Industrial Agglomeration and Development: A Survey of ...hypothesized, but that genuine industrial clusters are more characteristic of sectors and spaces that have been most deeply

295

In this article, we explore the issue ofindustrial agglomeration and its relationshipto economic development and growth in theless-developed countries of East Asia, withspecial reference to China. Our approachinvolves two main lines of attack. First, wepresent theoretical arguments and secondaryevidence as to why we should have strongexpectations about finding a positive rela-

tionship between agglomeration andeconomic performance in the less-developedcountries of East Asia, as in other parts ofthe world. This phase of our work is basedon the emerging body of ideas that is nowoften loosely referred to as the “neweconomic geography” (cf. Scott 1988,2002; Krugman 1991; Porter 2001). Second,we engage in a variety of statistical

Industrial Agglomeration and Development:A Survey of Spatial Economic Issues in East Asia

and a Statistical Analysis of Chinese Regions

C. Cindy FanDepartment of Geography, University of California–Los Angeles,1255 Bunche Hall, PO Box 951524, Los Angeles, CA 90095-1524

[email protected]

Allen J. Scott Department of Geography, University of California–Los Angeles,1255 Bunche Hall, PO Box 951524, Los Angeles, CA 90095-1524

[email protected]

Abstract: In this article, we explore the issue of industrial agglomeration and itsrelationship to economic development and growth in the less-developed countriesof East Asia. We present theoretical arguments and secondary empirical evidenceas to why we should have strong expectations about finding a positive relationshipbetween agglomeration and economic performance. We also review evidence fromthe literature on the roles of formal and informal institutions in East Asian regionaleconomic systems. We then focus specifically on the case of China. We argue thatregional development in China has much in common with regional developmentin other East Asian economies, although there are also important contrastsbecause of China’s history of socialism and its recent trend toward economic liber-alization. Through a variety of statistical investigations, we substantiate (in part)the expected positive relationship between agglomeration and economic perfor-mance in China. We show that many kinds of manufacturing sectors are character-ized by a strong positive relationship between spatial agglomeration and produc-tivity. This phenomenon is especially marked in sectors and regions whereliberalization has proceeded rapidly. We consider the relevance of our commentsabout industrial clustering and economic performance for policy formulation inChina and the less-developed countries of East Asia.

Key words: industrial clusters, agglomeration, regional development, new economicgeography, East Asia, China.

#9905—ECONOMIC GEOGRAPHY—VOL. 79 NO. 3—79304-Fan_Scott

Economic Geography 79(3): 295–319, 2003.© 2003 Clark University. http://www.clarku.edu/econgeography

This research was supported by the World Bank. We express our thanks to Wenfei Wang for herable research assistance on this project; to Chase Langford for his cartographic work; and to GeorgeLin, Henry Yeung, and four anonymous referees for their comments on earlier drafts of the article.

296 ECONOMIC GEOGRAPHY JULY 2003

investigations that seek to document andsubstantiate the expected positive relation-ship between agglomeration and economicperformance. Our discussion is focusedspecifically on an examination of thespatial distribution of manufacturing activityin China.

Most of the literature on industrialagglomeration and economic performancehas been concerned with empirical situa-tions in the advanced Western capitalistworld. In this article, we hope to demon-strate not only that the cluster approach isas useful in the East Asian context as it isin Western situations, but that it can alsoshed important light on critical dilemmas ofdevelopment that are specific to non-Western countries. We emphasize thattheory needs to address both the economicbases of regional development and socio-cultural factors, such as formal and informalinstitutions. In China, in particular, a historyof central planning and the recent transitionfurther complicate the applicability ofeconomic theories that are based primarilyon Western experiences. In this article, weshow that in China, the positive relationsbetween industrial agglomeration andeconomic performance are generally ashypothesized, but that genuine industrialclusters are more characteristic of sectorsand spaces that have been most deeply trans-formed by economic reforms and marketorientation.

In the first half of the article, we lay outa brief statement on the relations betweenindustrial agglomeration and economicperformance and consider the ways in whichthese relations throw light on the East Asiancase, and vice versa. In the second half, wefocus on regional development in China, firstby giving a short account of how recent polit-ical economic changes have affected thespace-economy and then by engaging in anumber of statistical exercises to demon-strate that there is a significant link betweenagglomeration and economic performancein Chinese regions. Unfortunately, empir-ical work of this sort commonly encounterssevere constraints resulting from the defi-ciencies of available industrial data. Our

statistical analyses accordingly fall short of afull-blown test of the theoretical ideas laidout in the first half of the article. Nevertheless,in so far as they go, the analyses are consis-tent with our main conceptual claims, andthe results provide an encouraging basisfor moving forward into new rounds ofstatistical research in the future.

Locational Agglomeration andRegional IndustrialPerformance

The notion of the region as a nexus of crit-ical developmental and growth processes haslong been familiar to heterodox economists,such as Hirschman (1958), Jacobs (1969),Kaldor (1970), Lampard (1955), Myrdal(1959), and Perroux (1961), who took exter-nalities and increasing returns seriously—inpart because they explicitly or implicitlyrecognized the importance of the intellec-tual legacy of Marshall (1890, 1919)—andwho saw that one major expression ofthese phenomena can be found in local-ized complexes of economic activity. Sincethe early 1980s, an enormous surge in theliterature in economic geography andallied fields has greatly expanded on thisearlier work. All this literature has repeat-edly suggested that selected regions—espe-cially those in which industries are organizedin transactions-intensive networks—arecapable of exerting powerful push effects onnational economic development.

Many kinds of industrial systems are trans-actions intensive both in a static sense (trans-actions are numerous and finely grained)and in a dynamic sense (transactions aresubject to constant flux because of rapidlychanging production arrangements andmarket instabilities). These features are espe-cially marked today in industries like elec-tronics, communications equipment, special-ized machinery and components, toys,watches, garments, furniture, software,and business services, where many smallestablishments with narrowly defined corecompetencies perform critical functionswithin constantly shifting and information-


VOL. 79 NO. 3 INDUSTRIAL AGGLOMERATION AND DEVELOPMENT 297

intensive systems of interlinkages. Mutualproximity is often crucial to the success ofthe individual establishments that makethese industries up, partly because it reducesthe costs of transacting and partly becauseof a series of other factors that intensify local-ized positive externalities. Benefits ofthese sorts are identifiable more generallyas agglomeration economies, although incertain cases, agglomeration diseconomiescan also appear as regions grow in size.Agglomeration effects, in turn, are oftencategorized as so-called localizationeconomies (i.e., efficiency-boostingphenomena that come from the clusteringof firms in a given sector) and as urbaniza-tion economies (i.e., efficiencies that resultfrom the agglomeration of many differentkinds of activities in a given region). In thevocabulary of more recent analysts, theseeconomies are often referred to, respectively,as Marshall-Arrow-Romer externalitiesand Jacobs externalities (e.g., Beardselland Henderson 1999).

In practice, these types of externalities canbe further broken down into more detailedcategories. The following five points summa-rize the main issues:

1. Interfirm transactions that are small inscale, variable in content, and subjectto frequent readjustment usually incurhigh costs per unit of distance. Themutual proximity of firms in networksmade up of transactions like these is animportant factor in preventing costs fromspiraling out of control and reducing therisks of any failure to establish promptinterfirm contact (Scott 1988).

2. Dense local labor markets representspatial concentrations of job seekers andjob vacancies, and high levels of mutualproximity make it relatively easy toacquire, process, and act on informationabout relevant opportunities.

3. Transactional relations also involve flowsof certain kinds of business informationor knowledge spillovers. Untraded inter-dependencies of these sorts are all themore important because they tend tounderpin many small-scale processes of

learning and innovation whose cumula-tive effects greatly reinforce localcompetitive advantages.

4. The clustering of many differentproducers can significantly enhance theformation of beneficial business alliancesand organizations that help to augmentlocal competitive advantage. Equally,agglomeration promotes the develop-ment of distinctive business cultures inparticular places, thus facilitating thetasks of interfirm communication andunderstanding.

5. Significant economies can be obtainedwhen the consumption of infrastructuralartifacts is spread out over many indi-viduals in any one place. Large localizedclusters of firms and workers make itpossible to construct disproportionatelydense and rich infrastructures with manypositive effects on local competitiveadvantage.

Regions that have some or all these attrib-utes, combined with dynamic learning capac-ities, stand some chance of becoming signif-icant articulations of value-adding activityand entrepreneurial energy in the new globaleconomy. They are apt to exhibit strongincreasing-returns effects that lead to ever-superior competitive advantages (asexpressed, for example, by rising quality/costratios for final products) as they grow. Inpractice, given statistical measurement diffi-culties, these effects are typically analyzedby economists in the form of aggregateproductivity levels per worker (operationally,value-added per worker), and in the analysispresented here, we pursue the same strategy.Much of the econometric work in this regardhas been focused on cases in the more devel-oped countries (see, e.g., Shefer 1973;Kawashima 1975; Sveikauskas 1975; Carlino1979), and only in recent years has therebeen a concerted attempt to apply similarapproaches to regions in less-developedcountries (e.g., Henderson 1986, 1988; Y.Chen 1996; Shukla 1996; Lee and Zang1998). We build on this latter literature witha detailed multilevel analysis of regionaldevelopment across China as it moves



from central planning to an economic orderthat is increasingly influenced by liberal-ization policies.

