INTERCOMMUNITY RELATIONSHIPS AND COMMUNITY …haiyang/SMJ zone 2.pdf1986, which formalized China’s...

Strategic Management JournalStrat. Mgmt. J., 30: 163–183 (2009)

Published online 4 November 2008 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/smj.727

Received 1 August 2007; Final revision received 23 July 2008

INTERCOMMUNITY RELATIONSHIPS ANDCOMMUNITY GROWTH IN CHINA’S HIGHTECHNOLOGY INDUSTRIES 1988–2000

YAN ZHANG,1 HAIYANG LI,1* and CLAUDIA BIRD SCHOONHOVEN2

1 Jesse H. Jones Graduate School of Management, Rice University, Houston, Texas,U.S.A.2 The Paul Merage School of Business, University of California, Irvine, California,U.S.A.

In this study, we examine how intercommunity relationships affect the growth of organizationalcommunities. Using a unique panel dataset on 53 technology development communities in Chinaspanning 1988–2000, we found that regional community density, a community’s geographicproximity to the nearest community and its domain overlap with the nearest community havean inverted U-shaped relationship with the community’s growth. These non-monotonic resultssuggest that adjacent communities have both mutualistic and competitive effects on each other.Theoretical and managerial implications are discussed. Copyright 2008 John Wiley & Sons,Ltd.

INTRODUCTION

Scholars from several disciplines have paidincreasing attention to the emergence and growthof organizational communities.1 According toPorter (1998a: 78), organizational communities are

Keywords: intercommunity relationships; high technol-ogy industries; China; technology clusters∗ Correspondence to: Haiyang Li, Jesse H. Jones GraduateSchool of Management, Rice University, 6100 Main Street,Houston, TX 77005, U.S.A.E-mail: [email protected] Scholars have used a plethora of terms to describe the orga-nizational community phenomenon such as organizational com-munities (Aldrich and Ruef, 2006; Astley, 1985; Freeman andAudia, 2006; Wade, 1995, 1996), regional industrial districtsor clusters (Krugman, 1991; Piore and Sabel, 1984; Porter,1998a, 1998b; Romanelli and Khessina, 2005; Tallman et al.,2004), incubator regions (Schoonhoven and Eisenhardt, 1993),industrial systems (Saxenian, 1994), and science parks and incu-bators (Phan, Siegel, and Wright, 2005). The common themein this research stream is that each describes a geographicallybounded locale within which multiple populations or industriesexist in a community of relationships. For example, Freeman andAudia (2006: 145) conceptualize community as a set of relationsbetween organizational forms or places where organizations arelocated in resource space or in geography. Similarly, Aldrichand Ruef (2006: 243) define an organizational community as a

‘geographic concentrations of interconnected com-panies and institutions in a particular field,’ andthey encompass an array of linked industries andother entities important to competition. It has beenargued that the creation of organizational com-munities is a vehicle for developing technologicalcompetitiveness and catalyzing economic growthat the nation, state, and city levels (Porter, 1998a;Romanelli and Khessina, 2005). Not surprisingly,several governments have placed the developmentof organizational communities at the center of theirnational development programs (Enright, 1999;Mathews, 1997; Perez-Aleman, 2005).

Considering the economic and technological sig-nificance of organizational communities, what fac-tors can affect their growth? Some theorists havefocused on the external resource conditions andargued that a panoply of superior natural, indus-trial, and institutional resources (Aldrich and Ruef,2006; Chiles, Meyer, and Hench, 2004; Krug-man, 1991) and social networking with access

set of coevolving organizational populations joined by ties ofcommensalism and symbiosis.

Copyright 2008 John Wiley & Sons, Ltd.

164 Y. Zhang, H. Li, and C. B. Schoonhoven

to cutting-edge information (Sorenson, 2003; Stu-art and Sorenson, 2003) give rise to particularregional capacities for community growth. Alter-natively, other scholars have emphasized the roleof within-community relationships in supportingcommunity growth. For example, Saxenian (1994)observed that despite beginning from ‘relativelysimilar’ starting points after World War II, theSilicon Valley region in northern California out-performed Boston’s Route 128 because the SiliconValley is a regional network-based industrial sys-tem that creates greater regional flexibility andtechnological dynamism, which in turn promotescollective learning among firms (Saxenian, 1994:2–9). Perez-Aleman (2005) argued in the contextof two successful communities in Chile that com-munity growth depends on building institutionsthat enable coordinated learning among firms toimprove capabilities, processes, and products.

While these studies have contributed substan-tially to our nascent understanding of communitygrowth, there are gaps in the extant literature. Asnoted above, previous studies have focused eitheron external resource conditions of a specific com-munity (e.g., Krugman, 1991) or endogenous fac-tors within a specific community (Perez-Aleman,2005; Saxenian, 1994). Beyond these two views,in cases of multiple communities, it is likely thatintercommunity relationships will have importantconsequences for community growth. For example,Porter (1998a: 89) noted that an industrial clus-ter could affect the productivity of other clusters.Tallman and Phene (2007) found that knowledgeflows (in terms of patent citations) within regionalclusters are not significantly different from thosebetween regional clusters in a domestic context.This finding implies that the boundaries of com-munities are open and porous and do not pre-vent knowledge from flowing from one commu-nity to another. Furthermore, Saxenian and Hsu(2001) noted that the external connections of theHsinchu Science District in Taiwan with SiliconValley in the United States (through the flowsof people, information, and know-how) providedHsinchu with an additional impetus for its sus-tained growth. More generally, Barnett and Carroll(1987: 400) noted that organizational interdepen-dence can exist between communities of organiza-tions although most existing research has focusedon interdependence between individual organiza-tions.

In this study, we adopt an ecological perspective(Aldrich and Ruef, 2006; Astley, 1985; Barnett andCarroll, 1987; Freeman and Audia, 2006; Hannanand Freeman, 1977, 1989) to explore how inter-community relationships affect community growth.From an ecological perspective, two or more com-munities are interdependent if the presence of oneaffects the outcomes of the other. We argue thatorganizational communities have both mutualis-tic and competitive effects on one another, andwe delineate three dimensions of intercommunityrelationships: regional community density (i.e., thenumber of communities in the same region), acommunity’s geographic proximity to the near-est community, and a community’s domain over-lap with the nearest community. We propose thateach of these dimensions will have an invertedU-shaped relationship with a focal community’sgrowth due to the joint effects of mutualism andcompetition. We explore these ideas in the contextof all 53 national technology development zonescreated in China between 1988, when the first wasfounded in Beijing, and 2000. National technologydevelopment zones are conceptualized as technol-ogy communities that contain several technology-related populations of firms.

This study contributes to a greater understandingof community phenomena, especially the impact ofintercommunity relationship on communitygrowth. Prior studies of organizational communi-ties have focused on either the external resourceconditions of a specific community or on endoge-nous factors within a specific community. Hence,the literature has implicitly treated communities asif they are independent of one another. In contrast,this research assumes that organizational com-munities are interdependent. We delineate threedimensions of intercommunity relationships (i.e.,density, geographic distance, and domain over-lap) and examine how these dimensions affect thegrowth of geographically dispersed organizationalcommunities containing multiple interrelated pop-ulations of firms in the context of China’s nationaltechnology development zones. Thus, this studyelaborates the spectrum of possible explanationsfor community growth.

Our study differs from existing ecology stud-ies and extends the ecology perspective in twoimportant ways. First, most existing ecology stud-ies have focused on interdependence between firmswithin a population (i.e., industry) and have exam-ined the growth or decline of the population or

Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 30: 163–183 (2009)DOI: 10.1002/smj

Community Growth in China’s High Technology Industries 165

the growth or decline of firms within a population(Aldrich and Ruef, 2006: 37). In their summaryof the extensive population ecology research pub-lished between 1983 and 2006, Hannan, Polos, andCarroll (2007: 31) observed that gaining the req-uisite institutional knowledge about multiple pop-ulations poses a formidable empirical challenge.Not surprisingly research on interpopulation rela-tions within a community has progressed slowly(Aldrich and Ruef, 2006: 250) as have studiesof community-level processes for similar reasons.This study advances the literature by applying anecological perspective to examine the overall eco-nomic growth of geographical clusters of numer-ous interrelated populations of firms in the contextof China’s national technology development zones.Building on the few exemplary studies of com-munities that exist (e.g., Ruef, 2000; Wade, 1995,1996), we have created a multicommunity, longi-tudinal dataset that enables analyses of how inter-community relationships affect community growthover time. Therefore, this study extends existingknowledge by testing the extent to which an eco-logical perspective can be applied to communitiesthat are geographic clusters of multiple populationsof firms (instead of a single population).

Second, our study also contributes to the ecol-ogy literature by simultaneously examining themultiple dimensions of intercommunity relation-ships described above. In contrast, previous ecol-ogy studies have typically examined one of thesedimensions (primarily density) at the organiza-tional or population level. Furthermore, whilethe inverted U-shaped effect of density has beenwidely examined in the population ecology litera-ture (we also develop a hypothesis on density toclosely link our study to the existing ecology lit-erature), existing studies of geographic proximity(Sorenson and Stuart, 2001; Stuart and Sorenson,2003) and domain overlap between organizations(Baum and Mezias, 1992; Baum and Singh, 1994a,1994b; Dobrev, Kim, and Hannan, 2001; Podolny,Stuart, and Hannan, 1996) have primarily exam-ined their monotonic effects on a variety of out-comes. In comparison, this study proposes thatgeographic proximity and domain overlap betweenadjacent communities will have both mutualisticand competitive effects, and thus we expect toobserve an inverted U-shaped relationship of eachwith community growth.

