Habitat International 57 (2016) 110e120

Contents lists available at ScienceDirect

Habitat International
journal homepage: www.elsevier.com/locate/habitatint

Spatial access to pedestrians and retail sales in Seoul, Korea

Chang-Deok Kang
Dept. of Urban Planning and Real Estate, Chung-Ang University, 84 Heukseok-ro, Dongjak-gu, Seoul, 156-756, South Korea

a r t i c l e i n f o a b s t r a c t

Article history: The retail sector plays a significant role in urban economic development, smart growth policies with
Received 1 April 2016 mixed land use, and retail desert issues. In addition, urban policies and designs for pedestrian-friendly
Received in revised form urban settings have become a dominant paradigm to reshape urban environments. As retail sales
16 June 2016
depend considerably on the volume of neighborhood pedestrians, the link between pedestrian mobility
Accepted 19 July 2016
and retail sales has been a key issue in studies, designs, and economic development of urban areas.
Despite the importance of the topic, few empirical studies have examined how pedestrian volumes along
street networks affect retail prosperity in urban spaces. Hence, this study investigates the effect of
Keywords:
Pedestrian
accessibility and centrality to walking volume on retail sales in Seoul by considering pedestrian volumes
Accessibility and street network configuration concurrently. Multilevel regression models confirm that spatial access
Centrality to pedestrians has differing effects on retail sales according to the type of retail sector. Specifically, a
Retail sales higher accessibility to and visibility of retail stores for walkers tend to improve the sales performance of
Street all retail types. Further, the sales of the Medical Services and Education sectors are remarkably sensitive
Seoul to the combined effects of pedestrian and street configuration, unlike those of the three other sectors,
namely Food, Retail, and Services.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction Creating walkable cities has been an important urban issue

The retail sector plays a significant role in urban economics and
spatial structures, smart growth policies with mixed land use, and
retail desert issues. Several studies have confirmed that the
development and decline of retail activities change local economic
well-being, spatial land-use patterns, and urban mobility (Glaeser,
Kolko, & Saiz, 2001; Lang, 2003). On the other hand, as cities face
chronic and serious urban issues such as traffic congestion, energy
overconsumption, environmental pollution, and climate change,
mixed land use has become an effective tool to facilitate the access
to different land use and to promote walking as an alternative to
automobiles (Handy, 2005; Song & Knaap, 2004). Further, the ex-
istence of retail deserts implies less access to retail stores in low-
income and socially disadvantaged areas (Schuetz, Kolko, &
Meltzer, 2012). As public goods, retail stores constitute local ame-
nities for residents and workers. Less access to retail stores gener-
ates severe social equity issues for the elderly, low-income groups,
and disabled consumers (Guy, 2006). Accessibility to retail stores is
a key measure of the adequacy of retail services and must be
considered in urban policies and designs that address retail deserts.
E-mail address: cdkang@cau.ac.kr.
because such cities offer certain advantages such as lower auto-
mobile usage, more opportunities for local shopping and social
cohesion, and improved health for citizens (Kang, 2015). Many ur-
ban researchers, planners, and designers have focused on the de-
terminants of walkability and pedestrian-friendly cities (Talen &
Koschinsky, 2013). Recently implemented urban policies and de-
signs for such cities have provided a highly favorable urban setting
for retail sales. While developing policies to address retail devel-
opment and walkability, it is necessary to understand the spatial
relationship between retail economic performance and pedestrian
volume.
The endogeneity issues between pedestrian volumes and total
retail sales have been debated as an indicator of the urban eco-
nomic fabric. Although the relationship between the two is inter-
active, and not unidirectional, most studies focus on the effects of
pedestrian flows and walking environments on retail locations and
sales owing to the importance and power of walking behavior and
environments in shaping a city (New York Department of
Transportation, 2013; Sullivan & Adcock, 2002). Further, a few
policy studies have also confirmed that higher walking activity and
better walking conditions increase retail sales in the neighborhood
(Litman, 2014). Studies on the relationship between walking envi-
ronments and retail economic performance implicitly assume that