At the same time, current theories ineconomic geography pay attention not onlyto the economic foundations of industrialagglomerations, but also to the importantrole played by institutional factors inpromoting localized growth and develop-ment. It is generally acknowledged that thesefactors tend to be extremely place specificand often vary greatly from one location toanother (Amin and Thrift 1995; Storper andScott 1995). In most East Asian countries,formal and informal institutions have beencritical in shaping regional economicoutcomes. In addition, governments haveplayed a notably directive role in assigninginvestment to different locations and insetting up export-processing zones and otherlocal development schemes. China presentsan especially interesting case in this regardbecause of its history of centralized economicplanning. An important research andpolicy question raised by any investigationof industrial districts in China is whetheragglomerations that begin as governmentalprojects are able to develop into genuinegrowth centers with a strong endogenousdynamic of growth. In our empirical analysis,we show that viable industrial regions havemade their historical and geographic appear-ance in those parts of China where economicliberalization has been most prominent, butthat the record is much less positive in otherparts of the country. Informal and trust-based relationships, too, are important inChina, all the more so since economic liber-alization has relied heavily on ethnic andkinship ties with Hong Kong, Taiwan, andoverseas Chinese.

Geographic Agglomeration inthe Less-Developed Countriesof East Asia

Although the phenomenon of industrialagglomeration has been well documented inthe advanced economies, including Japan(e.g., Nakamura 1985; Patchell 1993;Kanemoto, Ohkawara, and Susuki 1996;

Fujita and Tabuchi 1997), our concern hereis with the less-developed parts of East Asia.As such, our effort can be seen as a contri-bution to knowledge about agglomerationin the non-Western world in general, and inlow- and middle-income countries in partic-ular (see also Nadvi and Schmitz 1994; andScott 2002).

In these countries, dense industrial clus-ters derive from both local entrepreneurialefforts and foreign direct investment. Innumerous cases, clusters can be shown tohave their roots in traditional artisanal formsof production, as in the gem and jewelryindustry of Bangkok, the furniture industryof the Philippines, the pottery and silk indus-tries in China, and the shoe industry of Agrain India. At the same time, and especially inthe case of East Asia, we need to remainalert to the political context and the actionsof governments at various levels of authorityin molding regional patterns of development.In East Asia, moreover, as in the economi-cally developed parts of the world, large city-regions are the locations of many of the mostvibrant industrial districts. Singapore, HongKong, Seoul, Beijing, Shanghai, and KualaLumpur, for example, harbor many special-ized industrial districts that draw on denselocal supplies of skilled labor, infrastructure,educational and research facilities, and soon. In South Korea, high-technology firmshave been shown to prefer the Seoul metro-politan region because of high levels ofaccess to technical labor (Park 1994). Inaddition, foreign investors tend to favormetropolitan areas that offer a diversity ofinfrastructural and input services, as wellas a predictable regulatory environment(Scott 1987; Leung 1993, 1996; Wong andGoldblum 2000). The largest of Asia’s city-regions sometimes exhibit agglomerationdiseconomies because of congestion,pollution, crime, and so on, although thisproblem is not absolute, in the sense thatappropriate policy and planning interven-tions can invariably clear away some of thebarriers to further growth. In any case, farfrom seeing large-scale urbanization as anaberration, as some development theoristshave done in the past (e.g., Lipton 1977),our argument is that it is one of the impor-



tant channels through which agglomera-tion economies are achieved and acceler-ated development occurs. Smaller urbancenters, too, can play significant roles in thedevelopment process, especially in Asiawhere many traditional craft industriesoutside the main metropolitan areas arebeginning to show signs of modern entre-preneurial energy and orientation to widermarkets. We now examine more preciselyhow these relationships are worked out onthe ground.

Industrial Linkages, Subcontracting,and Flexible Production

There is much evidence on the impor-tance of industrial linkages as a factor inagglomeration processes in East Asia. Forthe case of South Korea, Park (1991) arguedthat polarized regional economies with adiversity of industries, information, and tech-nical workers generate innovative firms andspin-offs, which, in turn, encourage furthergrowth. Park (1993, 1994) further noted thatvertical disintegration and the clustering ofsmall plants increase the productivity oflabor-intensive industries and that subcon-tracting is an important organizationalstrategy for lowering wages and deflectingindustrial disputes. Similar arguments havebeen made for the case of Taiwan, wheresubcontracting networks and cluster-basedmanufacturing were found to be importantstrategies for increasing the competitivenessof firms (Shieh 1990; W.-H. Chen 1999).Focusing on Taiwan’s footwear industry,Levy (1991) contended that the transactions-costs hypothesis provides a powerful expla-nation for the emergence of localizedsubcontracting networks.

Detailed empirical studies on industriallinkages in the more peripheral parts of EastAsia are relatively scarce. In a study of thesemiconductor industry in Southeast Asia,Scott (1987) showed that production unitsin Manila’s semiconductor complex in themid-1980s were clustered close to oneanother and were intricately linked tominimize transactions costs. Scott (1994)demonstrated that the gem and jewelryindustry of Bangkok has been able to

thrive on the basis of its vibrant transactionalnetworks, its low wages, and the skilful polit-ical maneuvering of its representatives.Rasiah’s (1994) study of the machine-toolindustry in Malaysia underscored theconnection between subcontracting andlocalization. A number of studies onIndonesia have highlighted positive exter-nalities and productive effects of industrialclusters (Poot, Kuyvenhoven, and Jansen1990; Smyth 1992; Sandee, Rietveld,Supratikno, and Yowono 1994; Nadvi andSchmitz 1994; Sandee 1995; Weijland 1999;Sandee and Rietveld 2001).

Learning and Knowledge

Many studies have pointed to the impor-tance of the relation between learning andproximity in the cities of developing coun-tries in general, and East Asia in particular.At the same time, formal learning institu-tions, such as universities and research infra-structures, are located mostly in large urbanareas in Asia. In Taiwan, the close spatialrelations between the Industrial TechnologyResearch Institute and firms that produceintegrated circuits in the Hsinchu Science-Based Industrial Park are widely regardedas a major factor in the latter’s success(Chang and Hsu 1998). In a comprehensivearticle, Bell and Albu (1999) argued that“knowledge systems” are central to the long-term dynamism of industrial clusters andstressed that the diffusion of innovationwithin clusters, as well as the openness ofclusters to flows of knowledge from outside,are important in developing countries. Liu(1998) concluded that learning capabilityand human capital will determine thedurability of Taiwan’s industrial success.Concomitantly, the absence of these quali-ties in any region means that barriers todevelopment are likely to be rapidly encoun-tered, even if the region possesses otherkinds of assets. Sandee (1995) demonstratedhow collaboration in a rural industrial clusterin Indonesia fosters the diffusion of inno-vation and the deepening of technologicalcapabilities. Johansson and Nilsson (1997)showed that clustering in export processingzones in Malaysia stimulates local firms to



learn from foreign investors how to produce,market, sell, and distribute manufacturedgoods on the world market.

Cluster Development and theInternational Division of Labor

For the most part, East Asia has been atthe receiving end of the decentralization ofmanufacturing production from advancedindustrialized economies to developingcountries. For example, until the mid-1970s,the four newly industrializing economies ofSouth Korea, Taiwan, Hong Kong, andSingapore accounted for the bulk of theoffshore assembly work of U.S. electronicscorporations (Scott 1987). As theseeconomies developed, branch plants insearch of cheap labor moved to yet moreperipheral parts of East Asia, includingMalaysia, the Philippines, Indonesia,Thailand, and China (Tsay 1993; Ho 1994;Park 1994; Tang 1996; Eng 1997; Wong andGoldblum 2000). In the same way, consid-erable subcontracting now occurs fromNorth America and Western Europe to EastAsian producers in such sectors as clothing,furniture, and machinery. It is importantto note that large urban centers in East Asiaare especially attractive as locations for thiskind of manufacturing decentralization. Suchdevelopment typically draws on the cheap,surplus labor in cultural and political settingswhere workers are willing to tolerate longhours, accelerated rhythms of work, andround-the-clock manufacturing schedules.As manufacturing decentralization to EastAsian cities occurs, moreover, local inputindustries often spring up in the local envi-ronment, providing backup services andstrengthening local agglomerationeconomies.

Institutions and Regulation inEast Asian RegionalDevelopment

Formal and informal institutions play aprominent role in the structuring of EastAsian economies. Indeed, as we have indi-

cated, many proponents of the neweconomic geography have sought to extendthe theory of agglomeration by emphasizingthe institutional bases of regional economicdevelopment, particularly the role of publicagencies in regulating local market failuresand other impediments to rapid forwardmomentum (e.g., Peck 1992; Amin 1999).This approach has been greatly influencedby institutional and evolutionary economicsand economic sociology, which stress thateconomic activity is always sociallyembedded and therefore context specificand path dependent (Granovetter 1985;Hodgson 1988).

