1 / 36

Core-Periphery Model of Urban Economic Growth:

Empirical Evidence from Chinese City-Level Data (1990-2006)

Zhao Chen, Ming Lu and Zheng Xu

Abstract: The geographic nature and openness of China since 1990s make it a feasible

application of the Core-Periphery (CP) model, which has hardly explained urban

systems in existing studies. Using Chinese city- level data from 1990 to 2006, this

paper estimates the impact of spatial interactions in China’s urban system on urban

economic growth. Our results verify the non- linearity of the CP Model of urban

system, present agglomeration shadow in Chinese urban economies. We also find

“border effect” of administrative boundaries among Chinese provinces, which

prevents urban economic activities from being absorbed by big cities in other provinces,

while limiting inter-city agglomeration.

Key Words: Core-Periphery (CP) Model, border effect, urban system, agglomeration

shadow, China

Zhao Chen: China Center for Economic Studies, and Center for Industry Development Studies, Fudan University,

Shanghai, 200433, Email: zhaochen@fudan.edu.cn; Ming Lu: School of Economics, Fudan University and

Zhejiang University, Email: lm@fudan.edu.cn. Zheng Xu: Corresponding author, China Center for Economic

Studies, Fudan University, Shanghai, 200433, E-mail address: zhengxu@fudan.edu.cn. Financial support from the National Social Science Foundation (08BJL008), the Shanghai Leading Academic Discipline Project (B101) and

“985” project of Fudan University are greatly appreciated. We thank Jacques-François Thisse, Thierry Mayer and

seminar/conference participants at Bank of Finland, Fudan University and Peking University for their useful

comments. The content of this article is the sole responsibility of the authors.

2 / 36

I. Introduction

Krugman (1993) suggests that the presence of a particular site settled may skew

further development in vicinity in its favor, via agglomeration, which has been proved

by Core-Periphery (CP) Model of the New Economic Geography (NEG) theory in

many theoretical literatures (e.g. Fujita et al., 1999b). Although CP model has been

widely used in explaining urban, regional and international development, and even as

the backbone of the NEG literature (Krugman, 1991), few empirical studies investigate

its success in explaining current urban systems (Partridge et al., forthcoming).

Our paper will show how spatial interactions among cities affect China’s urban

economic growth, within the CP structure. As China has been rapidly opened to the

world after 1994, we expect the geographic nature and openness make China a perfect

application of the CP model. The fast opening process is a natural experiment to see

whether the greater access to the global market will make the distance to major ports

matter in city growth and shift the system of cities. Thus, we contribute new evidence

to the theory of Core-Periphery (CP) Model of urban system.

Our empirical study assumes: there are two hierarchical urban systems in

China-- one is the national urban system, cores of which are major ports, like

Shanghai and Hong Kong; the other is the regional urban system, cores of which are

big cities in China, like Guangzhou, Chongqing, etc. Furthermore, the spatial

interactions within each urban system show the Core-Periphery Structure, which will

be tested by our empirical results.

In the next section, we briefly review the theoretical background and empirical

studies on spatial interactions among cities; Section III introduces China’s urban

system and explains why China is a feasible application of the theory. Section IV

discusses the data and our econometric approach. Section V and VI present the results

and robustness checks. Concluding remarks are provided in the last section.

II. Literature Review

3 / 36

The NEG theory emphasizes the interplay of agglomeration and dispersion

forces in determining urban system (Overman & Ioannides, 2001). Krugman &

Elizondo (1996) even said that any interesting model of economic geography involves

a tension between the "centripetal" forces that tend to pull population and production

into agglomerations and the "centrifugal" forces that tend to break such agglomerations

up. Here, centripetal forces can include both pure external economies and a variety of

market scale effects, such as the forward and backward linkages, knowledge spillovers

(Krugman, 1991), while centrifugal forces include pure external diseconomies such as

congestion and pollution, urban land rents, transportation costs, and the interests of

moving away from highly competitive urban locations to less competitive rural ones

(Tabuchi, 1998).

The CP model formalizes the role of agglomeration and dispersion in the

dynamic formation of an urban system, of which a prominent feature is the

emergence of a hierarchy of cities based on regional market potential, featuring a

symbiotic relationship among cities (Partridge et al., forthcoming). But CP model is

astoundingly difficult to manipulate analytically and indeed most results in the

literature are derived via numerical simulation, which as simulated by Fujita and

Mori (1997), Fujita et al. (1999a), is a ∽-shaped curve between distance to regional

center and local market potential in a single-core urban system. This curve shows that

with the distance to a central city increasing, the market potential declines first, and

later rises, then declines again.

In the model of Fujita and Mori (1997), close proximity to suppliers of

intermediate inputs and customers lowers firm’s transportation costs (Venables, 1996),

in which scale economies may exist in the production of non-traded intermediate

inputs (Fujita, 1988). However, increased competition associated with close proximity

of economic activity acts as a dispersal force (Krugman, 1991; Combes, 2000),

limiting agglomeration. Firms in the areas closest to agglomeration centers find

themselves in what is referred to as “Krugman’s agglomeration shadow” (Dobkins

and Ioannides, 2001, Ioannides and Overman, 2004), in which they are able to

produce only the most basic goods and services for which there is less competition.

4 / 36

Small cities may thus serve only a local population, while larger cities serve wider

geographic markets that include small cities (Krugman, 1996), leading to an urban

hierarchy consistent with CP structure (Fujita and Thisse, 1996).

Due to the specific implications of the CP theory such as agglomeration shadow,

Dobkins & Ioannides (2001) noted that non- linearity (such as the “∽” shape)

might show up in the spatial interactions in urban system. So to directly examine the

spatial interactions among cities relating to their geographic distance, urban

economies, or their place in the urban hierarchy has generally become one of the

streams of empirical studies not only on agglomeration (Hanson, 2001), but also on

spatial distribution of cities (Partridge et al., forthcoming).

“Buttressing the approach with empirical work” is one of the most important

directions for the new economic geography future researches as Fujita and Krugman

(2004) suggests. However, “due to the highly nonlinear nature of geographical

phenomena”, it’s not easy “to make the models consistent with the data” (Fujita and

Krugman, 2004). Empirical studies on spatial interactions have partial testified the CP

model and agglomeration force in urban systems, and find that, more or less, closer is

better, but the “non- linearity” of CP model and “agglomeration shadow” remains to be

unverified.

Brülhart and Koenig (2006) analyze the internal spatial wage structures of the

Czech Republic, Hungary, Poland, Slovakia and Slovenia, using regional data for

1996–2000. They find that real wage falls with distance from national capitals and

from Brussels, but none of these estimates are statistically significant.

Related empirical evidences based on U.S. urban data are more complex and

puzzling. Black and Henderson (1999) examine the correlation between population

growth in U.S. metropolitan areas and initial conditions over the period 1950-1990.

Population growth is faster in cities with closer proximity to a coast and to cities with

larger initial populations, with this effect weakening as neighboring population

masses become larger. Dobkins and Ioannides (2000, 2001), Ioannides and Overman

(2004)，using the U.S. metropolitan data 1900-1990, find that distance from the

5 / 36

nearest higher-tier city is not always a significant determinant of size and growth and

no evidence of persistent non-linear effects of either size or distance on urban growth.

Partridge et al. (forthcoming) explore whether proximity to higher-tiered urban

centers affected 1990-2000 U.S. county population growth. Rather than

agglomeration shadow, their results suggest that larger urban centers promote growth

for more proximate places of less than 250,000 people. They think NEG theory (CP

model) only partially explains current U.S. urban growth, suggesting a need for a

broader framework.

