May 2021
Copyright 2021 Eunjee Kwon
Acknowledgments
This dissertation summarizes my journey to be an urban and real estate economist, which
was only possible with numerous people’s support, guidance, and love. The word "thank
you" is not enough to express my gratitude to my advisors, mentors, friends, colleagues, and
family. Albeit hardship, my years at USC have been cheerful and beautiful, thanks to my
people.
Above all, it is a genuine pleasure to express my deep sense of thanks and gratitude to my
advisors, Professor Matthew Kahn, Richard Green, and Paulina Oliva, who took a chance
on me. They are not mere academic advisors to me. They are my life mentors, role models,
and great friends. Prof. Kahn’s overwhelming attitude to help me had been responsible for
my dissertation. He trained me in a way that none of the advisors could do. Prof. Green
discovered the potential in my research when I doubted it. His enormous support and guid-
ance kept me moving, and with that, I am now where I am. Prof. Oliva completed me as
a researcher and became my role model. She grew me with sharp feedback on my research
with a warm heart.
I have heavily benefited from conversations with faculty from the University of South-
ern California. I want to give massive appreciation to prof. Jorge De la Roca and Andrii
Parkhomenko, who selflessly invested hours discussing my ideas and guided me. Special
thanks to prof. Vittorio Bassi and Jeff Weaver, who vigorously led applied microeconomics
groupsintheeconomicsdepartmentandsetgreatexamplesasjuniorfaculty. Finally, Iwould
like to acknowledge Prof Jeff Nugent, John Strauss, Roger Moon, Geert Ridder, Jinkook Lee,
Manochehr Rashidian, Michael Magill, and Joel David, for training me as an economist.
ii
Six years at USC grew me not only as a researcher, but also as a better human being.
Six years at USC gave me so many friends who made Los Angeles home. I want to thank
my cohorts, Andreas, Bada, Qin, Youngmin, Jeehyun, Jason, Yinan, Yiwei, Lidan, Yimeng,
Yinqi, Kanika, and Sherly. We have gone through all the obstacles together since our joining
in 2015 August. I was also fortunate to meet two brilliant postdocs, Kilian and CK, as
my friends and mentors. I got great support from fellow Ph.D. students, such as Rashad,
Brian, Usman, Jake, Rachel, Amy, Yue, Ruozi, Rajat, Tushar, Juan, Dawoon, Seungwoo,
Bada, Chris, Gene, Amanda, Tao, Bardia, Jay, Yi-Ju, Hayun, Minsoo, and Greg, who I am
truly indebted. Finally, I am very fortunate to meet Yoon and Jeongwhan, who gave me
tremendous supports at my most challenging time.
Without exceptional supports from the staff members at the USC, this dissertation would
not have existed. For every significant decision that I made, I heavily relied on Young’s wis-
dom and heart. Having Alex in the department enormously benefited all of us, including
myself. I need to thank Irma, Nina, and Anna, for their administrative supports with a
warm heart.
I should not forget to mention that I was fortunate to spend my 2019 Fall at the Uni-
versity of Barcelona, thanks to the generous invite by professor Elisabet Viladecans-Marsal.
During my visit, I met terrific researchers and friends, including Pierre, Makdaliama, Ro-
drigo, Filippo, Kinga, and Alessio, who shared unforgettable memories with.
I also want to take this opportunity to thank USC Lusk Real Estate Center, Urban Eco-
nomics Association, American Real Estate and Urban Economics Association, USC Korean
Studies Institute, and Haynes Foundation for financially supporting my research.
LA Sarang Church community played a significant role and kept me sane. Pastor Kang,
Pastor Jang, and Pastor Hwang grew my faith in Jesus through their prayers and time. I
send my sincere gratitude to my church friends, including Yeun, Haeun, Kyounghee, Hy-
iii
oungsoon, Jaechul, Hankook, David, for being with me at my lowest moments.
Finally, I am grateful to have unconditional support from my loving family. I always
believe that God, who knows me better than I do, sent this family to me. Whenever I need
wisdom, support, and love, my father, who is my superman, has always been there for me.
He has been my motivation thanks to his warm heart, diligence, and humbleness. My mom
has been with me in every step, with her tears and ceaseless prayers. I am very proud of my
mom, who taught me to be happy and have a positive mindset, even at the darkest moment.
My best buddy, my sister, understands me more than anyone else. I am fortunate to have
her as a life-long supporter and friend.
Among all, I glorify the name of Jesus Christ, my savior, and Lord, who gift the oppor-
tunities to meet these people, and who always provides more than I pray. By the grace of
God, I am what I am.
"Now to him who is able to do immeasurably more than all we ask or imagine, according to
his power that is at work within us" - Ephesians 3:20 NIV
iv
Contents
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv
1 Why Do Improvements in Transportation Infrastructure Reduce the Gen-
der Gap in South Korea? 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.1 Economic Geography of South Korea . . . . . . . . . . . . . . . . . . 7
1.2.2 Gender Gap in Labor Market . . . . . . . . . . . . . . . . . . . . . . 8
1.2.3 Why Women Sort into Different Sectors from Men? . . . . . . . . . . 9
1.2.4 High-Speed Rail Expansion . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.1 Location Choice of Households . . . . . . . . . . . . . . . . . . . . . . 13
1.3.2 Firm: Gender-Segmented Intermediate Good Industries . . . . . . . . 14
1.3.3 Housing Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.4 Equilibrium: Gender Gap in Labor Market . . . . . . . . . . . . . . . 16
1.3.5 Prediction from the Model . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3.6 Comparative Statics: Impact of HSR . . . . . . . . . . . . . . . . . . 17
1.4 Road Map: From Model to Empirics . . . . . . . . . . . . . . . . . . . . . . 18
1.5 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.6 Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
v
1.6.1 OLS Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.6.2 Instrumental Variable . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.7 Main Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.7.1 Effect of HSR construction on Local Employment and Population . . 24
1.7.2 Impact of HSR on Employment and Wages by Genders . . . . . . . . 25
1.8 Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.8.1 Predictions from the Model . . . . . . . . . . . . . . . . . . . . . . . 26
1.8.2 Impact of HSR Across Industries with Different Gender Intensity . . 27
1.8.3 Impact of HSR on Endogenous Amenity . . . . . . . . . . . . . . . . 28
1.8.4 Impact of HSR on Migration Decisions of Singles and Couple . . . . . 28
1.9 Quantifying the Mechanism: Structural Model . . . . . . . . . . . . . . . . . 32
1.10 Robustness Check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.10.1 Subset Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.10.2 Nevo and Rosen (2012)’s Identification with Imperfect Instruments . 34
1.11 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2 How Do Cities Change When We Work from Home? 65
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.2.1 Workers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.2.2 Firms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.2.3 Developers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.2.4 Externalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.2.5 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.3 Data and Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.4 Counterfactuals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2.4.1 Spatial Reallocation . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.4.2 Commuting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
2.4.3 Wages and Floorspace Prices . . . . . . . . . . . . . . . . . . . . . . . 80
2.4.4 Accounting for Counterfactual Changes . . . . . . . . . . . . . . . . . 82
vi
2.4.5 Role of Endogenous Productivities, Amenities, and Congestion . . . . 84
2.4.6 Welfare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3 The Effects of Ability Tracking on the Academic Performance in the Sec-
ondary School: Evidence from South Korea 89
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.2 Evolution of High school Allocation in South Korea . . . . . . . . . . . . . . 92
3.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.3.1 College Scholastic Ability Test . . . . . . . . . . . . . . . . . . . . . . 95
3.4 The Empirical Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.6 Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.6.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.6.2 Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.6.3 Teacher Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.6.4 Peer Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.6.5 Hours of Studying and Self Evaluation . . . . . . . . . . . . . . . . . 102
3.7 Robustness Check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.7.1 Test for Tiebout Migration . . . . . . . . . . . . . . . . . . . . . . . . 102
3.8 Discussion: Equity and Efficiency . . . . . . . . . . . . . . . . . . . . . . . . 103
3.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Bibliography 121
A Appendix to Chapter 1 129
A.1 Data Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
A.1.1 Census on Establishment . . . . . . . . . . . . . . . . . . . . . . . . . 129
A.1.2 Population and Housing Census . . . . . . . . . . . . . . . . . . . . . 129
A.1.3 Internal Migration Statistics . . . . . . . . . . . . . . . . . . . . . . . 130
A.1.4 Korean Labor and Income Panel Study, KLIPS . . . . . . . . . . . . 130
vii
A.2 Data Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
A.2.1 District-level Wage Index . . . . . . . . . . . . . . . . . . . . . . . . . 131
A.2.2 A Travel Time Matrix between Cities to Cities . . . . . . . . . . . . . 131
A.3 Regression on Migration Flow . . . . . . . . . . . . . . . . . . . . . . . . . . 134
B Appendix to Chapter 2 143
B.1 Data Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
B.1.1 Property Price Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
B.1.2 Wage Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
B.1.3 Commuting Time Data . . . . . . . . . . . . . . . . . . . . . . . . . . 146
B.1.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
B.2 Model Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
B.2.1 Floorspace Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
B.2.2 Factor Incomes and Transfers . . . . . . . . . . . . . . . . . . . . . . 151
B.2.3 Welfare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
B.3 Structural Residuals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
B.4 Additional Results of Counterfactual Experiments . . . . . . . . . . . . . . . 155
B.4.1 Land Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
B.4.2 Job access . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
B.4.3 Breakdown of residential and job changes by worker type . . . . . . . 157
B.5 Elasticity of Speed to Traffic Volume . . . . . . . . . . . . . . . . . . . . . . 160
B.6 Accounting for Spatial Variation in Outcomes . . . . . . . . . . . . . . . . . 163
C Appendix to Chapter 3 168
C.1 Summary Statistics for KEEP . . . . . . . . . . . . . . . . . . . . . . . . . . 168
C.2 CSAT Score and the Hourly Wage 10 Years Later . . . . . . . . . . . . . . . 170
viii
List of Tables
1.1 Descriptive Statistics of Demographics . . . . . . . . . . . . . . . . . . . . . 48
1.2 Model Prediction: Impact of HSR . . . . . . . . . . . . . . . . . . . . . . . . 49
1.3 Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
1.4 First Stage of Instrumental Variable Regression . . . . . . . . . . . . . . . . 51
1.5 Impact of HSR on Local Population, Employment and the number of Estab-
lishments (Two-way Fixed Effects Model) . . . . . . . . . . . . . . . . . . . . 52
1.6 Impact of HSR on Local Population, Employment and the number of Estab-
lishments (IV) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
1.7 Impact of HSR on Population, Employment and Employment Rate across
Gender (IV) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
1.8 Impact of HSR on Wages, and Gender Wage Gaps (IV) . . . . . . . . . . . . 55
1.9 Summaries of Reduced Form Findings . . . . . . . . . . . . . . . . . . . . . 56
1.10 Mechanism 1.(1). Effect of HSR on Sectoral Employment . . . . . . . . . . . 57
1.11 Mechanism 2. HSR’s Impact on Endogeneous Amenities in Non-core . . . . . 58
1.12 Effect of Travel Time between Origin and Destination Districts on Migration:
Single . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
1.13 Effect of Travel Time between Origin and Destination Districts on Migration:
Couple . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
1.14 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
1.15 Quantitative Decomposition of the Impact of HSR . . . . . . . . . . . . . . . 61
1.16 Robustness Check: Sub-sample Analysis, without Sejong Special City . . . . 62
1.17 Robustness Check: Sub-sample Analysis, without Districts in Big Cities . . . 63
ix
1.18 Robustness Check: Sub-sample Analysis, without Districts close to North
Korean Border . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.1 Breakdown of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.1 Summary Statistics: CSAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.2 Mixing Effects on College Enrollment Test Total Scores: CSAT . . . . . . . . 113
3.3 Mixing Effects on College Enrollment Test Subject Scores: CSAT . . . . . . 114
3.4 Tracking Effects on Teacher-Pupil Interaction . . . . . . . . . . . . . . . . . 115
3.5 Tracking Effects on Teacher’s Quality and Class Atmosphere . . . . . . . . . 116
3.6 Direct Peer Effects on Possible Mechanism . . . . . . . . . . . . . . . . . . . 117
3.7 Tracking Effects on Study Hours . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.8 The Low Types Overestimate Themselves Under Tracking . . . . . . . . . . 119
3.9 Evidence that high school enrollment is not affected by treatment, GOMS . . 120
A.1 Summary Statistics of Wage Data (Source: KLIPS) . . . . . . . . . . . . . . 131
A.2 History of the KTX Station Expansion . . . . . . . . . . . . . . . . . . . . . 137
A.3 The Inferred impact of HSR on Relative productivity and female labor par-
ticipation costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
A.4 The Inferred impact of HSR on Relative productivity and female labor par-
ticipation costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
A.5 Mechanism 1.(2). Effect of HSR on Sectoral Sex Ratio (Male-to-Female Ratio)140
A.6 Residential Space, Housing Ownership, and Housing expenditure (Family Sur-
vey, 2012) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
A.7 Changes in estimations according to the value of (%) . . . . . . . . . . . . 142
B.1 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
B.2 Number of Transactions by County and Property Type . . . . . . . . . . . . 144
B.3 Number of observations in each earnings bin . . . . . . . . . . . . . . . . . . 145
B.4 Descriptive statistics: the estimated tract-level earnings, by county . . . . . 146
B.5 Commuting time coverage, by distance . . . . . . . . . . . . . . . . . . . . . 147
B.6 Commuting time coverage, by N. of commuters . . . . . . . . . . . . . . . . 147
x
B.7 Commuting time bins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
B.8 Data overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
B.9 Accounting for counterfactual floorspace price changes . . . . . . . . . . . . . 165
B.10 Accounting for counterfactual employment changes . . . . . . . . . . . . . . 166
B.11 Accounting for counterfactual residence changes . . . . . . . . . . . . . . . . 166
C.1 Summary Statistics: KEEP . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
xi
List of Figures
1.1 Spatial Distribution of Population and Employment in South Korea . . . . . 38
1.2 Cross-Country Gender Wage Gap (year=2015) . . . . . . . . . . . . . . . . . 39
1.3 Gender Differences in Time Use across Years . . . . . . . . . . . . . . . . . . 40
1.4 Wage Differences between Men and Women across Years . . . . . . . . . . . 41
1.5 Spatial Distribution of Employment Sex Ratio . . . . . . . . . . . . . . . . 42
1.6 Life Cycle of Work and Marriage Status . . . . . . . . . . . . . . . . . . . . 43
1.7 Occupation Composition for Gender and Age Group . . . . . . . . . . . . . 44
1.8 KTX Network in South Korea . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.9 Annual Ridership across Different Transportation Modes . . . . . . . . . . . 46
1.10 KTX Network and Old Railroad Stations Constructed during Japanese Colo-
nial Era . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.1 Changes in residence, jobs, and real estate prices . . . . . . . . . . . . . . . . 79
2.2 Commuting, wages, and prices . . . . . . . . . . . . . . . . . . . . . . . . . . 81
2.3 Quantiles of Centrality and Counterfactual Reallocations . . . . . . . . . . . 82
3.1 Enrollments in Seoul National University (SNU) . . . . . . . . . . . . . . . 106
3.2 Kernel Distribution of Test Score: Treatment Group . . . . . . . . . . . . . . 107
3.3 Treated and Controls Region . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.4 Pre-Trend Analysis: Lowess Graph, CSAT . . . . . . . . . . . . . . . . . . . 109
3.5 Relationship Between Student’s Own Rank and Peers’ Rank . . . . . . . . . 110
3.6 Evidence that refutes Tiebout migration, CSAT . . . . . . . . . . . . . . . . 111
A.1 Population-weighted Centroid . . . . . . . . . . . . . . . . . . . . . . . . . . 135
xii
A.2 KTX Network Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
B.1 Structural residuals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
B.2 Land Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
B.3 Land Use Specialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
B.4 Access to jobs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
B.5 Residence changes for continuing commuters, old telecommuters, and new
telecommuters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
B.6 Job changes for continuing commuters, old telecommuters, and new telecom-
muters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
B.7 Commuting time and distance . . . . . . . . . . . . . . . . . . . . . . . . . . 161
B.8 House prices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
B.9 Commuter market access, wages, and land prices . . . . . . . . . . . . . . . . 162
B.10 Welfare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
B.11 Quantiles of Centrality and Initial Allocations . . . . . . . . . . . . . . . . . 164
xiii
Abstract
This dissertation studies the impact of place-based policy and structural changes on the
socio-demographic implications of the system of the cities.
Thefirstchapterinvestigateswhethertheexpansionofhigh-speedrail(HSR)helpsreduce
the gender gap in labor market outcomes in South Korea. Using an instrumental variable
strategy that leverages historical railroads constructed in Korea during the Japanese colo-
nial era, I demonstrate empirically that the gender gap in the South Korean labor market
decreased with the expansion of HSR.
The second chapter studies how the shape of our cities would change if there were a
permanent increase in working from home. We study this question using a quantitative
model of the Los Angeles metropolitan area featuring local agglomeration externalities and
endogenous traffic congestion. We find three important effects: (1) Jobs move to the core of
the city, while residents move to the periphery. (2) Traffic congestion eases and travel times
drop. (3) Average real estate prices fall, with declines in core locations and increases in the
periphery.
The third chapter studies the impact of ability tracking, which creates a homogeneous
learningenvironmentwithintheclassroom,bysegregatinghigh-performingandlow-performing
students into different groups. This paper exploits a unique policy change in South Korea
combined with comprehensive micro dataset to measure the causal consequences of ability
tracking, compared to the allocating students to mixed ability groups. We find that track-
ing benefits high performing students who study more and gain from a superior learning
environment, with the expense of the academic achievement of the low-performing students.
xiv
Chapter 1
Why Do Improvements in
Transportation Infrastructure Reduce
the Gender Gap in South Korea?
1.1 Introduction
The gender gap in the labor market narrows with economic development, yet the dis-
parities remain salient even in developed countries like the U.S., Japan, and South Korea
(Blau and Kahn (2017)). The labor force participation costs for women are generally higher,
mainly due to the more substantial burden of childcare (Maurer-Fazio et al. (2011)) or home
production (Greenwood et al. (2005)). As a result, women tend to seek jobs with more flex-
ible work arrangements and choose different industries and occupations than men (Goldin
and Katz (2016)).
The previous literature has shown that technological shocks can affect the gender gap in
labor market outcomes. Technological progress, such as electrification in household sectors
reduces the domestic the burden on women (Greenwood et al. (2016); Vidart (2020)) and
helps reduce the gender gap in labor market by encouraging them to join the labor force.
1
However, Aksoy et al. (2019) show that robotization unintentionally increases the gender
pay gap, as jobs that are male-dominant disproportionately benefit from robots, which ulti-
mately increases the gender disparities in the labor market.
This study extends the literature by investigating whether improvements in transporta-
tion infrastructure could have different labor market impacts on men and women—, a ques-
tion for which there is limited prior evidence. Specifically, this study investigates the impact
of South Korea’s high-speed rail (HSR
1
) on gender-specific local labor market outcomes. As
investment in such infrastructure involves considerable government spending,
2
understand-
ing its heterogeneous impacts across different demographic groups is necessary. Despite the
fast-growing literature on the impact of HSR,
3
there is only limited evidence of its impact
on gender disparities in labor market outcomes. This study fills this gap.
HSR is likely to affect men and women differently through both the labor demand and
supply channels. On the one hand, HSR can create shocks in local gender-specific labor
demand. As a transportation mode mainly for carrying passengers, HSR reduces the cost
of moving people, but it does not necessarily reduce the cost of moving goods. Naturally,
industries that rely more on the cost of moving people than other sectors would benefit more
from HSR (Lin (2017)). If, for example, women are disproportionately hired more in such
sectors, then the increases in the labor demand for female would occur, somewhat uninten-
1
HSR, a relatively recent innovation in intercity transit, has become common in both developed and
developing countries. Since the construction of the first HSR, Japan’s Shinkansen, in 1964, HSRs have been
built across the world, including in many European as well as East Asian countries. Discussion on HSR
construction is ongoing in the United States.
2
2 For example, for the HSR in California, which links Los Angeles–San Francisco–Las Vegas, the total
construction cost is estimated to be $7 billion. The total cost to construct HSR in South Korea, which links
Seoul to Busan and several other major cities, is $15 billion (2015).
3
The literature finds that HSR enhances interaction among scientists living in different cities (Dong et al.
(2020)) and enhances the spatial agglomeration of cities (Zheng and Kahn (2013); see Qin (2017) for Chinese
cases; for German examples, see Ahlfeldt and Feddersen (2018)). HSR also affects the location decisions
of firms (Charnoz et al. (2018) for French firms), and the impacts are heterogeneous across industries (Lin
(2017)). Furthermore, HSR expansion can affect workers’ decisions on where to live and where to work. For
example, in Germany,workers prefer jobs in smaller cities while residing in bigger towns (Heuermann and
Schmieder (2019)).
2
tionally, with HSR. The disproportional increases in labor demand for women would lead to
the decreases in gender gap in labor market outcomes. On the other hand, the labor supply
decisions of men and women could be affected differently. As economic activities are redis-
tributed with HSR, amenities are endogenously changed (Gorback (2020)). Improvement
in local amenities, such as in childcare facilities, would liberate women from the burden of
childcare and encourage them to join the labor force. Combining both demand- and supply-
side impacts, HSR could generate different labor market implications for men and women.
The massive construction of HSR in South Korea, together with the country’s distinc-
tive gender inequality and a heavy geographical concentration of economic activities, makes
South Korea an ideal laboratory setting to study the research question. The gender pay
gap of 34% in South Korea is much higher than the OECD average of 13.1%. Only about
45% of the women in Korea are employed, compared to 80% of men. Furthermore, men and
women sort into different industries especially after marriage, as women disproportionally
choose to work in local service sectors, that require less work experience and provide flexible
work arrangement. Finally, economic activities are concentrated in a few districts in the
Seoul metropolitan area (i.e., Seoul, Incheon, and Gyeonggi provinces). Korea Train eXpress
(KTX) was first introduced in 2004, connecting Seoul with areas all the way down to the
South edge of the South Korean territory. An interesting feature of KTX is that it is not
primarily used as a commuter train: most trips are made for visiting relatives and family,
for leisure and business, or for enjoying services in other cities.
Estimating the causal relationship between improvement in HSR and local economic
outcomes, such as in employment, population, and gender gap in labor market outcomes, is
complicated further by endogeneity concerns. For example, districts that expect to gain the
most from HSR are more likely to be favored by the South Korean government or to lobby
for HSR development more aggressively, thereby leading to the non-random placement of
3
HSR stations. To address this concern, I estimate a two-way fixed effects model using dis-
trict and year fixed effects to control for unobservable variables. I leverage the fixed effects
estimators with the existence of old railroad stations constructed during the Japanese inva-
sion (1894—1945) as an instrumental variable for HSR stations’ location. The identifying
assumption behind this instrumental variable is that using the existing stations and routes
might reduce HSR construction costs (i.e., relevance condition), and after about 100 years,
the old railroad stations built during the colonial era would not impact the growth of the
regional economic outcome (i.e., exclusion restriction).
My empirical findings using the old stations as instruments show that HSR reduces the
gender gap in labor market outcomes. HSR narrows the gender employment gap in both core
(i.e., districts in the Seoul metropolitan areas) and non-core areas (i.e., districts outside of
the Seoul metropolitan areas) and decreases the wage disparities between men and women
in the core areas. In addition, the expansion of HSR changes the distribution of people
and jobs, decentralizing both the population and employment: population and employment
increase with the KTX expansion in districts outside the core areas and decrease in the core
areas.
To understand the mechanisms underlying the decrease in gender gaps with the expan-
sion of HSR, I use a class of spatial general equilibrium models with the gender-segmented
labor market, following Chauvin (2017).
4
In equilibrium, the gender gap in labor market
outcomes is a function of gender-segmented industries’ relative productivity and women’s
labor participation costs. The construction of HSR changes these two aspects for each site.
The equilibrium condition allows us to decompose the mechanisms behind the HSR’s
4
“Households” in the model consist of a woman and a man who jointly decide where to live. Each member
decides whether to work based on their labor force participation costs, which are higher for women. Firms
hire women and men in different intermediate sectors, which are used to produce final goods.
4
impact on the gender employment and wage gaps into the labor demand channels (rela-
tive productivity) and the labor supply channels (the labor force participation costs). The
quantitative decomposition combining the structural estimation and the empirical findings
shows that overall, HSR increases the labor demand of female-intensive sectors and decreases
women’s labor participation costs. Specifically, in core areas, the labor demand impacts are
more significant than the labor supply impacts, whereas the labor supply impacts are more
salient in non-core areas.
Next, I empirically test the model predictions to understand the mechanisms underly-
ing the empirical findings. First, to test relative productivity increases in female-intensive
sectors, I examined how industries with different gender ratios have been affected. It is
shown that male-intensive sectors, such as transportation, public administration, and man-
ufacturing, are negatively affected by HSR. In contrast, female-intensive sectors such as
retail, education, medical services, and restaurants are positively affected by HSR expan-
sions. These empirical results are consistent with previous findings that HSR has different
impacts across sectors. I then test whether the amenity level
5
in non-core areas increases
with the HSR expansion. Focusing on the retail, education, restaurant, and medical service
sectors, I show that HSR increases the amenity level in non-core areas but not in core areas.
Additionally, HSR’s impact on the number of workers per child in the education sector shows
that education amenities in non-core areas increase significantly with HSR. This result im-
plies that HSR potentially contributes to decreased female labor force participation costs in
non-core areas by reducing the childcare burden. All these findings are consistent with the
model predictions.
The contribution of this study is twofold. First, this study is related to the literature on
how structural changes could have different labor market impacts on men and women. A
5
Following Diamond (2016), I define the local service amenity as the number of establishments per
resident.
5
growing body of research has investigated the effect of technological shocks on gender gaps in
labor market outcomes either from the supply side (Greenwood et al. (2016); Vidart (2020))
or the demand side (Aksoy et al., 2019). This study extends the literature by examining how
an improvement in transportation technology could impact both the demand and supply side
of the labor market. The demand channels this study finds are similar to those in Aksoy
et al. (2019): HSR reduces both the gender pay and employment gap as it increases the la-
bor demand in female-intensive sectors (e.g., local services) more than in industries that hire
men more intensively. This is also similar to how the famous Bartik shock (Bartik (1991))
works. On the other hand, women’s labor supply is promoted by improvements in local
amenities, particularly in education and childcare, which reduce women’s childcare burden.
Furthermore, this study contributes to the literature by focusing on improving transporta-
tion infrastructure as a technology shock, which few previous works document.
Second, the study is closely related to the literature on the impact of transportation
infrastructure.
6
Its heterogeneous effects across spaces,
7
as well as across different socio-
economic groups (Tsivanidis (2018)), have been a focal point in recent research. A growing
body of research in urban economics has documented that gender differences exist in loca-
tion choices and commuting decisions (Farré et al. (2020), Chauvin (2017), Rosenthal and
Strange (2012), Black et al. (2014), Liu and Su (2020a), and Le Barbanchon et al. (2020)). I
contribute to this rich literature by focusing on the impact of transportation infrastructure
on gender gaps in the labor market, which is a crucial yet under-researched aspect.
6
Improvement in transit systems increases the trade of goods across spaces and reduces interregional
price gaps (Duranton et al. (2014); Donaldson (2018); Donaldson and Hornbeck (2016)). Another strand
of literature finds that transportation reduces the cost of moving people and promotes interregional mi-
gration (Morten and Oliveira (2016)) or changes workers’ decisions on where to live and work (Baum-Snow
(2007); Tsivanidis (2018); Severen (2018)). Redding and Turner (2014) provides the theoretical and empirical
literature on the relationship between spatial distribution of economic activity and transportation costs.
7
The literature examines the differences both across and within the regions. For example, within the
city-level, Baum-Snow (2007) showed that when highways in the United States connect city centers and
suburbs, it decentralizes economic activities. Investigating the transportation infrastructure across cities,
Baum-Snow et al. (2020) find that regional highways in China increase economic output and population in
regional primates at the expense of hinterland prefectures in China.
6
Therestofthispaperisorganizedasfollows. Section1.2providesbackgroundinformation
on the HSR expansion in South Korea and its economic geography. Section 1.3 outlines the
model that guides the empirical predictions. Section 1.4 provides a road map which links the
model to empirical exercises. The data sets used and empirical methodologies in this study
are described in Sections 1.5 and 1.6, respectively. Section 1.7 presents the empirical results
of the impact of HSR on the gender wage and employment gap. Section 1.8 empirically
tests some of the driving mechanisms derived from the model, and Section 1.9 quantitatively
estimatesthemechanisms. Section1.10presentsrobustnesschecksandtheregressionresults,
and finally, Section 1.11 concludes.
1.2 Background
1.2.1 Economic Geography of South Korea
Hometo51millionpeople,SouthKoreaisknownfornotonlyitsfasteconomicgrowthbut
also its high urbanization rate and heavy geographical concentration of economic activities.
The urbanization rate was 81.7% in 2018, which is similar to that of developed countries
such as the United States, Canada, and France. Half of the South Korean population lives
in the Seoul metropolitan areas
8 9
. The population of Seoul proper in 2016 was estimated
at 10.29 million; however, the sprawling metropolitan area is much larger at 25.6 million. In
contrast, Busan, the second-largest city in South Korea, is much smaller, with a population
of 3.4 million. Figure 1.1 shows the spatial distribution of population and employment in
South Korea.
8
SeoulmetropolitanareasaredefinedasSeoulandthesurroundingareasincludingGyeonggiandIncheon
provinces.
9
With a population density of 16,425 per square kilometer (Statistics Korea, 2015), Seoul (the capital
of South Korea) ranks second in population density among urban areas of more than 5 million people.
Considering the average population density of South Korea at 528 persons per square kilometer, Seoul
metropolitan areas are disproportionately large.
7
Figure 1.1 about here
1.2.2 Gender Gap in Labor Market
The demographic structure and labor market situation in South Korea are remarkable.
The gender pay gap in South Korea is the highest among OECD countries. Female full-time
employees on average earn 37.2% less than a male employee as shown in Figure 1.2. In the
labor market, gender disparities still exist even after adjusting for human capital differences.
This means that the gender pay gap in South Korea likely reflects a lack of opportunity,
rather than a lack of ability, for women. In addition, the disparity may be a result of other
factors such as more substantial domestic burdens, discrimination, or social norms. Many
women stop working when they have children. In the 2000s, fewer young people got married,
and this demographic had fewer babies. The fertility rate has become close to 1 since the
early 2000s, and the marriage rate has halved compared to that of the 1980s.
Figure 1.2 about here
Figure 1.3 presents the gender differences in time use over the years (Korea Income and
Labor Panel Survey, 2000—2015 ). Female employment has seen slight increases, but the
gender gap has not narrowed significantly. Around 80% of men responded that “working
full-time” is their main activity, whereas only 40—50% of women answered so, despite the
slight increases over the years. Another 40—50% of women answered “doing domestic work”
as their main activity, and about 10% of women partly worked and did domestic work, which
has decreased over time. Domestic burdens do not seem to be a significant concern for men.
Less than 5% of men answered that domestic work was their primary activity with almost
no changes in this number over the years.
Figure 1.3 about here
InSouthKorea,genderwagegapshavebeenconsistentovertheyears,evenafteradjusting
forthehumancapital(education)andexperiencelevel. Figure1.4showsgenderdifferencesin
8
wages over the years (Korea Income and Labor Panel Survey, 2000—2015). The unadjusted
wage differences between men and women have remained at around 40—45% over the years.
Adjusted wage differences have become even more considerable and persistent at around
40—44%.
Figure 1.4 about here
Figure 1.5 shows the spatial distribution of sex ratio in employment, defined as male
employment over female employment. Overall, the core areas or big cities have a higher
male-to-female employment ratio in 2000. Changes in the sex ratio in Panel B show that
female employment increased more than male employment, especially in core areas.
Figure 1.5 about here
Table 1.1 presents distinctive demographic characteristics in core and non-core areas. In
Panel A., differences in labor force participation rate between men and women are signif-
icant. The share of working men does not differ between core and non-core groups, and
approximately 75.5% of men work. However, the labor force participation rate for women is
much lower than that for men, and even lower in the core areas (39.72%) than in non-core
areas (46.96%). Finally, for both women and men, a disproportionately high number of
singles live in core areas, which is consistent with the marriage market hypothesis.
In terms of gender disparities in labor market outcomes (Panel B), the employment gap
is slightly larger in core areas, whereas the wage gap is slightly higher in non-core areas.
However, overall, the gender gap in both intensive and extensive margins is not negligible.
Table 1.1 about here
1.2.3 Why Women Sort into Different Sectors from Men?
Understanding what causes women and men to be sorted into different sectors in South
Korea is a topic of interest. Figure 1.6 shows the life cycle of men and women in terms
9
of their marriage and working status. Men and women tend to show similar labor force
participation rates until their mid-twenties. However, in their mid-twenties, as women begin
getting married, they tend to exit the labor market due to the childcare and other domestic
burdens. Some women reenter the labor force in their mid-thirties, but the number of women
never reaches that of men.
Figure1.6 about here
Moreover, women tend to work in different sectors or occupations when they reenter
the labor market after marriage. Figure 1.7 compares occupation compositions of each age
group of men and women before their exit from the market (ages 25–34) and after reentry
(ages 35–44). Women and men at earlier life stages (ages 25–34) show a similar pattern in
the compositions of occupations; generally, they work in white-collar jobs such as managers,
professionals, or other office workers. However, after women’s return to the labor market at
the age of 35–45, a significant number of women work in sales and local services (Figure 1.7,
(b) and (d)) compared to that of men, which does not change much in terms of occupation
compositions. This shows that an inordinate number of women work in local service sectors
after their marriage as the jobs require relatively less work experience and provide more
flexibility.
Figure1.7 about here
This trend is related to the work of Goldin and Katz (2016), which shows that the gender
gap in the pharmacy sector decreased over time in the United States, as the jobs have less
penalty in the pay for part-time work and meet women’s demand for flexibility in working
hours.
1.2.4 High-Speed Rail Expansion
Korea Train eXpress (KTX), the HSR in South Korea, was first introduced in 2004 and
experienced massive expansion from 2010 to 2012. The construction plans for KTX were
10
first made in the 1980s to reduce both unequal development across spaces and the problems
that became acute in areas near Seoul, such as unaffordable housing problems and traffic
congestion.
The bullet train not only links Seoul and Busan but also connects small towns, which
were previously only poorly connected to big cities. The KTX expansion process is depicted
in Figure 1.8 and Table A.2. In 2004 (stage 1), the Gyeongbu-line, which connects Seoul
to the southeastern region, and the Honam-line, which connects Seoul to the southwestern
part of South Korea, were opened. In 2010–2012 (stage 2), the Jeonla-line began operation,
the Gyeongbu-line was extended from Daegu to Busan, and some stations in Gyeongjeon
opened. In 2012, KTX had covered 22% of the total territory and served 56% of the total
South Korean population (KOSIS).
Figure 1.8 about here
Table A.2 about here
KTX makes round-trips between cities feasible for single-day travel. Travel from Seoul to
Busan (417.4 km), two metropolitan areas located at opposite edges of South Korea, takes 6
hours by car or non-high-speed train, but only around 2 hours by KTX at the fastest speed.
The train is mainly used for transporting people rather than goods. More than 80% of KTX
passengers use KTX to visit friends or relatives or to go on business trips. Fewer than 1%
of the passengers use KTX for their daily commutes because of the expensive ticket price.
10
As observed in Figure 1.9 (a), KTX ridership has increased significantly since its opening.
KTX accounted for 3.7% of interregional ridership (buses, cars, KTX or non-KTX train) in
10
A one-way train ticket from Seoul to Busan is around $40. As the average household net-adjusted
disposable income per capita was USD 19,372 per year in 2015 (USD 1,614.33 per month), train tickets are
too expensive to be used for daily commutes.
11
2004 and reached 10% in 2015
11
, while ridership on other transportation modes (except for
domestic flights) has decreased over time. Considering the total passenger-km riderships in
Figure 1.9 (b), the increase for KTX is even more significant. Total passenger-km of KTX
outweighed that of non-KTX trains from 2006, which drives the increases in total passenger-
km from 2011. Combining 1.9 (a) and (b), in 2015, average travel distance per ticket =
15,000,000,000 (passenger-km) / 60,000,000 (n. tickets)= 250 km, which is certainly over
the average commute distance in South Korea.
Figure 1.9 about here
The direct competitors ofKTX are intercitybuses or existing train networks. However, in
South Korea, domestic flights are not considered significant interregional transits as airports
are located far from city centers. Therefore, flying is regarded as a less-efficient means of
transit.
12
. Indeed, the ridership on any domestic flights (disregarding flights to Jeju Island)
was almost stagnant during the sample period of 2000–2015, whereas ridership for non-KTX
trains or intercity buses decreased significantly over the sample period.
1.3 Model
This section provides a model for explaining how the expansion of HSR can affect the
gender gap in the labor market and the distribution of economic activities. Here, the stan-
dard spatial general equilibrium framework as Roback (1982) is utilized; with local wages,
housing rents, and amenities. The general equilibrium approach captures both the direct ef-
fects of the shock as well as endogenous adjustments of factor prices and quantities (Moretti
(2010)). Besides, I borrow the notion of the gender-segmented labor market and location
choice of a married couple framework from Chauvin (2017). I advance the idea by allowing
HSR to change the equilibrium outcomes through the changes in (1) productivity of different
11
Korea Transport Database of the Korea Transport Institute (KTDB).
12
Speedy trains and planes are generally competitive if travel time is less than 1,000 kilometers (621
miles). (Bloomberg, 2018)
12
intermediate goods, (2) amenities of the locations, and (3) labor force participation costs of
females.
In the model, a household, consisting of a wife and a husband, chooses where to live
(and they work in the same location). Each household member decides whether to work
or not, given their labor force participation costs and the wage level at each location. As
is well known in the literature, the cost of labor force participation is higher for women.
13
Women and men are hired by different local industries that produce intermediate goods.
These intermediate goods are aggregated locally with imperfect substitutions as a nationally
produced good. Finally, housing is supplied locally. A spatial equilibrium consists of a set of
local wage levels for each gender, housing prices, a mapping of households to cities, locations,
and sectors so that firms maximize profits, households maximize utility, and labor markets
in each gender-segmented sector and housing markets are clear.
1.3.1 Location Choice of Households
Indirect utility of a household i, living in j at t
14
is as follows:
U
jt
= +w
net
jt
r
jt
+A
jt
(1.1)
with total household earning (w
net
jt
=w
M;net
jt
+w
W;net
jt
, where w
G;net
jt
indicates local wage
level of each gender group G2fWomen(W );Men(M)g), rent (r
jt
), amenities (A
jt
), and
some constant .
Specifically, local wage of each gender group (w
G;net
jt
) is as follows:
13
The higher cost of labor force participation for women could be due to commuting, childcare, or home
production.
14
The indirect utility function is driven from a household’s Cobb-Douglas utility function. Household
collaboratively derive utility from the consumption of tradable good (C
jt
) priced one, housing (H
jt
) rented
at r
jt
, and local amenities (A
jt
).
13
w
G;net
jt
=
8
>
>
<
>
>
:
w
G
jt

