Transport Policy ()

Contents lists available at ScienceDirect

Transport Policy
journal homepage: www.elsevier.com/locate/tranpol

Density-oriented versus development-oriented transit investment:

Decoding metro station location selection in Shenzhen
Jiawen Yang a,n, Junxian Chen b, Xiaohui Le b, Qin Zhang c
a
School of Urban Planning and Design, Shenzhen Graduate School, Peking University, Building E, Room 326, Shenzhen 518055, China
b
School of Urban Planning and Design, Shenzhen Graduate School, Peking University, Building E, Room 319, Shenzhen 518055, China
c
School of Computer Science and Engineering, South China University of Technology, Room 338, Building B3, South Campus (HEMC), Panyu District,
Guangzhou 510006, China

ar t ic l e i nf o a b s t r a c t

Article history: The transit-oriented development literature has focused on the built environment around stations in
Received 1 July 2015 operation, largely neglecting how the station location was selected. We hypothesize that city govern-
Received in revised form ments in China are likely to put stations outside established suburban centers. By putting metro stations
5 April 2016
at relatively underdeveloped places, city governments can lower right-of-way cost and gain more rev-
Accepted 8 April 2016
enue from future land transactions. Using Shenzhen as a case study, we test this hypothesis with metro
planning examples and land transaction data from 2000 to 2014. We found that metro alignment and
Keywords: station placement has bypassed the core of established communities. This planning practice is supported
Metro investment by a strong real estate market that appreciates transit accessibility, despite the high transit operation
Station location
subsidy.
Transit-oriented development
& 2016 Elsevier Ltd. All rights reserved.
Shenzhen

1. Introduction
The enormous cost involved in the establishment of a metro
system (above and below ground rail guided urban transit net-
work) justies detailed planning, which includes route alignment
and location of metro stations (Samanta and Jha, 2008). Many
researchers have pointed to the relatively high density land use or
high-priced properties that surround metro stations as positive
impact of metro operation on land development (Dorantes et al.,
2011; Hess and Almeida, 2007). Most literature on transit invest-
ment and transit-oriented development (TOD), however, has fo-
cused on describing the built environment surrounding stations,
but not the potential siting that planners have considered. The
existence of high-density development around metro stations
could result from two different paths: either the station is built to
serve pre-existing high-density areas or the station has attracted
high-density (re)development. A better conception of these alter-
native development paths can help us to understand the char-
acteristics of planning, design and operation associated with me-
tro systems in different parts of the world, including those in
China.
When stations are placed to serve pre-existing high density
areas, the advantage is relatively high ridership, which helps to
n
Corresponding author.
E-mail address: yangjw@pkusz.edu.cn (J. Yang).
boost fare box revenue and reduce operational subsidy. One par-
ticular example is the American federal government's New Start
program. In evaluating grant proposals from various locations, the
Federal Transit Administration uses cost-effectiveness measures,
such as the expenditure to produce one-hour travel time saving, as
major project ranking criteria (FTA, 2012). Since xed guide-way
projects placed in relatively high density corridors can produce
relatively robust ridership and thus more travel time savings,
density-oriented transit planning is typically prioritized to receive
federal grants. In contrast, places like Hong Kong have emphasized
a development-oriented approach. New metro stations are placed
at places where new town development is expected to happen
(Lee and Leung, 1994). One particular concern underlying Hong
Kong's approach is to support metro investment and operation
with revenue from the real estate sector.
From the perspective of municipal nance, the choice between
the density-oriented approach and the development-oriented
approach is a tradeoff between land-sector revenue and fare-box
revenue, which is strongly affected by the scal environment for
metro investment. The American federal government which does
not receive any direct revenue from land development around
stations, has elected to prefer fare-box revenue to land-sector
revenue. In contrast, the Hong Kong government, which owns the
land and collects revenue from land development activities, has
emphasized land development revenue. These two approaches
can, of course, co-exist. For example, the US Federal Transit Ad-
ministration's New Start program also examines land development

http://dx.doi.org/10.1016/j.tranpol.2016.04.004
0967-070X/& 2016 Elsevier Ltd. All rights reserved.

