Social and spatial differentiation of high and low income groups’ out-of-home activities in Guangzhou, China

Please download to get full document.

View again

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
  With China’s rapid urbanisation driving its growing economy, the enlarging socio-spatial inequalities in the cities have received wide attention. Rather than following the largely studied residential spaces, this paper focuses on socio-spatial
  Social and spatial differentiation of high and low income groups’out-of-home activities in Guangzhou, China Suhong Zhou a,b, ⇑ , Lifang Deng c , Mei-Po Kwan d , Ruogu Yan e a School of Geography and Planning, Sun Yat-sen University, Guangzhou, China b Guangdong Key Laboratory for Urbanization and Geo-simulation, Guangzhou, China c The Urban Planning Institute of Foshan Chancheng, Foshan, China d The University of Illinois at Urbana-Champaign, Urbana, IL, USA e Guangdong Academy of Social Sciences, Guangzhou, China a r t i c l e i n f o  Article history: Received 16 June 2014Received in revised form 1 March 2015Accepted 1 March 2015 Keywords: Spatiotemporal behaviourSocial diversityChina a b s t r a c t With China’s rapid urbanisation driving its growing economy, the enlarging socio-spatial inequalities inthe cities have received wide attention. Rather than following the largely studied residential spaces, thispaper focuses on socio-spatial differentiation based on the spaces of one’s out-of-home activities. Usingdata of 1006 individuals collected by door to door questionnaires, this paper sets up the spatial andtemporal autocorrelation GT coefficient to examine the spatial heterogeneity characteristics of high-and low income groups’ out-of-home activities in a continuous spatiotemporal framework. The factorsand different mechanisms influencing the clustering of the activities are discussed to better understandsocialdiversityinpost-reformurbanChina. Theresultssuggestthatthereisobviousspatialandtemporalvariation in high- and low income groups’ out-of-home activities, indicating that differing social spacesare not just limited to the macro-static residence-based living space, but also exist in the individual’sdaily-activities space. Both high- and low income people have drastically different activity spaces andthey may not interact much with each other. This is socially very significant because it means that thereisconsiderablesocial isolationor segregationforbothgroups. Theresultsalsoshowthatwithinthesameincome group there exists a divisive cluster with different formation mechanisms, including the job–housing relationship, the correlation of activity opportunities with those surrounding residential areasand the individual’s ability to access activities (that is, space–time accessibility). Structural transitioncan also impact on activities choices of various social groups.   2015 Elsevier Ltd. All rights reserved. Introduction China’s urbanisation level rose from 17.9% in 1978 to 49.68% in2010 with a 3.24% annual growth rate (National Bureau of Statistics of China, 2008, 2011). This represents an unprecedentedspeed and scale of urbanisation. Based on GDP data in 2010, Chinahasbecometheworld’ssecondlargesteconomywithanimportantrole in the world economic system. Whist rapid economic growthhas considerable potential to improve people’s living standards,income inequality in China is becoming increasingly serious.Various statistical sources indicate that China’s Gini coefficientwas 41.5 in 2011, which exceeded the internationally acceptablewarning level of 40. The quintile income ratio is now 8.4, whichmeansthattheaverageincomeoftherichest20%ofthepopulationis 8.4 times higher than the average income of the poorest 20% of the population (United Nations Development Programme, 2013).Chinahasbecomeoneoftheworld’scountrieswithseriousincomeinequality, which indicates that the benefits of China’s rapideconomic growth have not been equally shared by different levelsof society. As social difference in China becomes increasingly ser-ious (Gu & Christian, 1997; Yang, 2005), concern for social justicehas attracted much attention and discussion.It is widely recognised that post-fordist cities are often charac-terisedbysocio-spatial polarisation(Li, Wu, &Lu, 2004;Marcuse&Kempen, 2000). However, many post-socialist cities are also facingincreasing socio-spatial divisions (Ruoppila, 2005). As China hastransformed to a socialist market system in the last three decadesor so, some new socio-spatial phenomena have emerged in thecountry (Ma, 2002). These include new gated communities (Wu, 2005), peasant enclaves, and  chengzhongcun  (villages in the city)for migrants (Ma & Xiang, 1998; Zhang, Zhao, & Tian, 2003).   