http://dx.doi.org/10.1016/j.habitatint.2016.07.006
0197-3975/© 2016 Elsevier Ltd. All rights reserved.
 C.-D. Kang / Habitat International 57 (2016) 110e120
more pedestrian-friendly environments attract more walkers,
generating higher retail sales. However, researchers are yet to
arrive at a consensus that better walking conditions always in-
crease pedestrian volumes. Two recent studies on walking in Seoul
found different effects of walking environments on walking from
those suggested by previous studies (Kang, 2015; Sung, Go, & Choi,
2013). Thus, empirical studies need to control for walking activities
and walking conditions concurrently.
A literature review reveals that few studies have investigated
the different effects of walking activities and walking conditions,
controlling for other relevant variables. This study is unique in that
it attempts to analyze the effects of accessibility and centrality to
pedestrians on retail sales, controlling for walking conditions and
urban environments.
As emphasized previously, walking, as an activity, and walk-
ability, as a condition for walking, are different issues (Litman,
2003). Higher walkability does not always generate more
walking. Thus, we review the relevant literature on both walkability
and walking. Many studies estimate the effects of walkability on
economic value in areas such as property value, personal health
and income, mixed urban development, and retail sales. Recent
studies have shown that a walkable urban setting is associated with
higher property value (Sohn, Moudon, & Lee, 2012). Further, a
higher Walk Score, which is a measure of walkability, increases the
value of residential and commercial property (Cortright, 2009; Pivo
& Fisher, 2011). Other effects of walkable neighborhoods on eco-
nomic value involve the foreclosure risk of households with more
cars and a lower access to public transport (Moudon et al., 2006;
Rauterkus, Thrall, & Hangen, 2010). Furthermore, a walkable ur-
ban setting leads to higher mixed land use and housing develop-
ment (Ellen & Voicu, 2005; Tatian, Kingsley, Parilla, & Pendall,
2012). The section titled “Description of Variables” introduces the
specific conditions for walking according to prior studies.
While a few studies investigate the relationship between
walkable streets and the spatial variation of retail sales, most focus
on the association between street configuration and retail location
for profit maximization. A walker-oriented street network config-
uration with lesser automobile usage generates spatial variation in
retail turnover (Hass-Klau, 1993). As the location and economic
performance of retail stores depend on local pedestrian volume
along street networks, the effects are different between large and
small stores. On the other hand, a higher proximity to retail stores
and owning bikes promote walking and cycling behavior (Krizek &
Johnson, 2006). As pedestrian access to retail stores is important for
the economic performance of local businesses, downtown devel-
opment plans utilize pedestrian-friendly urban structures
(Robertson, 1993). Such urban structures tend to increase people's
willingness to walk to retail stores (Schneider, 2015). A few studies
have analyzed how pedestrian-friendly urban structures in the
United States alter retail sales. New York City's Department of New
York Department of Transportation (2013) confirmed that a project
for pedestrian-friendly streets increased retail sales in the relevant
area according to a comparison of retail performance between
impacted and non-impacted zones. Further, Leinberger and Alfonzo
(2012) show that walkable neighborhoods generate higher retail
sales and rent by attracting more pedestrians.
Other studies relevant to retail location have confirmed that
higher centrality and connectivity along street networks substan-
tially explain the spatial distribution of retail activities (Scoppa &
Peponis, 2015). Specifically, higher betweenness centrality is
associated with more clustering of retail and service activities
(Porta et al., 2009). The response of retail location to street
configuration varies by retail type. Among retail stores, specialty
stores and department stores considerably preferred to have a
more accessible location (Wang, Chen, Xiu, & Zhang, 2014). Another
111
study verified that locations with busier streets and higher building
density were favorable for retail and food stores (Sevtsuk, 2014).
The local contexts generate different connections between spatial
patterns of retail and street network structure. While retail stores in
older cities tend to cluster in main roads and streets, the stores in
newly planned towns are more likely to spread across planned
areas (Omer & Goldblatt, 2015).
We need to examine the relationship between walking activities
and retail sales, controlling for walking conditions and other fac-
tors, for several reasons. First, city leaders and citizens increasingly
recognize the influence of pedestrian behavior and environments
on direct and indirect benefits of higher urban mobility, trans-
portation cost savings, efficient land use, neighborhood livability,
better public health, local economic development, and social eq-
uity. Second, creating pedestrian-friendly environments for urban
revitalization and sequentially increasing walking activities have
local impacts on not only residents and local businesses but on a
city as a whole. Thus, it is critical to measure the positive and
negative effects of creating effective urban designs for pedestrians
and to predict the outcomes of relevant urban policies and designs.
Finally, as an indicator of local economic vitality, retail sales clearly
respond to changes in pedestrian volume and a walkable street
configuration.
To summarize, while numerous studies show that walkability
has a positive effect on neighborhood property value, personal
health, individual income, urban development, and retail sales, few
studies investigate how walking activities combined with the street
configuration determines the health of local businesses. This un-
explored hypothesis is empirically testable because walkable cities
and local economic vitality have been central issues to urban policy
and design. Further, few studies have tested the combined effects of
pedestrian volume and street network configuration on retail rev-
enue by types.
This study aims to fill this gap in the literature by analyzing the
combined effects of pedestrian volume and street configuration.
Furthermore, we compare the various effects of spatial accessibility
and centrality to walking volume on retail sales in Seoul and test
five unique spatial accessibility and centrality effects. To study the
combined effects of walking volume and street network layout, we
apply the spatial accessibility and centrality indices to identify the
differing effects on retail sales under a single research framework.
Moreover, we analyze the different effects of the five indices on
retail sales according to the type of retail sector. Few studies have
attempted to separately analyze the effects of walking activities and
walking conditions on local business performance. This study
controls for walking volume and walking conditions in one model
to isolate the effects of pedestrian volume on retail sales.
The rest of this paper has four sections. The first section ad-
dresses the study area and data sources used for the econometric
tests. The second section provides the motivation behind
measuring spatial accessibility and centrality and details about
variables in the multilevel regression models. The third section
interprets the model results, focusing on how spatial accessibility
and centrality to pedestrians affect retail sales. The final section
summarizes the results and suggests policy implications.
2. Study context and data sources
Seoul's main economic sector evolved from secondary to ter-
tiary since the 1970s. The 2011 industry data show that the tertiary
sector accounted for 88.9% of Seoul's industry. In the tertiary in-
dustry, wholesale and retail, accommodation, and food and bev-
erages have become dominant industries. In 2011, wholesale and
retail accounted for 32.0%; accommodations and food and
beverage, 17.7%; transportation, 13.6%; other services, 10.3%; and
112 C.-D. Kang / Habitat International 57 (2016) 110e120
business services, 6.0% of the tertiary sector (Seoul Institute, 2015).
Thus, retail activities are important to Seoul's industrial structure
and local economy.
On the other hand, the Seoul Metropolitan Government (SMG) is
focused on pedestrian-friendly and sustainable cities. The SMG has
implemented multiple policies to expand pedestrian and public
transit-oriented urban settings, such as the replacement of free-
ways with urban greenways, the development of car-free streets,
and the creation of transit malls (Seoul Metropolitan Government,
2014). As Seoul city has enhanced its local economy and quality of
life through retail development and walkable urban structures, it is
a suitable case to empirically test a strong relationship between
pedestrian volume and retail sales.
This study analyzes unique data on retail sales and pedestrian
volume to test the effects of pedestrian access to retail stores on the
sales. SKT Geovision, a division of SK Telecom that is a mobile
phone service company, provided the data for three regions in
Seoul (see Fig. 1). The amount for sales per Small Area was deter-
mined by credit card payments, Value Added Network (VAN), and
Points of Sales (POS) in April, June, September, and December 2013.
The data, however, does not offer information on the number of
retail stores. The retail sales data were categorized according to the
type of retail sectors, such as Education, Food, Medical Services,
Retail, and Services. SKT Geovision also provided the pedestrian
volume data per pCell unit developed by the company. A pCell is a
50-by-50 m spatial unit for aggregating pedestrian volume data
using mobile signals. We assume that accessibility and centrality to
walkers measured at the pCell points affect the retail sales at the
nearest centroids of the Small Area because these two factors
indicate neighborhood conditions in the pCell units.
As Fig. 1 illustrates, the three regions are the Traditional Urban
Fig. 1. Location of the study areas.
Core (TUC) area with the Central Business District (CBD), the New
Urban Core (NUC) area with a sub-CBD (Kangnam Business District,
KBD), and the Mixed Core (MC) area with a sub-CBD (Youido
Business District, YBD) (Jang & Kang, 2015). The section on abbre-
viations details the information on regional classification. The 2010
retail and service statistics show that the share of employment and
firms respectively was 26% and 18% in the NUC, 16% and 15% in the
TUC, and 8% and 5% in the MC (Statistics Korea, 2012).
As mentioned previously, we used the retail sales data for five
types of retail sectors (see Table 1). We have matched the Korean
retail sectors with the North American Industrial Classification
System for a better understanding.
Fig. 2 compares the concept of the Korean census tract and Small
Areas. The Korean census tract represents the smallest statistical
and spatial units for aggregating socioeconomic data. Statistics
Korea identifies these units using homogeneous features such as
demographic attributes, economic status, and living conditions.
Seoul had 16,671 units with about 500 inhabitants, on average, in
2010 (Kang, 2015; Statistics Korea, 2012). To aggregate retail sales
data, SKT Geovision subdivides the Korean census tracts into Small
Areas, each with a daytime population of at least 1,000 individuals.
To control for other factors that change retail sales, we used
geographical information such as location and transportation, land
usage attributes of census tracts, socioeconomic attributes, retail
type, and time of sales. The SMG provides geographical information
on road and street networks, CBD and sub-CBD locations, subway
stations, bus stops, major retail centers, parks, schools, and the total
building area of land used per Korean census tract. Statistics Korea,
as the Census Bureau in the United States, offers population and
employment information for each census tract.
 C.-D. Kang / Habitat International 57 (2016) 110e120 113

Table 1
Retail sector types and descriptions.

Type Description

Education Child day care service; Exam preparation and tutoring services
Food Full service restaurants; limited service restaurants; cafeterias, grill buffets, and buffets; snack and nonalcoholic
beverage bars; beer, wine, and liquor stores
Medical Services General medical and surgical hospitals; pharmacies and drug stores
Retail Furniture stores; electronic stores; food supplement stores; transportation; book stores; hobby, toy, and game
stores; office supplies and stationery stores; other miscellaneous stores; clothing accessories stores; clothing stores
Services Beauty salons; accommodation; recreation services; all other personal services

Note: Obtained from SKT Geovision (2013).