Induced by the empirical studies above, we think it’s the critical time to clarify

whether the CP model is suitable for explaining the current urban system changes.

We believe our empirical study might give strong evidence to the CP model of the

NEG theory, because compared with the U.S. and other countries, China has several

unique advantages in its city- level data, which will be explained in the next session.

And instead of only being derived via numerical simulation (Baldwin et al., 2003),

the non- linearity of the Core-Periphery model captured by China’s data would

partially explain some of the puzzles in the studies using U.S. data, such as

agglomeration shadow.

III. China’s Urban System Evolution during Economic Opening

China may be an interesting application of the CP model, because, with the force

of spatial agglomeration, the quick development of both international and domestic

trade has reconstructed China’s regional and urban economies, especially since

China’s rapid openness to the world in 1994. The nature of Chinese urban system and

its evolution makes it an interesting case to testify the CP model of urban system. The

opening process is a natural experiment to see the role of the distance to main ports

and international markets in shifting the urban system.

3.1China’s Urban System

In China, “city” is defined as a local administrative and jurisdictional entity.

There are three different administrative levels of cities in China’s urban system:

6 / 36

municipalities, prefecture- level cities and county- level cities. Small settlements with

townships or lower administrative levels are not treated as “cities”. The major

administrative criteria distinguishing cities, towns and rural places is the scale of

urban population, in particular there is a lower population boundary for cities. The

economic and political importance of an urban agglomeration is also one of the

government’s considerations in defining cities. However, the definition of cities has

generally been consistent since 1949, when the People’s Republic of China was

established (Anderson and Ge, 2005).

Information of both prefecture-level and county- level cities is reported by

China’s National Bureau of Statistics (NBS), but only prefecture- level and above

cities data are used in our research, for the county- level city boundaries and numbers

have fluctuated greatly. (See Table 1) The number of prefecture-level cites has also

increased a lot since 1990, so the prefecture- level cites boundaries might also have

changed. However, for cities of the prefecture level and above, NBS reports

information of both “Diqu” (urban area plus rural counties within the same

administration) and “Shiqu” (only urban area). “Shiqu” is a good definition of the

metropolitan area by international standards (Fujita et al., 2004), and the entities of

new cities usually happen outside “Shiqu” (urban area), so “Shiqu” information is

used in our study.

Table 1: Number of China's Cities

year total municipalities prefecture- level

cities

county-level

cities

1990 467 3 182 279

1995 640 3 207 427

2000 663 4 255 400

2005 657 4 282 370

Source: NBS, Chinese Urban Statistical Yearbooks (1991–2006).

Note：Our database does not include information of Hong Kong，Macao, Taiwan and

Tibet.

The definition of China’s urban hierarchy is another problem. As we mentioned

7 / 36

above, three administrative levels of cities are included in China’s urban system:

municipalities (or province- level cities), prefecture- and county- level cities, and

provincial capitals also conduct a special administrative level in prefecture-level cities.

But according to the CP model, our study focuses on the spatial agglomeration in

urban system, for which urban size (e.g. population) and market potential are much

more important than policy. Though the effects of policy and agglomeration may

correlate with each other, we try to distinguish them by different definitions as shown

by Table 2.

Table 2: China’s Urban Hierarchy

Year

Prefecture-

level and

above cities

Provinci

al

capitals

Non-

agricultural

population>2

million

Non-

agricultural

population>1.

5 million

Non-

agricultural

population>

1 million

1990 188 29 9 14 31

1995 213 29 10 16 32

2000 263 30 13 18 38

2005 290 30 22 31 54

Source: NBS, Chinese Urban Statistical Yearbooks (1991–2006).

Note：Provincial capitals include both municipalities and other provincial capitals.

3.2China’s Reform and Openness to the World

Hanson (1998) finds in Mexico, after trade reform and openness to the world,

especially the U. S., firms locate in regions with good access to foreign markets.

International trade changes the reference market for firms in a country, shifts

resources to locations with low-cost access to foreign markets such as border regions

and port cities, during which agglomeration matters. Fujita and Mori (1996) propose

an evolutionary model of spatial economic development in which agglomeration

economies and the hub-effect of transport nodes interplay. Their simulation of market

potential in a spatial system also shows somehow the non- linearity of core-periphery

pattern.

In fact, as for China, the economic reform has gone through three stages (Ho and

8 / 36

Li, 2008): 1978-84, reform of “household responsibility system” that linked

remuneration to output for agriculture; 1985-91, reform of state-owned enterprises

(SOEs); and 1992-present, a widespread opening up of markets and a commitment to

a market economy. The increasing importance of international trade in China is shown

in Figure 2, especially after 1992. 1994 was the year of unifying the official exchange

rate and black market exchange rate of RMB. Since then China has formed a n

export- led growth pattern and has gained trade surplus up to now.

Figure 1: International Trade in China during Post–reform Era (1978-2008)

Source: NBS, China Statistical Abstract 2009.

Ever since the mid-1990s, China has become increasingly open to the

international market. International trade is also more and more important for China’s

national and urban economic growth. Wei (1995) finds that in China cities with large

export sectors grow fast. Furthermore, international trade strengthens the

agglomeration force of major ports in China and dynamically rebuilds China’s

national spatial system as well as the urban economic activities. Wei (1995) provides

some evidence that for obvious reasons, many emergence cities locate in coastal areas

near Hong Kong, the Pearl River Delta. More and more empirical studies find that

agglomeration changes the distribution of China’s regional and urban economies, so

9 / 36

coastal area and big cities have higher economic growth rates (Chen et al., 2008; Bao

et al., 2002; Ho and Li, 2008).

China’s open-door policy is an exogenous shock to city growth and the urban

system. Different from the 1980s, when China selected several coastal cities to open,

in the 1990s, open-door was a national policy. Especially after China depreciated the

exchange rate of RMB in 1994, and the Chinese economy followed an export- led

growth pattern, preferential policies have not been the major force for cities to enjoy

the benefits of international trade and FDI inflow. The geographic difference between

cities, mainly their different distance to the ports that determines international trade

costs plays an essential role in city growth. Therefore, the opening process in the

1990s is a natural experiment for economists to see whether and how the urban

system evolves following a core-periphery model.

3.3 China’s Geography and Growth

Partridge et al. (2007) believe agglomeration economies may have a greater

geographic scope than usually assumed by economists. From this view, economic

geography, in particular the role of nonlinear spatial agglomeration may require a wide

range of space, long time of accumulation, while national boundaries, geographical

boundaries, wars and other factors often limit the free flow of resources, so it is difficult

to use the empirical econometric model and real-world data to portray how

agglomeration works. Therefore, it is studies on stably-developing large countries,

such as U.S. and China after 1978, which seems particularly important for the

empirical researches of the NEG theory.

Researches using China data may contribute to the empirical studies of the NEG,

mainly based on the following reasons: 1) China has a vast territory, as well as a large

population. The vast territory provides not only space required by agglomeration, but

also plenty of samples for empirical studies. Meanwhile, a large population provides

sufficient market potentials. 2) Compared with the United States, China has a larger

interregional geographical diversity and more obvious geographical heterogeneity,

because of the concentration of the ports distribution along the eastern coast, which

10 / 36

leads to a bigger variance within cross-section samples. In the United States, which has

big ports on both the western and eastern sides, interregional geographic differences

are reduced. 3) China has a rapid economic development in the last thirty years,

especially after 1992, with observably temporal changes of the spatial distribution of

economic activities, which allow us to observe the impact of spatial agglomeration on

urban economy in a longer term with data of recent years.

