Content uploaded by Linchuan Yang
Author content
All content in this area was uploaded by Linchuan Yang on Jan 18, 2023
Content may be subject to copyright.
1
COVID-19 moderates the association
between to-metro and by-metro accessibility and house prices
Linchuan Yang a, Yuan Liang a,b, Baojie He c,d,e,f, Hongtai Yang g, Dong Lin h*
aDepartment of Urban and Rural Planning, School of Architecture, Southwest Jiaotong
University, Chengdu, China
bDepartment of Geography, Hong Kong Baptist University, Kowloon Tong, Hong Kong,
China
cCentre for Climate-Resilient and Low-Carbon Cities, School of Architecture and Urban
Planning, Chongqing University, Chongqing, China
dInstitute for Smart City of Chongqing University in Liyang, Chongqing University, Liyang,
Jiangsu, China
eKey Laboratory of New Technology for Construction of Cities in Mountain Area, Ministry
of Education, Chongqing University, Chongqing, China
fNetwork for Education and Research on Peace and Sustainability (NERPS), Hiroshima
University, Hiroshima, Japan
gSchool of Transportation and Logistics, Southwest Jiaotong University, Chengdu, China
hSchool of Engineering, University of Aberdeen, Aberdeen, United Kingdom
E-mail: yanglc0125@swjtu.edu.cn (L. Yang), yuanliang@my.swjtu.edu.cn (Y. Liang),
baojie.he@cqu.edu.cn (B. He), yanghongtai@swjtu.cn (H. Yang), dong.lin@abdn.ac.uk (D. Lin)
* Corresponding author
To cite: Yang, L., Liang, Y., He, B., Yang, H., & Lin, D. (2023). COVID-19
moderates the association between to-metro and by-metro accessibility and house
prices. Transportation Research Part D: Transport and Environment,114, 103571.
Abstract: Previous studies extensively examined the role of accessibility to metro in
shaping house prices but largely overlooked the contribution of accessibility by metro.
In addition, limited studies examined the moderating effect of COVID-19 on the price
effects of to-metro and by-metro accessibility. Based on multilevel hedonic price and
quantile regression models, this study scrutinizes the association between to-metro
accessibility, by-metro accessibility, and house prices in Chengdu, China, and
examines the moderating role of COVID-19 in this association. We show that
by-metro accessibility significantly influences house prices. COVID-19 significantly
influences the value of to-metro accessibility but marginally affects that of by-metro
accessibility. The value of to-metro accessibility is disproportionately affected by the
pandemic. Specifically, small or low-priced houses are less affected than big or
high-priced houses. In other words, the flattening of the to-metro price gradient is
more discernible for big or high-priced houses. The changing preference of residents
has also been verified by the decreases in house transaction volume in metro-adjacent
areas.
Keywords: transit accessibility, moderator, COVID-19 pandemic, property price,
apartment price, housing market
2
1. Introduction
The transport sector, including ground, maritime, and air transport, produces
approximately a quarter of greenhouse gas emissions worldwide (Linton et al., 2015).
Therefore, reducing transport-related greenhouse gas emissions is crucial. As such,
green transport, including mass transit (e.g., metro, commuter rail, and bus) and active
transport (e.g., walking and biking), is extensively advocated worldwide (Cao and
Porter-Nelson, 2016; Mao et al., 2022; Yang et al., 2021). The promotion of mass
transit has long been regarded as an essential transport planning objective (Bao et al.,
2022).
As a typical mass transit mode, the metro has gained wide currency for its
functions of enhancing urban mobility (e.g., reducing commuting time), ameliorating
traffic congestion, guiding urban development, and promoting transit-oriented
development (TOD) practice (Lin et al., 2022a; Lin et al., 2021). The development of
metro networks could contribute to sustainable urban development by incorporating
TOD. It is always effective in promoting economic growth, mixed land use, and real
estate development near stations (Wang et al., 2022). These impacts of metro
development are compatible with the core principles and goals of TOD. Therefore,
TOD needs to be adopted and incorporated into metro development, wherever
possible, to achieve sustainable urban development outcomes. In addition, the metro
has discernible socioeconomic impacts, such as increasing the value of nearby
properties and improving residents’ life satisfaction (Chen et al., 2022; Lin et al.,
2022b; Yin et al., 2021).
Metro accessibility can be broadly categorized as “transit accessibility.” Transit
accessibility includes not only to-transit accessibility (access to nearby transit stations)
but also by-transit accessibility (ease of accessing essential destinations by transit)
(Yang et al., 2020a). By-transit accessibility has tremendous implications for
travel/activity behavior (Moniruzzaman and Páez, 2012), mobility level (Lessa et al.,
2019), transit ridership (Pan et al., 2017), and property price (He, 2020; Liu et al.,
2022; Yang et al., 2020a). A summary of the distinction, measurement, and
terminologies of to-transit and by-transit accessibility can be found in the research by
Yang et al. (2020a). However, previous studies on the connection of metro
accessibility with property prices primarily focused on the role of to-metro
accessibility, which is normally assessed as the Euclidean or network distance to the
nearest station or a dummy variable (= 1 for a distance shorter than a cut-off value and
0 otherwise), but largely overlooked the role of by-metro accessibility. Exceptions are
limited (see the ensuing section).
The coronavirus disease 2019 (COVID-19) pandemic hit the globe at breakneck
speed and put it on pause periodically. This disruptive and unprecedented event
affects the lives of people everywhere. Unlike many medical, pharmacological, and
public health studies relevant to COVID-19, studies focusing on its socioeconomic
implications are very limited in number. Moreover, COVID-19 makes people
reconsider the valuation of property attributes such as access to the city center and
neighborhood density (Cheung et al., 2021; Gupta et al., 2022; Huang et al., 2022).
Ample evidence has suggested that the COVID-19 pandemic “weakens” the positive
association between access to the city center/neighborhood density and property
prices. In the same vein, the pandemic is likely to change the value of (or residents’
willingness to pay for) metro accessibility (Liu and Su, 2021; Yang et al., 2022b)
because the metro suffers an eclipse during the pandemic, which is directly evidenced
by its shrinking ridership (Yang et al., 2022b). In other words, the pandemic depresses
the attraction of the metro and thus may decrease the value of its accessibility.
3
Empirically, the interaction between to-metro and by-metro accessibility and property
prices in the COVID-19 setting remains to be fully scrutinized.
Using numerous house transaction records (N = 72,187) before and after the
explosion of COVID-19 (spanning two years), this study analyzes the association
between metro accessibility (specifically, to-metro and by-metro accessibility
combined) and house prices in Chengdu, a megacity in southwest China. In addition
to examining global or average effects using hedonic price models, this study
scrutinizes the heterogeneous effects (more specifically, in small- and big-house
submarkets and across the entire distribution of house prices) using separate modeling
or quantile regression models. This study is the pioneer in examining the
heterogeneity in the value of transit accessibility on the conditional distribution of
house prices in the era of COVID-19. Based on the above endeavor, this study calls
for the use of time-varying and category-specific value-capture tools.
The principal contributions of this paper include: (1) introducing the two-part
transit accessibility analysis approach to the “metro-property price” research field; (2)
assessing and valuing by-metro accessibility, an under-studied house attribute; (3)
investigating the moderating role of COVID-19 on the value of to-metro and by-metro
accessibility; (4) examining the heterogeneity in the moderating role; and (5)
confirming that COVID-19 decreases the value of to-metro accessibility for big or
high-priced houses by a larger extent than for small or low-priced ones. This finding
can be explained by the inelastic demand of economically restrained people
(represented by the purchasers of small and low-priced houses) for metro use.