Fig. 2. Comparison of the Korean census block and small area unit concepts.
Sources: Statistics Korea (2012), SKT Geovision (2013)

3. Methodology 3.2. Description of variables

3.1. Motivation
Distinctly from previous studies, we note that pedestrian vol-
ume and street networks matter in spatial retail sale variation.
Intuitively, pedestrian retail access and relative spatial concentra-
tion generate local variation in retail sales. Thus, the concurrent
measure of street configuration and walking volume in a single
index provides more reliable and accurate information on the
relationship between spatial pedestrian patterns and retail eco-
nomic performance. A few related studies show that a comfortable
street network configuration with retail activities is a favorable
urban setting for walking (Ozbil, Peponis, & Stone, 2011; Peponis,
1989). However, our knowledge of the integrated pedestrian vol-
ume and street network configuration effects on retail sales is
limited. This study notes that the Urban Network Analysis (UNA)
measure of spatial accessibility and network centrality is an effec-
tive tool in considering street network structure and walking vol-
ume concurrently. Related studies have used the concept of
accessibility and centrality as the key determinants of street net-
works to explain population density, residential property price and
rents, and commuting behavior (Chiaradia, Hillier, Schwander, &
Barnes, 2013).
To identify the different effects of spatial accessibility and cen-
trality to walking on retail sales, we measure Reach, Gravity Index,
Betweenness, Closeness, and Straightness (see Table 2). While
Reach indicates the total pedestrian volume within a given network
distance from retail spots, Gravity Index represents Reach com-
bined with distance friction based on Hansen's accessibility
concept. With respect to centrality indices, Betweenness is defined
as the passing of a specific location in total passing pedestrian
volume over the neighborhoods, Closeness indicates the total
pedestrian volume multiplied by the network distance from retail
spots, representing the centrality of the retail spots, and Straight-
ness measures the extent to which paths from retail spots to other
pedestrian points are straight lines. Previous studies have generally
used these indices, which justifies our use of the measures. Gravity
Index is widely used to measure spatial variation of accessibility
(Hansen, 1959; Sevtsuk, 2014). Betweenness, Straightness, and
Closeness are popular indices in studies assessing the relative
importance of urban networks and flows (Kuby, Tierney, Roberts, &
Upchurch, 2005; Xiao, Webster, & Orford, 2014). We expand these
concepts by combining pedestrian volumes and street networks,
and measure the sophisticated connection between them.
We mobilize three components for the empirical study to
114 C.-D. Kang / Habitat International 57 (2016) 110e120
Table 2
Accessibility and centrality to pedestrians and calculation methods.
Indices
Reach
Gravity Index
Betweenness
Straightness
Closeness
Note:
i: The pCell points nearest to the centroids of Small Areas (retail sites).
j: The pCell points (pedestrian spots).
G: Network.
r: Network radius (500 m).
d[i, j]: Shortest network distance between origin i and destination j (meter).
d[i, j]: Euclidian distance between origin i and destination j (meter).
njk[i]:Number of paths passing through node i with j and k in the network radius r from i.
njk: Number of paths between nodes j and k.
Beta (b): 0.00217.
W[j]: Number of walkers in j (persons).
Source: Modified from Sevtsuk and Mekonnen (2012).
examine how spatial accessibility to walking affects retail sales:
spatial data, the Urban Network Analysis (UNA) toolbox installable
for the ArcGIS program, and a multilevel regression model in the
Stata program. The dependent variable is the 2013 data for total
retail sales per Small Area. The independent variables that we
adopted with a reference to previous studies are (1) five accessi-
bility and centrality indices related to pedestrians, (2) location and
transportation, (3) land-use attributes in Seoul's census tract units,
(4) socioeconomic features in the census tract units, (5) retail types
(dummy variables with reference to Education), and (6) time of
sales (dummy variables with reference to April 2013).
3.2.1. Pedestrian accessibility and centrality
This study applied UNA to measure Reach, Gravity Index,
Betweenness, Closeness, and Straightness from the nearest pCell to
each centroid of the Small Areas (retail sites) to each pedestrian
spot within a given radius (500 m). Table 2 shows the calculations
of accessibility and centrality. The UNA was originally developed by
the MIT City Form Lab (Sevtsuk & Mekonnen, 2012). In our analysis,
the origin is a nearest pCell to the centroids of the Small Areas, and
the destinations are the pedestrian-counted spots. We note that
this approach incorporates a third element, walking volume, with
the nodes and edges of typical street networks. While most pre-
vious analyses of street networks ignore spatial variation of walking
activities along the network, the UNA in this study measures the
spatial accessibility and centrality incorporating the two elements,
namely of street network configuration from origins and destina-
tions, and pedestrian volumes at the destinations (Sevtsuk &
Mekonnen, 2012). Thus, this tool effectively generates more accu-
rate information and realistic results for urban studies, planning
and policy, and design.
Spatial accessibility is measured by the Reach and Gravity Index.
While Reach indicates the sum of pedestrians within a 500-m
radius along the shortest distance between origin i and destina-
tion j, the Gravity Index, first used by Hansen (1959), represents the
Reach concept combined with distance decay (beta value). This
study sets the beta value to 0.00217 for meter distance, as sug-
gested by Handy and Niemeier (1997). Two reasons permit the use
of this value. First, there is no empirically measured beta value that
fits the local context. Second, the value was calculated in the
context of convenience shopping travel, which involves short
Calculation methods
P
Reachr ½i ¼ W½j
j2Gfig;d½i;jr
r P W½j
Gravity ½i ¼ εb$d½i;j
j2Gfig;d½i;jr
P njk ½i
Betweennessr ½i ¼ njk $W½j
j2Gfig;d½i;jr
P d½i;j
Straightnessr ½i ¼ d½i;j
$W½j
j2Gfig;d½i;jr
Closenessr ½i ¼ P 1
ðd½i;j$W½jÞ
j2Gfig;d½i;jr
commutes, similar to the ones considered in this study. A few
studies confirmed that the Gravity Index substantially explains the
variations of land-usage patterns, retail store locations, and local
employment patterns (Hansen, 1959; Huff, 1963; Waddell &
Ulfarsson, 2003).
Network centrality incorporates Betweenness, Closeness, and
Straightness. Betweenness represents the number of passings at a
specific node i relative to total available passings between two pairs
of other origins k and destinations j, weighted by the walking
volume of the destinations. Empirical studies have confirmed that
Betweenness significantly alters the spatial patterns of retail and
service stores (Porta, Crucitti, & Latora, 2005; Sevtsuk, 2014).
Closeness, a popular centrality index, measures how close each
origin is to the adjacent destination within a given threshold path
distance. This measure can identify the pedestrian volumes
weighted by the shortest path distances between origins and des-
tinations close to the origin location within a given radius. In the
calculation, a lower Closeness value means more total pedestrian
volume weighted by the shortest network distance within a given
threshold. Thus, lower Closeness coefficients indicate highly posi-
tive effects on retail sales. The final network centrality index,
Straightness, measures the similarity between the shortest
network distance from the origin i to the destination j within a
given radius and the straight-line distance (Porta et al., 2005). In