Please cite this article as: Yang, J., et al., Density-oriented versus development-oriented transit investment: Decoding metro station
location selection in Shenzhen. Transport Policy (2016), http://dx.doi.org/10.1016/j.tranpol.2016.04.004i
2 J. Yang et al. / Transport Policy ()
potential around planned stations when ranking projects, albeit
with a weight lower than the factor of ridership and time-saving.
The development-oriented approach in Hong Kong also targets
future high ridership, expecting that development around stations
should catch up quickly, so that the subsidy required for metro
operation will soon be reduced.
Observing how China's city governments fund metro projects,
we hypothesize that the scal environment shapes the choice of
Chinese local governments and the development-oriented ap-
proach is more likely to be emphasized.. First, unlike the US metro
systems which are heavily subsidized by the federal government,
metro construction and operation in China in general receives no
funding from the central or provincial governments. City govern-
ments fund metro projects with general revenue sources or with
bank loan guarantees for metro corporations. In addition, in con-
trast to USs private land ownership, China's city governments re-
present the national state in each city and claims signicant
benet stemming from urban land development. A typical practice
for city governments to raise revenue has been to transfer multi-
decade land use rights to the private sector (Yang, 2006; Yang
et al., 2007). Placing transit stations in undeveloped parcels is
likely to raise signicant revenue for the city government as
property developers tend to bid higher prices for parcels closer to
stations.
Furthermore, placing stations in a pre-existing high-density
area not only implies high right-of-way-costs, but also a loss of
opportunity to boost land-sector revenue (Yang, 2006). In devel-
oped areas, property value appreciation stemming from metro
operations benets the private sector more than the city govern-
ment because China does not have an American-style property tax.
Property value appreciation does not lead to an increase in prop-
erty tax revenue. The lack of an annual property tax, therefore,
discourages China's city governments from placing stations in pre-
existing high-density areas.
This hypothesis is tested with a case study of metro planning,
particularly metro station placement, in Shenzhen, the fth city in
mainland China to operate a metro system. By examining planned
metro lines and those in operation, we will review how concern
Fig. 1. Three Phases of Shenzhen Metro System.
Please cite this article as: Yang, J., et al., Density-oriented versus development-oriented transit investment: Decoding metro station
location selection in Shenzhen. Transport Policy (2016), http://dx.doi.org/10.1016/j.tranpol.2016.04.004i
about land-sector revenue has affected project ranking, route de-
sign and station placement. By examining the price of transacted
land parcels in relation to their distance to the nearest metro
stations, we aim to nd out whether station placement in general
has bypassed community centers.
2. Funding metro in Shenzhen
Shenzhen is located on a southern tip of China's Guangdong
Province and on the east bank of the Pearl River, neighboring Hong
Kong. Occupying 1991 square kilometers, the city had a total re-
gistered population of 10.8 million in 2014. As China's rst and one
of the most successful special economic zones, Shenzhen has
grown from a tiny border town into a modern metropolis with a
gross foreign export value that ranks rst among all Chinese cities.
In 2014, Shenzhen's GDP had reached 1600 billion RMB (US$
258.06 billion), fourth among all cities in China, making it one of
the most developed cities in China and one of the fastest-growing
cities in the world (SZGOV, 2014).
Similar to other megacities, trafc congestion raises the need
for high capacity rapid transit systems. Following Beijing, Tianjin,
Shanghai and Guangzhou, Shenzhen is the fth mainland China's
city to operate a metro system. Shenzhen's metro system was
developed in three phrases (Fig. 1), with different funding ar-
rangements for each phase (Table 1). In the rst phase, the Luobao
Line and the rst part of the Longhua Line were planned in 1998
and began operation in 2004, with a length of 22 km and an in-
vestment of 11.5 billion RMB (US$ 1.85 billion). The city govern-
ment funded 70% of the construction cost in cash and the re-
maining 30% through bank loans.
In the second phase, the network added three new lines (She-
kou Line, Longgang Line and Huanzhong Line) and extended the
existing Luobao and Longhua Lines. These lines were completed in
2011, with a total length of 142 km and a total investment of 68.8
billion RMB (US$ 11.10 billion). The proportion of the construction
cost covered by Shenzhen government's general revenue shrunk
from 70% to 50%. The remaining 50% was funded by Shenzhen
 J. Yang et al. / Transport Policy () 3

Table 1
Phases of metro system development and nancing.

Phase Total length (kilometers) Investment (billion RMB) Time period Government funding Source of the rest fund

Ratio Form

Phase 22 11.5 Operation began in 2004 70% Government cash Bank loan
Phase II 142 68.8 Operation began in 2011 50% Government cash Bank loan and other
modes
Phase III 169.6 81.2 Scheduled to operate in 2016 50% Property land transfer income Self-raised

Table 2
Project ranking criteria and ranking results.

Criteria Weight Line 4 Line 6 Line 7 Line 8 Line 9 Line 10 Line 11 Line 12

Urban trafc pressure (ridership) 0.2 8 6 2 3 1 5 4 7

Network strategy (connectivity) 0.2 8 4 2 5 3 6 1 7
Urgency 0.2 8 5 1 6 2 3 7 4
Coordinating with urban development 0.2 8 2 7 3 4 5 1 6
Operational revenue 0.1 6 3 7 5 8 4 1 1
Land-sector revenue 0.1 2 5 6 7 8 4 3 1
Weighted sum 1.0 7.2 4.2 3.7 4.6 3.6 4.6 3 5
Overall ranking 8 4 3 5 2 6 1 7

Source: SZGOV, Short-term metro system plan of Shenzhen (20112016), 2011. p. 150.