G
ifw
G
jt
>
G
(working)
0 ifw
G
jt
<
G
(notworking)
(1.2)
with the cost of labor force participation,
W
>
M
, which makes men more likely to
work.
Following Chauvin (2019), men and women draw an exogenous and stochastic labor force
participation cost, after they move to a new region. The CDF of labor participation costs
follow a power law (F (
i
) = (

i

min
)

). Women’s labor participation cost CDF has higher
support: the support for men is
M
2 (1;
max
) and for women is
W
2 (1 +T
jt
;
max
), for
T
jt
> 0. The supports for men and women are determined locally. (Wo)men will work if

i
< W
Gjt
. This gives female labor supply function of N
Wjt
= N
jt
(
W
Wjt
1+T
)

, and male labor
supply function of N
Mjt
=N
jt
(W
Mjt
)

.
Wage gap from female and male labor supply functions is:
W
M
jt
W
W
jt
=
1
(1 +T
jt
)
(
N
Mjt
N
Wjt
)
1

(1.3)
1.3.2 Firm: Gender-Segmented Intermediate Good Industries
In the gender-segmented labor market, women and men (G2fWomen(W );Men(M)g)
are hired in different intermediate good sectors:
Y
Gjt
=
Gjt
N

Gjt
K
1
jt
Intermediategoodsarecombinedwithconstantelasticityofsubstitutionintoanationally-
14
traded final good priced:
Y
jt
=[(Y
Mjt
)

+ (Y
Wjt
)

]
1

Y
jt
=[(
Mjt
N

Mjt
)

+ (
Wjt
N

Wjt
)

]
1

K
1
jt
This gives the following labor demand equations in each sector:
W
G
jt
=
Gjt
N
1
Gjt
L
1

jt
whereL
jt
= [(
Mjt
N

Mjt
)

+ (
Wjt
N

Wjt
)

]
1

and the local gender wage gap is:
W
M
jt
W
W
jt
= (

Mjt

Wjt
)

(
N
Mjt
N
Wjt
)
1
(1.4)
1.3.3 Housing Market
The housing supply closely follows that in Glaeser et al. (2008). In every locationj, there
is an absentee landlord who buy the housing from developers and rent it to local residents
at r
jt
. Profits for the developers are given by:

jt
=
1
t=0
r
jt
(1 +i
t
)
CC
jt
(1.5)
where i
t
is the national interest rate and CC
jt
are the local construction costs. The free
entry and zero-profit conditions give the local housing supply.
The housing demand function is from the household’s Cobb-Douglas utility function,
which is aggregated at the number of households located in location j at time t (N
jt
). In
equilibrium, the local housing market clears for every location j at time t.
15
1.3.4 Equilibrium: Gender Gap in Labor Market
Equating labor supply (equation 1.3) and labor demand (equation 1.4) functions, in equi-
librium, the following equations regarding the gender gap in labor market outcome hold.
The gender employment gap (male-to-female ratio,
N
M
jt
N
W
jt
) becomes
N
M
jt
N
W
jt
= (1 +T
jt
)

(

Mjt

Wjt
)

(1.6)
The gender wage gap (male-to-female ratio,
W
M
jt
W
W
jt
) becomes
W
M
jt
W
W
jt
= (1 +T
jt
)
(1)

(

Mjt

Wjt
)

(1.7)
where 1 +T
jt
is labor force participation cost for women in location j at time t,

Mjt

Wjt
is
relative productivity between male- (
Mjt
) and female-intensive sector (
Wjt
). Given the
range of parameters; the labor share in Cobb-Douglas production of the intermediate good
(), the elasticity of substitution of final good production (), and the CDF parameter of
the female labor force participation cost function (),
=
1
1

+1
> 0 and ( 1)
< 0.
The two equilibrium equations in the gender gap provide intuitive interpretations. Note
that in Equation 1.6 and 1.7, parameters in blue is positive value, and in red is negative
value. In Equation 1.6, the gender employment gap increases with labor force participation
cost for women. As women participate in the labor force only if the market wage is higher
than their costs, higher labor force participation cost lowers women’s labor participation
rate, which leads to the increases in the gender employment gap. On the other hand, as
in Equation 1.7, the higher cost would increase the overall wage level of women who join
the labor force, which decreases the gender wage gaps. Meanwhile, the increase in relative
productivity between male- and female-intensive sector would worsen the gender gap in both
16
employment and wage, as the labor demand for men increases more than that for women.
Equations 1.6 and 1.7 give us the following predictions regarding the gender employment
and employment gap.
15
1.3.5 Prediction from the Model
Table 1.2 summarizes the changes in the gender employment and wage gap, when rel-
ative productivity and female labor participation cost change simultaneously due to HSR
expansion. For example, if both the relative productivity of the male-intensive sector and the
female labor force participation cost decrease, the gender gap in employment would decrease.
Simultaneously, the changes in the gender wage gap would be indeterminate.
Table 1.2 about here
1.3.6 Comparative Statics: Impact of HSR
With HSR, location-specific amenities, productivities, and the labor force participation
cost for women could be affected:
1. Changes local amenities is expressed as follows:
U
ijt
= +w
net
jt
r
jt
+
jt
2. Changes in relative productivity of intermediate goods is expressed as follows:
W
Mjt
W
Wjt
= (

Mjt

Wjt
)

(
N
Mjt
N
Wjt
)
1
15
Population in area j at time t is N
jt
can be expressed with respect to all the exogenous parameters
and local amenities (
jt
), relative productivity of intermediate good sectors (

Mjt

Wjt
), and labor participation
costs of women (T
jt
).
17
3. Changes in labor force participation cost for women is expressed as follows:
W
M
jt
W
W
jt
=
1
(1 +T
jt
)
(
N
Mjt
N
Wjt
)
1

1.4 Road Map: From Model to Empirics
Taking derivatives of Equation 1.6 and 1.7 with respect to HSR, the following equations
are obtained.
For the changes in the gender employment gap:
@(
N
M
jt
N
W
jt
)
@HSR
=

@(1 +T
jt
)
@HSR
+

@(

Mjt

Wjt
)
@HSR
(1.8)
and for the gender wage gap:
@(
W
M
jt
W
W
jt
)
@HSR
= ( 1)

@(1 +T
jt
)
@HSR
+

@(

Mjt

Wjt
)
@HSR
(1.9)
where
=
1
1

+1
and equation in blue is observable, and in orange is unobservable.
Empirically, changesinlocalemploymentgap(
@(
N
M
jt
N
W
jt
)
@HSR
)orlocalwagegap(
@(
W
M
jt
W
W
jt
)
@HSR
)inEqua-
tion 1.8 and 1.9 can be directly observed. However, the relative productivity between male-
and female-intensive sectors (
@(

Mjt

Wjt
)
@HSR
) or labor force participation cost for women (
@(1+T
jt
)
@HSR
)
are not easily observed from the data. Additionally, the range of the three parameters (;,
and ) are determined in the model.
Given the availability of data and parameters, the model predictions to the empirics were
linked as follows:
18
1. Section 1.7 empirically estimates the impact of HSR on the local gender gaps in the
labor market outcomes (
@(
N
M
jt
N
W
jt
)
@HSR
,
@(
W
M
jt
W
W
jt
)
@HSR
), which are directly observed from the data.
2. Section 1.8 infers the impact of HSR on the relative productivity and female labor
force participation costs based on the results of Section 1.7 and the model predictions
in Table 1.2.
3. Section 1.8 continues to investigate whether the inferred predictions can be confirmed
with suggestive evidence from the data.
4. Finally, Section 1.9 structurally decomposes the impact of HSR on the relative produc-
tivity (
@(

Mjt

Wjt
)
@HSR
) and labor force participation cost for women (
@(1+T
jt
)
@HSR
), based on the
calibrations and the results in Section 1.7.
1.5 Data
Thissectiondescribesthedatasetsconsideredinanalyzingthecausalimpactsofthemas-
sive bullet train expansion on South Korea’s economic geography. To construct a geocoded
16-year panel data set at the district-level, the Census on Establishment, the Population and
Housing Census, and the Internal Migration Statistics collected by Statistics Korea are uti-
lized. These administrative data sets are combined into a time-varying travel-time matrix,
which is constructed using the algorithm explained in the Appendix. Note that the mi-
crolevel data for the Census on Establishment and the Population and Housing Census are
exclusively provided through the subscription to the Micro Data Integrated System (MDIS),
operated by Statistics Korea. Throughout this study , regions are defined at district-level.
In 2010, there were 228 districts consisting of 15 provinces
16
. By combining the data sets
explained above, a (nearly) balanced panel data set at the district-level (with the number of
16
As our main interests are railroad transportation, districts on Jeju Island and Ulleng Island were ex-
cluded in this paper.
19
districts equal to 228) from 2000 to 2015 (t=16) was constructed. The details of each data
set are described in Appendix ??. Summary statistics are presented in Table 1.3.
Table 1.3 about here
1.6 Empirical Strategy
The following hypotheses from the model are empirically tested in this section. Does the
construction of HSR affect the gender gaps in the labor market outcomes, such as wage levels
and employment? Does the construction of HSR shift population and employment from core
to non-core areas or does it induce more concentration of economic activities in core areas?
1.6.1 OLS Regression
In a districti within a provinces at timet, the econometric model estimating the impact
of the KTX expansion on the outcome variable of interest is:
log(y
it
) = +Treat
i
Post
t
+X
it
+
st
+f
i
+
t
+
it
; (1.10)
where y
it
is a dependent variable including the gender gap in labor market outcomes,
employment, population, or wage-level of a county i of a province s at time t. Treat
i
Post
t
is the treatment dummy equal to one, if the centroid of a district i is within 15km of any
KTX station at time t. A set of control variables (X
it
), district fixed effect (f
i
), year fixed
effect (
t
) and province-time varying fixed effect (
st
) are included. can be interpreted as
the percentage change in the outcome of interest with new construction of a KTX station in
a district. Here, all the standard errors are clustered at district-level. Instead of using the
spatial center as centroids, this study calculates the population-weighted centroids of each
districts, based on the zip-code level population data. Refer to Appendix for the graphical
representation of the calculation.
20
To identify the heterogeneous impacts between core and non-core areas, consider the
following specifications:
log(y
it
) = +
c
Treat
i
Post
t
D
c
i
+
nc
Treat
i
Post
t
D
nc
i
+X
it
+
st
+f
i
+
t
+
it
; (1.11)
where D
c
i
(D
nc
i
) = 1 if a district i is located in the core (non-core). In terms of the
institutional setting in South Korea, core areas are defined as districts located in the Seoul
metropolitan areas (Seoul, Gyeonggi, and Incheon provinces) and non-core areas are districts
in the rest of the provinces.
1.6.2 Instrumental Variable
Non-random Placement of HSR
Studying the causal effect of transportation requires careful inspection due to selection
issues as this is the nature of place-based policies. The central inference problem researchers
are facing in the transportation literature is that transportation is not assigned randomly but
is determined based on both observed or unobserved location characteristics and forecasts
of expected benefits (Redding and Turner (2014)).
In the context of this study, although the location of high-speed train stations in South
Korea was determined by the central government 14 years prior to the actual opening (Baek
and Park (2015)), possible endogeneity concerns remain. For example, districts that expect
to gain the most from HSR are more likely to lobby more aggressively to have the station in
their districts . This makes the location choice of HSR nonrandom. The unobserved factors
that could affect the actual timing of the opening can compromise the identification of the
regression. To address these issues, this study suggests the following instrumental variable
estimation, following the literature in urban economics.
21
Instrumental Variable: Old Railroad Stations in Japanese Colonial Rule
Existing literature has dealt with this endogeneity issue by using either planned (Baum-
Snow (2007); Duranton and Turner, 2012; Duranton et al. (2014)) or historical (Morten
and Oliveira (2016); Garcia-López et al. (2015); Tsivanidis (2018)) railroads, or least-cost
corridors (Faber (2014)) as instrumental variables. In this study, the historical railroad sta-
tions, constructed during Japanese colonial rule (before 1945) were used as the instrumental
variable for the HSR stations.
In1894,JapanassumedrailroadconstructionintheKoreanterritoryaspartoftheir“modernization
process.” ThefirstrailroadnetworkinKorea, Gyeongin-line, wasopenedin1899. Fromthen,
the Japanese railway ministry expanded railroad networks over the years. From 1904–1906,
the Gyeongbu-line and the Gyeongui-line were added to serve as military pathways to China
and Russia; in 1910, the Honam-lines were added to help transport natural resources and
agricultural products from South Korea to Japan.
Figure 1.10 about here
The assumption behind using the historical railroad stations as instrumental variables
for HSR is that the existing station platforms and routes might be more cost efficient, which
satisfies the relevance condition of IV(Pr(D = 1jZ = 0)6= Pr(D = 1jZ = 1)), but at
the same time they might have no impact on current economic situations in the counties
(exclusion restriction, E(jZ = 1) =E(jZ = 0)).
Table 1.4 shows that the relevance condition of the IV is satisfied. The first stage of the
regressions in Table 1.4 shows high F-statistics, which are above the typical rule of thumb
(F-statistic of more than 10, Stock and Yogo (2002)). The high F-statistics confirm that
the existence of the old railroad stations in districts predicts a higher probability of the
construction of HSR in the districts.
22
Table 1.4 about here
Ontheexclusionrestriction,oldrailroadstations,constructedbefore1945,intheJapanese
colonialperiod, wouldaffectthegrowthoftheeconomiesofcountiesintheearly2000s. Then,
trains were mainly used to transport resources from South Korea to Japan or as assets of
the Russian-Japanese war, rather than for supporting the economic development of the re-
gion. Even if one argues that the existence of old railroad stations initially determines the
economic situation of counties, the district fixed effects (f
i
) control for such an initial level
difference across counties. Here, the identification strategy comes from the condition that
the old railroad station might have an impact on the level of the outcome variables, but not
on their growth.
IV estimation
As the instrument variable is time-invariant (Z
j
), the long-difference between the final
year (2015) and the initial year (2000) of Equation 1.10 is computed as follows:
log(y
i
) =Treat
i
+X
i
+
s
+
j
(1.12)
Here, Treat
i
with Z
i
= 1 is used as the instrument variable, if a railroad station was
built in district i during the Japanese colonial era.
Finally,forcapturingheterogeneousimpactsacrosslocations,asinEquation1.11,treat
i
post
t
D
k
j
is instrumented by Z
i
D
k
i
, for k2fcore;noncoreg.
23
1.7 Main Results
1.7.1 Effect of HSR construction on Local Employment and Popu-
lation
The OLS regression results of Equations 1.10 and 1.11 are presented in Table 1.5. If
all the coefficients are correctly identified, each coefficient can be interpreted as the per-
centage change in the outcome variables when the HSR station is located within 15 km of
the centroid of districts. In columns (1)—(3) of the table, any aggregated treatment effects
of HSR on population, employment, or the number of establishments cannot be observed.
Conversely, once the heterogeneous treatment effects across core and non-core districts are
considered (columns (4)–(6)), we observe that HSR construction is associated with 7.3% and
3.9% increases in population and employment in non-core areas, respectively, and 5.6% (with
no statistical significance) and 13.6% decreases in population and employment in core areas,
respectively. The results show that HSR induces the redistribution across space by moving
economic activities from the core areas to non-core areas.
The OLS regression results of equation 1.10 and 1.11 are presented in table 1.5. If all
the coefficients are correctly identified, each coefficient can be interpreted as the percentage
change in the outcome variables when the HSR station is located within 15km of the centroid
of districts. In column (1)-(3) of Table 1.5, we do not see any aggregated treatment effects
of HSR on population, employment, or the number of establishments. In contrast, once we
consider the heterogeneous treatment effects across core and non-core districts (column (4)-
(6)), we see that HSR construction is associated with 7.3% and 3.9% increases in population
and employment in non-core areas respectively, and 5.6% (with no statistical significance)
and 13.6% decreases in population and employment in core areas respectively. The results
show that HSR induces the redistribution across space, by moving economic activities from
the core areas to the non-core areas.
24
Table 1.5 about here
Table 1.6 summarizes the IV estimations using the old railroad stations as instrumental
variables. The qualitative results of IV estimates are consistent with the OLS estimates in
Table 1.5, with magnitudes of two to three times and more statistical strength. Meanwhile,
areas with high economic concentrations lose population, employment, and numbers of es-
tablishments, whereas the less-developed provinces benefit from increased population and
jobs after HSR stations open.
Table 1.6 about here
We can infer the direction of the bias by comparing the OLS and the IV estimates.
Upward and downward biases in OLS estimates are shown in core and non-core areas, re-
spectively. These biases mean that HSR stations were constructed in the districts that are
expected to grow in core areas, whereas in non-core areas, the districts that are expected to
shrink were likely to get HSR stations.
1.7.2 Impact of HSR on Employment and Wages by Genders
Table 1.7 shows the heterogeneous treatment effects across genders in population and em-
ployment. The female and male population are affected in similar ways both quantitatively
and qualitatively. However, in terms of the sex ratio in employment (i.e., male employment
over female employment), female employment is observed to greatly increase in both core
and non-core areas with HSR expansions.
Table 1.7 about here
25
The effects of HSR on land price and wages are reported in Table 1.8. Land prices in
non-core areas increased by 5.8%, whereas those in core areas do not change with HSR ex-
pansion. Interestingly, an examination of the HSR’s differential wage impacts across gender
shows that the female wage level in core areas increases by approximately 20.3% in core
areas, which contributes to the decrease in the wage gap in these core areas, by 15.8%. Con-
versely, was no changes were observed in the wage level in non-core areas, both for women
and men, resulting in no changes in the wage gap in these non-core areas.
Table 1.8 about here
1.8 Mechanism
Table 1.9 summarizes the empirical findings of the impact of HSR are. HSR expansions
in South Korea reduce the concentration in core areas while increasing the population and
employment rate in non-core areas. Moreover, interesting gender differences for labor mar-
ket outcomes are found. HSR contributes to the decrease in gender earning disparity in
core areas. Moreover, the employment of women is higher in both core and non-core areas,
compared to that of men.
Table 1.9 about here
1.8.1 Predictions from the Model
Combining the empirical findings in Table 1.9 and the predictions from the model pre-
sented in Table 1.2, the impact of HSR on female labor force participation costs and relative
productivity between female- and male-intensive sector can be predicted in core and non-core
areas.
26
1. In the core areas, in order to have the decrease in the male-to-female wage gap, but no
changes in the gender gap in employment, the relative productivity of female intensive
sectors should increase (
@

Mjt

Wjt
@HSR
< 0). The impact on the labor cost participation cost is
indeterminate, which is an empirically testable question, (
@T
jt
@HSR
?0). See table A.3 for
proof.
2. In the non-core areas, in order to have no changes in the male to female wage gap,
but the increase in the employment rate of female, the relative productivity of female
intensivesectorsshouldincrease (
@