The remainder of this paper is organized as follows. Section 2 reviews two
strands of research: (1) accessibility to rail, by rail, and property prices, and (2)
COVID-19 impacts on the value of property attributes. Section 3 succinctly describes
Chengdu’s metro system and the data. Section 4 shows the methodology. Section 5
presents the modeling results. Section 6 provides conclusions and lists research
limitations.
2. Related studies
2.1. Accessibility to rail, by rail, and property prices
Due to its policy/practice relevance, the capitalization effect of transit has long
been a popular research topic (So et al., 1997). Its theoretical ground is the
Alonso-Mills-Muth model. Among the investigated transit modes, urban rail transit
(including metro and commuter rail) has received much more scholarly attention than
other modes (e.g., bus rapid transit and conventional bus transit) (Bowes and
Ihlanfeldt, 2001). Among the investigated land/property categories, attention to
houses far exceeds that to office, retail, and industrial land/properties. A few
researchers (Bartholomew and Ewing, 2011; Debrezion et al., 2007; Wu et al., 2020)
provided valuable summaries on this strand of research.
Since at least the 1970s, various studies have been performed to explore the
intricate connection between to-rail accessibility and property (or land) prices (Davis,
1970). North American, European, and Oceanian cities have received the most
attention. Boyce et al. (1972) analyzed the impact of the Lindenwold transit line,
Phase 1, on suburban property prices and found that the line positively affects
property prices. Damm et al. (1980) developed a battery of hedonic price models to
explore the association of accessibility to the Washington metro with house and retail
shop prices. They concluded that to-metro accessibility impacts property prices
considerably, especially retail shop prices. Bowes and Ihlanfeldt (2001) predicted
property prices in Atlanta as a function of to-rail accessibility and determined
4
significant metro access (to-metro accessibility) premiums. They suggested high price
premiums exist within one to three miles of rail stations. Moreover, they proposed that
the holistic or full influence of rail on property prices is a sum of positive effects
(reducing commuting costs and attracting retail activities) and negative effects
(offering nuisances such as pollution and increasing neighborhood crimes). To put it
simply, many possible channels for the transmittal of price externalities exist. Cervero
and Duncan (2002) affirmed that accessibility to light rail and commuter rail
significantly affects commercial land prices in a Californian city. Wu et al. (2015)
concluded that residential and commercial land parcels experience price premiums
attributable to increases in to-metro accessibility. Of the two parcel types, the former
benefits slightly more than the latter. Mulley et al. (2018) adopted geographically
weighted regression models to analyze the interaction between accessibility to light
rail and property prices in Sydney, Australia, and revealed that house values benefit
from accessibility to light rail, and the rail has a greater impact in the places far from
the city center.
The association of to-rail accessibility with property (or land) prices has entered
public discourse in Asian cities in the past decades (Yang et al., 2018). Bae et al.
(2003) concentrated on the connection of to-metro accessibility with house prices in
Seoul, South Korea, and indicated that to-metro accessibility had a significant effect
only before the metro began operation (anticipatory effect). Pan and Zhang (2008)
identified significant metro access premiums in Shanghai based on hedonic price
modeling. Xu et al. (2016) investigated the commercial property price impacts of
to-metro accessibility based on spatial autoregressive models and found that metro
access premiums within 100 m of stations are approximately twice those within
100-400 m. Dai et al. (2016) found that transfer stations of the metro have greater
values than regular stations in Beijing, and premium differentials in the suburban area
exceed those in the central city. Diao et al. (2017) examined the to-metro accessibility
effect on property prices in Singapore using difference-in-differences (DID) models
and revealed approximately 8.6% metro access premiums. Based on hundreds of land
transaction observations, Yang et al. (2016) found limited evidence to suggest the
significant effect of to-metro accessibility on land prices in Shenzhen. By contrast,
Yang et al. (2020b) observed that to-metro accessibility influences property prices in a
non-linear way in Shenzhen. They identified a non-linear trend over space.
In a departure from the overwhelming majority of studies focusing solely on the
connection of to-rail accessibility with property prices, a handful of studies captured
by-metro accessibility and reached mixed, even contradictory, conclusions. Bajic
(1983) analyzed the influence of by-rail accessibility (quantified by rail ride time to
five essential destinations) on house prices in Toronto, Canada, and revealed that
by-rail accessibility significantly shapes property prices. Armstrong and Rodriguez
(2006) developed a time-based by-rail accessibility measure and examined its
connection with single-family house values in a US region. They revealed that the
relationship varies across the functional forms of hedonic price models. Debrezion et
al. (2011) scrutinized the association between a composite by-rail accessibility
measure and property prices in a handful of metropolitan areas in the Netherlands.
They showed that by-rail accessibility is important in two out of three areas. Pan et al.
(2014) modeled the connection of by-rail accessibility with house prices in Houston
and suggested that this connection is significant. Yang et al. (2016) assessed the
relationship between a distance-based by-metro accessibility measure and property
prices in Shenzhen using hedonic price models. They suggested that the
distance-based by-metro accessibility measure plays a significant role. Li et al. (2019)
5
concluded that cumulative opportunities-based by-rail accessibility variables
significantly affect property prices in Beijing. He (2020) minutely analyzed the
importance of incorporating by-metro accessibility in real estate valuation studies and
used multilevel hedonic price models and DID models to identify the property price
effect of gravity-based by-metro accessibility in Hong Kong. The author demonstrated
that by-metro accessibility plays an essential role in determining property prices.
2.2. COVID-19 impacts on the value of property attributes
The alteration of COVID-19 to the value of property attributes (e.g., regional
location and population density) has been explored in various studies. A few property
attributes are studied, and many findings are obtained. For example, COVID-19
weakens the association between property prices and property attributes such as
access to the city center, high density, and transit accessibility (Rosenthal et al., 2022).
Location attributes (e.g., access to the city center) have received the greatest
scholarly attention. Gupta et al. (2022), Pawson et al. (2022), and Rosenthal et al.
(2022) found that premiums for access to the city center diminished in many places
during the pandemic and concluded “the flattening of the bid-rent curve.” In other
words, the pandemic makes residents flee from city centers to suburbs or rural areas
because of (1) low population density and perceived contagious risks in suburbs or
rural areas (Gupta et al., 2022; Huang et al., 2022) and (2) the rapid rise of working
from home (remote working), which reduces demand for downtown living (D'Lima et
al., 2022). Rosenthal et al. (2022) further pointed out the heterogeneity in the
pandemic effect on the premiums and revealed that transit-dependent cities (those
relying heavily on transit) are greatly impacted by the pandemic, while car-oriented
cities are marginally affected.
Neighborhood attributes, such as population density and proximity to hospitals,
have also been studied in several previous studies. Cheung et al. (2021), Liu and Su
(2021), D'Lima et al. (2022), and Rosenthal et al. (2022) suggested that premiums for
high population or employment density abated, providing additional evidence to
support the decreasing value of access to the city center. The reason is that access to
the city center is highly correlated with neighborhood density. Moreover, Liu and
Tang (2021), Qian et al. (2021), and Huang et al. (2022) determined that in many
cities in China, properties located in neighborhoods with COVID-19 infection are
valued less than others. Furthermore, Huang et al. (2022) identified that in 60 Chinese
cities, the pandemic negatively influences the properties located closer to hospitals
more than others because the former has higher infection risks than the latter.
Moreover, Li et al. (2021) determined that houses in a gated community are valued
more during the pandemic than those in an open community. Finally, Cheung et al.
(2021) suggested that proximity to the COVID-19 epicenter (Hunan Seafood Market)
adversely influences property prices in Wuhan, China.
In addition to location and neighborhood attributes, the structural attributes of a
property have also garnered scattered attention. Pawson et al. (2022) demonstrated
that the prices of detached houses increased faster compared to apartments after the
COVID-19 pandemic in many countries. Huang et al. (2022) concluded that the
pandemic decreases the value of tall buildings because such buildings involve high
interaction opportunities and infection risks. Huang et al. (2022) also indicated that
residents value high-floor properties more (compared to low-floor properties) during
the pandemic.