reality, higher Straightness indicates the existence of a direct route
between the surrounding pedestrian spots and a higher visibility of
the origin, which influence retail sales (Lynch, 1960).
3.2.2. Location and transportation
Locational advantage and nearness to transportation change the
spatial patterns of retail sales. Many empirical models verify that
access to urban centers and sub-centers, transportation networks,
public transit, retail centers, education facilities, and parks as
common walking environments influence walkability (Cerin,
Macfarlane, Ko, & Chan, 2007; Forsyth, Hearst, Oakes, & Schmitz,
2008; Tiwari, Cervero, & Schipper, 2011; Vojnovic, Jackson-
Elmoore, Holtrop, & Bruch, 2006; Wey & Chiu, 2013). Several
studies adopt a street-focused approach, concentrating on topo-
logical and physical attributes of street configurations, street den-
sity, connectivity, and spatial patterns of reachable destinations
(Agrawal & Schimek, 2007; Baran, Rodriguez, & Khattak, 2008;
 C.-D. Kang / Habitat International 57 (2016) 110e120
Boarnet & Crane, 2001; Crane & Crepeau, 1998; Ewing & Cervero,
2001; Lee & Moudon, 2006; Song & Knaap, 2004). To control for
location and transportation attributes, we measure the distance to
the CBD and five sub-CBDs, subway stations and bus stops, streets
and roads, main retail areas, schools, and parks. Further, we control
for the density of subway stations, bus stops, street and road con-
nectivity, and park areas per Korean Census Block unit.
3.2.3. Census tract land usage attributes
Neighborhood land use is also an important part of the urban
setting that affects the relationship between the spatial variation of
walking and local retail sales. A few studies confirmed that higher
access to land use and higher land use density tend to be positively
associated with a higher pedestrian volume (Frank & Pivo, 1994;
Zhao, 2014). Thus, this study focuses on mixed land use, balanced
between residential and nonresidential use, overall land develop-
ment density, and four types of land use density: residential,
commercial, office, and industrial space. The Entropy index mea-
sures mixed land use in Korean census tract units. It is applied to
measure the mix of five types of land-use: residential, commercial,
office, industrial, and others, where a value close to 1 indicates
mixed land use based on formula (1):
Pn
i¼1 Pi lnðPi Þ
Mixi ¼ (1)
lnðnÞ
where n is the number of land-use types (here, n ¼ 5), and Pi is the
proportion of each land use i.
As another land use index, we calculate the balance indexes of
residential and nonresidential (commercial, office, industrial, and
other) land use from the 2009 Seoul Building Register data. The
balance index based on formula (2) ranges from 0 to 1, where a
higher value indicates a higher balance between pairs of land use
types (Cervero & Duncan, 2003; Kang, 2015).
Res  Nonresi
Balancei ¼ 1  j j (2)
Res þ Nonresi
To measure land use development density, we measure the net
area with the Korean census tracts. The net area represents the total
biotope for residential, commercial, office, industrial, and infra-
structural space in each census tract from the 2010 Seoul Biotope
data (Kang, 2015). While the overall development density is the
total building area divided by the net area, the land use density is
measured as the area of each building in the total residential,
commercial, office, and industrial space divided by the respective
net area.
3.2.4. Socioeconomic attributes
Dense population and employment environments are favorable
urban settings for the provision of various services and land use
within walking distance (Chen, Jia, & Lau, 2008; Forsyth et al., 2008;
Loo & Chow, 2006). Thus, we control for the two variables
measured by the density of residents and workers, respectively.
3.2.5. Retail types
Retail sales vary by sector type. To control for the effects of
sector types, we construct a dummy variable for each retail type
with reference to Education. Thus, we include the dummy variables
for the Food, Medical Services, Retail, and Services sectors in the full
model.
3.2.6. Time of sales
Because of seasonal consumption, the timing of sales affects the
volume of retail sales. As the sales data were recorded in April, June,
115
September, and December 2013, we contain the dummy variables
for June, September, and December 2013 in the full model with
reference to April 2013.
3.3. Multilevel regression models
Since we analyze different units of data, namely, Small Area
(Level 1) and Korean census tract (Level 2), we apply multilevel
regression models. In the model framework, while the Small Area
data include retail sales, spatial accessibility and centrality to pe-
destrians, distance to main transport systems and urban facilities,
retail types, and time of sales, the geographical data in the Korean
census tracts contain net density of main transportation facilities
and connectivity, park density, land use attributes, and socioeco-
nomic attributes. Because the units of the data are different, we use
multilevel regression models to help correct the under or over-
estimation of coefficients by applying ordinary least squares
regression (Rabe-Hesketh & Skrondal, 2008). Our multilevel
regression uses the following equation:
Pij ¼ g00 þ b1 Aijk þ b2 Sijk þ b3 Cijk þ b4 Lijk þ b5 Eijk þ b6 Tijk þ m0j
þ εij
(3)
where Pij is the retail sales in Small Area i (Level 1) in census tract j
(Level 2), bk refers to the model coefficients of variables (k ¼ 1, 2, 3,
…, m), r00 denotes the model constants, Aijk is a vector of pedestrian
accessibility and centrality in Small Area i and census tract j, Sijk is a
vector of location and transportation in Small Area i and census
tract j, Cijk is a vector of attributes of land use and transportation in
Small Area i and census tract j, Lijk is a vector of socioeconomic
attributes in Small Area i and census tract j, Eijk is a vector of retail
types in Small Area i, Tijk is a vector of sales time in Small Area i,
and u0j and eij are the residual error terms of census tract units and
Small Areas, respectively.
We take a log-log functional form and convert the dependent
and independent variables into natural logarithms, except for the
dummy variables for retail types and sales time. We choose the log-
log form because, first, there is no consensus on the selection of
specific functional forms (Duncan, 2011). Second, we apply the
untransformed, semi-log, or log-log models, and choose the log-log
form because the models generate better results from theoretical
and empirical perspectives. Finally, the log-log form helps interpret
the coefficients as mean attribute elasticity regardless of variable
attributes (Shyr, Andersson, Wang, Huang, & Liu, 2013).
Before our empirical analysis, we use an intra-class correlation
(ICC) test to check whether a multilevel regression is appropriate.
An ICC value exceeding 0.05 justifies the use of multilevel regres-
sion. Otherwise, we should use typical regression models.
Furthermore, we use the variance inflation factor (VIF) test and
correlation matrix to check for correlation between the explanatory
variables.
Table 3 presents the variables and descriptive statistics. The
sales range from USD 0.05 to USD 204 million (USD 1 ¼ 1,161 KRW)
per Small Area. The table reports the values of five accessibility and
centrality measures using the UNA tool. Location and trans-
portation attributes contain the density of each transport facilities,
connectivity of roads and streets, and park within the Korean
census tracts and the distance from urban centers, transportation
facilities and neighborhood attributes to the Small Areas. The
census tract attributes of land use include land use mix, the balance
between residential and nonresidential use, and the net density of
total development and individual land use. As reported under so-
cioeconomic attributes, the mean of net employment density is
116 C.-D. Kang / Habitat International 57 (2016) 110e120