Mjt

Wjt
@HSR
< 0)andimpactonthelaborcostparticipation
cost should decrease with the expansion of HSR (
@T
jt
@HSR
> 0). See table A.4 for proof.
The following sets of regressions empirically test the propositions.
1.8.2 Impact of HSR Across Industries with Different Gender In-
tensity
Whether HSR increases the relative productivity of female-intensive sectors compared to
male-intensive sectors was verified. Table 1.10 presents the effect of HSR across different sec-
tors with different gender intensities in employment (i.e., male-to-female employment ratio).
The most male-intensive industry is transportation, with a ten-to-one male-to-female gender
ratio, whereas the most female-intensive sectors are the lodging and restaurant industries,
with a male-to-female ratio of 0.5. Interestingly, sectors with higher male workers intensity,
in general, show no effects on the growth in their employment, except for the construction in-
dustry. On the other hand, the increase in employment in non-core areas was shown in most
of the female-intensive sectors such as retail, education, medical, and restaurant industries.
Table 1.10 about here
Table A.5 investigates whether the changes in sex ratio within industries with HSR ex-
pansion are not driven by the changes in sex ratio. Sectors positively affected with HSR, such
27
as retail, education, medical service, and restaurant industries, do not show extreme changes
in male-to-female employment ratio with the HSR expansion, except for the education sector
in non-core areas. In summary, this confirms that HSR positively affects female-intensive
sectors compared to the male-intensive sectors, thus increasing labor demand for women.
1.8.3 Impact of HSR on Endogenous Amenity
Whether HRS has effects on the local amenity level was verified. Specifically, following
Diamond (2016), local amenity level is defined as the number of establishments per resident
of certain service industries. Table 1.11 shows the impacts of HSR on the local amenity level.
Here, the total establishments of retail, medical services, and education sectors per resident
in non-core areas are positively affected in non-core areas, with HSR expansion. Conversely,
the local amenities are not affected by HSR in the core areas.
Table 1.11 about here
Notably, theeducationamenitylevelincreasessignificantlyinnon-coreareas. Specifically,
whether workers working in the education sector per kids (under 13 years old, which is
elementary school age) changes with HSR has been verified in column (5) of Table 1.11.
Interestingly, an increase in workers per kid in the education sector is only observed in non-
core areas. This indirectly implies that childcare burden for women decreases with HSR,
which can potentially reduce the labor force participation costs of women.
1.8.4 Impact of HSR on Migration Decisions of Singles and Couple
Finally, whether the HSR changes the demographic compositions (single vs. couples) in
different areas was verified. The composition changes in demographics indirectly indicate
location characteristics, with the revealed preference argument. For example, if singles and
couples are segregated in different regions, this indicates that each region provides a location
28
benefit favored by different demographic groups.
The migration flow of individuals and households with different marital statuses was
examined. Population is a stock variable that impairs our understanding of the underlying
migration flows of individuals. Hence, the impact of the reduction in travel time between
origin (o) and destination (d) counties due to KTX networks on the number of migrants for
each county-pair, which provides a better idea on the movement of people. As reduced travel
time improves both physical and emotional accessibility between districts, individuals more
easily sort themselves into more favorable locations.
Data
The data set mainly used in this section is from the Internal Migration Statistics. From
the household level of 88,248,353 observations over 15 years (2001—2015), origin–destination
county-pair level data is aggregated, which consists of 849,600 origin–destination pairs. Ad-
ditionally, to observe the various patterns across the population, this study investigates
whether the migration patterns are affected differently by travel time among different mari-
tal status groups (singles vs. couples). The details of the data set are provided in Appendix
??.
Empirical Strategy: Gravity Model of Migration Flow
A gravity model guides us in analyzing the relationship between changes in travel time
and number of migrants. The model is often used to explain the commuting patterns (Heuer-
mann and Schmieder (2019)) as well as the migration patterns (Morten and Oliveira (2016))
between regions. From the model, each empirical specification is derived as follows. The
details of the model derivation are provided in Appendix A.3
Considering the potential endogeneity issue, the main regression equation is expressed
29
as
17
log(M
odt
) = +
log(TravelTime
odt
) +
od
+
ot
+
dt
+
odt
(1.13)
where, log(TravelTime
odt
) is the travel time between origin and destination counties at
time t,
od
is the origin-destination pair fixed effect,
ot
is the origin-year fixed effect, and

dt
is the destination-year fixed effect.
Fixed effects of the regression control for the potential endogeneity in the placement of
HSR, which are parsimonious enough to capture those unobserved heterogeneities.
od
cap-
turesthesystematictime-unvaryingrelationshipbetweenoriginandthedestinationcounties;
time-varying origin (
ot
) and destination (
dt
) fixed effects capture all potential unobserved
changes in origin and destination counties. For example, changes in inflow or outflow of
migrants, due to an economic downturn or boom of origin (or destination), can be controlled
by the time-varying fixed effects. Finally,
odt
is clustered at the origin–destination pair-level.
The specific interests in this study are whether the migration flow between core areas and
non-core areas are affected once HSR reduces travel time between the two areas. To capture
theheterogeneityintheelasticityofmigrationconsideringthetraveltimereduction,thethree
interaction terms are placed in the regression: (origin: core; destination: non-core), (origin:
non-core; destination: core), and (both origin and destination are (non)-core). Thereafter,
the regression becomes
log(M
odt
) = +
log(TravelTime
odt
)[
SN
D
o:S;d:N
+
NS
D
o:N;d:S
+
within
D
o;d:within
]
+
od
+
ot
+
dt
+
odt
(1.14)
17
The same empirical specification is also used in Morten and Oliveira (2016).
30
The changes in the number of migrants in response to the changes in travel time (i.e.,
migration elasticity with respect to travel time changes) for each direction are follows:
•
SN
: Elasticity of migration from core to non-core districts
•
NS
: Elasticity of migration from non-core to core districts
•
within
: Elasticity of migration within core counties or non-core districts
Results
Elasticities of Migration with Respect to the Travel Time Reduction
The estimation results of Equation 1.14, for each demographic group are presented in
Tables 1.12 and 1.13.
18
Table 1.12 about here
Table 1.13 about here
As travel time is reduced by the bullet train, single households generally migrate to the
core areas (Table 1.12), whereas couples move to non-core areas (Table 1.13), leading to
demographic segregation. This result might be related to the marriage market hypothesis
that single households, who are active in the marriage market, are more likely to commit
to or move to core areas, which have a larger marriage market pool. In contrast, couples
with children moved out from core to non-core areas. One of the reasons behind the couples’
movement is that couples might want to consume larger spaces as it is shown in column (3)
of Table A.6, so they move out to non-core areas as they are cheaper.
18
In the data, we cannot determine the legal relationship between people, who move together. Men and
woman who have moved in together are considered couples if both individuals are older than 18 and if their
age differences are less than 12 years.
31
1.9 Quantifying the Mechanism: Structural Model
Let’s recall Equation 1.8 and 1.9 in Section 1.3.
For the changes in the gender employment gap,
@(
N
m
jt
N
fe
jt
)
@HSR
=

@(1 +T
jt
)
@HSR
+

@(

Mjt

Wjt
)
@HSR
(1.15)
and for the gender wage gap,
@(
W
m
jt
W
fe
jt
)
@HSR
= ( 1)

@(1 +T
jt
)
@HSR
+

@(

Mjt

Wjt
)
@HSR
(1.16)
where
=
1
1

+1
.
The structural decomposition using Equation 1.15 and 1.16 is as follows.
@(
W
m
jt
W
fe
jt
)
@HSR
and
@(
N
m
jt
N
fe
jt
)
@HSR
can be directly estimated, from the reduced form estimations using the instrumental
variables in Section 1.7. Under the assumptions of a set of parameters (;;), Equation
1.15 and 1.16 become a system of two linear equations with two unknown variables,
@(1+T
jt
)
@HSR
and
@(

Mjt

Wjt
)
@HSR
.
Table 1.14 summarizes three parameters used in the quantitative estimation. The choice
of elasticity of substitution in the production function between male- and female-intensive
inputs () and the labor share in Cobb-Douglas Function in each male- and female-intensive
sectors () are borrowed from the Bank of Korea estimates (Bae (2014)). CDF parameter
of labor force participation cost () is calibrated. Details of the calibration are provided in
Appendix A.7.
32
Table 1.14 about here
Table 1.15 presents the results of the quantitative estimation of the impact of HSR,
solving Equations 1.8 and 1.9 given the set of parameters in Table 1.14. HSR reduces
labor force participation costs for women in core areas by 4.74% and in non-core areas by
16.71%. In contrast, the relative productivity of male-intensive sectors decreased by 30.59%
and 10.63%, respectively, in core and non-core areas. The estimation results show that
HSR mainly benefits women in core areas through increased relative productivity in female-
intensive sectors. Women in non-core areas benefit more from the decreases in labor force
participation costs, with the expansion of HSR.
Table 1.15 about here
1.10 Robustness Check
1.10.1 Subset Analysis
In this section, robustness checks are performed to validate the results found in the
previous sec-tions. One of the main concerns is if outliers may drive the empirical results.
To address this, the same sets of regression in Section 1.7.1 with subsets of the sample are
used: (1) without districts designated as a special district by the government (i.e., Sejong
City); (2) without districts located in big cities; (3) without districts located close to the
North Korean border. Neither of the subsample analyses change the main findings of the
results in Section 1.7.1.
Without Sejong cities
First, the subsample without Sejong City, an autonomous city in South Korea, is con-
sidered. Since 2012, the South Korean government has relocated numerous ministries and
agencies to Sejong. If this new city’s influence drove the redistribution of local economic
33
outcomes, this could compromise our primary regression’s causal relationship. As we can
observe from Table 1.16, the estimates barely changed in the estimation without Sejong city,
which confirms the validity of the main findings.
Table 1.16 about here
Without Districts in Big Cities
Next, Table 1.17 contains a set of regressions without the following “big” cities: Seoul,
Busan, Daejeon, Daegu, and Gwangju. All the results are quantitatively similar to our
main regressions, except for the gender gap in employment (the third column in the upper
panel). The impact of HSR on the reduction in the gender employment gap in non-core areas
dissipate if we remove the big cities in the non-core areas from the variables. This means that
the decline in the gender gap in non-core areas mainly occurs in the highly industrialized
areas, whereas reduction was not significant in rural areas.
Table 1.17 about here
Without Districts near North Korean Border
Finally, Table 1.18 presents the subsample analysis without districts closer to the North
Korean border, which could follow a different trajectory in growth due to the political uncer-
tainty. The regression results are qualitatively and quantitatively close to the main results.
Table 1.18 about here
1.10.2 Nevo and Rosen (2012)’s Identification with Imperfect In-
struments
Using old railroad stations as an instrument strategy is widely accepted in the urban eco-
nomics literature, but this approach has limitations. For instance, if the path dependency
34
in local development exists, even if Japanese construction of the railroad did not target the
local economy growth after 100 years, the existence of old railroad stations could contribute
to today’s growth of the districts. To address this issue, in this section, Nevo and Rosen
(2012)’s identification with imperfect instrument idea was used as a robustness check of the
empirical estimates using instrumental variables.
Nevo and Rosen (2012) allow the instrumental variable to be correlated with the error
term; however, they assume the correlation between the instrumental variable and the error
term has the same sign as the correlation between the endogenous regressor and the error
term. Moreover,theinstrumentalvariableisassumedtobelesscorrelatedwiththeerrorterm
than is the endogenous regressor. These assumptions provide the bounds for the parameters
using the instrumental variable instead of the exact estimates. Qualitatively, the bounds
estimated using Nevo and Rosen (2012) are similar to our main regression results in Section
1.7.1.
1.11 Conclusion
This study investigates the effect of HSR on the redistribution of economic activities and
the gender differences in labor market outcomes. Considering construction of the old rail-
road in the Japanese invasion era as an instrumental variable, this study demonstrated that
population and employment move from core areas to non-core areas due to the construction
of the HSR in South Korea. Women benefit more from HSR expansion than men because the
local service sector wherein women mainly work benefits more from reduction in travel time
with HSR as it reduces the cost of moving people rather than that of goods. Moreover, with
changes in endogenous local amenities with HSR, especially changes in education amenities,
women benefit more from HSR as it potentially reduces the burden of childcare and other
domestic burdens, which is not a significant concern for men.
35
With the construction of the old railroad in the Japanese invasion era as an instrumen-
tal variable, this paper empirically shows that HSR narrows the gender employment gap in
both core areas (i.e., districts in the Seoul metropolitan areas) non-core areas. The wage
disparities between men and women in core areas decrease with HSR.
The mechanisms are structurally decomposed using a spatial general equilibrium model.
The HSR’s impact on the gender employment and wage gaps can be decomposed into the
labor demand and the labor supply channels in the model. The quantitative decomposition
shows that overall, HSR increases the labor demand of female-intensive sectors and decreases
women’s labor participation costs. Specifically, in core areas, the labor demand impacts are
more significant than the labor supply impacts, whereas, in non-core areas, the labor supply
impacts are more distinctive.
Finally, the model-predicted mechanisms were empirically explored. Women benefit more
from HSR expansion than men, as the local service sector in which women mainly work ben-
efits more from reducing travel time with HSR than the sectors where men mostly work.
Additionally, the improvement in local amenities, particularly in education and childcare
facilities, reduces women’s childcare burden and encourages women to join the labor force.
This paper contributes to the literature by focusing on the impact of transportation in-
frastructure on the gender gap in labor market outcomes, which there exists limited prior
evidence. Furthermore, this study extends the literature by examining how improved trans-
portationtechnologycouldimpactboththedemandandsupplysideofthegender-segmented
labor market. This paper’s findings shed light on the importance of understanding the het-
erogeneous impacts of infrastructure investments across different demographic groups, which
is an essential consideration, given the considerable spending on the transportation infras-
36
tructure.
This study is not without limitations. First, the study does not cover much of the overall
welfareformenandwomen. Moreover, ittestshowtheHSRaffectstheeconomicopportunity
for women somewhat indirectly; however, did not provide definite evidence of why women’s
labor force participation costs would be decreased. Future studies related to this topic could
address these questions.
37
Figure 1.1: Spatial Distribution of Population and Employment in South Korea
(a) Population (2010) (b) Population Growth (2003-2012)
(c) Employment (2010) (d) Employment Growth (2003-2012)
38
Figure 1.2: Cross-Country Gender Wage Gap (year=2015)
Note: The gender wage gap is defined as the difference between median earnings of men and women relative
to median earnings of men. Data refer to full-time employees on the one hand and to self-employed on the
other.
Source: OECD (2020), Gender wage gap (indicator)
39
Figure 1.3: Gender Differences in Time Use across Years
40
Figure 1.4: Wage Differences between Men and Women across Years
41
Figure 1.5: Spatial Distribution of Employment Sex Ratio
(a) Sex Ratio (2000) (b) Changes in Sex Ratio (2000-2015)
Note: Sex ratio of employment is defined as local male employment over female employment. Source: Census on Establishment (Statistics Korea)
42
Figure 1.6: Life Cycle of Work and Marriage Status
(a) (%) Married
(b) (%) Working
43
Figure 1.7: Occupation Composition for Gender and Age Group
(a) Men, age 25-34 (b) Men, age 35-44
(c) Women, age 25-34 (d) Women, age 35-44
44
Figure 1.8: KTX Network in South Korea
(a) Phase 1 KTX Station (2004) (b) Phase 1 Area Covered by KTX (2004)
(c) Phase 2 KTX Station (2012) (d) Phase 2 Area Covered by KTX (2012)
Note: Area covered by KTX are districts with a centroid distanced less than km from any KTX stations
45
Figure 1.9: Annual Ridership across Different Transportation Modes
(a)
(b)
Source: Korea Transportation Database (KTDB)
46
Figure 1.10: KTX Network and Old Railroad Stations Constructed during Japanese Colonial Era
Note: Color represents the population density of districts in 2010. Districts within the thick red lines are
defined as core areas, districts in Seoul metropolitan areas, which are districts of Incheon, Seoul, or Gyeonggi
provinces. Shaded districts are ’treated’ districts, where the population-weighted centroid of districts is
within 15km of any high-speed rail stations; and zero, otherwise. Red dots are the location of old railroad
stations constructed during the Japanese colonial era.
47
Table 1.1: Descriptive Statistics of Demographics
Panel A. Share of Working Population and Single
Working (%) Single (%)
Female Male Female Male
Core 39.72% 75.68% 12.40% 12.96%
Non-core 46.96% 75.53% 6.92% 7.50%
‘Working’ includes any types of employment(e.g. for-
mal, informal). Statistics are share of (fe)male, who
work (Column (1), Column (2)); and who are not mar-
ried, i.e. single (Column (3), Column (4)). Statistics
are mean in 2005, over age 25. Census on Population
and Household, Statistics Korea.
Panel B. Gender Gap in Labor Market Outcomes
Employment Gap Wage Gap
(male-to-female,%) (male-to-female,%)
Core 40.91% 38.8%
Non-core 38.63% 40.3%
Employment gap is defined as local male employment
over female employment. Wage gap is local male-to-
female wage ratio. The statistics are 3 year averages
(2000-2003). Census on Establishment, Korea Labor In-
come Panel Survey (KLIPS).
48
Table 1.2: Model Prediction: Impact of HSR
Panel A. Impact on the Gender Employment Gap (Male/Female)
Female Labor Participation Cost (1 +T
jt
)
Relative Productivity (

Mjt

Wjt
) Decrease No Changes Increase
Decrease (-) (-) (?)
No Changes (-) (.) (+)
Increase (?) (+) (+)
Panel B. Impact on the Gender Wage Gap (Male/Female)
Female Labor Participation Cost (1 +T
jt
)
Relative Productivity (

Mjt

Wjt
) Decrease No Changes Increase
Decrease (?) (-) (-)
No Changes (+) (.) (-)
Increase (+) (+) (?)
49
Table 1.3: Summary Statistics
(1) (2) (3) (4) (5)
Variable Mean Std Dev Min Median Max
log(Employment) 10.639 1.065 7.954 10.789 13.475
log(Establishment) 9.154 0.939 6.811 9.253 11.206
Sex ratio (Employment) 1.460 0.364 0.764 1.383 4.044
Sex ratio (Population) 1.006 0.041 0.898 1.005 1.308
log(Landprice) 87.267 9.316 46.527 90.003 105.563
log(Wage, female) -0.399 0.254 -2.377 -0.382 0.538
log(Wage, male) 0.038 0.240 -1.664 0.038 1.184
log(Wage, male)- log(Wage, female) 0.437 0.240 -0.596 0.434 2.201
log(Employment): Transportation 7.613 1.286 4.477 7.767 10.815
log(Employment): Construction 7.720 1.095 5.182 7.637 11.136
log(Employment): Public Admin 7.578 0.719 5.011 7.540 10.325
log(Employment): Manufacturing 8.817 1.403 4.382 8.802 12.440
log(Employment): Retail Store 7.024 0.932 4.407 7.064 9.616
log(Employment): Education 8.067 1.078 5.357 8.137 11.140
log(Employment): Medical Service 7.217 1.240 4.060 7.286 10.386
log(Employment): Restaurant 8.545 0.967 6.174 8.684 11.107
Establishment per population: Retail 0.003 0.002 0.001 0.003 0.011
Establishment per population: Education 0.003 0.001 0.001 0.002 0.012
Establishment per population: Medical Service 0.001 0.001 0.000 0.001 0.005
Establishment per population: Restaurant 0.016 0.008 0.005 0.014 0.059
Education Workers Per Kids (age< 14) 9.417 2.482 14.295 87.968 16.707
Notes: All the statistics are the district-level, from 2000 to 2015.
50
Table 1.4: First Stage of Instrumental Variable Regression
(1) (2) (3)
First Stage
VARIABLES D=1: Treat D=1: Treat*Core D=1: Treat*Non-core
Old Railroad Station 0.564***
(0.63)
*Non-core areas 0.000 0.542***
(0.042) (0.054)
*Core areas 0.687*** -0.000
(0.100) (0.128)
State Fixed Effect X X X
Observations 238 238 238
R-squared 0.354 0.612 0.485
F-Statistics 15.69 39.87 23.93
Notes: *** p<0.01, ** p<0.05, * p<0.1. Column (1)-(3) reports the first stage
of the instrumental variable regression. Column (1) is the first stage for the
regression without heterogeneous treatment effects and column (2), (3) are
with heterogeneous treatment effects. Old Railraod Station variable is equal
to one, if the population-weighted centroid of districts is within 15km of any
Japanese constructed train stations; or zero, otherwise. Core areas are defined
as districts in Seoul metropolitan areas, which are districts of Incheon, Seoul,
or Gyeonggi provinces. Non-core areas are all other districts.
51
Table 1.5: Impact of HSR on Local Population, Employment and the number of Establishments (Two-way Fixed Effects Model)
(1) (2) (3) (4) (5) (6)
log of Population Employment Establishment Population Employment Establishment
Treat 0.033* -0.002 0.022
(0.017) (0.020) (0.020)
*Non-core areas 0.073*** 0.039* 0.066***
(0.018) (0.021) (0.020)
*Core areas -0.056 -0.097** -0.078*
(0.036) (0.042) (0.042)
Year FE X X X X X X
District FE X X X X X X
Province-Year FE X X X X X X
Observations 3,806 3,806 3,806 3,806 3,806 3,806
R-squared 0.994 0.993 0.992 0.994 0.993 0.992
N. Districts 238 238 238 238 238 238
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are in parentheses. The coefficients are the
regression results of instrumental variable estimation in Section 1.6.1. Dependent variable of column (1),
(4) is log of population, column (2), (5), is log of employment, and column (3), (6) is log of the number of
establishments respectively. Treat variable is equal to one, if the population-weighted centroid of districts
is within 15km of any high-speed rail stations; or zero, otherwise. Core areas are defined as districts in
Seoul metropolitan areas, which are districts of Incheon, Seoul, or Gyeonggi provinces. Non-core areas are
all other districts. Standard errors are clustered at district-level.
52
Table 1.6: Impact of HSR on Local Population, Employment and the number of Establishments (IV)
(1) (2) (3) (4) (5) (6)
log of Population Employment Establishment Population Employment Establishment
Treat 0.114 0.032 0.055
(0.087) (0.120) (0.125)
*Non-core areas 0.188*** 0.132** 0.161***
(0.045) (0.052) (0.047)
*Core areas -0.207** -0.407*** -0.410***
(0.085) (0.086) (0.090)
Province FE X X X X X X
1st stage F-stat 15.69 15.69 15.69
(Non-core) 39.87 39.87 39.87
(Core) 23.93 23.93 23.93
Observations 238 238 238 238 238 238
R-squared 0.161 0.092 0.094 0.191 0.043 0.018
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are in parentheses. The coefficients are the
regression results of instrumental variable estimation in Section 1.6.2. Dependent variable of column (1),
(4) is log of population, column (2), (5), is log of employment, and column (3), (6) is log of the number of
establishments respectively. Treat variable is equal to one, if the population-weighted centroid of districts
is within 15km of any high-speed rail stations; or zero, otherwise. Core areas are defined as districts in
Seoul metropolitan areas, which are districts of Incheon, Seoul, or Gyeonggi provinces. Non-core areas are
all other districts. Standard errors are clustered at district-level.
53
Table1.7: Impact of HSR on Population, Employment and Employment Rate across Gender (IV)
(1) (2) (3) (4) (5) (6)
Population Employment
Male Female Sex Ratio Male Female Sex Ratio
Treat
*Non-core 0.177*** 0.200*** -0.025* 0.120* 0.209*** -0.162**
(0.046) (0.045) (0.013) (0.060) (0.059) (0.068)
*Core -0.243** -0.168* -0.083 -0.438*** -0.295*** -0.199***
(0.095) (0.084) (0.067) (0.070) (0.090) (0.055)
Province FE X X X X X X
1st stage F-stat
(Non-core) 39.87 39.87 39.87 39.87 39.87 39.87
(Core) 23.93 23.93 23.93 23.93 23.93 23.93
Observations 238 238 238 237 237 237
R-squared 0.178 0.202 0.014 0.034 0.094 0.095
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are in parentheses. The co-
efficients are the regression results of instrumental variable estimation in Section 1.6.2.
Dependent variable of column (1), (2) is log of population of male and female, and column
(3)issexratioofpopulation(malepopulation/femalepopulation). Column(4), (5), islog
of employment of male and female, and column (6) is sex ratio of employment (male em-
ployment/ female employment). Treat variable is equal to one, if the population-weighted
centroid of districts is within 15km of any high-speed rail stations; and zero, otherwise.
Core areas are defined as districts in Seoul metropolitan areas, which are districts of In-
cheon, Seoul, or Gyeonggi provinces. Non-core areas are all other districts. Standard
errors are clustered at province-level.
54
Table 1.8: Impact of HSR on Wages, and Gender Wage Gaps (IV)
(1) (2) (3)
IV estimates
logwage (male) logwage (female) wage gap (male/female)
Treat
*Non-core areas 0.072 0.128 0.003
(0.257) (0.175) (0.096)
*Core areas 0.046 0.203** -0.158***
(0.100) (0.085) (0.017)
Province FE X X X
1st stage F-stat
(Non-core) 39.87 39.87 39.87
(Core) 23.93 23.93 23.93
Observations 179 166 162
R-squared 0.046 0.011 0.057
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are in parentheses.
The coefficients are the regression results of instrumental variable estimation
in Section 1.6.2. Dependent variable of column (1), (2), is log of wage of men
and women, respectively, and column (3) is gender wage gap, defined as local
wage of men relative to local wage of women. Treat variable is equal to one,
if the population-weighted centroid of districts is within 15km of any high-
speed rail stations; and zero, otherwise. Core areas are defined as districts in
Seoul metropolitan areas, which are districts of Incheon, Seoul, or Gyeonggi
provinces. Non-core areas are all other districts. Standard errors are clustered
at province-level.
55
Table 1.9: Summaries of Reduced Form Findings
Population (Employment) Gender Gap (male/female)
Employment Wage
Core (-) (-) (-)
Non-core (+) (-) (.)
56
Table 1.10: Mechanism 1.(1). Effect of HSR on Sectoral Employment
Panel A. Male Intensive Sector
(1) (2) (3) (4)
Log of Employment
(Transportation) (Construction) (Public Admin) (Manufacturing)
Treat
*Non-Seoul 0.241 0.251* -0.016 -0.316**
(0.137) (0.134) (0.114) (0.136)
*Seoul -0.735** -0.490** -0.189*** -0.925***
(0.310) (0.205) (0.062) (0.269)
Sex Ratio(2000) 10.09 7.56 4.40 2.36
Observations 237 237 237 237
Panel B. Female Intensive Sector
(5) (6) (7) (8)
Log of Employment
(Retail) (Education) (Medical Service) (Restaurant)
Treat
*Non-Seoul 0.251*** 0.333*** 0.474*** 0.205***
(0.060) (0.071) (0.085) (0.061)
*Seoul -0.328* -0.048 -0.090 -0.327***
(0.172) (0.087) (0.182) (0.025)
Sex Ratio(2000) 0.88 0.76 0.54 0.50
Observations 237 237 236 237
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are in parentheses. The coefficients are the second stage
regression results of instrumental variable estimation in Section 1.6.2. Dependent variable of column (1) - (8) is log
of employment in each sector. Core areas are defined as districts in Seoul metropolitan areas, which are districts
of Incheon, Seoul, or Gyeonggi provinces. Non-core areas are all other districts. Standard errors are clustered at
province-level.
57
Table 1.11: Mechanism 2. HSR’s Impact on Endogeneous Amenities in Non-core
(1) (2) (3) (4) (5)
SECTOR Retail Stores Medical Service Restaurant Education Education
VARIABLE Establishment per Residents Workers per Child (age< 13)
Treat
*Non-Seoul 0.219*** 0.161*** 0.052 0.249*** 0.016*
(0.059) (0.050) (0.049) (0.055) (0.009)
*Seoul 0.065 0.118 -0.054 0.131 0.036
(0.129) (0.151) (0.181) (0.179) (0.037)
Province FE X X X X X
Province FE X X X X X
1st stage F-stat
(Non-core) 39.87 39.87 39.87 39.87 39.87
(Core) 23.93 23.93 23.93 23.93 23.93
Observations 237 236 236 237 237
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are in parentheses. The coefficients are the second
stage regression results of instrumental variable estimation in Section 1.6.2. Dependent variable of column
(1), (2), (3), (4) is log of the number of retail (KSIC G471: department stores, supermarkets, grocery stores,
etc); education (KSIC P: Private and Public education institutes); medical service(KSIC Q86: hospitals and
all types of medical services); and restaurant (KSIC I56) establishments per resident respectively, following
Diamond (2016). Dependent variable of column (5) is number of workers in education sector per child below
age of 13. Core areas are defined as districts in Seoul metropolitan areas, which are districts of Incheon, Seoul,
or Gyeonggi provinces. Non-core areas are all other districts. Standard errors are clustered at province-level.
58
Table 1.12: Effect of Travel Time between Origin and Destination Districts on Migration: Single
(1) (2) (3)
VARIABLES log of Number of Migrants
(Single, all) (Single, Male) (Single, Female)
Panel A.
log(Travel Time) -0.050*** -0.050*** -0.045***
(0.013) (0.013) (0.013)
Observations 849660 849660 849660
R2 0.930 0.912 0.898
Panel B.
log (Travel Time)
*(origin:core/destination:non-core) -0.028 -0.026 -0.020
(0.021) (0.022) (0.023)
*(origin:non-core/destination:core) -0.101*** -0.111*** -0.127***
(0.020) (0.021) (0.022)
* Within Region -0.023 -0.016 0.007
(0.020) (0.020) (0.020)
Origin-Dest Fixed Effect X X X
Year Fixed Effect X X X
Origin-year Fixed Effect X X X
Destination-year Fixed Effect X X X
Observations 849,660 849,660 849,660
R-squared 0.935 0.919 0.906
N. origin-destination pair 56,644 56,644 56,644
Notes: *** p<0.01, ** p<0.05, * p<0.1. Dependent variable of column (1)-(3) is log of the
number of migrants in each district-pair for each demographic category. Core areas are
defined as districts in Seoul metropolitan areas, which are districts of Incheon, Seoul, or
Gyeonggi provinces. Non-core areas are all other districts. Standard errors are clustered
at origin-destination-pair-level.
59
Table 1.13: Effect of Travel Time between Origin and Destination Districts on Migration: Couple
(1) (2) (3)
VARIABLES log of Number of Migrants
(Couple) (with Kid(s)) (without Kids(s))
Panel A.
log(Travel Time) -0.019 -0.053*** 0.006
(0.015) (0.015) (0.013)
Observations 849660 849660 849660
Adj R2 0.907 0.903 0.852
Panel B.
log(Travel Time)
*(origin:core/destination:non-core) -0.069*** -0.119*** -0.013
(0.025) (0.026) (0.021)
*(origin:non-core/destination:core) -0.031 -0.043* -0.026
(0.023) (0.023) (0.020)
* Within Region 0.039* -0.002 0.055***
(0.022) (0.021) (0.019)
Origin-Dest Fixed Effect X X X
Year Fixed Effect X X X
Origin-year Fixed Effect X X X
Destination-year Fixed Effect X X X
Observations 849660 849660 849660
Adj R2 0.907 0.903 0.852
N. origin-destination pair 56,644 56,644 56,644
Notes: *** p<0.01, ** p<0.05, * p<0.1. Dependent variable of column (1)-(3) is log of the
number of migrants in each district-pair for each demographic category. Core areas are
defined as districts in Seoul metropolitan areas, which are districts of Incheon, Seoul, or
Gyeonggi provinces. Non-core areas are all other districts. Standard errors are clustered
at origin-destination-pair-level.
60
Table 1.14: Parameters
Parameter Value Source
Elasticity of Substitution in Production Function = 0.82 Bae (2014)
Labor Share in Cobb-Douglas Function = 0.66 Bae (2014)
Labor Force Participation Cost CDF Parameter = 0.9 Calibrated
Table 1.15: Quantitative Decomposition of the Impact of HSR
Core Non-core
(Reduced Form) Impact on the Gender Employment Gap (
@(
N
m
jt
N
fe
jt
)
@HSR
) -19% -16.2%
(Reduced Form) Impact on the Gender Wage Gap (
@(
W
m
jt
W
fe
jt
)
@HSR
) -16% 0%
(Estimated) Impact on Labor Force Participation Cost (
@(1+T
jt
)
@HSR
) -4.75% -16.71%
(Estimated) Impact on Relative Productivity (
@(

