1、No 2007 22DecemberEconomic Geography, Spatial Dependence andIncome Inequality in China_Laura HeringSandra PoncetEconomic Geography, Spatial Dependence andIncome Inequality in ChinaLaura HeringSandra PoncetNo 2007 -22DecemberEconomic Geography, Spatial Dependence and Income Inequality in ChinaContent
2、s1 Introduction 82 Theoretical framework: geography and income level 102.1 Demand side . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.2 Supply side . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 The role of spatial dependence 134 Data and construction
3、of variables 144.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144.2 Construction of market access . . . . . . . . . . . . . . . . . . . . . . . . 154.3 Calculation of the spatially lagged income . . . . . . . . . . . . . . . . . . 175 Empirical estimation result
4、s 185.1 Benchmark estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185.2 The heterogeneous influence of market access . . . . . . . . . . . . . . . . 206 Conclusion 217 References 223CEPII, Working Paper No 2007-22ECONOMIC GEOGRAPHY, SPATIAL DEPENDENCE AND INCOME INEQUALITY INCHIN
5、ANON-TECHNICAL SUMMARYOver the last two decades, China has benefited from unprecedented income growth but itsdevelopment process has come with large and increasing income level differences withinChina (Meng et al., 2005). With regions with low per capita income predominantly locatedat the geographic
6、al periphery while those with the highest income locate at the center, acore-periphery structure is clearly emerging within China.Recent studies show that this evolution is coherent with predictions from the New Eco-nomic Geography (NEG) theory (De Sousa and Poncet, 2007) and confirm the validity of
7、the “NEG wage equation”(Hering and Poncet, 2006; Lin, 2005; Ma, 2006): locations closerto consumer markets (i.e. with a higher “market access”) experience lower transport costsand enjoy higher income levels (Fujita et al., 1999).But some critics can be addressed to these analyzes. First, these studi
8、es do not properlycontrol for differences in endowments, policies or institutions across locations since noneinclude location fixed effects. Another possible shortcoming of these studies is that they as-sume each location to be an isolated entity. But individual geographical units tend to be rela-ti
9、vely integrated due to migration, interregional trade, technology and knowledge spilloversas well as institutions (Buettner, 1999), what can lead to a spatial dependence between thelocations. Spatial dependence is acknowledged to be an important force in the process ofconvergence (Rey and Montouri,
10、1999) and ignoring this dimension in the estimation couldthus result in a serious misspecification (Abreu et al., 2005).The current paper contributes to a better understanding of the relationship between marketaccess and spatial inequality in China by exploring regional variations. We use a panel da
11、tasetcovering195Chinesecitiesbetween1995and2002, thatallowsustoincludefixedeffectsby city to control for scale economies and factor endowments.A final contribution of this paper consists in the analysis of whether the impact of marketaccess depends on province-level characteristics such as the inten
12、sity of immigration or theincome level.Our results suggest that access to sources of demand is indeed an important factor in shapingthe spatial income structure in China. While spatial relationships between Chinese cities alsomatter significantly they do not bias our estimates of the impact of marke
13、t access on income.As an attempt to determine whether labor supply under the form of internal immigrationplays a role in this dynamics we investigate whether the relationship between income andmarket access is valid only for locations with low or high immigration. According to ourtheoretical model,
14、in case of quasi-infinite labor supply for the manufacturing sector, wageswould respond only little to changes in the demand emanating from international and localmarkets. Our results are very consistent with the NEG model, which predicts that the rela-tionship between market access and income will
15、be weaker as migration is stronger. We thusconfirm previous results (De Sousa and Poncet, 2007; Hering and Poncet, 2006) on that fur-ther liberalization of internal migration may help to mitigate the widening of spatial incomedisparities fueled by the further opening of the country.4Economic Geograp
16、hy, Spatial Dependence and Income Inequality in ChinaABSTRACTThis paper contributes to the analysis of growing income disparities within China. Basedon a structural model of economic geography using data on per capita income, we evaluatethe extent to which market proximity and spatial dependence can
17、 explain growing incomeinequality between Chinese cities. We rely on a data set of 195 Chinese cities between 1995and 2002. Our econometric specification incorporates an explicit consideration of spatialdependence effects in the form of spatially lagged per capita income. We provide evidencethat the