Governance Structures and RegionalDevelopment

East Asian governments have had signif-icant impacts on economic developmentthrough the adoption of export-orientedindustrialization policies and the establish-ment of industrial parks and export-processing zones. Kaohsiung in Taiwan,Masan in South Korea, Penang and Johorein Malaysia, and Shenzhen in China wereamong the first generation of government-sponsored industrial zones for fosteringexport-oriented industrialization. Yuan andEden (1992), in their comparative study ofTaiwan, South Korea, and China, under-scored the role of governmental policy indetermining the performance of export-processing zones. Governments in Taiwanand Singapore have played particularly crit-ical roles in fostering technological advance-ment in these zones (Yuen 1991; Ho 1994;Xue 1997).

In many cases, however, government-initi-ated development zones have not yet reallytaken off, in the sense that they generatelimited agglomeration economies. Incontrast to the Marshallian formulation, inwhich endogenously driven growth and localembeddedness sustain agglomerations, theseclusters are more often than not “hub andspoke” or “satellite platform” industrialdistricts (Markusen, Lee, and DiGiovanna1999). For example, in South Korea, Pohangis anchored by a steel company, and Ulsan



is anchored by a petrochemical plant and anautomobile plant, whereas Kumi,Changwon, and Ansan are satellite districtswith branch plants whose external linkagesare mostly extralocal (Markusen and Park1993; Park 1996). Park (1993) observed thatthe South Korean government has increasedsupport for collaboration between large firmsand small or medium-sized firms so thatsatellite clusters can benefit from the forma-tion of local networks and linkages.Furthermore, Park (1996) showed that asfirms begin to lose the advantages thataccrue from governmental incentives, theytend to pursue strategies such as theformation of local linkages, cooperative inter-firm relations, and contracting-out activities,while spin-off activities and new start-upsalso tend to increase (e.g., in TaeduckScience Town). Through this process, Parkcontended, initial satellite and hub-and-spoke industrial districts begin to evolvetoward systems marked by local networksand embeddedness, as illustrated by the caseof Kumi.

Informal Bases of the Local Economy

Trust-based business relationships are ofgreat significance in many East Asiancultures and are reflected not only in localtransactions, but also in the intricate tiesbased on the Asian diaspora that existbetween East Asia and other parts of theworld (Hamilton 1991; Smart 2000; Yeungand Olds 2000). Since the 1990s, anemerging body of work on East Asia hassought to highlight the role of interpersonalrelations and ethnic and kinship ties inregional economic development. Thus, Olds(2001) argued for a relational geography thatfocuses on networks and flows, linkages,interdependence, connections, and mutu-ality. This kind of relational geography ischaracteristic of “Chinese capitalism” and ofthe operations of large Chinese-controlledinternational conglomerates.

The importance of interpersonal relationsin East Asia further reinforces the notionthat economic behavior is embedded, in part,in social networks that are defined by mutu-

ality, trust, and cooperation. Ties of prox-imity and association are especially impor-tant in promoting learning effects and thediffusion of know-how (Amin and Thrift1995; Sunley 1996; Storper 1997). Face-to-face contacts, in particular, help to engenderrelational assets or untraded interdepen-dencies in the form of information, reputa-tional capital, economically useful sensibil-ities and forms of habituation, and so on.Economic agents and firms in regions withrich traditions of reciprocity benefit fromthe sharing of know-how and expertise,and by doing so, they increase the local stockof competitive advantages (Cooke andMorgan 1998). Transactions, therefore, areto be evaluated not only in terms of theircosts, but also in terms of their qualitativesocial characteristics (Harrison 1992).

Geographic Agglomerations inChina

China has much in common with otherEast Asian economies, from the many bigcities scattered across its landscape to theprominent roles of governments and socialnetworks in economic life. However, itsdevelopment history has also been heavilyshaped by socialist ideology and, morerecently, by the trend toward economicliberalization.

China has a mixed economy, one in whichthe government’s role and institutional lega-cies from the former command economyhave had profound effects on the geographyof production. Prior to the 1980s, theChinese government’s blueprint for nationaleconomic development, with its stress onlarge-scale vertically integrated units ofproduction, was not especially conduciveto the formation of dynamic industrialdistricts rich in positive externalities. Thesocialist economy was focused on state-owned enterprises, domestic investment andmarkets, and a locational rationale thatemphasized political goals more thaneconomic efficiency. From the 1950s tothe early 1970s, the Chinese governmentdeliberately discouraged investment in



coastal areas and promoted the growth ofinland cities. It also selected remote inlandsites for key sectors, such as cars and ironand steel, but discouraged spatial clustering(Huang 1999; Naughton 1988). Sit and Lu(2000) showed how the geography of China’scar industry is still strongly influenced bythese past locational decisions. This influ-ence is reinforced by regulatory structuresthat restrict foreign investments in the carindustry to joint ventures with domesticfirms.

Since the economic reforms of the 1980s,however, the Chinese government hasactively pursued alternative policies that aredirected to economic liberalization. Itpromoted export-oriented industrializationprograms by designating special economicand open zones, most notably in the easterncoastal zone (Fan 1995, 1997). New policyincentives, infrastructural investments, andsocial relationships that include ties withoverseas Chinese (Hsing 1996; Leung 1993)have paved the way for the eastern coastalzone to be integrated quickly with Westerncapitalist production. Fan (2001, 2002)contended that the government has usedsocialist-type policies to foster rural-urbanmigration and, by doing so, has engendereda migrant labor regime and increased thesupply of cheap rural labor to coastal areas.In contrast to industries that are heavilyinfluenced by Maoist legacies, the consumerelectronics and garments industries, whichtend to dominate in these areas, are char-acterized by high levels of foreign invest-ment, export orientation, and functional flex-ibility. We argue that it is in these spaces,where economic reforms and market orien-tation are the most deeply rooted, thatagglomeration economies appear most force-fully in China.

Some government-sponsored industrialareas in China’s open economic zonesinitially resembled satellite platforms ofthe South Korean type, where many firmswere tied to overseas markets but had onlyweak local linkages. Tong and Wang’s (2002)study of the personal computer industry inDongguan, Guangdong, for example,showed that during the 1980s, printed-circuitassembly plants with foreign capital invest-

ment depended on parts brought fromTaiwan. Since the mid-1990s, however,strong local networks underpinning theproduction of printed circuits have emerged,with interrelated factories constituting agrowing spatial agglomeration. As a result,a complete computer can now be assembledfrom parts made entirely in factories locatedwithin 50 kilometers of one another. Park(1996) also noted that small firms in satel-lite industrial districts in China, such asShenzhen, are becoming increasingly inte-grated into local network structures.

In contrast to the socialist period, whenthe government had strict control overeconomic activities, economic liberalizationsince the 1980s has boosted the role of themarket and has been conducive to industriallinkages that respond to market signals.Perhaps the most vivid example is the hosierycluster in Zhejiang province, where morethan 10,000 village households in 120 villagesin and near Datang have become China’sbiggest center of the production of socks andstockings and cluster development is nowstrongly under way (Wang 2001). Eachenterprise specializes in a particular aspectof production, including manufacturing,machinery, marketing, and services, and thisnetworked organization is considered to bethe primary reason for the success of thecluster (Guo and Cai 2000). The intricatedivision of labor and face-to-face networkingin the system enable enterprises to respondquickly to changes in the market and toensure high levels of efficiency and mini-mization of risks. This system is stronglyembedded in local social relationships in thatinterfirm linkages depend on trust and socialcapital. Wang (2001) argued that the closeand specific relations among enterprises, thelow barriers to entry and exit, and externaleconomies that are due to interlinkagescreate a situation that is similar to the ThirdItaly model of flexible specialization.Similarly, flexible production systems in thegarment, textile, toy, and footwear/leatherindustries in south China permit fastturnover and quick response to marketchanges—key factors in firms’ competi-tiveness in the global market (Christersonand Lever-Tracy 1997). Also prominent in



south China are firms in the computerindustry that have attained to high levels ofcompetitiveness because agglomerationenables them to ensure that their productsare high quality through rapid access toinformation and inputs (Tong and Wang2002). Even many small rural areas in thePearl River Delta have grown considerablybecause of subcontracting orders from over-seas. Here, as in some other parts of EastAsia, industrialization has blurred the distinc-tion between the urban and the rural, givingrise to the desakota phenomenon describedby McGee (1991) and Lin (2001).

Another case is the agglomeration of high-technology firms in Zhong’guancun, Beijing,which benefits not only from a local poolof skilled labor, but also from the proximityof a large number of research establish-ments, such as Beijing University andQinghua University (Wang and Wang 1998).Many firms in Zhong’guancun are spin-offsfrom state-owned enterprises and continueto receive strong support from parent insti-tutions or universities. Wang (2001)observed that a club culture has emerged inthe area, which further stimulates interac-tion among entrepreneurs, managers,engineers, and professionals and promotesa climate of entrepreneurship, innovation,and risk taking. This case suggests that locallyembedded cultures, norms, and social rela-tions are major factors in the dynamism ofagglomerations in China.