Mjt

Wjt
)
@HSR
) -30.59% -10.63%
61
Table 1.16: Robustness Check: Sub-sample Analysis, without Sejong Special City
(1) (2) (3)
IV Results
log(Employment)
(Male) (Female) (Sex Ratio)
Treat*Non-core 0.117
*
0.207
***
-0.164
**
(0.061) (0.059) (0.069)
Treat*Core -0.438
***
-0.295
***
-0.199
***
(0.070) (0.090) (0.055)
Obs 236 236 236
(4) (5) (6)
IV Results
log(Wage)
(Male) (Female) (Wage Gap)
Treat*Non-core 0.072 0.128 0.003
(0.257) (0.175) (0.096)
Treat*Core 0.046 0.203
**
-0.158
***
(0.100) (0.085) (0.017)
Obs 178 165 161
(7) (8) (9)
IV Results
log of
Population Employment Establishment
Treat*Non-core 0.180
***
0.151
**
0.158
***
(0.044) (0.054) (0.048)
Treat*Core -0.207
**
-0.377
***
-0.410
***
(0.085) (0.075) (0.090)
Obs 237 236 237
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are in
parentheses. The coefficients are the second stage regression results
of instrumental variable estimation in Section 1.6.2. Core areas are
defined as districts in Seoul metropolitan areas, which are districts
of Incheon, Seoul, or Gyeonggi provinces. Non-core areas are all
other districts. Standard errors are clustered at province-level.
62
Table 1.17: Robustness Check: Sub-sample Analysis, without Districts in Big Cities
(1) (2) (3)
IV Results
log(Employment)
(Male) (Female) (Sex Ratio)
Treat*Non-core 0.254
***
0.313
***
-0.130
(0.045) (0.044) (0.089)
Treat*Core -0.461
***
-0.353
***
-0.127
***
(0.000) (0.000) (0.000)
Obs 163 163 163
(4) (5) (6)
IV Results
log(Wage)
(Male) (Female) (Wage Gap)
Treat*Non-core -0.052 0.090 0.060
(0.384) (0.280) (0.144)
Treat*Core 0.156
***
0.284
***
-0.127
***
(0.000) (0.000) (0.000)
Obs 107 93 90
(7) (8) (9)
IV Results
log of (IV Results)
Population Employment Establishment
Treat*Non-core 0.231
***
0.272
***
0.229
***
(0.024) (0.034) (0.032)
Treat*Core -0.080
***
-0.411
***
-0.423
***
(0.000) (0.000) (0.000)
Obs 164 163 164
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are in
parentheses. Big Cities includes Seoul, Busan, Daejeon, Daegu,
Gwangju, Ulsan, Incheon. The coefficients are the second stage
regression results of instrumental variable estimation in Section
1.6.2. Core areas are defined as districts in Seoul metropolitan
areas, which are districts of Incheon, Seoul, or Gyeonggi provinces.
Non-core areas are all other districts. Standard errors are clustered
at province-level.
63
Table 1.18: Robustness Check: Sub-sample Analysis, without Districts close to North Korean
Border
(1) (2) (3)
IV Results
log(Employment)
(Male) (Female) (Sex Ratio)
Treat*Non-core 0.115
*
0.204
***
-0.159
**
(0.060) (0.058) (0.067)
Treat*Core -0.415
***
-0.274
***
-0.199
***
(0.059) (0.075) (0.053)
Obs 214 214 214
(4) (5) (6)
IV Results
log(Wage)
(Male) (Female) (Wage Gap)
Treat*Non-core 0.030 0.258
**
-0.051
(0.258) (0.087) (0.082)
Treat*Core 0.053 0.219
**
-0.166
***
(0.101) (0.091) (0.011)
Obs 167 156 153
(7) (8) (9)
IV Results
log of (IV Results)
Population Employment Establishment
Treat*Non-core 0.189
***
0.149
**
0.162
***
(0.045) (0.054) (0.047)
Treat*Core -0.200
**
-0.355
***
-0.384
***
(0.091) (0.062) (0.083)
Obs 215 214 215
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are in
parentheses. Districts closer to North Korea include all districts
in Gangwon Province and districts in Gyeonggi Province (Paju,
Pochun, Yeonchun, Gapyung, Dongduchun). The coefficients are
the second stage regression results of instrumental variable esti-
mation in Section 1.6.2. Core areas are defined as districts in
Seoul metropolitan areas, which are districts of Incheon, Seoul, or
Gyeonggi provinces. Non-core areas are all other districts. Stan-
dard errors are clustered at province-level.
64
Chapter 2
How Do Cities Change When We Work
from Home?
2.1 Introduction
The potential savings in overhead costs and commuting time from remote work are sig-
nificant.
1
Technological conditions have been improving steadily for years, yet the fraction of
Americans working from home has remained small. In 2019, just 4.2% of all workers worked
from home. In 2020, COVID-19 social distancing requirements forced many companies and
organizations to pay a part of the fixed cost of transition to remote work. Abundant survey
evidence suggests that many now plan to continue remote work at much higher rates even
after the pandemic is over.
2
A lasting increase in working from home could have far-ranging consequences for the
0
Joint Work with Co-authored with Matthew J.Delventhal and Andrii Parkhomenko
0
ThisarticlewaspublishedinJournalofUrbanEconomics,Delventhal,MatthewJandKwon,Eunjeeand
Parkhomenko, Andrii, "How Do Cities Change When We Work from Home?." Journal of Urban Economics
p.103331., Copyright Elsevier (Delventhal et al. (2021))
1
Mas and Pallais (2020) provide an overview of the current state of research in telecommuting. Bloom
et al. (2015) present experimental evidence that telework increases employee work satisfaction without nec-
essarily reducing their productivity.
2
A May 2020 survey by Barrero et al. (2020) finds that 16.6% of paid work days will be done from home
after the pandemic ends, compared to 5.5% in 2019. Results of a survey by Bartik et al. (2020) also indicate
that remote work will be much more common after the pandemic.
65
distribution of economic activity inside urban areas.
3
One of the critical factors driving
workers’ location choices is the need to commute between their job and their residence.
Increasingthenumberoftelecommutersmakesthistrade-offmootforasignificantfractionof
the workforce. In this paper, we quantify the potential impact of this change using a general
equilibrium model of internal city structure. The model features employment, residence, and
real estate development choices, as well as local agglomeration and congestion externalities,
and endogenous traffic congestion across 3,846 non-rural census tracts of the Los Angeles-
Long Beach combined statistical area.
We calibrate our model to match residence and employment patterns prevalent in Los
Angelesduringtheperiod2012–2016, withanaverageof3.7%ofworkersworkingfromhome.
We then conduct a counterfactual exercise in which we gradually increase the fraction of
telecommutersallthewayto33%,whichaccordingtoDingelandNeiman(2020),corresponds
to the share of L.A. metro area workers whose jobs could be performed mostly from home.
The effects on city structure over the long run can be broken into three categories.
First, jobs relocate to the core of the urban area, while residents move to the periphery.
The largest driver of this effect is workers who previously had to commute and can now
work at home. They tend to move farther away from the urban core to locations with more
affordable houses. This increases the demand for real estate in peripheral locations and
lowers the demand in the core, pushing jobs from the suburbs into more central locations.
Second, average commuting times fall, while commuting distances increase. Since fewer
workers commute, traffic congestion eases, which increases average speed of travel. Com-
muters take advantage of this and also move farther away from their workplaces to live in
locations with lower real estate prices.
Third, average real estate prices fall. As many workers move into distant suburbs, prices
in the periphery increase. However, these price increases are more than offset by the decline
3
A study by Upwork in October 2020 finds that since the beginning of the pandemic 2% of survey
participants had already moved residences because of the ability to work at home and another 6% planned
to do so (Ozimek, 2020).
66
of prices in the core. This decline is driven by two factors. The first is the decline in demand
for residential real estate in core locations. The second is the reduced demand for on-site
office space from workers who now telecommute. In the counterfactual where 33% of workers
telecommute, average house prices fall by nearly 6%.
In addition to these three broad trends, our quantitative model predicts considerable
heterogeneity in outcomes that is not accounted for by the simple core-periphery continuum.
Within the core, locations with high productivity gain jobs while less productive locations
lose them. At all distances from the center, locations with better exogenous residential
amenities either gain more or lose fewer residents than less attractive equi-distant locations.
Overall, thesinglemonocentricdimensionofdistancefromthecenteronlyaccountsforabout
half of all variation in predicted outcomes.
The shift to telecommuting implies changes in the income both of workers and the owners
of real estate. On the one hand, labor productivity is pushed upward as jobs leave periph-
eral areas, and employment in the most productive tracts increases. Productivity receives a
further boost from the accompanying increase in spatial agglomeration externalities. Simul-
taneously, labor productivity is pushed downward because more employees work at home
and teleworkers do not contribute to agglomeration. In our quantitative exercise, these two
effects offset each other almost completely, leading to very small increases in average wages.
At the same time, changes in the spatial distribution of real estate demand and the reduced
need for office space lead to lower real estate prices and thus a reduction in the income earned
by landowners and property developers.
Our results conform fundamentally with previous theoretical findings by, for example,
Safirova (2003), Rhee (2008), and Larson and Zhao (2017). Recent work by Lennox et al.
(2020)explorestheeffectsofworkingfromhomeinanAustraliancontextusingaquantitative
spatial equilibrium model. A related study of ours, Delventhal and Parkhomenko (2020),
extends the analysis to the entire U.S. and multiple types of telecommuters. Behrens et al.
(2021) and Davis et al. (2021a) develop stylized city models with on-site and remote work
67
in order to study the implications of greater work from home on the demand for floorspace,
productivity, income inequality, and city structure. In line with the predictions of our model,
Althoff et al. (2020) document a rellocation of residents from the densest to the least dense
locations in the U.S. during the COVID-19 pandemic, while Gupta et al. (2021) and Liu
and Su (2020b) show that bid-rent gradients in U.S. cities have flattened reflecting higher
demand for low density and lower demand for urban amenities.
This paper also follows a number of recent efforts to assess the impact of urban policies
andtransportinfrastructureoncitystructure, suchasthosebyAhlfeldt et al.(2015), Tsivani-
dis (2019), Owens et al. (2020), Anas (2020), and Severen (2021). Our paper uses a similar
framework to assess the impact of a change to the underlying technology of production on
urban structure.
The remainder of the paper is organized as follows. Section 2.2 describes the model.
Section 2.3 provides an overview of how we calibrate the model. Section 2.4 describes and
discusses the counterfactual exercises. Section 2.5 concludes.
2.2 Model
Consider an urban area that consists of a finite setI of discrete locations, each populated
by workers, firms, and floorspace developers. Total employment is fixed and normalized to
1.
Workers supply their labor to firms and consume residential floor space and a numeraire
consumption good. Workers suffer disutility from time spent in commuting between home
and work, and this time depends endogenously on aggregate traffic volume. Their choice
of residence and employment locations depends on the commuting time, wages at the place
of employment, housing costs and amenities at the place of residence, and idiosyncratic
location preferences. Residential amenities depend on agglomeration spillovers, which are
increasing in the residential density of nearby locations. Firms use labor and commercial
68
floorspace to produce the consumption good, which is traded costlessly inside the urban area.
Firms’ total factor productivity depends on agglomeration spillovers, which are increasing
in the density of employment in nearby locations. Developers use land and the numeraire to
produce floorspace, which can be put to residential or commercial use. The supply of floor
space in each location is restricted by zoning regulations that limit commercial development
and overall density.
We introduce work from home by proposing a second type of worker–the telecommuter.
Telecommuters only come to their worksite a small fraction of workdays and thus suffer
much less disutility from commuting. On the days that they are not in the office, they do
not use commercial floorspace and instead produce output using “home office” floorspace in
their residence location. Working from home uses floorspace less intensively than on-site
work, has a different total factor productivity, and neither contributes to nor benefits from
agglomeration spillovers.
This model is similar in many respects to Ahlfeldt et al. (2015). The remainder of the
section presents the model and Appendix B.2 provides additional details.
2.2.1 Workers
Commuters and Telecommuters
Before choosing where to work and where to live, workers draw their commuter type.
With probability 0, a worker becomes a “telecommuter.” With probability 1 , the
workerbecomesa“commuter.” Thetwotypesdifferinthefractionofworkdaystheycommute
to work,. Commuters must come daily and therefore have = 1, while telecommuters have
=
T
< 1.
69
Preferences
A worker n who resides in location i2I, works in location j2I, and has to commute
from i to j a fraction of time, enjoys utility
U
ijn
(;c;h) =
z
ijn
d
ij
()

c
1

1

h

; (2.1)
wherez
ijn
represents an idiosyncratic preference shock for the pair of locations i andj, and
d
ij
() is the disutility from commuting given byd
ij
() = (1)+e
t
ij
. Individuals consume
c units of the final good andh units of housing. The share of housing in expenditures is given
by
, and consumption choices are subject to the budget constraint (1 +)w
ij
() =c +q
i
h.
In this constraint,w
ij
() is the wage earned by a worker who commutes fromi toj a fraction
of days, and q
i
is the price of residential floorspace in location i. In addition to wages,
workers also earn proportional transfers,w
ij
(), which distribute income from land and the
consumption good sold to real estate developers equally among all city workers.
Idiosyncratic shocksz
ijn
are drawn from a Frèchet distribution with c.d.f. F
z
(z) =e
z

.
The indirect utility of worker n who lives in location i and works in location j is given by
u
ijn
() =z
ijn
v
ij
(), where
v
ij
()
X
i
E
j
w
ij
()
d
ij
()q

i
(2.2)
is the utility obtained by a worker, net of the preference shock. In the above formulation,
X
i
is the average amenity derived from living in location i, and E
j
is the amenity derived
from working in location j.
Commuting time is a function of total vehicle miles traveled and road capacity in the
entire city: t
ij
=t
ij
(VMT;Cap). We assume that the capacity is fixed and the elasticity of
time on each link (i;j) with respect to total volume is a constant"
V
. Appendix B.5 provides
more details.
70
Location Choices
Optimal choices imply that the probability that a worker with a given chooses to live
in location i and work in location j is

ij
() =
(X
i
E
j
w
ij
())

(d
ij
()q

i
)

P
r2I
P
s2I
(X
r
E
s
w
rs
())

(d
rs
()q

r
)

: (2.3)
As a result, the equilibrium residential population of workers with a given in location i,
and the equilibrium employment in location j are given by
N
Ri
() =
X
j2I

ij
() and N
Wj
() =
X
i2I

ij
(): (2.4)
Finally, totalresidentialpopulationisN
Ri
= (1 )N
Ri
(1)+ N
Ri
(
T
), andtotalemployment
is N
Wj
= (1 )N
Wj
(1) + N
Wj
(
T
).
2.2.2 Firms
Production
In each location, there is a representative firm that hires both on-site and remote labor
and produces a homogeneous consumption good which is traded costlessly across locations.
The total output of the firm in location j is Y
j
= Y
C
j
+Y
T
j
, where Y
C
j
and Y
T
j
are the
amounts produced on-site and remotely, respectively.
4
The on-site production function is
given by
Y
C
j
=A
j

N
C
Wj

H
C
Wj

1
; (2.5)
whereN
C
Wj
= (1 )N
Wj
(1)+
T
N
Wj
(
T
) is the supply of on-site labor,H
C
Wj
is commercial
floorspace, and is the labor share. The remote production function is also Cobb-Douglas
4
In our model, on-site and remote production are perfectly substitutable. Davis et al. (2021b) entertain
a model with imperfect substitution between on-site and remote work, yet they estimate a high elasticity of
substitution.
71
and it combines workers from different locations as follows:
Y
T
j
=A
j
X
i2I

N
T
ij

T

H
T
ij

1
T
: (2.6)
In this specification, N
T
ij
= (1
T
)
ij
(
T
) is the supply of remote labor of telecommuters
who reside in location i and work for a firm in location j, whereas H
T
ij
is the amount of
home office space the firm rents on behalf of these workers in the place of their residence.
5
Parameter is the productivity gap between on-site and remote work, common to all workers
and firms. We let the labor share in remote production,
T
, to be different from the labor
share in on-site production.
6
Wages
Firms take wages and floorspace prices as given, and choose the amount of on-site labor,
telecommuting labor, and floorspace that maximize their profits. Equilibrium payments for
on-site work at location j and remote work for a firm in location j while living in location i
are, respectively,
w
C
j
=A
1

j

1
q
j
1

and w
T
ij
=
T
(A
j
)
1

T

1
T
q
i

1
T

T
; (2.7)
where q
j
is the local price of floorspace. The take-home wage of a worker with a given
is the weighted average of payments to his commuting labor and his telecommuting labor:
w
ij
() =w
C
j
+ (1)w
T
ij
.
7
5
We assume that the firm rents the floorspace that remote workers need in order to work from home,
however this specification is isomorphic to the one in which the firm only pays for labor services of a
telecommuter and the telecommuter uses his labor income to rent additional floorspace in his house.
6
One may expect that<
T
because telecommuters tend to work in jobs that require little floorspace.
While we do not impose this inequality in our theoretical analysis, it holds in our calibration.
7
Note that the wage of a commuter does not depend on her location of residence i. However, the wage
of a telecommuter depends on his location of residence i because production uses home-office floorspace.
72
2.2.3 Developers
There is a large number of perfectly competitive floorspace developers operating in each
location. Floorspace is produced using the following technology:
H
i
=K
1
i
(
i
(H
i
)L
i
)

; (2.8)
whereL
i

i
andK
i
are the amounts of land and the final good used to produce floorspace,
and is the share of land in production.
i
is the exogenous supply of buildable land, and
in equilibrium it is optimal for developers to use all buildable land, i.e., L
i
=
i
. Function

i
(H
i
) 1
H
i

H
i
determines the local land-augmenting productivity of floorspace developers.
8
Parameter

H
i
determines the density limit in tract i. When H
i
approaches

H
i
,
i
(H
i
)
approaches zero. As a result, it becomes very costly to build due to regulatory or political
barriers, such as zoning, floor-to-area ratios, or local opposition to development.
Floorspacehasthreeuses: commercial,residential,andhomeoffices. Commercialfloorspace
can be purchased at price q
Wj
per square foot. Residential and home office floorspace is lo-
cated in the same structure (e.g., a house) and each can be bought at price q
Ri
. Developers
sell floorspace at price q
i
minfq
Ri
;q
Wi
g to either residential or commercial users. However,
the effective price that residents or firms pay for floorspace may differ from q
i
due to zoning
restrictions. The wedge between prices for residential and commercial floorspace is denoted
by parameter
i
> 0. If
i
> 1, regulations increase the relative cost of supplying commercial
floorspace. Thus, the relationship between residential and commercial floorspace prices is
9
q
Wi
=
i
q
Ri
: (2.9)
Thedemandforcommercialfloorspace(H
C
Wj
)andhomeofficefloorspace(H
T
ij
)arisesfrom
profit-maximizing choices of firms. The demand for residential floorspace (H
Ri
) comes from
8
This function was also used in Favilukis et al. (2019) to model density limits.
9
This equality does not need to hold if the supply of commercial or residential floorspace in a given tract
is zero. In our quantitative model, however, these corner cases do not occur.
73
utility-maximizing choices of residents. Equilibrium selling price q
i
equalizes the demand
and the supply of floorspace:
H
C
Wj
+
X
i2I
H
T
ij
+H
Rj
=H
i
: (2.10)
Finally, since developers optimally use all land available for development,
i
, equilibrium
land prices are given by
l
i
=

i
(q
Ri
(H
Ri
+H
Ti
) +q
Wi
H
Wi
): (2.11)
Appendix B.2 provides more details.
2.2.4 Externalities
Local total factor productivity and residential amenities depend on density. In partic-
ular, the productivity in location j is determined by an exogenous component, a
j
, and an
endogenous component that is increasing in the density of on-site labor in this location, as
well as every other location s, weighted inversely by the travel time from j to s:
A
j
=a
j
"
I
X
s=1
e
t
js
N
C
Ws

s
#

: (2.12)
Parameter> 0 measures the elasticity of productivity with respect to the density of work-
ers, while parameter accounts for the decay of spillovers from other locations. Productive
externalities may include learning, knowledge spillovers, and networking that occur as a re-
sult of face-to-face interactions between workers. Hence, we assume that only commuters
and telecommuters who are on-site on a given day contribute to these externalities.
Similarly, the residential amenity in locationi is determined by an exogenous component,
x
j
, and an endogenous component that depends on the density of residence in every other
74
location, weighted inversely by the travel time to that location from i:
X
i
=x
i
"
I
X
s=1
e
t
is
N
Rs

s
#

: (2.13)
Parameter> 0 measures the elasticity of amenities with respect to the density of residents,
and is the decay of amenity spillovers. The positive relationship between density and
amenities represents, in reduced form, the greater propensity for both public amenities, such
as parks and schools, and private amenities, such as retail shopping, to locate in proximity to
greaterconcentrationsofpotentialusersandcustomers. Alltypesofworkers, commutersand
telecommuters, contribute equally to amenity externalities at their location of residence.
10
2.2.5 Equilibrium
An equilibrium consists of residential and workplace employment of commuters and
telecommuters, N
Ri
() and N
Wj
(); wages of commuters and telecommuters, w
C
j
and w
T
ij
;
residential and commercial floorspace prices,q
Ri
andq
Wj
; land prices,l
i
; local productivities,
A
j
; and local amenities,X
j
; such that equations (2.4), (2.7), (2.9), (2.10), (2.11), (2.12), and
(2.13) are satisfied.
2.3 Data and Calibration
The Los Angeles-Long Beach Combined Statistical Area had a total population of 18.7
million in 2018, distributed across a total land area of 88,000 square kilometers (U.S. Census
Bureau, 2020). It comprises five counties (Los Angeles, Orange, Riverside, San Bernardino,
and Ventura) and 3,917 census tracts. To omit nearly empty desert and mountain tracts with
large land areas, we exclude any tracts that are below the 2.5th percentile of both residential
and employment density. This excludes less than 1% of workers and leaves us with 3,846
10
It would also be possible that telecommuters, by spending more time in the area of their residence,
contribute more to local amenities than commuters.
75
tracts that constitute the set of locations in the model. We focus on the five years between
2012 and 2016. To construct tract-level data on the residential and workplace employment,
we use the LEHD Origin-Destination Employment Statistics (LODES) data for the years
2012 to 2016. Tract-level wages are constructed using the American Community Survey
(ACS) and the Census Transportation Planning Products (CTPP). We also use CTPP to
estimate bilateral commuting times. Finally, prices of residential and commercial floorspace
come from the universe of transactions provided by DataQuick. Refer to Appendix B.1 for
more details on the data.
The baseline probability of telecommuting, , is set to 0.0374. This number corresponds
to the fraction of workers who report that they primarily work from home in the 2012–2016
individual-level data from the American Community Survey for the Los Angeles-Long Beach
CSA. The fraction of time that telecommuters spend at an on-site workplace, , is set to
0.114, based on Global Work-from-Home Experience Survey (Global Workplace Analytics,
2020).
11
The elasticity of commuting time with respect to total traffic volume in the city,
"
V
, is set to 0.2, following Small and Verhoef (2007). In Appendix section B.5 we discuss
the robustness of results to different values of "
V
.
We calibrate the relative TFP of telecommuters, , so that the wages of commuters and
telecommuters are identical in the benchmark economy.
12
The floorspace share of telecom-
muters,
T
, is calibrated so that, on average, the home office of a telecommuter constitutes
20% of her house.
13
The calibrated values of and
T
are equal to 0.71 and 0.934, respec-
11
The survey asked the number of days an employee worked from home per week. We classify workers as
telecommuters if they work from home three or more days per week. According to the survey, 9% of workers
work from home five days per week, 2% do this four days a week, and 3% work from home three days per
week. Based on these numbers, we calculate the fraction of time spent on-site as 1 [0:09 (5=5) + 0:02
(4=5) + 0:03 (3=5)]=[0:09 + 0:02 + 0:03] = 0:114.
12
Our empirical analysis finds that wages of telecommuters are higher than those of commuters, however,
the wage premium disappears once we control for age, education, industry, and occupation. It is also unclear
how the wage gap between the two types will change if many more workers start working remotely.
13
The average house size was 2,430 square feet in 2010, according to Muresan (2016). Home-based tele-
workers have, on average, 500 square feet larger homes than other workers (Nilles, 2000). Hence, telecom-
muters’ houses are about 20% larger. This gap may reflect differences in income, location within a city, and
the need for designated workspace within a house. All of these factors are also present in our model. More
recent work by Stanton and Tiwari (2021) estimates a smaller house size difference of about 5–7%.
76
tively.
We borrow values for the remaining city-wide parameters from previous studies. The
share of housing in expenditures,
, is equal to 0.25, following Davis and Ortalo-Magné
(2011). The labor share in production, , is 0.8 (Valentinyi and Herrendorf, 2008), and the
land share in construction, , is 0.25 (Combes et al., 2018). Parameters that determine the
strength of agglomeration forces and decay speed for productivity and residential amenities
are borrowed from Ahlfeldt et al. (2015). In particular, we set = 0:071, = 0:3617,
= 1:0326, and = 0:7595.
14
We also take the variance of the Frèchet shocks and the
elasticity of utility with respect to commuting from Ahlfeldt et al. (2015), and set = 6:6491
and = 0:0105.
15
Besides city-wide parameters, in order to solve the model, we also need to know vectors
of structural residuals: E, x, a, , and