To our knowledge, only two studies examined the link of accessibility to rail
with property prices in the context of the pandemic. Yang et al. (2022b) developed
6
multilevel DID models to assess the value of to-metro accessibility before and after
the pandemic explosion in Chengdu, China. Rosenthal et al. (2022) presented an
empirical analysis of the association of the pandemic with commercial property
premiums for accessibility to rail in dozens of US cities and concluded that the
pandemic reduces the value of accessibility to rail. The above two studies provide a
prima facie case on accessibility to rail and property prices in the COVID-19 setting.
Their centerpiece is that the pandemic flattens the to-metro price gradient. A
compelling reason is that the pandemic weakens the attraction of the metro, which is
supported by real-world ridership data.
3. Data
3.1. Chengdu’s metro system
Chengdu is a megacity in Sichuan Province, a national central city, and a
megacity integrating ancient civilization and modern civilization in southwest China.
It is the center and window of the land of abundance (tianfu zhi guo). By the end of
2021, the city had a landmass of 14,335 km2(5535 mile2), a permanent population
(changzhu renkou) of 21.19 million, and an urbanization rate of 79.48%. Its GDP
(gross domestic product) reached 1.992 trillion yuan.
7
Fig. 1. Location of Chengdu and spatial distribution of house samples.
Source: self-elaboration.
Chengdu’s rapid development demands an efficient and high-quality metro
system. As of 2021, Chengdu had 12 metro lines (totaling 518.54 km) and 338
stations. Surprisingly, Chengdu simultaneously opened five new metro lines on 18
December 2020. The city has recently experienced rapid metro development (opening
500 km of metro lines in ten years) and is now regarded as the “fourth city” in terms
of urban rail-based transit in China. Tables 1 and 2 provide the details of the Chengdu
8
metro projects as of 2021 and the projects under construction, respectively.
Table 1
Details of the Chengdu metro projects as of 2021
Metro project
Length
(km)
Number of
stations
Ground broken
time
Operation time
Line 1, Phase 1
18.52
17
2005.12
2010.9.27
South extension of
Line 1
5.41
5
2012.8
2015.7.25
Section 1 of Line 1,
Phase 3
10.95
8
2014.12
2018.3.18
Section 2 of Line 1,
Phase 3
6.12
5
2015.8
2018.3.18
Line 2, Phase 1
22.38
20
2007.12
2012.9.16
West extension of
Line 2
8.77
6
2010.8
2013.6.8
East extension of
Line 2
11.12
6
2011.12
2014.10.26
Line 3, Phase 1
20.36
17
2012.4
2016.7.31
Line 3, Phases 2 & 3
29.54
20
2015.11
2018.12.26
Line 4, Phase 1
22.00
16
2011.12
2016.11
Line 4, Phase 2
21.48
14
2014.4
2017.6.2
Line 5, Phases 1 & 2
49.01
41
2015.9
2019.12.27
Line 6, Phases 1 & 2
46.80
38
2016.10
2020.12.18
Line 6, Phase 3
22.08
18
2017.3
2020.12.18
Line 7
38.63
31
2013.9
2017.12.6
Line 8, Phase 1
29.10
25
2016.12
2020.12.18
Line 9, Phase 1
22.18
13
2016.12
2020.12.18
Line 10, Phase 1
10.94
6
2014.5
2017.9.6
Line 10, Phase 2
27.04
10
2016.12
2019.12.27
Line 17, Phase 1
26.15
9
2017.2
2020.12.18
Line 18, Phase 1
41.20
8
2016.8
2020.9.27 (first
round)
2020.12.18
(second round)
Line 18, Phase 2
28.79
5
2017.6
2021.6.27
Sum
518.54
338
-
9
Table 2
Details of the Chengdu metro projects under construction
Metro project
Length
(km)
Number of
stations
Ground broken
time
Expected operation
time
Line 10, Phase 3
5.87
5
2019.12
2024
Line 13, Phase 1
29.07
21
2019.12
2024
Line 17, Phase 2
24.76
18
2019.12
2024
Line 18, Phase 3
13.55
5
2019.12
2024
Line 19, Phase 2
42.87
12
2019.11
2023
Line 8, Phase 2
7.81
7
2020.5
2024
Line 27, Phase 1
24.86
23
2020.5
2024
Line 30, Phase 1
26.28
23
2020.5
2024
Ziyang Line
38.7
7
2020.11
2024
Sum
213.77
121
-
3.2. Data description
This study obtains house transaction data between January 2019 and December
2020 from the official website of two well-known real estate agencies, Lianjia and
Beike. The data include detailed information on the price, residential district (juzhu
xiaoqu) location, and many structural parameters (e.g., size and house orientation) of
transacted houses. The houses that are too far from the metro (> 3 km), which are
deemed irrelevant to our analysis, are excluded (Chu et al., 2021). Moreover, those
adjacent to metro stations on the lines opened in 2019 and 2020 are also discarded to
eliminate the effect of the metro-opening shock on the house market. Altogether,
72,187 houses are retained for analysis. Fig. 1 reveals the spatial distribution of house
samples. It suggests that the observations are distributed evenly in space.
The above data have limited hedonic variables. As such, an ArcGIS framework
incorporating multi-source data is established for variable quantification. We
geo-code house samples and measure some neighborhood variables and all location
variables. The POI data are obtained from OpenStreetMap (www.openstreetmap.org).
Mobile phone signal data, which is used to infer the distribution of employment
opportunities, are obtained from China Unicom Smart Steps (zhongguo liantong
zhihui zuji), which is given by one of the three largest mobile operators (China
Unicom Limited) in China. Dense mobile phone base stations in the central city
ensure that the data can accurately reflect the characteristics of human activities.
Moreover, the data used in the current study have high accuracy in assessing urban
dynamics. China Unicom Limited tested the correlation of population parameters
derived from the data and the official Street-level population data (the 7th National
Population Census of China). The fairly high correlations obtained reveal the high
quality of the mobile phone signal data. Fig. 2 shows the spatial distribution of
employment opportunities, which is derived from the above data.
10
Fig. 2. Distribution of employment opportunities in Chengdu.
Source: Self-elaboration based on mobile phone signal data
4. Methodology
4.1. Hedonic price model
The hedonic price model hinges upon the assumption that the price of a
commodity is determined by its attributes. As a result, commodity prices are an output
of component prices (implicit prices, shadow prices, or hedonic prices). Two
researchers—Lancaster (1966) and Rosen (1974)—provided a sound theoretical basis
for the hedonic price model. Lancaster invented the “characteristics theory” of
consumer demand and innovatively proposed the hedonic hypothesis that goods can
be characterized by their component attributes, or what determines utility (or
consumer preference orderings) is not goods themselves but their characteristics (i.e.,
attributes). Based on the hedonic hypothesis, Rosen (1974) sketched a model of
product differentiation and modeled consumer and producer decisions as market
equilibrium.
In real estate valuation studies, the hedonic price model is often developed, with
the predicted variable being the property price and predictor variables being
observable property attributes (Lai et al., 2021; Zou et al., 2022). There are three
categories of hedonic attributes: structural (micro-scale), neighborhood (meso-scale),
and location (macro-scale) attributes. These categories are often adopted in studies
using various property attributes to explain variations in property prices (Wu and
Dong, 2014).
Basically, the hedonic price model can take three forms, namely linear, semi-log
(the predicted variable converted to natural logarithmic form), and double-log models
(the predicted and predictor variables converted to natural logarithmic form). It can be
expressed as follows.
Linear model:
11
Semi-log model:
Double-log model:
where is the transaction price of house i,is a constant (intercept), is the
n-th hedonic variable of house i,is the coefficient of the n-th hedonic variable, N
is the number of hedonic variables, and is an error term.