Besides, local statistical bureaus in China have for years collected data on all

enterprises in their local area; and report GDP figures at the level of the appropriately

defined metropolitan area (Fujita et al., 2004; Au and Henderson, 2006a). Though

doubts may be expressed to the quality of statistical data in China, these GDP figures

are of high quality as discussed by Au and Henderson (2006a), especially when

econometric models find highly significant results consistent with economic theories.

IV. Model Specification and Data

4.1 Model Specification

The reduced-form estimation is widely used in the former empirical studies on

spatial interactions among cities (Brülhart and Koenig, 2006; Dobkins and Ioannides,

2000). We use cross-section OLS regression as our basic reduced-form CP model of

urban system, based on the economic growth model of Barro (2000). Since the urban

system is evolving during the opening process, the shift of urban system will be

reflected in different growth rates of cities in different locations. Since all the

explanatory variables are either exogenous geographic factors or initial values of

those control variables, the potential endogeneity problem of OLS estimation is not a

major concern. The model specification is as follows:

0 0 0 0 0 0 0 0(ln , , , , , ; ; ...)it i i i i i i i iDgdp f gdp inve lab edu gov fdi con geo

Here, per capita GDP growth is the dependent variable. Urban per capita GDP

growth, a measure of the urban economic activities is also used in existing literatures

(Glaeser et al., 1995; Dobkins and Ioannides, 2000; etc.). The reasons why we do not

use real wage, income, population or per capita GDP level are: First, wage and

11 / 36

income data are not reported at the China’s city level. Second, in early years, urban

population in China did not include migrants without local household registration

identity (hukou) (Au and Henderson, 2006b), so it does not well represent urban

economy or size. Third, as we want to show the dynamics of urban economies after

China’s openness to the world, per capita GDP growth, instead of its level, would be a

better measurement.

In our studies, we highlight the roles of openness and international trade in

determining the distribution of China’s urban economic activities, so we conduct the

dataset for years 1990-2006. As discussed above, the third period of reform started in

1992, but most information about China’s cities in 1992 and 1993 are missing in

Chinese Cities Statistical Yearbook (National Bureau of Statistics, 1991-2007). We

will also use the data of 1994-2004 to check the robustness of our results after 1994,

the year of drastic openness.

Hanson (2001) pointed out three issues may trouble for identifying

agglomeration effects in empirical studies: unobserved regional characteristics,

simultaneity in regional data, and multiple sources of externalities, of which the

former two are particularly crucial when it comes to the spatial interaction of cities.

As to “unobserved regional characteristics”, we will try to control as many

variables as the data allow according to the literatures of economic growth and

empirical studies of China. “Simultaneity in regional data” may be a minor issue in

our study, since the essential explanatory variables in our model are geography and

will not change in our observations. However, we will use the initial state of other

explanatory variable in 1990 to alleviate simultaneity bias of the model, so the

estimated results would represent the long-term impact of the explanatory variables

on urban economic growth.

Initial advantage is a challenging concept for modelling. Where did initial

advantage come from in our story? History (Krugman, 1991) or geography (Fujita and

Mori, 1996)? This question is out of the scope of this paper and won’t be answered

directly by our empirical results, though some related evidences would be presented.

12 / 36

4.2 Data

In this paper, we used China’s urban area (Shiqu) data (1990-2006) complied

from Chinese Cities Statistical Yearbook (National Bureau of Statistics, 1991-2007),

including 286 prefecture- level and above cities① from 30 provinces of mainland

China. The information of Hong Kong, Macao, Taiwan and Tibet is not included.

Other data sources and constructions of variables are as follows.

Urban Economic Growth

The dependent variable itDgdp is the the average annual growth rate of real per

capita GDP from 1990-2006 deflated by provincial urban CPIs, respectively, for city i

year t. The theoretical assumptions of the NEG theory emphasize the agglomeration

of manufacture and service (Krugman 1991), so we removed the agricultural output

out of the GDP indicator and the agricultural population out of the population

indicator.

Spatial Interactions among Cities

Hanson (1998, 2005) approximates the access of each considered region to its

principal markets by geographic distance. Therefore, geographic distance as well as

driving distance or road/railway distance is used to approximate spatial interaction

among cities (Dobkins and Ioannides, 2000; Brülhart and Koenig, 2006); Partridge et

al., forthcoming). In particular, we use straight geographic distance as our

measurement of spatial interaction, because it’s exogenous as second nature, which

avoids the potential endogenous bias brought by traffic distance.

Geographic Distances to the Nearest Big City and Major Ports

Our hypothesis is that there are two hierarchical urban systems in China. One is

the national urban system, cores of which are major ports, like Shanghai and Hong

Kong. The distance to these major ports measures a city’s remoteness to the global

market. The other is the regional urban systems, cores of which are big cities in China,

① We have information of 286 cities in 2006, while 211 cities in 1990.

13 / 36

like Guangzhou, Chongqing, etc. Furthermore, the spatial interaction within each

urban system shows the Core-Periphery Structure, which will be testified by our

empirical results. So we use distances to the nearest big city (distbig) to measure the

interaction in regional urban systems, and distance to the nearer major port (disport)

to measure the interaction among national urban systems.

As to exploit how agglomeration of major ports affects urban economic activities

in China’s spatial system, we need to define “major ports” in China.

Our information about ports cargo throughput is complied from China Statistical

Abstract 2009 (National Bureau of Statistics, 2009). Though it contains information of

ports of mainland China and Hong Kong in 1990, they are measured in different ways

and hard to compare. So we have to use the recent year (e.g. 2008) cargo throughput

of ports which could be transferred in the same measurement as “ton”. According to

China Statistical Abstract 2009, in 2008, as to major ports in China: Hong Kong

accounted for 10.8% of China’s overall cargo throughput, Shanghai 10.6%, Tianjin

7.4%②; as to ports of different regions, the Bohai Sea ports (Tianjin, Qingdao,

Qinhuangdao, Dalian) totally accounted for 24%; the Pearl River Delta ports (Hong

Kong, Guangzhou, Shenzhen) 22.4%, the Yangtze River Delta ports (Shanghai,

Ningbo-Zhoushan) 21.4%. The data above clearly show that cargo throughput in

China mainly concentrates in those three areas, but the distribution of the Bohai Sea

ports is comparatively disperse, therefore, the agglomeration of Tianjin might be not

as strong as Shanghai and Hong Kong. So we define Shanghai and Tianjin as “major

ports” of China in our study, and also take Tianjin as major ports when testing the

robustness.

In Table 2, we try to find a proper definition of high- level cities by their

non-agricultural population. To construct time-consistent data, we will only use the

information of 1990 to conduct such definition for “big cities”, and information after

② Some may argue that the rank of Shanghai port is higher than Hong Kong in 2008 according to some

international index, which is due to the different measurements of cargo throughput. It’s not a big issue since both

are major ports in our study.

14 / 36

1990 is just to show how quickly China’s urban sizes evolve. Finally, we choose the

1.5 million-non-agricultural-population as the lower boundary of high- level cities in

China’s urban hierarchy. Reasons for such criteria is that: first, in 1990, there were

nine cities with a non-agricultural population greater than 2 million, eight of which

located in coastal area, so this lower boundary (if so), would not well represent the

urban hierarchy in inland area; second, in 1990, there were thirty-one cities with a

non-agricultural population larger than 1.5 million, most of which are municipalities

and provincial capitals, therefore high- level cities according to this definition may

highly correlate with municipalities and provincial capitals in China. If we use 1

million non-agricultural population as the lower boundary, there will be too many

“big cities”.