Table 3
Descriptive statistics.

Variable description Variable code Mean Min. Max.

Retail sales
Retail sales (Korean Won) logsales 117,000,000 61 237,000,000,000
Accessibility and centrality to pedestrians
Reach logreach 7,167.54 0.00 125,699.50
Gravity Index loggra 4,665.85 0.00 96,309.28
Betweenness logbet 52,032.98 0.00 3,679,777.00
Closeness logclo 0.0001 0.00 1.55
Straightness logstra 7,014.08 0.00 125,583.50
Location and transportation attributes
Distance to CBD logcbd_dis 8,485.04 115.70 17,961.61
Distance to sub-CBD logscbd_dis 3,663.82 96.32 11,505.59
Distance to subway stations logsub_dis 479.57 30.06 5,701.90
Distance to bus stops logbus_dis 113.53 1.70 1,016.90
Distance to roads logro_dis 85.85 0.03 1,488.36
Distance to streets logst_dis 25.84 0.00 552.68
Distance to main retail centers logret_dis 629.21 1.26 4,625.01
Distance to schools logsch_dis 342.79 2.08 3,073.55
Distance to parks logpark_dis 239.06 0.44 1,056.51
Net density of subway stations, number of stations per net area logsub_den 0.02 0.00 1.64
Net density of bus stops, number of bus stops per net area logbus_den 0.45 0.00 34.14
Net density of street connectivity, number of connectivity per net area logst_con_~n 13.90 0.00 161.19
Net density of road connectivity, number of connectivity per net area logro_con_~n 1.03 0.00 30.72
Total park area per net area logpark_ar~n 86,456.22 0.00 4,090,000,000.00
Census tract attributes of land use
Entropy index of residential, commercial, office, industrial, and other land use newmix5 0.51 0.00 0.96
Balance index for residential and non-residential use resi_nonresi 0.43 0.00 1.00
Net density of building development, total building area per net area logdev_den 25,100.09 65.81 26,300,000.00
Net density of residential development, total residential area per net area logresi_den 9,017.74 0.02 392,129.90
Net density of commercial development, total commercial area per net area logcom_den 7,794.76 0.03 26,300,000.00
Net density of office development, total office area per net area logoff_den 7,220.57 0.02 113,544.80
Net density of industrial development, total industrial area per net area logind_den 171.67 0.02 25,921.95
Socioeconomic attributes
Net population density logpop_den 242.21 0.00 3,134.90
Net employment density logemp_den 447.74 0.00 165,985.80
Retail types
Education (Reference group, yes ¼ 1, no ¼ 0) type_dum1 0.04 0.00 1.00
Food (yes ¼ 1, no ¼ 0) type_dum2 0.45 0.00 1.00
Medical Services (yes ¼ 1, no ¼ 0) type_dum3 0.09 0.00 1.00
Retail (yes ¼ 1, no ¼ 0) type_dum4 0.30 0.00 1.00
Service (yes ¼ 1, no ¼ 0) type_dum5 0.13 0.00 1.00
Time of sales
April 2013 (Reference group, yes ¼ 1, no ¼ 0) month_dum1 0.25 0.00 1.00
June 2013 (yes ¼ 1, no ¼ 0) month_dum2 0.25 0.00 1.00
September 2013 (yes ¼ 1, no ¼ 0) month_dum3 0.25 0.00 1.00
December 2013 (yes ¼ 1, no ¼ 0) month_dum4 0.25 0.00 1.00

higher than the net population density, indicating a higher con-
centration of workers. The retail type and time of sales serve as
dummy variables with the reference groups.
4. Results and discussion
This study validated and calibrated all models from theoretical
and empirical perspectives as well as through methodological tests.
We checked the sign and statistical significance of the coefficients
by referring to existing studies. Further, the test results of the ICC
for the multilevel regression models, VIFs, and a simple correlation
matrix justified the selection of the variables. In Table 4, the full
model analysis includes 119,702 retail sales per Small Area (Level 1)
and 2,619 Korean census tracts (Level 2). The R-square values are
0.16 (overall), 0.07 (within), and 0.36 (between). The R-square
values are relatively low due to the following reasons. First, the
excluded nonspatial factors such as the service level, price, and
other components affect the retail sales performance. Second, we
use the retail sales aggregated by Small Area units. Thus, the models
exclude the factors concerning the sales of individual retail stores.
Each model's ICC (Rho) is 0.2, which is more than 0.05, the refer-
ence value, which justifies the use of multilevel regression models.
This study limits the variable coefficients within the 5% statistical
significance level. The VIF ranges from 1.29 to 6.37, indicating no
multicollinearity. We additionally analyze and compare the effects
by retail type. Owing to the higher correlation among the five
accessibility and centrality indices in the correlation matrix of all
variables used, we identify the effects of the indices in the indi-
vidual models as shown in Table 4.
4.1. Effects of spatial accessibility and centrality to pedestrians
Fig. 3 compares the marginal effects of pedestrian accessibility
and centrality on retail sales by type. The marginal effects are
derived from the coefficients of the full models and each individual
model by retail type. The figure compares the five indices in all
models and shows that Reach and Gravity Index have almost
similar effects whereas Straightness has remarkably greater effects
than Closeness and Betweenness.
Thus, the higher accessibility of pedestrians, indicated by the
Reach and Gravity Index, and the higher centrality, indicating vis-
ibility to walkers, as measured by Straightness, considerably tend to
increase the sales volume of all retail types. The effects of Closeness
and Betweenness on retail economic performance are relatively
weaker, contrary to propositions of previous studies. Since we
consider both street configuration and pedestrian volume for each
 C.-D. Kang / Habitat International 57 (2016) 110e120 117

Table 4
Multilevel regression models for all retail types.

Variables Model 1 Model 2 Model 3 Model 4 Model 5 VIF

Coefficients Coefficients Coefficients Coefficients Coefficients

Accessibility and centrality to pedestrians

logreach 0.0202*** 1.33
loggra 0.0210*** 1.34
logbet 0.00595*** 1.36
logclo 0.0126*** 1.27
logstra 0.0201*** 1.34
Location and transportation
logcbd_dis 0.0193 0.0192 0.0187 0.0191 0.0189 1.91
logscbd_dis 0.0273 0.0271 0.0273 0.029 0.0272 1.21
logsub_dis 0.0837*** 0.0831*** 0.0870*** 0.0861*** 0.0835*** 1.36
logbus_dis 0.0479*** 0.0480*** 0.0506*** 0.0483*** 0.0478*** 1.36
logro_dis 0.0625*** 0.0627*** 0.0645*** 0.0628*** 0.0624*** 1.46
logst_dis 0.00111 0.000881 0.00813 0.0000241 0.000997 1.41
logret_dis 0.198*** 0.198*** 0.201*** 0.198*** 0.198*** 1.63
logsch_dis 0.0976*** 0.0980*** 0.0963*** 0.0968*** 0.0975*** 1.19
logpark_dis 0.0368** 0.0369** 0.0351* 0.0366** 0.0368** 1.26
logsub_den 0.0732 0.0733 0.0747 0.0728 0.0732 1.53
logbus_den 0.129*** 0.129*** 0.126*** 0.129*** 0.128*** 1.45
logst_con_~n 0.214*** 0.214*** 0.223*** 0.215*** 0.214*** 2.36
logro_con_~n 0.0281 0.0283 0.0346 0.0281 0.0283 1.43
logpark_ar~n 0.00437 0.00439 0.00417 0.00422 0.00442 1.39
Census tract attributes of land use
newmix5 0.651*** 0.650*** 0.630** 0.655*** 0.650*** 5.02
resi_noresi 0.188 0.189 0.198 0.186 0.188 2.82
logdev_den 0.247*** 0.247*** 0.255*** 0.249*** 0.247*** 2.48
logresi_den 0.0332* 0.0332* 0.0330* 0.0328* 0.0332* 3.13
logcom_den 0.0198* 0.0198* 0.0210* 0.0197* 0.0198* 2.89
logoff_den 0.00285 0.00283 0.00254 0.00266 0.00285 3.78
logind_den 0.00323 0.00333 0.00447 0.0028 0.00324 1.54
Socioeconomic attributes
logpop_den 0.195*** 0.195*** 0.194*** 0.195*** 0.195*** 3.97
logemp_den 0.0383* 0.0383* 0.0404** 0.0384* 0.0382* 3.62
Retail types
type_dum2 0.0540* 0.0540* 0.0552* 0.0541* 0.0540* 7.3
type_dum3 0.505*** 0.505*** 0.505*** 0.505*** 0.505*** 3.03
type_dum4 0.533*** 0.533*** 0.534*** 0.533*** 0.533*** 6.36
type_dum5 1.067*** 1.067*** 1.069*** 1.067*** 1.067*** 3.91
Time of sales
month_dum2 0.00587 0.00585 0.00661 0.00619 0.00588 1.51
month_dum3 0.0729*** 0.0730*** 0.0706*** 0.0719*** 0.0729*** 1.5
month_dum4 0.0326* 0.0325* 0.0350** 0.0339** 0.0326* 1.51
Constant 16.21*** 16.20*** 16.30*** 16.24*** 16.21***
Rho 0.2 0.2 0.2 0.2 0.2
Number of cases 119,702 119,702 119,702 119,702 119,702
Number of groups 2,619 2,619 2,619 2,619 2,619
R-square
Within 0.07 0.07 0.07 0.07 0.07
Between 0.36 0.36 0.35 0.36 0.36
Overall 0.16 0.16 0.16 0.16 0.16

Note: *p < 0.05, **p < 0.01, ***p < 0.001.