In the following pages, we describe the con-text of the study by explaining how technology

development zones were first created and havesubsequently grown in China. Then, we presentour conceptual framework and hypotheses, dis-cuss the research design and measures, and reportour empirical findings. We conclude by discussingimplications of these results for theory, futureresearch, and managerial implications.

CONTEXT OF THE STUDY

The creation of technology communities has beenviewed as a powerful vehicle for developing tech-nological competitiveness and catalyzing economicgrowth at the nation, state, and city levels (Porter,1998a, 1998b). In an attempt to duplicate thesuccess of the U.S.’s Silicon Valley in develop-ing high-technology industries, the Chinese centralgovernment launched its 863 Program in March1986, which formalized China’s intent to establishnational technology development zones to encour-age local entrepreneurship in high-technology in-dustries as a means of building China’s future tech-nology capabilities. The first national technologydevelopment zone was established in 1988 in Bei-jing, and by 1998 an additional 52 national tech-nology development zones were created through-out China.

All national technology development zones aregoverned by State Council regulations (i.e., Rel-evant Policies and Regulations on National Tech-nology Development Zones, 1991). The regulationsrequire that zones foster collaboration betweena university-based research center, an innovationcenter that will provide applied technology forproduct development, and commercial firms thatcan provide product manufacturing and market-ing (DFL International, 1999: 23–24). Zones areopen to both domestic and foreign high-technologyinvestors and are composed of a mixture of spe-cific industrial populations in certain technologyindustries that are considered new and ‘high tech-nology’ in China. These include electronic infor-mation, integrated optical and advanced manufac-turing, biotech and pharmaceuticals, new materi-als, new energy, aeronautical engineering, oceantechnology, high technology agriculture, environ-mental protection, and nuclear applications.

Firms in the national technology zones enjoypreferential policies that include tax reductions,facility and land use rights, and import privileges,among others. For example, firms in the zones pay



an income tax of 15 percent, which is less thanhalf the normal tax rate of 33 percent. Taxes fornew entrants are waived for the first three years,with an additional 50 percent reduction in taxesover the subsequent three years. Because the pri-mary purpose of the national technology develop-ment zones is to promote technological innovation,only qualified firms are allowed to enter the zones.To qualify for entry, a firm must be certified as‘high-tech’ by the Administrative Committee of azone by conducting business activities in targetedhigh-technology industries, by having a top man-agement team composed of engineers or scientists,having 20 percent of employees be college gradu-ates, and having at least three percent of sales spenton research and development (Li and Atuahene-Gima, 2001). The high-tech status of entrants isfurther monitored and renewed by the Administra-tive Committee of a zone on an annual basis.

As a group, China’s technology developmentzones have grown dramatically in their first decadeof existence. Figure 1 reveals that revenues forall zones reached 460 billion (in 1990’s renminbi[RMB] value) between 1988 and 2000. How-ever, zones have grown differentially. For exam-ple, although both were founded in 1991, theShanghai Zone reached RMB 75.1 billion in rev-enues in 2000, whereas the Taiyuan Zone in Shanxiprovince reached only RMB 7.7 billion in the sameyear. We may ask: What factors account for thesezones’ differential growth rates?

Because our analysis includes geographic prox-imity as a dimension of community interdepen-dence, the reader may find it useful to visualizethe geographic distribution of national technologydevelopment zones in China. Figure 2 shows that

all are located in cities, typically formed where ear-lier organizing has concentrated one or more orga-nizations or institutions specializing in science-based technology research. For example, the Bei-jing Technology Development Zone is located inthe Haidian District of Beijing, which is hometo the Chinese Academy of Science, Peking Uni-versity, Tsinghua University, and other researchinstitutes and government think tanks.

Figure 2 also reveals that China’s national tech-nology development zones are not evenly dis-tributed throughout the country. All provinces,municipality cities, and autonomous regions con-tain at least one zone, with the exception ofthree of China’s innermost province/autonomousregions (Qinghai Province, the Tibet, and NingxiaAutonomous Regions), which are mountainous andsparsely populated. Also, some provinces havemore zones than others, ranging from one to sixper province. Furthermore, some zones are in closeproximity to others, whereas others are relativelydistant from the nearest zone. The variable geo-graphic distribution of national technology zonesthroughout the country provides an opportunity toexamine how intercommunity relationships affectcommunity growth. In the next section, we drawupon an ecological perspective to develop theoryand research hypotheses.

THEORY AND HYPOTHESES

Interdependence between communities:mutualism and competition

In ecological theory, organizations are consideredto be interdependent when they affect each other’s

050

100150200250300350400450500

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

Year

To

tal s

ales

rev

enu

e in

RM

Bb

illio

ns

Figure 1. Total sales revenue for all zones 1988–2000 (in 1990 RMB)



Figure 2. Geographic locations of national technology development zones in China∗

fates (e.g., growth and mortality). There are twogeneric forms of organizational interdependence:competition and mutualism. As Barnett and Car-roll (1987) noted, ‘When organizations negativelyaffect one another, they are competitive. Whenthey enhance each other’s viability, organizationsare mutualistic’ (Barnett and Carroll, 1987: 400).The density dependence model of the populationecology literature captures competition and mutu-alism between individual organizations within apopulation (Barnett, 1990; Barnett and Carroll,1987). This model, originated by Hannan andFreeman (1989), proposes that an initial increasein the number of organizations in a populationimproves survival chances of the individuals, indi-cating mutualism between organizations. Mutual-ism occurs because organizations ‘making similardemands on the environment combine their efforts,intentionally or otherwise’ to improve an emergingpopulation’s position (Aldrich, 1999: 302). How-ever, as a population increases beyond a certainpoint, competition for similar resources increasesmortality, due to increased competition betweenorganizations. The mutualistic benefits of an initialincrease in density, combined with the competi-tive effects of further increases, create an inverted

U-shaped effect of population density on organi-zational outcomes.

As Barnett and Carroll (1987) noted, ‘Organiza-tional interdependence can exist at several levels:between individual organizations, between pop-ulations of organizations, and between commu-nities of organizations. For the most part, cur-rent organizational researchers think only of theorganizational level’ (Barnett and Carroll, 1987:400, italics added). In this study, we focus oninterdependence between communities and exam-ine how intercommunity relationships can affectcommunity growth. Drawing upon prior work oninterdependence between organizations, we pro-pose that interdependence between communitieshas two forms: mutualism and competition. In ourresearch context (i.e., national technology devel-opment communities), mutualism between tech-nology communities derives from the greater andmore generalized attention that multiple relatedcommunities can attract from external audiencesto their locations. It has been noted that although

∗ By permission from Xiaohong Quan, author. From Saxenian,A., and Quan, X. 2005. In The Software Industry in EmergingMarkets, Commander S (ed). Edward Elgar Publishing Limited:Cheltenham, U.K.; 73–132



natural, industrial, and institutional resources arecritical for community emergence and growth, sub-stantial uncertainty surrounds the nature as well asthe location of the relevant resources for organiz-ing at the outset of new industries (Arthur, 1990;Rauch, 1993). Hence, prospective entrepreneursand investors may have only a tangential under-standing of actual resources present in a givenlocale. For this reason, Romannelli and Khessina(2005) argued that it is perception rather thanactual resources that forms the basis for externalaudiences’ understandings about a region’s attrac-tive characteristics and thus their investment deci-sions. In particular, organizational communitiesare ‘the principal, observable features of regionalindustrial identities, informing the perceptions ofaudiences about the region and, therefore, thesalient public indicators of regional suitability forparticular kinds of business activity’ (Romannelliand Khessina, 2005: 345).

Romannelli and Khessina (2005) further arguedthat the presence of multiple related communitiescan attract greater and more generalized attentionfrom external audiences. The reason is that whenimportant external audiences such as suppliers,buyers, and venture capitalists interact with organi-zations in one community, they are more likely tobecome aware of organizations in other communi-ties if these communities are located in a proximatenetwork of cross-community exchanges (Roman-nelli and Khessina, 2005). As a result, the pres-ence of multiple technology communities withina specific region can enhance the region’s capac-ity for technology development. This can affectindividuals’ decisions about where to locate theirtalents, entrepreneurs’ decisions about where tolocate businesses, and investors’ decisions aboutwhere to invest financial resources, which in turncan lead to mutual benefits for these technologycommunities.

Competition among organizations generallyarises from the joint dependence of multiple orga-nizations on the same set of finite resources (Han-nan and Freeman, 1977, 1989). At the communitylevel, because key resources sought by technologycommunities such as technology entrepreneurs,scientists, engineers, technology project managers,and venture capitalists are in short supply in China,technology communities in a specific region arein a state of competitive interdependence. Ruef(2000) defined carrying capacity as ‘the maxi-mum number of organizations having some identity

(potential or realized) that can be supported by theenvironment at a particular point in time’ (Ruef,2000: 678, italics in original). When environmen-tal carrying capacity is greater than that required,the surplus can support greater demand and onecan anticipate increased community growth. Incontrast, when community size reaches the envi-ronment’s carrying capacity, increased competitionwill likely decrease community growth. In ourresearch context, the combined resource require-ments of multiple technology communities aregreater than the resource requirements of a sin-gle community alone. Hence, as the joint resourcerequirements of multiple communities approach aregion’s carrying capacity for technology devel-opment, one can anticipate decreased communitygrowth.

In summary, we argue that interdependencebetween communities has two forms: mutual-ism and competition, which will jointly affectcommunity growth. In this study, we focus onthree dimensions of intercommunity relationships:regional community density, a focal community’sgeographic proximity to the nearest community,and a focal community’s domain overlap with thenearest community. Regional community densityis defined as the number of organizational com-munities in a specific region (a province or equiv-alently autonomous region and municipality city).Geographic proximity captures the spatial distancebetween a focal community and the nearest neigh-boring community. Domain overlap captures afocal community’s industry specialization relativeto the nearest community. By systematically exam-ining the effects of these three dimensions, weare able to offer a more complete picture of therole of intercommunity relationships in communitygrowth.