H. The model provides equilibrium relationships
that allow us to identify these residuals from observed prices and quantities. Appendix B.3
provides more details.
2.4 Counterfactuals
The COVID-19 outbreak in early 2020 has forced many individuals to work from home.
While before the epidemic around 4% of workers in the Los Angeles metropolitan area
worked from home, Dingel and Neiman (2020) estimate that as many as 33% of workers in
Los Angeles have jobs that can be done remotely.
In this section, we study how a permanent reallocation from working on-site to working
at home would affect the urban economy of Los Angeles. We simulate this increase by
14
Note that the parameter in our model corresponds to the product of the variance of the Frèchet shocks
and the elasticity of residential amenities with respect to density in Ahlfeldt et al. (2015).
15
Thesetwoparameters, aswellasthefourparametersthatdetermineamenityandproductivityspillovers,
were estimated for the city of Berlin, and we leave the estimation for Los Angeles for future work. Nonethe-
less, similar structural models with parameters estimated for other cities are characterized by similar mag-
nitudes of productivity agglomeration effects and spatial spillovers. At the same time, the estimates of
amenity agglomeration effects and spatial spillovers differ substantially across studies. See Berkes and Gae-
tani (2021),Tsivanidis (2019), and Heblich et al. (2020), among others.
77
permanently raising the probability of telecommuting, . The maximum permanent increase
we consider is all the way to 0.33. We also calculate results for a range of intermediate values.
As the number of teleworkers increases, both firms and workers change their locations
within the urban area. In response, the city also experiences endogenous adjustments in the
supply of commercial and residential floorspace, as well as commuting speeds. In what fol-
lows, we describe the effects on the spatial allocation of workers and firms, floorspace prices,
commuting patterns, wages and land prices. Then we discuss the drivers of counterfactual
changes, the role of endogenous productivity and amenities, and welfare effects.
2.4.1 Spatial Reallocation
Whenworkersarefreedfromtheneedtocommutetotheirworkplace, theytendtochoose
residences farther from the urban core in locations with more affordable housing. As the
share of telecommuters rises, this drives a reallocation of residents from the core of the urban
area towards the periphery.
16
The top panel of Figure 2.1 maps the predicted reallocation
of residents when the fraction of telecommuters rises to 33%.
As residents decentralize, employment centralizes. There are three main factors driving
this reallocation. First, the flipside of a telecommuter being able to access jobs even if
they live far away, is that employers can access the labor of telecommuters even if they are
located far from where they live. Therefore, employment shifts from locations which are less
productive but closer to workers’ residences, toward locations closer to the core which have
higher estimated exogenous productivity and benefit from greater productivity spillovers.
Second, the reallocation of residents increases the demand for floorspace in the periphery
and reduces it in the core, creating a cost incentive for jobs to move in the opposite direction.
Third, the fact that teleworkers require less on-site office space further improves the cost-
efficiency of firms in core locations with high productivity but high real estate prices. The
16
In line with our model’s predictions, Althoff et al. (2020) document a reallocation of residents from the
densest locations to the least dense locations in the U.S. during the COVID-19 pandemic.
78
Figure 2.1: Changes in residence, jobs, and real estate prices
1500
1000
500
0
500
1000
1500
residents per sq. km.
3000
2000
1000
0
1000
2000
3000
workers per sq. km.
< -20
-15
-10
-5
0
5
10
15
> 20
% change
Note: Absolute change in residential density (top), job density (middle) and % change in floorspace prices
(bottom).
79
middle panel of Figure 2.1 maps the predicted reallocation of jobs.
17
The net effect of these reallocations is to reduce the price of floorspace in core locations
and increase it in the periphery.
18
The bottom panel of Figure 2.1 maps predicted changes
in real estate prices when the fraction of telecommuters rises to 33%.
2.4.2 Commuting
Ashifttotelecommutingbringslargebenefitstothoseworkerswhodonothavetocometo
the office every day anymore and therefore suffer less disutility from commuting. However,
those who still have to commute benefit too, as traffic congestion drops and commuting
speeds increase. As the upper left panel of Figure 2.2 shows, with lighter traffic and faster
speeds, the average commuting time for those who still commute falls from 31 to 30 minutes.
At the same time, the average commute distance for commuters increases by nearly 1 km,
as they relocate farther away. This can be seen in the upper right panel. However, the total
amount of kilometers traveled falls by 29%, which suggests possible environmental benefits
of the increase in telecommuting.
19
The magnitudes of these effects depend importantly on
the elasticity of speed with respect to traffic volume, "
V
. Simulations for alternative values
of "
V
can be found in Appendix B.5.
2.4.3 Wages and Floorspace Prices
When the share of telecommuters increases, two opposite forces influence average wages.
On the one hand, jobs are being reallocated to more productive locations that also benefit
from agglomeration externalities. On the other hand, a larger fraction of the workforce does
not contribute to these externalities. In our calibration, these two forces almost perfectly
balance each other. As can be seen in the lower left panel of Figure 2.2, a full increase in
17
For a breakdown of residence and job changes by worker type, see Appendix B.4.
18
This prediction is supported by the evidence in Gupta et al. (2021) who document a flattening of
bid-rent curves in major U.S. metropolitan areas during the COVID-19 pandemic.
19
These benefits may be curbed by possible countervailing effects of greater urban sprawl (Kahn, 2000).
80
Figure 2.2: Commuting, wages, and prices
Commute times Commute distances
0.0374 0.1 0.15 0.2 0.25 0.3 0.33
Fraction of telecommuters ( )
22
24
26
28
30
Commuting time, minutes
All workers
Commuters
0.0374 0.1 0.15 0.2 0.25 0.3 0.33
Fraction of telecommuters ( )
22
24
26
28
30
Commuting distance, km
All workers
Commuters
Wages and land prices Floorspace prices
0.0374 0.1 0.15 0.2 0.25 0.3 0.33
Fraction of telecommuters ( )
-8
-6
-4
-2
0
Change, %
Wages
Land prices
0.0374 0.1 0.15 0.2 0.25 0.3 0.33
Fraction of telecommuters ( )
-6
-5
-4
-3
-2
-1
0
Change, %
Residential floorspace price
Commercial floorspace price
Note: Upper left: average commuting time for all workers and commuters. Upper right: average commuting
distance. Lower left: percentage change in average wages and land prices. Lower right: percentage change
in floorspace prices. All variables are plotted as a function of the share of teleworkers.
the fraction of telecommuters to 33% leads to a 0.3% increase in average wages.
As can also be seen in the lower left panel of Figure 2.2, an increasing share of telecom-
muters is decisive for the average price of land. Residents reallocate themselves to less ex-
pensive locations, and firms with more telecommuters need less office space. If the fraction
of telecommuters rose to 33%, the income of landowners would fall by 8%.
The lower right panel of Figure 2.2 shows that the value of both types of real estate
falls by about 6%.
20
The relative decreases in residential and commercial prices depend
20
In this model, residential and commercial prices in a given location move one to one; see equation (2.9).
However, changes in average prices of each type may differ due to changes in the supply of each type of real
estate.
81
on the fraction of telecommuters. When the change in the amount of telecommuting is
relatively small, the decrease in residential prices is somewhat larger. After the fraction of
telecommuters passes 28%, commercial prices are hit harder.
2.4.4 Accounting for Counterfactual Changes
Whatarethemainfactorswhichdrivetheseresults? Wefindthatasubstantialpartofthe
variation in predicted changes is accounted for by simple measures of centrality such as the
distance to the central business district. In this way, there is substantial overlap between our
predictions and the predictions that could be obtained from a uni-dimensional “monocentric
city” model. We also find that there is significant heterogeneity in predicted outcomes
between tracts that are roughly the same distance from downtown L.A. This additional
heterogeneity reflects the differences in exogenous local characteristics and transport network
connections which our quantitative model allows us to account for.
Figure 2.3: Quantiles of Centrality and Counterfactual Reallocations
1 0.8 0.6 0.4 0.2 0
-75%
-50%
0
50%
100%
200%
400%
N R, i
i
(% change)
local mean
1 0.8 0.6 0.4 0.2 0
-75%
-50%
0
50%
100%
200%
400%
N W, i
i
(% change)
local mean
1 0.8 0.6 0.4 0.2 0
-75%
-50%
0
50%
100%
200%
400%
q R, i (% change)
local mean
Note: The x-axis is scaled to quantiles of the centrality measure, weighted by land area. The size of each
circle is proportional the land area of the tract.
This heterogeneity is highlighted in the three panels of Figure 2.3. Each panel plots
predicted changes on the y-axis against the land-area weighted centrality rank of a tract on
the x-axis–a centrality rank of 0 represents the most distant tract, a centrality rank of 1
82
represents the tract closest to the center of the metropolitan area.
21
In the middle panel, we
seethatwhilethereisanunambiguouspredictionofjoblossesintheperiphery, roughlyequal
numbers of tracts gain and lose jobs from the 60
th
percentile and higher of centrality. In the
left panel, we see that while peripheral tracts are projected to gain residents, predictions are
much more ambiguous once the centrality is higher than the 60
th
percentile. In the right
panel, we see that real estate price changes fall systematically as we move towards the center
ofthecity. Allthreepanelsshowsubstantialdisparitiesinpredictedoutcomesbetweentracts
of very similar centralities.
22
What can account for this variation?
To help answer this question, we perform a Shapley-style decomposition of the variation
in predicted outcomes between centrality, exogenous local productivity and exogenous local
employment and residential amenities.
23
We find that distance from the center can account
for at most 60% of the variation in changes in floorspace prices, around 40% of the variation
in changes in employment, and 50% of the variation in changes in residence across space.
Two of the key takeaways from this exercise are that (1) locations with higher exogenous
residential amenities have bigger resident gains and smaller resident losses, all else equal;
and (2) locations with higher exogenous productivity have larger job gains and smaller job
losses, all else equal.
24
21
We calculate an eigenvector centrality from the II matrix of inverse commuting disutilities, where
I is the number of model locations. This measure is highly correlated with both straight-line distance from
downtown Los Angeles, and travel time from downtown Los Angeles–the correlation is higher than 0.97 in
both cases. More details can be found in Appendix B.6.
22
We can also see this variation between equidistant tracts if we return to look at Figure 2.1. It is perhaps
most striking in the middle panel of Figure 2.1, where we can see that one set of tracts that are close to
the downtown experience strong gains in employment, while other tracts, equally close or even closer to
downtown, lose jobs. The bottom panel of Figure 2.1 shows large differences in the size of real estate price
reductions between different tracts close to downtown.
23
Full details of this decomposition are provided in Appendix B.6.
24
Maps of structural residuals are shown in Figure B.1 in Appendix B.3.
83
2.4.5 Role of Endogenous Productivities, Amenities, and Conges-
tion
In the baseline counterfactual, we assume that local productivities and amenities are
endogenous. We also assume that commuting speeds fall as total vehicle miles traveled goes
down, and that these increased speeds also mean that spillovers have a broader reach.
Whatistheroleofthesespecificationchoicesindrivingourresults? Weturneachofthem
offandoninturn, andshowtheresultsinTable2.1. Column(6), whenallmarginsareturned
“on,” corresponds to the benchmark scenario. It turns out that in none of these permutations
are our main results significantly altered. Commuting times go down, floorspace prices fall,
and overall welfare goes up, all in roughly the same proportions, no matter which set of
assumptions is turned on. There are, however, some variations which illustrate the role of
different model mechanisms in shaping the results.
First, let us compare columns (1) and (2) of Table 2.1 with columns (3) and (4). If
local productivities do not adjust endogenously, wages increase. This is primarily because
telecommuters do not contribute to productivity spillovers. If these adjust, the locations
which lose in-person workers–nearly every location–see a fall in productivity. As a result,
average wages fall.
Second, let us compare columns (1) and (3) with columns (2) and (4). If residential
amenities do not adjust, there is a bigger reduction in travel times and distances. This
is because allowing amenities to follow telecommuters out to the periphery increases the
attractiveness of peripheral locations for regular commuters, making them willing to put up
with longer commutes.
Finally, let us compare columns (5) and (6) with column (4). We see that endogenous
congestion leads to larger reductions in time spent commuting. It also flips small reductions
in distance traveled into small increases–increased speeds allow workers to travel further
while spending less time on the road. Comparing columns (5) and (6), we can see that
allowing the reach of spillovers to increase when travel speeds go down gives a small but
84
Table 2.1: Breakdown of results
Endogenous productivities: no no yes yes yes yes
Endogenous amenities: no yes no yes yes yes
Endogenous congestion: no no no no yes yes
Spillovers affected by congestion: n/a n/a n/a n/a no yes
(1) (2) (3) (4) (5) (6)
Wages of all workers, % chg 1.77 1.79 -0.39 -0.41 -0.37 0.31
Wages of commuters, % chg 2.66 2.80 0.34 0.45 0.51 1.21
Wages of telecommuters, % chg -0.13 -0.38 -1.98 -2.26 -2.26 -1.61
Residential floorspace prices, % chg -4.37 -5.03 -5.75 -6.16 -6.23 -5.63
Commercial floorspace prices, % chg -6.43 -7.39 -6.14 -6.86 -6.97 -6.41
Time spent commuting, all workers, % chg -31.42 -30.69 -31.46 -30.80 -32.23 -32.13
Time spent commuting, commuters, % chg -1.46 -0.43 -1.52 -0.57 -2.63 -2.49
Distance traveled, all workers, % chg -31.91 -30.69 -31.96 -30.85 -28.96 -28.82
Distance traveled, commuters, % chg -2.18 -0.42 -2.24 -0.65 2.06 2.27
Welfare by source, % chg
consumption 2.30 2.52 0.45 0.55 0.64 1.17
goods only 0.84 0.85 -1.26 -1.29 -1.24 -0.57
housing only 4.75 5.70 3.95 4.58 4.73 4.79
+ commuting 11.71 11.54 9.74 9.46 10.02 10.56
+ amenities 11.90 14.38 10.16 12.23 12.80 14.15
+ Frèchet shocks 17.97 19.67 15.28 16.79 17.14 18.91
Welfare by commuter type, % chg
commuter 1.94 2.06 -0.48 -0.41 0.82 2.24
telecommuter -3.33 -1.69 -5.52 -4.06 -3.94 -2.47
Note: Columns (1)–(6) present results from specifications with different combinations of endogenous produc-
tivities, amenities and congestion, and whether spillovers increases when traffic congestion goes down. Each
column reports the results of a counterfactual experiment with an increase of the fraction of telecommuters
to 0.33.
significant boost to wages.
2.4.6 Welfare
The lower half of Table 2.1 shows that the increase in telecommuting to 33% of the
workforce results in significant welfare gains, which we measure as consumption-equivalent
changes in expected utility (see Appendix section B.2.3 for details). We find that reduced
commuting is the single biggest driver of welfare improvements, even when traffic congestion
remainsfixedatthebenchmarklevel. Focusingoncolumn(6): whencommutingisaccounted
85
for in addition to the 1.2% gain from consumption, welfare gains rise by over 9 percentage
points. After this, improved access to amenities adds another 3.6 percentage points, while
workers’ improved ability to fulfill their idiosyncratic preferences contributes less than 5
percentage points.
One important driver of welfare gains for commuters is access to jobs. In large, sprawled
and congested cities, such as Los Angeles, good jobs are often inaccessible for households
who live on the periphery. To study how a shift to telecommuting impacts job access,
we calculate commuter market access for each tract as CMA
i
=
P
j
(w
j
e
t
ij
)

. We find
that an increase in the fraction of telecommuters improves average job access for those who
keep commuting by 16%, largely thanks to lower traffic congestion. We also find that the
elasticity of floorspace prices with respect to market access at the tract level falls, meaning
that places with better access to jobs command a lower price premium. Further details of
these calculations as well as other results can be found in Appendix B.4.
The utility of the average telecommuter is significantly higher than that of the average
commuter, due to reduced disutility from commuting, access to lower-cost housing, and
access to better-paying jobs and amenities. As a result, the shift of workers from commuting
to telecommuting is an important source of the welfare increases. Workers who remain
commuters or telecommuters, see their welfare change only marginally. Commuters who
continue to commute benefit from reduced time commuting, access to lower-cost housing,
and access to better-paying jobs and amenities, and see their welfare rise by more than 2%.
At the same time, telecommuters who were already telecommuting do not benefit from the
increase in their mode of work. On the contrary, they need to compete with an increasing
fraction of the workforce for residence and job sites that were previously accessible only to
them. Their welfare falls by about 2.5%.
86
2.5 Conclusion
In this paper we used a detailed quantitative model of internal city structure to study
what would happen in Los Angeles if telecommuting becomes popular over the long run. We
find substantial changes to the city structure, wages and real estate prices, and commuting
patterns. We also find that more widespread telecommuting could bring significant welfare
benefits.
Our analysis necessarily omits several important channels which could dampen or amplify
our findings. First, in our model all workers are ex-ante identical and have the same chances
of being able to telecommute. In reality, the ability to telecommute is correlated with
occupation, industry and income. Accounting for this would likely have two effects. First,
it would center the large shifts in jobs and residence even more on the high-density center-
city locations where the share of skilled, telecommute-ready workers is likely to be highest.
Second, there would be more downward pressure on the average wage. This is because these
center-city locations have higher than average local productivity. If these locations lose
proportionally more in-person workers, their reduction in productivity from spillovers will
be greater, as will the impact on aggregate average wages. It is also likely that different skill
levels of workers differ in their contribution to productivity externalities.
25
If higher-skilled
workers are also more likely to telecommute, the effect of this detail would be similar to the
previous one: additional downward pressure on wages.
Second, we calibrated the productivity gap between commuters and telecommuters to
ensure that their average wages are the same in the benchmark economy, and assume that
this parameter remains constant in the counterfactual. We also assume that telecommuters
do not contribute at all to productivity spillovers. However, as telecommuting becomes more
widespread, technological changes might increase the relative productivity of telecommuters
and allow them to contribute more to productivity spillovers even without literal face-to-face
interaction. This would put upward pressure on wages, as we find in a related paper of ours,
25
This is a finding of, e.g., Rossi-Hansberg et al. (2019).
87
Delventhal et al. (2021).
Third, we do not take account of non-commuting travel.
26
If we did, we would probably
find an increase in local traffic congestion in the peripheral areas that telecommuters relocate
to, alongside the reduction in congestion along the main commuting arteries. This would
mitigategainsfrommovingtotheperipheryandleadtolessdecentralizationoverall. Wealso
do not distinguish between transportation modes in the model. The reduction in congestion
brought by more telecommuting could be offset if some transit users start commuting by car.
Finally, wedonotallowmigrationinandoutofthecity. Inpractice, assomeworkersgain
the ability to work remotely, they may choose to leave Los Angeles and move to a different
city, or even a different country. On the other hand, telecommuters from elsewhere may
move into Los Angeles to enjoy local amenities. Indeed, this is what we find in Delventhal
et al. (2021), which expands the scope of analysis to include the entire U.S.
One more caveat is recommendable in interpreting our predictions for welfare. We model
telecommutingasafactimposedexogenouslyonworkers. Theyloveitbecausetheycommute
less. Most welfare gains come through this channel. In reality, some workers may dislike
remote work. If in our model telecommuting were a choice in which workers balance the
benefits against their individual dislike, welfare gains would almost surely be smaller than
what we report.
26
Couture et al. (2018) estimate that work-related trips account for only 40% of vehicle miles traveled in
the U.S.
88
Chapter 3
The Effects of Ability Tracking on the
Academic Performance in the Secondary
School: Evidence from South Korea
3.1 Introduction
Whether to group students based on their academic performance has been in the center
of policymakers and researchers’ attention
1
. Under the ability tracking
2
system, students
are exposed to a homogeneous learning environment compared to the grouping mechanism
that students are mixed, irrespective of their academic records (i.e., ability-mixing). Since
different compositions of groups create variations in peer groups and teachers’ pedagogical
strategies, a comprehensive understanding of each grouping’s impact is key to successful
education policies.
Although there are fairly well-developed literature on tracking effects, the most credible
0
Joint Work with Seungwoo Chin
1
Austria, Germany, Hungary, Singapore, and the Slovak Republic track students to a secondary school
based on their performance. By contrast, Canada, Japan, Norway, Sweden, South Korea, the UK, and the
US keep lower secondary education mixing (Hanushek et al. (2006), Brunello and Checchi (2007))
2
Tracking is often defined as assigning students to a different class (or school) according to their prior
achievement (e.g., Slavin (1987)).
89
papers, withrespecttocausalidentification, typicallyusearegressiondiscontinuityapproach
to examine the impact of being placed in a high-ranked school(or class) relative to a a lower-
ranked school(or class) at the margin(Pop-Eleches and Urquiola (2013), Card and Giuliano
(2016), Dustmann et al. (2017)). As such, the literature has relatively little to say about
the causal effects of tracking policy relative to a counterfactual where there is no tracking,
or on the impacts for inframarginal students. In addition, some randomized experiments
provide rigorous tracking effects estimates that explore how students’ academic performance
can be improved (or worsened) if they move to a different peer group (e.g. reallocation from
a non-honors group to an honors group). These papers, however, have focused only on a
fraction of the population in primary school students (Duflo et al. (2011)), United States Air
Force AcademyCarrell et al. (2013), university students(Booij et al. (2017), Garlick (2018)).
This limits our nation-wide understanding of the impact of the grouping policy on the overall
human capital accumulation.
The main contribution of our paper is to measure the impact of transition from track-
ing to mixing on the whole distribution of test scores. This paper makes use of using a
difference-in-difference research design using unique natural experiment called “Equalization
Policy" in South Korea. Before the policy implementation, students applied to high schools
within the districts they lived in and went through a competitive admissions process based
on their municical-level high school entrance test score. The competitive admission process
provides a homogeneous learning environment to students as students with similar academic
performance were assigned to the same high school. Under the new policy, students were
randomly assigned to high schools regardless of academic performance in the middle school.
This creates significant variations in academic performance amongst students within schools.
The removal of tracking system occurred at a differential pace across districts. Seoul im-
plemented the new highschool admission regime in 1974, and a few major cities such as
Busan, Daegu, Gwangju and Incheon followed in 1975. Ever since then, the “Equalization
Policy" has been expanding further. The variations in the timing of the policy implemen-
90
tation provide a difference-in-difference research design. This paper takes advantage of the
policy change at Ulsan in 2000
3
.
A wide range of micro data sets and the temporal variation in the policy implementation
across districts allow us to investigate various outcomes and the underlying mechanisms
using a difference-in-difference research design.
Empirical findings show that tracking benefits high-performing students who study more
and gain from a superior learning environment, with the expense of low-performing students’
academicperformance. ThepolicytransitiontoamixedsystemcausedaconvergenceinSAT
scores’ distribution to the mean and decreased SAT scores in the treated regions. We find
no teachers’ adjustment in response to their student mix whereas, changes in peer dynamic
seem to play an important role when the grouping mechanism changes. We see neither
clear evidence that teacher-pupil interaction improved nor the teaching effectiveness affected.
However, it is found that the higher the rank of the peers is, the better class atmosphere is
created. As a result, under the tracking system, students at the top 25% of the distribution
increase their effort level.
The teacher’s response to the grouping mechanism is similar to Booij et al. (2017) that
teachers do not adjust their teaching in response to whom they teach
4
. However, how the
peer dynamics affect the educational outcomes is different from what Booij et al. (2017) was
found.
The paper’s main contribution is to measure the impact of (the removal of) tracking
throughout the whole distribution of test scores. Albeit there is a well-developed literature
on Tracking, most of the papers typically focus on the fraction of the population, primarily
high achievers, which hampers our understanding of the overall impact of the designing of
education policy. In this paper, we make use of a drastic policy implementation, which com-
3
Tracking effects in South Korea have been explored in Hahn et al. (2008), Kim and Lee (2003), Kang
et al. (2007) Kim et al. (2008), Wang (2015).
4
This is in contrast with Duflo et al. (2011), who study the effect of grouping students from the lowest
and highest ability tertiles in their setting of primary school students in Kenya. They find that tracking
benefits both high-performing and low-performing students. It enables a teacher to provide more-tailored
instruction, which plays a more significant role than peer effects.
91
pletely changed the grouping mechanism in South Korean high schools
5
. This reform offers
a rare opportunity to speak to a tracking policy’s causal effects relative to a counterfactual
where there is no tracking. The mechanisms we found were different from the previous lit-
erature, which again highlights the importance of understanding the impact of tracking on
the general population.
The rest of the paper is organized as follows. Section 3.2 describes the background.
Section 3.3 outlines the data. Section 3.4 describes the empirical strategy. Section 3.5
discusses the results. Section 3.6 explores the suggestive mechanisms through which ability
tracking affects students’ performance. Section 3.7 provides robustness check and Section
3.8 discusses the policy implication. Finally, Section 8 concludes the paper.
3.2 Evolution of High school Allocation in South Korea
South Korea is well known for its booming economic growth and level of high human
capital
6
. Despite the fast-growing average educational attainment level, improving educa-
tional equity has always been the main issue for South Korean policymakers. Specifically,
policies regarding high school allocation have been reformed to improve the unequal environ-
ment across high schools. Before 1974, in every district, high school allocation was based on
tracking. Students took a district-level examination at the end of their middle school years
(grade 9). Based on the test score, students applied to schools within the district where they
took the exam. If the number of applicants exceeded a school’s maximum capacity, students
would be allocated according to the selection process that honors applicants with higher
scores. Even though most high schools in South Korea were public schools, this allocation
mechanism resulted in strict rankings of high schools within districts in terms of academic
5
as of 2008, more than 61% of high schools were under EP.
6
The nation was one of the OECD’s top performers in mathematics, reading, and science in PISA 2012.
More than 92% of students in 2009 entered high school, which was a noticeable improvement: in 1980, only
50% of students went into high school (KOSIS).
92
performance
7
. This process naturally resulted in ‘across-school tracking,’ as students with
similar academic scores were assigned to the same schools(Wang (2015)). However, this
tracking system had created a public outcry, as most parents wanted to send their kids to
top-performing high schools of the district, but such spaces were limited. Competition for
better high schools had become highly fierce, which also induced an extreme demand increase
in private tutoring
8
.
The “Equalization Policy” (hereafter, EP) was introduced in 1974 to address the concerns
for the fierce competition and rising inequality regarding high school allocation. The EP
abolished the competitive high school entrance exam and randomly assigned students to high
schools in their district, regardless of their prior academic achievement in middle school.
The EP’s assignment procedure is as follows: all students submit a preference ranking of
high schools within districts where they live
9
. In the first round, if the number of students
who applied for a school as a first choice exceeds the school’s maximum capacity, the system
randomly selects students up to the school’s maximum capacity. In the second round, schools
that are not filled up by students and students who did not get their first choice are matched
based on students’ second choices. Then the same assigning mechanism works as in the first
round. If the students’ second choices were already filled up in the first round, they are
matched to their third choices. This process is repeated until either no slots of schools or
students are unmatched
10
.
The timing of the policy implementation differed across districts; starting from districts
in Seoul and Busan in 1974, it had been extended to other cities until 1981.
11
The EP did
7
Speficially, SAT scores or number of students who were admitted to top universities, and the rank across
schools was known to the public
8
Kim and Lee (2010) investigates the private tutoring in South Korea.
9
A student should explicitly mention his/her most preferred school. Then his/her second and third most
preferred high schools should follow. A student may list up to 13 high schools.
10
This is similar to the Boston Student Assignment Mechanism in the United States (Abdulkadiroglu
and Sönmez (2003), and Abdulkadiroğlu et al. (2005)). The Boston Student Assignment Mechanism is
not strategy-proof to the extent that each student has an incentive to misrepresent a school preference to
avoid too a large risk. This misrepresentation also leads to Pareto inefficiency (Abdulkadiroglu and Sönmez
(2003)).
11
Daejeon, Chungju, Suwon, Chuncheon, Jeonju, and Jeju adopted the EP in 1979, while Changwon,
Sungnam, Wonju, Gunsan, Iksan, Mokpo, Ahndong, and Jinju adopted the EP in 1980.
93
not expand between 1981 and 2000, but all districts adopt it in Ulsan and seven districts in
Gyeonggi Province in 2000 and 2002, respectively. As a result, 61.05% of high schools were
affected by the EP as of 2008 (Korean Educational Statistics Service).
The policy reduced the variation of academic performances between schools to a large
extent. Figure 3.1 shows the distribution of the number of students who were admitted to
Seoul National University
12
from each high school, before and after the adaptation of the
EP. Before the EP(1999-2004, in the seven cities in Gyeonggi Province, and 1999-2002, all
districts in Ulsan), the distribution is right-skewed, meaning the gaps across high schools
were salient. For example, before the EP, the top-rank high schools in Gyeonggi Province
had more than 30 students admitted to Seoul National University, while the low-rank high
schools had none. However, after the policy implementation, the distribution has changed,
and the gap between high schools vanished considerably.
It is worth noting that there is no significant difference between public and private high
schools in terms of curriculum and students’ quality, especially between the late 1990s and
early 2000s. Regardless of whether it is public or private, all high schools are listed in the
same pool based on the districts that they are located. This makes the selection into private
or public schools impossible, as the random assignment rule applies no matter it is a public
or private school
13
.
12
The number of students who go to Seoul National University, which is considered to be the top-ranked
university, has been a metric of school quality in South Korea.
13
There are special-purpose high schools for science or foreign language in South Korea. They select
their students with different admission procedures, meaning that ability-tracking still takes place to some
extent. However, in the empirical analysis, we drop individuals who graduate from the special-purpose
schools because the inclusion confounds our definition of treatment.
94
3.3 Data
3.3.1 College Scholastic Ability Test
For evaluation of the change in tracking policy in Ulsan Metropolitan City in 2000, we
utilize College Scholastic Ability Test data (CSAT) between 2000 and 2007.
14
These are pop-
ulation data that have all scores of students who took the exam. Around 600,000-700,000
students take the exam annually. The percentage of students who did not show up at the
exam date is less than 5% on average between 2002 and 2007 (Korea Institute for Curricu-
lum and Education).
15
We make use of a percentile rank of Korean, math, and English.
We define students in Ulsan Metropolitan City as the treatment group. Students who took
the exam between 2000 and 2002 were under the tracking system. Table 3.1 reports the
summary statistics of the CSAT data. The average percentile rank for the whole sample is
47.89, close to that for the control groups.
What draws our attention is a significant increase in all treated districts’ scores since
the EP was implemented. At first glance in Table 3.1, it seems that students in the treated
districts benefited from the mixing system on average.
Figures 3.2, which depicts the distribution of test scores, provides another aspect. Before
EP,thedistributionwasslightlyright-skewedandunevenlydistributed. AfterEP,theoverall
distribution became a bell-shaped distribution. Students are more likely to attain the test
score around the mean under the mixing system than under the tracking system. This
description suggests that the mixing system is better at fostering students’ performance on
average. However, it is still unclear whether this is a causal effect of the EP.
14
The government changed the structure of the exam significantly in 2008 so that we do not include scores
from 2008 onwards.
15
Despitethelargesamplesize, themajordisadvantageofthedataisthatnoinformationofsocioeconomic
status, besides gender, exists. As each student takes different electives, results from electives are excluded
from our analysis. We restrict the sample to the scores of the student who takes the exam for the first time.
95
3.4 The Empirical Approach
In this section, we describe the empirical strategy employed. As noted by a large body of
literature (e.g., Bayer et al. (2007) and Choi et al. (2014)), endogenous school choice or resi-
dential sorting results in a simple ordinary least squares (OLS) estimate bias. Following Kim
et al. (2008) and Wang (2015), our approach is based on the linear difference-in-difference
model.
Specifically, this paper takes advantage of the adaptation of EP in Ulsan city, which was
implemented in 2000. Figure 3.3 illustrates the treated and control regions in our setting.
The non-shaded areas indicate always-mixing WHEREAS, the shaded area shows always-
tracking, while the dashed area means where the EP was implemented. The always-mixing
and always-tracking regions are the control group, whereas the dashed regions are considered
as the treatment group.
For an individual i who was born in year t, graduated from high school in district j, the
college enrollment test score is
Y
ijt
=
1
Treat
j
Post
t
+
2
Gender
ijt
+
j
+
t
+"
ijt
; (3.1)
whereTreat
j
takes on value one if a student is in treated district, and zero otherwise. Post
t
is one if a student took an exam before 2003, and zero otherwise. Due to a lack of covariates
in the data, gender is the only control variable. Equation (3.1) also includes district fixed
effect (
j
) and year fixed effect (
t
). We exploit a percentile rank as a dependent variable
(Y
ijt
). We use not only total score but also Korean, math, and English scores to explore
which subject is influenced by the EP the most.
In equation (3.1),
1
captures the mixing effect on the college entrance test score in the
treatment region. For
1
to be valid, the parallel trajectory in the absence of the policy
intervention and no migration assumption should be met. We will conduct the pre-trend
analysis in section 6.1. To investigate whether the transition from the tracking to the mixing
96
system triggers any change in population composition or Tiebout migration, we compare the
young population size and educational attainments in district level between 2000 and 2005
in Appendix C using the census data. Our analysis suggests that the likelihood that the
treatment affects other demographics is rather small. Robust standard errors are clustered
at the district level.
3.5 Results
For the linear difference-in-difference approach to be valid, both the treated and control
groupswouldhavefollowedparalleltrajectoryintheabsenceofthetreatment. Analyzingthe
pre-trend, we demonstrate that the treated and control groups are not significantly different.
Figure 3.4 illustrates the both ex-ante and ex-post trend using CSAT data. A vertical
line indicates when the policy was implemented. In this figure, for all four outcome variables,
both the treatment and the control group moved along the same path before the EP. They
began to diverge after the intervention. This suggests that they are comparable and that
both the treatment and the control group would have followed a similar path in the EP’s
absence. We also run a similar regression analysis.
16
Table 3.2 reports the impacts of the EP on the college entrance test score. It shows that
inthetreatmentregion, fewerstudentsareinthetoptier(topfivepercentortoptenpercent)
at the exam, while, simultaneously, students are less likely to be left behind. The probability
to score higher than the top five percent or the top ten percent decreases by 0.34 and 0.73
percentage points, respectively. Meanwhile, the chance to be in the bottom ten percent or
in the bottom five percent decreases by 5.92 and 3.19 percentage points, respectively. Our
estimates imply that students are more likely to score around the mean. As the positive
effect on being out of the bottom outweighs the negative impact that drags students away
from the right tail of the distribution, the total score increases by 2.3 percentage points, as
16
We estimateY
ijt
=
P
2007
Y =2001