Taking above reasons into consideration, we define “big cities” in China’s urban

hierarchy as those with a non-agricultural population more than 1.5 million in 1990.

Our assumption is that: via agglomeration, those big cities had or would generally

become the regional centers of domestic and regional market and thus, play a central

role in China’s urban system of each region within the CP structure. We will testify

such assumption, and also use “capitals” which as our definition of both

municipalities and provincial capitals to check the robustness of our empirical results.

The distribution of big cities in China is in Figure 2③ below.

As the CP model shows the non-linear nature as discussed before, we also add

the square and cube of distance in some of our regressions to capture such

non- linearity. The geographic distance is measured in China Map 2008, developed by

Beijing Turing Software Technology Co., Ltd, China Transport Electronic &

Audio-Video Publishing House.

For the defined fourteen big cities, their distances to the nearest big city are

defined as zero. Their economic growth may benefit from their city size, so we use a

dummy bigcity to control it. This would not be a big problem to major ports, for Hong

③ They are Beijing, Tianjin, Shenyang, Wuhan, Guangzhou, Ha’erbin, Chongqing, Xi’an, Nanjing, Dalian,

Chengdu, Changchun and Taiyuan.

15 / 36

Kong is not in our sample and Shanghai would have already been controlled by

bigcity. Besides, GDP level of the nearest big city in the initial year will also be

controlled by gdpofbig, which is ln(GDP) in 1990, exclusive of agricultural output.

Border effect

Policy is an important factor impacting spatial agglomeration in the NEG theory.

When it comes to studies on Chinese urban system, an important policy variable

which affects regional economic development cannot be overlooked, that is provincial

boundary. Studies have suggested that among Chinese provinces there is serious

market segmentation (Young, 2000; Ponect, 2005; Chen, et al., 2007). This division is

likely to add the actual distance between cities in different provinces, and limit the

spatial interaction among cities. Our studies will use a dummy variable to find out

whether the administrative boundary would limit spatial interaction within the urban

system. Denoted as samepro, the dummy variable represents whether a city is in the

same province where its nearest big city locates. If the “border effects” of China’s

province does not affect the spatial interaction among cities, and thus urban economic

growth, this variable would be insignificant; otherwise, significantly positive or

negative.

Geography

Actually, spatial interaction is only the second nature of cities. For studies on

China’s urban economy, there are some other geographic factors of first nature should

be controlled for potential initial advantages (Krugman, 1993).

Mainland China usually is divided into three regions: East, Center and West.

Eastern region has eight provinces and three municipalities, Central China has seven

provinces while West has elenven provinces and one municipality. Due to the

concentration of the ports distribution, there are significant inter-regional differences

in climate, access to water, topography, and other features of the environment (Ho and

Li, 2008), as well as economic development. So we use a dummy “west” or “center”

16 / 36

to capture inter-regional differences of geography④. Actually, the regional division of

mainland China does not strictly follow the geographic location. For example,

Guangxi Province locates in the southeast coastal area but was divided into the West

by the Strategy of Western Region Development launched by China government in

1999, for its economic performance.

The distribution of China’s provinces is presented in the Figure 2.

Figure 2: Distribution of Big Cities, Major Ports and Provinces in Mainland China

Note: (1) Deltas and squares, respectively, are big cities and major ports in our

definition.

④ West includes cities of 12 provinces and municipality: Chongqing, Sichuan, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang, Inner Mongolia, Guangxi, Guizhou; Center includes cities of 7 provinces: Hebei,

Anhui, Jiangxi, Heinan, Hubei, Hunan and Shanxi. Cities of other 12 provinces and municipalities belong in the

East. They are Heilongjiang, Jilin, Liaoning, Tianjin, Beijing, Shandong, Jiangsu, Shanghai, Zhejiang, Fujian,

Guangdong and Hainan.

17 / 36

(2) Provinces in square brackets locate in the west, while those in

parentheses in the center, and others in the east.

Besides, cities with ports in coastal area or with river ports along big rivers

might benefit from access to international or domestic markets, so we also use dummy

seaport or riverport to control such initial geographic advantage. The list of cities

with sea ports or river ports is from “The First China Port City Mayors (International)

Summit Forum 2006” held by the Development & Research Center, the Ministry of

Communications, the Municipal Government of Tianjin, and the China

Communications & Transportation Association.⑤

Initial State of Economic Growth Factors

Following the traditional economic growth literatures (Barro, 2000) and studies

on China’s economic growth, we will also measure some other factors that may

contribute to China’s urban economic growth, such as initial economic performance,

investment, labor, education, etc.

0ln igdp is the logarism of per capita output of manufacturing and service in

1990, which is controlled to observe whether Chinese economy experiences

conditional convergence at city level. 0iinve is the ratio of investment to GDP, where

GDP is the total output of manufacturing and service in 1990.

We control the ratio of employee to total population ( 0ilab ) in 1990 to proxy

labor. China’s urban labor market reform was drastically pushed ahead in the

mid-1990s, so in 1990 urban employees contain lots of hidden unemployment in

state-own enterprise. Therefore, higher labor ratio may not be significantly positive

for urban economic growth. We use the ratio of teachers to students in primary and

⑤There are thirty-two cities with sea ports: Qingdao, Yantai, Weihai, Rizhao, Haikou, Sanya, Tianjin, Tangshan,

Qinhuangdao, Zangzhou, Dalian, Jinzhou, Yinkou, Lianyungang, Fuzhou, Xiamen, Quanzhou, Zhangzhou, Guangzhou, Shenzhen, Zhuhai, Shantou, Zhanjiang, Zhongshan, Shanghai, Linbo, Wenzhou, Zhoushan, Taizhou,

Beihai, Fangchenggang, qinzhou; while twenty-two with river ports: Ha’erbin, Jiamusi, Wuhu, Ma’anshan,

Tonglin, Anqin, Yueyang, Nanjing, Wuxi, Suzhou, Nantong, Yangzhou, Zhenjiang, Foshan, Dongwan, Luzhou,

Wuhan, Yichang, Nanchang, Jiujiang, Nanning, Wuzhou, Chongqing.

18 / 36

junior schools (0iedu ) in 1990 to control the effect of education, which may be not a

good measurement but the best we could find.

And the government expenditure-to-GDP ratio and FDI-to-GDP ratio in 1990, as

usually controlled by economic growth literatures, are denoted as 0igov and

0ifdi in

this paper, where GDP is the total output of three sectors for we cannot distinguish the

ultimate flows of government expenditure or foreign direct investment among sectors.

The effects of government expenditure and foreign direct investment on economic

growth are difficult to predict (Barro，2000; Clarke, 1995; Partridge, 1997).

0icon represents some other control variables related to Chinese urban economic

growth, including: the ratio of the non-agricultural population to the total population

(urb) accounting for the level of urbanization; the population density (density) and its

square (den_2) accounting for internal population agglomeration in urban areas; the

ratio of service GDP to manufacturing for the industrial structure of urban sectors

(SM).