Similarly, the strong ethnic and kinshipties of China with Hong Kong, Taiwan,and overseas Chinese in other countries haveheavily shaped spatial patterns of foreigndirect investment and have been a crucialfactor in the development of new indus-trial agglomerations in the eastern andsouthern coastal areas. Labor-intensiveindustrial clusters in Fujian and the PearlRiver Delta in Guangdong, for example, arestrongly dependent on such ties. Much ofthe growth of these regions has been dueto the relocation of manufacturing produc-tion from Taiwan and Hong Kong and tosubcontract orders from manufacturers ofChinese origin in the United States (cf.Bonacich and Appelbaum 2000). Leung(1993, 1996) stressed the same point in an

analysis of subcontracting activities in thePearl River Delta that highlighted the roleof cultural and kinship ties between distantinvestors and local producers (see also Hsing1995, 1996, 1998). Christerson and Lever-Tracy (1997), who also focused on Fujianand the Pearl River Delta, argued that thefamily and social relationships of overseasChinese investors have been of major impor-tance in the growth of the garment industryin south China.

Y. Chen (1996) made a major econometriccontribution to the study of industrialagglomeration in China. He estimatedsectoral value added as a function of capital,labor, and a regional factor multiplier thatconsists of five components, includingagglomeration. The impact of agglomerationon productivity was found to be positive andhigh for the machinery industry but lowerfor less technology-intensive industries suchas food manufacturing. In the next section,we propose a framework similar to Chen’sand on this basis seek to push the study ofindustrial agglomeration in China forwardone more notch. Even though our analysiswas greatly hampered by the lack of data,we have nevertheless been able to bring avariety of different perspectives and scalesof analysis to bear on the central problem.

Agglomeration and IndustrialProductivity in China

We begin our analysis by probing gener-ally into the relations between industriallocation and productivity in China. One wayof measuring the broad locational charac-teristics of an industry is to compute aHerfindahl index (hereafter H-index) toassess the industry’s overall level of spatialagglomeration or dispersal. Thus, for Sectori, let pij be the proportion of total activitylocated in the jth region or province. Theterm pij is defined more explicitly as xij / Xi,where xij is the amount of activity in industryi in province j and Xi is the total amount ofactivity in industry i in China as a whole (i.e.,Xi = Σjxij). The H-index for Sector i is thencomputed as Hi = Σjpij

2. When all the activityin Sector i is concentrated in one province,the index is equal to one; when all activity



is evenly dispersed, the index convergesasymptotically to zero for a sufficiently largenumber of provinces; intermediate locationalpatterns are represented by values of theindex between these two extremes. Wecomputed two sets of H-indexes for all two-digit manufacturing sectors in China (exceptarmaments, for which the published statis-tics are heavily censored), using the numberof establishments and employment as rawdata inputs. Data for this purpose were takenfrom the statistical yearbooks of the Chineseprovinces for 2000. In most instances, thenumber of establishments and employmentfigures for two-digit sectors could be

obtained for all 31 provinces in China,although for 10 sectors, data were availablefor only 30 provinces, and for 2 sectors, theywere available for 29 provinces. The qualityof these data leaves much to be desired, butthe results we obtained appear, with certainreservations, to be reasonably reassuring.

Table 1 shows H-indexes for this exercise,ordered from high to low values. The indexfor establishments is usually smaller than theindex for employment, a statistical effect thatseems to arise because of the occurrenceof large establishments in many sectors (andall the more so given the labor-hoardingpropensities of state-owned enterprises in


Table 1

H-Index Values for Two-Digit Industrial Sectors in Chinese Provincesa

Values of H-Index

Industrial Sector Establishments Employment

Stationery, education, and sporting goods 0.1381 0.3316Electronics and telecommunications 0.1274 0.1970Furniture manufacturing 0.1259 0.1183Garments and other fiber products 0.1225 0.1849Metal products 0.1182 0.1074Instruments, meters, educational and office machinery 0.1180 0.0926Leather, furs, down, and related products 0.1044 0.2612Chemical fibers 0.1016 0.1082Other manufacturing 0.0974 0.1551Electric equipment and machinery 0.0959 0.1178Plastic products 0.0933 0.1476Textiles 0.0885 0.1095Printing and record pressing 0.0859 0.0770Ordinary machinery manufacturing 0.0846 0.0925Petroleum processing and coking products 0.0800 0.0977Special-purpose equipment 0.0763 0.0877Rubber products 0.0755 0.0925Transportation equipment manufacturing 0.0721 0.0689All manufacturing 0.0676 0.0810Papermaking and paper products 0.0646 0.0934Chemical raw materials and chemical products 0.0639 0.0742Timber processing, bamboo cane, palm fiber, and straw products 0.0592 0.0798Smelting and pressing of nonferrous metals 0.0585 0.0658Tobacco processing 0.0574 0.0828Nonmetallic minerals 0.0560 0.0756Food manufacturing 0.0548 0.0784Smelting and pressing of ferrous metals 0.0544 0.0910Food processing 0.0507 0.0869Medical and pharmaceutical products 0.0468 0.0638Beverage manufacturing 0.0450 0.0778

a Analysis based on data from Chinese provincial yearbooks.


China), thus causing a spurious incrementto apparent agglomeration when the employ-ment measure is used. It is spurious, at least,in the sense in which the term agglomera-tion is used in this investigation (i.e., to desig-nate a dense geographic clustering of func-tionally interrelated producers orestablishments, in contrast to large concen-trations of employment that are due simplyto the existence of one or two giant plants).However, the simple correlation betweenthe indexes for establishments and employ-ment is 0.73, which means that each canbe approximately substituted for the other.

The information in Table 1 provides thefirst analytical glimpse of the geography ofmanufacturing industries in China. At thetop of the list are sectors that are relativelyagglomerated; at the bottom are industriesthat are much more dispersed. A rough firstassessment of the main structural featuresof these two groups suggests that the formercomprises relatively labor-intensive small-plant industries with a proclivity to formvertically disintegrated networks (electronics,furniture, garments, and so on); the latterseems to be much more focused on capital-intensive, large-plant sectors, often with amarket- or resource-oriented locational logicand with significant levels of state owner-ship (smelting and pressing of ferrous metals,nonmetallic minerals, beverages, and so on).Note, however, that food manufacturing andfood processing, which have relatively lowH-indexes, are actually labor intensive interms of their actual capital-labor ratios.Presumably these industries are at leastoriented, to some degree, to more dispersed,rural locations.

In an attempt to flesh out these firstimpressions, we examined the relationsbetween various measures of industrial struc-ture and the H-indexes presented in Table1. Given the small number of two-digitsectors (which constitute the observationsin this analysis) and the paucity of availabledata, this exercise proved to be of limitedvalue, but one set of simple regressionsemerged as being of particular interest. Therelevant information is displayed in Table 2,where H-indexes for establishments andemployment are regressed against an inde-pendent variable (K/L), defined opera-tionally as total gross investment per worker,measured in units of 100 million yuan. Allvariables are transformed into natural loga-rithms. The regression coefficient attachedto K/L is negative and highly significant. Thisfinding confirms the idea that labor-inten-sive producers tend to be relatively agglom-erated (despite the anomalous cases of foodmanufacturing and food processing), whilecapital-intensive producers are relativelydispersed in their overall locational struc-ture. From what we know of industrial organi-zation processes in general, we speculatethat there is some likelihood in principle thatproducers in the former group are charac-terized by tighter, more variable, and morefinely grained interindustrial networksthan are those in the latter group. As ithappens, both H-indexes are also negativelyand significantly correlated with average sizeof establishment, though the weight of thisvariable is entirely swamped when we alsoadd K/L to the analysis as an independentvariable. Clearly, there are likely to be manyother forces at play in structuring locational


Table 2

Regression Analysis of H-Indexes for Two-Digit Industries in Chinese Provinces

Values of Regression Statistics

H (Establishments) H (Employment)

Constant –2.6192 –0.3584Coefficient of K/L –0.2214** –0.2619**Adjusted R2 0.2323 0.2256

Note: Dependent variables are the H-indexes.** p < .01.


outcomes in these industries (including, inChina, forms of ownership), but what is ofinterest at this point is that our results so farsuggest that in a general sense, differentialpatterns of agglomeration and dispersal may,indeed, have some fundamental relationshipto the shape and form of industrial activityas we proposed earlier. We now need toinquire more explicitly whether there areproductivity effects that flow from thelocational structure of Chinese industry asexpressed in the H-indexes. Productivity ismeasured as value added per worker, andthe hypothesis to be tested is that agglom-eration (for all the complex reasons discussedearlier) has a significantly positive impact onthis variable.