Metro Corporation (SZMC), a state-owned enterprise. SZMC tried
to meet the huge funding demand by applying diversied modes
of nancing (SZMC, 2015a), including collaborating with Hong
Kong Mass Transit Corporation (HKMTR) to construct the Longhua
Line extension, following a build-operate-transfer (BOT) protocol
(Yang et al., 2014). The whole system carried 2.1 million passen-
gers per day in 2013 (SZMC, 2013).
In the midst of Phase II construction, the Shenzhen city gov-
ernment and SZMC launched a rail property model of land
development, which largely follows Hong Kong's practice, but is
tailored to Chinese regulations. In 2006, China's central govern-
ment instituted an auction requirement for urban land transac-
tions and restrained land transfer by negotiation. Shenzhen's city
government circumvents this regulation by transferring land
through a restrained auction, which assures SZMC to obtain land
parcels through the auction process. Shenzhen city government
then turns over the land lease revenue to its State-owned Assets
Supervision and Administration Commission (SASAC), a govern-
ment division that manages government investment for state-
owned enterprises. After receiving this land lease payment, SASAC
transfers it back to SZMC in the form of registered capital (Li,
2011). Fig. 2 depicts this whole process, with the land transaction
at Qianhai station as an example. In 2008, SZMC obtained the
usage rights to this parcel (0.55 million square meters) with a
payment of 1.74 billion RMB (US$ 0.28 billion).
SZMC's Division of Property Development was established in
March 2007 to undertake real estate development by working
with banks and experienced developers. Commercial housing
projects have been pushed forward along Luobao Line, Shekou Line,
Huanzhong Line and Longgang Line. For instance, the Shanhaijin
project, which uses the air rights of the west train depot at Shekou
station, is a commercial housing project with one- and two-bed-
room condos. The condos were sold in the fourth quarter of 2013
Fig. 2. Rail Property Model, Qianhai Example.
and SZMC received sales revenue of 0.4 billion RMB (US$ 64.5
million) (SZMC, 2013).
The third phase includes three lines under construction (Airport
Line, Xili Line and Meilin Line), which are scheduled to begin op-
eration in 2016, and another two lines in the planning stage
(Guangming Line and Yantian Line). These ve lines have a total
length of 169.6 km, with a total investment of 81.2 billion RMB (US
$ 13.10 billion) (SZMC, 2015b). Already burdened by the opera-
tional subsidy for the system currently in use, the Shenzhen city
government is committed to the rail property model to nance
these lines The municipal government plans no cash appropriation
for Phase III projects. Instead, it has prepared 5.66 million square
meters of land for SZMC to recoup its construction cost and op-
eration decits (Zhao, 2013). By 2014, the total amount of land
transferred to SZMC for this purpose had reached 3.74 million
square meters (SZMC, 2015a). The Qianhai case reported here is
representative of many more such developments that are expected
in years to come.
3. Metro planning geared toward land development revenue
The above brief history of funding for Shenzhen's metro reveals
a strengthened link between metro investment and land-sector
revenue. Under this scal environment, metro planning and de-
sign has shifted away from a pure transport concern.
3.1. Project ranking drifts away from ridership concern
According to the Shenzhen metro system construction plan in
2007, metro lines are ranked by ve factors: urban trafc pressure
(ridership), network strategy (connectivity), urgency, coordination
with urban development, operational revenue and land-sector
revenue. Table 2 lists the weight assigned to each factor and the
ranking results for eight metro projects (SZGOV, 2011a). A higher
number indicates a lower priority.
With this exercise, a 10% weight is assigned to land-sector
revenue and a much higher weight is given to trafc and ridership.
The Line 4 (Longhua Line extension) is the least preferred project
according to this criterion. In reality, however, it was the rst to be
built among the eight. One possible reason for this decision is its

Please cite this article as: Yang, J., et al., Density-oriented versus development-oriented transit investment: Decoding metro station
location selection in Shenzhen. Transport Policy (2016), http://dx.doi.org/10.1016/j.tranpol.2016.04.004i
4 J. Yang et al. / Transport Policy ()
Fig. 3. Longhua Line follows under-developed corridor.
high land development potential: it was ranked second by land
sector revenue. This discrepancy in planning and project im-
plementation indicates that project implementation has employed
a more pragmatic approach and emphasized land sector revenue,
which suggests a drift away from the ideal planning exercise that
emphasizes ridership and trafc mitigation.
3.2. Metro alignment bypasses established communities
A detailed examination of the metro alignment or station pla-
cement can shed more insight on these issues. Again, we will use
line 4 (Longhua Line) as an example Line 4t connects the city
center and the suburban Longhua district. It starts from the Futian
Checkpoint, at Hong Kong's border, and extends northward. When
the Longhua Line goes northward into the suburbs, it deliberately
turns west and bypasses Minzhi, a high-density suburban settle-
ment (Fig. 3).
What complicates the story a bit is the placement of the
Shenzhen North Station, which is not only a transfer station be-
tween Longhua Line and Huanzhong Line, but also a high-speed
train station. It is located in Shenzhen's suburban Longhua District,
9.3 km from the city center. Its construction began in 2008 and
was completed in 2011. Unsurprisingly, such a complex needs a
huge amount of funding and a big piece of land. China's Ministry of
Railway (MOR), which shoulders these capital expenditures, pre-
ferred to locate the station at the urban periphery to minimize the
cost of the right-of-way. Shenzhen Municipal Government shared
the same idea, hoping the transit investment would boost the land
development around the station, thus generating more land-lease
revenue. As a result, this huge station was also put in an under-
developed area.
3.3. Station placement favors land-sector revenue at expense of
trafc mitigation
A more recent example comes from the planning for Guangmin
Line, line 6, which is a Phase III project. It mainly serves suburban
communities, starting from Shenzhen North Station, where
Please cite this article as: Yang, J., et al., Density-oriented versus development-oriented transit investment: Decoding metro station
location selection in Shenzhen. Transport Policy (2016), http://dx.doi.org/10.1016/j.tranpol.2016.04.004i
passengers can transfer to the Longhua and Huanzhong lines. De-
cisions on metro station location have sparked many debates. The
initial plan was made in 2008, and then revised in 2012. Fig. 4
shows the metro alignment of the two plans. The locations of
several stations have been adjusted. But both plans bypass den-
sely-populated areas and run along the edge of communities.
One debate has centered on the placement of the Dalang
community station. In the 2008 plan, the municipal government
located a Shilong metro station within Dalang community (gray dot
in Fig. 4). It was put there because the city governments have
several empty parcels there, which implies both low right-of-way
costs before station construction and high land lease potential
following metro completion. However, residents in Dalang Com-
munity strongly opposed this plan, claiming that this alignment
bypassed Dalang community's center where heavy trafc was a
daily problem. Instead, they proposed a Dalang station as a re-
placement, which is located in the center of Dalang community
(the white dot to the east of gray dot in Fig. 4). Regardless of the
strong opinion from the local residents, the revised plan of 2012
did not change the station location. Ironically, it renamed Shilong
station to Dalang station. The evidence presented here suggests
that the city government's preference for land lease revenue and low
right-of-way cost leads to favorable consideration of under-developed
areas, at the expense of community desire.
While other factors such as geological conditions may also
matter in station placement, it is unlikely the major cause here.
The alignment alternatives and the station alternatives are only a
few kilometers apart. Their geological conditions are quite similar.
In addition, no government report or media release points to
geological or morphological factors when explaining the align-
ment and station placement.
Note that this scal oriented approach for planning not only
exists in metro planning, but also in urban road planning. For
example, the densely spaced ring roads in China's megacities have
their roots in land lease revenue. Comparing two different align-
ments of ring roads outside a pre-existing built-up area, one dis-
tanced from the urban center by 10 km and the other by 20 km,
the former can surely help the city government to prot more on
 J. Yang et al. / Transport Policy () 5