2015 Elsevier Ltd. All rights reserved. ⇑ Corresponding author at: 135 Xingang Xi Road, School of Geography andPlanning, Sun Yat-sen University, Guangzhou 510275, China. Tel.: +8613825044799. E-mail address: (S. Zhou).Cities 45 (2015) 81–90 Contents lists available at ScienceDirect Cities journal homepage:  Many studies have examined the socio-spatial differentiation of post-reform Chinese cities and observed that post-reform urbanChina is characterised by residential segregation (Feng & Zhou,2008;Gu&Liu,2001;Li&Wu,2008;Wu,2002).Whilstitisimpor-tanttoexaminesocialdifferentiationinresidentialspaces,itisalsonecessary to investigate whether social segregation happens inother places where people perform their out-of-home activitiessuch as work, shop, and play (Kwan, 2013). This broadened focuswill help us better understand socio-spatial differentiation andthe complex relationships between social justice and spatialstructure.Individual behaviour has spatiotemporal characteristics.Examiningindividual spatial behaviour at themicro-level providesan important perspective and method for urban research. It ana-lyzes the interaction between individual behaviour and urbanspace by considering how individual characteristics and demandsimpact on the activities system of a whole city (Chai, Liu, & Li,2002; Zhou & Deng, 2010). Current research focuses on a series of particular behaviours, such as commuter and consumer beha-viour (Giuliano & Small, 1993; Kwan, 1999; Vandersmissen,Villeneuve, & Thériault, 2003; Wachs et al., 1993; Shen, Kwan, &Chai, 2013); time allocation among different activities (Huff &Hanson, 1986; Pas & Sundar, 1995); the time allocation andchanges between family members (Bhat, 1996; Golob & McNally,1997; Mannering, Murakami, & Kim, 1994; Niemeier & Morita,1996); multi-purpose travelling (Krizek, 2003; Nishii & Kondo,1992); decision-making simulation (Kitamura, 1984); and the rea- lisation of 3D geographic visualisation (Forer, 1998; Huisman &Forer,1998;Kwan,2000,2004;Yu,2006).However,muchresearchto date has used panel data for analysis, and studies that take intoaccount individual behaviours spatiotemporally in a continuousspace–time framework are still very limited.Spatial behaviour and the associated activity spaces thatunfold in a continuous space–time framework are important per-spectives that reveal the quality-of-life and social equity issues inurban areas. Individual activity spaces are determined by threeimportant determinants, such as home, regular activities, and tra-vel between and around the pegs (Golledge & Stimson, 1997:279). Thus, it consists of the visited locations, and the routesand areas one has travelled through (Schönfelder & Axhausen,2003). The location and status of residential areas, and the fre-quently visited activity locations such as work or shopping wouldbe important ‘‘anchors” of daily activity spaces. In urban China,the location and status of these ‘‘anchors” are deeply impactedby historical path dependence and urban social and economictransformation.The widening social and income inequality in China mentionedabove also manifests itself in the spatiotemporal behaviour of peo-ple (Shen et al., 2013). Recent research has found that poor peoplemay work longer hours (thus facing more space–time constraints)and have lower mobility (because of a lack of access to privatevehicles). Thus they are less free to move around and access vari-ous urban opportunities, such as shops and social and recreationalfacilities(Zhou&Deng,2010).However,howwouldsocialinequal-ity affect people’s spatiotemporal activities? What are the charac-teristics of the out-of-home activities clusters of different socialgroups? Why has the cluster formed? This paper will use casestudies to examine these issues.Income inequality intensified social differentiation during thepost-reform era in China. Income is an important index for socialstratification, where groups with different incomes can be usedfor socialclassification. Studies of howthe dailyactivities of differ-ent income groups form different space–time clusters will help toreveal their social spatiotemporal segregation from a space–timeperspective. This study uses spatial and temporal autocorrelationto examine the sociospatial