We argue that each of these dimensions affectsthe levels of mutualism and competition betweencommunities. The functional form of mutualismand competition between communities that weexpect draws upon the logic of the density depen-dence model of the population ecology literature(Barron, West, and Hannan, 1994; Hannan andFreeman, 1977, 1989; Haveman, 1993). The den-sity dependence model assumes that legitimacy(which leads to mutualism) grows with density at adecreasing rate, while competition grows with den-sity at an increasing rate (Haveman, 1993: 594).Similarly, we propose that mutualism between



communities grows with regional community den-sity, geographic proximity, and domain overlap,at a decreasing rate, while competition betweencommunities grows with these dimensions at anincreasing rate.

More specifically, mutualism between commu-nities grows with these dimensions at a decreas-ing rate because there is a ceiling on the pro-cess when multiple related communities gener-ate more generalized attention from external audi-ences. As a result, the marginal mutualistic bene-fit becomes smaller as these dimensions increase.Furthermore, in our theory competition betweencommunities comes from constraints arising fromthe joint dependence of these communities onthe same set of finite resources. When resourcedemands of communities are far below the carryingcapacity of the environment, the marginal increasein competition between communities associatedwith increase in these dimensions is limited. How-ever, as these dimensions further increase, themarginal increase in competition between com-munities becomes greater as available resourcesdecrease and resource demands of these commu-nities are approaching the carrying capacity ofthe environment. As a result, competition betweencommunities grows at an increasing rate with anincrease in these dimensions. Therefore, at lowlevels, increases in these dimensions serve pri-marily to enhance mutualism between communi-ties. At high levels of these dimensions, increasesstrengthen competition far more than mutualism.Therefore, we expect to observe an inverted U-shaped effect of these dimensions on communitygrowth. However, we acknowledge that since thisecology logic has been mainly tested at the orga-nizational level in previous studies, the questionof whether this logic holds at the community levelis still empirically open. Thus, the predictions ofthis study are partially exploratory, and we aim toempirically test this logic in the context of China’snational technology development communities.

Regional community density and communitygrowth

As noted earlier, previous studies have applied thedensity dependence model to examine the impactof density on organizational founding, growth, andmortality rates and have found substantial support

for this model in a variety of organizational popu-lations (Carroll and Hannan, 2000). Given the con-sistency of empirical results supporting an invertedU-shaped relationship between density and organi-zational outcomes, it is reasonable to predict thatcommunity density may have a similar effect oncommunity growth. This has yet to be tested, butit is an important empirical question. The reasonis that the number of technology communities ina specific region not only will reflect the competi-tion among the communities but will also provideopportunities for resource flows and leveragingacross community boundaries (Porter, 1998b; Tall-man and Phene, 2007).

We argue that an initial increase in regionalcommunity density will have a positive impact oncommunity growth due to the mutualistic benefitsdiscussed above. When the number of technologycommunities in a region is low, increases in den-sity heighten recognition that a given region isappropriate for technology development—or, touse Romanelli and Khessina’s (2005) term, theregion’s industrial identity for technology devel-opment. As a result, important external audi-ences such as prospective entrepreneurs, tech-nology talents, and investors will increasinglyassociate the region with technology develop-ment activities and direct their investment deci-sions toward the region accordingly. This willmutually benefit all of the technology commu-nities—e.g., China’s national technology devel-opment zones—in the region. This conjecture isconsistent with previous research. For example, ina study of organizational form evolution amongdisk array producers, McKendrick and colleagues(2003) observed that the presence of relativelylarge groups of similar organizations in a regionhelps to draw the attention of external observersand thus promotes the development of a new orga-nizational form.

However, as regional community density con-tinues to increase, competition between communi-ties in the region will increase, which will gradu-ally erode the benefits of mutualism. Competitionwith others is likely to undermine an individualcommunity’s growth because communities in thesame region draw upon and compete for a com-mon resource pool. In our research context, it isnot uncommon that national technology develop-ment zones in a specific region (e.g., a province)compete for resources and support from both thecentral and provincial governments. Also, these



zones tend to attract resources from a limited setof entrepreneurs and investors. Thus, as the num-ber of technology communities in a region fur-ther increases, combined resource requirements ofthese communities are more likely to reach theregion’s carrying capacity (Ruef, 2000). Resourcesrequired in common become scarce, making itdifficult for the individual communities to con-tinuously grow. Combined, the mutualistic bene-fits of increased numbers when regional commu-nity density is low plus the competitive pressuresplaced on communities when density increases willjointly create an inverted U-shaped relationshipbetween regional community density and commu-nity growth. Thus, we propose the first hypothesisof this study:

Hypothesis 1: Regional community density willhave an inverted U-shaped relationship with acommunity’s growth.

Geographic proximity to the nearestcommunity and community growth

The ecology literature has paid a fair amount ofattention to geographic proximity/distance at theorganizational level. The basic argument in thisstream of research is that geographic proximitybetween organizations will facilitate resource flowand knowledge spillovers by providing opportu-nities for both planned and serendipitous interac-tions (e.g., Baum and Mezias, 1992; Sorenson andStuart, 2001). At the community level, the impor-tance of geographic proximity between communi-ties has also been discussed by several scholars(e.g., Porter, 1998a, 1998b; Tallman and Phene,2007). For example, Tallman and Phene (2007)argued that geographic proximity plays an impor-tant role in knowledge flows across geographicboundaries of clusters. However, our knowledgeof how geographic proximity may affect commu-nity growth is still limited, and empirical evidenceis particularly lacking.

Further, some of the earlier ecology studies didnot directly measure geographic distance betweenorganizations. Instead, they utilized binary densitymeasures and examined the density dependencemodel on different geographic scales (e.g., den-sity at the national level versus density at the locallevel) (Stuart and Sorenson, 2003: 239). In gen-eral, these studies found that the effect of densityis stronger when density is measured on a more

limited geographic scale (e.g., Baum and Mezias,1992; Carroll and Wade, 1991; Kuilman and Li,2006; Lomi, 1995; Sorenson and Audia, 2000).There are some exceptions, however. In their studyof Manhattan hotels, Baum and Haveman (1997)found that a new entrant’s geographic distance toexisting hotels is negatively related to its size dif-ference with existing hotels and positively relatedto its price difference with existing hotels. In astudy of the spatial distribution of venture capi-tal investments, Sorenson and Stuart (2001) foundthat geographic distance between a venture cap-italist’s main office and the location of a targetfirm reduces the likelihood that the venture capi-talist will invest in the target firm. Furthermore, inanother study of biotechnology firms, Stuart andSorenson (2003) found that geographic proximityto other biotechnology firms, biotechnology patentinventors, venture capital firms, and leading uni-versities have a positive impact on founding rates.

By directly measuring geographic proximitybetween adjacent communities, in this study weexamine how spatial heterogeneity affects com-munity growth. More importantly, while previousstudies have only examined the monotonic impactof geographic proximity/distance, we propose thatgeographic proximity between adjacent commu-nities will have an inverted U-shaped effect oncommunity growth. We argue that at low levelsof geographic proximity between a focal technol-ogy community and the nearest community (i.e.,when the focal community is distantly locatedfrom others), increases in geographic proximitycan produce mutualistic benefits. First, increasesin geographic proximity between adjacent tech-nology communities increase the chance of inter-community learning. Porter (1998a) observed thatindustry clusters located in close geographic prox-imity to others have a greater chance of learn-ing about and deploying cutting-edge informationabout markets and technologies than more isolatedclusters. To illustrate, he discussed the locationof multiple, related clusters including vineyards,wineries, winemaking equipment producers, wine-related university research, etc. in the Napa wineregion as a key source of the region’s ongoingeconomic vitality. Second, increases in geographicproximity between adjacent technology communi-ties also enable communities to draw greater atten-tion to themselves from external audiences. Asgeographic proximity between adjacent commu-nities increases, external audiences’ search costs



can be reduced. As a result, when external audi-ences (e.g., buyers, suppliers, and venture capi-talists) interact with organizations in one commu-nity, they are more likely to learn about organiza-tions in another community if these communitiesare located proximately (Romanelli and Khessina,2005).

However, as a focal technology community’sgeographic proximity to the nearest communityfurther increases, competition between these com-munities is likely to increase. This is becauseclosely located communities will draw upon re-sources from the same geographic locations, thuscreating greater competition for limited resources.For example, technology communities like thenational technology development zones in Chinaimplicitly compete with one another to create inno-vative technologies, which requires the recruit-ment of engineering talent as well as technical‘stars’ (Owen-Smith and Powell, 2004). Accordingto Ruef (2000), as geographic proximity betweenadjacent populations further increases, their com-bined size is more likely to reach the carryingcapacity of the common location from which theydraw resources.

In summary, we argue that at low levels ofgeographic proximity between adjacent technol-ogy communities, increases in geographic prox-imity produce mutualistic benefits. At high levelsof geographic proximity between adjacent technol-ogy communities, increases in geographic proxim-ity increase intercommunity competition. The neteffect produces mutualism at low levels of geo-graphic proximity and the effect shifts to com-petition at high levels of geographic proximity.Accordingly, we propose the following hypothesis:

Hypothesis 2: A focal community’s geographicproximity to the nearest community will havean inverted U-shaped relationship with the focalcommunity’s growth.