Fig. 4. Guangming Line's Alignment and Dalang Station.

the basis of per-km road construction. The primary reason: the
land closer to the city center is much more expensive while the
additional right-of-way cost per linear kilometer is relatively
minor. As a result, ring roads in urban China are quite dense (Yang,
et al., 2007).
Also note that the plan to put metro stations at relatively un-
derdeveloped areas in Shenzhen does not suggest that there is no
interest in TOD by the city government. Indeed, there is an ex-
pectation that TOD should happen. For example, the city govern-
ment's manual of planning and design guidance awards a density
bonus to station inuence areas. It divides each station inuence
area into two distance bands, an inner band within 200 m of the
station and an outer band between 200 and 500 m. Around metro
stations, oor-area-ratio (FAR) in the inner band can be as high as
160% of base FAR outside the inuence area, and those in the outer
band can be as high as 140% of base FAR (SZGOV, 2011b).
4. Evidence from land markets
The examples described above appear to conrm our hypoth-
esis that metro planning driven by a desire to boost land sector
revenue tends to overlook ridership and emphasize land
development potential. Is this true in general throughout the city?
Can we see any evidence from the land market? If this is indeed
the case, one would see that metro stations were placed in areas
with lower land prices, or at least without higher land prices. That
is to say, land prices should have no station orientation, i.e., the
closer to the station, the higher the land price per square meter,
before the relevant metro plan was announced.
4.1. Station orientation of land prices
The website of the Shenzhen Land and Real Estate Exchange
Center records all land parcel transactions from 2000 to 2014. This
data set includes price of the parcel transaction, year of transac-
tion, parcel location, parcel size, planned land use type, and
transaction mode (auction, tendering or listing). Fig. 5 shows the
geo-referenced parcels and the metro stations in operation today.
In order to show whether parcel price has a station orientation,
we calculated air-distance from each parcel to its nearest metro
station and the land transaction price per square meter. We use
air-distance rather than route distance because we are mainly
interested in whether station-orientation exists before metro plan
announcement. Many roads today around metro station might not
have existed at the time of the transaction. We also divide all land

Please cite this article as: Yang, J., et al., Density-oriented versus development-oriented transit investment: Decoding metro station
location selection in Shenzhen. Transport Policy (2016), http://dx.doi.org/10.1016/j.tranpol.2016.04.004i
6 J. Yang et al. / Transport Policy ()

Fig. 5. Land parcels and metro lines.

parcels into three groups according to its temporal relationship
with the planning and operation of its closest metro stations.
Group A includes land transactions happening before the plan for
its nearest station was announced. Group C includes land trans-
actions happening after the relevant metro station was put into
service. Group B includes land transactions happening during the
interval between plan announcement and station operation.
Among the total 640 parcels in our database, 145, 293 and 202
parcels fall into groups A, B and C respectively.
Fig. 6 plots land price (thousand yuan per square meter) for

(before) (during) (after)

Fig. 6. Land parcels price and FAR around the nearest station. Note: land price in Shenzhen appreciate faster year by year. In order to make the graphs more readable, the
price here is already adjusted by land price index. So land transaction price in different years can be compared in a visual format. The horizontal lines are added to make it
easy for inter-graph comparison.

Please cite this article as: Yang, J., et al., Density-oriented versus development-oriented transit investment: Decoding metro station
location selection in Shenzhen. Transport Policy (2016), http://dx.doi.org/10.1016/j.tranpol.2016.04.004i
 J. Yang et al. / Transport Policy () 7