Fig. 3. Marginal effects of accessibility and centrality to pedestrians on retail sales by retail type.
118 C.-D. Kang / Habitat International 57 (2016) 110e120
index within the relatively short radius of 500 m, our results differ
from those of related studies.
A comparison between the five indices measuring pedestrian
effects implies multiple and actual connections between walking
volume and retail sales. As the Reach and Gravity Index effects
indicate, a larger volume of pedestrian traffic within a 500-m radius
tends to increase retail sales in the neighborhoods. Because of weak
spatial impedance in the Gravity Index, we did not observe any
remarkable difference between the effects of Reach and Gravity
Index effects on retail sales. These results imply higher retail sales
in the neighborhood of dense networks and higher walking vol-
umes. Furthermore, as the Straightness effects suggest, retail stores
with higher visibility and direct routes that facilitate larger walking
volumes enjoy a superior economic performance. Unlike previous
studies that focused on Betweenness and Closeness without
incorporating pedestrian volumes, this study finds that the effects
of Betweenness and Closeness on retail sales are weaker than those
of Reach, Gravity Index, and Straightness. Thus, dense streets,
higher pedestrian volumes, and direct routes to pedestrians, which
are factors that indicate higher retail store visibility, matter in
spatial retail sales variation.
Interestingly, the indices produced heterogeneous effects on
sales by retail type. The greatest positive effects of accessibility and
centrality to pedestrians are on Medical Services sector sales, fol-
lowed by Education sector sales. Except for Betweenness effects on
the Retail sector, all retail types (full models), and the Food, Retail,
and Services sectors have almost similar effects of spatial accessi-
bility and centrality. Usually, retail stores near highly dense street
networks with higher pedestrian volumes have higher sales reve-
nues. In contrast, Betweenness has no statistically significant effect
on the Retail sector selling various goods. In Seoul, large and small
stores serve wide areas, including roads and subway stations near
residential areas. Thus, the positive and negative effects of
Betweenness vary across store locations.
From the heterogeneous attributes of retail stores, we infer that
spatial accessibility and centrality to pedestrians generate different
effects by retail type. Table 2 indicates that the Medical Services and
Education sectors differ from the other sectors in terms of service
and customers. While Medical Services offer routing and urgent
medical care and medicines, Education supports tutoring and
preparing students from elementary to high schools and even
universities. Usually, hospitals and pharmacies form spatial clusters
because patients with prescriptions tend to buy medicines near
hospitals. A Korean study on hospital locations confirmed that
higher daytime accessibility is a key determinant of a hospital's
location (Lee, 2005). Further, private educational institutions also
co-locate within limited areas owing to the frequent mobility of
students among private educational institutes. Thus, these sectors
tend to locate near highly accessible and large-scale residential
areas to attract potential patients and students (Han & Lee, 2012).
In addition, the Medical Services and Education sectors are highly
sensitive to pedestrian accessibility and centrality because their
clustered locations along well-connected main roads and streets
serve relatively wider zones of potential customers. In contrast, the
Food, Retail, and Service sectors serve a larger volume of general
customers who buy various goods and services. These three sectors
are located widely along street networks and mainly near roads and
key transportation nodes such as subway stations and bus stops.
These spatial patterns imply that the sectors are usually exposed to
dense networks and higher pedestrian volumes.
4.2. Effects of other variables
Other controlled variables partly determine retail sales. The
distance to the CBD and sub-CBDs has no statistically significant
effect on retail sales. Retail stores closer to subway stations, bus
stops, roads, and main retail zones have more sales, contrary to
those closer to schools and parks. This clearly shows that retail
stores near main transportation nodes and networks enjoy a su-
perior economic performance by offering greater access for pe-
destrians. Higher mix of diverse land use, development density, and
commercial density consistently generate higher retail sales. Thus,
we confirm higher mixed land use, dense development, and retail
concentration are favorable urban settings for retail economic
performance. While employment density is positively associated
with sales, population density has the opposite effect. Since most
retail stores share their location with offices in Seoul, higher
employment density is considerably associated with higher retail
sales. As previously mentioned, the models contain dummy vari-
ables for the Food, Medical Services, Retail, and Services sectors
with Education sector as the reference. Medical Services had
highest sales among the retail sectors. Compared with April 2013,
the December 2013 sales were higher but the September 2013 sales
were lower. Generally, Koreans tend to spend more during holiday
seasons such as Christmas and New Year day owing to their larger
volumes of seasonal consumption.
5. Conclusion and policy implications
This empirical study isolated the effects of pedestrian accessi-
bility and centrality on retail sales, controlling for other variables.
We highlight the key findings of this study, which are different
from those of previous studies, in terms of four aspects. First, the
UNA tool presented five accessibility and centrality indices of
pedestrian volume and street configuration concurrently. Distinctly
from previous studies, our indices introduce perspectives on the
spatial connection between walking volume, street network layout,
and retail sales. Second, we compared the various effects of
accessibility and centrality indices to walking volume on retail sales
under a single framework. Higher accessibility to pedestrians,
indicated by the Reach and Gravity Index, and higher visibility to
walkers, indicated by Straightness, tend to increase sales perfor-
mance of all retail types. Closeness and Betweenness showed
weaker effects on retail economic performance, contrary to previ-
ous studies. Third, we attempted to identify how pedestrian volume
along street networks affects retail sales performance, controlling
for location and transportation attributes, land use features, so-
cioeconomic features, retail type, and time of sales. In particular,
this study confirmed that higher walking activities along more
accessible and central streets tend to increase retail sales. Finally,
the indices generated heterogeneous effects on revenue by retail
type. Accessibility and centrality to pedestrians generated the
greatest positive effects on Medical Services sector sales followed
by Education sector. Except for Betweenness effects on the Retail
sector, all the retail types (full models) and the Food, Retail, and
Services sectors’ effects were almost similar for the spatial acces-
sibility and centrality indices. Typically, retail stores near dense
street networks and pedestrian volumes earned higher sales rev-
enue. The strong preference for walkable urban settings and ten-
dency to consume more goods and services at the retail stores along
walkable streets of high income groups may explain this pattern.
However, Betweenness shows no statistically significant effects on
the Retail sector selling various goods.
Our core findings offer a few insightful policy implications for
urban planners and designers seeking to create pedestrian-friendly
urban structures and ensure dynamic retail development. First, the
econometric models verified that the spatial association between
pedestrian volume and street configuration affects retail sales per-
formance. Thus, an urban fabric that efficiently combines walking
volume along street networks and local retail policies can enhance
 C.-D. Kang / Habitat International 57 (2016) 110e120 119