Domain overlap with the nearest communityand community growth

In the organization literature, an organization’sdomain consists of the claims it makes with respectto products offered, services provided, and popu-lations served (Levine and White, 1961; Thomp-son, 1967). The overlap of two organizations’domains refers to the fraction of the focal orga-nization’s domain duplicated by the domain of the

other (Baum and Singh, 1994a, 1994b; MacArthur,1972; McPherson, 1983). A standard postulate ofthe ecology perspective is that the intensity ofthe competitive pressure exerted by one organiza-tion on another is proportional to domain overlapbetween these organizations (Hannan and Free-man, 1989; MacArthur, 1972). In studying Cana-dian day-care centers, Baum and Singh (1994a,1994b) operationalized domain overlap as overlapin markets served (age of children served). Theyfound that the number of organizations present inthe focal organization’s domain (i.e., overlap den-sity) is negatively related to organizational found-ing and positively related to organizational mor-tality. Dobrev et al. (2001) operationalized domainoverlap as overlap in automobile producers’ spreadof engine capacity and found that overlap den-sity has a positive effect on organizational mortal-ity. Operationalizing domain overlap as overlap inpatents and patent citations, Podolny et al. (1996)found that a firm’s domain overlap with others inthe population is negatively related to its growth.More formally, Hannan et al. (2007) theorized that‘The expected intensity of the competitive pres-sure exerted by one organization on another nor-mally equals zero if their fundamental niches donot overlap. Otherwise, the expected intensity ofthe competitive pressure rises monotonically withthe thickness of the overlap of their fundamentalniches’ (Hannan et al., 2007: 195–196).

Domain overlap is also an important element inunderstanding intercommunity relationships,although empirical research on this issue is verylimited. For example, Porter (1998b) illustratedcluster intersections (i.e., industry overlaps) byobserving that in Massachusetts such interactions‘have proven to be fertile breeding grounds fornew companies’ (Porter, 1998b: 241). In this study,we define a focal community’s domain overlapwith its adjacent community as the extent to whichthe focal community’s major industries correspondto those of the nearest community. We are inter-ested in how domain overlap between two adjacentcommunities may affect the growth of the focalcommunity.

As noted earlier, prior ecology research hasmainly focused on the competition among orga-nizations with overlapping domains, suggesting amonotonic effect of domain overlap. However,Hannan et al. (2007: 197) raised the possibility thatthe competitive effect of domain overlap may be



overridden by the legitimation effect (i.e., legiti-mation leads to mutualism between organizations)when legitimation is low. In other words, whenlegitimation is low, an increase in overlap (due toan associated rise in density) can increase legiti-mation of the organizational form. Hence, domainoverlap should have a nonmonotonic relationshipwith organizational outcomes. However, to the bestof our knowledge no research has explicitly exam-ined this possibility. In our context, while legiti-mation may not be a concern for technology com-munities (because the Chinese government createdthem and assigned strong economic incentives tomotivate firms to locate within them), we arguethat domain overlap between adjacent communi-ties may affect community growth through bothmutualistic and competitive effects and thus willhave an inverted U-shaped impact on communitygrowth.

More specifically, at low levels of domain over-lap between adjacent communities, increases indomain overlap can produce mutualistic bene-fits. First, adjacent communities with overlap-ping domains are more likely to engage in cross-community communication, resource flows, andinformation exchange. Managers and entrepreneursmay easily share information with others acrosscommunities. The common need for skilledemployees creates mobility opportunities for em-ployees to move easily from one community toanother, and this promotes learning between adja-cent communities and helps with the discovery andimplementation of new ideas, which translate intohigher growth. For example, Porter (1998a) arguedthat the economic benefits of clustering dependupon the presence of multiple interrelated industryclusters with complementary interests that promoteinformation sharing, innovation, and entrepreneur-ship. Second, adjacent communities with overlap-ping domains also attract greater and more general-ized attention from external audiences (Romanelliand Khessina, 2005: 352). Sorenson and Stuart(2001) showed that venture capitalists are morelikely to invest in industries outside their nor-mal industry experience if they have previouslypartnered with other venture capitalists with expe-rience in these industries. Extending this findingfrom venture capitalists to other external audi-ences, we argue that external audiences of organi-zations in one community are more likely to inter-act with organizations in the adjacent communityif these communities have overlapping industry

domains. This is because when adjacent commu-nities have overlapping industry domains, externalaudiences’ prior experience with organizations inone community can be applied in the other com-munity.

However, at high levels of domain overlapbetween adjacent communities, further increases indomain overlap will lead to competition betweenadjacent communities as the two are likely torequire similar resources (e.g., technical expertsand project managers in a particular industry).Their combined requirements for these resourcesare more likely to reach the location’s carry-ing capacity for these particular resources (Ruef,2000), and as a result community growth willdecrease. Furthermore, when domain overlapbetween adjacent communities is very high, adja-cent communities are homogeneous. In general,homogeneity can restrict creativity, innovation,and the range of strategic responses (e.g., Abra-hamson and Fombrun, 1994). At the communitylevel, when adjacent communities have excessivedomain overlap, resources attracted from externalaudiences or generated within these communitieswill become more homogeneous (Romanelli andKhessina, 2005). As a result, the pace of inno-vation in these communities is likely to decline,which can also lead to lower community growth.

In summary, we argue that initial increasesin domain overlap between adjacent communi-ties produce mutualistic benefits but that furtherincreases in domain overlap create more homo-geneous communities competing for the same oroverlapping resources. Combined, we propose aninverted U-shaped relationship between domainoverlap and community growth.

Hypothesis 3: A focal community’s domain over-lap with the nearest community will have aninverted U-shaped relationship with the focalcommunity’s growth.

Controls

Thus far we have argued that three dimensions ofintercommunity relationships will have an invertedU-shaped relationship with community growth. Totest these hypotheses, we also control for alterna-tive explanations for community growth. At thecommunity level, we control for community age,institutional origin, community research intensive-ness, and export intensiveness. We also control



for attributes of the city in which a communityis located, including the city’s political status,gross domestic product (GDP), population, indus-try structure, number of higher education institu-tions, and foreign direct investment (FDI). Further-more, we control for calendar year dummies in ourmodels. The rationale for each is discussed in themeasurement section.

METHODS

Research design and data sources

Our data include all 53 national technology devel-opment zones that were founded in China fromtheir inception through the year 2000. Data werecollected from several sources. One is a proprietaryreport (2001) provided by the Chinese Ministry ofScience and Technology (MST), which providesdata on each zone’s annual sales revenue, exportrevenue, number of employees, and number ofR&D personnel, from the first year that the zonewas founded through the year 2000. We collectedadditional information on China’s national tech-nology development program and on each zonefrom the MST Web site (http://www.most.gov.cn/English/index.htm) as well as the individual zones’Web sites. We also employed a research assis-tant from Renmin University in Beijing to tele-phone zone administrators throughout the countryto verify information obtained from public sources,and these administrators’ responses helped resolvequestions when information conflicted acrosssources. To develop a better understanding ofChina’s technology development program, we con-ducted exploratory and semi-structured interviewswith zone administrators and a range of entre-preneurs in four different technology developmentzones (Beijing, Xi’an, Shanghai, and Shenzhen).

Furthermore, we studied China’s economicgrowth programs and policies since the 1980sin order to distinguish additional variables thatmight influence zone growth. Relying on datafrom China’s Statistical Yearbooks for the rele-vant years, we identified longitudinal data on eachcity’s annual GDP, its population size, the numberof universities and colleges, the size of the city’sFDI, and its industry structure.

As these national technology development zonesare well delineated geographically by the gov-ernment itself, there is no empirical ambiguity

regarding geographic location of a zone or whichindustrial populations are included within it. Thesezones also have a well-defined origination date dueto policy actions of the government. With a clearorigination date for each zone, the research designemployed here avoids left-censoring a zone’s his-tory because we capture data from the first yearthat each new zone was founded; this also allowsus to avoid model misspecification and biased con-clusions regarding patterns of growth (Hunt andAldrich, 1998). Annual data from the foundingyears were collected for the zones and their citycontexts.

Finally, because our data are a yearly time series,observations for each of the 53 national technologydevelopment zones were pooled. In our model esti-mations, the dependent variable is lagged by oneyear behind the independent and control variables.The final data for analysis include 434 zone years.

Measures of independent variables

Following Barnett’s (1990: 45) measure of localpopulation density, we calculated regional com-munity density separately for each community foreach year by using the number of national tech-nology development zones within a focal com-munity’s provincial location. We focused on theprovince level to measure regional communitydensity because national technology developmentzones located in the same province are subjectto the provincial government’s administration andsupport. Hence, all zones in a given provinceshare common rules and regulations that governtheir operations. Empirically, a province may havemultiple national technology development zones,while a city can have only one at maximum, andthus only the specification of a region at the provin-cial level provides variation in regional communitydensity. The values of this variable ranged from 1to 6 in our data.

Consistent with previous studies on geographicproximity (e.g., Baum and Haveman, 1997; Stu-art and Sorenson, 2003), we measured geographicproximity of a community to the nearest commu-nity as follows. We first measured the geographicdistance between communities as the natural logof the distance, in kilometers, between the citywhere the focal zone is located and the city wherethe next closest zone is located. As China’s mainform of intercity transportation is the railroad, dis-tances between the paired cities were derived from



China’s Railroad Bureau data on railroad trans-portation distances, in kilometers. The (logged)values of geographic distance ranged from 3.40 to7.58. To transform geographic distance into a mea-surement of geographic proximity, we subtractedthe values of geographic distance from the maxi-mum value of 7.58 in the data to obtain the valuesof geographic proximity. The values of geographicproximity ranged from 0 to 4.18.