Policy

Policy can never be ignored in empirical studies on agglomeration or urban

economies. Besides the administrative boundary mentioned above, there might be

some other policy advantages affecting China’s urban economic growth as controlled

by literatures (Bao et al., 2002; Ho and Li, 2008)

As introduced before, in our database, there are still different administrative

levels of cities: municipalities and prefecture- level cities which also include

provincial capitals and other cities. Because municipalities and provincial capitals

may enjoy special policy from the central and provincial governments, we use dummy

capital to capture such policy benefits. Chongqing became a municipality in 1997

from an ordinary prefecture- level city. Take the special history of Chongqing into the

consideration⑥, we define it as a capital in 1990.

⑥ Chongqing once was the capital of Republic of China during the World War II, and is an important big city in

western China.

19 / 36

In the process of China’s reform and openness, four Special Economic Zones

(SEZs) were established since 1980 in China’s coastal area, and later 14 cities were

chosen as Open Coastal Cities in 1994⑦. Those Special Economic Zones and Open

Coastal Cities were established to attract foreign direct investment and access to

international market and have special policy advantages. We use the dummies SEZ

and open to measure the effects of the preferential policy. Since they may highly

correlate with FDI, dummy seaport or some others, we don’t expect significant

coefficients of them.

Descriptive statistics of all variables are in Table 3⑧.

Table 3: Descriptive statistics

Variable Obs Mean Std. Dev. Min Max

dgdp(%) 208 8.36 2.82 -.37 17.10

distbig(km) 286 291.33 251.35 0 2351.8

disport(km) 286 896.75 542.90 0 3526.4

bigcity 286 0.07 0.26 0 1

gdpofbig(100

million yuan)

286 114.93 80.74 37.36 344.47

samepro 286 0.41 0.49 0 1

seaport 286 0.11 0.32 0 1

riverport 286 0.08 0.28 0 1

center 286 0.35 0.48 0 1

west 286 0.29 0.46 0 1

gdp per

capita(yuan)

210 3581.295 2120.255 382.23 19820.83

inve(%) 210 25.91 16.67 1.54 127.88

⑦ Four Special Economic Zones established in 1980 are: Shenzhen, Zhuhai, Xiamen, Shantou; fourteen Open

Coastal Cities in 1994 are: Tianjin, Shanghai, Dalian, Qinhuangdao, Yantai, Qingdao, Lianyungang, Nantong, Ningbo, Wenzhou, Fuzhou, Guangzhou, Zhanjiang, Beihai. ⑧ We drop the value of labor of Baicheng, Jinlin Province and Shenzhen, Guangdong Province, and drop the value

of gov of Siping, Jinlin Province, because they are abnormal. In fact, our results are robust whether or not with

those observations included.

20 / 36

labor(%) 209 57.60 7.79 30.01 93.79

edu(%) 207 4.68 1.26 3.02 18.58

gov(%) 199 10.26618 3.821074 1.04 29.16

fdi(%) 145 3.28 9.78 0.01 100.99

urb(%) 211 59.40 23.62 8.05 96.53

density (per

km2) 210 11849.77 4180.96 3631 32920

SM(%) 210 59.53 31.95 7.14 188.51

capital 286 0.11 0.31 0 1

open 286 0.05 0.21 0 1

SEZ 286 0.02 0.15 0 1

V．Estimation Results

Estimated results of the model are in Table 4. In addition to the traditional

economic growth factors, we add the geographical variables we concern in equation

(1)-(3). In equation (1) we only add the linear item of distance to big city and major

port, while we add the squares and cubes of distances in equation (2) and drop the

insignificant cube of distance to big city in equation (3).

Table 4: Distance and Urban Economic Growth

(Dependent variable is the compound average growth between 1990-2006.)

(1) (2) (3)

dgdp dgdp dgdp

distbig -0.00169 -0.00499 -0.00802**

(0.00149) (0.00735) (0.00352)

distbig_2 6.76e-06 1.32e-05**

(1.48e-05) (5.40e-06)

distbig_3 3.57e-09

(7.59e-09)

disport -0.000215 -0.0129** -0.0112**

(0.000716) (0.00555) (0.00429)

disport_2 1.50e-05** 1.28e-05***

21 / 36

(6.79e-06) (4.82e-06)

disport_3 -4.98e-09** -4.17e-09**

(2.36e-09) (1.60e-09)

bigcity 0.754 0.883 0.724

(0.918) (0.988) (0.925)

gdpofbig0 -0.549 -0.742 -0.763

(0.625) (0.637) (0.633)

samepro -1.947*** -2.165*** -2.211***

(0.666) (0.707) (0.697)

seaport 1.978** 2.022** 1.962**

(0.838) (0.843) (0.831)

riverport 0.752 0.730 0.614

(0.649) (0.693) (0.645)

center -1.391* -0.989 -1.025

(0.797) (0.812) (0.806)

west -0.980 -1.349 -1.330

(0.878) (0.878) (0.874)

lngdp -0.845 -1.249* -1.201

(0.726) (0.745) (0.735)

inve -0.0345* -0.0278 -0.0274

(0.0201) (0.0201) (0.0200)

labor 0.0427 0.0416 0.0411

(0.0336) (0.0344) (0.0343)

edu 0.552*** 0.590*** 0.584***

(0.172) (0.171) (0.170)

gov 0.0439 0.0269 0.0307

(0.0743) (0.0753) (0.0746)

fdi 0.0305 0.0146 0.0132

(0.0256) (0.0262) (0.0259)

urb 0.00701 0.00351 0.00460

(0.0135) (0.0136) (0.0133)

density 0.000215 0.000270 0.000280

(0.000207) (0.000208) (0.000206)

den_2 -5.07e-09 -7.77e-09 -7.99e-09

(6.00e-09) (6.05e-09) (6.01e-09)

SM -0.00887 -0.00516 -0.00475

(0.0106) (0.0107) (0.0106)

capital 1.051 0.598 0.548

(0.775) (0.792) (0.782)

open -0.348 -0.287 -0.282

(1.043) (1.031) (1.027)

SEZ -1.798 -1.976 -2.005

(1.436) (1.416) (1.410)

Constant 13.06* 19.72** 19.39**

22 / 36

(7.626) (8.290) (8.230)

Observations 132 132 132

R2 0.387 0.430 0.428

Notes: (1) Standard errors in parentheses;

(2) * p < 0.10,

** p < 0.05,

*** p < 0.01.

Non-linearity of the Core-Periphery Structure in China’s Urban System

In equation (1), both of the distances are insignificantly negative, which means

urban economic growth decreases away from big city and major port. As we

discussed before, the non- linearity of the CP model may be shown in China’s urban

system. If so, à la Fujita et al. (1996, 1999b), the “∽” shaped correlation between

distance and economic activities might present in our estimated results. So the squares

and cubes of distances are added in equation (2)

Results of equation (2) show that distance to major port and its square and cube

are all significant, which partially proves the non- linearity in China’s urban system.

However, neither of the three terms of the distance to big city is significant. We guess

a “∽” shaped curve could only occur when distance to big cities is long enough,

otherwise only a U-shape can be seen in the real data. So we remove the cube of

distance to big city from equation (3). Obviously, in equation (3), all distance

variables are significant: distance to big city is negative, its square item positive;

distance to major port is negative, while the square item positive and the cubic term

negative.

Based on estimated results, we simulate the correlation between distances to the

major ports or big cities and urban economic growth rate, respectively, in Figure IV.

The horizontal axis represents the distance (kilometers) away from major ports, and

the vertical axis means the urban economic growth rate (%). We do find a

core-periphery pattern of urban system evolution.

23 / 36

Figure 3: Distance to Major port and Urban Economic Growth.