Y
Treat
j
Year
Y
+
2
Gender
ijt
+
j
+
t
+"
ijt
, and present each
Y
in
figure 10. It implies how treatment effects have changed annually with the year of 2000 as a baseline.
97
seen in Column (7).
We explore mixing effects on each subject: Korean, math, and English. Our results, in
Table 3.3, show that the EP increased the probability to be out of the bottom regardless
of subject, and all magnitude in Column (4)-(6) are similar. On the other hand, only math
shows statistically significant positive mixing effects on a top tier at a 10 percent level. Our
results show that the chance to get a math score higher than the top five percent decreased
by 0.49 percentage points if students moved from tracking to mixing. As documented by
Gamoran (1992), instruction for math is more flexible than that for language. The significant
instructional differentiation in math might cause heterogeneity between subjects.
Corrected figure 1/37 Untitled document Saved 44 readability score DOCUMENT EDIT
ACCOUNT PREMIUM Eunjee Kwon eon.kwon@gmail.com
3.6 Mechanism
In this section, we explore suggestive mechanisms under which the tracking system affects
students’ academic performances. Different group compositions may affect teacher’s peda-
gogical strategies and spillovers from peers (Duflo et al. (2011)). To shed light on plausible
pathways through which peers and teachers respond to the school-wide ability-tracking, we
test how the EP affects students’ academic performance through various channels; studying
effort, teacher’s quality, class atmosphere, and more.
3.6.1 Data
Weusethe2004and2005wavesoftheKoreanEducationandEmploymentPanel(KEEP)
for this section’s analysis. This is an ongoing longitudinal survey that selected 2,000 students
from grade 9 (final grade in middle school) and 4,000 students from grade 12 (final grade
in high school) in 2004. The subjects are followed up by more than ten years to facilitate
research about the education and labor markets for the younger generation. We use the
98
baseline survey for 2,000 students in grade 9 in 2004 and the follow-up survey in 2005
because we intend to estimate an instant effect of tracking in the short term. If key variables
are not surveyed in the 2005 wave, we replace those with the 2006 wave. For example, the
query about teacher-pupil interaction was not carried out in the 2005 wave, so we exploit the
informationinthe2006wave. Thisdataincludesarichsetoffamilybackground, information
on schools and teachers, studying behaviors, and whether students go to high school in a
tracking region. The summary statistics of the data are presented in Appendix C.1.
3.6.2 Empirical Strategy
To investigate behavioral changes for both students and teachers caused by the tracking
system, we exploit the unique educational context in South Korea. Due to the recogni-
tion that tracking across middle schools increases education costs considerably and exposes
students to too fierce competition at an early age, South Korea’s government abolished all
tracking systems across the nation in 1971. It started assigning students to middle school
regardless of how well they studied in primary school. However, tracking across high schools
remains in some districts.
Figure 3.6 illustrates relationships between a student’s rank and peers’ rank in the same
school in our data. Note that all ranks presented in Figure 3.6 are ranks in middle schools,
not in high school, meaning that the ranks of two bottom graphs can be interpreted as
prior ranks. Given the history of tracking and mixing in South Korea, we expect no salient
association between one’s own rank and peers’ rank both at the middle and high school levels
under the mixing system. Conversely, we hope a positive relationship between those in high
school under the tracking system.
Thus, we utilize another difference-in-difference framework, with students in the region
under a high school tracking system defined as a treatment group. Students in the other
regions are considered to be a control.
17
17
Similarly, Wang (2015) estimates the tracking effects on non-academic outcomes. Unlike his work,
99
For an individual i who was surveyed in year t, studied in high school in district j, her
academic outcome is
Y
ijt
=
1
Treat
j
High
t
+
2
Treat
j
High
t
Above
i
+
3
High
t
+
i
+"
ijt
;
(3.2)
where Treat
j
takes a value of one if a student goes to high school in a tracking district
and zero otherwise. High
t
takes on value one if a student was surveyed when she was in high
school, and zero if she was surveyed in middle school. Above
i
represents a dummy variable
thatshowswhetherastudent’srankinmiddleschoolisabovethemedian. WeincludeAbove
i
to explore heterogeneity in tracking effect across prior ranks, as predicted by the model in
section 2. Equation (8) includes individual fixed effect, and "
ijt
is an unobservable distur-
bance. As the individual fixed effect considers all time-invariant individual characteristics,
including gender parents’ educational level and genetic information, additional covariates
are not included. Our outcome variables (y
ijt
) of interest include study hours, teacher-pupil
relationship, teacher qualification, and class atmosphere. All robust standard errors are
clustered at the district level.
3.6.3 Teacher Effect
First, we use several subjective measures
18
to see whether tracking changes teacher-pupil
interaction. Overall, Table 3.4 shows that interaction between teachers and students is not
affected by the group composition.
we explicitly add a triple interaction term to explore the heterogeneity between high and low performing
students
18
These measures are based on a scale of five. We let the outcome of interest take on value one if it is
four and above and take on zero, otherwise.
100
Next, we investigate the association between tracking and teacher quality. Due to no
consensus on what defines a better teacher (Hanushek and Rivkin (2006)), we use a few
relevant outcome variables. Following a large body of literature (Birdsall (1985), Summers
and Wolfe (1977), Rice (2003)), we exploit whether or not a teacher attains a masters
degree or higher, total years of experience as a teacher, and subjective evaluation by an
administrator in the school. Besides, results presented in Column (1) - (3) of Table 3.4
shows that teachers’ characteristics are no different regardless of whether a high school is
under the tracking system or not. This finding is consistent with Booij et al. (2017) in the
sense that, unlike Duflo et al. (2011), that teachers at the University of Amsterdam do not
respond significantly to the composition of their student bodies.
3.6.4 Peer Effect
What tracking system affects is classroom atmosphere. Column (4) of Table 3.4 shows
that students in the top half are more likely to experience a better education atmosphere
under tracking.
Weinvestigatedirectpeereffectsonpossiblemechanismsexploredintheprevioussection.
We estimate, for a student i at time t, Y
it
=
1
Peer
it
+
2
High
t
+
i
+"
it
, where Peer
it
is the mean of peer rank
19
evaluated at middle school, excluding oneself in the school that
one attends, andHigh
t
takes on value one if a student was surveyed in high school and zero
otherwise. As it includes the individual fixed effect, this reveals how peers are composed of
study hours and other school characteristics.
Table 3.6 shows the higher the prior rank of the peers are, the better class atmosphere
is created. This finding reinforces the claim that how tracking affects students’ academic
outcomes is mainly through the better performing peers, who create better classroom atmo-
spheres.
19
We define the rank in the way that a higher percentage represents a better performance. It implies that
the 100 percentile rank means the highest, while zero means the lowest.
101
3.6.5 Hours of Studying and Self Evaluation
Finally, we explore how this tracking leads to changes in studying behaviors and self-
evaluation. Table 3.7 presents changes in students’ self-studying hours depending on whether
they are under tracking or not. Table 3.7 shows that students in the bottom half reduce
their study hours under the tracking system, whereas those in the top half increase their
study hours.
Table 3.8 provides a possible behavioral motivation through which the tracking system
changes self-studying hours. We utilize a self-evaluation for Korean, math, and English,
respectively, as outcome variables and see if they are affected under the tracking system.
Table 3.8 shows that students in the bottom half of school in grade 9 overestimate themselves
under the tracking system. In contrast, those in the top half estimate themselves in the same
way as they did in middle school or slightly underestimate themselves, especially for English.
These results are consistent with the claim that the high types consider themselves to be
relatively worse because they are exposed to high-quality peer classrooms.
3.7 Robustness Check
3.7.1 Test for Tiebout Migration
To make our difference-in-difference framework not biased, the policy change from the
tracking to the mixing system should not trigger migration. As documented by Tiebout
(1956) and Benabou (1993), it is plausible for residents to migrate to seek better educational
services. If this were true, then the estimation could be biased in both directions, depending
on who migrates between the high and low types. Due to the data’s limitations, we cannot
access which type moves, but we can test whether the mixing system’s change causes high
school enrollment in the treated region.
With high school enrollment in a district each year as an outcome variable, we estimate
102
equation 3.1 by OLS. Column (1) in table A.1 shows no statistically significant difference
between the control and the treatment group at 10% is found. We also estimate Y 1
ijst
=
P
1988
Y =1984

Y
Treat
j
Year
Y
+
2
X
ijst
+
j
+
st
+"
ijst
with the same outcome variable to
compare annual change in high school enrollments with that in 1999 as a baseline. Column
(2) in Table 3.9 implies that the intervention does not cause migration between districts in
the GOMS case.
As high school enrollment data before 1999 are not available, we cannot test if the
intervention causes that in the CSAT case. Hence, we use the census data and compare the
control and treatment district in terms of population size and education level in districts;
we utilize the population of 10- to 14-year-olds and 15- to 19-year-olds and the number of
residents who attain a bachelor’s degree and who receive a master’s degree. Figure A.1 shows
that both the control and treatment group move in the same direction from 2000 to 2005.
This implies that the possibility of Tiebout migration is marginal.
3.8 Discussion: Equity and Efficiency
We view our main results from an argument of equity and efficiency with a back-of-the-
envelope calculation in this section. Viewed from equity grounds, on the one hand, our
empirical findings show that the tracking system causes educational inequality and strat-
ification within a city to the extent that it is good at fostering a top-ranked student and
that, at the same time, more students are likely to lag. This is consistent with prior lit-
erature in the sense that tracking or educational peer effect worsens inequality (Gamoran
and Mare (1989), Benabou (1993), Benabou (1996)). On the other hand, it is worthwhile
to investigate whether across-school tracking increases or decreases economic growth or pro-
ductivity from the perspective of efficiency. Throughout our entire work about efficiency,
we rule out possible economic inefficiency due to high inequality for simplicity. We define,
instead, productivity as a mean income: the higher a mean income is the better productivity
103
a city shows. Although the tracking system deteriorates inequality to some extent, it can be
justified on the grounds of efficiency if it increases the city’s mean income.
Using KEEP data,
20
we plot the relationship between the college enrollment test score
in 2004 and hourly wages 10 years later in Figure C.2. It shows that higher test scores
are positively associated with higher hourly wages in the future, reinforcing arguments that
educational inequality can be persistent. We present how they are related graphically in
figure 12. The relationship between test scores and future hourly wages appears to be
an increasing convex curve. This feature justifies that a widened equality contributes to
increasing a mean income. A city where everyone attains the mean score earns less income
in total than that where the distribution is polarized. Putting these together, debates on
whether tracking or mixing is superior hinges considerably on the social and cultural context
acityfaces. Trackingcanbeeitherbeneficialorharmful, dependingonhowthesocialwelfare
function is shaped.
3.9 Conclusion
This paper provides evidence on the effects of across-school tracking using two EP cases
in South Korea. Our difference-in-difference design in the first case shows that the transition
from the tracking to the mixing system causes students to be less likely to advance to college.
It hurts the top-ranked college enrollments harder than the four-year college enrollments. In
our second case, in which we use college entrance test score, we show that the EP changes
the score distribution to be more bell-shaped. More students are distributed around the
mean, while fewer students are distributed at both tails under the mixing system. We also
provide evidence on the behavioral changes triggered by across-school tracking. Students in
the top half exert more effort and benefit from a suitable classroom atmosphere for studying
under the tracking system. Students in the bottom half reduce self-studying hours. This
20
We use a sample of students in grade 12 (final grade in high school) in 2004 and take advantage of the
longitudinal data.
104
might cause educational inequality between the high and the low types to be exacerbated
and persistent. However, our simple analysis shows that the polarized academic performance
could not be the worst scenario if policymakers wanted to maximize a reference group’s mean
income. The increasing convex relationship between college entrance test scores and future
income implies that more inequality might increase social wealth. Hence, whether or not
across-school tracking is desirable depends on where society puts more weight on.
Across-school tracking is good at training high-achievement students at the expense of
widened educational inequality. The primary motivation behind the equalization policy in
South Korea was an outcry against overly fierce competition at an early age and a consider-
able education burden. However, South Korea performs well in many education indicators.
21
However, it should be recalled that no panacea exists. Coping with inequality might weaken
peer effects and vice versa.
21
The share of 25- to 34-year-olds in 2015 with at least an upper secondary education (high school) is
above the Organisation for Economic Co-operation and Development (OECD) average (98%, compared to
the OECD average of 84%). Take school performance, for example. Korean students are performing better
than most of the students in other OECD countries; according to the OECD’s Programme for International
Student Assessment (PISA) in 2012, South Korea led OECD countries in math, was second to Japan in
reading, and was in the top seven for science (OECD Observer, 2017).
105
Figure 3.1: Enrollments in Seoul National University (SNU)
Note: High school graduates in Gyeonggi Province and in Ulsan have been affected
by the equalization policy since 2005 and 2003, respectively.
106
Figure 3.2: Kernel Distribution of Test Score: Treatment Group
107
Figure 3.3: Treated and Controls Region
108
Figure 3.4: Pre-Trend Analysis: Lowess Graph, CSAT
109
Figure 3.5: Relationship Between Student’s Own Rank and Peers’ Rank
Note: This is constructed, using KEEP data. The average rank on the vertical axis
in each graph represents an average of rank of peers, excluding oneself, in a reference
group. All ranks mean how well students study in middle school.
110
Figure 3.6: Evidence that refutes Tiebout migration, CSAT
111
Table 3.1: Summary Statistics: CSAT
Variable Whole Sample Controls Treated P-value (A = B)
Before (B) After (A)
Total Score 47.894 47.83 47.366 49.747 0.00***
(25.51) (25.558) (26.728) (23.226)
Korean 48.091 48.04 47.477 49.722 0.00***
(28.222) (28.258) (28.96) (26.709)
Math 48.575 48.441 48.026 52.121 0.00***
(27.998) (27.987) (29.237) (26.825)
English 47.678 47.661 46.675 48.931 0.00***
(28.098) (28.175) (28.486) (25.937)
Female (= 1) 0.48 0.48 0.475 0.477 0.8022
(0.5) (0.5) (0.499) (0.499)
Observations 1,105,705 1,018,790 40,129 46,786
112
Table 3.2: Mixing Effects on College Enrollment Test Total Scores: CSAT
(1) (2) (3) (4) (5) (6) (7)
VARIABLES Top 5% Top 10% Top 25% Bottom 25% Bottom 10% Bottom 5% Test Score
Treat Post -0.0021*** -0.0048** -0.0077 -0.0727*** -0.0390*** -0.0117*** 2.3158***
(Mixing Effect) (0.0004) (0.0012) (0.0050) (0.0099) (0.0053) (0.0018) (0.3709)
Female (=1) -0.0020** -0.0010 0.0203*** -0.0654*** -0.0397*** -0.0172*** 3.9661***
(0.0006) (0.0016) (0.0041) (0.0058) (0.0071) (0.0036) (0.0834)
Observations 1,105,705 1,105,705 1,105,705 1,105,705 1,105,705 1,105,705 1,105,705
R-squared 0.0008 0.0013 0.0043 0.0166 0.0117 0.0057 0.0165
Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses. The standard errors are
clustered at the district level.
113
Table 3.3: Mixing Effects on College Enrollment Test Subject Scores: CSAT
(1) (2) (3) (4) (5) (6) (7)
VARIABLES Top 5% Top 10% Top 25% Bottom 25% Bottom 10% Bottom 5% Test Score
Korean
Treat Post -0.0016 -0.0019 0.0029 -0.0547*** -0.0442*** -0.0281*** 2.2425***
(Mixing Effect) (0.0012) (0.0020) (0.0032) (0.0076) (0.0068) (0.0042) (0.3628)
Female (=1) 0.0093*** 0.0197*** 0.0512*** -0.0823*** -0.0522*** -0.0342*** 5.9898***
(0.0013) (0.0024) (0.0038) (0.0050) (0.0068) (0.0054) (0.1911)
Math
Treat Post -0.0049** -0.0066** -0.0020 -0.0542*** -0.0381*** -0.0232*** 2.5639***
(Mixing Effect) (0.0013) (0.0023) (0.0040) (0.0071) (0.0041) (0.0027) (0.3494)
Female (=1) -0.0138*** -0.0209*** -0.0171*** -0.0479*** -0.0283*** -0.0166*** 1.4273***
(0.0006) (0.0004) (0.0008) (0.0050) (0.0043) (0.0029) (0.1966)
English
Treat Post 0.0007 0.0006 0.0014 -0.0579*** -0.0450*** -0.0258*** 2.3524***
(Mixing Effect) (0.0010) (0.0028) (0.0067) (0.0086) (0.0068) (0.0040) (0.4443)
Female (=1) 0.0045* 0.0092* 0.0325*** -0.0708*** -0.0455*** -0.0277*** 4.6636***
(0.0021) (0.0037) (0.0065) (0.0034) (0.0050) (0.0036) (0.2050)
Observations 1,105,705 1,105,705 1,105,705 1,105,705 1,105,705 1,105,705 1,103,346
Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses. The standard errors are
clustered at the district level.
114
Table 3.4: Tracking Effects on Teacher-Pupil Interaction
Students ( ) a teacher Teachers ( ) a student
(1) (2) (3) (4)
VARIABLES Respect Like Get Interested in Understand
Treat High -0.0365 -0.1197 0.0786 -0.0447
(Tracking effect) (0.0860) (0.0772) (0.0602) (0.1078)
Treat High Above half -0.0283 0.0536 -0.1205 0.0289
(0.0817) (0.0616) (0.0972) (0.0823)
Observations 2,134 2,134 2,134 2,132
R-squared 0.5854 0.6157 0.5918 0.5998
Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses.
The standard errors are clustered at the district level.
115
Table 3.5: Tracking Effects on Teacher’s Quality and Class Atmosphere
(1) (2) (3) (4)
VARIABLES Teacher has Teacher’s Teacher’s Classroom
M.A. Degree Experience Ability Atmosphere
Treat High -0.0192 -0.0731 -0.0243 -0.0976
(Tracking effect) (0.0853) (0.1196) (0.1101) (0.1272)
Treat High Above half -0.0498 -0.0175 -0.0678 0.1645**
(0.0665) (0.0616) (0.0737) (0.0741)
Above half 0.0140 0.0591 0.0176 0.0241
(0.0224) (0.0461) (0.0108) (0.0212)
Observations 1,722 1,719 2,061 2,068
R-squared 0.0747 0.1984 0.2486 0.1973
Notes: ***p<0.01,**p<0.05,*p<0.1. Robuststandarderrorsinparentheses.
The standard errors are clustered at the district level.
116
Table 3.6: Direct Peer Effects on Possible Mechanism
(1) (2) (3) (4)
VARIABLES Study Hours Study Hours Teacher-Pupil Class
15 hours 20 hours Closeness Atmosphere
Direct Peer Effect 0.0024 0.0019 0.0014 0.0072***
(0.0019) (0.0016) (0.0018) (0.0026)
Mean of Peers’ Rank (%) 59.89 59.89 59.97 59.97
Observations 1,718 1,718 1,617 1,617
R-squared 0.6838 0.6469 0.6275 0.6756
Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses.
The standard errors are clustered at the district level.
117
Table 3.7: Tracking Effects on Study Hours
(1) (2)
VARIABLES 20 hours 20 hours
Treat High -0.1031** 0.1080*
(Tracking effect) (0.0439) (0.0596)
Treat High Above half 0.1511***
(0.0441)
Observations 2,271 2,271
R-squared 0.5345 0.5367
Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust stan-
darderrorsinparentheses. Thestandarderrorsareclus-
tered at the district level.
118
Table 3.8: The Low Types Overestimate Themselves Under Tracking
(1) (2) (3)
VARIABLES Korean Math English
Treat High 0.1201 0.1253** 0.1150**
(Tracking effect) (0.0721) (0.0519) (0.0483)
Treat High Above half -0.1226* -0.1228*** -0.1271**
(0.0613) (0.0422) (0.0503)
Observations 2,274 2,274 2,273
R-squared 0.6522 0.7000 0.6736
Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard
errors in parentheses. The standard errors are clustered at
the district level.
119
Table 3.9: Evidence that high school enrollment is not affected by treatment, GOMS
(1) (2)
VARIABLES High school Enrollments High school Enrollments
Treatment After 306.8419
(282.8117)
Treatment Year 2000 -23.1128
(177.1945)
Treatment Year 2001 118.9426
(206.3606)
Treatment Year 2002 110.4352
(288.9192)
Treatment Year 2003 342.2591
(390.1286)
Treatment Year 2004 563.9044
(434.2090)
Observations 388 388
Data: KESS
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
120
Bibliography
Abdulkadiroğlu, A., Pathak, P. A., Roth, A. E. andSönmez, T. (2005). The boston
public school match. American Economic Review, pp. 368–371.
Abdulkadiroglu, A. and Sönmez, T. (2003). School choice: A mechanism design ap-
proach. The American Economic Review, 93 (3), 729–747.
Ahlfeldt, G. M. andFeddersen, A. (2018). From periphery to core: measuring agglom-
eration effects using high-speed rail. Journal of Economic Geography, 18 (2), 355–390.
—, Redding, S. J., Sturm, D. M. and Wolf, N. (2015). The economics of density:
Evidence from the berlin wall. Econometrica, 83 (6), 2127–2189.
Akbar, P. A., Couture, V., Duranton, G. and Storeygard, A. (2018). Mobility
and congestion in urban India. Working paper.
Aksoy, C. G., Berkay, O. and Philip, J. (2019). Robots and the gender pay gap: Evi-
dence from europe. In 2nd IZA/CREA Workshop: Exploring the Future of Work, vol. 5.
Althoff, L.,Eckert, F.,Ganapati, S.andWalsh, C.(2020).Thecityparadox: Skilled
services and remote work.
Anas, A. (2020). The cost of congestion and the benefits of congestion pricing: A general
equilibrium analysis. Transportation Research Part B, 136, 110–137.
Bae, B. (2014). Report on construction of new bok-dsge model for economic forecasting and
policy analysis.
Baek, J. and Park, W. (2015). The impact of improved passenger transport system on
manufacturing plant productivity. KDI School of Pub Policy & Management Paper, (15-
02).
Barrero, J. M., Bloom, N. and Davis, S. J. (2020). Covid-19 is also a reallocation
shock. Tech. rep., National Bureau of Economic Research.
Bartik, A. W., Cullen, Z. B., Glaeser, E. L., Luca, M. and Stanton, C. T. (2020).
What jobs are being done at home during the COVID-19 crisis? Evidence from firm-level
surveys. Tech. rep., National Bureau of Economic Research.
Bartik, T. J. (1991). Who benefits from state and local economic development policies?
121
Baum-Snow, N. (2007). Did highways cause suburbanization? The Quarterly Journal of
Economics.
—, Henderson, J. V., Turner, M. A., Zhang, Q. and Brandt, L. (2020). Does
investment in national highways help or hurt hinterland city growth? Journal of Urban
Economics, 115, 103124.
Bayer, P., Ferreira, F. and McMillan, R. (2007). A unified framework for measuring
preferencesforschoolsandneighborhoods.Journal of Political Economy,115(4), 588–638.
Behrens, K., Kichko, S. and Thisse, J.-F. (2021). Working from home: Too much of a
good thing?
Benabou, R.(1993).Workingsofacity: location,education,andproduction. The Quarterly
Journal of Economics, 108 (3), 619–652.
— (1996). Equity and efficiency in human capital investment: the local connection. The
Review of Economic Studies, 63 (2), 237–264.
Bento, A., Hall, J. D. and Heilmann, K. (2020). Estimating congestion externalities
using big data, working Paper.
Berkes, E.andGaetani, R.(2021).Income segregation and rise of the knowledge economy.
Tech. rep.
Birdsall, N. (1985). Public inputs and child schooling in brazil. Journal of development
Economics, 18 (1), 67–86.
Black, D. A., Kolesnikova, N. and Taylor, L. J. (2014). Why do so few women work
in new york (and so many in minneapolis)? labor supply of married women across us
cities. Journal of Urban Economics, 79, 59–71.
Blau, F. D. and Kahn, L. M. (2017). The gender wage gap: Extent, trends, and expla-
nations. Journal of Economic Literature, 55 (3), 789–865.
Bloom, N., Liang, J., Roberts, J. and Ying, Z. J. (2015). Does working from home
work? evidence from a chinese experiment. The Quarterly Journal of Economics, 130 (1),
165–218.
Booij, A. S., Leuven, E. and Oosterbeek, H. (2017). Ability peer effects in university:
Evidence from a randomized experiment. The Review of Economic Studies, 84 (2), 547–
578.
Brunello, G. and Checchi, D. (2007). Does school tracking affect equality of opportu-
nity? new international evidence. Economic Policy, 22 (52), 782–861.
Card, D. and Giuliano, L. (2016). Can tracking raise the test scores of high-ability
minority students? The American Economic Review, 106 (10), 2783–2816.
122
Carrell, S. E., Sacerdote, B. I. and West, J. E. (2013). From natural variation
to optimal policy? the importance of endogenous peer group formation. Econometrica,
81 (3), 855–882.
Charnoz, P., Lelarge, C. and Trevien, C. (2018). Communication costs and the in-
ternal organisation of multi-plant businesses: evidence from the impact of the french
high-speed rail. The Economic Journal, 128 (610), 949–994.
Chauvin, J. P. (2017). Gender-segmented labor markets and the effects of local demand
shocks. Job Market Paper, Harvard University.
Choi, E. J., Moon, H. R. and Ridder, G. (2014). Estimation of an education produc-
tion function under random assignment with selection. The American Economic Review,
104 (5), 206–211.
Combes, P.-P., Duranton, G. and Gobillon, L. (2018). The Costs of Agglomeration:
House and Land Prices in French Cities. The Review of Economic Studies, 86 (4), 1556–
1589.
Couture, V., Duranton, G. and Turner, M. A. (2018). Speed. Review of Economics
and Statistics, 100 (4), 725–739.
Davis, M. A.,Ghent, A. C.andGregory, J. M.(2021a).The Work-at-Home Technology
Boon and its Consequences. Tech. rep., National Bureau of Economic Research.
—, — and — (2021b). The Work-at-Home Technology Boon and its Consequences. Working
Paper 28461, National Bureau of Economic Research.
— and Ortalo-Magné, F. (2011). Household expenditures, wages, rents. Review of Eco-
nomic Dynamics, 14 (2), 248 – 261.
Delventhal, M. and Parkhomenko, A. (2020). Spatial implications of telecommuting.
Available at SSRN 3746555.
Delventhal, M. J., Kwon, E. and Parkhomenko, A. (2021). How do cities change
when we work from home? Journal of Urban Economics, p. 103331.
Diamond, R. (2016). The determinants and welfare implications of us workers’ diverging
location choices by skill: 1980-2000. American Economic Review, 106 (3), 479–524.
Dingel, J. I. and Neiman, B. (2020). How Many Jobs Can be Done at Home? Working
Paper 26948, National Bureau of Economic Research.
Donaldson, D. (2018). Railroads of the raj: Estimating the impact of transportation
infrastructure. American Economic Review, 108 (4-5), 899–934.
— and Hornbeck, R. (2016). Railroads and american economic growth: A “market access”
approach. The Quarterly Journal of Economics, 131 (2), 799–858.
123
Dong, X., Zheng, S. and Kahn, M. E. (2020). The role of transportation speed in
facilitating high skilled teamwork across cities. Journal of Urban Economics, 115, 103212.
Duflo, E., Dupas, P. and Kremera, M. (2011). Peer effects, teacher incentives, and
the impact of tracking: Evidence from a randomized evaluation in kenya. The American
Economic Review, 101 (5), 1739–1774.
Duranton, G., Morrow, P. M. and Turner, M. A. (2014). Roads and trade: Evidence
from the us. Review of Economic Studies, 81 (2), 681–724.
Dustmann, C., Puhani, P. A. and Schönberg, U. (2017). The long-term effects of early
track choice. The Economic Journal, 127 (603), 1348–1380.
Faber, B. (2014). Trade integration, market size, and industrialization: evidence from
china’s national trunk highway system. Review of Economic Studies, 81 (3), 1046–1070.
Farré, L., Jofre-Monseny, J. and Torrecillas, J. (2020). Commuting time and the
gender gap in labor market participation.
Favilukis, J., Mabille, P. and Van Nieuwerburgh, S. (2019). Affordable Housing and
City Welfare. Working Paper 25906, National Bureau of Economic Research.
Gamoran, A. (1992). The variable effects of high school tracking. American Sociological
Review, pp. 812–828.
— and Mare, R. D. (1989). Secondary school tracking and educational inequality: Com-
pensation, reinforcement, orneutrality? American journal of Sociology,94(5), 1146–1183.
Garcia-López, M.-À., Holl, A. and Viladecans-Marsal, E. (2015). Suburbanization
and highways in spain when the romans and the bourbons still shape its cities. Journal of
Urban Economics, 85, 52–67.
Garlick, R. (2018). Academic peer effects with different group assignment policies: Res-
idential tracking versus random assignment. American Economic Journal: Applied Eco-
nomics.
Glaeser, E. L. et al. (2008). Cities, agglomeration, and spatial equilibrium.
Global Workplace Analytics (2020). Work From Home Experience Survey, 2020.
Global Workplace Analytics.
Goldin, C. and Katz, L. F. (2016). A most egalitarian profession: pharmacy and the
evolution of a family-friendly occupation. Journal of Labor Economics, 34 (3), 705–746.
Gorback, C. (2020). Essays in Urban and Real Estate Economics. Ph.D. thesis, University
of Pennsylvania.
Greenwood, J., Guner, N., Kocharkov, G. and Santos, C. (2016). Technology and
the changing family: A unified model of marriage, divorce, educational attainment, and
married female labor-force participation. American Economic Journal: Macroeconomics,
8 (1), 1–41.
124
—, Seshadri, A. and Yorukoglu, M. (2005). Engines of liberation. The Review of Eco-
nomic Studies, 72 (1), 109–133.
Gupta, A., Peeters, J., Mittal, V. and Van Nieuwerburgh, S. (2021). Flattening
the curve: Pandemic-induced revaluation of urban real estate. Available at SSRN 3780012.
Hahn, S., Sung, T. and Baek, J. (2008). Mixing versus sorting: Entering top universities.
Economics Letters, 100 (1), 43–46.
Hanushek, E. A. and Rivkin, S. G. (2006). Teacher quality. Handbook of the Economics
of Education, 2, 1051–1078.
— et al. (2006). Does educational tracking affect performance and inequality? differences-
in-differences evidence across countries. The Economic Journal, 116 (510).
Heblich, S., Redding, S. J. and Sturm, D. M. (2020). The making of the modern
metropolis: evidence from london. The Quarterly Journal of Economics, 135 (4), 2059–
2133.
Heuermann, D. F. and Schmieder, J. F. (2019). The effect of infrastructure on worker
mobility: evidence from high-speed rail expansion in germany. Journal of economic geog-
raphy, 19 (2), 335–372.
Kahn, M. E. (2000). The environmental impact of suburbanization. Journal of Policy Anal-
ysis and Management, 19 (4), 569–586.
Kang, C., Park, C. and Lee, M.-J. (2007). Effects of ability mixing in high school on
adulthood earnings: quasiexperimental evidence from south korea. Journal of Population
Economics, 20 (2), 269–297.
Kim, S. and Lee, J.-H. (2003). The secondary school equalization policy in south korea.
Unpublished paper, University of Wisconsin-Milwaukee.
— and — (2010). Private tutoring and demand for education in south korea. Economic
Development and Cultural Change.
Kim, T., Lee, J.-H. and Lee, Y. (2008). Mixing versus sorting in schooling: Evidence from
the equalization policy in south korea. Economics of Education Review, 27 (6), 697–711.
Larson, W. andZhao, W. (2017). Telework: Urban form, energy consumption, and green-
house gas implications. Economic Inquiry, 55 (2), 714–735.
Le Barbanchon, T., Rathelot, R. and Roulet, A. (2020). Gender differences in job
search: Trading off commute against wage.
Lennox, J. et al. (2020). More working from home will change the shape and size of cities.
Centre of Policy Studies/IMPACT Centre Working Papers g-306, Victoria University,
Centre of Policy Studies/IMPACT Centre.
125
Lin, Y. (2017). Travel costs and urban specialization patterns: Evidence from china’s high
speed railway system. Journal of Urban Economics, 98, 98–123.
Liu, S. and Su, Y. (2020a). The geography of jobs and the gender wage gap. Available at
SSRN.
— and — (2020b). The impact of the covid-19 pandemic on the demand for density: Evi-
dence from the us housing market. Available at SSRN 3661052.
Mas, A. and Pallais, A. (2020). Alternative work arrangements. Tech. rep., National
Bureau of Economic Research.
Maurer-Fazio, M., Connelly, R., Chen, L. and Tang, L. (2011). Childcare, eldercare,
and labor force participation of married women in urban china, 1982–2000. Journal of
Human Resources, 46 (2), 261–294.
Moretti, E.(2010).Local labor markets.Tech.rep.,NationalBureauofEconomicResearch.
Morten, M. and Oliveira, J. (2016). Paving the way to development: Costly migration
and labor market integration. Tech. rep., National Bureau of Economic Research.
Muresan, A. (2016). Who lives largest? the growth of urban american homes in the last
100 years. Press Release.
Nevo, A. and Rosen, A. M. (2012). Identification with imperfect instruments. Review of
Economics and Statistics, 94 (3), 659–671.
Nilles, J. (2000). Telework in the us: Telework america survey 2000. Washington, DC:
International Telework Association and Council.
Owens, R., Rossi-Hansberg, E. andSarte, P.-D. (2020). Rethinking detroit. American
Economic Journal: Economic Policy, 12 (2), 258–305.
Ozimek, A. (2020). Remote workers on the move, upwork Economist Report.
Pop-Eleches, C. and Urquiola, M. (2013). Going to a better school: Effects and behav-
ioral responses. The American Economic Review, 103 (4), 1289–1324.
Qin, Y. (2017). ‘no county left behind?’the distributional impact of high-speed rail upgrades
in china. Journal of Economic Geography, 17 (3), 489–520.
Redding, S. J. and Turner, M. A. (2014). Transportation Costs and the Spatial Orga-
nization of Economic Activity. CEP Discussion Paper, (No. 1277).
Rhee, H.-J. (2008). Home-based telecommuting and commuting behavior. Journal of Urban
Economics, 63 (1), 198–216.
Rice, J. K. (2003). Teacher quality: Understanding the effectiveness of teacher attributes.
ERIC.
126
Roback, J. (1982). Wages, rents, and the quality of life. Journal of political Economy,
90 (6), 1257–1278.
Rosenthal, S. S. and Strange, W. C. (2012). Female entrepreneurship, agglomeration,
and a new spatial mismatch. Review of Economics and Statistics, 94 (3), 764–788.
Rossi-Hansberg, E., Sarte, P.-D. and Schwartzman, F. (2019). Cognitive hubs and
spatial redistribution. Tech. rep., National Bureau of Economic Research.
Safirova, E. (2003). Telecommuting, traffic congestion, and agglomeration: a general equi-
librium model. Journal of Urban Economics, 52 (1), 26–52.
Severen, C. (2018). Commuting, labor, and housing market effects of mass transportation:
Welfare and identification.
— (2021). Commuting, labor, and housing market effects of mass transportation: Welfare
and identification, working Paper.
Shorrocks, A. (2013). Decomposition procedures for distributional analysis: a unified
framework based on the shapley value. The Journal of Economic Inequality, 11 (1), 99–
126.
Slavin, R. E. (1987). Ability grouping and student achievement in elementary schools: A
best-evidence synthesis. Review of Educational Research, 57 (3), 293–336.
Small, K. A. and Verhoef, E. T. (2007). The economics of urban transportation. Rout-
ledge.
Spear, B. D. (2011). NCHRP 08-36, Task 098 Improving Employment Data for Trans-
portation Planning. American Association of State Highway and Transportation Officials.
Stanton, C. T. and Tiwari, P. (2021). Housing Consumption and the Cost of Remote
Work. Tech. rep., National Bureau of Economic Research.
Stock, J. H. and Yogo, M. (2002). Testing for weak instruments in linear IV regression.
Tech. rep., National Bureau of Economic Research.
Summers, A. A. and Wolfe, B. L. (1977). Do schools make a difference? The American
Economic Review, 67 (4), 639–652.
Tsivanidis, N. (2018). The aggregate and distributional effects of urban transit infrastruc-
ture: Evidence from bogotá’s transmilenio. Job Market Paper.
— (2019). The aggregate and distributional effects of urban transit infrastructure: Evidence
from bogotá’s transmilenio, working Paper.
U.S. Census Bureau (2020). American Community Survey, 2012-2016. Washington, DC:
U.S. Census Bureau.
127
Valentinyi, A. and Herrendorf, B. (2008). Measuring factor income shares at the
sectoral level. Review of Economic Dynamics, 11 (4), 820–835.
Vidart, D. (2020). Human capital, female employment, and electricity: Evidence from the
early 20th century united states. Unpublished Working Paper, UCSD.
Wang, L. C. (2015). All work and no play? the effects of ability sorting on students’ non-
school inputs, time use, and grade anxiety. Economics of Education Review, 44, 29–41.
Zheng, S. and Kahn, M. E. (2013). China’s bullet trains facilitate market integration and
mitigate the cost of megacity growth. Proceedings of the National Academy of Sciences,
110 (14), E1248–E1253.
128
Appendix A
Appendix to Chapter 1
A.1 Data Source
A.1.1 Census on Establishment
Begining in 1994, the Census on Establishments is an annual population survey that provides
business information on all 3.3 million enterprises and establishments in South Korea. This adminis-
trative dataset includes each establishment’s business information, such as its’ 5-digit industry code,
number of employees, and geographic information. This detailed information makes it possible to
construct a panel data set of job creation
1
and job destruction
2
with detailed geographic and estab-
lishment information. For the time being, I focus on the number of employees and the number of
establishments at the district-level across different industries or establishments with different sizes.
A.1.2 Population and Housing Census
The Population and Housing Census is an administrative survey that collects information on
all 50 million Koreans and foreign residents in South Korea every five years. The survey includes
information on demographics (age, gender, education level), socio-economic status (job status), as
well as residential location, job location, commuting patterns, and the residential location of the
respondent five years prior to the survey.
1
The number of jobs created each year due to either firms’ entrance or expansion of employment of
incumbents.
2
The number of jobs that disappeared each year due to either exit of firms or declines in employment of
surviving firms.
129
This study uses 4 waves (2000, 2005, 2010, 2015) of the Census data, which contain a 2%
random sample from the population survey. The key variables used in this paper are current
residential location, residential location 5 years prior, commute location, commuting mode, and
other socio-demographic information such as education level, gender, and marriage.
A.1.3 Internal Migration Statistics
The Internal Migration statistics is an administrative dataset of residential migration for estab-
lishing population policies. The data is collected when a household moves to a zip-code outside of
its original zip-code. This makes it possible to track all the across-zipcode migrants in South Korea.
The information is collected at the household-level with demographic information (gender and age)
on all household members who move together. Importantly, this data contains information on the
origin zip-code of each migrant, which lets us track area-to-area migration flows each year.
By aggregating 88,248,353 observations of annual (2001-2015) household-level data to annual
county-level origin-destination (o/d) data, this paper uses of 849,600
3
observations of o/d migra-
tion flows. I focus on the number of migrants, the number of female/male migrants, the number
of migrants across different age groups (under 20s, 20s, 30s, 40s, 50s, and over 60s) to see the
heterogeneous migration patterns across different demographics.
A.1.4 Korean Labor and Income Panel Study, KLIPS
KLIPS (Korean Labor & Income Panel Study) is a longitudinal survey of the labor market and
income activities of households and individuals residing in urban areas. The 1st Wave of the KLIPS
was launched by the KLI (Korea Labor Institute) in 1998, and the survey is conducted annually.
Around 5,000 households and 15,000 individuals are surveyed every year.
3
849,600 = 238 counties *238 counties *15 years. Note that within county migration is also included in
the observation.
130
Table A.1: Summary Statistics of Wage Data (Source: KLIPS)
Variable Obs Mean Std. Dev. Min Max
logwage (pre-HSR) 23,517 4.61 0.61 0.69 7.53
logwage (post-HSR) 28,156 5.29 0.66 1.79 8.29
A.2 Data Construction
A.2.1 District-level Wage Index
Since the administrative wage data during the sample period is not available, wage index at
district-level data is imputed using Korea Labor and Income Panel Study (KLIPS). The individual
panelsurveyprovidesindividuals’richinformationaboutwherepeopleliveandwork,socio-economic
characteristics as well as annual labor earnings. To compute the wage index at district-level for
before and after HSR construction, I use individual-level wage data, and get the district-fixed
effects as the wage index. I make use of four years (2000-2003) for pre-HSR period, and five years
(2011-2015) for post-HSR period. Table A.1 presents the summary statistics for the original source
data.
For an individuali, living in districtj, at timet2fbefore;afterg, the district-level wage ( ^ w
jt
)
is estimated as follows:
w
ijt
= +X
ijt
+ ^ w
jt
+
ijt
(A.1)
where X
ijt
are the vector of individual characteristics such as years of education, age, gender,
and experience.
A.2.2 A Travel Time Matrix between Cities to Cities
In this study, the reduction in transportation costs induced by the KTX expansion is proxied
by the reduction in travel time between cities. I calculate travel time between each county-pair over
each year, which consists of a time-varying travel time matrix with an element located in the rowj;
the columnk represents the minimum travel time from a countyj to a countyk, at a given timet.
4
4
The timed travel between the two districts is assumed to be symmetric.
131
As of 2010, there are 228 districts, which in turn result in 51,756
5
unidirectional relations between
the counties. The travel time within districts is assumed to be zero, as the focus of this paper is on
the inter-city movement rather than intra-city movement.
I simplify the travel time calculation by assuming that the distance between each county pair is
time fixed, whereas the speed of the transportation modes changes over time as the KTX networks
expand. Under this simplifying assumption, calculating the travel time before KTX expansion is
relatively easy as the travel time is the linear distance between the centroid of origin and destination
county divided by the average speed of an automobile or bus, which is assumed to be 90km/hour,
the highway speed limit in South Korea.
6
The changes in travel time induced by KTX expansion are calculated as follows. First, I con-
struct KTX networks by using the Korean Transportation Database, collected by the Korea Trans-
portInstitute(KTDB).Thedatabaseprovidesinformationontheexactlocationofalltrainstations,
railroad networks, and the opening date of each station. Combining this information with the KTX
train schedule available online, I calculate estimated travel time from a station to every other sta-
tion connected by KTX networks.
7
Finally, the travel time from an origin county to a destination
county is calculated as the sum of 1) the estimated time from the centroid of the origin county to
departure KTX station; 2) the estimated travel time by KTX from the departure station to the
arrival station; and 3) the estimated travel time from the arrival KTX station to the centroid of the
Destination county.
Finally, the time-varying minimum travel time matrix is constructed by using the following
algorithm. Whenever there was an expansion in KTX networks (e.g., 2004, 2010, 2011, 2012) I
compare the travel time before the expansion of KTX to the travel time after the expansion and
take the smaller value for every county-pair. Before 2004, as KTX had not yet been introduced,
the minimum travel time is calculated as the linear distance divided by the speed of an automobile.
For 2004, when the KTX was first introduced, the previous travel time is compared to the travel
5
(district-district pairs) - (pairs with its own) = (228*228) - 228.
6
The underlying assumption is that the motor is the closest substitute of the KTX. This is because travel
by other transportation modes such as traditional railway or domestic flights takes longer than by car, in
the South Korean context (KTI, 2013).
7
To simplify the calculation, I first calculate the linear distances between each station and divide the
distance by the average train speed between each departure-arrival station.
132
time accounting for the new KTX networks and the smaller travel time is taken between the two.
Whenever the KTX network is expanded, the algorithm repeats the same procedure for every
county-pair to create the time-varying minimum travel time matrix.
133
A.3 Regression on Migration Flow
Let M
odt
be the number of migrants from region o to region d in year t. The gravity model for
M
odt
takes the following form:
M
odt
=aPop
o
ot
Emp
o
ot
Pop