(the vector with its elements , n = 0, 1, … , N) is estimated as
by
minimizing the following objective function:
.
In the traditional hedonic price model, predictor variables are at only one level.
However, in China, houses are clustered in gated or open communities. Overlooking
the hierarchical (tree-like) structure and using the single-level model would lead to
biased estimates and erroneous interpretations (Yang et al., 2022a). Therefore, the
multilevel hedonic price model should be used to avoid the biases inherent in the
single-level model. It is displayed below.
Multilevel linear model:
Multilevel semi-log model:
Multilevel double-log model:
where is the transaction price of house ithat is located in residential district j,
is a constant, is the n-th hedonic variable of house ithat is located in residential
district j,is the coefficient of the n-th hedonic variable, Nis the number of
hedonic variables, is the random intercept for residential district j, and is an
error term.
4.2. Quantile regression model
In contrast with the above models estimating the conditional mean of the
predicted variable, the quantile regression model is concerned with predicting the
conditional quantiles (e.g., 0.25 quantile, median, and 0.75 quantile) of the predicted
variable. The quantile regression model can assess the effects of predictors on the
predicted variable along the distribution of the predicted variable. To put it simply, it
allows the effects of predictor variables to differ over the quantiles, thereby obtaining
additional information and offering a more comprehensive picture of the connection
12
between house attributes and house prices than traditional regression models.
Moreover, the quantile regression model is much more robust to outliners than the
above models. It is applied in numerous research areas, such as finance, economics,
bio-medicine, geography, engineering, and transport (Cheng et al., 2022).
Quantile regression models in this study can be expressed as
where is a constant, is the coefficient of the n-th hedonic variable
associated with the -th quantile, and other variables are defined as above.
(the vector with its element , n = 0, 1, … , N) is estimated as
by
minimizing the following objective function:
.
Asymmetrical weights are normally used. Only median regression (q= 0.5)
employs symmetrical weights. If q= 0.5, the above function is equivalent to
.
4.3. Predicted and predictor variables
Tables 3 and 4 show the description and summary, respectively, of the variables.
The predictor variables can be classified into four groups: structural, neighborhood,
location, and time variables.
Structural variables include size, age, bedroom dummies, living room dummies,
bathroom dummies, floor dummies, and house orientation. Floor dummies are created
based on the information obtained from the real estate agency websites. For example,
for an 18-floor building, properties between the first and the sixth floors (both
inclusive) are defined as “low-floor,” those between the seventh and the twelve floors
as “intermediate-floor,” and those between the thirteenth and the eighteenth floors as
“high-floor” (Huang et al., 2022).
Neighborhood variables include floor area ratio, green space ratio, access to
schools, access to hospitals, access to shopping centers, and distance to the metro
station.
Location variables include distance to Tianfu Square, distance to Financial City
(jinrong cheng), and employment accessibility by metro. Tianfu Square is the
traditional and geometric city center. Financial City is the new CBD of the city, and it
is located at the junction of five traditional central districts (wu chengqu) and
Chengdu Hi-Tech Industrial Development Zone.
Given that COVID-19 considerably disrupts the house market (D'Lima et al.,
2022; Gupta et al., 2022), we create a COVID-19 dummy. Moreover, to capture the
fixed year-month effects, we create twenty-four time dummies (2019M01,
2019M02, …, 2020M11, and 2020M12).
The quantile plot of the predicted variable is provided in Fig. 3. The house price
distribution is heavily skewed.
13
Fig. 3. Quantile plot of the predicted variable
Source: self-elaboration.
The explanatory variables of metro accessibility are carefully chosen. The most
intuitive and extensively-used variable–distance to the metro station–is chosen to
evaluate to-metro accessibility. Moreover, informed by previous studies (Cervero and
Duncan, 2002; Giuliano et al., 2010; Ko and Cao, 2013; Li et al., 2019), a cumulative
opportunities-based accessibility measure, which implies that more accessible
opportunities translate to more choices for residents, is chosen to assess by-metro
employment accessibility. The choice of the 45-min cut-off value is informed by
Cervero and Duncan (2002) and Li et al. (2019).
Table 3
Description of the variables
Variables
Description
Unit
Predicted variable
House price
The total transaction price of a house
104yuan
Predictor variable
Size
Gross floor area
m2
Age
-
year
Bedroom1
= 1 if having one bedroom and 0 otherwise (serving as the
baseline)
Bedroom2
= 1 if having two bedrooms and 0 otherwise
Bedroom3
= 1 if having three bedrooms and 0 otherwise
Bedroom4+
= 1 if having four or more bedrooms and 0 otherwise
Living room
0&1
= 1 if having at most one living room and 0 otherwise
(serving as the baseline)
Living room2+
= 1 if having two living or more rooms and 0 otherwise
Bathroom 0&1
= 1 if having zero or one bathroom and 0 otherwise (serving
as the baseline)
Bathroom2+
= 1 if having two or more bathrooms and 0 otherwise
Low floor
= 1 if located on the low floor and 0 otherwise (serving as
the baseline)
Intermediate
floor
= 1 if located on the intermediate floor and 0 otherwise
14
High floor
= 1 if located on the high floor and 0 otherwise
South
orientation
= 1 if facing south and 0 otherwise
Floor area ratio
Residential district-level variable
Green space
ratio
Residential district-level variable
Access to
schools
The number of schools within 500 m
Access to
hospitals
The number of hospitals within 500 m
Access to
shopping
centers
The number of shopping centers within 500 m
Distance to
Tianfu Square
Euclidean distance to Tianfu Square (residential
district-level variable)
m
Distance to
Financial City
Euclidean distance to Financial City (residential
district-level variable)
m
Distance to the
metro station
Distance to the nearest station (residential district-level
variable)
m
Employment
accessibility by
metro
Number of accessible employment opportunities within
45-min metro travel time (residential district-level variable)
COVID-19
= 1 if transacted after the explosion of COVID-19 (Feb.
2020) and 0 otherwise
15
Table 4
Summary of the variables
Variables
Mean
St. D.
Min
(0 quartile)
0.25 quantile
(1 quartile)
0.5 quantile
(2 quartile)
0.75 quantile
(3 quartile)
Max
(4 quartile)
Predicted variable
House price
140.35
76.00
7
94
124
165
1800
Predictor variables
Size
86.97
31.41
13
68
84
98
602
Age
9.18
6.13
1
5
8
12
56
Bedroom 1
12.7%
-
0
-
-
-
1
Bedroom 2
41.0%
-
0
-
-
-
1
Bedroom 3
39.6%
-
0
-
-
-
1
Bedroom 4+
6.7%
-
0
-
-
-
1
Living room 0&1
53.9%
-
0
-
-
-
1
Living room 2+
46.1%
-
0
-
-
-
1
Bathroom 0&1
70.7%
-
0
-
-
-
1
Bathroom 2+
30.3%
-
0
-
-
-
1
Low floor
25.7%
-
0
-
-
-
1
Intermediate floor
35.4%
-
0
-
-
-
1
High floor
38.9%
-
0
-
-
-
1
South orientation
26.9%
-
0
-
-
-
1
Floor area ratio
3.51
1.382
0.20
2.55
3.50
4.20
12.41
Green space ratio
0.33
0.10
0
0.30
0.30
0.38
0.87
Access to schools
1.01
1.26
0
0
1
2
9
Access to hospitals
0.09
0.72
0
0
0
0
24
Access to shopping centers
1.09
2.81
0
0
0
1
33
16
Distance to Tianfu Square
10,426
6201
245
5473
8750
15,916
27,960
Distance to Financial City
12,657
5989
256
8240
11,615
15,323
30,479
Distance to the metro
station
842
572
7
452
702
1035
2997
Employment accessibility
by metro
24,687,00
459,060
996,075
2,233,251
2,546,684
2,820,160
3,778,317
COVID-19
55.3%
-
0
-
-
-
1
17
5. Results
5.1. Revealing the association between to-metro and by-metro accessibility and house
prices
Multilevel double-log models are used in the following analysis because they
outperform the other two basic functional forms (linear and semi-log). The original
form of dummy variables is utilized because they cannot be converted to natural
logarithmic form. Moreover, because three access count variables (i.e., access to
schools, access to hospitals, and access to shopping centers) have a minimum value of
0, a transformation of adding one and taking the natural logarithm is used.