Figure 3 suggests that the impact of distance to major port on urban economic

growth has basically the same shape with the market potential curve of CP model in

urban system (Fujita et al., 1996, 1999b). While a city is located away from major

ports within about 600 kms, the closer it is to the major port and international market,

the larger market potential and the higher economic growth rate it has. While the

distance is farther away, international market access is no longer that important.

Therefore, a location far away from ports might promote the accumulation of regional

and domestic market potential, as well as the the development of local economies.

When distance is long enough, farther than 1500 kms, cities far way from both

domestic and international market would suffer from low market potential and

economic growth rate. Our result gives convincing evidence of the CP model of city

structure in the case of China.

In Figure 3, we mark major cities in China in accordance with their distance to

major port. To make those geographic distances more clear, we mark the two turning

points of economic growth curve on the map of mainland China in Figure 4. 600 kms

away from Hong Kong and Shanghai, denoted by a dashed curve, is the place with the

lowest average growth, while 1500 kms away, marked by a solid curve, is the place

Urban Economic Growth (%)

Distance (km)

Shenzhen Hefei Wuhan Zhengzhou Chongqing Chengdu Xi’ning

Suzhou Nanjing Changsha Taiyuan Xi’an Urumchi

Shanghai

Hong Kong

24 / 36

with second highest growth rate.

Figure 4: Geographic Distance and Urban Economic Growth.

In Figure 5, we simulate the correlation between distance to big city and urban

economic growth based on our regression.

Figure 5 Distance to big city and Urban Economic Growth

Urban Economic Growth (%)

Distance (km)

25 / 36

When it’s close to big cities, scale effects and other external economies related

with spatial agglomeration promote big cities to absorb economic resources from

surroundings, which is the significant centripetal force. So the closer to the central

cities, the faster a city grows. But when it’s far away from big cities, instead of the

centripetal force, the centrifugal force plays a major role. So the farther the distance,

the faster a city grows. Our estimated result shows the turning point is about 300

kilometers, which means within a scope of 300 kilometers, interplay among cities

shows a strong centripetal force, which is similar with Hanson (2005). The difference

is that we also find that when the distance is greater than 300 kilometers, because of

transportation cost and other external diseconomies, the spatial interaction among

cities performs is dominated by the centrifugal force.

Comparing Figure IV and Figure VI, we find that: First, the impact of major

ports is stronger than big cities on urban economies, because within a certain distance,

the economic growth decreases much faster as distance from major ports increases

than from big city. Second, in Figure 4, the curve is U-shaped rather than “∽”-shaped,

which is not surprising. In Figure 3, the complete “∽”-shaped curve requires at least

1,400-km distance, while the real distance to big city is not long enough. In fact,

China’s urban system today is the result of a long-term evolution, thus new big cities

might have emerged wherever there was large market potential due to the spatial

agglomeration of old big cities, as the numerical simulation of Fujita and Mori (1997),

and Fujita et al. (1999a). So in Figure 5, we only can see the left part of the

“∽”-shaped curve instead of the whole.

From above, we may conclude that the national urban system in China has

presented a complete Core-Periphery Structure, due to the adjustment of urban

economies to the international market; and the importance of big cities and regional

urban systems is less significant but also verified.

Agglomeration Shadow

Something about agglomeration shadow (Krugman, 1993) has also been seen in

26 / 36

Figure 3-5. In Figure 3, cities located 400-600 kms away from major ports suffer from

the absorptive effect of major ports. So compared with cities located even farther,

their economic growth rates are even slower. Figure VI also presents some similar

results.

Partridge et al. (forthcoming) think current economic realities generally predict a

positive relationship between economic activities and close proximity to the largest

agglomeration centers, which is in contrast to Krugman’s NEG agglomeration shadow.

But our results may give evidence to such argument that agglomeration shadow exists

in different geographic scope. With distance to agglomeration centers increased to

farther enough, agglomeration shadow would appear.

As the results in Table III shows, initial economic scale of big cities (gdpofbig) is

insignificantly negative, which is similar with the finding in Partridge et al.

(forthcoming). This suggests the possibility that the larger economic scale a cities has,

the stronger its absorption to its vicinities. But this needs further evidence.

Border effect

The same-province dummy (samepro) is always significantly negative, which

seems the spatial agglomeration of big cities differs whether or not smaller cities are

in the same province with its nearest big city. If a city is in the same province with the

nearest big city, the absorptive effects from big cities will be larger, and consequently,

the city grows slowlier. On the contrary, this means that the "border effect" similar to

the findings in Parsley and Wei (2001) and Poncet (2005) exists among Chinese

provinces, which is likely to increase the actual distances between cities in different

provinces. Based on our estimates, China's inter-provincial "border effect" is

equivalent to as much as 260 kms⑨ for two neighboring cities. Poncet (2005) finds

that the inter-provincial border in China is just like international border in Europe, so

our estimate of the inter-provincial border effect is acceptable.

The "border effect" in this paper presents as the distortion of the spatial

⑨ We estimate this “border effect” by dividing the coefficient of the same-province dummy by that of the

distance to the nearest big city in Equation (3).

27 / 36

concentration. We think it’s relevant with Chinese province- level market

segmentation (Young, 2000; Ponect, 2005) and migration restrictions (Au and

Henderson, 2006a, 2006b). Chen et al. (2007) argue that the provincial governments

have incentives to restrict the agglomeration effects from big cities in other provinces

by administrative forces, so as to protect their own economic development. This

argument is also consistent with the finding in this paper. Although for the cities of

the same province, such segmentation can prevent them from the absorptive effects of

large cities in other provinces, but market segmentation always brings loss of resource

allocative efficiency, thus a lower growth rate of the whole economy, resulting in

underdeveloped city scale (Au and Henderson, 2006a), and smaller scale inequality

among Chinese cities (Fujita et al., 2004).

Geography and Policy

We also notice that in equation (1)-(3), the center-west dummy or the capital

dummy is not significant, which obviously differs from previous researches, such as

Bao et al. (2002), Ho and Li（2008). This is because with the spatial interaction among

cities controlled in our paper, no obvious growth disadvantages exist in western or

central regions, and no other obvious advantages in provincial capitals or

municipalities. In other words, spatial agglomeration in China’s urban system

contribute most to the interregional economic disparities of China

Though correlated with the dummy variables of open and SEZ, the dummy

seaport is significantly positive, while open and SEZ insignificant This implies the

possibility that geography is more important than policy for China’s urban economies.

But this need further study. Besides, the dummy, riverport, is insignificantly positive.

Other Economic Growth Factors

In our regressions, we also control other urban economic growth factors, the

impacts of which don’t vary significantly among equation (1)-(3), and most of factors

are insignificant except for the education. Education (edu) is significantly positive.

This is because education investment in the initial stage promotes the economic

28 / 36

growth in the long run.

Our regression results also imply that, the impact of investment (inve) is

insignificantly negative which may be because China cities with high investment have

no obvious economic advantages in long the long term, because, as a whole, Chinese

economy suffers from low efficiency spawned by over- investment (Zhang, 2003).

Some other factors, like labor (lab), government expenditure (gov), FDI (fdi),

urbanization (urb), population density (density, den_2), industrial structure (SM) all

have insignificantly impacts on urban economic growth in the long run in our

regression. There are three possible explanations: First, those factors have no

long-term impacts on urban economic growth; Second, the distributions of those

factors may correlate with geographic distances controlled by our model. Third, the

measurement errors of the variables may have reduced the significance level. Actually,

the t-statistics for the significance test of coefficients of labor and population density

are greater than 1.

Finally, the initial level of per capita GDP has an insignificant negative impact

on China’s urban economic growth, showing no significant trend of conditional

convergence.