differentiation between the activityclusters of high- and low-income groups in a contiguous spa-tiotemporal framework. It presents a case study of Guangzhoubased on a survey of a sample of residents’ behaviour to constructa spatiotemporal autocorrelation GT coefficient. It seeks to revealthe underlying mechanisms and factors contributing to such dif-ferentiation. Results of this research will be useful for addressingsociospatial equity issues and also optimising the allocation of urban resources. Historical path dependence and socio-spatial differentiation of daily activities in urban China As mentioned above, the location and status of activity‘‘anchors”, mostly residential areas and workplaces, are deeplyimpacted by historical path dependence in urban China. Thebuilt environment formed in the socialist period such as  danwei and in the post socialist era such as commercial and socialhousing and new industrial area, have some impact on currentsocio-spatial differentiation and thus impact on personal dailyactivities.Fordecades, danwei wasnotonlythebasicunitofeconomicandsocial organisation, but also the spatial organiser in urban China(Bjorklund, 1986; Bray, 2005). According to Bjorklund (1986),  dan-wei  is a spatial framework in which social life, economic activity,and political control are integrated. Apart from offering a job, danwei  provides its employees with a comprehensive package of welfare services for daily life, which deeply impact on not onlyresidential but also people’s out-of-home activity choices.AftertherevolutioninChinain1949, especiallyafterthesocial-ist reform of the publicly-owned system, almost all properties andproduction were organised by  danwei , with  danwei  responsible forsettingupeconomicandsocialunits. Danwei  madedecisionsaboutthe daily needs of individual households and became an ‘invisiblehand’ which organised people’s daily activities. However, thehistoric path of the  danwei  system still plays its role in the currenturban structure and thus on people’s out-of-home activities,especially for the employees of   danwei .In the socialist era, the  danwei  compound, a mix of housing,workplaces, and daily services was one of the main working andliving units. People living in a compound mostly commuted bywalking, resulting in mixed land use in ‘walking’ neighbourhoods.At the same time, with the development of the cities, and the lackof large pieces of land for compound construction, more and more danwei  constructedtheirresidentialhousingoutofthecompoundsin areas called Danwei Residential Areas ( 单 位 生 活 区 ). Although itresultedinsomegeographicalseparationofjobs,housingandotherservices, the optimal strategy of the jobs-housing connectionmostly came from the  danwei , which tried to make this separationas small as possible. Afterthe reformof the late1990s, bothlabourand housing became more and more mobile and individual house-holds had more decision-making autonomy. However, the  danwei system still plays its role in the current urban structure and dailyactivities.Instead of the much more socially diverse market system,social status in the  danwei  system is much less differentiated.Through the  danwei  compound providing jobs, housing and dailyservices, the development of mixed land use in the city is stimu-lated. The Danwei Residential Areas, however, encouraged someworkplaces to relocate near to the large residential areas, whichin turn may stimulate the mix of land use development aroundthe living places. With the reform of the  danwei  system, somegovernment-financed institutions and government offices, whoseemployees are mostly professionals and highly educated, left andthe compounds of these  danwei  became one of the high-incomegroups’ daily activity clustering zones. At the same time, some 82  S. Zhou et al./Cities 45 (2015) 81–90  state-owned companies closed during this reform, and left theirworkers with no jobs and low incomes with subsequent lowmobility in the places where some Danwei Residential Areaswere located.Thus, the much less socially differentiated status of people in danwei  and the location of traditional  danwei  compounds andDanwei Residential Areas carry some impacts on residents’out-of-home activity even today, and become the activity cluster-ing zones with high-income or low-income groups.On the other hand, with the market-oriented reformation from1978, urban China has faced social and spatial reorganisation. Thetransformation and increasing role