each parcel in relation to its distance to the closest metro station.
We also plot (FAR) in relation to distance to the nearest metro
station. FAR is calculated as the ratio of the maximum allowed
oor area to parcel size. Observation suggests that operating metro
stations have noticeable impacts on parcel price within a 1 km
radius. We double that threshold distance and include all cases
within 2 km of their nearest stations for detailed analysis. This is
to make sure that a station orientation of land price or FAR can be
observed, if it indeed exists. We exclude parcels beyond 2 km
because they may not be particularly useful for our analysis, which
centers on the development situation at station proximity.
Fig. 6 indeed demonstrates a strong station orientation for land
transaction happening after station construction (group C in
Fig. 6), which indicates a market appreciation of transit accessi-
bility. This is no surprise and is consistent with many empirical
studies of housing and land prices in China (Feng et al., 2011; Gu
and Zheng, 2010; Pan and Zhong, 2008; Zhang et al., 2012). Im-
portant to this study, we nd a much weaker or even no station-
orientation for parcels transacted before metro plan announce-
ment (group A in Fig. 6), indicating that station placement might
have bypassed community centers. Since many other factors also
affect price, we have to control for other variables in further
analysis.
4.2. Spatial econometric models
Our regression models use a hedonic approach. The dependent
variable used in our models is the logarithm of the land parcels
transaction price (per sq meter), and the explanatory variables
include the distance attributes, time attributes, land use types,
transaction types and access to amenities (Table 3). Various studies
have concluded that attributes like land use type (Cervero and
Landis, 1993), land transaction mode (Li and Wu, 2010), and ac-
cessibility to other amenities such as hospitals, schools and service
facilities (Linneman, 1980) strongly inuence real estate price and
should be included as explanatory variables.
First, we include three distance variables. We dene variable
D1 as the distance from the land parcel to its nearest metro sta-
tion. We add variable D2, which is the distance from the nearest
station to the nearest transfer stations, as proximity to the transfer
station implies higher convenience when travelling by metro.
When D2 equals zero, the closest metro station is a transfer sta-
tion. Shenzhen has 13 transfer stations in total. We also include a
variable D3, which is the distance from the nearest station to the
city center's Civic Center Station. It serves as a measure of acces-
sibility to the city's peak land value. While D1 is air-distance, both
D2 and D3 are distances along the metro lines.
Second, regarding the time points when transactions happen in
relation to station planning and construction, we consider three
variables: groups A, B and C, as dened above. A large number of
studies have demonstrated that impacts on land price may be-
come noticeable following metro plan announcement and metro
operation (Bae et al., 2003; Grimes and Young, 2013; Knaap et al.,
2011). Here, we are more interested in knowing what had hap-
pened before station location was announced.
Third, we added variables relevant to other amenities, such as
distance to nearest hospitals, schools, parks, and shopping malls.
We have data in GIS format, showing the location of those points
of interest. Fourth, land price is always affected by land use type. In
China, residential and commercial land parcels are typically priced
higher than industrial use. We use a dummy variable to indicate
each of the three land use categories. Fifth, transaction modes may
also affect land price. Dummy variables are used to indicate the
mode used for each parcel transaction (auction, tendering and
listing). Listing is widely recognized as the least competitive
transaction mode, as it allows transaction to happen when there is
only one bidder.
Since land parcels located close to each other share the char-
acteristics of the nearby area, the transaction prices are inevitably
spatially dependent. Moran's I can be used to detect whether there
is spatial autocorrelation among the regression residuals (Anselin,
1995). After testing that, we developed both spatial lag and spatial
error models to address the issue of spatial dependence. Models
are estimated with GeoDa. The robust Lagrange multipliers of
spatial lag models are higher than those of spatial error models.
For example, model 2's Lagrange multiplier is 5.87 for spatial lag
model but only 1.00 for the spatial error models. Table 4 thus lists
only results of spatial lag models. All three models use log-trans-
formed price as the dependent variable. Different models are es-
timated with different data records or different combination of
independent variables.
Regression model:
3 2 3
LnP = P 0 + di Di + ti GroupPi + i LDi
i=1 i=1 i=1
2 5
+ mi TRi + Pi POIi + W ln P +
i=1 i=1

Table 3
Specication of variables.

Variable Name Denition Form Maximum Minimum Average

Dependent Ln(P) Land parcels transaction prices Logarithm 12.73 3.17 8.63
Distance attributes D1 Distance from land parcel to its nearest metro station. Continuous (km) 1.99 0 0.73
D2 Distance along the metro lines from land parcel's nearest Continuous (km) 24.78 0 4.01
station to its nearest transfer station.
D3 Distance along the metro lines from land parcel's nearest Continuous (km) 37.08 0 16.54
station to the city center.
Time attributes: (Reference: Group B Transaction after plan announcement, but before operation Dummy (1 if yes,0 if 1 0 0.35
Group A) (during) no)
Group C (after) Transaction after operation Dummy 1 0 0.32
Land use types (Reference: COM_LD Commercial land Dummy 1 0 0.30
others) RES_LD Residential land Dummy 1 0 0.28
IND_LD Industrial land Dummy 1 0 0.25
Transaction types (Reference: AUC Auction Dummy 1 0 0.07
listing) TEN Tendering Dummy 1 0 0.03
POI HOSP Hospital within 1 km Dummy 1 0 0.35
PARK Park within 1 km Dummy 1 0 0.20
PRIM Primary school within 1 km Dummy 1 0 0.66
MIDD Middle school within 1 km Dummy 1 0 0.61
COLL College within 2 km Dummy 1 0 0.25

Please cite this article as: Yang, J., et al., Density-oriented versus development-oriented transit investment: Decoding metro station
location selection in Shenzhen. Transport Policy (2016), http://dx.doi.org/10.1016/j.tranpol.2016.04.004i
8 J. Yang et al. / Transport Policy ()