Consistent with Podolny and colleagues’ (1996)measure of niche overlap at the organizationallevel, we measured a community’s domain overlapwith the nearest community as the extent to whicha focal zone’s major industries corresponded withthose of the next closest zone. We first identified azone’s primary industries based on three sources:(1) the MST’s Web site, (2) individual technologyzones’ Web sites, and (3) a publication that intro-duces each of these zones, including their industryfoci (Sun and Zhang, 2001). Next, our researchassistant in Beijing telephoned each zone’s admin-istrators to verify the industry information gatheredfor each zone. We asked whether and when a zoneexperienced significant changes in its major indus-tries from its founding through the year 2000.Among the 53 zones, 9 zones experienced sig-nificant changes, and these occurred when a newmajor industry emerged in a zone. Data on a zone’smajor industries were then updated to reflect thetime-based change in its industry mix. Domainoverlap was calculated as the percentage of thefocal zone’s major industries that were also presentin its next closest zone in the prior year. This mea-sure varies from 0 (i.e., none of a focal zone’smajor industries were present in the next closestzone) to 100 percent (i.e., all of a focal zone’smajor industries were present in the next closestzone).

Measures of control variables

As noted earlier, we have controlled for the fol-lowing variables that could provide alternativeexplanations for community growth. Communityage was measured as the number of years thathave transpired between a zone’s founding yearand the current year, calculated annually. Com-munity institutional origin refers to the fact thatsome of the 53 national technology developmentzones were initially founded by the central gov-ernment, whereas others (e.g., Xi’an Zone and

Nanjing Zone) were initially founded by provin-cial governments and later upgraded to nationalstatus by the central government. Community insti-tutional origin was coded as 1 if a zone was ini-tially founded as a national zone with sponsorship,recognition, and support from the central Chinesegovernment, and 0 otherwise. Community researchintensiveness was calculated as the ratio of R&Dpersonnel to total personnel in all firms in a zoneat the prior year’s end. Community export inten-siveness was measured as the ratio of the value ofexport sales to all sales for all firms within a zonein the prior year. These variables were updatedannually.

To control for the political importance of a com-munity’s city locale, we took advantage of the factthat in China there exists a clear political hierar-chy of cities. The administrative areas in China aredivided into provinces, autonomous regions, andmunicipalities directly under the central govern-ment. Whereas provinces and autonomous regionsmaintain their own local governments situated inthe capital cities, the municipality cities reportdirectly to the central government in Beijing.Provincial capitals are the political center of eachprovince and autonomous region and have theirown provincial resource bases. All other cities—called subprovincial cities—are contained within aprovince or an autonomous region and are subjectto political control from its provincial government.We created two dummy variables by using sub-provincial cities as the base comparison group:municipality city and provincial capital city.

Local city GDP (in RMB 10,000) was controlledbecause it indicates the size of the local economy,the growth of which could in turn influence growthof the local technology zone. The measure was cor-rected for inflation and log transformed in the prioryear. We also controlled for the local city’s popula-tion as an indicator of the size of the local supplyof labor. Local city population was measured in10,000s and log transformed in the prior year. Wecontrolled for local city industry structure, mea-sured as the proportion of service industries in thecity’s GDP in the prior year. This measure is usedin China to capture the extent to which a city isindustrialized.

We also controlled for the number of higher edu-cation institutions (i.e., universities and colleges)in a city (log transformed), which could influencecommunity growth by providing educated work-ers for the technology development zones. We



controlled for city foreign direct investment (FDI),measured as capital invested in a city by sourcesnot from China but rather from a company head-quartered outside of China. FDI includes all for-eign capital invested in a given city in the prioryear, and data were updated annually. As FDI datain China’s Statistical Yearbook are in U.S. dollars(US $10,000), data were transformed to Chinesecurrency (RMB 10,000) using the exchange rateat the end of the corresponding year. The measurewas further corrected for inflation and finally logtransformed.

Furthermore, China has experienced substantialeconomic, social, and institutional change since thefirst zone was founded in 1988. To account for thepossibility that the growth of China’s high technol-ogy development zones may vary systematicallyover years, our models controlled for calendar yeardummy variables (Podolny et al., 1996). The inclu-sion of the calendar year dummy variables can alsodistinguish the effects of zone age from the effectsof calendar time.

Analysis of community growth

To examine the effects of intercommunity relation-ships on community growth, we estimate mod-els of growth in terms of a zone’s annual salesrevenues. These data are collected by the zoneadministrators from the resident firms, then theyare aggregated to the zone level and reported annu-ally to the MST, which publishes the data. Salesrevenue data were updated annually (in ChineseRMB 1,000) and corrected for inflation (usingRMB value in 1990).

Following prior research on organizationalgrowth (e.g., Barron et al., 1994; Baum andMezias, 1992; Podolny et al., 1996; Sorensen,1999), we model community growth in sales rev-enue as a function of a community’s sales revenueand a number of covariates that can affect com-munity growth:

Si,t+1/Sit = (Sit )α−1 exp(βxit + εi,t+1), (1)

where S is a time-varying measure of communitysales revenue, α is an adjustment parameter thatindicates how growth rates depend on communitysales revenue, and β is a vector of parameterscharacterizing the effects of covariates (xit ). If wetake the log of Equation 1 and rearrange terms, wehave the log-linear model:

ln(Si,t+1) = α ln(Sit ) + βxit + εi,t+1. (2)

The data are arranged in the form of a pooledcross-section time series dataset, with each zonecontributing a time series of observations of differ-ing lengths. The length of each zone’s time seriesdiffers because these zones may be founded in dif-ferent years. In a pooled cross-sectional dataset,zones have multiple observations corresponding toeach year of observation. However, these obser-vations may not be independent of one another.A robust variance estimator for cluster data cancorrect for nonindependence. It essentially treatseach cluster (i.e., all observations associated withone zone) as a super-observation that contributesto the variance estimate and thus generates robustestimates (Westphal and Khanna, 2004). Thus, weincluded the robust option in our models to calcu-late robust standard errors for coefficients (Stata,2003: 328).

RESULTS

Table 1 reports means, standard deviations, andcorrelations of all variables except calendar yeardummies used in the analysis. Table 2 presents theestimates of the models on community growth.Model 1 includes controls only, Model 2 addsthe effects of regional community density and itssquared term, and Model 3 includes the effects ofgeographic proximity and its squared term. Finally,Model 4 includes the effects of domain overlap andits squared term.

Hypothesis 1 predicts that regional communitydensity has an inverted U-shaped relationship withcommunity growth. In Model 2 in Table 2, thecoefficient for regional community density is pos-itive and significant (b = 0.15, p < 0.01), and thecoefficient for its squared term is negative and sig-nificant (b = −1.4E-2, p < 0.05). Thus, Hypothe-sis 1 is supported.

Based upon the results of Model 2, if all othervariables take their mean values, a zone’s expectedsales revenue is 1.31 billion (in 1990 RMB) whenregional community density is 1. The highestexpected zone sales revenue is 1.80 billion (in1990 RMB), which occurs when regional commu-nity density is 5. However, the expected zone salesrevenue would be 1.72 billion (in 1990 RMB)



Tabl

e1.

Mea

ns,

stan

dard

devi

atio

ns,

and

corr

elat

ions

ofal

lva

riab

les

inan

alys

esa

Var

iabl

esM

ean

S.D

.1

23

45

67

89

1011

1213

14

1.C

omm

unity

sale

s(l

og)

14.3

11.

30—

2.R

egio

nal

com

mun

ityde

nsity

2.73

1.65

0.12

—3.

Geo

grap

hic

prox

imity

toth

ene

ares

tco

mm

unity

2.66

0.93

0.25

0.57

—4.

Dom

ain

over

lap

with

the

near

est

com

mun

ity0.

800.

160.

070.

330.

40—

5.C

omm

unity

age

6.37

2.68

0.63

−0.0

8−0

.03

−0.0

7—

6.C

omm

unity

inst

itutio

nal

orig

in0.

680.

46−0

.01

0.09

−0.0

40.

04−0

.20

—7.

Com

mun

ityre

sear

chin

tens

iven

ess

0.12

0.06

−0.0

5−0

.31

−0.1

5−0

.11

−0.1

80.

01—

8.C

omm

unity

expo

rtin

tens

iven

ess

0.11

0.14

0.20

0.31

0.22

0.04

0.14

0.03

−0.3

4—

9.M

unic

ipal

ityci

ty0.

070.

250.

35−0

.26

0.07

−0.2

00.

080.

190.

140.

05—

10.

Prov

inci

alca

pita

lci

ty0.

440.

49−0

.07

−0.2

9−0

.26

0.06

0.10

−0.2

10.

27−0

.31

−0.2

4—

11.

City

GD

P(l

og)

14.4

30.

820.

670.

160.

300.

020.

400.

040.

030.

220.

45−0

.08

—12

:C

itypo

pula

tion

(log

)8.

340.

680.

39−0

.21

0.08

−0.0

50.

130.

080.

29−0

.24

0.42

0.12

0.55

—13

.C

ityin

dust

ryst

ruct

ure

0.37

0.09

0.31

−0.0

9−0

.19

−0.2

20.

400.

040.

120.

050.

170.

370.

200.

01—

14.

No.

ofhi

gher

educ

atio

nin

stitu

tions

inth

eci

ty(l

og)

2.26

0.94

0.25

−0.4

6−0

.25

−0.2

20.

110.

000.

50−0

.30

0.40

0.55

0.33

0.61

0.41

—15

.C

ityFD

I(l

og)

10.5

92.

930.

370.

280.

360.

090.

22−0

.14

−0.1

80.

320.

240.

030.

410.