d
dt
Emp

d
dt
D

odt
(A.2)
where Pop
ot
(Pop
dt
) is the population of an origin o (destination d) at time t, Emp
ot
(Emp
dt
)
is the total employment of origin o (destination d) at time t, and D
odt
means travel costs (in our
context, travel time) between origion (o) and destination (d). The intuition for the model is that
the migration flow between an origin and a destination county-pair increases if the size (population,
employment) of the origin or the destination county is big and the distance between the two is close
enough. Taking the logarithms of equation A.2 we get:
log(M
odt
) =log(a) +
o
log(Pop
ot
) +
d
log(Pop
dt
) +
o
log(Emp
ot
) +
d
log(Emp
dt
) +
log(D
odt
)
(A.3)
Our coefficient of interest in equation (A.3) is
, which can be interpreted as the percentage
changes in the number of bilateral migrants associated with a 1% change in travel time.
However, a naive estimation of equation (A.3) would cause an endogeneity problem. For exam-
ple, if the placement of HSR specifically was targetting counties which grow or decline, then the
causal inference between the travel time reduction and migration flow would be threatened.
Keeping in mind the potential endogeneity issue, the main regression equation is
8
log(M
odt
) = +
log(TravelTime
odt
) +
od
+
ot
+
dt
+
odt
(A.4)
where, log(TravelTime
odt
) is the travel time between origin and destination counties at time
t,
od
is the origin-destination pair fixed effect,
ot
is the origin-year fixed effect, and
dt
is the
destination-year fixed effect.
8
The same empirical specification is also used in Morten and Oliverira (2017)
134
Figure A.1: Population-weighted Centroid
Note: Color of the map represents zip-code-level 2010 population.
135
Figure A.2: KTX Network Expansion
Note: Color represents the population density of districts in 2010. Districts within the thick red lines are
defined as core areas; districts in Seoul metropolitan areas, which are districts of Incheon, Seoul, or
Gyeonggi provinces.
136
Table A.2: History of the KTX Station Expansion
Date Line From/To Project Type Station
Apr/2004
Gyeongbu
Seoul/Daegu* New line
Seoul, Hangshin, Youngdengpo, Gwangmeng
Suwon, Cheonan-Asan, Daejeon, Dong-Daegu
Daegu/Busan** New line Gupo, Milyang, Busan
Honam** Daejeon/Mokpo New line
Yongsan, Seo-Daejeon, Gyeryong, Nonsan
Iksan, Jeongeup, Gwangju-Songjeong, Naju, Mokpo
Dec/2010
Gyeongbu*
Seoul/Daegu Add-station Osong
Daegu/Busan
Add-station Shin-Gyeongju, Gimcheon(Gumi), Ulsan
Improve speed Busan
Gyeongjeon** Milyang/Masan New line Jinyoung, Changwon-Joongang, Changwon, Masan
Oct/2011 Jeonla** Ilsan/Yeosu New line
Jeonju, Namwon, Guryegu, Sooncheon, Yeocheon
Yeosu-expo, Goksung
Dec/2012 Gyeongjeon** Masan/Jingu Add-station Jinju
*High Speed Railroad (V
max
= 305km=h)
**Electrified conventional railway directly connected with HSR (V
max
< 180km=h)
Source: Korail
137
Table A.3: The Inferred impact of HSR on Relative productivity and female labor participation
costs
A. Core Areas
A.1 Impact on the Gender Gap in Employment
Female Labor Participation Cost (1 +T
jt
)
Relative Productivity (

Mjt

Wjt
) Decrease No Changes Increase
Decrease (-) (-) (?)
No Changes (-)
Increase (?)
A.2 Impact on the Gender Gap in Employment
Female Labor Participation Cost (1 +T
jt
)
Relative Productivity (

Mjt

Wjt
) Decrease No Changes Increase
Decrease (?) (-) (-)
No Changes (-)
Increase (?)
A.3 Inferred Impact of HSR
Female Labor Participation Cost (1 +T
jt
)
Relative Productivity (

Mjt

Wjt
) Decrease No Changes Increase
Decrease X X X
No Changes
Increase
138
Table A.4: The Inferred impact of HSR on Relative productivity and female labor participation
costs
B. Noncore Areas
B.1 Impact on the Gender Gap in Employment
Female Labor Participation Cost (1 +T
jt
)
Relative Productivity (

Mjt

Wjt
) Decrease No Changes Increase
Decrease (-) (-) (?)
No Changes (-)
Increase (?)
B.2 Impact on the Gender Gap in Wage
Female Labor Participation Cost (1 +T
jt
)
Relative Productivity (

Mjt

Wjt
) Decrease No Changes Increase
Decrease (?)
No Changes (.)
Increase (?)
B.3 Inferred Impact of HSR
Female Labor Participation Cost (1 +T
jt
)
Relative Productivity (

Mjt

Wjt
) Decrease No Changes Increase
Decrease X
No Changes
Increase
139
Table A.5: Mechanism 1.(2). Effect of HSR on Sectoral Sex Ratio (Male-to-Female Ratio)
Panel A. Male Intensive Sector
(1) (2) (3) (4)
Sex Ratio
(Transportation) (Construction) (Public Admin) (Manufacturing)
Treat
*Non-Seoul 0.379 1.576** 1.038** -0.828**
(0.595) (0.704) (0.368) (0.340)
*Seoul -5.237 4.825** 1.922 -0.621***
(4.481) (2.104) (2.216) (0.127)
Sex Ratio(2000) 10.09 7.56 4.40 2.36
Observations 237 237 237 237
Panel B. Female Intensive Sector
(5) (6) (7) (8)
Sex Ratio
(Retail) (Education) (Medical Service) (Restaurant)
Treat
*Non-Seoul 0.027 0.422*** -0.053 0.035
(0.024) (0.107) (0.037) (0.025)
*Seoul -0.015 0.991 -0.040 0.002
(0.027) (0.913) (0.093) (0.014)
Sex Ratio(2000) 0.88 0.76 0.54 0.50
Observations 237 237 236 237
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are in parentheses. The coefficients are the second
stage regression results of instrumental variable estimation in Section 1.6.2. Dependent variable of column (1)-
(8) is sex ratio of each sector, defined as male employment over female employment. Core areas are defined as
districts in Seoul metropolitan areas, which are districts of Incheon, Seoul, or Gyeonggi provinces. Non-core
areas are all other districts. Standard errors are clustered at province-level.
140
TableA.6: ResidentialSpace, HousingOwnership, andHousingexpenditure(FamilySurvey, 2012)
(1) (2) (3)
Residential Space Home ownership Housing expenditure
(not home-owners)
Omitted Group: Couple, living in non-core
Core
Couple -2.068 -0.148*** 36.461***
(1.339) (0.014) (11.981)
Single Men -42.792*** -0.544*** -33.199*
(1.571) (0.026) (19.667)
Single Women -38.482*** -0.478*** -37.101*
(2.539) (0.032) (20.670)
Others (single mom/dad) -22.065*** -0.361*** 21.562
(1.678) (0.023) (19.883)
Non-Core
Single Men -32.255*** -0.423*** -88.567***
(2.300) (0.028) (15.653)
Single Women -22.529*** -0.270*** -78.640***
(2.537) (0.030) (17.976)
Others (single mom/dad) -15.946*** -0.224*** -44.592***
(1.592) (0.022) (12.502)
Observations 7,755 7,755 3,495
R-squared 0.082 0.096 0.020
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
141
Table A.7: Changes in estimations according to the value of (%)
0 0.1 0.3 0.5 0.7 0.9 1
core LFP cost -1874.03 -166.39 -43.16 -19.80 -10.12 -4.75 -2.83
core Relative Productivity -42.31 -39.42 -35.24 -32.82 -31.39 -30.59 -30.35
non-core LFP cost -1611.42 -154.92 -49.23 -29.16 -21.01 -16.71 -15.27
non-core Relative Productivity -19.52 -17.70 -14.88 -13.01 -11.66 -10.63 -10.20
142
Appendix B
Appendix to Chapter 2
B.1 Data Appendix
B.1.1 Property Price Data
Our commercial and residential property price data comes from DataQuick, which consists of
public records on the population of property transactions and the characteristics of individual prop-
erties. The dataset covers 2,354,535 properties over 2007–2016 in the Los Angeles-Long Beach com-
bined statistical area. The data provides information such as sales price, geographical coordinates,
transaction date, property use, transaction type, number of rooms, number of baths, square-footage,
lot size, year built, etc.
We categorize properties as commercial or residential based on their reported use. Examples of
residential use include “condominium”, “single family residence”, and “duplex”. Examples of com-
mercial use include “hotel/motel”, “restaurant”, and “office building”. Table B.1 provides descriptive
statistics. Table B.2 provides the number of observations in each county over the period of 2007–
2016. Note that the commercial transactions are far less frequent than residential transactions.
We then use the transactions data to estimate hedonic tract-level residential and commercial
propertyindices. Foraresidentialtransactionofapropertyp, intractj inyear-montht, weestimate
ln(P
pjt
) = +X
p
+
t
+
j
+
pjt
; (B.1)
143
Table B.1: Descriptive Statistics
Panel A. Residential Properties
County sqft (mean) sqft (median) sales price, $ (mean) sales price, $ (median)
Los Angeles 1752.25 1499 774734.19 389000
Orange 1969.92 1578 714043.38 495000
Riverside 2046.06 1855 489885.35 246649
San Bernardino 1759.41 1584 345662.41 200000
Ventura 1860.88 1626 569042.40 410000
Panel B. Commercial Properties
County sqft (mean) sqft (median) sales price, $ (mean) sales price, $ (median)
Los Angeles 20687.28 5203 5661399.99 1300000
Orange 16447.48 5329 3879699.73 1260000
Riverside 1329.38 1201 1813988.76 590000
San Bernardino 19486.08 3541 2472923.09 522000
Ventura 12087.09 4565 3513023.97 982500
Table B.2: Number of Transactions by County and Property Type
Los Angeles Orange Riverside San Bernardino Ventura Total
Residential 909,954 330,689 557,204 363,173 105,518 2,266,538
Commercial 47,408 12,084 14,045 11,099 3,361 87,997
where P
pjt
is the price per square foot; X
p
contains property characteristics including property
use, transaction type, number of rooms, number of baths, lot size, and year built; and
t
is the
year-month fixed effect. Then the residential price index in tractj corresponds to
j
, the tract fixed
effect.
Becausecommercialtransactionsarelessnumerous, formanyCensustractsweonlyobservevery
few or no transactions in the period of interest. To overcome this issue, we calculate commercial
property indices at the Public Use Microdata Area (PUMA)-level.
1
There are 123 PUMAs in our
sample. For a commercial property transaction of a propertyp, in tractj of PUMAg in year-month
t, we estimate
ln(P
pgjt
) = +X
p
+
t
+
g
+
pgjt
; (B.2)
where P
pgjt
is the price per square foot; X
p
is property characteristics including property use; and

t
is the year-month fixed effect. The commercial price index in PUMA g corresponds to
g
, which
1
PUMA is a geographic unit used by the US Census for providing statistical and demographic informa-
tion. Each PUMA contains at least 100,000 people.
144
is the PUMA fixed effect.
B.1.2 Wage Data
Our source of wage data is the Census Transportation Planning Products (CTPP). CTPP data
sets produce tabulations of the American Community Survey (ACS) data, aggregated at the Census
tract level. We use the data reported for years 2012 to 2016. We use the variable “earnings in the
past 12 months (2016 $), for the workers 16-year-old and over,” which is based on the respondents’
workplace locations. The variable provides the estimates of the number of people in each earning
bin in each workplace tract. Table B.3 provides an overview of the number of observations in each
bin for the five counties included in our study.
Table B.3: Number of observations in each earnings bin
Income Bin Los Angeles Orange Riverside San Bernardino Ventura
$1 to $9,999 or loss 416,469 147,484 86,219 85,854 34,973
$10,000 to $14,999 279,132 90,871 51,959 52,605 21,143
$15,000 to $24,999 541,649 168,284 97,184 97,059 40,458
$25,000 to $34,999 440,298 146,337 79,994 81,911 34,829
$35,000 to $49,999 493,434 170,364 77,170 87,969 37,487
$50,000 to $64,999 387,533 138,932 57,409 62,487 27,979
$65,000 to $74,999 176,079 63,244 24,869 27,687 13,895
$75,000 to $99,999 308,994 114,436 39,159 44,409 23,871
$100,000 or more 486,179 189,108 44,925 43,158 36,346
No earnings 520 134 144 85 55
Earnings in the past 12 months (2016$) (Workers 16 years and over), based on workplace location, Source:
CTPP.
We calculate mean tract-level labor earnings as
^ w
j
=

b
nworkers
b;j
^ meanw
b

b
nworkers
b;j
; (B.3)
where nworkers
b;j
is the number of workers in bin b in tract j, and ^ meanw
b
is mean earnings in
bin b for the entire L.A. urban area, calculated from the ACS microdata.
Next, to control for possible effects of workers’ heterogeneity on tract-level averages, we run the
145
following Mincer regression,
^ w
j
= +
1
age
j
+
2
sexratio
j
+
r

2;r
race
r;j
+
i

3;i
ind
i;j
+
o

4;o
occ
o;j
+
j
; (B.4)
whereage
j
is the average age of workers;sexratio
j
is the proportion of males to females in the labor
force;race
r;j
is the share of racer2fAsian;Black;Hispanic;Whiteg;ind
i;j
is the share of jobs in
industryi;occ
o;j
is share of jobs in occupationo in tractj.
2
Finally, the estimated tract-level wage
index corresponds to the sum of the constant and the tract fixed effect, ^ + ^
j
. Table B.4 presents
summary statistics for the estimated tract-level earnings.
Table B.4: Descriptive statistics: the estimated tract-level earnings, by county
Obs Mean Std. Dev. Min Max
Los Angeles 2,339 61203.81 13589.54 21376.82 170987.1
Orange 582 63455.76 11197.14 24120.39 113428.8
Riverside 452 61477.51 13606.08 17286.49 138802.9
San Bernardino 369 59823.33 12741.2 21101.49 132544.9
Ventura 172 61034.83 10709.51 29174.4 89796.23
Earnings in the past 12 months (2016$) (Workers 16 years and over), based on workplace location, Source:
CTPP.
B.1.3 Commuting Time Data
The CTPP database provides commuting time data for 270,436 origin-destination tract pairs in
theLosAngeles-LongBeachCombinedStatisticalAreafor2012-2016. Thereare15,342,889possible
trajectories, and the LODES data for 2012-2016 reports positive commuting flows for 5,647,791 of
them. We follow the practice recommended by Spear (2011) and use LODES data as a measure of
2
We use the following industry categories: Agricultural; Armed force; Art, entertainment, recreation,
accommodation; Construction; Education, health, and social services; Finance, insurance, real estate; Infor-
mation; Manufacturing; Other services; Professional scientific management; Public administration, Retail.
We use the following occupation categories: Architecture and engineering; Armed Forces; Arts, design, en-
tertainment, sports, and media; Building and grounds cleaning and maintenance; Business and financial
operations specialists; Community and social service; Computer and mathematical; Construction and ex-
traction; Education, training, and library; Farmers and farm managers; Farming, fishing, and forestry; Food
preparation and serving related; Healthcare practitioners and technicians; Healthcare support; Installation,
maintenance, and repair; Legal; Life, physical, and social science; Management; Office and administrative
support; Personal care and service; Production;Protective service; Sales and related.
146
commuting flows and CTPP data to provide information on commute times.
Table B.5: Commuting time coverage, by distance
N. of trajectories % covered by time data % w/ observed positive flows N. of commuters
< 1 km 10,105 60.8% 96.4% 239,188
< 2 km 36,205 40.5% 93.3% 410,571
< 5 km 188,047 24.4% 86.9% 1,088,797
< 10 km 649,005 15.0% 79.9% 2,248,646
< 20 km 2,099,417 8.2% 69.8% 3,995,134
< 40 km 5,549,775 4.3% 54.4% 5,508,736
< 80 km 10,752,785 2.5% 43.4% 6,515,595
All 15,342,889 1.8% 36.8% 6,935,765
Table B.6: Commuting time coverage, by N. of commuters
N. of trajectories % covered by time data N. of commuters
> 100 commuters 1,778 94.4% 259,259
> 50 commuters 8,678 89.9% 723,849
> 25 commuters 27,833 82.2% 1,380,081
> 10 commuters 96,177 63.7% 2,417,561
> 5 commuters 220,555 46.5% 3,289,529
> 1 commuters 1,108,755 17.9% 5,247,370
All > 0 5,647,791 4.8% 6,935,760
TableB.5summarizesCTPPdatacoveragebytrajectorydistance. TableB.6summarizesCTPP
data coverage by trajectory and the number of commuters observed using that trajectory. These
tables show that CTPP has the greatest coverage of high-volume short-distance trajectories, just
as Spear (2011) observes and just as would be expected from a dataset based on a partial sample.
The CTPP data places commuting times into 10 bins: less than 5 minutes, 5 to 14 minutes, 15 to
19 minutes, 20 to 29 minutes, 30 to 44 minutes, 45 to 59 minutes, 60 to 74 minutes, 75 to 89 minutes,
90 or more minutes, and work from home. In order to get as accurate commute times as possible
for the set of primitive connections of the network, we drop all home-workers, who are irrelevant
for transit times. We drop workers in the top time bin, because this bin has no upper bound and
so the mean may vary substantially across trajectories. We assign mean commute times to all the
remaining bins as the mid-points between the bin bounds. We then drop all observations which
report an average commuting speed that is either less than 8 kilometers per hour, a brisk walking
pace, or more than 70 miles per hour (112.7 kilometers per hour), the standard rural freeway speed
147
limit in the United States. Finally, we calculate tract-pair mean commuting times as the average
of the mean commuting times in each bin weighted by the share of commuters on that tract-pair
reporting times in each bin. Table B.7 provides a summary of the overall share of commuters in
each bin before and after the cleaning steps described above, and the mean commute time assigned
to each bin.
Table B.7: Commuting time bins
share in raw data share in cleaned data bin mean time
< 5 min 1.6% 0.9% 5
5-14 min 19.4% 18.9% 10
15-19 min 14.0% 15.7% 17
20-29 min 19.1% 22.5% 25
30-44 min 20.5% 24.4% 37
45-59 min 8.0% 9.6% 52
60-74 min 6.1% 6.9% 67
75-89 min 0.9% 1.0% 82
> 90 min 2.8% 0 ??
work from home 7.6% 0 n/a
The previous cleaning steps eliminate observations for 36,279 trajectories, and we are left with
commuting time data for 234,157 origin-destination pairs. We then find that there are 211,521 paths
for which a commuting time estimate exists for the outbound route but not the reverse. We impute
commute times for these missing return journeys, assuming that they can be completed in the same
time as the outbound trajectories. This set of connections is then almost enough to connect all
tracts–there are only a set of eight tracts that are still detached from the rest of the network. In
order to remedy this, we create a connection at the mean travel speed of 31.3 kilometers per hour
between these left-out tracts and any tracts within a radius twice as large as the hypothetical radius
of the tract if its land area formed a circle.
3
The final directed network contains 447,277 directed paths. We use the Dijkstra’s algorithm to
calculate the fastest path through this network for each origin-destination pair. We assume that
these calculated times represent the time required to travel from tract centroid to tract centroid.
We then add time to each trajectory to represent the time need to travel from place of residence
within tract to residence tract centroid, and from workplace tract centroid to workplace within the
3
2
p
landarea=
148
tract. Naturally, these times are proportional to tract land area–larger tracts should on average
require more internal travel time. Specifically, we assume that the “internal” distance traveled on
each end of the trip is equal to the hypothetical average straight-line distance from any point in the
tract to the tract centroid, if the tract were a circle.
4
We then assume that each of these distances
is traveled at twice the overall average commuting speed in the cleaned data of 31.3 kilometers
per hour. For the vast majority of tracts this adds a negligible amount to commuting time–two
minutes or less. For a handful of very large tracts it adds considerable travel time–up to half an
hour. We think that this is reasonable given the time that is required to travel within these much
larger tracts. These origin-destination distribution effects are also applied to self-commute times,
so that a worker that lives and works in the same tract will still have to spend some time traveling
to their workplace–more time for larger tracts.
B.1.4 Summary
Table B.8 gives summary statistics by tract for seven key variables: residential density; em-
ployment density; wage by workplace weighted by employees; average constant-quality price of one
square foot of residential floorspace; average constant-quality price of one square foot of commercial
floorspace; average commute time by residence tract; and average commute distance by residence
tract.
Table B.8: Data overview
Mean Median St. Dev. Max. N. Obs.
Residents/km
2
1,621.0 1,380.2 1,376.6 15,929.3 3,847
Workers/km
2
1,285.5 578.7 3,960.5 157,995.7 3,847
Wages ($$, weight by employees) 59,342 57,293 17,903 125,871 3,847
Res. price/sq ft ($$, weight by residents) 177 159 169 4,549 3,847
Comm. price/sq ft ($$, weight by employees) 345 316 170 3,510 3,847
Av. commute time (min, weight by residents) 28.4 26.4 6.9 96.9 3,847
4 2
3
p
landarea=
149
B.2 Model Details
B.2.1 Floorspace Markets
Floorspace Supply
Land-market clearing and profit maximization imply that the equilibrium supply of floorspace
is
H
i
=
i
(H
i
) ((1) q
i
)
1