A set of multilevel hedonic price models are developed without considering the
moderating effects of the pandemic. A model without the two metro accessibility
variables (Model 1) is initially estimated. The two metro accessibility variables are
then added (Model 2). Notably, in the two models, we use twenty-three time dummies
(instead of the COVID-19 dummy) to capture the time effects on house prices.
Table 5 shows the hedonic modeling results. It indicates that all variance
inflation factors (VIFs) are < 10. This outcome suggests that collinearity is absent in
the data (Dormann et al., 2013). In other words, the variance of regression parameters
is not inflated. Therefore, we can conclude that this regression result is reliable and
can be used for interpretation. Moreover, nearly all hedonic variables are significant at
the 1% level, and the coefficient estimates have anticipated signs and reasonable
magnitudes. Only two variables–high floor and access to hospitals–fail to statistically
correlate with house prices, indicating that houses located on the high floor do not
value differently from those on the low floor, and houses with high access to hospitals
are valued similarly to those with low access.
18
Table 5
Hedonic modeling results without interaction terms
Variables
Model 1
Model 2
Coefficient
z-statistic
Coefficient
z-statistic
VIF
Size
0.819**
249.77
0.819**
249.75
3.84
Age
-0.277**
-37.52
-0.279**
-37.73
1.56
Bedroom2
0.102**
50.56
0.102**
50.59
4.04
Bedroom3
0.160**
61.39
0.160**
61.43
6.06
Bedroom4+
0.207**
56.37
0.207**
56.41
3.38
Living room2+
0.007**
7.47
0.007**
7.46
1.34
Bathroom2+
0.019**
14.01
0.019**
14.02
2.19
Intermediate floor
0.008**
8.15
0.008**
8.15
1.54
High floor
0.001
1.03
0.001
1.02
1.54
South orientation
0.006**
7.00
0.006**
7.00
1.03
Floor area ratio
0.049**
5.77
0.048**
5.64
1.20
Green space ratio
0.082**
10.38
0.082**
10.35
1.22
Access to schools
0.026**
3.28
0.029**
3.65
1.31
Access to hospitals
0.007
0.58
0.010
0.86
1.12
Access to shopping centers
0.016**
2.68
0.018**
2.99
1.21
Distance to Tianfu Square
-0.223**
-32.52
-0.211**
-29.46
3.41
Distance to Financial City
-0.174**
-19.45
-0.166**
-18.34
1.81
Distance to the metro station
-
-
-0.023**
-3.58
2.67
Employment accessibility by metro
-
-
0.039**
3.81
1.16
Time dummies
Yes
Yes
Constant
5.155**
57.10
4.573**
22.84
Random effects
19
Variance(Residential district)
Estimate
95% confidence interval
Estimate
95% confidence interval
0.055
[0.052, 0.058]
0.055
[0.052, 0.057]
Performance statistic
AIC
-113,149.7
-113,172.8
BIC
-112,754.6
-112,759.3
Note: * Significant at the 5% level. ** Significant at the 1% level. The estimates of 23 time dummies (2019M01 as the baseline) are omitted to conserve space.
20
Many interesting findings can be obtained. First, the magnitude of the price
elasticity of distance to Tianfu Square (0.211) is greater than that of distance to
Financial City (0.166). This result indicates that the to-Tianfu Square price gradient is
steeper than the to-Financial City gradient. To put it simply, Tianfu Square is still the
stronger city center and does not lose its appeal in the rapid urban expansion process.
Second, houses on the intermediate floor are more expensive than those on the low or
high floor. This finding is in concordance with the literature (Huang et al., 2022; Yang
et al., 2022b).
Model 2 has lower AIC and BIC than Model 1, indicating that the two newly
added metro accessibility variables help explain the variation in house prices. Houses
with high to-metro and by-metro accessibility are found to outshine others (in terms
of prices). On the one hand, the distance-based to-metro accessibility variable is
significant at the l% level and has a negative coefficient (-0.023), indicating that
to-metro accessibility positively affects house prices. This outcome agrees with a
large body of the existing literature (Ingvardson and Nielsen, 2018). For a 100%
increase (decrease) in the distance to the station, house prices depreciate (appreciate)
by 2.3%.
Employment accessibility by metro is significant at the l% level, and its
coefficient is larger than zero (0.039). This result means that houses with high
by-metro accessibility outshine others (in terms of prices). Specifically, for a 100%
increase (decrease) in the number of accessible employment opportunities within a
pre-specified metro ride time, house prices appreciate (depreciate) by 3.9%.
5.2. Verifying the moderating effect of COVID-19
We add interaction terms between the COVID-19 time dummy and metro
accessibility variables to Model 2 to verify the moderating effect of COVID-19. Table
6 reports the estimation outcomes of the new model (Model 3). It reveals that the
interaction term between to-metro accessibility and the COVID-19 time dummy is
significant at the 1% level, and its coefficient is positive. This outcome suggests that
COVID-19 flattens the to-metro price gradient, a result that can be attributed to the
diminishing attraction of the metro (Rosenthal et al., 2022; Yang et al., 2022b). To put
it simply, to-metro accessibility has become less valuable after the pandemic
explosion. More explanations for this interesting finding can be seen in the work of
Yang et al. (2022b) and Rosenthal et al. (2022). Moreover, the interaction term
between employment accessibility by metro and the COVID-19 time dummy is
insignificant at the 5% level. This result means that COVID-19 marginally affects the
value of employment accessibility by metro.
Table 6
Results for testing the moderating effect of COVID-19
Variables
Model 3
Coefficient
z-statistic
Size
0.819**
249.75
Age
-0.279**
-37.72
Bedroom2
0.102**
50.63
Bedroom3
0.160**
61.47
Bedroom4+
0.207**
56.45
Living room2+
0.007**
7.56
Bathroom2+
0.019**
14.05
21
Intermediate floor
0.008**
8.13
High floor
0.001
1.01
South orientation
0.006**
7.00
Floor area ratio
0.048**
5.65
Green space ratio
0.082**
10.34
Access to schools
0.029**
3.65
Access to hospitals
0.010
0.86
Access to shopping centers
0.018**
3.02
Distance to Tianfu Square
-0.211**
-29.40
Distance to Financial City
-0.166**
-18.27
Distance to the metro station
-0.027**
-4.11
Employment accessibility by metro
0.042**
3.89
Interaction terms
Distance to the metro station × COVID-19
0.006**
5.67
Employment accessibility by metro × COVID-19
-0.001
-0.11
Time dummies
Yes
Constant
4.538**
21.77
Random effects
Variance(Residential district)
Estimate
95% confidence interval
0.055**
[0.052, 0.057]
Performance statistic
AIC
-113,201.4
BIC
-112,769.7
Note: * Significant at the 5% level. ** Significant at the 1% level. The estimates of 23 time dummies
(2019M01 as the baseline) are omitted to conserve space.