18、 geography of access to markets is statistically significant in explaining variation inper capita income in China, especially so in provinces with low migration inflows which iscoherent with NEG theory.JEL classification: E1, O1, O5, R1.Keywords: Income inequality, Economic geography, Spatial depend
19、ence, China.5CEPII, Working Paper No 2007-22GOGRAPHIE CONOMIQUE, DPENDANCE SPATIALE ET INGALIT DE REVENUEN CHINERSUM NON TECHNIQUEAu cours des deux dernires dcennies, la Chine a bnfici dune croissance du revenusans prcdent mais le dveloppement conomique sest accompagn dcarts importants etcroissants
20、des revenus au sein de son territoire (Meng et al., 2005). Les rgions faiblerevenu par tte se localisent principalement en priphrie alors que celles ayant les revenusles plus levs sont situes au centre de sorte quune structure centre-priphrie mergeclairement lintrieur de la Chine.Des tudes rcentes m
21、ontrent que cette volution est cohrente avec les prdictions de laNouvelle Economie Gographique (NEG) (De Sousa et Poncet, 2007) et confirment la va-lidit de lquation de salaire de la NEG (Hering et Poncet, 2006; Lin, 2005; Ma, 2006): leslocalits proches des marchs de consommation (i.e. avec un “accs
22、 au march” suprieur)subissent des cots de commerce plus faibles et profitent de revenus suprieurs (Fujita et al.,1999).Plusieurs critiques peuvent tre nanmoins mises sur ces analyses. Tout dabord, ellesne tiennent pas compte des diffrences de dotations, politiques ou institutions entre les lo-calits
23、 puisquelles ne contrlent pas pour les effets spcifiques des localits. Une autrelimite est quelles supposent que chaque localit est une entit isole. Pourtant les unitsgographiques individuelles ont tendance tre relativement intgres en raison de flux mi-gratoires, dchanges commerciaux, deffets de dif
24、fusion technologique ou de connaissanceet dinstitutions communes (Buettner, 1999), qui induisent une dpendance spatiale entre leslocalits proches. La dpendance spatiale est ainsi reconnue comme une force importantedans le processus de convergence (Rey et Montouri, 1999) de sorte que lignorer dans la
25、procdure destimation pourrait produire un problme srieux de spcification (Abreu et al.,2005).Cet article contribue une meilleure comprhension de la relation existant entre laccs aumarch et les ingalits spatiales en Chine travers ltude des diffrences de revenu entrevilles. Nous utilisons une base de
26、donnes de panel sur 195 villes chinoises ce qui permetdinclure des effets fixes par ville et ainsi de contrler les effets dchelle et linfluence desdotations de facteurs.Nos rsultats suggrent que laccs la source de demande est vritablement un facteurimportant de la dtermination de la structure spatia
27、le de revenu en Chine. Si les relationsspatiales entre les villes chinoises ont un effet significatif, elles ne biaisent pas les estima-tions de limpact de laccs au march sur le revenu. Dans loptique de dterminer si loffrede travail sous la forme dimmigration interne joue un rle dans cette dynamique
28、, nous tu-dions si le lien entre accs au march et revenu est valide uniquement pour les localits avecimmigration faible ou leve. Daprs notre modle thorique, en prsence doffre quasi-ment infinie de main doeuvre dans le secteur manufacturier, les salaires devraient rpondreseulement marginalement aux c
29、hangements de demande issue des marchs locaux et interna-tionaux. NosrsultatssonttrscohrentsaveclemodledelaNEG,quiprditquelarelationentre accs au march et revenu sera plus faible en cas de migration intense. Nous confir-mons ainsi des rsultats prcdents (De Sousa et Poncet, 2007; Hering et Poncet, 20
30、06) surle fait que la libralisation accrue de la migration interne permettrait dattnuer les ingalitsspatiales de revenu induites par louverture internationale du pays.6Economic Geography, Spatial Dependence and Income Inequality in ChinaRSUM COURTDans ce travail, nous tudions les dterminants de la c
31、roissance des disparits de revenu lintrieur de la Chine. Sur la base dun modle de la Nouvelle Economie Gographiqueappliqu des donnes de revenu par tte, nous valuons dans quelle mesure la proxim-it au march et la dpendance spatiale contribuent au creusement des ingalits de revenuentre les villes chin
32、oises. Notre base de donnes porte sur 195 villes chinoises sur la pri-ode 1995 2002. Notre spcification conomtrique incorpore une considration explicitedes effets de dpendance spatiale sous la forme du revenu par tte spatialement dcal.Nous montrons que la gographie de laccs au march explique de mani
33、re significativelvolution des revenus per tte en Chine, notamment dans les provinces caractrises pardes flux dimmigration faibles en cohrence avec la thorie de la NEG.Classification JEL : E1, O1, O5, R1.Mots Clefs : Ingalit de revenu, Gographie conomique, dpendance spatiale, Chine.7CEPII, Working Pa