Table 4 though the metro stations were placed in relatively under-devel-

Spatial lag model estimation. oped areas.
Other control variables have meaningful parameters. For ex-
Model Model 1 Model 2 Model 3
ample, commercial parcels have higher prices than residential
Data Group A, All groups, All groups, parcels, and industrial land has the lowest prices. Whereas D2 is
within 2 km within 2 km within 1 km not signicant in the models, probably indicating that hub stations
Distance D1  0.21  0. 16 0.32
in Shenzhen have no noticeable advantage over other stations
attributes D2 0.03 0.02 0.02
D3  0.05***  0.04*  0.04***
when attracting development. In addition to the estimated para-
Time Group B 0.97*** 1.12*** meters, the models have robust goodness of t measures. The
(during) resulting Moran's I indicators show that the spatial lag models
Attributes Group C 1.60*** 1.65*** have successfully addressed the spatial autocorrelation issue.
(after)
Land use types COM_LD 1.59*** 2.33*** 2.52***
RES_LD 1.02* 1.07*** 1.11***
IND_LD 0.55 0.26 0.26 5. Development around stations
Transaction types AUC  0.26 0.12  0.01
TEN 0.45 0.87 0.84
Qualitative and quantitative analysis both indicate that
POI HOSP  0.24  0.11  0.03
PARK 0.22 0.28 0.25 Shenzhen's metro placement tends to bypass the core of estab-
PRIM 0.71*** 0.36 0.52* lished communities, behind which is a desire of the city govern-
MIDD  0.37  0.37*  0.35 ment to boost land development around metro stations and collect
COLL 0.29 0.25 0.25 land development revenue. This linkage between land develop-
Intercept CONS 5.40*** 5.56*** 6.16***
Spatial W_LnP 0.24* 0.19** 0.07
ment revenue and government planning decisions has been re-
dependence ported in China's housing and land development literature (Tian
Statistic values AIC 286.90 1134.82 893.15 and Ma, 2009; Lin and Yi, 2011). In the decision-making for metro
Log  128.45  550.41  429.57 planning, this desire for land development revenue is supported
Likelihood
by the market appreciation of transit proximity. Parcels closer to
Moran's I 0.005  0.001  0.008
Number 103 314 236 metro stations appreciate much faster than those more distant.
Behind this planning practice is an assumption that fast de-
(*) for pr 0.05, (**) for pr 0.01, (***) for p r 0.001. velopment will happen around planned metro stations. This
build-it-and-they-will-come philosophy sometimes proves itself
Of central interest, model 1 indicates that land transaction price quickly, but sometimes it does not. In Shenzhen, it worked much
before metro plan announcement has no station-orientation Even better for the Longgang and Longhua Lines with a radial orienta-
though land price decreases as distance to the city center in- tion, but much slower for the Huanzhong Line. What makes the
creases, it does NOT decrease as distance to the metro station in- difference, perhaps, is that the former two both have terminals in
creases. The lack of a station-orientation suggests that metro the urban core. Improved access to the urban core brings fast
placement indeed has bypassed the core areas of established development around the suburban stations. When the Longgang
communities, where land price tends to be higher. Note that our Line was rst opened in 2011, the trains were almost empty.
conclusion here is not inconsistent with studies that tend to sug- Starting from 2014, however, it has been badly congested at peak
gest positive or negative impacts of metro operation (Cervero and periods in the peak direction.
Kang, 2011; Cervero et al., 2011; Pan and Zhang, 2008; RICS, 2002; When development fails to catch up, the expense is low transit
Ryan, 1999; Debrezion et al., 2006; Pennington et al., 1990). Model ridership, a high operating subsidy and possible public criticism
1 here includes only land transaction happening before metro plan about government investment policies. The Shekou Line in
announcement. It reects government's locational choice of metro Shenzhen, for example, was planned to foster land development
stations, rather than the impacts of metro investment. and most stations were outside densely populated area. As a re-
We can of course reproduce regression results similar to ex- sult, ridership did not meet expectations and its operational decit
isting studies. We developed model 2 and model 3 by adding the reached 481 million RMB (US$ 77.58 million) since its operation
two time attribute variables. Note that model 3 uses a subset of began in 2011 (Peng, 2014).
model 2's data, those transactions within 1 km of metro stations. This development-oriented practice for metro planning is not
With model 2, using Group A (transaction before metro plan an- unique to Shenzhen. It is also observable in other Chinese cities
nouncement) as the benchmark, price becomes 2.64(exp(0.97) such as Hangzhou, Wuxi, Nanning and Chengdu (Lin, 2011; Lv,
2.64) times higher after metro plan announcement and further 2012). Despite affordable transit fares, low ridership and huge
becomes 4.95 (exp(1.60) 4.95) times higher following metro decits have been reported in these cities. Taking Shenzhen's
operation. With model 3, following metro plan announcement, Hangzhou's Line 1 as an example, the average annual operating
parcel price within 1 km of metro stations becomes 3.06 times cost was 578 million RMB (US$ 93.2 million), while the fare-box
(exp(1.12) 3.06) higher, compared to 2.64 in model 2. Following revenues in 2013 amounted to only 304 million RMB (US$ 49.0
metro operation, price becomes 5.21 (exp(1.65) 5.21) times million), resulting in a decit of 274 million RMB (US$ 44.2 mil-
higher, compared to 4.95 in model 2. A comparison of these two lion). According to local news sources, the low ridership has causes
models thus shows a faster price increase at station proximity. in metro alignment and station location selection. Line 1 was
The signicance of time attributes is consistent with ndings planned to connect the city center and the inner suburbs. Driven
from other studies. For example, Grimes and Young, who used by a desire to leverage more development and more land lease
Auckland Western Line as a case, found a noticeable increase in revenue, the municipal government revised the plan and extended
land prices after the announcement of a metro upgrade strategy the line to Xianghu Station, which was still largely rural (Lv, 2012).
(Grimes and Young, 2013). Similarly, parcels within 1 km of the
metro station had 3040% higher prices after the publication of 6. Conclusions and discussion
MAX light rail's plan in Portland, Oregon (Knaap et al., 2011).
Detection of similar impacts of metro investment in Shenzhen The co-location of transit stations and relatively high-density
points out that land markets indeed produce positive returns, even clusters is generally praised as an ideal practice to attain higher

Please cite this article as: Yang, J., et al., Density-oriented versus development-oriented transit investment: Decoding metro station
location selection in Shenzhen. Transport Policy (2016), http://dx.doi.org/10.1016/j.tranpol.2016.04.004i
 J. Yang et al. / Transport Policy ()
transit ridership and stable fare box revenue. When planning for
metro projects in China, many local governments, however, have
put metro stations at relatively under-developed places, with an
expectation that development will quickly follow once station
locations are announced and the metro is in operation. Behind this
development-oriented approach, is a pressure to fund metro
projects with land lease revenue, in addition to lowering right-of-
way cost. As a result, project ranking drifts from ridership concern
toward land development potential. Metro alignment follows un-
der-developed corridors. When interests in land development
collides with public demand for efcient transportation, govern-
ment's preference for land lease revenue leads to favorable con-
sideration of under-developed station locations, at the expense of
ridership and community transit preferences. This planning prac-
tice is conrmed with evidences of decision-making on metro
planning and land market analysis in Shenzhen.
Note that this development-oriented approach may not be a
problem in the long-run, considering that the share of urban po-
pulation in China is still less than 55% and urbanization is pro-
ceeding rapidly. The market appreciation for metro accessibility
indicates that station areas are indeed hot spots for new devel-
opment. The development-oriented approach thus provides an
opportunity for metro investment to shape the spatial pattern of
urban growth. The aim of this paper is not to criticize the devel-
opment-oriented approach, but to point out the incentive struc-
ture behind decision-making on metro planning and ultimately
the implications for spatial patterns of urban development.
A system of property taxation is expected in China. The es-
tablishment of property tax may restructure the funding system
for city government and thus change the practice of metro plan-
ning. With a property tax, the ability of transit investment to boost
real estate value will lead to increased tax revenue for the city
government, which will lower city governments resistance to
placing stations at pre-existing high-density areas. By contrast,
placing stations among empty parcels will bring much lower or
even no property tax for city governments. The exact impact of
this tax reform will depend critically on the proportion of city
government revenue that is raised by property tax.
Note that city governments preference for land-sector revenue
over fare-box revenue in metro planning has its roots not only
within China's scal institutions, but also in its political traditions.
In a democratic society, where city government leadership is se-
lected via popular votes, the community may not tolerate empty
trains going back and forth between sparsely populated stations.
Such a white elephant would be politically expensive and would
motivate planners and elected ofcials to choose the routes of
highest ridership to appeal to the largest number of voters. This
community pressure exists in Shenzhen, as evidenced by the case
of Dalang station.
China's political system, however, has enabled the city gov-
ernment to act without much regard for community interests. The
city leadership is appointed by a higher level of government rather
than by election by the grassroots. They will respond to the per-
formance indicator of economic development, which is imposed
by the higher-level government, more than the desire for efcient
transportation by the local residents. In fact, places with less de-
mocratic practices, including Hong Kong and Singapore, have ex-
celled in encouraging development through mass transit invest-
ment. Route alignment and station placement are less driven by
political tradeoffs. New political measures that empower com-
munities in Chinese cities can surely produce signicant impacts
on metro planning.
Please cite this article as: Yang, J., et al., Density-oriented versus development-oriented transit investment: Decoding metro station
location selection in Shenzhen. Transport Policy (2016), http://dx.doi.org/10.1016/j.tranpol.2016.04.004i
9
Acknowledgement
This research was funded by China's National Natural Science
Foundation Grant 41371168 and the Shenzhen Overseas High-Ca-
liber Personnel Grant KQCX20130628093909157.
References
Anselin, L., 1995. Local indicators of spatial associationlisa. Geogr. Analysis 27,
93115.
Bae, C.C., Jun, M., Par, H., 2003. The impact of Seoul's Subway Line 5 on residential
property values. Transp. Policy 2 (10), 8594.
Cervero, R., Kang, C.D., 2011. Bus rapid transit impacts on land uses and land values
in Seoul, Korea. Transp. Policy 18, 102116.
Cervero, R., Landis, J., 1993. Assessing the impacts of urban rail transit on local real-
estate markets using quasi-experimental comparisons. Transportation Research
Part a-Policy and Practice 27, 1322.
Debrezion, G., Pels, E., Rietveld, P., 2006. The impact of rail transport on real estate
prices: an empirical analysis of the Dutch housing market. Tinbergen Institute
Discussion Paper.
Dorantes, L., Paez, A., Vassallo, J., 2011. Analysis of house prices to assess the eco-
nomic impacts of new public transport infrastructure: Madrid metro line 12.
Transp. Res. Rec.: J. Transp. Res. Board, 2245.
Feng, C., Li, W., Zhao, F., 2011. Inuence of rail transit on nearby commodity housing
prices: a case study of Beijing subway line ve (Chinese). Acta Geogr. Sin.
(Chin.) 66 (8), 10551062.
FTA., 2012. Capital Investment Program FY 2012 Evaluation and Rating Process.
(Retrieved from) www.fta.dot.gov/documents/FY12_Evaluation_Process(1).
pdf. Accessed May 10, 2015.
Grimes, A., Young, C., 2013. Spatial effects of urban rail upgrades. J. Transp. Geogr.
30, 16.
Gu, Y., Zheng, S., 2010. The Impacts of Rail Transit on Property Values and Land
Development Intensity: The Case of No.13 Line in Beijing (Chinese). Acta Geogr.
Sin. (Chin.) 65 (2), 213223.
Hess, D.B., Almeida, T.M., 2007. Impact of proximity to light rail rapid transit on
station-area property values in buffalo, New York. Urban Stud. 44 (5),
10411068.
Knaap, G.J., Ding, C., Hopkins, L.D., 2011. Do plans matter? The effects of light rail
plans on land values in station areas. J. Plan. Educ. Res. 21, 3239.
Lee, E.S.W., Leung, C.K., 1994. Transport planning for new town development: ex-
perience of Hong Kong. J. Chin. Geography 4 (1), 113.
Li, X., 2011. Shenzhens Property Rial is not easy and the budgetary burden is
really heavy. (Retrieved from) http://nance.sina.com.cn/china/dfjj/20111226/
140511064735.shtml?qq-pf-to pcqq.c2c. Accessed May 4, 2015.
Li, Z., Wu, S., 2010. The relationship between land transaction system and land
prices (Chinese). Spat. Zone Econ. (Chin.) 9, 301302.
Lin, G., Yi, F., 2011. Urbanization of capital or capitalization on urban land? Land
development and local public nance in urbanizing China. Urban Geogr. 32 (1),
5079.
Lin, Y., 2011. Professors said the average cost of metro line per kilometer is
0.5 billion. (Retrieved from) http://news.163.com/11/0721/04/
79F8HCHL00014AED.html. Accessed June 8, 2015.
Linneman, P., 1980. Some empirical results on the nature of the hedonic price
function for the urban housing market. J. Urban Econ. 8 (1), 4768.
Lv, M., 2012. Decit of Line 1 happens in Hangzhou and the local oppose the high
transit fee. (Retrieved from) http://news.ifeng.com/shendu/nfzm/detail_2012_
11/02/18775016_0.shtml. Accessed May 17, 2015.
Pan, H., Zhong, B., 2008. The impact of metro on property price: a case study of
Shanghai (Chinese). Urban Plann. Forum (Chin.) 2, 6269.
Pan, H., Zhang, M., 2008. Rail Transit impacts on land use: evidence from Shanghai,
China. Transp. Res. Rec.: J. Transp. Res. Board 812, 1621.
Peng, Y., 2014. Shenzhen Line 2s Planning is lack of consideration, with a decit of
481 billion RMB in 3 years. (Retrieved from) http://bj.loupan.com/html/news/
201412/1523905.html. Accessed June 25, 2015.
Pennington, G., Topham, N., Ward, R., 1990. Aircraft noise and residential property
values adjacent to manchester international airport. Journey Transp. Econ.
Policy 1 (25), 4959.
RICS, 2002. Land value and public transport.: the Royal Institution of Chartered
Surveyors. Ofce of the Deputy Prime Minister.
Ryan, S., 1999. Property values and transportation facilities: nding the transpor-
tation-land use connection. J. Plan. Lit. 4 (13), 412427.
Samanta, S., Jha, M.K., 2008. Identifying feasible locations for rail transit stations
two-stage analytical model. Transp. Res. Rec.: J. Transp. Res. Board 2063, 8188.
Shenzhen Government (SZGOV). 2011a. Short-term metro system plan of Shenzhen
(20112016). Shenzhen: Urban Planning, Land and Resources Commission of
Shenzhen Municipality. (Retrieved from) http://www.ditiezu.com/thread-
367550-1-1.html. Accessed June 25, 2015.
Shenzhen Government (SZGOV). 2011b. Shenzhen Urban Planning Standards and
Guidelines. (Retrieved from) http://www.szsmb.gov.cn/content.asp?
id 16648. Accessed June 26, 2015.
Shenzhen Government (SZGOV). 2014. The introduction of Shenzhen. (Retrieved
from) http://www.sz.gov.cn/cn/zjsz/szgl/201408/t20140812_2544674.htm.
10 J. Yang et al. / Transport Policy ()

Accessed June 26, 2015. Transp. Res. Board 1954, 714.

Shenzhen Metro Corporation (SZMC). 2013. Annual Report of 2013. (Retrieved
from) http://www.szmc.net/public/cms/public/upload/infor/attach/2014-06-
27/1403861082016.pdf. Accessed June 28, 2015.
Shenzhen Metro Corporation (SZMC). 2015a. Investment introduction of Shenzhen
metro system. Accessed June 28, 2015.
Shenzhen Metro Corporation (SZMC). 2015b. The plan of Shenzhen metro system.
(Retrieved from) http://www.szmc.net/page/plan/index.html. Accessed April
20, 2015.
Tian, L., Ma, W., 2009. Government intervention in city development of China: a
tool of land supply. Land Use Policy 26 (3), 599609.
Yang, J., 2006. Transportation implications of land development in a transitional
economy: evidence from housing relocation in Beijing. Transp. Res. Rec.: J.
Yang, J., Feng, C., Cao, G., 2007. Land and transportation development in china: an
economic analysis of government behavior. Transp. Res. Rec.: J. Transp. Res.
Board 2038, 7883.
Yang, J., Luan, X., Lin, X., McGuinness, E., 2014. Emerging public-private partner-
ships in Chinas rail mass transit: the case of Shenzhen. Transp. Res. Rec.: J.
Transp. Res. Board 2450, 127135.
Zhang, W., Li, H., Duan, X., 2012. The impacts of rail transit on property valuesthe
case of No.1 line in Beijing (Chinese). Econ. Geogr. (Chin.) 32 (2), 4665.
Zhao, P., 2013. Shenzhen metro systems development market-oriented mode and it
still needs 200 billion RMB. (Retrieved from) http://nance.sina.com.cn/roll/
20130218/145514572951.shtml?qq-pf-to pcqq.c2c. Accessed July 25, 2015.

Please cite this article as: Yang, J., et al., Density-oriented versus development-oriented transit investment: Decoding metro station
location selection in Shenzhen. Transport Policy (2016), http://dx.doi.org/10.1016/j.tranpol.2016.04.004i