To begin with, we conducted a statisticalanalysis based on a simple production-func-tion approach (with two-digit industries asobservations), as indicated in Table 3. Thedependent variable is defined as VA/L(i.e., value added divided by employment),and the independent variables are K/L andthe H-index. In Regression 1, the H-indexis defined for establishments; in Regression2, it is defined for employment. Again, allvariables are transformed into natural loga-rithms. As in Table 2, both regressions runmore or less parallel to one another. In eachcase, there is a (perhaps suspiciously) lowbut positive and extremely significant coef-ficient attached to K/L. The H-indexesperform as expected. They indicate thatsignificant productivity effects emerge as

industries become increasingly clustered. Ifwe add average size of establishment to thesetwo regressions, we find that the associatedcoefficient is positive but not significant.However, the simple correlation betweenaverage size of establishment and K/L is0.61, a finding that encourages us to inferthat agglomerated industries are marked bycomparatively smaller plant sizes on averagethan are more dispersed industries. In otherwords, despite the admittedly crude andaggregate nature of the analysis thus far, wesuggest that agglomerated industrial clus-ters in China tend to be made up of small,labor-intensive establishments and that itis these kinds of industries whose produc-tivity is enhanced by the convergence ofproducers around their own centers ofgravity.

A More Disaggregated Analysis

The regressions presented in Table 3represent an extremely aggregated level ofinvestigation. It is therefore important toinquire whether similar relationships can befound using a more spatially disaggregatedapproach. To do so, we evaluated modelsbased on a production-function approach fortwo-digit sectors with the observationsdefined in terms of individual provinces.

A synoptic view of the results of thisapproach is presented in Table 4 for 29 two-digit manufacturing sectors. The dependentvariable is VA/L as previously defined. The


Table 3

Regression Analysis of the Relations between Productivity and Locational Structure forTwo-Digit Industries in Chinese Provinces

Values of Regression Statistics

Regression 1 Regression 2

Constant 2.6729 3.1164Coefficient of K/L 0.3231** 0.5067**Coefficient of H 0.3828* 0.5691**Adjusted R2 0.4131 0.5123

Note: VA/L is the dependant variable in both regressions.* p < .05, ** p < .01.



Table 4

Industrial Productivity in Chinese Provinces by Two-Digit Sectors, 2000: RegressionResultsa

Sector

Food processingFood manufacturingBeverage manufacturingTobacco processingTextilesGarments and other fiber

productsLeather, furs, down, and related

productsTimber processing, bamboo cane,

palm fiber, and straw productsFurniture manufacturingPapermaking and paper productsPrinting and record pressingStationery, education, and

sporting goodsPetroleum processing and coking

productsChemical raw materials and

chemical productsMedical and pharmaceutical

productsChemical fibersRubber productsPlastic productsNonmetallic mineralsSmelting and pressing of ferrous

metalsSmelting and pressing of

nonferrous metalsMetal productsOrdinary machinery

manufacturingSpecial purpose equipmentTransportation equipment

manufacturingElectric equipment and

machineryElectronics and

telecommunicationsInstruments, meters, educational

and office machineryOther manufacturingAll manufacturing

a Analysis based on data from Chinese provincial yearbooks. * p < .05, ** p < .01.

ConstantValue

–0.3239–1.8230**–1.60093.6634

–2.7573–5.3746**

–4.4040

–4.6338

–4.5040*1.6282

–3.4975**1.0806

0.4736

1.7837

3.1377*

2.1169–2.4372–7.0235**–0.5699–2.9597*

4.9269

–1.75051.5330

–2.4027–2.8779

–3.9346

2.5839

7.8152**

–2.37890.0268

Capital-LaborRatio

0.13270.27670.83871.1595**1.1235**0.9050*

0.9757

1.3171

1.1043**0.6970**0.9663**0.0355

1.1248**

0.4556

0.7768*

0.4993**1.8424**1.8577**0.9257**0.6599*

0.3853

1.0496*0.7400

0.83871.3677**

1.6158**

0.6029

–0.4436

0.6215**1.0047**

AverageSize of

Establish-ment

–0.14250.18380.0372

–0.4419*0.06490.4770

0.5464

0.6262

0.5099*–0.3589*0.3086

–0.3524

–0.2590*

–0.3370*

–0.5881**

–0.36720.31380.7580**

–0.16310.0844

–0.4360**

–0.0840–0.1452

0.0372–0.0652

0.4266

–0.2988

–0.2513

0.16570.0290

LocationQuotient

0.1283–0.00030.31240.08170.5400**0.2540

0.6901*

–0.0211

–0.13370.2133

–0.06590.5355*

–0.1424

0.0618

0.4320*

–0.24200.17200.26150.21590.2504

–0.0598

–0.09530.5390

0.31240.1103

0.7617**

0.4982*

1.1143*

–0.00232.4893*

Populationof Largest

City

0.3608*0.2439**0.2173

–0.23630.12960.4296**

0.1522

–0.1275

0.1440–0.05990.1752**0.2576

–0.1176

0.0039

0.0036

0.0074–0.44250.04620.02680.2979

–0.3699

0.1861–0.2971

0.21730.1969

–0.1162

–0.0942

–0.7878

0.2678–0.2223

AdjustedR2

0.20750.65130.76220.72570.78420.5221

0.2071

–0.0565

0.27750.61120.67790.1973

0.8791

0.3185

0.5644

0.60520.66500.82470.45480.4663

0.4779

0.20760.0965

–0.00160.7255

0.8191

0.5411

0.3578

0.64840.4932

Numberof

Observa-tions

202020182020

20

18

20192019

19

20

20

1819201920

19

1920

1920

19

18

19

1920


independent variables are the capital-laborratio (K/L, as before), the average size ofestablishments, the location quotient, andthe population of the largest city in eachprovince. The location quotient for anyindustry, i, in any province, j, is defined asπij/Πi, with πij = xij/Σixij and Πi = Xi/ΣXi, wherexij is the number of establishments inindustry i in province j and Xi is the numberof establishments in industry i in the wholeof China (i.e., Xi = Σjxij). The locationquotient (which can take on any valuegreater than zero) thus identifies the degreeto which any given province is specialized inany given industry. In view of the specificconcept of agglomeration proposed earlier,the location quotients used in this phase ofthe investigation are defined for establish-ments, rather than for employment. In prin-ciple, location quotients are supposed tocapture the effects (if any) of so-called local-ization economies. The variable representedby the population of the largest city in eachprovince is a putative measure of the effectsof urbanization economies. These two vari-ables are conceptually unsatisfactory, but inthe absence of more adequate data, they arethe best we have to work with. The largest-city variable used here is especially inade-quate, given that it does not account forthe full extent of urbanization in eachprovince. We experimented with a corre-sponding variable defined as the percentageof the total provincial population living incities, but this variable performed badly asan explanation of agglomeration, as we mightanticipate, given that it does not express amass but a ratio. Complete sets of data foreach industry are not always available, andhence the number of observations(provinces) underlying each regressiondiffers from sector to sector, as shown inTable 4. Unfortunately, the number ofdegrees of freedom is restricted in allcases. All variables are measured in terms ofnatural logarithms.

The first thing to note about the resultspresented in the table is the wide variationin the values of the coefficients attached tothe capital-labor ratio; in some cases, thesevalues are so low as to be close to unbeliev-

able (even when the inefficiencies of state-owned and collective-owned industries aretaken into account). These low values maystem from measurement errors in the dataused. In about half the cases, however, thecoefficients have values that range from0.7 to 1.3, and these values seem to be morecredible, at least in comparison witheconometric studies carried out in otherparts of the world. At the same time, mostof the statistically significant coefficientsattached to the average-size-of-establish-ment variable (which is intended to captureinternal economies and diseconomies ofscale that are due to plant size) are negative,which may not be unreasonable in morelabor-intensive sectors, such as furniture,but seems anomalous in more capital-inten-sive sectors, such as tobacco processing,chemical raw materials, or nonferrousmetals. Again, in the absence of furtherinvestigation, we cannot know whether thisapparent anomaly is due to data errors, togenuine inefficiencies (e.g., in the state-owned sector), or to some other cause.

We now turn to the specifically geographiccontent of the regressions laid out in Table4. This content is represented by the effectsof the location-quotient variable and thelargest-city variable on value added perworker. As expected, the location-quotientvariable has significant positive impacts onproductivity in a number of sectors wherecapital-labor ratios are relatively low andvalues of the H-index high, namely, textiles,leather, stationery, electric equipment andmachinery, electronics and telecommuni-cations, and instruments. In turn, the largest-city variable has a significant and positiveimpact on productivity in garments andprinting, which is in line with what we knowabout the economic geography of thesetwo sectors in other countries. The sameindependent variable is significantly andpositively correlated with productivity infood processing and food manufacturing. Inall four cases, the positive impact of thelargest-city variable may be explained notonly by the basic organizational features ofthese four industries, but also by the orien-tation of a significant number of firms within



them to market locations. If we now corre-late K/L for each two-digit sector (at the levelof the Chinese economy as a whole) with theentire set of regression coefficients attachedto the location-quotient and largest-city vari-ables, we find that in the former case, thesimple correlation is –0.18, which has theexpected sign but is not significant, and inthe latter case, it is –0.50, which is signifi-cant at the .01 level. This outcome suggestsstrongly that labor-intensive industries, inaddition to their inclination to intrasectoralagglomeration, have a particular predilec-tion for locations in large cities.