34、per No 2007-22ECONOMIC GEOGRAPHY, SPATIAL DEPENDENCE ANDINCOME INEQUALITY IN CHINALaura HERING 1Sandra PONCET21 IntroductionThis paper provides a case study of how economic geography contributes to incomedisparities. Over the last two decades, China has benefited from unprecedented in-come growth bu
35、t its development process has come with large and increasing incomelevel differences within China (Meng et al., 2005). With regions with low per capitaincome predominantly located at the geographical periphery while those with thehighest income locate at the center, a core-periphery structure is cle
36、arly emergingwithin China.Recent studies show that this evolution is coherent with predictions from the NewEconomic Geography (NEG) theory (De Sousa and Poncet, 2007). This theory ex-plains the emergence of a heterogeneous economic space on the basis of increasingreturns to scale and transport costs
37、 (Krugman, 1991 and Krugman and Venables,1995). In NEG models, the spatial distribution of demand is a key determinant ofeconomic outcomes. Recent findings on Chinese data (Hering and Poncet, 2006;Lin, 2005; Ma, 2006) confirm the validity of the “wage equation”: locations closerto consumer markets (
38、i.e. with a higher “market access”) experience lower transportcosts and enjoy higher income levels (Fujita et al., 1999).Most of the studies on China cited above rely on province-level data on market ac-cess and income to identify that greater market proximity is associated with higheraverage provin
39、cial wages. These findings are however confirmed at the micro level.Hering and Poncet (2006) use a data set covering around 6,000 Chinese workersfrom 56 cities in 11 provinces in 1995 and find that a significant fraction of inter-individual differences (within provinces) in terms of return to labor
40、can be explainedby the geography of access to markets.Some critics can be addressed to these analyzes. First, these studies can be criticizedfor not properly controlling for differences in endowments, policies or institutionsacross locations since none include location fixed effects.3Another possibl
41、e shortcoming of these studies is that they assume each location to bean isolated entity. But individual geographical units tend to be relatively integrateddue to migration, interregional trade, technology and knowledge spillovers as well1Universit Paris 1, CNRS2Universit Paris 1, CNRS and CEPII.3 H
42、ering and Poncet (2006) add fixed effects to account for specific features at the province-sectorlevel, but cannot include fixed effects at the city level.8Economic Geography, Spatial Dependence and Income Inequality in Chinaas institutions (Buettner, 1999), what can lead to a spatial dependence bet
43、ween thelocations. The concept of spatial dependence refers to the absence of independencebetween geographic observations, and is defined as the correlation of a variable withitself proceeding from the geographic distribution of data. This means, economiccharacteristics, as for example income, may b
44、e correlated with those of neighboringlocalities.4Spatial dependence is acknowledged to be an important force in the process of con-vergence (Rey and Montouri, 1999) and ignoring spatial dependence in the estima-tion could therefore result in serious misspecification (Abreu et al., 2005). Indeed,Yin
45、g (2003) estimates an empirical model of output growth for China using crossprovincial data over the 1978-1998 period and shows that not controlling for spatialdependence leads to a misspecification. Several studies on foreign direct investmentin China show the importance of spatial dependence at th
46、e provincial (Cheung andLin, 2004; Coughlin and Segev, 2000) and city-level (Madariaga and Poncet, 2007)for the Chinese economy.Eventhoughspatialeconometricsliteraturehasreceivedincreasingattentionoverthelastyears, pastresearchontheimpactofaccesstomarketsonincome(whetherinthecontext of China or not5
47、) mainly ignores these potential problems and, as a result,previously measured parameter estimates and statistical inferences are questionable.Hanson (2005) and Mion (2004) were the first to address spatial dependence in aNEG framework, based on the Krugman-Helpman model. Fingleton (2006) uses thesa
48、me theoretical model to test NEG theory versus urban economics theory, showingthat taking into account spatial dependence can render its market potential variablenon significant when looking at a very fine geographical level.The current paper contributes to a better understanding of the relationship
49、 betweenmarketaccessandspatialinequalityinChina. Itaimsatexploringregionalvariationswithin a single country (China), mitigating the two common shortcomings of theliterature listed above, by relying on a panel data set covering 195 Chinese citiesbetween 1995 and 2002. With data for several years for a high number of locations ata fine geographical level, our regressions can i