played by the market such asthediversificationofjobopportunities,thedevelopmentoftherealestate market, and the improvement of household income haveimpacted greatly on both people’s daily activities and thus theurbanstructure.Themarkethasbecomeoneofthemostimportantfactors determining resource distribution, with opportunities dis-tributed more according to individual ability. A greater diversityof high and low incomes together with different levels of mobilityformed, resulting in the diversity of out-of-home activities.Due to this dynamic system, some new high-end service andhigh-technology industrial areas attracted high-income earners.At thesame time, older places withpoorlivingandworkingcondi-tions became areas attracting low-income earners who mostlyengage in their out-of-home activities near their living placesbecause of time constraints and low levels of mobility.Under this dual system, planning in China still plays its role ininfluencingurbanstructureandimpactsonpeoples’dailyactivitiesthrough the housing supply and some new working places. Afterthe reform of the economic system, China set up a dual housingprovision system with subsidised social housing provided by localgovernment for low income households and commercial housingprovided by the market for middle- and high-income households(Wang & Murie, 2000). Because of the low level of mobility of itsresidents, some social housing areas became low-income groups’out-of-home activity places. At the same time, with the power of planning, some new workplaces with high standards were formedand attracted high-income earners to these areas for their out-of-home activities. Data and methods The study area for this research is Guangzhou, one of SouthChina’s major cities and the capital of Guangdong Province. It islocated on the Pearl River Delta adjacent to Hong Kong andMacao, the frontier region of China’s reform and open-door policy.Guangzhou has been one of the fastest growing cities duringChina’s development changes. Census data from 2010 show thatthe total population of Guangzhou was 12.70million comparedwith9.94millionin2000, showinganincreaseof 27.74%. The totalarea is 7434square kilometres, within which the urbanised areascomprise more than 900square kilometres. In recent years, a lotof infrastructure has been constructed. The average road area perperson has improved from 3.70m 2 in 1990 to 11.20m 2 in 2010,and eight underground rail lines have come into service inGuangzhou. However, traffic congestion is still a serious issue inthe city. It is obvious that traffic demand has increased faster thanthesupplyoftransportinfrastructure. Exploringthecharacteristicsof people’s daily behaviour will help to better understand thedynamics of traffic demand.The data used for this study come from a random sample of household questionnaires completed in Guangzhou between Mayand August in 2007. 11 Neighbourhoods approximately 1km 2 insize were chosen. They were selected from the central, transitionand marginal districts in Guangzhou (Fig. 1), after consideringthe history, location, and housing types in the districts, such as Fig. 1.  Location of study neighbourhoods in Guangzhou, China. S. Zhou et al./Cities 45 (2015) 81–90  83  traditional self-built housing, welfare housing,  danwei  com-pounds, mixed residential areas, commercial housing, and urbanvillages. 800 Families, comprising 1006 individuals in total werechosen randomly. One or two questionnaires were completedrandomly in each family, excluding children younger than schoolage and unemployed people. Excluding those who did not com-plete the travel dairies, 982 samples were valid. Data were col-lected on basic social attributes, spatial information from thelatest travel dairies on weekdays (travel purpose, time andmode), and so on. The places of out-of-home activities for eachindividual were mapped along with the attributes supplied bythe questionnaires.The spatial and temporal autocorrelation coefficient of GT wasset up to examine the spatial and temporal clustering of thehigh- and low-income groups’ activities. It was derived from thespatial autocorrelation coefficient that refers to the potentialinterdependence of the same variable among the observed datain a geographic area or region (Griffith, 1987). The spatial autocor-relationcoefficientis oneofthespatialstatisticalmethodsusedforassessing the vicinity correlation and dependence across spatiallocations. The integration of GIS and spatial statistics has