Thus, in a rough way, the analysis indi-cates that localization and urbanizationeconomies are indeed at work in the Chinesespace-economy and that they have statisticalprofiles that conform broadly to theoreticalexpectations. That said, we are sharply awarethat the analyses presented here are char-acterized by many statistical shortcomingsand interpretative ambiguities and thatfurther probing of the issues is obviously inorder. In the next section, therefore, weattempt to explore further many of the ideaswe developed earlier with the aid of muchmore detailed information on industrialproduction. This exploration involves, first,a shift from a provincial geographic base toa more finely grained county base andsecond, a search for statistical correlationsbetween spatially and functionally relatedsectors.

Industrial Clusters in ChineseCounties

The data set we explored in this phase ofthe investigation consisted of informationfor 510,381 firms that are defined as inde-pendent accounting units registered at orabove the township level of authority,1 as

recorded in China’s 1995 Industrial Census.These firms accounted for 68.2 percent and59.7 percent of the output and employment,respectively, of all China’s industries. Notethat the locations of a number of firms couldnot be ascertained, and when we excludedthese and nonmanufacturing firms (inmining and public utilities), we were leftwith a total of 459,108 firms.

Data on all firms with known locationswere aggregated by county. We defined loca-tion quotients as before, except that theindex i now runs over 534 four-digit indus-trial sectors and j runs over 2,325 coun-ties,2 as opposed to provinces. Again, wefocus on establishments rather than employ-ment. With a full set of location quotients inhand, we computed correlation coeffi-cients representing the spatial associationbetween pairs of four-digit sectors. Then,we used simple cluster analysis in an attemptto identify spatial complexes of industrialactivity.

The correlation matrix we worked withis large, defined as over 543 different four-digit sectors, and a great many differentindustrial clusters can be extracted from it.In what follows, we focus on one large andmultifaceted cluster, which dominates theentire analysis and provides important repre-sentation of both sectoral and spatialtrends in the new Chinese economy (seeFigure 1). In fact, this dominant clusterbreaks down into two overlapping


1 The definitional criteria are specific to theChinese industrial statistics system. Firms thatare independent accounting units are notfinanced by or ancillary to any other institu-tions. Firms are also identified in terms of theadministrative rank of the places in which theyare registered. The data set used here excludes

firms identified as being below the township level,such as those established by villages. In addition,firms registered at or above the township levelinclude the bulk, if not all, of state-owned andshare-holding enterprises and joint venturesbetween Chinese and overseas investors.Collective-owned enterprises and private- andindividual-owned enterprises are less highlyrepresented in the data set. As such, the dataset is probably biased toward larger firms. In total,the 510,381 firms in the data set accounted for86.2 percent of all firms at or above the townshiplevel.

2 County-level units include counties, county-level cities, and urban districts of prefecture-leveland provincial-level cities. In this article, we referto all county-level units as counties.


subgroups. The first subgroup is comprisedof consumer electronics and garments(including cotton knitting) industries, withthe former breaking further down intosectors like radios, television sets, cameras,toys, and clocks; these are among the mostprominent export-oriented industries inChina. The second subgroup consists ofcomputers, electronic equipment and instru-ments industries, including semiconductorsand medical equipment. Unfortunately, ourdata do not allow us to identify functionallinkages between the sectors under exami-

nation, but there is strong prima facieevidence, based on our general knowledgeof how these industries work, that at leastsome of them are both transactionally inter-connected and strongly associated withone another in geographic terms. In Figures2 and 3 we map out values of the number ofestablishments and location quotients forestablishments in the two subgroups. Theagglomerative tendencies of these industriesare clearly evident.

The consumer electronics and garmentsgroup is concentrated in the Pearl River


Figure 1. Cluster analysis of establishments in the consumer electronics and garments industriesand the computers, electronic equipment and instruments industries, China, 1995. Figure based onfour-digit data by county from the 1995 Industrial Census provided by the State Statistical Bureau,

People’s Republic of China.

#9905—ECONOMIC GEOGRAPHY—VOL. 79 NO. 3—79304-Fan_Scott—Pg. 311

Figure 2. Distribution of establishments in the consumer electronics and garments industries bycounty, China, 1995. Figure based on four-digit data from the 1995 Industrial Census provided bythe State Statistical Bureau, People’s Republic of China. Note: Counties in which the number of

establishments is less than 0.04 percent of all consumer electronics and garments establishments inChina are excluded.

0 600 km

0 600 mi

Tianjin

Beijing

JIANGSU

ZHEJIANG

GUANGDONG

Shenzhen

Pearl River Delta

Shanghai

FUJIAN

Location QuotientNumber of

Establishments

1001-1500

1—250251—500

501—1000

≥ 4.50

3.00—4.49

1.50—2.99


Figure 3. Distribution of establishments in the computers, electronic equipment and instrumentsindustries by county, China, 1995. Figure based on four-digit data from the 1995 Industrial Censusprovided by the State Statistical Bureau, People’s Republic of China. Note: Counties in which the

number of establishments is less than 0.04 percent of all computers, electronic equipment andinstruments establishments in China are excluded.

Tianjin

Beijing

Shenzhen

Shanghai

Wuhan

Jinan

Xian

Chongqing

Hongyuan

Shenyang

≥ 4.50

3.00—4.49

1.50—2.99


Establishments

201-500

1—5051—100

101—200

0 600 km

0 600 mi


Delta, eastern coastal Guangdong andcoastal Fujian, Shanghai and nearby areasin southern Jiangsu and northern Zhejiang,and in Beijing and Tianjin and adjacentareas. Industries in this subgroup are char-acterized by relatively low capital-laborratios, which as we have already shown is asign of the tendency toward locationalagglomeration. In addition, these industriesare among the largest export industries inChina, and the Chinese government hasactively reinforced their concentration in thePearl River Delta. These industries are muchfavored by overseas Chinese investors, whosesocial ties and networks have helped to boosteconomic development in the area(Chisterson and Lever-Tracy 1997). ThePearl River Delta has benefited from bothgovernmental policy and strong social tiesto overseas Chinese investors. The largenumber of adjacent counties with high loca-tion quotients and/or a large number ofestablishments in that area suggests not onlythat the industries are spatially concentrated,but also that institutional forces have effec-tively fostered functional linkages andagglomeration economies. Similar processesappear to have begun in and near Beijing,Tianjin, and Shanghai.

The computer, electronic equipment andinstruments subgroup also displays a degreeof clustering in the Pearl River Delta,especially in and near Shenzhen, and in areasnear Shanghai, Beijing, and Tianjin, butthere are a number of other clusters scat-tered throughout the country. In a fewisolated cases, the high location quotientsobserved for this subgroup are a statisticalartifact resulting from the small total numberof industrial establishments in certain coun-ties. Hongyuan county in western China, forexample, has a high location quotient of 10.1but is the location of only six industrial estab-lishments in total. More important, the manydifferent agglomerations of establishmentsin this subgroup reflect their affinity for largecities, which, in addition to the cities of thecoastal areas, include Chongqing, Xian,Wuhan, Jinan, and Shenyang. This patterncan be related, in part, to the combinedeffects of market forces and the past loca-

tional policies of the Chinese governmentin promoting the growth of inland cities. Inaddition, the work of Wang and Wang (1998)on the Zhong’guancun high-technologydistrict adjacent to Beijing University under-lines the significance of institutions of highereducation in helping to anchor agglomera-tions of industries in the second subgroup.Beijing’s large number of establishments andhigh location quotient again point to theimportant role of institutions of highereducation. This factor appears to be relatedalso to the large number of establishmentsin Tianjin and Shanghai.

As we indicated in Table 1, the garmentsand electronics and telecommunicationsindustries (at the two-digit level and withprovinces as units of observation) are highlyconcentrated in spatial terms. On the basisof our county-level data, we recomputed theH-indexes for all industries and again foundthat garments and electronics and telecom-munications are close to the top of thesectoral rankings. The analysis presentedin Table 4 also suggested that the clusteringof these two industries has statistically signif-icant effects on their productivity.Specifically, we found that urbanization hasa positive impact on productivity in thegarments sector, and localization has a posi-tive impact in electronics and telecommu-nications. The Chinese space-economy is ahighly complex one, however, and forsome industries spatial concentration doesnot necessarily or always lead to unusuallyhigh levels of productivity. This seems to beespecially the case in transportation-equip-ment manufacturing, which is marked bydense spatial concentration, especially inareas around Shiyan in central China,Changchun in the northeast, and Chongqingin the southwest, as shown in Figure 4.There are undoubtedly some productivityeffects in these areas that are due to agglom-eration, but these industries were initiallylocated in these relatively inaccessible placesduring the Maoist period for politicalreasons. During the 1950s, the First AutoWorks was located in Changchun so as to beclose to the former Soviet Union (Sit andLiu 2000). After the Sino-Soviet split and



Figure 4. Distribution of establishments in the transportation-equipment manufacturing industry bycounty, China, 1995. Figure based on four-digit data from the 1995 Industrial Census provided bythe State Statistical Bureau, People’s Republic of China. Note: Counties in which the number ofestablishments is less than 0.04 percent of all transportation-equipment manufacturing establish-

ments in China are excluded.