5.3. Comparing the moderating effect of COVID-19 for small and big houses
We classify the sample into two groups according to size: small houses (size ≤
median) and big houses (size > median) to detect differences in the moderating role of
COVID-19 in different house submarkets. The sample sizes of small and big houses
are 37,118 (in 2686 residential districts) and 35,069 (in 2730 residential districts),
respectively. Multilevel hedonic price models are separately calibrated for the two
sub-samples. Model 4 is estimated for small houses, while Model 5 is for big ones. A
Watson and Westin pooling test, which compares the log-likelihoods of Model 3 and
the sum of the log-likelihoods of Models 4 and 5, is conducted to identify whether
variations in pricing behavior existed in the two house submarkets. The outcome
confirms the presence of such variations (Watson and Westin, 1975).
Table 7 shows the estimation results of Models 4 and 5. We find that the control
variables perform well in the two models in terms of significance and magnitude.
Their coefficient estimates mirror those presented in Table 6. In addition, to-metro and
by-metro accessibility have positive associations with house prices in the two
submarkets. Moreover, the interaction term between employment accessibility by
metro and the COVID-19 time dummy is insignificant at the 5% level, indicating that
COVID-19 marginally moderates the relationship between by-metro accessibility and
small/big house prices, which is consistent with Model 3. In other words, there is no
significant evidence of changes in the value of by-metro accessibility for both small-
and big-house submarkets.
22
Interestingly, the interaction term between to-metro accessibility and the
COVID-19 time dummy is significant at the 5% level in Models 4 and 5. Moreover,
the coefficient of the interaction term is much larger in the big-house submarket
(0.009) than in the small-house submarket (0.004). This observation suggests that the
purchasers of small houses are less disturbed by the pandemic in terms of metro travel,
or their demand for the metro is more inelastic. A possible explanation is that such
purchasers, who are more economically restrained, have lower car ownership and
fewer opportunities to use personalized transport and thus are more heavily reliant on
the metro.
Table 7
Results for modeling small and big houses separately
Variables
Model 4 (small houses)
Model 5 (big houses)
Coefficient
z-statistic
Coefficient
z-statistic
Size
0.838**
194.81
0.803**
137.03
Age
-0.252**
-31.03
-0.283**
-31.66
Bedroom2
0.087**
41.53
0.139**
7.10
Bedroom3
0.136**
44.62
0.201**
10.22
Bedroom4+
0.295**
22.26
0.251**
12.58
Living room2+
0.006**
4.76
0.006**
4.30
Bathroom2+
0.039**
10.95
0.015**
7.98
Intermediate floor
0.007**
5.57
0.009**
6.54
High floor
0.001
0.57
0.003
1.94
South orientation
0.003*
2.52
0.009**
6.52
Floor area ratio
0.054**
5.66
0.060**
5.49
Green space ratio
0.050**
5.87
0.125**
11.03
Access to schools
0.041**
4.66
0.029**
2.98
Access to hospitals
0.013
1.04
-0.002
-0.11
Access to shopping centers
0.032**
4.76
0.004
0.45
Distance to Tianfu Square
-0.195**
-23.20
-0.209**
-23.41
Distance to Financial City
-0.140**
-12.85
-0.186**
-17.28
Distance to the metro station
-0.030**
-4.08
-0.030**
-3.83
Employment accessibility by metro
0.053**
3.95
0.044**
2.88
Interaction terms
Distance to the metro station ×
COVID-19
0.004**
2.82
0.009**
5.14
Employment accessibility by metro ×
COVID-19
-0.002
-0.35
0.005
0.81
Constant
3.820**
14.91
4.802**
16.64
Random effects
Variance(Residential district)
Estimate
95%
confidence
interval
Estimate
95%
confidence
interval
0.049**
[0.046,
0.051]
0.060**
[0.057,
0.063]
Performance statistic
AIC
-67,217.9
-50,009.8
23
BIC
-66,817.4
-49,612.0
Note: * Significant at the 5% level. ** Significant at the 1% level. The estimates of 23 time dummies
(2019M01 as the baseline) are omitted to conserve space.
5.4. Additional evidence to support the changing preference of residents: transaction
volume
In addition to the shadow price, transaction volume can be used to infer residents’
preferences (Chmielewska et al., 2022). Inspired by the literature (Chmielewska et al.,
2022), we decide to compare the proportion of the transaction volume of
metro-adjacent houses (within 500 m of the station) to the total transaction volume (N
= 72,187) before and after the explosion of the pandemic. Fig. 4 shows the proportion
of metro-adjacent house transactions in each month of 2019 and 2020. It reveals that
this proportion plummeted from January to February 2022 (33.2% vs. 26.3%).
Moreover, the average proportion before and after the pandemic explosion is 30.90%
and 29.82%, respectively. A stark difference can be observed. Moreover, we conduct
two-sample t-tests to verify whether the difference in the means of the two groups is
statistically significant. The results suggest inequality in the means of the two groups
is significant at the 5% level. To recap, the above outcomes reveal that after the
explosion of the pandemic, purchasers tend not to choose houses adjacent to metro
stations, leading to the decreasing value of to-metro accessibility.
Fig. 4. Proportion of metro-adjacent house transactions in each month of 2019 and
2020.
5.5. Comparing the moderating effect of COVID-19 for houses with different values
We develop a set of quantile regression models to capture the heterogeneous
effect of metro accessibility on house prices across price quantiles. The results show
that to-metro and by-metro accessibility have a positive price effect over the entire
conditional distribution.
Table 8 displays the estimates of the two interaction terms. The coefficient
estimate of the interaction term between to-metro accessibility and the COVID-19
time dummy is very diffused. Specifically, it is statistically insignificant at the 5%
24
level for low-priced houses (0.05, 0.1, and 0.25 quantiles) but significant for
high-priced houses (0.5, 0.75, 0.9, and 0.95 quantiles). Differences in these coefficient
estimates are tested to be significant at the 1% level. These observations clearly
illustrate that COVID-19 reduces the value of to-metro accessibility for high-priced
houses but not for low-priced ones.
The ranking of the pandemic effect on the value of to-metro accessibility is:
highest-priced houses (0.9 and 0.95 quantiles) > high-priced ones (0.5 and 0.75
quantiles) > low-priced ones (0.05, 0.1, and 0.25 quantiles). COVID-19
insignificantly affects the value of to-metro accessibility for low-priced houses but
substantially decreases this value for high-priced houses. A possible explanation is
that high-priced house purchasers have more economic freedom and can choose their
travel options at will. Note that this finding regarding the heterogeneity concurs with
the result of Models 4 and 5. The only difference is that one focuses on the total house
price, while the other emphasizes size. The two variables are indicative of economic
strength. The above outcomes jointly demonstrate that economically advantaged
people have high (travel) choice freedom, and in the event of a shock such as
COVID-19, they can make changes swiftly. In comparison, economically
disadvantaged people have few opportunities to use personalized transport and are
still heavily dependent on the metro. All in all, we conclude that COVID-19 decreases
the value of to-metro accessibility, especially for big or high-priced houses.
25
Table 8
Quantile regression results
Variables
0.05 quantile
0.1 quantile
0.25 quantile
Median
0.75 quantile
0.9 quantile
0.95 quantile
Coeffici
ent
t-statistic
Coeffici
ent
t-statistic
Coeffici
ent
t-statistic
Coeffici
ent
t-statistic
Coeffici
ent
t-statistic
Coeffici
ent
t-statistic
Coeffici
ent
t-statistic
Distance to the metro
station × COVID-19
-0.008
-1.79
-0.006
-1.63
0.003
1.10
0.009**
3.07
0.010**
3.04
0.011**
2.49
0.015**
2.68
Employment
accessibility by metro
× COVID-19
-0.033*
-2.07
-0.031*
-2.56
-0.021*
-2.23
0.009
1.00
0.033**
2.97
0.041**
2.76
0.048**
2.60
Performance statistic
Pseudo R2
0.540
0.540
0.537
0.527
0.530
0.544
0.553
Note: * Significant at the 5% level. ** Significant at the 1% level. The estimates of other variables are omitted to conserve space.