120.

240.

02

N=

434

zone

year

s.a

Cor

rela

tions

equa

lto

orla

rger

than

0.10

are

sign

ifica

ntat

the

leve

lof

p<

0.05

.

when regional community density further increasesto 6 (the largest value in the data). In other words,all else being equal, a zone’s expected sales rev-enue would be 37 percent (= 1.80/1.31-1) greaterif regional community density were to change from1 to 5, and the expected sales revenue wouldbecome smaller as regional community density fur-ther increases.

Hypothesis 2 predicts that a focal community’sgeographic proximity to the nearest community hasan inverted U-shaped relationship with the focalcommunity’s growth. The results of Model 3 showthat the coefficient for geographic proximity (b =0.48, p < 0.001) is positive and significant and thatthe coefficient for its squared term (b = −0.08,p < 0.01) is negative and significant. These resultssupport Hypothesis 2. An examination of Figure 1shows that the Urumqi zone (which is located inthe far northwest of China) is exceptionally distantfrom the nearest zone. To test the robustness of theproximity findings, we dropped the Urumqi zoneand reestimated the model, and the results wereconsistent with the original finding: the coefficientfor geographic proximity is 0.59 (p < 0.001), andthe coefficient for its squared term is −0.09 (p <

0.01). Again, these results support Hypothesis 2.Based upon the results of Model 3, if all other

variables take their mean values, a zone’s expectedsales revenue is 0.61 billion (in 1990 RMB) whenits geographic proximity to the nearest zone is 0(i.e., a distance of 1,959 kilometers—the largestgeographic distance in the data). The highestexpected zone sales revenue is 1.64 billion (in1990 RMB), which occurs when a zone’s geo-graphic proximity to the nearest zone is 3 on theproximity scale (i.e., a distance of 98 kilometers).However, the expected zone sales revenue wouldbe 1.55 billion (in 1990 RMB) when a zone’s geo-graphic proximity to the nearest zone is 4.18 (i.e.,a distance of 30 kilometers—the smallest geo-graphic distance in the data). In other words, allelse being equal, a zone’s expected sales revenuewould be 169 percent (= 1.64/0.61-1) greater ifits geographic distance to the nearest zone were tochange from 1,959 kilometers to 98 kilometers andthe expected sales revenue would become smalleras the geographic distance further decreases.

Hypothesis 3 predicts that a focal community’sdomain overlap with the nearest community hasan inverted U-shaped relationship with the focalcommunity’s growth. In Model 4, the coefficientfor domain overlap is positive and significant (b =



Table 2. Models of growth of China’s national technology development zones

Variables Model 1 Model 2 Model 3 Model 4

PredictorsRegional community density 0.15∗∗

(0.06)Regional community density squared −1.4E-2∗

(6.5E-3)Geographic proximity to the nearest community 0.48∗∗∗

(0.14)Geographic proximity squared −0.08∗∗

(0.03)Domain overlap with the nearest community 2.30∗∗

(0.75)Domain overlap squared −1.22∗∗

(0.45)ControlsLagged community sales (log) 0.71∗∗∗ 0.69∗∗∗ 0.67∗∗∗ 0.68∗∗∗

(0.04) (0.05) (0.05) (0.04)Community age 0.00 −0.01 0.00 −0.01

(0.03) (0.03) (0.03) (0.02)Community institutional origin −0.09 −0.11† −0.07 −0.12∗

(0.05) (0.06) (0.05) (0.05)Community research intensiveness 0.88† 1.09∗ 0.53 0.84†

(0.52) (0.53) (0.52) (0.50)Community export intensiveness 0.01 0.01 −0.02 0.03

(0.15) (0.15) (0.15) (0.14)Municipality city 0.02 0.32∗ 0.03 0.00

(0.09) (0.13) (0.08) (0.09)Provincial capital city −0.14∗ −0.08 −0.15∗ −0.25∗∗

(0.07) (0.07) (0.07) (0.08)City Population (log) 0.06 0.04 0.01 0.05

(0.05) (0.06) (0.05) (0.05)City GDP (log) 0.13∗∗ 0.09∗ 0.12∗ 0.12∗

(0.05) (0.04) (0.05) (0.05)City industry structure 0.65 0.41 0.93† 0.86

(0.59) (0.51) (0.56) (0.57)City higher education institutions (log) 0.01 0.06 0.07 0.08

(0.06) (0.06) (0.06) (0.06)City FDI (log) 0.02∗ 0.02∗ 0.02∗ 0.02∗

(0.01) (0.01) (0.01) (0.01)Calendar year dummies Included Included Included IncludedConstant 1.67∗∗ 2.64∗∗∗ 1.82∗∗ 0.97

(0.56) (0.62) (0.62) (0.64)F-value 198.38∗∗∗ 181.67∗∗∗ 193.60∗∗∗ 196.24∗∗∗

R-Squared 0.90 0.91 0.91 0.91

N = 434 zone years. Robust standard errors are reported in parentheses.Significance levels: ∗∗∗ p < 0.001, ∗∗ p < 0.01, ∗ p < 0.05, †p < 0.10 (two-tailed tests).

2.30, p < 0.01), and the coefficient for its squaredterm is negative and significant (b = −1.22, b <

0.01). These results thus support the prediction ofHypothesis 3.

Based upon the results of Model 4, if all othervariables take their mean values, a zone’s expectedsales revenue is 0.49 billion (in 1990 RMB) whenits domain overlap with the nearest zone is 0. Thehighest expected zone sales revenue is 1.64 billion

(in 1990 RMB), which occurs when a zone’sdomain overlap with the nearest zone is 94 percent.However, the expected zone sales revenue wouldbe 1.49 billion (in 1990 RMB) when its domainoverlap with the nearest zone is 100 percent. Inother words, all else being equal, a zone’s expectedsales revenue would be 235 percent (=1.64/0.49-1)greater if its domain overlap with the nearest zonewere to change from 0 to 94 percent, and the



expected sales revenue would become smaller asthe domain overlap further increases.

The results reported above support our argu-ments regarding the nonmonotonic effects ofregional community density, geographic proxim-ity, and domain overlap on community growth.As a supplementary analysis, we also estimatedzone level fixed-effect models by including 52zone dummy variables (there are 53 zones in total)in the models. The results of this analysis arereported in the Appendix. These results show thatthe coefficients for geographic proximity and itssquared term are significant. However, regionalcommunity density and its squared term, as wellas domain overlap and its squared term, are notsignificant. These nonsignificant results are likelydue to the fact that regional community densityand domain overlap did not vary substantially overtime in this study. Thus, the effects of these vari-ables are not distinguishable from the zone levelfixed effects (Judge et al., 1985) (c.f. Jensen andZajac, 2004: 514–516). As a consequence, thezone level fixed-effect models may not be appro-priate for testing the hypothesized relationships.Indeed, all of the models with zone-level fixedeffects cannot produce F values and the associatedp values, suggesting that the coefficients of pre-dictors estimated in these fixed-effect models maynot be reliable. Therefore, we interpret our find-ings based upon the cross-sectional results reportedin Table 2. Nonetheless, we acknowledge that ourfindings may stem from cross-sectional variationin the data.

DISCUSSION AND IMPLICATIONS

In this study, we theoretically articulate and empir-ically test how three dimensions of intercommu-nity relationships are expected to affect communitygrowth. With a unique dataset on all national tech-nology development zones founded in China fromtheir inception through the year 2000, we foundthat regional community density, a focal commu-nity’s geographic proximity to and domain overlapwith the nearest community have an inverted U-shaped relationship with the focal community’sgrowth. These findings support our argument thatorganizational communities are interdependent andthat interdependence between communities, whichincludes both mutualism and competition, has asignificant impact on community growth. While

several studies have examined the extent to whichpopulations within a community are interdepen-dent (e.g., Ruef, 2000; Wade, 1995, 1996), webelieve this is one of the first empirical studies todemonstrate that organizational communities, eachof which contains several populations of firms,have an impact on one another’s growth.

Implications for ecology arguments

The existing literature on organizational ecol-ogy has provided consistent empirical support foran inverted U-shaped relationship between den-sity and organizational outcomes. This study hasdemonstrated that regional community density alsohas an inverted U-shaped relationship with com-munity growth. This finding supports the argumentthat the number of technology communities in aspecific region not only reflects competition amongcommunities, but it also provides opportunities forresource flows and leveraging across communityboundaries (Porter, 1998b; Tallman and Phene,2007).

The significant effects of geographic proxim-ity and domain overlap have important implica-tions for ecology arguments. We have simultane-ously examined geographic proximity and domainoverlap at the community level, whereas priorecology research has looked at either one or theother—primarily the latter—and only within orga-nizational populations. More importantly, extantecology research has only examined the mono-tonic effects of geographic proximity and domainoverlap between organizations on organizationaloutcomes. These prior studies link increased geo-graphic proximity and domain overlap with in-creased competition among organizations, and sothey have been shown to adversely affect organi-zations (e.g., Dobrev et al., 2001; Podolny et al.,1996; Sorenson and Stuart, 2001; Stuart and Soren-son, 2003). In contrast, we proposed and found anonmonotonic, inverted U-shaped effect of geo-graphic proximity and domain overlap betweencommunities on community growth. These resultssupport the argument that geographic proximityand domain overlap between adjacent technologycommunities are important dimensions of inter-community relationships, and that both mutualisticand competitive forces play out between commu-nities in ways that jointly affect community out-comes. Our nonmonotonic theoretical argumentsand consistent empirical findings add new insights



to the effects of geographic proximity and domainoverlap and hopefully will inspire future studies toexamine these effects in other organizational andcommunity contexts.