led toimportant developments in spatial autocorrelation analysis, andis now widely used in many areas of research (e.g., see Basu &Thibodeau, 1998; Yu & Zhang, 2010). Cliff and Ord (1981) present an improved algorithm to calculate spatial autocorrelation indexthat can be used to test the statistical significance of spatialautocorrelation. They also define a local spatial autocorrelationindex that can indicate high/low value clusters. López and Chasco(2007)proposedaunivariatespatiotemporalautocorrelationindexand expanded the srcinal static single time-point focus to furtherreveal the effect of geographic correlation with dynamic changes.However, although many scholars have expanded spatial autocor-relation to a space–time model, they have adopted a given timeperiod (such as a day or year) as the time dimension to interceptthe time cross-section as the self-correlation, rather than con-sidering time as a continuous variable, which inevitably makesthe study of spatiotemporal autocorrelation asymmetric.Although adding the time variable to emphasise dynamic changes,it is only part of the interception time point, thus weakening theimpact of the temporal dimension.This study comprehensively improves upon these scholars’definition of spatiotemporal autocorrelation. The purpose of thiscase study is to modify the partial spatial autocorrelation coeffi-cients G * .Thetemporalautocorrelationcoefficient T   isdefinedwithreference to  G * , and added to  G * to build a spatial and temporalautocorrelation coefficient of GT. This study focuses on thespace–time clustering of the daily behaviour of groups withdifferent incomes. The average individual monthly income is theobserved variable. Three coefficients are defined as follows: Spatial partial autocorrelation coefficient G * Getis and Ord (1992) proposed the partial  G * coefficient as apartial indicator of spatial autocorrelation for detecting high-valueaggregation (‘high–high’ type) and low-value aggregation (‘low–low’ type). A significant positive  G  i  implies that large values of   x  j (values above the mean  x  j ) are within distance  d  of point  i . Asignificant negative  G  i  means that small values of   x  j  are withindistance  d  of point  i . Calculations were performed in ArcGIS withthe Hot Spot Analysis tool (Getis-Ord  G  i ) as follows: G  i  ¼ P n j ¼ 1 w ij  x  j    X  P n j ¼ 1 w ij S   ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi n P n j ¼ 1 w 2 ij  P n j ¼ 1 w ij   2   n  1 v uut ð 1 Þ  X   ¼ P n j ¼ 1  x  j n  ð 2 Þ S   ¼  ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiP n j ¼ 1  x 2  j n   ð  X  Þ 2 s   ð 3 Þ where  x  j  is the attribute value for the feature  j ;  w ij  is the spatialweight between features  i  and  j ; and  n  is equal to the total numberof features.  G  i  is simply a standard deviation. In our study,  n  is thetotal number of out-of-home activities’ places of sample residents;  x i  is the personal monthly income of sample residents’ activities;and  w ij  is the distance weight between the activity  i  and  j , thusreflecting the degree of mutual influence between individual beha-viours. We used the ‘inverse distance squared’ (IDS) to calculate  w ij inthis study. Asignificant positive  G  i  indicated spatial clustering of high income samples’ out-of-home activities, whilst a significantnegative  G  i  indicated spatial clustering of low income groupssamples’ activities’. Temporal autocorrelation coefficient T  The temporal autocorrelation coefficient  T  i  corresponds to thedefinition of spatial autocorrelation coefficient  G  i  (Formula (1)).Each activity has an event start time  t  i 1  and end time  t  i 2 , thus eachactivity point corresponds to a spatial coordinate  t  i 1 ,  t  i 2 ). The spa-tial autocorrelation coefficient  G * mainly measures the distancerepresenting the relevant local interaction between spatial units.Thesmallerdistancebetweentwofeatureswill havemoreinterac-tion between each other. This study defines time–distance  z   simi-larly with distance variables to measure the time similarity andsimultaneity between two activities of two people. The smallertime–distance  z   examines the overlap of two activities’ start timesand end times. We define time distance of activities  i  and  j  as  z  ij  ¼  ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð t   j 1    t  i 1 Þ 2 þ ð t   j 2    t  i 2 Þ 2 q   ð 4 Þ Referring to the local spatial autocorrelation coefficient of   G  i (Formula (1)), the time autocorrelation coefficient  T  i  is calculatedas follows: T  i  ¼ P n j ¼ 1 w Tij  x  j    X  P n j ¼ 1 w Tij S   ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi n P n j ¼ 1 w 2 Tij  P n j ¼ 1 w Tij   2   n  1 v uut ð 5 Þ Notethat w Tij  isthetimeweightbetweenactivity i  andactivity  j  –itequals  z  ij  here. The other variables are the same as for Formula (1). T  i  is used to measure the time clustering of activities for differentincome groups. A significant positive  T  i  represents the time cluster-ingof highincomegroups’ activities; asignificant negative T  i  repre-sents the time clustering of low income groups’ activities. Spatio and temporal autocorrelation coefficient GT  i Space and time are the two continuous dimensions with thesame influence, so the spatial autocorrelation coefficient partially G  i  and temporal-autocorrelation coefficient  T  i  should have equalweight in constructing the spatial and temporal autocorrelationcoefficient GT i . Both  G  i  and  T  i  are standardised to be  N  Gi  and  N  Ti ,andthespatiotemporalautocorrelationcoefficientGT i  iscalculatedas following: GT i  ¼  N  Gi   þ  N  Ti  ð 6 Þ Asignificant positiveGT i  indicatesspatial andtemporalcluster-ing of high income groups’ activities, whilst a significant negative 84  S. Zhou et al./Cities 45 (2015) 81–90  result indicates the spatial and temporal clustering of low incomegroups’ activities. Spatial and temporal patterns of the high- and low-incomegroups’ out-of-home activities Taking personal monthly income as the observed variable, thestandardised spatial autocorrelation coefficient  N  Gi  , the standard-ised temporal autocorrelation coefficients NT i , and the spatio andtemporal autocorrelation coefficients GT i  of each activity pointswere calculated and visualised in ArcMap and the Lorentz curveis plot to test the spatial unequal distribution of different groups(Figs. 2 and 3). These coefficients were divided into five levels bythe Jenks method. The first level with a significant positive indexrepresents the clustering of high income individuals’ out-of-homeactivities, and the 5th level with a significant negative indexrepresents the clustering of low income individuals’ out-of-homeactivities. The figures showthat the clustering of high andlowval-ues is much more significant with  N  Ti  indexes than  N  Gi  ones. Bothspatial and temporal dimensions are important for out-of-homeactivities and Fig. 3 is the result of GT i  indexes.TheresultofLorentzcurveinFig.3showedthatthedistributionof high incomes’ out-of-home activities is the most unequal,following with the low incomes’. Both high- and low-income peo-ple have drastically different out-of-home activity spaces and maynot interact much with each other. This is socially very significantbecause it means that there is considerable social isolation orsegregation for both groups.The whole sample included three categories of out-of-homeactivities, such as work, daily living (buying food and taking chil-dren to school) and recreation and shopping (entertainment, visit-ing friends and shopping). The distributions are different amongthem (Fig. 4). Work activities of the high-income groups are nearto the new city centre, some governmental administrative officedistricts, and some newly developed industrial areas. Work activi-ties of the low-income groups are mostly located in old city centreand some places near to some traditional communities, that is, thedegradedworker-communitiesandurbanvillagesindowntown,aswell as the affordable housing in the outer ring of the city centre.Daily living activities of the high-income groups are near to somework units and some communities. Daily living activities of thelow-income groups are mostly near to some traditional communi-ties. Recreation and shopping activity places for high incomegroups are near to the new city centre, whilst those of the low-in-come groups are near to the low-income housing communities. Spatial and temporal clustering of the high- and low-incomegroups’ out-of-home activities To examine the distribution pattern of the spatial and temporalclustering of high- and low-income groups’ out-of-home activities, Fig. 2.  The  N  Gi  and  N  Ti  index of out-of-home activities. Fig. 3.  The GT i  index and the Lorentz curve of out-of-home activities. S. Zhou et al./Cities 45 (2015) 81–90  85
Related Search
Similar documents
View more
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!