26
6. Conclusions
This study probes the association between to-metro and by-metro accessibility
and house prices in Chengdu, China, and the moderating effect of COVID-19 on this
association using a variety of hedonic price and quantile regression models. Notably,
it considers to-metro and by-metro accessibility together in assessing metro
accessibility.
This study obtains nuanced findings: (1) Not only to-metro accessibility but also
by-metro accessibility determine house prices; (2) COVID-19 affects the pricing of
hedonic variables such as metro accessibility; (3) COVID-19 significantly decreases
the value of to-metro accessibility (flattening the to-metro price gradient) but weakly
affects the price of by-metro accessibility; and (4) The value of to-metro accessibility
is disproportionately affected. Specifically, the flattening of the to-metro price
gradient is more observable for big or high-priced houses than for small or low-priced
houses. A compelling reason for these findings is the inelastic demand for metro use
by economically restrained people (represented by the purchasers of small or
low-priced houses). In light of the above findings, this study supports the use of
time-varying and category-specific value capture tools to recoup the benefits provided
by large-scale transport infrastructure (e.g., the metro) and to strengthen the
“beneficiary-to-pay” rationale.
In a departure from most existing real estate valuation studies focusing solely on
to-metro accessibility, this study introduces by-transit accessibility into the utility
functions of house purchasers and advocates its use in transport studies, especially
those pertaining to human behavior and psychology. Moreover, previous studies have
rarely examined the relationship between transit accessibility and house markets
during disruptive events, such as disasters and epidemics (e.g., SARS, Ebola, and
COVID-19). By contrast, this study examines the disruptive role of the unprecedented
COVID-19 pandemic and suggests that COVID-19 is a moderator in the connection
between metro accessibility and house prices.
This study has some limitations. First, more diverse by-metro accessibility
measures, such as gravity-, utility-, and floating catchment area-based approaches, can
be used in further studies. Second, using spatial econometric methods to address
spatial autocorrelation or heterogeneity is encouraged in further studies to enrich our
understanding of the interaction between metro accessibility, house prices, and the
pandemic. Third, the pandemic continues to disrupt the lives of people everywhere.
This study cannot indicate whether the COVID-19-induced changes in the pricing of
hedonic variables are short-lived. In other words, we are uncertain of whether the
connection between hedonic variables and house prices will bounce back to the
pre-pandemic state. The lack of data covering a long period prevents us from
conducting related research. We, however, believe that more comprehensive studies
spanning many years should be conducted in the future to explore this issue.
Acknowledgments
This study was supported by the Sichuan Science and Technology Program (No.
2022JDR0178), Sichuan Youth Science and Technology Innovation Team Funding
(No. 2022JDTD0005), and the National Natural Science Foundation of China (No.
U20A20330). The authors are grateful to the reviewers for their constructive
comments.
References
Armstrong, R.J., Rodriguez, D.A., 2006. An evaluation of the accessibility benefits of
27
commuter rail in eastern Massachusetts using spatial hedonic price functions. Transportation
33(1), 21-43.
Bae, C.-H.C., Jun, M.-J., Park, H., 2003. The impact of Seoul's subway Line 5 on residential
property values. Transport Policy 10(2), 85-94.
Bajic, V., 1983. The effects of a new subway line on housing prices in metropolitan Toronto.
Urban Studies 20(2), 147-158.
Bao, Z., Ou, Y., Chen, S., Wang, T., 2022. Land use impacts on traffic congestion patterns: a
tale of a Northwestern Chinese City. Land 11, 2295.
Bartholomew, K., Ewing, R., 2011. Hedonic price effects of pedestrian-and transit-oriented
development. Journal of Planning Literature 26(1), 18-34.
Bowes, D.R., Ihlanfeldt, K.R., 2001. Identifying the impacts of rail transit stations on
residential property values. Journal of Urban Economics 50(1), 1-25.
Boyce, D.E., Allen, B., Mudge, R.R., Slater, P.B., Isserman, A.M., 1972. Impact of rapid
transit on suburban residential property values and land development.
Cao, X.J., Porter-Nelson, D., 2016. Real estate development in anticipation of the Green Line
light rail transit in St. Paul. Transport Policy 51, 24-32.
Cervero, R., Duncan, M., 2002. Transit’s value-added effects: Light and commuter rail
services and commercial land values. Transportation Research Record: Journal of the
Transportation Research Board 1805(1), 8-15.
Chen, S., Wang, T., Bao, Z., Lou, V., 2022. A path analysis of the effect of neighborhood built
environment on public health of older adults: A Hong Kong study. Frontiers in Public Health
10, 861836.
Cheng, L., Wang, K., De Vos, J., Huang, J., Witlox, F., 2022. Exploring non-linear built
environment effects on the integration of free-floating bike-share and urban rail transport: A
quantile regression approach. Transportation Research Part A: Policy and Practice 162,
175-187.
Cheung, K.S., Yiu, C.Y., Xiong, C., 2021. Housing market in the time of pandemic: A price
gradient analysis from the COVID-19 epicentre in China. Journal of Risk and Financial
Management 14(3), 108.
Chmielewska, A., Ciski, M., Renigier-Biłozor, M., 2022. Residential real estate investors'
motives under pandemic conditions. Cities 128, 103801.
Chu, J., Duan, Y., Yang, X., Wang, L., 2021. The last mile matters: Impact of dockless bike
sharing on subway housing price premium. Management Science 67(1), 297-316.
D'Lima, W., Lopez, L.A., Pradhan, A., 2022. COVID‐19 and Housing market effects:
Evidence from us shutdown orders. Real Estate Economics 50(2), 303-339.
Dai, X., Bai, X., Xu, M., 2016. The influence of Beijing rail transfer stations on surrounding
housing prices. Habitat International 55, 79-88.
Damm, D., Lerman, S.R., Lerner-Lam, E., Young, J., 1980. Response of urban real estate
values in anticipation of the Washington Metro. Journal of Transport Economics and Policy
14(3), 315-336.
Davis, F.W., 1970. Proximity to a rapid transit station as a factor in residential property values.
Appraisal Journal 38, 554–572.
Debrezion, G., Pels, E., Rietveld, P., 2007. The impact of railway stations on residential and
commercial property value: a meta-analysis. The Journal of Real Estate Finance and
28
Economics 35(2), 161-180.
Debrezion, G., Pels, E., Rietveld, P., 2011. The impact of rail transport on real estate prices:
An empirical analysis of the Dutch housing market. Urban Studies 48(5), 997-1015.
Diao, M., Leonard, D., Sing, T.F., 2017. Spatial-difference-in-differences models for impact
of new mass rapid transit line on private housing values. Regional Science and Urban
Economics 67, 64-77.
Dormann, C.F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., Marquéz, J.R.G.,
Gruber, B., Lafourcade, B., Leitão, P.J., 2013. Collinearity: A review of methods to deal with
it and a simulation study evaluating their performance. Ecography 36(1), 27-46.
Giuliano, G., Gordon, P., Pan, Q., Park, J., 2010. Accessibility and residential land values:
Some tests with new measures. Urban Studies 47(14), 3103-3130.
Gupta, A., Mittal, V., Peeters, J., Van Nieuwerburgh, S., 2022. Flattening the curve:
Pandemic-induced revaluation of urban real estate. Journal of Financial Economics 146(2),
594-636.
He, S.Y., 2020. Regional impact of rail network accessibility on residential property price:
Modelling spatial heterogeneous capitalisation effects in Hong Kong. Transportation
Research Part A: Policy and Practice 135, 244-263.
Huang, N., Pang, J., Yang, Y., 2022. JUE insight: COVID-19 and household preference for
urban density in China. Journal of Urban Economics, 103487.