the health of local retail activities. Second, for developing effective
designs, urban designers should consider the features of the retail
type, the local context of street networks, and retail performance. As
our study results indicate, these complex features combining
pedestrian patterns and street connectivity substantially affect local
economic vitality. Third, guidance and advice based on information
on retail fabric, neighborhood socioeconomics, and policies for
transportation networks and land use would support retailers in
their decision-making process on retail location and type. Further,
sophisticated methods and analysis on the link between walking and
retail sales provide effective knowledge to evaluate how urban policy
and design change local retail business, controlling for other relevant
variables. Fourth, attractive place-making and walkable neighbor-
hood strategies promote economic prosperity of local retail. Well-
known places with better amenities and events attract a signifi-
cantly larger number of residents, firms, and tourists, enhancing
neighborhood retail performance. In addition, a well-organized
street network configuration that appeals to pedestrians would be
a highly effective solution to draw various retail stores and increase
the potential demand of residents and visitors. Finally, we need more
expanded perspectives on urban form indicating the physical pattern
of land use and transportation infrastructure and individuals’
behavior. Specifically, the components essential to create pedestrian-
friendly and vibrant retail neighborhoods are a holistic plan and
design allowing close connection between walking and public
transit, comfortable and attractive urban street design, and transport
performance. Nevertheless, further studies on walking and retail
for a longer period and in other cities would help find a more so-
phisticated relationship between walking behavior and retail sales.
This study provides an applicable framework with which to eval-
uate and predict urban policy and design related to pedestrians and
retail development for other cities. Increasingly, city leaders are
turning their attention to developing pedestrian-friendly cities and
ensuring economic prosperity of local businesses. Thus, the meth-
odological approach and empirical results of this study can deepen
their understanding on walking behavior and retail business per-
formance, and support them in creating relevant and effective ur-
ban policies and designs for sustainable and livable cities.
The limitations of the models suggest new topics for future
research. First, we focused on three among 25 wards in Seoul owing
to our limited access to relevant data. Thus, expanding research
over Seoul and other cities should yield more general and appli-
cable conclusions. Second, future studies should investigate aspa-
tial factors, such as quality of service and relative price, to partly
affect retail sales. Because we examine the effects of spatial attri-
butes on retail sales, further study may compare the effects of
spatial and aspatial factors on the variation of economic perfor-
mance of retails. Finally, we may apply the research framework to
examine the effects of multi-modal transportation networks on
urban economic performance. For instance, the combination of
public transit and walking, driving and walking, or bicycling and
walking may significantly explain and predict local retail sales.

Abbreviations

Term Description Administrative districts

Traditional Urban Core A concentration of core government functions Jongro-gu, Joong-gu, Yongsan-gu
(TUC) Various traditional and cultural buildings
Constantly declining populations and employment
Old and deteriorating residential areas
Mixed Core (MC) The National Assembly Building in Youido Kangseo-gu, Guro-gu,
Except for Youido and nearby areas, most districts are mixed, and the development of residential, Yangcheon-gu, Youngdungpo-gu,
commercial, and industrial facilities is unplanned Keumchun-gu,
Lack of transit services and amenities for residents and workers Kwanak-gu, Dongjak-gu
New Urban Core (NUC) Areas planned by the central government Seocho-gu, Songpa-gu Kangnam-gu,
A new business district for global, information technology (IT), and finance firms Kangdong-gu
A residential concentration of high-end buildings having high-income residents
Central Business Area for central government offices, conglomerate headquarters, and large- and small-scale retail Jongro-gu, Joong-gu
District (CBD)
Youido Business Area for financial offices, security trading, broadcasting activities, media companies Youido-dong, Youngdungpo-gu
District (YBD)
Kangnam Business Large-scale areas for office, retail, and service firms Kangnam-gu
District (KBD)

Sources: Jang and Kang (2015), Seoul Metropolitan Government (2014).