Implications for understanding communitygrowth and cluster development

This study has contributed to a better under-standing of community growth in several ways.First, to the best of our knowledge, this studyis among the first empirical investigations of therole of intercommunity relationships in the growthof organizational communities. While some schol-ars (e.g., Porter, 1998a, 1998b; Romanelli andKhessina, 2005; Saxenian and Hsu, 2001) haveobserved the importance of connections betweencommunities, we contribute to the literature bytheoretically delineating three dimensions of inter-community relationships and empirically examin-ing how these different dimensions affect com-munity growth. Our findings on the significantimpact of regional community density, intercom-munity geographic proximity and domain overlapon community growth demonstrate that organi-zational communities are interdependent and areparticularly affected by their relationships withneighboring communities. Porter (1998b: 241) hassuggested that cluster development often becomesvibrant at the intersection of clusters becauseinsights, skills, and technologies from differentfields and directions merge, thus sparking newbusinesses and stimulating innovation. Our studyadds greater specificity to Porter’s (1998b) insightsby showing that cluster intersection can occuralong two dimensions: geographic distance andindustry overlap between clusters. This study hasshown that changes in geographic proximity anddomain overlap between adjacent communitiesalter the outcomes obtainable by a focal commu-nity.

Second, this study examined the extent to whichboth mutualistic and competitive forces coex-ist between organizational communities. The fewscholars who have addressed intercommunity rela-tionships (Porter, 1998a, 1998b; Romanelli andKhessina, 2005; Saxenian and Hsu, 2001; Tall-man and Phene, 2007) have primarily focused onmutualistic effects in the form of intercommunitylearning, resource and knowledge exchanges, and

enhanced visibility to external audiences. How-ever, the possibility that intercommunity rela-tionships are characterized by the coexistence ofmutualism as well as competition has not beenotherwise studied. Our study advances this lineof inquiry by demonstrating that intercommunityrelationships are characterized by a combinationof mutualism and competition between organi-zational communities. Vigorous competition canoccur in such areas as acquiring scarce resourcesand attracting and retaining employees. The pres-ence of multiple communities with overlappingdomains and within a certain distance enhancesthe intensity of competition between communities.Meanwhile, mutualistic benefits also accrue, par-ticularly when two communities have an optimallevel of geographic distance and industry overlap.Therefore, we provide a more complete picture ofhow organizational interdependence operates at thecommunity level.

Practical implications

While we expected that intercommunity relation-ships would have a significant impact on a focalcommunity’s growth, the magnitude of the effectsexceeded our expectations. As we saw in theresults section of this article, all else being equal,a zone’s expected sales revenue would be 37 per-cent greater if regional community density wereto change from 1 to 5, and sales revenue wouldbe smaller as regional community density fur-ther increases. All else being equal, a commu-nity’s expected sales revenue would be 169 percentgreater if its geographic distance to the nearestcommunity were to decrease from 1,959 kilome-ters to 98 kilometers, and sales revenue wouldbe smaller as the geographic distance furtherdecreases. Finally, all else being equal, a commu-nity’s expected sales revenue would be 235 percenttimes greater if its domain overlap with the near-est zone were to increase from 0 to 94 percentand would be smaller if the domain overlap furtherincreases.

These findings are especially important consid-ering that most of the variables included in thisstudy are beyond zone administrators’ and policymakers’ ability to influence, at least in the shortterm, because geographic, economic, and socialdifferences between communities and regions can-not be changed quickly. In contrast, the extent ofdomain overlap with other communities represents



a strategic variable for zone administrators becausethey can select and modify a zone’s industry mixby selectively admitting firms in targeted indus-tries. Similarly, when a new technology devel-opment zone is to be founded, manipulating thezone’s location and its geographic proximity to thenearest zone is a strategic variable available to pol-icy makers. The location choices for new zones canaffect not only a region’s community density butalso the new zones’ geographic proximity to adja-cent zones as well as existing zones’ geographicproximity to adjacent zones (i.e., one of the newlyfounded zones may be their adjacent zone). Hence,the strategic implications of our findings are sig-nificant, suggesting that zone administrators andpolicy makers must attend not only to the inter-nal dynamics within a specific zone but also to thezone’s relationship with other zones, particularlythe number of other zones in the region and thefocal zone’s geographic proximity to and domainoverlap with adjacent zones.

Limitations and directions for future research

The preceding discussion should be consideredin light of the study’s limitations. One is that itwas conducted within a single country during aperiod of economic transition from a planned econ-omy to a market economy. Other research (e.g.,Perez-Aleman 2005) has shown that governmentactions in several countries can play an impor-tant role in facilitating the growth of technologycommunities. It may be that the government ofChina has also played a role in community growthbeyond the policy it enacted to create the zonesand the multiple incentives (e.g., tax, land) it putin place to encourage the founding of technol-ogy ventures within the zones. Thus, it is possiblethat differential growth rates across these tech-nology development zones may simply be dueto unmeasured but different reactions to govern-ment policies. In this study we have addressedthis alternative explanation in two ways. First, wecontrolled for calendar year dummy variables tocapture the overall growth in the Chinese economyover the period of observation and to account forthe possibility that the Chinese government mayhave different policy priorities over time, encour-aging technology development in lieu of othereconomic development options. Second, we con-trolled for zones’ institutional origins (whether a

specific zone was initially founded by the cen-tral government or by a provincial government)to account for the possibility that the initial pol-icy differences in their founding conditions couldhave an imprinting effect that subtly influencesthese zones’ growth. Nonetheless, we cannot com-pletely rule out the possibility that unspecified andunmeasured government policies may have a sig-nificant impact on the growth of these technologyzones.

Further, while we have controlled for manyalternative explanations for technology communitygrowth in our research context, an extension of thisline of research could make cross-country compar-isons to enhance variation in national economies(emerging, developed, etc.), and institutional dif-ferences in the governance of such countries. Forexample, Porter (1998b: 230) has argued that thedepth and breadth of industrial clusters in devel-oped economies are usually greater than those indeveloping economies. Also, cross-country com-parisons could include wholly commercial com-munities like science parks established to make aprofit for the developers rather than to serve gov-ernment priorities.

Second, in this study our arguments are devel-oped at the community level of analysis, andwe consider commensalistic relationships betweencommunities because these communities are ‘like’social units, and so symbiotic relationships shouldnot apply, at least in the early days of the evolutionof these communities (Aldrich, 1999: 301–302).However, one could imagine in the future thatcommunities might become more specialized andthus become ‘unlike’ one another in fundamentalways. While all would be expected to follow a‘high technology’ trajectory (in accordance withChina’s regulations), one can imagine that somecommunities will evolve to specialize in subsets ofthe focal industries that differ in significant ways.For example, among the technologies targeted byChina for development are ocean technologies andnuclear applications of technology. It could beargued that research and development in oceantechnologies and industrial applications of nucleartechnology are sufficiently different to present thepotential for symbiotic relationships between com-munities thus specialized.

In conclusion, to our knowledge no other large-scale study of organizational communities hasexamined intercommunity relationships to thisextent, as most community ecology research has



focused on within-community population dynam-ics (e.g., Ruef, 2000; Wade, 1996). In this study,we used a unique dataset on all 53 national tech-nology development zones founded in China fromtheir inception through the year 2000 to investi-gate how intercommunity relationships affect thegrowth of organizational communities. We foundthat regional community density and a commu-nity’s geographic proximity to and domain overlapwith the nearest community have an inverted U-shaped effect on the focal community’s growth.Our findings demonstrate that intercommunity rela-tionships have both mutualistic and competitivecomponents.

ACKNOWLEDGEMENTS

We would like to thank Editor Will Mitchell, andthe two anonymous referees for their constructivesuggestions and insightful comments. The arti-cle benefited significantly from the comments ofGautam Ahuja, Raffi Amit, Mark Kennedy, OlgaKhessina, Alan Meyer, Nandini Rajagopalan, andother participants of the first Grief Entrepreneur-ship Research Symposium at the University ofSouthern California (2008). We would also liketo thank Xudong Gao, Michael Hitt, MarjorieLyles, Sam Park, Laszlo Tihanyi, Weiying Zhang,Changhui Zhou, Li-An Zhou, and other partic-ipants of the research roundtable at the secondconference on ‘China-U.S. Relations’ in Beijing(2005) for their comments on an earlier version ofthe article. We are grateful for the help of GuoqingGuo, Yu Li, and Hui Zheng in data collection.

REFERENCES

Abrahamson E, Fombrun CJ. 1994. Macrocultures: deter-minants and consequences. Academy of ManagementReview 4: 728–755.

Aldrich HE. 1999. Organizations Evolving . Sage Publi-cations: Thousand Oaks, CA.

Aldrich HE, Ruef M. 2006. Organizations Evolving (2ndedn). Sage Publications: London, UK.

Arthur WB. 1990. ‘Silicon Valley’ locational clusters:when do increasing returns imply monopoly?Mathematical Social Sciences 19: 235–251.

Astley WG. 1985. The two ecologies: population andcommunity perspectives on organizational evolution.Administrative Science Quarterly 30: 224–241.

Barnett WP. 1990. The organizational ecology of a tech-nological system. Administrative Science Quarterly35: 31–60.

Barnett WP, Carroll GR. 1987. Competition and mutual-ism among early telephone companies. AdministrativeScience Quarterly 32(3): 400–421.

Barron DN, West E, Hannan MT. 1994. A time to growand a time to die: growth and mortality of creditunions in New York, 1914–1990. American Journalof Sociology 100: 381–421.

Baum JAC, Haveman HA. 1997. Love thy neighbor?Differentiation and agglomeration in the Manhattanhotel industry, 1898–1990. Administrative ScienceQuarterly 42: 304–338.