Ingvardson, J.B., Nielsen, O.A., 2018. Effects of new bus and rail rapid transit systems–an
international review. Transport Reviews 38(1), 96-116.
Ko, K., Cao, X.J., 2013. The impact of Hiawatha Light Rail on commercial and industrial
property values in Minneapolis. Journal of Public Transportation 16(1), 47-66.
Lai, Z., Liu, X., Li, W., Li, Y., Zou, G., Tu, M., 2021. Exploring the spatial heterogeneity of
residents' marginal willingness to pay for clean air in Shanghai. Frontiers in Public Health 9,
791575.
Lancaster, K.J., 1966. A new approach to consumer theory. Journal of Political Economy
74(2), 132-157.
Lessa, D.A., Lobo, C., Cardoso, L., 2019. Accessibility and urban mobility by bus in Belo
Horizonte/Minas Gerais–Brazil. Journal of Transport Geography 77, 1-10.
Li, L., Wan, W.X., He, S., 2021. The heightened ‘security zone’ function of gated
communities during the COVID-19 pandemic and the changing housing market dynamic:
Evidence from Beijing, China. Land 10(9), 983.
Li, S., Chen, L., Zhao, P., 2019. The impact of metro services on housing prices: a case study
from Beijing. Transportation 46(4), 1291-1317.
Lin, D., Broere, W., Cui, J., 2022a. Metro systems and urban development: Impacts and
implications. Tunnelling and Underground Space Technology 125, 104509.
Lin, D., De Vos, J., Maruyama, T., Bobylev, N., Cui, J., 2022b. Metro-related transfers: A
review of recent literature. Journal of Urban Planning and Development 148(3), 03122002.
Lin, D., Nelson, J.D., Beecroft, M., Cui, J., 2021. An overview of recent developments in
China’s metro systems. Tunnelling and Underground Space Technology 111, 103783.
Linton, C., Grant-Muller, S., Gale, W.F., 2015. Approaches and techniques for modelling CO2
emissions from road transport. Transport Reviews 35(4), 533-553.
Liu, Q., Zhao, P., Xiao, Y., Zhou, X., Yang, J., 2022. Walking accessibility to the bus stop:
29
Does it affect residential rents? The case of Jinan, China. Land 11(6), 860.
Liu, S., Su, Y., 2021. The impact of the COVID-19 pandemic on the demand for density:
Evidence from the US housing market. Economics Letters 207, 110010.
Liu, Y., Tang, Y., 2021. Epidemic shocks and housing price responses: Evidence from China's
urban residential communities. Regional Science and Urban Economics 89, 103695.
Mao, Z., Wang, F., Wang, D., 2022. Attitude and accessibility on transit users’ travel
satisfaction: A person-environment fit perspective. Transportation Research Part D:
Transport and Environment 112, 103467.
Moniruzzaman, M., Páez, A., 2012. Accessibility to transit, by transit, and mode share:
Application of a logistic model with spatial filters. Journal of Transport Geography 24,
198-205.
Mulley, C., Tsai, C.-H.P., Ma, L., 2018. Does residential property price benefit from light rail
in Sydney? Research in Transportation Economics 67, 3-10.
Pan, H., Li, J., Shen, Q., Shi, C., 2017. What determines rail transit passenger volume?
Implications for transit oriented development planning. Transportation Research Part D:
Transport and Environment 57, 52-63.
Pan, H., Zhang, M., 2008. Rail transit impacts on land use: Evidence from Shanghai, China.
Transportation Research Record: Journal of the Transportation Research Board 2048(1),
16-25.
Pan, Q., Pan, H., Zhang, M., Zhong, B., 2014. Effects of rail transit on residential property
values: Comparison study on the rail transit lines in Houston, Texas, and Shanghai, China.
Transportation Research Record: Journal of the Transportation Research Board 2453(1),
118-127.
Pawson, H., Martin, C., Aminpour, F., Gibb, K., Foye, C., 2022. COVID-19: Housing market
impacts and housing policy responses-an international review.
Qian, X., Qiu, S., Zhang, G., 2021. The impact of COVID-19 on housing price: Evidence
from China. Finance Research Letters 43, 101944.
Rosen, S., 1974. Hedonic prices and implicit markets: product differentiation in pure
competition. Journal of Political Economy 82(1), 34-55.
Rosenthal, S.S., Strange, W.C., Urrego, J.A., 2022. JUE insight: Are city centers losing their
appeal? Commercial real estate, urban spatial structure, and COVID-19. Journal of Urban
Economics 127, 103381.
So, H.M., Tse, R.Y.C., Ganesan, S., 1997. Estimating the influence of transport on house
prices: evidence from Hong Kong. Journal of Property Valuation and Investment 15(1),
40-47.
Wang, F., Zheng, Y., Wu, W., Wang, D., 2022. The travel, equity and wellbeing impacts of
transit-oriented development in Global South. Transportation Research Part D: Transport
and Environment 113, 103512.
Watson, P.L., Westin, R.B., 1975. Transferability of disaggregate mode choice models.
Regional Science and Urban Economics 5(2), 227-249.
Wu, W., Dong, G., 2014. Valuing the “green” amenities in a spatial context. Journal of
Regional Science 54(4), 569-585.
Wu, W., Dong, G., Wang, B., 2015. Does planning matter? Effects on land markets. The
Journal of Real Estate Finance and Economics 50(2), 242-269.
30
Wu, W., Zheng, S., Wang, B., Du, M., 2020. Impacts of rail transit access on land and housing
values in China: a quantitative synthesis. Transport Reviews 40(5), 629-645.
Xu, T., Zhang, M., Aditjandra, P.T., 2016. The impact of urban rail transit on commercial
property value: New evidence from Wuhan, China. Transportation Research Part A: Policy
and Practice 91, 223-235.
Yang, H., Huo, J., Bao, Y., Li, X., Yang, L., Cherry, C.R., 2021. Impact of e-scooter sharing
on bike sharing in Chicago. Transportation Research Part A: Policy and Practice 154, 23-36.
Yang, H., Huo, J., Pan, R., Xie, K., Zhang, W., Luo, X., 2022a. Exploring built environment
factors that influence the market share of ridesourcing service. Applied Geography 142,
102699.
Yang, J., Bao, Y., Zhang, Y., Li, X., Ge, Q., 2018. Impact of accessibility on housing prices in
Dalian city of China based on a geographically weighted regression model. Chinese
Geographical Science 28(3), 505-515.
Yang, J., Chen, J., Le, X., Zhang, Q., 2016. Density-oriented versus development-oriented
transit investment: Decoding metro station location selection in Shenzhen. Transport Policy
51, 93-102.
Yang, L., Chau, K.W., Szeto, W.Y., Cui, X., Wang, X., 2020a. Accessibility to transit, by
transit, and property prices: Spatially varying relationships. Transportation Research Part D:
Transport and Environment 85, 102387.
Yang, L., Chen, Y., Xu, N., Zhao, R., Chau, K.W., Hong, S., 2020b. Place-varying impacts of
urban rail transit on property prices in Shenzhen, China: Insights for value capture.
Sustainable Cities and Society 58, 102140.
Yang, L., Liang, Y., He, B., Lu, Y., Gou, Z., 2022b. COVID-19 effects on property markets:
The pandemic decreases the implicit price of metro accessibility. Tunnelling and
Underground Space Technology 125, 104528.
Yin, J., Cao, X.J., Huang, X., 2021. Association between subway and life satisfaction:
Evidence from Xi’an, China. Transportation Research Part D: Transport and Environment 96,
102869.
Zou, G., Lai, Z., Li, Y., Liu, X., Li, W., 2022. Exploring the spatial heterogeneity of residents'
marginal willingness to pay for clean air in Shanghai. Economics of Transportation 31,
100272.
