大数据与城市规划 (16).pdf

ArticleUrban Analytics andCity ScienceDoes block size matter?The impact of urbandesign on economicvitality for Chinese citiesAbstractThe influence of urban design on economic vitality has been analyzed by a number of researchers andis also a key focus of many planning/design theories.However,most quantitative studies are based onjust one city or a small set of cities,rather than a large number of cities that are representative of anentire country.With the increasing availability of new data,we aim to alleviate this gap by examiningthe impact of urban design upon economic vitality for the 286 largest cities in China by looking at agrid of geographical units that are 1km by 1km.We use these units and a set of new data(emergingbig data and new data that reflecting urban developments and human mobility)to look at the impactof urban form indicators,such as intersection density(urban design),level of mixed use,and accessto amenities and transportation,on economic vitality represented by activities using social mediadata.Our results show that these urban design indicators have a significant and positive relationshipwith levels of economic vitality for cities at every administrative level.The results contribute to aholistic understanding of how to improve economic vitality in cities across China at a detailed level,particularly at a time when Chinas economic growth will depend largely on growth of the servicesector in urban areas.We think these results can help decision makers,developers,and planners/designers to improve economic vitality in cities across China.KeywordsConsumption vitality,intersection density,block size,big/open data,ChinaIntroductionThe contribution of urban design principles to vibrant and prosperous cities,particularly itsimpact on urban vitality,1has become common belief.Pioneering urbanists like JacobsCorresponding author:Ying Long,Tsinghua University,Beijing 100084,China.Email:EPB:Urban Analytics and CityScience2019,Vol.46(3)406422!The Author(s)2017Article reuse guidelines: Lefebvre(1995)were among the first to specify the features of good urban design(e.g.small blocks)and how it cultivates community,sociocultural vibrancy,and healthyneighborhoods.Other influential urbanists such as Gehl(1971),Lynch(1981),Montgomery(1998),and Whyte(1980)also established new ground on ways to understand urbanvibrancy.Urbanists like Jacobs and Gehl,along with the New Urbanists,broadlypromoted pedestrian-friendly,compact,walkability,and mixed-use neighborhoods(Katzet al.,1994).Since the advent of these urbanists,there has been extensive and qualitativediscussion on how good urban design promotes the formation of lively city and urbanvitality.The variables that we quantitatively analyze in this paperintersection density,level of mixed use,access to amenities and transportationare aimed at using data toconfirm the beliefs of these great urban thinkers.This paper aims to quantitatively explore the relationship of urban design by looking atthe impact of variables such as intersection density,level of mixed use,and access toamenities and transportation,with economic vitality.We are particularly interested inexploring the answers to the following two questions:(1)Do urban design principlespositively associate with vibrant cities in terms of economy?How much is a variable suchas intersection density correlating with economic vitality compared to other factors?(2)Doesthis relationship vary across cities of different administrative levels?It should be noted that itis not easy to derive any causal relationship between the variables and economic vitality dueto data source limitation in this paper.We believe the correlation analysis in this paper is afirst step toward this final destination.This paper aims to fill a gap in existing studies by using granular data to analyze 286 ofthe largest cities in China.Of all the countries in the world,Chinas urbanization rate isamong the highest and its urban population is also among the largest.We use ordinary leastsquares(OLS)to look at the comparative impact of each urban design variable on economicvitality.We use social media comments,sign-ins,and housing price data as proxies foreconomic vitality.Our analysis includes spatial variables such as intersection density,level of mixed use,access to amenities,and access to transportation.Our results showthat intersection density has a significant and unique impact on economic vitality,evenwhen controlling for variables such as point of interest(POI)density and populationdensity.We also find that other spatial variables such as level of mixed use,access toamenities,and access to transportation also have a significant effect on economic vitality.This paper is organized as follows.The next section provides a review of the literature onthistopic,lookingatrelevantliteraturewhilealsodemonstratingtheemergingconsensus around the results of our analysis.Data section describes our dataset indetail and Results section presents our