VI. Robustness Checks

Several models based on equation (3) are estimated to test the robustness of our

key findings in Table 4. In equation (4)-(5), we change the definition of big cities to

test the hypothesis about the regional urban systems. We redefine big cities as the

provincial capitals and municipalities, thus, distance to the nearest big city is replaced

by distance to its provincial capital (distcap), therefore, municipalities do not have

any spatial interactions in this definition. The geographic distance is also measured

using China Map 2008 (Beijing Turing Software Technology Co., Ltd, China

Transport Electronic & Audio-Video Publishing House, 2008). We put only the linear

term of distance to its provincial capital in equation (4) and add the square term into

equation (5). To be consist with the variable we control, we drop the dummy samepro

and bigcity, and replace the variable gdpofbig by gdpofcap -- the measurement of the

29 / 36

economic scale of capitals, that is ln(GDP) in 1990 exclusive of agricultural output.

In equation (6), we add Tianjin into our definition of major ports and replace

distance to the nearer major port between Shanghai and Hong Kong (disport) by

distance to the nearest major port among Shanghai, Hong Kong and Tianjin

(disport*).

Time span is also an important issue for our study, since we highlight the role of

China’s openness and international trade in the shift of urban system. But lots of

information about cities in 1992 and 1993 are missing in Chinese Urban Statistical

Yearbooks (National Bureau of Statistics, 1993–1994). Therefore, we use the urban

data of 1994 and replace the dependent variable by the average annual growth rate of

real per capita GDP from 1994 to 2006 (dgdp*) deflated by provincial urban CPIs,

where agricultural output and population are also excluded. Another benefit from this

robustness check is the greater number of observations from 133 in equation (3) to

193 in equation (7). Results of robustness checks are in Table IV, and only variables

about spatial interaction are reported to save space.

Table 4: Robustness: Distance and Urban Economic Growth

(3) (4) (5) (6) (7)

dgdp dgdp dgdp dgdp dgdp*

distbig -0.00802** -0.00953** -0.00925*

(0.00352) (0.00430) (0.00540)

distbig_2 1.32e-05** 1.47e-05** 1.07e-05

(5.40e-06) (5.68e-06) (8.04e-06)

distcap -0.00748*** -0.0104

(0.00250) (0.00853)

distcap_2 7.05e-06

(1.98e-05)

gdpofcap -2.011*** -2.036***

(0.423) (0.430)

disport -0.0112** -0.00373 -0.00384 -0.0139***

(0.00429) (0.00250) (0.00253) (0.00527)

disport_2 1.28e-05*** 3.59e-06* 3.70e-06* 1.46e-05**

(4.82e-06) (2.10e-06) (2.13e-06) (6.19e-06)

disport_3 -4.17e-09** -8.54e-10* -8.76e-10* -4.28e-09**

(1.60e-09) (4.63e-10) (4.69e-10) (2.07e-09)

disport* -0.00881

30 / 36

(0.00554)

disport_2* 1.44e-05*

(7.55e-06)

disport_3* -7.42e-09**

(3.30e-09)

Other controls Y Y Y Y Y

Observations 132 132 132 132 192

R2 0.428 0.492 0.493 0.434 0.373

Notes: (1) Standard errors in parentheses;

(2) * p < 0.10,

** p < 0.05,

*** p < 0.01;

As shown in Table IV, in equation (4), we only add distance to the provincial

capital, which is significantly negative, suggesting the provincial capitals have strong

absorption effects on the surrounding cities. In equation (5), we add both the distance

to the provincial capital and its square, both of which are not significant. So provincial

capitals have only significant centripetal forces, which is not conducive to the

economic growth of remote cities. We think this centripetal force might be due to the

administrative factors by provincial government to promote economic agglomeration

of provincial economies. Besides, the larger initial economic scale of the provincial

capital the slower its urban economy grows, as gdpofcap is significantly negative.

This result shows a stronger absorptive effect for bigger central cities. Variables of

distance to major port are always significant.

In equation (6) with Tianjin as a major port of China’s urban system, distance

variables to major port are still significant, but the significance levels decrease, which

proves that the economic agglomeration of Tianjin is not as strong as Shanghai and

Hong Kong. And distance variables to big city are insignificant with the same sign as

in equation (3), which suggests that the importance and robustness of big cites and

regional urban systems are much less than major ports and the national urban system.

In equation (7) with the dependent variable changed, the significance of distance

variables to major port are almost the same as in equation (3), while the significance

of the distance to big cities decreased a lot. The shorter time span or fewer

observations could be the explanation. However, it’s also possible that ever since

China’s rapid openness to the international market in 1994, the agglomeration patterns

31 / 36

of China’s urban system change to meet the needs of developing international trade,

so that distance to major ports are increasingly important. However, to confirm this

possibility, we need further studies.

VII. Conclusions

This paper is an early attempt to verify the spatial pattern of China’s urban system

following the Core-Periphery Model. The nature of China’s geography and openness

as nature experiment makes China a feasible application and an experiment field for

testing the New Economic Geography theory of urban system.

The most important finding of this paper is to verify the “∽”-shaped non- linear

correlation between the geographical distance to major ports and urban economic

growth, which is consistent with the Core-Periphery Model of urban system in the

New Economic Geography theory. We find that China’s urban economies adjust to the

increasingly important international trade according to their access to global market,

especially after China’s drastic opening up since 1992.

Agglomeration shadow modeled by Krugman (1993) is also found in China’s

urban system, which means that being closer to the agglomeration centers is not

always good news for local economy. This finding adds a new evidence to solve the

paradox between the positive closeness-growth relationship in real data and the

theoretical hypothesis of “agglomeration shadow” (Partridge et al., forthcoming).

The "border effect" due to Chinese inter-provincial segmentation is equivalent to

increasing actual distance between cities, as well as limiting the spatial interaction

among China’s cities. Such administrative boundaries protect the economic growth of

small cities from the absorption of big cities in other provinces. Nevertheless, it also

leads to efficiency losses of the interregional agglomeration and scale economy.

As the first paper to study China’s urban system within the structure of

Core-Periphery Model, our exercises also direct some further empirical researches

based on China’s urban data.

In the modern era, agglomeration patterns can change with lower transportation

costs, improved communication technology, shifts in trade patterns, and industry

32 / 36

structural change (Partridge et al., forthcoming). Our paper emphasizes the role of

shifts in trade patterns, and our robustness check also suggests some evidence of

agglomeration pattern changes after 1994. However, without variables of

transportation costs and communication technology controlled, it’s hard to identify the

impact of international trade on agglomeration patterns, so further researches are

needed to convince the effect of international trade on spatial interaction patterns

among China’s cities.

33 / 36

References：

[1] Anderson, G. and Ying Ge, 2005, "The Size Distribution of Chinese Cities,"

Regional Science and Urban Economics, 35(6), 756-776.

[2] Au, Chun-Chung and J. Vernon Henderson, 2006a, “Are Chinese Cities Too

Small?” Review of Economic Studies, 73(3), 549-576.

[3] Au, Chun-Chung and J. Vernon Henderson, 2006b, "How Migration Restrictions

Limit Agglomeration and Productivity in China," Journal of Development

Economics, 80, 350-388.

[4] Baldwin, R. E., R. Forslid, P. Martin, G.I.P. Ottaviano and F. Robert-Nicoud, 2003,

Economic Geography and Public Policy, Princeton: Princeton University Press.