0 600 km

0 600 mi

Changchun

Chongqing

Tianjin

Beijing

Shiyan

Wuhan


Establishments

101-200

1—2021—50

51—100

≥ 4.50

3.00—4.49

1.50—2.99


because of efforts to build key industries ininland “third front” locations (Naughton1988), Shiyan—a remote and inaccessiblecounty—was selected as the site of theSecond Auto Works and the nation’s firstindependent auto manufacturing complex.Our statistical analyses (see Table 4) showthat productivity in the transportation-equip-ment sector is statistically related to thecapital-labor ratio but not to localization orurbanization. In fact, new car-manufacturingclusters in coastal areas around Shanghai,Beijing, and Tianjin have emerged since thereforms and as market forces have increas-ingly driven locational outcomes in thisindustry (Sit and Liu 2000). Similarly, marketforces underlie the recent trend to diffusetransportation-equipment manufacturingfrom Shiyan toward Wuhan and other moreaccessible locations.

ConclusionIn this article, we have sought to explore

the interconnections between geography andeconomic development in the less-devel-oped countries of East Asia, paying specialattention to the role of dense industrialregions as conduits of productivity effectsand developmental impulses. We havecarried out this exploration on the basis oftheoretical speculation, appeals to theliterature, and a series of empirical investi-gations focused on the Chinese case. Onceagain, we stress that the quality of the dataon which the latter investigations are basedleaves much to be desired. That said, ourmultifaceted approach to the central ques-tion appears to point with some plausibilityto a real connection between agglomeration,on the one side, and economic developmentand growth, on the other. This outcome isof special interest because it concerns situ-ations that are very different indeed fromthe usual case-study material found in theliterature, which is generally focused ondeveloped Western economies.

Our empirical analysis of Chinese indus-tries supports the argument that a positiverelationship can be found between indus-trial agglomeration and productivity in

economies that were formerly dominated bycentral planning, and it highlights, above all,sectors and spaces that are undergoingeconomic liberalization as those most proneto the formation of agglomerationeconomies. China has a mixed economy inwhich socialist legacies exist side by side withnew developmental forces as a result ofmarket-opening measures and both formaland informal institutions continue to playimportant roles in shaping the space-economy. China’s economic geographywas formerly heavily shaped by a socialistideology that downplayed agglomerationeconomies. Our empirical analysis identifiedtransportation-equipment manufacturing asa sector that exhibits a degree of spatial clus-tering but little in the way of agglomera-tion economies and suggests that this situa-tion may be a legacy of former stateplanning. Recent economic liberalization inChina, however, seems to have fostered boththe macroeconomic and local conditionsunder which viable industrial agglomerationscan emerge. During the past two decades,the Chinese government has promotedlabor-intensive industries, mostly for export,and has developed economic zones in whichthese industries can flourish. Our empir-ical investigations point to a strong rela-tionship between industrial clustering andproductivity, especially in industries likeconsumer electronics and garments andcomputers that have gained prominencesince the government’s drive towardeconomic liberalization, and especially in thecoastal provinces and large cities. In addi-tion to showing that the relationship betweenagglomeration and productivity is mediatedin peculiar ways by governmental policies,both old and new, we have argued that theinformal social networks that have tradi-tionally played such a strong role in theChinese economy are also helping to rein-force regional patterns of economic devel-opment.

Of course, much additional work is neces-sary, not only for its own sake, but also as aprelude to effective policymaking in themany and vastly different regional economicsystems (at various levels of development)



that are to be found in East Asia in general,and China in particular. The analysispresented here, despite its shortcomings,provides a reasonably convincing precedentfor probing more deeply into the questionsraised. As we have indicated, there is a strongconceptual and empirical case for claimingthat locational clustering and economicperformance are intertwined in complex butimportant ways. This research agenda is notonly of considerable importance and interestin its own right, but is all the more pressingbecause the insights to be gained offer thepossibility of greatly enhanced policymakingand hence of significant economic gains toparticular regions and countries. The stakesare especially high in those parts of East Asiathat are still struggling to attain to world stan-dards of industrial production and compet-itive advantage and to establish a durableposition for themselves on global markets.

ReferencesAmin, A. 1999. An institutionalist perspective on

regional economic development. InternationalJournal of Urban and Regional Research23:365–78.

Amin, A., and Thrift, N. 1995. Institutional issuesfor the European regions: From markets andplans to socioeconomics and powers of asso-ciation. Economy and Society 24:41–66.

Beardsell, M., and Henderson, V. 1999. Spatialevolution of the computer industry in the USA.European Economic Review 43:431–56.

Bell, M., and Albu, M. 1999. Knowledge systemsand technological dynamism in industrial clus-ters in developing countries. WorldDevelopment 27:1715–34.

Bonacich, E., and Appelbaum, R. 2000. Behindthe label: Inequality in the Los Angeles apparelindustry. Berkeley: University of CaliforniaPress.

Carlino, G. A. 1979. Increasing returns to scalein metropolitan manufacturing. Journal ofRegional Science 19:363–73.

Chang, P.-L., and Hsu, C.-W. 1998. The devel-opment strategies for Taiwan’s semiconductorindustry. IEEE Transactions on EngineeringManagement 45:349–56.

Chen, W.-H. 1999. Manufacturing strategies ofnetwork-based small firms: Observations onthe textile industry in Taiwan. Journal of SmallBusiness Management 37:46–62.

Chen, Y. 1996. Impact of regional factors onproductivity in China. Journal of RegionalScience 36:417–36.

Christerson, B., and Lever-Tracy, C. 1997. TheThird China? Emerging industrial districts inrural China. International Journal of Urbanand Regional Research 21:569–88.

Cooke, P., and Morgan, K. 1998. The associa-tional economy: Firms, regions, and innova-tion. Oxford, U.K.: Oxford University Press.

Eng, I. 1997. Flexible production in late indus-trialization: The case of Hong Kong. EconomicGeography 73:26–43.

Fan, C. C. 1995. Of belts and ladders: State policyand uneven regional development in post-MaoChina. Annals of the Association of AmericanGeographers 85:421–49.

———. 1997. Uneven development andbeyond: Regional development theory in post-Mao China. International Journal of Urbanand Regional Research 21:620–39.

———. 2001. Migration and labor marketreturns in urban China: Results from arecent survey in Guangzhou. Environment andPlanning A 33:479–508.

———. 2002. The elite, the natives, and theoutsiders: Migration and labor market segmen-tation in urban China. Annals of theAssociation of American Geographers92:103–24.

Fujita, M., and Tabuchi, T. 1997. Regional growthin postwar Japan. Regional Science and UrbanEconomics 27:643–70.

Granovetter, M. 1985. Economic action and socialstructure: The problem of embeddedness.American Journal of Sociology 91:481–510.

Guo, H., and Cai, C. 2000. Network structure:The organizational model of the sock-makingindustry in Datang [in Chinese]. ZhejiangEconomy (Zhejiang Jingji) 9:16–17.

Hamilton, G. 1991. Business networks andeconomic development in East and SoutheastAsia. Hong Kong: Center of Asia Studies.

Harrison, B. 1992. Industrial districts: Old winein new bottles. Regional Studies 26:469–83.

Henderson, J. V. 1986. Efficiency of resourceusage and city size. Journal of UrbanEconomics 19:47–70.

———. 1988. Urban development: Theory, fact,and illusion. New York: Oxford UniversityPress.

Hirschman, A. O. 1958. The strategy of economicdevelopment . New Haven, Conn.: YaleUniversity Press.

Ho, K. C. 1994. Industrial restructuring, theSingapore city-state, and the regional divi-



sion of labour. Environment and Planning A26:33–51.

Hodgson, G. M. 1988. Economics and institu-tions. Cambridge, U.K.: Polity Press.

Hsing, Y. 1995. Taiwanese guerilla investors insouthern China. Review of Urban and RegionalStudies 7:24–34.

———. 1996. Blood thicker than water:Interpersonal relation and Taiwanese invest-ment in southern China. Environment andPlanning A 28:2241–61.

———. 1998. Making capitalism in China: TheTaiwan connection . New York: OxfordUniversity Press.

Huang, Y. 1999. Shiyan: A factory in the wilder-ness, 1968 false alarm. Time International154:78.

Jacobs, J. 1969. The economy of cities. New York:Random House.

Johansson, H., and Nilsson, L. 1997. Exportprocessing zones as catalysts. WorldDevelopment 25:2115–28.

Kaldor, N. 1970. The case for regional policies.Scottish Journal of Political Economy17:337–48.

Kanemoto, Y.; Ohkawara, T.; and Suzuki, T. 1996.Agglomeration economies and a test foroptimal city sizes in Japan. Journal of theJapanese and International Economies10:379–98.

Kawashima, T. 1975. Urban agglomerationeconomies in manufacturing industries. Papersof the Regional Science Association 34:157–75.