and land use reform to improve walking mode choice based on
human behavior.
In conclusion, we confirm the positive effects of access to
pedestrian volumes along street networks on retail sales revenue. If
the link between walking and retail economic performance could
be tested in a rigorous and compelling manner and applied
generally, the results could help create more effective urban de-
signs for retail development and pedestrian-friendly cities. Retail
development and pedestrian-oriented policies are compatible and
even complementary for sustainable and viable cities.
The empirical results and discussion from this study are trans-
ferable to other cities in developing and developed countries. The
effects of access to walking volumes on retail sales varied with
pedestrian flows, street configuration, and retail types. Clearly, local
contexts of higher pedestrian flow, well-connected street networks,
and attractive retail location contribute to improved retail business
Acknowledgement
This research is based on the results of “The Seoul Institute
Research Competition 2015” and the data provided by The Seoul
Institute.
References
Agrawal, A. W., & Schimek, P. (2007). Extent and correlates of walking in the USA.
Transportation Research Part D: Transport and Environment, 12, 548e563.
Baran, P. K., Rodriguez, D. A., & Khattak, A. J. (2008). Space syntax and walking in a
new urbanist and suburban neighbourhoods. Journal of Urban Design, 13, 5e28.
Boarnet, M. G., & Crane, R. (2001). Travel by design: The influence of urban form on
travel. NC, USA: Oxford University Press.
Cerin, E., Macfarlane, D. J., Ko, H.-H., & Chan, K.-C. A. (2007). Measuring perceived
neighbourhood walkability in Hong Kong. Cities, 24, 209e217.
Cervero, R., & Duncan, M. (2003). Walking, bicycling, and urban landscapes: Evi-
dence from the San Francisco Bay area. American Journal of Public Health, 93,
120 C.-D. Kang / Habitat International 57 (2016) 110e120
1478e1483.
Chen, H., Jia, B., & Lau, S. (2008). Sustainable urban form for Chinese compact cities:
Challenges of a rapid urbanized economy. Habitat International, 32, 28e40.
Chiaradia, A., Hillier, B., Schwander, C., & Barnes, Y. (2013). Compositional and urban
form effects on residential property value patterns in Greater London. Pro-
ceedings of the ICE-Urban Design and Planning, 166, 176e199.
Cortright, J. (2009). Walking the walk: How walkability raises home values in U.S.
cities. Washington, D.C.: CEOs for Cities.
Crane, R., & Crepeau, R. (1998). Does neighborhood design influence travel?: A
behavioral analysis of travel diary and GIS data. Transportation Research Part D:
Transport and Environment, 3, 225e238.
Duncan, M. (2011). The impact of transit-oriented development on housing prices
in San Diego, CA. Urban Studies, 48, 1e25.
Ellen, I. G., & Voicu, I. (2005). Nonprofit housing and neighborhood spillovers.
Journal of Policy Analysis and Management, 25, 31e52.
Ewing, R., & Cervero, R. (2001). Travel and the built environment: A synthesis.
Transportation Research Record, 87e113, 1780.
Forsyth, A., Hearst, M., Oakes, J. M., & Schmitz, K. H. (2008). Design and destina-
tions: Factors influencing walking and total physical activity. Urban Studies, 45,
1973e1996.
Frank, L. D., & Pivo, G. (1994). Impacts of mixed use and density on utilization of
three modes of travel: Single-occupant vehicle, transit, and walking. Trans-
portation Research Record, 44e52, 1466.
Glaeser, E. L., Kolko, J., & Saiz, A. (2001). Consumer city. Journal of Economic Geog-
raphy, 1, 27e50.
Guy, C. (2006). Planning for retail development: A critical view of the British experi-
ence. Routledge.
Handy, S. (2005). Smart growth and the transportation-land use connection: What
does the research tell us? International Regional Science Review, 28, 146e167.
Handy, S., & Niemeier, A. D. (1997). Measuring accessibility: An exploration of issues
and alternatives. Environment and Planning A, 29, 1175e1194.
Han, Y., & Lee, S. (2012). Study on the reality of the private educational institute
street and the spatial range of its service in Pyeongchon, Anyang-si. Journal of
the Economic Geographical Society of Korea, 15, 721e734 (Korean version).
Hansen, W. G. (1959). How accessibility shapes land use. Journal of the American
Institute of Planners, 25, 73e76.
Hass-Klau, C. (1993). Impact of pedestrianization and traffic calming on retailing: A
review of the evidence from Germany and the UK. Transport Policy, 1, 21e31.
Huff, D. L. (1963). A probabilistic analysis of shopping center trade areas. Land
Economics, 39, 81e90.
Jang, M., & Kang, C.-D. (2015). Retail accessibility and proximity effects on housing
prices in Seoul, Korea: A retail type and housing submarket approach. Habitat
International, 49, 516e528.
Kang, C. D. (2015). The effects of spatial accessibility and centrality to land use on
walking in Seoul, Korea. Cities, 46, 94e103.
Krizek, K. J., & Johnson, P. J. (2006). Proximity to trails and retail: Effects on urban
cycling and walking. Journal of the American Planning Association, 72, 33e42.
Kuby, M., Tierney, S., Roberts, T., & Upchurch, C. (2005). A comparison of geographic
information systems, complex networks, and other models for analyzing trans-
portation network topologies. NASA Langley Research Center.
Lang, R. (2003). Edgeless cities: Exploring the elusive metropolis. Brookings Institution
Press.
Lee, K. S. (2005). Distribution characteristics of the medical services in Korea.
Journal of Korea Geographical Society, 40, 242e251 (Korean version).
Lee, C., & Moudon, A. V. (2006). The 3Ds þ R: Quantifying land use and urban form
correlates of walking. Transportation Research Part D: Transport and Environ-
ment, 11, 204e215.
Leinberger, C. B., & Alfonzo, M. (2012). Walk this way: The economic promise of
walkable places in metropolitan Washington. Washington, DC: DC. The Brookings
Institution.
Litman, T. (2003). Economic value of walkability. Transportation Research Record:
Journal of the Transportation Research Board, 1828, 3e11.
Litman, T. (2014). Economic value of walkability. Paper, Victoria Transport Policy
Institute. presented at the Transportation Research Board. In 82nd Annual
Meeting.
Loo, B. P., & Chow, S. Y. (2006). Sustainable urban transportation: Concepts, policies,
and methodologies. Journal of Urban Planning and Development, 132(2), 76e79.
Lynch, K. (1960). The image of the city. The MIT Press.
Moudon, A. V., Lee, C., Cheadle, A. D., Garvin, C., Johnson, D., Schmid, T. L., et al.
(2006). Operational definitions of walkable neighborhood: Theoretical and
empirical insights. Journal of Physical Activity & Health, 3, 99e117.
New York Department of Transportation. (2013). The economic benefits of sustainable
streets (New York).
Omer, I., & Goldblatt, R. (2015). Spatial patterns of retail activity and street network
structure in new and traditional Israeli cities. Urban Geography, 37(4), 1e21.
Ozbil, A., Peponis, J., & Stone, B. (2011). Understanding the link between street
connectivity, land use and pedestrian flows. Urban Design International, 16,
125e141.
Peponis, J. (1989). Space, culture and urban design in late modernism and after.
Ekistics, 56, 93e108.
Pivo, G., & Fisher, J. D. (2011). The walkability premium in commercial real estate
investments. Real Estate Economics, 39, 185e219.
Porta, S., Crucitti, P., & Latora, V. (2005). The network analysis of urban streets: A
primal approach. Environment and Planning B, 35, 705e725.
Porta, S., Strano, E., Iacoviello, V., Messora, R., Latora, V., Cardillo, A., et al. (2009).
Street centrality and densities of retail and services in Bologna, Italy. Environ-
ment and Planning B: Planning and Design, 36(3), 450e465.
Rabe-Hesketh, S., & Skrondal, A. (2008). Multilevel and longitudinal modeling using
STATA. STATA Press.
Rauterkus, S., Thrall, G., & Hangen, E. (2010). Location efficiency and mortgage
default. Journal of Sustainable Real Estate, 2, 117e141.
Robertson, K. A. (1993). Pedestrianization strategies for downtown planners: Sky-
walks versus pedestrian malls. Journal of the American Planning Association, 59,
361e370.
Schneider, R. J. (2015). Walk or drive between stores? Designing neighbourhood
shopping districts for pedestrian activity. Journal of Urban Design, 20, 212e229.
Schuetz, J., Kolko, J., & Meltzer, R. (2012). Are poor neighborhoods “retail deserts”?
Regional Science and Urban Economics, 42, 269e285.
Scoppa, M. D., & Peponis, J. (2015). Distributed attraction: The effects of street
network connectivity upon the distribution of retail frontage in the city of
Buenos Aires. Environment and Planning B: Planning and Design, 42(2), 354e378.
Seoul Institute. (2015). The Seoul research data service. Seoul.
Seoul Metropolitan Government. (2014). Seoul master plan. Seoul.
Sevtsuk, A. (2014). Location and agglomeration: The distribution of retail and food
businesses in dense urban environments. Journal of Planning Education and
Research, 34, 374e393.
Sevtsuk, A., & Mekonnen, M. (2012). Urban network analysis. A new toolbox for
ArcGIS. Revue Internationale de G eomatique, 22, 287e305.
Shyr, O., Andersson, D. E., Wang, J., Huang, T., & Liu, O. (2013). Where do home
buyers pay most for relative transit accessibility? Hong Kong, Taipei and
Kaohsiung compared. Urban Studies, 50, 2553e2568.
SKT Geovision. (2013). Data of pedestrian volume and retail sales.
Sohn, D. W., Moudon, A. V., & Lee, J. (2012). The economic value of walkable
neighborhoods. Urban Design International, 17, 115e128.
Song, Y., & Knaap, G.-J. (2004). Measuring urban form: Is Portland winning the war
on sprawl? Journal of the American Planning Association, 70, 210e225.
Statistics Korea. (2012). The annual firm survey and information on Korean census
tracts.
Sullivan, M., & Adcock, D. (2002). Retail marketing. Cengage learning EMEA.
Sung, H.-G., Go, D.-H., & Choi, C. G. (2013). Evidence of Jacobs's street life in the
great Seoul city: Identifying the association of physical environment with
walking activity on streets. Cities, 35, 164e173.
Talen, E., & Koschinsky, J. (2013). The walkable neighborhood: A literature review.
International Journal of Sustainable Land Use and Urban Planning, 1(1), 42e63.
Tatian, P. A., Kingsley, G. T., Parilla, J., & Pendall, R. (2012). Building successful
neighborhoods. Washington, DC: The Urban Institute.
Tiwari, R., Cervero, R., & Schipper, L. (2011). Driving CO2 reduction by integrating
transport and urban design strategies. Cities, 28, 394e405.
Vojnovic, I., Jackson-Elmoore, C., Holtrop, J., & Bruch, S. (2006). The renewed in-
terest in urban form and public health: Promoting increased physical activity in
Michigan. Cities, 23, 1e17.
Waddell, P., & Ulfarsson, G. F. (2003). Accessibility and agglomeration: Discrete-
choice models of employment location by industry sector. In 82nd annual
meeting of the transportation research board (Vol. 63)Washington, DC: Citeseer.
Wang, F., Chen,, C., Xiu, C., & Zhang, P. (2014). Location analysis of retail stores in
Changchun, China: A street centrality perspective. Cities, 41, 54e63.
Wey, W.-M., & Chiu, Y.-H. (2013). Assessing the walkability of pedestrian environ-
ment under the transit-oriented development. Habitat International, 38,
106e118.
Xiao, Y., Webster, C., & Orford, S. (2014). Identifying house price effects of changes in
urban street configuration: An empirical study in Nanjing, China. Urban Studies,
53(1), 112e131.
Zhao, P. (2014). The impact of the built environment on bicycle commuting: Evi-
dence from Beijing. Urban Studies, 51, 1019e1037.