[5] Bao, Shuming, Gene Hsin Chang, Jeffrey D. Sachs, Wing Thye Woo, 2002,

“Geographic Factors and China’s Regional Development under Market Reforms,

1978–1998,” China Economic Review, 13, 89-111.

[6] Barro, Robert J., 2000, “Inequality and Growth in a Panel of Countries,” Journal of

Economic Growth, 5(1), 87-120.

[7] Black, Duncan and J. Vernon Henderson, 1999, “Spatial Evolution of Population

and Industry in the United States,” American Economic Review, 89, 321-327.

[8] Brülhart, M. and P. Koenig, 2006, “New Economic Geography Meets Comecon,”

Economics of Transition, 14(2), 245–267.

[9] Chen, Min, Qihan Gui, Ming Lu and Zhao Chen, 2007, “Economic Opening and

Domestic Market Integration,” in Ross Garnaut and Ligang Song (eds.), China:

Linking Markets for Growth, Asia Pacific Press, 369-393.

[10] Chen, Zhao, Yu Jin and Ming Lu, 2008, “Economic Opening and Industrial

Agglomeration in China,” in M. Fujita, S. Kumagai and K. Nishikimi (eds.),

Economic Integration in East Asia, Perspectives from Spatial and Neoclassical

Economics, Edward Elgar Publishing, 276-31.

[11] Clarke, G. R. G., 1995, “More Evidence on Income Distribution and Growth,”

Journal of Development Economics, 47(2), 403-427.

[12] Combes, Pierre-Philippe, 2000, “Economic Structure and Local Growth: France,

34 / 36

1984-1993,” Journal of Urban Economics, 47, 329-355.

[13] Dobkins, L. and Y. Ioannides, 2000, “Dynamic Evolution of the Size Distribution

of U.S. Cities," in J. M. Huriot and J. F. Thisse (eds.), Economics of Cities,

Cambridge: Cambridge University Press, 217-260.

[14] Dobkins, L. and Y. Ioannides, 2001, “Spatial Interactions among U.S. Cities:

1900–1990,” Regional Science and Urban Economics, 31, 701–731

[15] Fujita, M., 1988. “A Monopolistic Competition Model of Spatial Agglomeration,”

Regional Science and Urban Economics, 18, 87-124.

[16] Fujita, M., V. Henderson, Y. Kanemoto, and T. Mori, 2004, “Spatial Distribution

of Economic Activities in Japan and China,” in V. Henderson and J. F. Thisse (eds),

Handbook of Urban and Regional Economics, vol.4, North-Holland, 2911-2977.

[17] Fujita, M. and P. R. Krugman, 2004, “The New Economic Geography: Past,

Present and the Future,” Regional Science, 83, 139-164.

[18] Fujita, M., P. R. Krugman, and T. Mori, 1999a, “On the Evolution of Hierarchical

Urban Systems,” European Economic Review, 43, 209-251.

[19] Fujita, M. P.R. Krugman, and A.J. Venables, 1999b, The Spatial Economy: Cities,

Regions and International Trade, Cambridge, Massachusetts: The MIT Press.

[20] Fujita, M. and T. Mori, 1996, “The Role of Ports in the Making of Major Cities:

Selfagglomeration and Hub-effect,” Journal of Development Economics, 49,

93-120.

[21] Fujita, M. and T. Mori, 1997, “Structural Stability and Evolution of Urban

System,” Regional Science and Urban Economics, 27, 299-442.

[22] Fujita, M. and J. F. Thisse, 1996, “Economics of Agglomeration,” Journal of the

Japanese and International Economies, 10(4): 339-78.

[23] Glaeser, E., J. Scheinkman, A. Shleifer, 1995, “Economic Growth in a

Cross-section of Cities,” Journal of Monetary Economics, 36, 117–143

[24] Hanson, Gordon H., 1998, “Regional Adjustment to Trade Liberalization,”

Regional Science and Urban Economics, 28, 419-444.

[25] Hanson, Gordon H., 2001, “Scale Economies and Geographic Concentration of

Industry,” Journal of Economic Geography, 1, 255-276.

35 / 36

[26] Hanson, Gordon H., 2005, “Market Potential, Increasing Returns, and Geographic

Concentration,” Journal of International Economics, 67, 1 –24

[27] Ho, Chunyu and Dan Li, forthcoming, “Spatial Dependence and Divergence across

Chinese Cities,” Review of. Development Economic.

[28] Ioannides, Y. M. and H. G., Overman, 2004, “Spatial Evolution of the US Urban

System,” Journal of Economic Geography, 4, 131-156.

[29] Krugman, P., 1991, “Increasing Returns and Economic Geography,” Journal of

Political Economy, 99, 483-499.

[30] Krugman, P., 1993, “First Nature, Second Nature, and Metropolitan Location,”

Journal of Regional Science, 33, 129-44.

[31] Krugman, P., 1996, “Confronting the Mystery of Urban Hierarchy,” Journal of the

Japanese and International Economies, 10(4), 399-418.

[32] Krugman, P., and R. L. Elizondo, 1996, "Trade Policy and the Third World

metropolis," Journal of Development Economics, 49(1), 137-150.

[33] Overman, Henry G. and Y. M. Ioannides, 2001, “Cross-Sectional Evolution of the

U.S. City Size Distribution,” Journal of Urban Economics, 49, 543-566.

[34] Parsley, D., Mark, and S. Wei, 2001, “Explaining the Border Effect: The Role of

Exchange Rate Variability, Shipping Cost, and Geography,” Journal of

International Economics, 55 (1), 87 -105.

[35] Partridge, M. D., 1997, “Is Inequality Harmful for Growth? Comment,” American

Economic Review, 87, 1019-1032.

[36] Partridge, M. D., R. Bollman, M. R. Olfert and A. Alasia, 2007, “Riding the Wave

of Urban Growth in the Countryside: Spread, Backwash, or Stagnation,” Land

Economics, 83, 128-152.

[37] Partridge, M. D., D. S. Rickman, K. Ali and M. R. Olfert, forthcoming, “Do New

Economic Geography Agglomeration Shadows Underlie Current Population

Dynamics across the Urban Hierarchy?” Regional Science.

[38] Poncet, S., 2005, “A Fragmented China: Measure and Determinants of Chinese

Domestic Market Disintegration,” Review of International Economics, 13(3), 409

-430.

36 / 36

[39] Tabuchi, T., 1998, “Unban Agglomeration and Dispersion: A Synthesis of Alonso

and Krugman,” Journal of Unban Economics, 44, 333-351.

[40] Venables, A. J., 1996, “Equilibrium Locations of Vertically Linked Industries,”

International Economic Review, 37, 341-59.

[41] Wan, Guanghua, Ming Lu and Zhao Chen, 2006, “The Inequality–Growth Nexus

in the Short and Long Runs: Empirical Evidence from China,” Journal of

Comparative Economics, 34(4), 654-667.

[42] Wei, S., 1995, "The Open Door Policy and China's Rapid Growth: Evidence from

City-level Data," in Takatoshi Ito and Anne O. Krueger (eds.), Growth Theories in

Light of the East Asian Experience, NBER-East Asian Seminar on Economics,

vol.4, University of Chicago Press.

[43] Young, A., 2000, “The Razor’s Edge: Distortions and Incremental Reform in t he

People’s Republic of China,” Quarterly Journal of Economics, 115(4), 1091

-1135.

[44] Zhang, Jun, 2003, “Investment, Investment Efficiency, and Economic Growth in

China,” Journal of Asian Economics, 14(5), 713-734.