regression results.The final section concludeswith suggestions for future research.Literature reviewThe widespread use of new data sources,including mobile phone traces,publictransportation smartcard records,social media data,and geo-tagged data,has created thepossibility to understand how urban design impacts economic vitality(Zhou and Long,2016).These data sources allow researchers to have a more granular and human-scaleunderstanding of how people experience cities with increased accuracy,consistency,anddetail(Dunkel,2015).Our literature review shows that while urban form indicators havebeen widely analyzed,there have only been a handful of studies that look at the impact ofurban design on urban or economic vitality2through quantitative analysis.More detailedreviews are as follows.Long and Huang407We find several specific studies that connect urban design with urban(especiallyeconomic)vitality.One study(De Nadai et al.,2016)targets six cities in Italy and aims toconfirm four of Jane Jacobs principles for urban vitality using mobile phone data.Two ofthe principles overlap with our indicators(mixed use and intersection density)and the studyfinds that both of these characteristics have a significant effect on economic vitality.JaneJacobss principles are also examined by Sung et al.(2015),using the conventional householdtravel survey,by examining their role on pedestrian activity and evidence that they have apositive impact are found in Seoul,Korea.Similar studies supported by urban GIS andemerging new data have been conducted in individual Chinese cities,like Beijing by Longand Zhou(2016)using mobile phone data,Wu et al.(2016)using mobile phone positioningrecords,and Zheng et al.(2016)using catering establishment data,Shenzhen by Yue et al.(2017)using mobile phone data,and Shanghai by Shen and Karimi(2017)using housingprice records.Most of the aforementioned publications in this paragraph are for a single cityor a limited number of cities.For example,De Nadai et al.(2016)looked at six cities,ratherthan having an aggregate view of all the cities in a country.These aforementioned studies areall addressing the role of urban form on population density represented social vitality,withShen and Karimi(2017)and Zheng et al.(2016)as two exceptions,rather economic vitalitywhich is the focus of this paper.Furthermore,we also notice that extensive studies usingspace syntax,focusing on a single city as well,are discussing the relationship betweenstreet pattern indicators like integration/depth/choice and urban vitality like Van Nes andShi(2009).There are also a number of studies that show there is a positive impact of indicatorsrelated to urban design,such as walkability,on property values and other economicindicators.While data such as property values can provide some insight into the role thaturban design can influence a citys economy,social media data provide more granularityand more information on residents direct participation in the economy.Li et al.(2015)and Loehr(2013)show that walkability can positively impact the property values of aneighborhood,while other researchers have shown that proximity to highways has anegative impact on property values(Madison and Kovari,2013;Tajima,2003).While quantitative analysis that looks at the correlation between urban design andeconomic vitality(especially for numerous cities rather a single city)is relatively new,3studies that look at the relationship between the urban built environment and intensity oftravel behavior are relatively mature.For example,there is research that shows streetdesign parameters such as block size have a positive impact on pedestrian volume(Hesset al.,1999).In addition to the extensive studies on urban form and travel from Ewing andCervero(2010),there are a number of studies that show that pedestrian activity increaseswith mixed use,access to amenities,POI density,and intersection density(Krizek,2003;Sung et al.,2015;Sunga et al.,2015).This paper goes a step further to integrate pedestrianactivity with participation in commercial or economic activity,which is our view ofeconomic vitality.DataFor this analysis,we have used a number of data sources,including social media data,geo-tagged data,property value data,and government-released economic data for controlvariables.In this section,we will review each dataset.Table 1 offers more information onthe data we use in this paper.In Table 1,WEIBO represents Sina Weibo,a social mediaplatform that has check-in data,showing where users tag their location on the platform.Sina Weibo,the Chinese version of Twitter combined with Foursquare,is the leading online408EPB:Urban Analytics and City Science 46(3)microblogging platform in China,which allows users to publish,share,and discuss shortpostings on their website(http:/).4DIANPING represents DazhongDianping,an aggregated social media tool used to rate restaurant and other serviceindustry companies in China(Liu,2014).5Table 1.List of variables and data sources used in analysis.Main typeNameDescriptionData sourceVitalityDIANPINGDianping comments for eachsquare kilometer grid in 2014(#/ha)http:/,and used by Long(2016)WEIBOWeibo count of each squarekilometer grid in 2014(#/ha)http:/HOUSING_PRICEAverage housing price for eachsquare kilometer grid in 2014(CNY)http:/www.SDesignINTERSECTIONNumber of intersections for eachsquare kilometer grid in 2014(#/ha)Long(2016)DensityPOI_DENSITYPoint of interest(POI)count ofeach square kilometer grid in2014(#/km2)Long(2016)POP_DENSITYPopulation count of each gridderived from the township levelpopulation density of China in2010(#/km)Mao et al.