L
i
: (B.5)
Solving this expression for H
i
and using the definition of construction efficiency
i
(H
i
), yields
H
i
=
((1) q
i
)
1

L
i
1 + ((1) q
i
)
1

L
i
=

H
i
: (B.6)
Floorspace Demand
From equation (2.3), the probability that an individual who commutes a fraction of days works
in j, conditional on living in i, is given by

ijji
() =
E
j
w
ij
()

d
ij
()

I
P
s=1
E
s
w
rs
()

d
is
()

: (B.7)
Define ~ w
i
as the average wage earned by residents of location i. This is given by
~ w
i

X
j2I
"
w
C
j

ijji
(1)N
Ri
(1)
N
Ri
+w
T
ij

ijji
(
T
)N
Ri
(
T
)
N
Ri
#
: (B.8)
Therefore, the demand for residential and home-office floorspace is given by
H
Ri
=

(1 +) ~ w
i
q
Ri
N
Ri
: (B.9)
150
The demand for home offices is
H
Ti
=

1
T
q
Ri
1

T
X
j2I
(A
j
)
1

T
N
T
ij
: (B.10)
Finally, the demand for commercial floorspace is given by
H
Wj
=

(1)A
j
q
Wj
1

N
C
Wj
: (B.11)
B.2.2 Factor Incomes and Transfers
The city-wide total land income is
X
i2I
l
i

i
: (B.12)
Income generated by land and the consumption good sold for the purposes of real estate devel-
opment is redistributed to all workers, proportionally to their incomes. The transfers increase labor
income by a fraction of which is equal to
=
P
i2I
(l
i

i
+K
i
)
P
i2I
~ w
i
N
Ri
+
P
i2I
(l
i

i
+K
i
)
: (B.13)
B.2.3 Welfare
The expected utility enjoyed by a resident of the city is given by
U

1

"
X
r2I
X
s2I
X
r
E
s
h
(1 )

e
trs
(1 +)w
C
s
q

Rr

+

(1 +e
trs
)(1 +)w
T
rs
q

Rr

i
# 1

;
(B.14)
where ()isthegammafunction. Notethattheexpectedutilityisdefinedbeforethetelecommuting
lottery and before the location preference shocks realize.
A consumption-equivalent measure of change in welfare is given by . This quantity represents
the percentage amount by which the composite consumption of goods and housing, c
1
h

, must
change in order to make the expected utility in the benchmark economy equal to the expected utility
151
in the counterfactual economy. Note that in this model the composite consumption is proportional
to wages. Let “~” denote variables in the counterfactual economy. Then must satisfy
"
X
r2I
X
s2I
X
r
E
s
h
(1 )

e
trs
(1 +)(1 + )w
s
q

Rr

+

(1 +e
trs
)(1 +)(1 + )w
T
rs
q

Rr

i
# 1

=
"
X
r2I
X
s2I
~
X
r
E
s
h
(1 )

e

~
trs
(1 + ~ ) ~ w
s
~ q

Rr

+

(1 +e

~
trs
)(1 + ~ ) ~ w
T
rs
~ q

Rr

i
# 1

:
It follows that the change in welfare is a function of the ratio of expected utilities in the coun-
terfactual and the benchmark economies,
=
~
U
U
1: (B.15)
B.3 Structural Residuals
The amounts of commuting workers and residents are related as
N
Wj
(1) =
X
i2I

ijji
(1)N
Ri
(1); (B.16)
Let
^
E
j
E
j

w
C
j

. From equations (B.7) and (B.16),
^
E
j
can be defined implicitly as:
^
E
j
=N
Wj
(1)
0
B
@
X
i2I
e
t
ij
P
s2I
^
E
s
e
t
is
N
Ri
(1)
1
C
A
1
; (B.17)
where N
Wj
and N
Ri
are observed tract-level employment and residential populations, and t
ij
are
observed average commuting times from tract i to tract j. Since we do not observe how many
workers telecommute in each tract and since the share of telecommuters in the data is small (3.74%
of workforce), we perform this and the following calculations assuming that all workers commute to
their jobs. A vector
^
E is solved recursively using equation (B.17) and then the vector of residuals
E is recovered as E
j
=
^
E
j

w
C
j

, using observed tract-level wages.
A similar procedure is applied to solve for vector X. First, let
^
X
j
X
j
q

Rj
.
^
X
j
can be defined
152
implicitly as:
^
X
i
=N
Ri
0
B
@
X
j2I
e
t
ij
P
r2I
^
X
r
e
t
rj
N
Wj
1
C
A
1
: (B.18)
The vector
^
X is solved recursively using equation (B.18) and then the vector of residuals X is
recovered as X
j
=
^
X
j
q

Rj
, using observed tract-level prices of residential floorspace. Then the
exogenous part of local amenities, x
j
, can be recovered using equation (2.13) and the data on local
residential population and land area.
The vector of local productivities A can be solved for using (2.7) and the data on wages and
commercial floorspace prices as follows:
A
j
=

w
C
j

!

q
Wj
1

1
: (B.19)
Thentheexogenousparta
j
canberecoveredusingequation(2.12)andthedataonlocalemployment
and land area.
Since we observe commercial and residential floorspace prices for all Census tracts, we can
calculate the zoning parameter
i
as

i
=
q
Wi
q
Ri
: (B.20)
To calculate
i
, we replaceq
Wi
andq
Ri
with tract-level quality adjusted indexes of commercial and
residential prices,
com
j
and
res
j
, respectively, as described in Appendix B.1.
Finally, inordertorecover

H
i
, weusemarketclearingconditionsforlandandfloorspace(L
i
=
i
and equation B.6). Combining them, we can recover

H
i
from the following relationship:

H
i
=

((1) q
i
)
1

i

((1) q
i
)
1

i
=H
i
1
; (B.21)
where
i
is the observed land area and H
i
=H
Ri
+H
Wi
+H
Ti
is the total demand for floorspace
in tract i.
Figure B.1 maps the recovered values for three key structural paramters: the exogenous com-
ponent of residential amenities, x
i
, the exogenous component of productivity, a
i
, and exogenous
employment amenities, E
i
.
153
Figure B.1: Structural residuals
10
8
6
4
2
0
2
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
10
8
6
4
2
0
Note: Exogenous residential amenities (top figure), exogenous productivities (middle figure) and exogenous
employment amenities (bottom figure).
154
B.4 Additional Results of Counterfactual Experiments
B.4.1 Land Use
When the fraction of telecommuters rises, land use becomes more specialized. Figure B.2 shows
that in the economy with more widespread telework, commercial development becomes relatively
more prevalent in core areas and less prevalent in the periphery. In addition, both types of develop-
ment become more concentrated in space. As a consequence, the numbers of primarily residential
and primarily commercial tracts increase, while the number of mixed tracts goes down (right panel
of Figure B.3).
5
5
We label a tract as commercial if the share of commercial floorspace in the tract is more than 3 times
the share of the average tract. Similarly, we label a tract as residential if the share of commercial floorspace
in the tract is less than 1/3 of the share of the average tract. All other tracts are labeled mixed.
155
Figure B.2: Land Use
0
5%
10%
15%
20%
40%
60%
80%
100%
..
0
5%
10%
15%
20%
40%
60%
80%
100%
..
Note: Benchmark (upper figure) and the = 0:33 counterfactual (lower figure). Maps show the fraction of
commercial floorspace in each tract, varying from 0 (green) to 1 (brown). See main text for details.
Figure B.3: Land Use Specialization
0.0374 0.1 0.15 0.2 0.25 0.3 0.33
Fraction of telecommuters ( )
100
120
140
160
180
200
1280
1300
1320
1340
# commercial tracts (LHS)
# residential tracts (RHS)
Note: The figure shows the number of commercial and residential tracts, as a function of the share of
teleworkers. See main text for details.
156
B.4.2 Job access
In large, sprawled and congested cities, such as Los Angeles, good jobs are often inaccessible for
households who live on the periphery. To study how a shift to telecommuting impacts job access,
we calculate commuter market access for each tract as CMA
i
=
P
j2I
(w
j
e
t
ij
)

. We find that, as
the number of teleworkers grows, the average job access increases for those who keep commuting
(left panel of Figure B.4). Moreover, in the counterfactual economy the elasticity of housing prices
with respect to the market access halves, meaning that places with better access to jobs command
a lower price premium (right panel of Figure B.4).
Figure B.4: Access to jobs
0.0374 0.1 0.15 0.2 0.25 0.3 0.33
Fraction of telecommuters ( )
-5
0
5
10
15
Change, %
CMA
1 2 3 4 5 6 7
Log commuter market access
-3
-2
-1
0
1
2
Log housing prices
Benchmark, elasticity: 0.31
Counterfactual, elasticity: 0.14
Note: Left panel shows the weighted-average commuter market access for each level of . Right panel shows
the relationship between CMA
i
and housing prices q
Ri
in the benchmark economy and the counterfactual
economy with = 0:33.
B.4.3 Breakdown of residential and job changes by worker type
In the context of the counterfactual exercise, there are three types of workers: continuing com-
muters, old telecommuters, and new telecommuters. In Figures B.5 and B.6, we show changes in
residence and jobs for each category separately.
157
Figure B.5: Residence changes for continuing commuters, old telecommuters, and new telecom-
muters
1000
500
0
500
1000
residents per sq. km.
1000
500
0
500
1000
residents per sq. km.
1000
500
0
500
1000
residents per sq. km.
Note: Absolute change in residential density for continuing commuters (top figure), old telecommuters
(middle figure) and new telecommuters (bottom figure). Relative to benchmark economy in counterfactual
with = 0:33.
158
Figure B.6: Job changes for continuing commuters, old telecommuters, and new telecommuters
2000
1000
0
1000
2000
workers per sq. km.
2000
1000
0
1000
2000
workers per sq. km.
2000
1000
0
1000
2000
workers per sq. km.
Note: Absolute change in job density for continuing commuters (top figure), old telecommuters (middle
figure) and new telecommuters (bottom figure). Relative to benchmark economy in counterfactual with
= 0:33.
159
B.5 Elasticity of Speed to Traffic Volume
We set the elasticity of commuting speed with respect to traffic volume is "
V
= 0:2, following
SmallandVerhoef(2007). Inthecounterfactualeconomy, wecalculatechangesincommutingspeeds
as
speed
CF
ij
speed
BM
ij
speed
BM
ij
="
V
VMT
CF
VMT
BM
VMT
BM
;
assuming that the road capacity remains unchanged and only taking into account the change in
total vehicle miles traveled (VMT) in the metropolitan area.
6
Then we recover commuting times
ast
CF
ij
=distance
ij
= maxfspeed
CF
ij
; 65mphg. The maximum operator caps speeds at 65 mph which
is the speed limit on most highways in California. Since t
ij
and VMT endogenously depend on
each other, when solving for an equilibrium in a counterfactual economy, we iterate the model until
VMT converges.
Robustness. Since the results of the counterfactual experiments described in Section ??
crucially depend on changes in commuting speeds, we investigate whether our results are robust to
the value of "
V
. While 0.2 is a standard value in the traffic modeling literature, other studies used
higher values.
7
At the same time, a low value of"
V
ensures that many of our counterfactual results
are conservative.
To understand how sensitive our results are to the value of "
V
, we compute the counterfactual
economy with fraction = 0:33 telecommuters at different levels of "
V
ranging from 0 to 1. Our
three main sets of results remain robust to the value of "
V
. First, regardless of the value of "
V
, the
economy exhibits the decentralization of residents and centralization of jobs. Second, commuters’
trips are characterized by shorter times and longer distances (Figure B.7). Third, residential and
commercial floorspace prices fall for all values of "
V
(Figure B.8).
6
Note that our methodology does not allow for the differential impact of changes in traffic on individual
routes.
7
For example, Akbar et al. (2018) used values of 0.2 and 0.3. Bento et al. (2020) estimate a value of
about 0.9 for peak-hour commuting in Los Angeles.
160
Figure B.7: Commuting time and distance
BM 0 0.2 0.4 0.6 0.8 1
Elasticity of speed to volume
20
25
30
Commuting time, minutes
All workers
Commuters
BM 0 0.2 0.4 0.6 0.8 1
Elasticity of speed to volume
20
25
30
Commuting distance, km
All workers
Commuters
Note: Left panel displays the average commuting time for all workers and commuters in the benchmark and
the counterfactual economies at different levels of the elasticity of commuting speed with respect to traffic
volume. Right panel shows the average commuting distance.
Figure B.8: House prices
0 0.2 0.4 0.6 0.8 1
Elasticity of speed to volume
-6.5
-6
-5.5
-5
-4.5
-4
Change compared to BM, %
Residential floorspace price
Commercial floorspace price
Note: The figure displays the counterfactual change in average residential and commercial floorspace prices,
as a function of the elasticity of commuting speed with respect to traffic volume.
At the same time, quantitative implications of more telecommuting for wages and welfare are
sensitive to the value of "
V
. In our main counterfactual with "
V
= 0:2, the average commuter
market access (CMA) increases by about 17%. However, as "
V
approaches 1, commutes become
speedier and the average CMA increases by nearly 80% (left panel of Figure B.9). In addition, the
higher the elasticity of speed, the stronger will be spatial productivity spillovers. Hence, when "
V
161
goes to 1, wage gains for commuters are much larger and wage losses for telecommuters turn into
small gains, resulting in larger average wage increases (right panel of Figure B.9).
Figure B.9: Commuter market access, wages, and land prices
0 0.2 0.4 0.6 0.8 1
Elasticity of speed to volume
0
20
40
60
Change compared to BM, %
CMA
0.0374 0.1 0.15 0.2 0.25 0.3 0.33
Fraction of telecommuters ( )
-8
-6
-4
-2
0
Change, %
Wages
Land prices
Note: Left panel displays the average commuter market access for commuters, as a function of the elasticity
of commuting speed with respect to traffic volume. Right panel shows average wages and land prices.
As a result, with higher values of "
V
, welfare gains are larger (Figure B.10). In particular, as
"
V
goes to 1, commuters see their welfare increase by almost 10% (compared to 2.2% at "
V
= 0:2),
telecommuters experience a 2% increase (compared to a 2.5% loss), and overall welfare increases by
nearly 25% (compared to 18.9%).
162
Figure B.10: Welfare
0 0.2 0.4 0.6 0.8 1
-10
0
10
20
30
Change compared to BM, %
Commuters
0 0.2 0.4 0.6 0.8 1
Elasticity of speed to volume
-10
0
10
20
30
Telecommuters
Total welfare
Consumption and commuting welfare
Consumption welfare
0 0.2 0.4 0.6 0.8 1
-10
0
10
20
30
All workers
Note: Left panel shows the change in total expected welfare of commuters (“total welfare”), welfare net
of preference shocks and amenities (“consumption and commuting welfare”), and welfare net of shocks,
amenities, and commuting costs (“consumption welfare”). Central and right panels report changes in welfare
for telecommuters and all workers, respectively.
B.6 Accounting for Spatial Variation in Outcomes
Centrality. Distance from the center is a key driver of outcomes in most theoretical models
of the city. When dealing with data on real cities, it has been customary to measure this factor
simply as the straight-line distance from a “central business district” whose location is determined
by convention. Our alternative, which uses information on the city’s transportation network, is the
eigenvector centrality of each tract. We calculate it by finding the eigenvector associated with the
largest eigenvalue of theII matrix whoseij
th
element is given by expf
ij
g and whereI is the
number of model locations. This measure reflects the total strength of a given tract’s connections,
taking into account not only its direct connections, but also the connections of its connections
(second order), and their (third order) connections, and so on ad infinitum.
Interestingly, this measure picks out downtown LA as the most central location on the map. It
also turns out to be highly correlated with both straight line distance and travel time to downtown
163
Figure B.11: Quantiles of Centrality and Initial Allocations
1 0.8 0.6 0.4 0.2 0
1
4
16
80
400
2k
10k
N R, i
i
(residents per sq. km.)
local mean
global median
1 0.8 0.6 0.4 0.2 0
1
4
16
80
400
2k
10k
50k
N W, i
i
(jobs per sq. km.)
local mean
global median
1 0.8 0.6 0.4 0.2 0
$50
$100
$200
$400
$800
$1,600
$3,200
q R, i ($$ per sq. ft.)
local mean
global median
Note: The x-axis is scaled to quantiles of the centrality measure, weighted by land area.
LA (Pearson’s correlation coefficient 0.97 for each). Figure B.11 shows the evolution of some key
variables along the centrality gradient.
8
Real estate prices, the density of employment, and the
density of residence all increase on average the closer one gets to the center. The time required to
reach downtown LA is also, naturally, lower near the center.
In Figure 2.3, we plot the changes that take place in the counterfactual exercise in the same
manner as in Figure B.11. Here again we see that on average jobs move towards the center and
residents move away from it, and that there are big property price increases in the periphery. We
can also see that there is a great deal of variation that is unexplained.
Accounting for counterfactual changes. In order to have a more complete idea of
what is driving the variation in counterfactual outcomes, we expand our view to consider not only
a location’s initial centrality, but also the change in centrality between the baseline and counter-
factual due to changes in average speed, and the exogenous local characteristics a
i
, E
i
and x
i
. We
run a multivariate regression at the tract level, weighted by land area, of these five variables on
the log differences between counterfactual and baseline floorspace prices, employment density, and
residential density. From the estimated coefficients of these regressions we can infer the sign of each
relationship. We then use the Shapley method to decompose the coefficient of determination (R
2
)
for each regression.
9
The share assigned to each explanatory variable is a measure of its importance
8
The x-axis is scaled to quantiles of the centrality measure, weighted by land area. In other words, 0.5
on the x-axis represents the single square meter of land area such that 50% of the land area in the metro
area is less central (and 50% is more central.
9
See, e.g., Shorrocks (2013).
164
in accounting for the variation across space in each counterfactual outcome.
Table B.9: Accounting for counterfactual floorspace price changes
Coeff. Var. expl.
constant 0.274
(0.063)
centrality 0.022 32.0%
(0.015)
centrality 3.918 32.1%
(0.389)
a
i
-0.270 3.4%
(0.012)
E
i
-0.018 1.8%
(0.001)
x
i
0.024 15.0%
(0.001)
Total 84.33%
Table B.9 shows the results of this exercise for the change in floorspace prices. The negative
estimated coefficient on centrality confirms the core-periphery gradient of price changes, with prices
falling in the core and rising in the periphery. Once this is accounted for, locations whose centrality
increases due to change in speed in the counterfactual also see a more positive overall change in
prices. The negative coefficients on a
i
indicates that the relative value of real estate in locations
with high productivity falls, which is to be expected as workers on average need much less worksite
floorspace than before. The positive coefficient on x
i
indicates that the premium for locations with
good natural amenities has increased in the counterfactual, driven by telecommuters who can now
choose their residence location more freely. We see that position relative to the core drives the lion’s
share of the action here: centrality and centrality together account for 64.1% of the variation in
outcomes. Overall, the factors we consider here account for about 84% of the total variation.
For employment density and residential density, we further break the overall changes down into
changes in the average choices made by three groups of workers. These groups are: those that
commute both in the baseline and the counterfactual (67% of all workers), those that switch from
commuting to telecommuting (29.3%), and those that telecommute both in the baseline and the
counterfactual (3.7%). Table B.10 shows the results for changes in employment density and Table
B.11 shows the same for changes in resident density. Workers who continue commuting take jobs
closer to the urban core and also choose residences that are, on average, closer to the core. New
165
Table B.10: Accounting for counterfactual employment changes
Always commuter New telecommute Always telecommute All
Coeff. Var. expl. Coeff. Var. expl. Coeff. Var. expl. Coeff. Var. expl.
constant -1.292 -7.139 -0.924 -3.835
(0.199) (0.362) (0.119) (0.251)
centrality -0.138 34.5% -1.384 5.0% -0.149 32.9% -0.684 20.6%
(0.049) (0.089) (0.029) (0.062)
centrality -15.619 35.0% -40.542 5.9% -10.146 33.9% -25.506 22.2%
(1.237) (2.253) (0.737) (1.561)
a
i
1.012 3.1% 2.508 18.7% 0.534 3.1% 1.596 8.3%
(0.039) (0.070) (0.023) (0.049)
E
i
0.057 1.7% 0.189 4.7% 0.022 1.8% 0.107 2.2%
(0.004) (0.008) (0.003) (0.006)
x
i
-0.041 10.0% -0.015 4.8% -0.010 7.3% -0.030 3.5%
(0.003) (0.005) (0.002) (0.003)
Total 84.36% 39.04% 78.95% 56.78%
Table B.11: Accounting for counterfactual residence changes
Always commuter New telecommute Always telecommute All
Coeff. Var. expl. Coeff. Var. expl. Coeff. Var. expl. Coeff. Var. expl.
constant 0.840 -0.078 9.789 -0.148
(0.119) (0.225) (0.502) (0.141)
centrality 0.153 29.5% 0.026 35.6% 1.038 12.4% -0.045 24.9%
(0.029) (0.055) (0.123) (0.035)
centrality -1.282 29.5% 15.681 36.0% 44.651 13.1% 3.514 24.9%
(0.737) (1.400) (3.122) (0.879)
a
i
0.171 3.1% -0.240 2.8% -8.139 17.2% 0.009 1.9%
(0.023) (0.044) (0.097) (0.027)
E
i
0.050 2.4% 0.036 2.6% -1.051 27.8% 0.046 5.5%
(0.003) (0.005) (0.011) (0.003)
x
i
0.004 5.8% 0.010 10.2% 0.338 7.6% -0.001 7.8%
(0.002) (0.003) (0.007) (0.002)
Total 70.33% 87.19% 78.08% 65.01%
telecommuters, with the new-found freedom, do the opposite: they choose jobs and residence that
are, on average, farther from the core than before. Continuing telecommuters make smaller shifts
overall, taking jobs a bit closer to the core and moving their residences a bit farther from it. Across
all categories of workers there is a strong shift from commercial to residential use of land in locations
where there is a larger increase in centrality due to commuting speed changes.
There is also some heterogeneity in the way that location-specific characteristics correlate with
changes in choices for the three groups. For example, those who telecommute both in the baseline
and the counterfactual move their residences out of high-a
i
and high-E
i
tracts, presumably to make
room for the overall shift of employment into those tracts, while this pattern isn’t seen for the other
166
two groups.
As with changes in land prices, initial centrality and changes in centrality together account for
the lion’s share of the explained variation: 42.6% out of 56.78% total for employment changes, and
49.8% out of 65.01% total for residence changes. The positive coefficient on a
i
for employment
changes, and its 8.3% share in the variation in outcomes, is consistent with an improvement in the
allocation of workers to high-productivity locations in the counterfactual. Overall, the included
factors account for less of the variation than in the case of floorspace prices. This is partly due to
opposing tendencies in the three different types of workers canceling each other out.
167
Appendix C
Appendix to Chapter 3
C.1 Summary Statistics for KEEP
168
Table C.1: Summary Statistics: KEEP
Variable Mean Std. Dev. N
Study hours (per week)
Less than 2 hours 0.391 0.488 2562
More than 15 hours 0.138 0.345 2562
More than 20 hours 0.085 0.279 2562
Private Tutoring
Tutoring (=1) 0.678 0.467 2565
Expense on Private Tutoring(US$ per month) 201.487 279.87 2515
Teacher-Pupil Interaction
Students respect a teacher (=1) 0.705 0.456 2410
Students like a teacher (=1) 0.516 0.5 2410
Teachers are interested in a student (=1) 0.708 0.455 2410
Teachers understand a student 0.563 0.496 2407
Teacher Quality
Master degree or above (=1) 0.356 0.479 1934
Experience (years) 13.406 7.779 1931
Experience on a certain subject 12.642 7.723 1930
Subjective evaluation on teacher (Good=1) 0.762 0.426 2323
Study Atmosphere
Teacher-pupil closeness 0.746 0.435 2334
Class Atmosphere (Good=1) 0.573 0.495 2334
169
C.2 CSAT Score and the Hourly Wage 10 Years Later
(a) The CSAT Score And The Future Wage Are Positively Correlated
(b) The Convex Association
Data Source: KEEP
170

Abstract (if available)

Abstract This dissertation studies the impact of place-based policy and structural changes on the socio-demographic implications of the system of the cities. The first chapter investigates whether the expansion of high-speed rail (HSR) helps reduce the gender gap in labor market outcomes in South Korea. Using an instrumental variable strategy that leverages historical railroads constructed in Korea during the Japanese colonial era, I demonstrate empirically that the gender gap in the South Korean labor market decreased with the expansion of HSR. The second chapter studies how the shape of our cities would change if there were a permanent increase in working from home. We study this question using a quantitative model of the Los Angeles metropolitan area featuring local agglomeration externalities and endogenous traffic congestion. We find three important effects: (1) Jobs move to the core of the city, while residents move to the periphery. (2) Traffic congestion eases and travel times drop. (3) Average real estate prices fall, with declines in core locations and increases in the periphery. The third chapter studies the impact of ability tracking, which creates a homogeneous learning environment within the classroom, by segregating high-performing and low-performing students into different groups. This paper exploits a unique policy change in South Korea combined with comprehensive micro dataset to measure the causal consequences of ability tracking, compared to the allocating students to mixed ability groups. We find that tracking benefits high performing students who study more and gain from a superior learning environment, with the expense of the academic achievement of the low-performing students.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The impact of digital transformation on urban communities, welfare, and distributional outcomes

PDF

Three essays in education and urban economics

PDF

Essays on the economics of climate change adaptation in developing countries

PDF

Essays in climate change adaptation: role of market power in incentivizing adaptation behavior

PDF

Essays on environmental economics: education, employment, and experiments

PDF

Essays on the dual urban-rural system and economic development in China

PDF

Essays in environmental economics

PDF

Two essays in economic policy: The influence of perceived comparative need on financial subsidy requests; &, Unintended consequences of the FOSTA-SESTA legislation

PDF

Essays in the economics of education and conflict

PDF

Essays on family and labor economics

PDF

Essays on narrative economics, Climate macrofinance and migration

PDF

Essays in labor economics: demographic determinants of labor supply

PDF

The EITC, labor supply, and child development

PDF

Essays on macroeconomics and income distribution

PDF

Essays on economics of education

PDF

Essays on development economics

PDF

Essays on education: from Indonesia to Los Angeles, CA

PDF

Essays on development and health economics: social media and education policy

PDF

Essays in health economics and provider behavior

PDF

Essays on economics of education and private tutoring

Asset Metadata

Creator Kwon, Eunjee (author)

Core Title Essays on urban and real estate economics

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Publication Date 04/23/2021

Defense Date 04/21/2021

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,Real estate,urban economics

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kahn, Matthew (committee chair), Green, Richard (committee member), Oliva, Paulina (committee member)

Creator Email eon.kwon@gmail.com,eunjeekw@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-450977

Unique identifier UC11668230

Identifier etd-KwonEunjee-9515.pdf (filename),usctheses-c89-450977 (legacy record id)

Legacy Identifier etd-KwonEunjee-9515.pdf

Dmrecord 450977

Document Type Dissertation

Rights Kwon, Eunjee

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA