Article

Urban Studies
2018, Vol. 55(12) 2586–2602
Ó Urban Studies Journal Limited 2017
Ungating the city: A permeability Article reuse guidelines:
sagepub.com/journals-permissions
perspective DOI: 10.1177/0042098017733943
journals.sagepub.com/home/usj

Guibo Sun
Faculty of Architecture, The University of Hong Kong, China

Chris Webster
Faculty of Architecture, The University of Hong Kong, China

Alain Chiaradia
Department of Urban Planning and Design, The University of Hong Kong, China

Abstract
China is seeking to prohibit the construction of any new gated communities and to gradually open
existing schemes after three decades of growth of large-block gated estates. In this article, we
use permeability analysis to explore the ‘what if?’ question posed by the policy: what if gated com-
munities became permeable? We ask the question in respect of non-motorised access. We use
two permeability metrics, closeness and betweenness, as outcome measures of gated and non-
gated versions of the city. We construct a bespoke complete pedestrian network, rather than
using the road network, for our permeability modelling. Nanchang, a medium-sized Chinese city
with widespread gated communities, is our study area. A series of permeability analyses with and
without gated communities is conducted using GIS and spatial design network analysis (sDNA).
On the basis of these analyses, we sequentially sort the gated compounds whose opening will
maximise permeability gains with minimum expropriation of property rights through coercive
ungating. We offer the analysis to urban scholars, planners and governments by way of a quanti-
fied simulation. This study and methodology, which is transferable without high data require-
ments, can assist urban practitioners in reconfiguring urban form to promote a healthier living
environment (more walking) and more economically viable local service centres (greater pedes-
trian footfall concentrations).

Keywords
gated community, geographic information system, pedestrian network, permeability, walking

Corresponding author:
Guibo Sun, Faculty of Architecture, The University of
Hong Kong, 4/F Knowles Building, Pokfulam Road, Hong
Kong 000000, China.
Email: [email protected]
Sun et al. 2587

᪈㾱
ѝഭ㓿শйॱᒤབྷ⡷ሱ䰝ᔿሿ४ᔪ䇮ਁኅਾˈ↓䈅മ⾱→ᔪ䇮ԫօᯠⲴሱ䰝ሿ४ˈᒦ൘䙀↕ᢃ
ᔰ⧠ᴹⲴሱ䰝ሿ४DŽᵜ᮷䘀㹼⑇䘿ᙗ࠶᷀ᶕ᧒䇘䘉а᭯ㆆᑖᶕⲴ䰞仈˖ྲ᷌ሱ䰝ᔿሿ४ਈᗇਟ
⑇䘿ˈՊਁ⭏ӰѸ˛ᡁԜӾ䶎ᵪࣘ䙊䚃Ⲵ䀂ᓖᨀࠪҶ䰞仈DŽ᮷ㄐ䟷⭘Ӣᇶᓖ઼ѝ䰤ᓖ䘉є亩⑇
䘿ᙗᤷḷˈ֌Ѫ㺑䟿ሱ䰝ර઼䶎ሱ䰝ර෾ᐲн਼㔃᷌ⲴᤷḷDŽᡁԜᶴᔪҶањᇊࡦⲴᆼᮤ㹼Ӫ
㖁㔌䘋㹼⑇䘿ᙗᔪ⁑ˈ㘼нᱟ֯⭘䚃䐟㖁㔌DŽ⹄ウሩ䊑ᱟѝഭই᰼䘉њࡠ༴ᤕᴹሱ䰝ᔿሿ४Ⲵ
ѝර෾ᐲDŽᡁԜ䘀⭘ GIS ઼オ䰤䇮䇑㖁㔌࠶᷀ (sDNA) ሩሱ䰝ᔿሿ४઼䶎ሱ䰝ᔿሿ४ᔰኅҶа
㌫ࡇ⑇䘿ᙗ࠶᷀DŽ൘䘉Ӌ࠶᷀Ⲵส⹰кˈᡁԜᧂࡇᮤ⨶Ҷ䛓Ӌሶ䙊䗷ᔰ᭮㘼ᴰབྷ〻ᓖᨀ儈⑇䘿
ᙗˈ਼ᰦᴰሿ〻ᓖഐᕪࡦᔰ᭮㘼‫ץ‬ঐӗᵳⲴሱ䰝ᔿ㔬ਸփDŽ䙊䗷䟿ॆ⁑ᤏⲴᯩᔿˈ䘉Ӌ࠶᷀ਟ‫׋‬
෾ᐲ⹄ウᆖ㘵ǃ㿴ࡂ㘵઼᭯ᓌᡰ⭘DŽᵜ亩⹄ウ઼⴨‫Ⲵޣ‬ᯩ⌅ᰐ䴰儈ḷ߶ᮠᦞণਟ〫⭘ˈਟԕᑞࣙ
෾ᐲᇎ䐥㘵䟽ᯠᐳተ෾ᐲᖒᘱˈԕ‫׳‬䘋ᴤ‫ڕ‬ᓧⲴ⭏⍫⧟ຳ˄ᴤཊ↕㹼˅઼ᴤާ㓿⍾ਟ㹼ᙗⲴൠᯩ
ᴽ࣑ѝᗳᔪ䇮˄ᴤབྷⲴ↕㹼㚊䳶ᓖ˅DŽ

‫ޣ‬䭞䇽
ሱ䰝ᔿሿ४ǃൠ⨶ؑ᚟㌫㔏ǃ㹼Ӫ㖁㔌ǃ⑇䘿ᙗǃ↕㹼

Received January 2017; accepted August 2017

Introduction argue that urban areas should strive for tran-

sit, pedestrian and bicycle systems access,
Gated community research has tended to while reducing dependence upon the auto-
focus on social cost-benefit calculus, with an mobile (Talen, 2013). Permeability, permit-
emphasis on radical social theoretic commen- ting ease of movement and avoiding urban
tary in the 1990s giving way to more prag- severance, is generally considered a positive
matic social commentary and a growing slate attribute of an urban design scheme (Melia,
of empirical studies measuring alleged social 2012). Gated communities, by definition and
and private costs and benefits (Le Goix, geometric and economic necessity, tend to
2005; Low, 2001; Webster, 2001). Much less segregate urban physical structure and to
effort has been given to normative design fragment urban spaces. Services, goods and
studies of gated estates or to empirically facilities are provided exclusively to commu-
based theories of the physical urban form nity members (Glasze et al., 2004; Miao,
associated with gated communities (Miao, 2003). Access restriction is hard-wired into
2003). Urban planners must understand not the business rationale of residential clubs.
only the functions and processes that create This means that the division between gated
urban patterns, but also the direction in and ungated communities creates new socio-
which urban form should change under vari- spatial patterns. The segregation raises
ous normative drivers (Talen, 2002). Social important questions about the future forms
science studies of urban planning should of urban development, since a gated city
deliver something back to practitioners. tends to have sharper social geography divi-
Private neighbourhoods with their gates, sions at a fine scale than traditional cities
impermeability and self-sufficient club (Manzi and Smith-Bowers, 2005; Scott,
economies have significant implications for 2002; Webster et al., 2018). One question
cities. A good city form is about having arises: is it possible to maintain an open and
permeable spatial patterns that can facilitate permeable city while accommodating the
accessibility and maximise connectivity seemingly unstoppable gated community
(Lynch, 1981; Talen, 2002). New urbanists genre of development, with its necessary
2588 Urban Studies 55(12)

degrees of enclosure? Answering this requires physical and mental health, sustainability of
us to be specific about the particular perfor- transport, reduction of traffic congestion,
mance domains of gated cities that might improvement of air quality and economic
need to be controlled. In the study, we focus wellbeing and inclusion of communities
on a single dimension of urban performance, (Alfonzo et al., 2014). Physical inactivity,
namely permeability. This allows us to overweightness and obesity in car-oriented
answer the question: how would accessibility gated compounds with exposure to heavy air
be re-distributed across the city if gates were pollution in daily life is popularly under-
removed? This, in turn, permits us to ask: stood to be a major risk factor in China. If
how could we prioritise gated communities urban planning is the biggest possible public
for ungating if we wanted to achieve maxi- health intervention, removal of the gates in
mum accessibility improvements with the Chinese cities could potentially have major
least possible social and legal disruption? health benefits through increasing walking
China is seeking to prohibit construction and cycling. The literature suggests that this
of any new gated communities and to gradu- might significantly help lower the risk of the
ally open existing schemes after three major non-communicable diseases such as
decades of growth of large-block gated coronary heart disease, type 2 diabetes, obe-
estates. New guidelines released by the cen- sity and shortened life expectancy (Lee et al.,
tral government in February 2016 call for 2012). We note, however, that the Chinese
more compact, livable, green and sustainable government’s interest in opening the gates is
cities (Normile, 2016). This is a turning not based on health performance. The policy
point in city form that breaks with the tradi- largely responds to the perceived negative
tions of both planned and market-led econ- effects on urban movement and congestion
omy periods. City form immediately post- of traffic – large closed blocks are thought to
1949 was heavily influenced by soviet- be inefficient in this respect.
inspired wide streets, huge blocks and work- In this article, we use a permeability per-
unit estates (danwei). US-style urban arter- spective to ask the question: ‘what if gated
ials and ringways were added from the late communities become permeable for non-
1990s, and the land between became rapidly motorised access?’. We propose two perme-
filled with commodity housing interspersed ability metrics, closeness and betweenness, to
with commercial and industrial buildings, use as outcome measures of gated and non-
most often planned in impermeable superb- gated versions of the city. We construct a
locks. Virtually all commodity housing in bespoke complete pedestrian network, rather
China has restricted entrances, and a grow- than the traditionally-used road network, for
ing number of developments are in super- the permeability modelling. Nanchang, a
block gated compounds. medium-sized Chinese city with widespread
There is research that supports the central gated communities, is our study area. A
government’s wish to ungate China’s urban series of permeability analyses of the city,
super-blocks, especially healthy city research with/without gating, is conducted using a
that emphasises the importance of street con- geographic information system (GIS) and
nectivity for health (Chen et al., 2016; Day spatial design network analysis (sDNA)
et al., 2013), under the assumption that (Cooper et al., 2016). On the basis of these
urban form can affect physical activity, espe- analyses, we sequentially sort the gated com-
cially walking and cycling. Other studies pounds whose ungating will create maximum
show that more highly connected and perme- effects. This allows a kind of mini-max anal-
able locations are associated with better ysis of the issue: which gates to remove to
Sun et al. 2589

maximise permeability gains with minimum communities from surrounding areas. Some
expropriation of property rights through work units belong to state-owned enterprises
coercive ungating? that are still viable or which have diversified
their enterprises and have renovated their
gated danwei properties to provide employ-
Gated communities and ees with updated living conditions (Lu,
car-oriented urbanism 2006). Many modern gated communities
have been built on inner-city industrial dan-
Physical layouts of gated communities in wei areas as brownfield redevelopments.
China Others have been built on converted farm-
Gated community in contemporary China land as the city has expanded.
presents a tangled picture in terms of devel-
opment process and physical, economic and
Car-oriented urbanism and gated
social outcomes. Two main types of gated
compound layouts coexist: legacy danwei compounds
estates; and privately developed, secured Legacy danwei systems still impact city
gated compounds. structure and travel behaviour (Sun et al.,
Although there was wide variation among 2016). Most danwei gates nowadays are
the soviet-style work-unit compounds, in open in the daytime and provide free access
terms of scale, facilities and social services, for outsiders by foot or bicycle. Some are
there was an overwhelmingly ubiquitous closed at night for security reasons. These
presence of walls and guarded gates. gates do not bother local people, who are
Facilities in danwei were exclusively for the familiar with them and their opening hours.
use of unit members. Through-traffic was However, they are barriers for strangers,
forbidden at any time. From a bird’s eye inhibiting through-travel (Xu and Yang,
view at the city scale, large urban com- 2009). The road system of work units has
pounds consisting of several to tens of acres never been integrated into the general road
are typical, distinguished by the regularity of network of the city. Whereas local distribu-
the block-house layouts and unified building tor roads in a western city will have been
heights of five to six walk-up floors, with designed to roughly optimise local and trunk
walkways laid out between the blocks and accessibility for local residents and other
only minimal connections with local streets users, local distribution networks within a
outside the compound (Glasze et al., 2004; danwei are more likely to be determined by
Wu, 2005). the gaps between buildings and internal cir-
Contemporary Chinese gated compounds culation logic, with no reference to the next
are usually large, involving hundreds of level of network outside.
acres and thousands of residents. The built Similarly, streets internal to contempo-
forms tend to be high-rise arranged in novel rary gated compounds are also isolated from
layouts, with much of the ground area city road networks. In the immediately sur-
devoted to elaborate landscaping and having rounding areas of gated communities, there
clearly identifiable boundaries and gated are often extensive lengths of ‘dead-frontage’
entrance ways. Their physical manifestations building setbacks abutting pavements and
include natural and human-made walls, or devoid of retail and amenities. This is espe-
outward-facing buildings constructed cially the case in the new development dis-
around the periphery, all of which have the tricts on the fringe of cities all over China.
effect of physically segregating gated This tends to give rise to limited local street
2590 Urban Studies 55(12)

vitality and an unstimulating experience for current level of gatedness. To measure per-
walking, and is thus associated with lower meability, we borrow theories from network
non-motorised rates of movement (Day analysis (Newman, 2010).
et al., 2013). Travelling between the superb-
locks of these communities involves walking Closeness: Measuring permeable
long distances because of the lack of perme- to-movement
ability. Pedestrian and cycle trips necessarily
have to be routed outside of the gated blocks For the to-movement component of perme-
and onto arterials that carry automobiles, ability we use closeness centrality. Closeness is
trucks and buses in growing numbers (Tao commonly used in social and other network
et al., 2010). Clustered gated compounds, studies. It measures the mean distance from a
with impermeability and mega-block struc- node to other nodes using the shortest path.
tures, are likely to induce greater car-driving We use Euclidean distance to measure the
behaviour (Normile, 2016). shortest path. It takes low values for nodes
that are separated from others by only a short
geodesic distance. Such nodes in an urban
Permeability and its measures street network are on average likely to have
Permeability has two important functional better access to traffic, services and informa-
tion, being more directly connected to and
characteristics in urban design analytics.
influenced by other nodes. The higher a
First, it permits pedestrians and cyclists from
node’s closeness, the more central it is in rela-
the wider area to access facilities, goods and
tion to all other nodes within the part of the
services within a neighbourhood centre. This
network included in the centrality analysis.
can be described as to-movement. Second,
Mathematically, we denote dij as the length
pedestrians and cyclists can travel through
of the shortest path from i to j. The mean dis-
the neighbourhood directly, rather than
tance from i to j, averaged over all nodes j in
detouring outside the neighbourhood, to
the network, is li . Closeness Ci is defined as:
reach other communities beyond. This can
be described as through-movement. Many 1 n
legacy danwei estates in Chinese cities allow Ci = =P
li j dij
non-motorised access to the community’s
internal facilities, meaning that they are
permeable in terms of pedestrian to-move-
Betweenness: Measuring permeable
ment. However, where there are gatekeepers
and a well-organised property management through-movement
company, pedestrian through-travel may be For the through-movement component of
restricted. Danwei communities therefore permeability we use betweenness centrality.
can be classified as urban areas with mixed Betweenness measures the density of short-
and unstable permeability. est paths between all nodes in the network
Allowing to-movement and through- (within a given radius) passing through a
movement for walking and cycling are link. It measures an important idea regard-
two characteristics of a neighbourhood with ing through-movement in network analysis,
good permeability. In this study we wanted that of the flow of information or other traf-
to measure permeability after the hypotheti- fic and of the influence that vertices might
cal removal of gated community barriers, as have over that flow. Nodes with highest
demanded by central government policy, betweenness are the ones most likely to be
and to compare it with permeability with the the location for activities necessitating
Sun et al. 2591

communications between a large number of to represent cycling distance and the regional
people and businesses because they lie on scale for our analysis (Zacharias, 2005).
the largest number of paths taken by traffic
flows, urban goods and information. These
places have the potential to attract more Construction of an urban
traffic and more people onto the street, and pedestrian network model
thus predict the location of commercial cen-
tres of varying scales. The radius at which Road centrelines tend to be used as a proxy
betweenness is measured can indicate the for pedestrian routes in walking studies, jus-
type of traffic likely to pass through a node tified on the grounds that most footpaths are
or link with a high betweenness value. pavements along roads (at least in North
Measured within a walking radius of 400 America and Australia). However, road net-
metres, for example, a high betweenness works often do not, in fact, reflect the finer-
value can be taken as an indication of high grained paths governing non-motorised
pedestrian through movement and is likely access (Iacono et al., 2010). A study in Perth,
to be correlated with locally-oriented com- Australia, dramatically found that measures
mercial land uses. of connectivity for some neighbourhoods
Mathematically, the betweenness central- improved by up to 120% when using pedes-
ity of a node (place) i is defined as the num- trian networks compared with using road
ber of paths that pass through i. Let nixy be 1 networks (Chin et al., 2008). In a study of
if vertex i lies on the shorted path from x to Davis, California, the pedestrian network
y and 0 if it does not or if there is no such within a five-minute walk of households was
path. Then the betweenness centrality Bi is underestimated by as much as 40% when
given by: using street network as a proxy (Tal and
Handy, 2012).
X There are few studies of pedestrian net-
Bi = nixy
xy works in Asian cities, where urban design
principles and density are strikingly different
from western counterparts. The typically
Geographic extent of permeability for missing pedestrian network elements from
official road network maps in China include
walking and cycling the following: (a) Small alleys, which pro-
As implied above, we need a reference point vide fine-grained permeability in the built
for empirical studies of behaviour to inter- environment and enhance non-motorised
pret permeability measured by network access; (b) Informal footpaths, which are
metrics. Following published studies, we use routes not official or legally sanctioned but
400 metres to represent an easy walking frequently used by local residents; and (c)
radius and the local scale for our analysis. Open accesses, which are those gates and
We use 1600 metres (around 1 mile) as the pedestrian paths in danwei compounds
longest walking distance. Empirical studies in familiar to local residents but unfamiliar to
Chinese cities have found that people would strangers. In addition, the large number of
walk longer if the walking environment was gated estates means that a significant part of
supportive (Sun et al., 2016). 1600 metres is the pedestrian network of a Chinese city is
also an easy urban cycling distance. Finally, not included in the road network. Therefore,
5000 metres is regarded as the regular toler- the official road centreline map would
able cycling distance in China and we use it clearly be inappropriate for our current
2592 Urban Studies 55(12)

study as gated-ungated differences would the before-after performance results, we

not be captured. sequentially sort the gated communities to dis-
Using high resolution aerial photographs, tribute the ungating intervention across the
street views and field visits with GPS ground city in a way that will create maximum gains
truthing techniques, we constructed a in permeability with minimum destruction of
bespoke pedestrian network of Nanchang existing property rights.
city, covering an area of 350 km2, a linear
network of 5572 km and comprising 85,857
road segments (links between junctions). Data and materials
The object was to create what we believe to Nanchang is a medium-sized Chinese city
be the first city-wide pedestrian network and the capital of Jiangxi province. In 2014,
map for a Chinese city and to use it to con- Nanchang’s urban population was 2.4 mil-
duct the comparison of permeability and lion and is projected to become 5.3 million
accessibility with and without gates. This by 2030 (Nanchang Statistics Bureau, 2015).
entailed mapping a progressive three-tier The legacy of danwei is well preserved in the
pedestrian network: (i) pedestrian network city, with large sized compounds built by for-
along the official road network, referred to mer and surviving state-owned enterprises.
in our discussion as the official pedestrian Private secured neighbourhoods have devel-
network; (ii) official pedestrian network plus oped apace in the city’s new development dis-
those missing elements listed above, referred tricts and suburbs. Nanchang is divided by the
to as gated pedestrian network; and (iii) Gan River. The traditional city area, located on
pedestrian network including that part of the eastern side of the river, has dense residen-
the network within closed gated commu- tial areas, commercial centres and public facili-
nities, which we refer to as the ungated ties. Development of the western side of the
pedestrian network. river started in 2000, under a new district
development plan, resulting in a mix of high-
Empirical study density office buildings, residential commu-
nities and a network of large avenues. The city
In the empirical study, we develop a metho- is experiencing rapid motorisation, bringing
dology that can be used for evaluating urban increasingly severe congestion to the city core
planning and design options, assessing the (Qu et al., 2014).
permeability of gated cities and conducting
‘what if?’ studies of the re-engineering pro-
posals. We use heuristic network analysis Distribution of gated communities in
metrics as surrogates for actual pedestrian Nanchang. To get a better understanding of
and motorised traffic flows, making this a how the city is segregated by gated com-
highly accessible technique suitable for rapid pounds, we firstly identified all gated com-
design appraisal and ongoing monitoring munities, including danweis and private
(Hajrasouliha and Yin, 2015). Our perfor- secured residential compounds. We used
mance measures capture two essential fea- Chinese map source services, including
tures of the city with and without gates: (a) Baidu (map.baidu.com), Gaode (gaode.com)
the network’s ability to make urban places and Tengxun (map.qq.com), to identify and
accessible for non-motorised travel (closeness); map gated communities. We also used high
and (b) the network’s ability to sustain the resolution aerial imagery and street views of
vitality of streets by allowing through-travel of the map sources, and complemented these
pedestrians and cyclists (betweenness). Using with field visits.
Sun et al. 2593

Figure 1. Distribution of the gated residential communities in Nanchang, China.

Note: The unmarked built up areas are occupied by industrial, commercial, institutional, company and public land uses.

Usually, Danwei compounds can be iden- gated compounds are recognisable from aer-
tified by their legacy names, such as a shor- ial imageries by distinctive geometries of
tened company name. For example, the physical layouts.
‘Jiangfang Residential Five’ is associated Using these methods, we identified all the
with the Jiangxi Textile Factory and ‘Five’ is gated residential communities in the city. In
one of a large area of numbered residential total, there are 159 remaining danwei com-
areas for this danwei. The Hongdu of munities and 417 privately secured gated
‘Hongdu Residential Second’ was built for communities. On average, contemporary
workers in the Jiangxi Hongdu Aviation gated compounds are over twice the area of
Industry Group Corporation Limited. Other danwei in geographic size (7.13 versus 3.27
strategies for mapping gated danwei estates hectares). The territorial and service scale of
included identifying danwei hospitals that contemporary gated compounds varies proj-
are still functioning (hospital services like ect by project, from just two or three apart-
most other urban services were provided by ment buildings in a small plot aimed at
large work units) and primary schools that specialised end-users, to a large residential
are still using the name of the danwei that district of over 130 hectares, comprising
built them for children of their workers. On high-, mid- and low-rise building types and
the other hand, contemporary gated com- targeting clients of mixed income levels.
pounds can usually be identified by their Figure 1 shows the distribution of
western style names using modern terms, for Nanchang’s gated communities. Danwei
instance: Honggu Triumphant Xiaoqu, estates are mostly in the traditional city
Weidong Garden Xiaoqu and Fenghe New centre since most were built before the mid-
Town; where Honggu, Weidong and Fenghe 1990s. A few are close to the present sub-
are location names of districts, former vil- urbs, such as Hongdu residential areas and
lages or large urban avenues. Regardless of Jiangfang residential areas which were
the names, both danwei and contemporary located on the suburb’s fringe before recent
2594 Urban Studies 55(12)

urban expansion. The danweis close to the network: gated pedestrian network and
suburbs generally remain intact, with very ungated pedestrian network. The two non-
few redevelopments, and are usually fully motorised travel standards of 400- and
open, without gates or walls and with full 5000-metre radius as local and regional
non-filtered permeability. In the city centre scales are applied for the ungated/gated sce-
danwei estates retain minimal gates and well- narios of the whole city separately. Thirdly,
preserved walls. Privately secured commu- we examine the change in betweenness and
nities dominate the new development district closeness when gates are ‘opened’. Pearson’s
and most of the suburbs. Most contempo- correlations, in quintiles of these changes,
rary gated communities were initially located are calculated. We identify the top 20 per-
on newly developing fringe areas, but many centage of road segments in terms of change
have become integrated within the city centre in betweenness and closeness and make a
due to radical expansion of the urban area. line density analysis, weighted by between-
ness or closeness change. We then identify
Three-tier pedestrian networks. As described the gated communities whose opening will
above, we constructed a three-tier pedestrian potentially have the minimum destruction
network of Nanchang. Using network seg- but approach the maximum effects.
ments (links) as a measure of network size, We used GIS (ArcGIS 10.3, ESRI,
the ungated pedestrian network is 18.3 times Redlands, California, USA) to calculate and
the size of the official pedestrian network (N visualise the effects of making the city
= 85,857 compared to N = 4699). Using permeable for walking and cycling. sDNA, a
road intersections, a measure of connectiv- plugin for GIS software, was used to calcu-
ity, the ungated pedestrian network is 16.8 late betweenness and closeness measures of
times the size of the official network. The connectedness. sDNA has been validated in
mean length of network segments in the offi- the literature and in empirical studies of
cial pedestrian network is 305 metres, com- non-motorised travel behaviour (Cooper
pared to 65 metres in the ungated pedestrian et al., 2016).
network. If Nanchang’s gates were opened,
there would be a 138% increase in total
pedestrian network length, with an extra Results
31,731 road segments and 17,974 road inter- Permeability measured by betweenness
sections, and average link size would reduce
and closeness
by 79%.
We use a large residential development in
the suburbs as our first case to illustrate
Analysis changes in closeness with and without gated
We deepen this initial analysis by comparing compounds. The distance from the area to
the networks using systemic performance the city centre is around 12 km. It contains
measurements associated with network func- 11 adjacent residential communities, of
tionality. We firstly calculate closeness and between 30 to 80 hectares each. Green
betweenness measures for long-walking but Mountain Villa, for example, is around 80
easy bicycle distance, using a 1600-metre hectares, while the Nanchang Residential
radius. From the results, we select two areas ‘Theme Park’ (juzhu zhutigongyuan) has three
for a detailed comparison with and without sub-communities each of around 33 hectares.
gates. Secondly, we calculate betweenness There are very few bus lines from this large
and closeness based on the two tiers of residential compound connecting to the job
Sun et al. 2595

Figure 2. Closeness and betweenness measures for the permeability: A comparison before and after
open the gates for pedestrians and cyclists in two selected study sites.

centres of the city. Residents heavily rely on a planning objective is to diversify the city
cars. There are few services and few work- and make use of the public benefits of streets
places. The large compound units are gated. in distributing economic choice and agglom-
Navigating to destinations within this large eration benefits, then these two under-used
residential area mostly requires travelling parts of the city’s street network would be
along arterials and community walls, before good candidates for the zoning of services
entering the gates for local navigation. centres and workplaces.
Figure 2-C1 shows the current pattern of We use a second case, located five min-
connectedness (closeness) and Figure 2-C2 the utes’ walking distance from the city centre,
pattern after removing the gates to make the to illustrate changes in betweenness with and
neighbourhood permeable for non-motorised without gates. This comprises large parcels
travels. There is a striking difference, with a of land with a mixture of legacy danwei and
denser and more connected pedestrian net- contemporary residential developments. It
work after the gates have been opened and includes danwei such as the Nanchang
with many more streets having high accessibil- Diesel Engine Factory Residential Areas and
ity values. From this analysis, it is likely that the Jiangxi Xinhua Print House Residential
at least two new community centres would Areas, and contemporary gated compounds
emerge without the gates, shown by the streets such as the Hengmao International Grand
with a much redder colour in Figure 2-C2. If Garden. It is striking for its juxtaposition of
2596 Urban Studies 55(12)

Table 1. Pearson’s correlation of changes in gated/ungated in quintiles (N road segment = 53,411).

Pearson’s correlation coefficients

Betweenness (400m): gated/ungated

5th percentiles 0.91
4th percentiles 0.28
3rd percentiles 0.18
2nd percentiles 0.10
1st percentiles 0.13
Betweenness (5000m): gated/ungated
5th percentiles 0.92
4th percentiles 0.28
3rd percentiles 0.09
2nd percentiles 0.04
1st percentiles 0.10
Closeness (400m): gated/ungated
5th percentiles 0.55
4th percentiles 0.32
3rd percentiles 0.31
2nd percentiles 0.34
1st percentiles 0.71
Closeness (5000m): gated/ungated
5th percentiles 0.88
4th percentiles 0.30
3rd percentiles 0.24
2nd percentiles 0.30
1st percentiles 0.47

gated commodity housing and danwei by non-motorised travel from roads that are
land redevelopment. currently suffering the most severe conges-
Betweenness was computed to measure tion in the city.
changes of through movement with and
without gates. Comparing Figure 2-B1 and
Figure 2-B2, the increase in potential for Changes in permeability after opening
through-travel is mapped. Several pedestrian gated communities
road segments start to play important roles Pearson’s correlation coefficients for a com-
in the simulated ungated city centre. One parison in quintiles between ungated and
way of interpreting these analytical maps is gated scenarios are shown in Table 1. We
that they quantify the amount of public found that the highest correlations of the
value that is removed from the public ungated/gated are in the top 20% percen-
domain as a result of gating. After ungating, tages (correlation coefficients are above
through-movement increases on nearly all 0.88, excluding closeness in 400 metres’
road segments in the neighbourhood, effec- radius). This means that the accessibility of
tively extending the city centre’s commer- places with high betweenness or closeness
cially viable area. As well as increasing the when gated is consistently enhanced by the
number of streets in the city centre with high opening of the gated communities. The other
potential commercial vitality, the figure percentiles of betweenness and closeness
shows why opening the gates would attract change inconsistently between ungated/gated
Sun et al. 2597

Figure 3. The changes in betweenness after opening the gated communities.

scenarios. The inconsistent effect of ungating betweenness scores based on the gated/
on these parts of the networks may be attrib- ungated pedestrian networks were produced
uted to variations in the size of the sur- at both the local and regional scale (Figure
rounding gated communities, the geographic 3). Using a 400-metre radius, we obtained
location of the road segments within the city betweenness of each road segment at the
grid (e.g. in the centre or on the fringe) and local scale. The red parts of the network are
the heterogeneity in pedestrian network lay- those that are well configured for walking.
outs within the calculation radius. As a The most highly permeable places post-gates
result of these morphological variations are mostly scattered in the traditional city
across gated communities, opening the gates area. The maps also reveal a few candidate
affects the urban grid connectivity and the local centres on the west bank of the river
associated positive and negative externalities and in the suburbs. These cluster in an origi-
to varying extents in different parts of the nally free-standing town before urban
city. This leads us to sort gated communities expansion, as well as in the southern part of
in order of the magnitude of the impact of the city, which is now a large residential
ungating (see next section). area. The maps identify candidate places
To visualise the city-wide effects of gating where new local centres could be located as
on urban centrality patterns, maps of a policy intervention to densify and diversify
 2598 Urban Studies 55(12)

Amount Average size (m2)

High gain in permeability (. above 80%)a Medium and above medium gains (. 40%) Low and above low gains walkable suburbs (Chiaradia et al., 2012). In
the second group of maps, with betweenness
measured within a 5000-metre radius, we see

100,426
28,188
26,116

36,517
30,323

44,132

93,286
23,174
a clear enlargement of the vitality potential
of the city following ungating. Notably,
greater pedestrian and cycling interaction
appears between the two riversides. The
121
109

174
133

109
importance of the three bridges is revealed.
24
24

50
Permeable bridges, only allowing pedes-
trians and cyclists but restricting cars, could
greatly enhance such connections.
Average size (m2)

How could we select gated communities

161,531

155,658

for ungating to maximise public benefits at

19,310
25,029

28,399
24,703

58,185

21,913

minimum private cost?

There are high political and potentially legal
and compensation costs in removing gates,
so it is prudent to ask which ungating inter-
Amount

ventions might yield the greatest benefits.

We generated a series of impact surfaces
44
46

84
99

31
11
6
9

using the top 20% percentage changes in

permeability (Chiaradia et al., 2012). The
calculations are made separately at the local
Average size (m2)
Table 2. A descriptive statistic of candidate communities for ungating.

(400-metre radius) and regional (5000-metre

radius) scales. We overlayed the surface with
the distribution maps of the gated commu-
Note: aThese high, medium and low ranges refer to the results in Figure 4.
447,739

250,961
14,694
23,132

31,707
25,137

nities and then sorted the gated communities

whose opening will bring the greatest perme-
-

ability enhancements to the city.

Figure 4 and Table 2 show the results,
indicating a partial ungating plan that opti-
Amount

mises public gains for a given private loss.

We found that the enhancements of
NA

NA
11

30

22
6

betweenness through ungating are predomi-

Contemporary gated compounds

Contemporary gated compounds

Contemporary gated compounds

Contemporary gated compounds

nantly located in the traditional city areas,

both at local and regional scales of analysis.
The gated communities within the darker
Betweenness (5000m)

blue surfaces are where most danwei legacy

Betweenness (400m)

Closeness (5000m)

estates are located. Our analytical prescrip-

Closeness (400m)
Danwei legacies

Danwei legacies

Danwei legacies

Danwei legacies

tion suggests that ungating danwei can be a

priority. The opening of gated danwei estates
might also have the advantage of facing less
resistance, since many danwei communities
allow mixed or unstable access, as already
described. As well as opening gates, physical
Sun et al. 2599

Figure 4. The gates to remove to maximise permeability gains at minimum social cost.

remodelling of these estates might focus on A combination of our maps provides use-
urban design improvements consistent with ful information for calibrating an ungating
adding the internal grids to the city-wide grid policy. The selected (high potential accessi-
to allow non-motorised through-travel. bility change) gated compounds could, for
Enhancements of closeness differ between example, be opened to non-motorised travel
local and regional scale radii. Closeness cen- at the same time as removing some of the
trality is difficult to improve at the local scale pedestrian/cycle through-routes from busier,
merely by removing gates where the length more polluted and more dangerous routes
of the gated communities’ walls is longer around the edges of gated large blocks. This
than 400 metres, which makes most destina- could also be accompanied by a number of
tions far away. At the regional scale, opening specific and fine-grained design implications
residential gated clubs in the new district and (Melia, 2012): a filtered permeable road net-
large residential areas on the city fringes work makes it easier to move by foot or on
would create more significant increases in bike instead of in a car; while non-filtered
permeability for non-motorised access and permeable roads allow cars, but with traffic
better access to destinations. calming techniques, can slow traffic speed.
2600 Urban Studies 55(12)

Discussion and conclusion new community centres; 2) promote the

vitality of the city; and 3) provide guidance
The release of new guidelines to open and
for an incremental ungating policy. We
prohibit gated communities is an interesting
explored intra-urban variations in perme-
twist in China’s urbanisation story. Now the
ability, asking how much overall variation
government would like to catastrophically
in permeability there is. We also explored
re-engineer the morphology that has evolved changes in permeability in ungated/gated
under the gated development regime of the scenarios, asking how it is distributed across
recent past. Whatever the private and politi- the city and which parts of the city can be
cal costs in terms of property rights appro- better improved by ungating. Further, we
priation, this is likely to have significant sequentially sorted the gated communities,
positive effects on spreading the urban local identifying those that will yield the biggest
vitality and commercial potential that comes changes for a given private cost if opened up
with pedestrian footfall and on health bene- – those associated with the top 20 percent of
fits, through a dramatic change in active enhancements in the city’s betweenness and
travel behaviour. closeness gains from ungating.
In this article, we explore the ‘what if?’ Our permeability analysis was based on a
question posed by this imminent natural detailed pedestrian network model. As far as
experiment. What if we open the gates? We we know, this is the first urban planning pol-
pose this as an intellectual thought experi- icy ‘what if?’ study using a pedestrian net-
ment, recognising the difficulties and dilem- work model for an entire city. The benefit of
mas in realising this policy. We do not take a this approach can be seen in the striking dif-
position for or against the opening of gates. ferences between the three tiers of the pedes-
We merely ask how the movement potential trian network, with an approximate 20-fold
of a heavily gated city would change if gates increase in road segments and intersections
were opened. We offer the analysis to urban in the ungated network compared to the
scholars, planners and Chinese governments official road network, which would be the
by way of a quantified simulation. typically used network for pedestrian-level
Specifically, we use a permeability per- analysis at the urban scale. We offer this as
spective to evaluate the through-movement a strong argument for building bespoke and
and to-movement improvements achieved accurate pedestrian network models as the
by ungating. Permeability is measured using basis for urban analysis of active travel
closeness and betweenness centrality mea- issues and policies.
sures of a city-wide pedestrian network The re-engineering of an urban form
model. Using an empirical study of dominated by gated communities faces mas-
Nanchang, these two measures captured sive challenges. It is not easy for current resi-
accessibility effects for pedestrians and dents to accept ungating if no detailed plans
cyclists, before and after the simulated accompany it, for instance to ensure security
ungating. This gives an indicator of the total and ongoing service provision plans at the
urban accessibility benefits, or pedestrian smaller neighbourhood or building scale.
route externalities foregone, as a result of Developers are likely to take a cautious ‘sit-
closed-gated compounds. Alternatively, it ting on the fence’ attitude with the policy,
can be taken as an indication of the social waiting for others to test out the govern-
costs of gating. ment’s negotiating stance. However, we note
We explored the potential of the city that a policy that proscribes gated projects
without gated compounds to: 1) bring forth provides a new kind of mechanism for
Sun et al. 2601

government planners to work more closely specific policy design guidance on which
with developers to achieve good planning. A gated estates to ungate. We offer a metho-
first result of the policy is likely to be the dology for Chinese planners to conduct
zoning of residential land in relatively ‘what if?’ evaluations as they approach this
smaller blocks than before. massive central policy directive. We have
The study has limitations and strengths. suggested how a subset of gated compounds
We did not separate out the pedestrian net- could be selected to maximise social gains
work from the cycle network in our analysis. for a given order of private costs. Finer,
The cycling mode-share in Nanchang is very more pragmatic and sequential plans can be
low and there are barely any cycling made on the basis of such strategic analysis.
facilities in the city (Nanchang News, 2010). Assuming the ungating policy is widely
Nonetheless, cycle route planning and the implemented, we plan to measure the actual
placement of cycling facilities could be effects of this massive natural experiment,
guided by our models with little additional including economic, health and traffic effects
effort. We used centrelines to represent the over time.
pedestrian network since this was sufficient
for our permeability analysis, based as it is Declaration of conflicting interests
on systemic connectivity. A finer pedestrian
The author(s) declared no potential conflicts of
network could comprise pavements on both
interest with respect to the research, authorship,
sides of a motorised road. We applied only a
and/or publication of this article.
crude cost-benefit analysis to the normative
part of this study. On the benefit side, we
could have calculated net benefits by taking Funding
into account the reduction in monetary value This study is supported by NSFC (41561035) and
of a residential location as a result of ungat- the 35th round of PDF/RAP Fellowship from
ing, using hedonic modelling. Our model is a University of Hong Kong.
simplification of reality, but we suggest that
it presents a powerful and policy-relevant References
analysis of the change in distribution of the Alfonzo M, Guo Z, Lin L, et al. (2014) Walking,
city’s accessibility as a result of gates being obesity and urban design in Chinese neighbor-
taken down. The study is limited by having hoods. Preventive Medicine 69(Suppl 1):
been conducted in only one city. However, S79–85.
we are confident that our methodology can Chen T, Hui ECM, Lang W, et al. (2016) People,
be reliably adapted to other cities across recreational facility and physical activity:
China and elsewhere, with additional consid- New-type urbanization planning for the
erations, if needed, for local context. healthy communities in China. Habitat Inter-
The study has several notable strengths. national 58: 12–22.
It is the first study using a quantified simula- Chiaradia A, Schwander C, Hillier B, et al. (2012)
Compositional and urban form effects on cen-
tion to test such a dramatic government pol-
tres in Greater London. ice | proceedings
icy reversal on gating. By quantifying the 165(DP1): 21–42.
effects of opening gated compounds in pur- Chin GKW, Van Niel KP, Giles-Corti B, et al.
suit of less congested and more accessible, (2008) Accessibility and connectivity in physi-
healthier urban form in China, we both give cal activity studies: The impact of missing
scientific justification to the policy (on the pedestrian data. Preventive Medicine 46(1):
benefits side) and at the same time provide 41–45.
2602 Urban Studies 55(12)

Cooper C, Chiaradia A and Webster C (2016) Newman M (2010) Networks: An Introduction.

Spatial Design Network Analysis( + ). Cardiff: Oxford: Oxford University Press.
Cardiff University. Normile D (2016) China rethinks cities. Science
Day K, Alfonzo M, Chen Y, et al. (2013) Over- 352(6288): 916–918.
weight, obesity, and inactivity and urban Qu L, Gao Y, Salzberg A, et al. (2014) Walkabil-
design in rapidly growing Chinese cities. ity analysis for Chinese transit oriented devel-
Health & Place 21: 29–38. opment: A case study in Nanchang. In:
Glasze G, Webster C and Frantz K (2004) Private Transportation Research Board 93rd Annual
Cities: Global and Local Perspectives. London: Meeting, Washington, DC, 12–16 January.
Routledge. Washington, DC: TRB 2014, pp. 1–15.
Hajrasouliha A and Yin L (2015) The impact of Scott A (2002) Global City-Regions: Trends, The-
street network connectivity on pedestrian vol- ory, Policy. Oxford: Oxford University Press.
ume. Urban Studies 52(13): 2483–2497. Sun G, Zacharias J, Ma B, et al. (2016) How do
Iacono M, Krizek KJ and El-Geneidy A (2010) metro stations integrate with walking environ-
Measuring non-motorized accessibility: Issues, ments? Results from walking access within
alternatives, and execution. Journal of Trans- three types of built environment in Beijing.
port Geography 18(1): 133–140. Cities 56: 91–98.
Le Goix R (2005) Gated communities: Sprawl Tal G and Handy S (2012) Measuring nonmotor-
and social segregation in Southern California. ized accessibility and connectivity in a robust
Housing Studies 20(2): 323–343. pedestrian network. Transportation Research
Lee I-M, Shiroma EJ, Lobelo F, et al. (2012) Record: Journal of the Transportation Research
Effect of physical inactivity on major non- Board 2299: 48–56.
communicable diseases worldwide: An analy- Talen E (2002) Pedestrian access as a measure of
sis of burden of disease and life expectancy. urban quality. Planning Practice and Research
Lancet 380(9838): 219–229. 17(3): 257–278.
Low SM (2001) The edge and the center: Gated Talen E (2013) Charter of the New Urbanism. 2nd
communities and the discourse of urban fear. edn. New York: McGraw-Hill Education.
American Anthropologist 103(1): 45–58. Tao W, Mehndiratta S and Deakin E (2010)
Lu D (2006) Travelling urban form: The neigh- Compulsory convenience? How large arterials
bourhood unit in China. Planning Perspectives and land use affect midblock crossing in
21(4): 369–392. Fushun, China. The Journal of Transport and
Lynch K (1981) Good City Form. Cambridge, Land Use 3(3): 61–82.
MA: MIT Press. Webster C (2001) Gated cities of tomorrow. Town
Manzi T and Smith-Bowers B (2005) Gated com- Planning Review 72(2): 149–170.
munities as club goods: Segregation or social Webster C, Ruan J and Sun G (2018) The
cohesion? Housing Studies 20(2): 345–359. Global Spread of Gated Communities. In:
Melia S (2012) Filtered and unfiltered permeability: Watanabe Y (ed.) Handbook of Cultural
The European and Anglo-Saxon approaches. Security. Cheltenham: Edward Elgar.
Project 4: 6–9. Wu F (2005) Rediscovering the ‘gate’ under mar-
Miao P (2003) Deserted streets in a jammed town: ket transition: From work-unit compounds to
The gated community in Chinese cities and its commodity housing enclaves. Housing Studies
solution. Journal of Urban Design 8(1): 45–66. 20(2): 235–254.
Nanchang News (2010) Push out cycling in Xu M and Yang Z (2009) Design history of Chi-
Nanchang. Available at: http://www.ncnews na’s gated cities and neighbourhoods: Proto-
.com.cn/xwzx/yw/t20100917_623592.htm (accessed type and evolution. Urban Design International
1 March 2016). 14(2): 99–117.
Nanchang Statistics Bureau (2015) Nanchang Sta- Zacharias J (2005) Non-motorized transportation
tistical Yearbook. Beijing: China Statistics in four Shanghai districts. International Plan-
Press. ning Studies 10(3–4): 323–340.