(2016)DiversityMIXED-USEMixed-use level determined pergrid from data in 2014Long(2016)using themethod in Liu and Long(2016)AmenitiesAMENITIESThe total number of bus stops,education and researchfacilities,governmental facilities,and convenient stores in eachsquare kilometer grid in 2014(#/km2)Derived from POIs in2014,see Long(2016)Access toTransitACCESSIBILITYThe average air distance to theclosest city center,subcenter,green space,shopping center,hospital,and subway and HSRstation in 2014(km)Derived from POIs in2014 using SpatialAnalyst of ArcGIS,seeLong(2016)City levelcontrolvariablesGDPGDP per capita of the city in 2014(CNY)MOHURD(2015)TERTIARYPercentage of GDP that thetertiary sector accounted for in2014INCOMEAverage income of each person inthe city in 2014CITY_LEVELThe administrative level of the cityin 2014(five for themetropolitan cities and two forthe prefectural cities)See details in Geographicdata sectionLong and Huang409Geographic dataIn this paper,our data come from all the major cities in China and we exclude data outsidegovernment-defined urban limits.The Chinese central government has specific boundariesand labels that demarcate and categorize Chinese cities and the data from the governmentsstatistical yearbooks are based on this system.The government defines five mainadministrative levels of cities,including four municipalities(MD,considered Tier 1),15subprovincial cities(SPC,Tier 2),17 other provincial cities(OPCC,Tier 3),250prefecture-level cities(PLC,Tier 4),and 367 county-level cities(CLC,Tier 5).As of 2014,there were a total of 653 Chinese cities from all these administrative levels6(Figure 1).Tier 1cities are generally the largest and have the most mature economies.Beijing,Tianjin,Shanghai,and Chongqing are the four Tier 1 cities.Tier 2 cities are cities such asChengdu and Wuhan,while Tier 3 cities include cities such as Hefei and Changsha.In general,Tier 1 and Tier 2 cities are larger and have more developed economiescompared to Tier 3 and Tier 4 cities,with Tier 4 cities having the smallest economies.Theadministrative boundaries shown in Figure 1 were divided into 782,263 one square kilometerunits to represent the main units of analysis used in this paper.7Since not all of the one square kilometer areas located in the official administrative areasshown in Figure 1 are urbanized,we use data from Landsat TM images in the year 2010 byLiu et al.(2014)to identify the urbanized areas.In China,there were a total of 63,425km2ofFigure 1.Chinese cities demarcated by administrative area(the polygon in color stands for theadministrative area of each city).Source:Long(2016).410EPB:Urban Analytics and City Science 46(3)urban areas in 2010,as indicated by the data.We categorize any area that has more than0.5km2of urban land areas as an urbanized area.8As a result,we determine that thereare 43,646 geographic units of urbanized land in total within administrative areas ofChinese cities.Economic vitality dataTo evaluate the economic vitality of each area,we collect data from Dianping.For thisanalysis,the main data we use as a dependent variable are the quantity of comments onDianping from the users per geographic unit and we interpret this as a proxy for economicvitality.9The diversified comments range from the evaluation on the meal,the restaurantenvironment,and even the feelings with the friends dining together.We aim to use Dianpingcomments to proxy consuming vitality,one of the main components of economic vitality.Rather for all activities,Dianping is dominated by restaurants and shops.While keeping inmind the limitation of the dataset,there are major advantages to using Dianping in thisstudy.First,there is data on all of the Tier 14 cities that are used in our analysis.Second,there are no major competitors to Dianping,making it the most comprehensive andrepresentative data source for both foot traffic in addition to actual engagement withcommercial establishments within the geographic unit.In total,we were able to collectdata from 13 million POIs from Dianping,and there are 1.9million POIs that have atleast one user comment.There are a total of 47 million comments for all POIs onDianping.In this study,each establishment in Dianping is assigned to an urbangeographic unit,and then each unit is linked with anywhere from a few to hundreds ofestablishments.We found that the Dianping data are not representative