Krugman, P. 1991. Geography and trade, GastonEyskens lecture series. Leuven, Belgium:Leuven University Press.

Lampard, E. E. 1955. The history of cities ineconomically advanced areas. EconomicDevelopment and Cultural Change 3:81–102.

Lee, Y. J., and Zang, H. 1998. Urbanization andregional productivity in Korean manufacturing.Urban Studies 35:2085–99.

Leung, C. K. 1993. Personal contacts, subcon-tracting linkages, and development in theHong Kong-Zhujiang Delta region. Annals ofthe Association of American Geographers83:272–302.

———. 1996. Foreign manufacturing invest-ment and regional industrial growth inGuangdong Province, China. Environment andPlanning A 28:513–36.

Levy, B. 1991. Transactions costs, the size of firmsand industrial policy. Journal of DevelopmentEconomics 34:151–78.

Lin, G. C. S. 2001. Metropolitan development ina transitional socialist economy: Spatial restruc-

turing in the Pearl River Delta, China.Urban Studies 38:383–406.

Lipton, M. 1977. Why poor people stay poor: Astudy of urban bias in world development.London: Temple Smith.

Liu, S.-J. 1998. Industrial development and struc-tural adaptation in Taiwan: Some issues oflearned entrepreneurship. IEEE Transactionson Engineering Management 45:338–48.

Markusen, A. R.; Lee, Y.-S.; and DiGiovanna, S.,eds. 1999. Second tier cities: Rapid growthbeyond the metropolis. Minneapolis: Universityof Minnesota Press.

Markusen, A., and Park, S. O. 1993. The stateas industrial locator and district builder: Thecase of Changwon, South Korea. EconomicGeography 69:157–81.

Marshall, A. 1890. Principles of economics.London: Macmillan.

———. 1919. Industry and trade: A study ofindustrial technique and business organiza-tion. London: Macmillan.

McGee, T. G. 1991. The emergence of desakotaregions in Asia: Expanding a hypothesis. InThe extended metropolis: Settlement transitionin Asia, ed. N. Ginsburg, B. Koppel, and T. G.McGee, 3–25. Honolulu: University of HawaiiPress.

Myrdal, G. 1959. Economic theory and under-developed regions . London: GeraldDuckworth.

Nadvi, K., and Schmitz, H. 1994. Industrial clus-ters in less developed countries: Review of expe-riences and research agenda. Discussion PaperNo. 339. Brighton, U.K.: Institute ofDevelopment Studies, University of Sussex.

Nakamura, R. 1985. Agglomeration economiesin urban manufacturing industries: A case ofJapanese cities. Journal of Urban Economics17:108–24.

Naughton, B. 1988. The third front: Defenseindustrialization in the Chinese interior. ChinaQuarterly 115:351–86.

Olds, K. 2001. Globalization and urban change:Capital, culture, and Pacific Rim mega-projects. Oxford, U.K.: Oxford UniversityPress.

Park, S. O. 1991. High-technology industries inKorea: Spatial linkages and policy implications.Geoforum 22:421–31.

———. 1993. Industrial restructuring andspatial division of labor: The case of theSeoul metropolitan region, the Republic ofKorea. Environment and Planning A 25:81–93.

———. 1994. Industrial restructuring in theSeoul metropolitan region: Major triggers and



consequences. Environment and Planning A26:527–41.

———. 1996. Networks and embeddedness inthe dynamic types of new industrial districts.Progress in Human Geography 20:476–93.

Patchell, J. 1993. Composing robot productionsystems: Japan as a flexible manufacturingsystem. Environment and Planning A25:923–44.

Peck, J. 1992. Labor and agglomeration: Controland flexibility in local labor markets. EconomicGeography 68:325–47.

Perroux, F. 1961. L’Economie du XXe siecle. Paris:Presses Universitaires de France.

Poot, H.; Kuyvenhoven, A.; and Jansen, J. 1990.Industrialization and trade in Indonesia.Yogyakarta, Indonesia: Gadjah MadaUniversity Press.

Porter, M. E. 2001. Regions and the neweconomics of competition. In Global cityregions: Trends, theory, policy, ed. A. J. Scott,139–57. Oxford, U.K.: Oxford University Press.

Rasiah, R. 1994. Flexible production systems andlocal machine-tool subcontracting: Electronicscomponents transnationals in Malaysia.Cambridge Journal of Economics 18:279–98.

Sandee, H. 1995. Innovation adoption in ruralindustry: Technological change in roof tile clus-ters in Central Java, Indonesia. Unpublisheddoctoral dissertation, Vrije Universiteit,Amsterdam.

Sandee, H., and Rietveld, P. 2001. Upgradingtraditional technologies in small-scale industryclusters: Collaboration and innovation adop-tion in Indonesia. Journal of DevelopmentStudies 37:150–172.

Sandee, H.; Rietveld, P.; Supratikno, H.; andYowono, P. 1994. Promoting small scale andcottage industries in Indonesia: An impactanalysis for Central Java . ResearchMemorandum 1994-10, Amsterdam:Department of Economics and Econometrics,Vrije Universiteit.

Scott, A. J. 1987. The semiconductor industryin South-East Asia: Organization, location andthe international division of labour. RegionalStudies 21:143–60.

———. 1988. Metropolis: From the division oflabor to urban form. Berkeley: University ofCalifornia Press.

———. 1994. Variations on the theme ofagglomeration and growth: The gem andjewelry industry in Los Angeles and Bangkok.Geoforum 25:249–63.

———. 2002. Regional push: The geography ofdevelopment and growth in low- and middle-

income countries. Third World Quarterly23:137–61.

Shefer, D. 1973. Localization economies inSMSAs: A production function analysis.Journal of Regional Science 13:55–64.

Shieh, G. S. 1990. Manufacturing “bosses”:Subcontracting network under dependent capi-talism in Taiwan. Unpublished Ph.D. disser-tation, University of California, Berkeley.

Shukla, V. 1996. Urbanization and economicgrowth. Delhi: Oxford University Press.

Sit, V. F. S., and Liu, W. 2000. Restructuring andspatial change of China’s auto industry underinstitutional reform and globalization. Annalsof the Association of American Geographers90:653–73.

Smart, A. 2000. The emergence of local capi-talisms in China: Overseas Chinese investmentand patterns of development. In China’sregions, polity, and economy: A study of spatialtransformation in the post-reform era, ed. S.Li and W. Tang, 65–95. Hong Kong: ChineseUniversity Press.

Smyth, I. 1992. Collective efficiency and selec-tive benefits: The growth of the rattan industryof Tegalwangi (Indonesia). IDS Bulletin23:51–56.

Storper, M. 1997. The regional world: Territorialdevelopment in a global economy. New York:Guilford Press.

Storper, M., and Scott, A. J. 1995. The wealthof regions. Futures 27:505–26.

Sunley, P. 1996. Context in economic geography:The relevance of pragmatism. Progress inHuman Geography 20:338–55.

Sveikauskas, L. 1975. The productivity of cities.Quarterly Journal of Economics 89:393–413.

Tang, H. K. 1996. Hollowing-out or internationaldivision of labour? Perspective from theconsumer electronics industry and Singapore.International Journal of TechnologyManagement 12:231–41.

Tong, X., and Wang, J. 2002. Global-localnetworking of PC manufacturing in Dongguan,China. In Knowledge, industry and environ-ment: Institutions and innovation in territo-rial perspective, ed. R. Hayter and R. Le Heron,67–86. London: Ashgate.

Tsay, C.-L. 1993. Industrial restructuring andinternational competition in Taiwan.Environment and Planning A 25:111–20.

Wang, J. 2001. Innovative spaces: Enterprise clus-ters and regional development [in Chinese].Beijing: Peking University Press.

Wang, J., and Wang, J. 1998. An analysis ofnew-tech agglomeration in Beijing: A new



industrial district in the making? Environmentand Planning A 30:681–701.

Weijland, H. 1999. Microenterprise clusters inrural Indonesia: Industrial seedbed and policytarget. World Development 27:1515–30.

Wong, T.-C., and Goldblum, C. 2000. The China-Singapore Suzhou Industrial Park: A turnkeyproduct of Singapore? Geographical Review90:112–22.

Xue, L. 1997. Promoting industrial R&D andhigh-tech development through science parks:The Taiwan experience and its implicationsfor developing countries. International Journalof Technology Management 13:744–61.

Yeung, H. W., and Olds, K. 2000. GlobalizingChinese business firms: Where are theycoming from, where are they heading? InGlobalization of Chinese business firms, ed.H. W.-c. Yeung and K. Olds, 1–28. London:Macmillan.

Yuan, J., and Eden, L. 1992. Export processingzones in Asia: A comparative study. AsianSurvey 32:1026–45.

Yuen, B. 1991. Planning and development ofindustrial estates in Singapore. Third WorldPlanning Review 13:47–68.



FPO—Per client’s

instructions, thispage is to be

printed with arunning head only

320 ECONOMIC GEOGRAPHY

Documents

Industrial Agglomeration and Development: A Survey of ...hypothesized, but that genuine industrial clusters are more characteristic of sectors and spaces that have been most deeply