Baum JAC, Mezias SJ. 1992. Localized competition andorganizational failure in the Manhattan hotel industry,1898–1990. Administrative Science Quarterly 37:580–604.

Baum JAC, Singh JV. 1994a. Organizational niches andthe dynamics of organizational founding. OrganizationScience 5: 483–481.

Baum JAC, Singh JV. 1994b. Organizational niches andthe dynamics of organizational mortality. AmericanJournal of Sociology 100: 346–380.

Carroll GR, Hannan MT. 2000. The Demography ofCorporations and Industries . Princeton UniversityPress: Princeton, NJ.

Carroll GR, Wade J. 1991. Density dependence in theorganizational evolution of the American brewingindustry across different levels of analysis. SocialScience Research 20: 271–302.

Chiles T, Meyer A, Hench T. 2004. Organizationalemergence: the origin and transformation of Branson,Missouri’s musical theatres. Organization Science 15:499–519.

DFL International. 1999. Report on U.S. CommercialTechnology Transfers to the People’s Republic ofChina . U.S. Bureau of Export Administration:Washington, D.C.

Dobrev SD, Kim TY, Hannan MT. 2001. Dynamics ofniche width and resource partitioning. AmericanJournal of Sociology 106: 1299–1337.

Enright M. 1999. Regional clusters and firm strategy. InThe Dynamic Firm: The Role of Technology, Strategy,Organization, and Regions , Chandler A, Hagstrom JP,Solvell O (eds). Oxford University Press: New York.

Freeman JH, Audia PG. 2006. Community ecology andthe sociology of organizations. Annual Review ofSociology 32: 145–169.

Hannan MT, Freeman J. 1977. The population ecologyof organizations. American Journal of Sociology 82:929–964.

Hannan MT, Freeman J. 1989. Organizational Ecology .Harvard University Press: Cambridge, MA.

Hannan MT, Polos L, Carroll GR. 2007. Logics of Orga-nization Theory: Audiences, Codes, and Ecologies .Princeton University Press: Princeton, NJ.

Haveman HA. 1993. Follow the leader: mimetic isomor-phism and entry into new markets. Administrative Sci-ence Quarterly 38: 593–627.

Hunt CS, Aldrich H. 1998. The second ecology: thecreation and evolution of organizational communitiesas exemplified by the commercialization of the WorldWide Web. In Research in Organizational Behavior



(Vol. 20), Staw BK, Cummings L (eds). JAI Press:Greenwich, CT; 267–302.

Jensen M, Zajac EJ. 2004. Corporate elites and corporatestrategy: how demographic preferences and structuralposition shape the scope of the firm. StrategicManagement Journal 25(6): 507–524.

Judge GG, Griffiths WE, Hill RC, Lee T-C. 1985. TheTheory and Practice of Econometrics . Wiley: NewYork.

Krugman P. 1991. Increasing returns and economic geog-raphy. Journal of Political Economy 99: 483–499.

Kuilman J, Li J. 2006. The organizers’ ecology: anempirical study of foreign banks in Shanghai.Organization Science 17: 385–401.

Levine S, White PE. 1961. Exchange as a conceptualframework for the study of interorganizationalrelationships. Administrative Science Quarterly 5:583–601.

Li H, Atuahene-Gima K. 2001. Product innovationstrategy and the performance of new technologyventures in China. Academy of Management Journal44: 1123–1134.

Lomi A. 1995. The population ecology of organizationalfounding: location dependence and unobservedheterogeneity. Administrative Science Quarterly 40:111–144.

MacArthur RH. 1972. Geographical Ecology: Patterns inthe Distribution of Species . Princeton University Press:Princeton, NJ.

Mathews JA. 1997. A Silicon Valley of the east:creating Taiwan’s semiconductor industry. CaliforniaManagement Review 39: 26–55.

McKendrick DG, Jaffee J, Carroll GR, Khesssina OM.2003. In the bud? Disk array producers as a (possible)emergent organizational form. Administrative ScienceQuarterly 48: 60–93.

McPherson JM. 1983. An ecology of affiliation.American Sociological Review 48: 519–535.

Owen-Smith J, Powell WW. 2004. Knowledge networksas channels and conduits: the effects of spillovers inthe Boston biotechnology community. OrganizationScience 15: 5–21.

Perez-Aleman P. 2005. Cluster formation, institutions andlearning: the emergence of clusters and development inChile. Industrial and Corporate Change 14: 651–677.

Phan PH, Siegel DS, Wright M. 2005. Science parksand incubators: observations, synthesis and futureresearch. Journal of Business Venturing 20: 165–182.

Piore M, Sabel C. 1984. The Second Industrial Divide.Basic Books: New York.

Podolny JM, Stuart TE, Hannan MT. 1996. Networks,knowledge, and niches: competition in the worldwidesemiconductor industry, 1984–1991. American Jour-nal of Sociology 102: 659–689.

Porter ME. 1998a. Clusters and the new economics ofcompetition. Harvard Business Review 76(6): 77–90.

Porter ME. 1998b. On Competition . Harvard BusinessSchool Press: Boston, MA.

Rauch JE. 1993. Does history matter when it only matterslittle? The case of city-industry location. QuarterlyJournal of Economics 108: 843–867.

Romanelli E, Khessina OM. 2005. Regional industrialidentity: cluster configurations and economic devel-opment. Organization Science 16(4): 344–358.

Ruef M. 2000. The emergence of organizational forms:a community ecology approach. American Journal ofSociology 106(3): 658–714.

Saxenian A. 1994. Regional Advantage: Culture andCompetition in Silicon Valley and Route 128 . HarvardUniversity Press: Cambridge, MA.

Saxenian A, Hsu J. 2001. The Silicon Valley-Hsinchuconnection: technical communities and industrialupgrading. Industrial and Corporate Change 10:893–920.

Schoonhoven CB, Eisenhardt KM. 1993. Entrepreneurialenvironments: incubator region effects on the birthof new technology-based firms. Advances in GlobalHigh-Technology Management 3: 149–175.

Sorensen JB. 1999. Executive migration and interorga-nizational competition. Social Science Research 28:289–315.

Sorenson O. 2003. Social networks and industrialgeography. Journal of Evolutionary Economics 13:513–527.

Sorenson O, Audia PG. 2000. The social structure ofentrepreneurial activity: geographic concentration offootwear production in the United States, 1940–1989.American Journal of Sociology 106(2): 424–462.

Sorenson O, Stuart TE. 2001. Syndication networks andthe spatial distribution of venture capital investments.American Journal of Sociology 106: 1546–1588.

Stata. 2003. Reference N-R. Stat Press: College Station,TX.

Stuart T, Sorenson O. 2003. The geography of opportu-nity: spatial heterogeneity in founding rates and theperformance of biotechnology firms. Research Policy32: 229–253.

Sun H, Zhang C. (eds). 2001. Venture Capital Market List(in Chinese). China Financial & Economic PublishingHouse: Beijing, China.

Tallman S, Jenkins M, Henry N, Pinch S. 2004. Knowl-edge, clusters, and competitive advantage. Academy ofManagement Review 29: 258–271.

Tallman S, Phene A. 2007. Leveraging knowledge acrossgeographic boundaries. Organization Science 18:252–260.

Thompson JD. 1967. Organizations in Action. McGraw-Hill: New York.

Wade J. 1995. Dynamics of organizational communitiesand technological bandwagons: an empirical investi-gation of community evolution in the microprocessormarket. Strategic Management Journal Summer Spe-cial Issue 16: 111–133.

Wade J. 1996. A community-level analysis of sources andrates of technological variation in the microprocessormarket. Academy of Management Journal 39:1218–1244.

Westphal JD, Khanna P. 2004. Keeping directors in line:social distancing as a control mechanism in thecorporate elite. Administrative Science Quarterly 48:361–398.



Appendix. Supplementary Analysis with Zone Level Fixed Effectsa,b

Variables Model 1 Model 2 Model 3 Model 4

PredictorsRegional community −0.06

density (0.20)Regional community 0.03

density squared (0.04)Geographic proximity to 2.29∗∗∗

the nearest community (0.50)Geographic proximity −0.48∗∗∗

squared (0.13)Domain overlap with the

nearest communityAutomatically

droppedDomain overlap squared −0.54

(0.61)ControlsLagged community sales 0.34∗∗∗ 0.34∗∗∗ 0.34∗∗∗ 0.34∗∗∗

(log) (0.05) (0.05) (0.05) (0.06)Community research 0.67 0.66 0.67 0.68

intensiveness (0.53) (0.53) (0.53) (0.53)Community export 0.19 0.15 0.19 0.19

intensiveness (0.22) (0.22) (0.22) (0.21)City population (log) −0.01 0.00 −0.01 0.00

(0.09) (0.08) (0.09) (0.08)City GDP (log) −0.01 −0.01 −0.01 −0.02

(0.05) (0.05) (0.05) (0.05)City industry structure 0.68 0.69 0.67 0.69

(0.78) (0.77) (0.78) (0.78)City higher education 0.15 0.16 0.15 0.15

institutions (log) (0.12) (0.12) (0.11) (0.11)City FDI (log) −0.01 −0.01 −0.02 −0.02

(0.01) (0.01) (0.02) (0.02)Zone dummies Included Included Included IncludedCalendar year dummies Included Included Included IncludedConstant 9.56∗∗∗ 9.58∗∗∗ 6.88∗∗∗ 9.77∗∗∗

(1.51) (1.61) (1.20) (1.50)F-value Can’t be estimated Can’t be estimated Can’t be estimated Can’t be

EstimatedR-squared 0.94 0.94 0.95 0.94

N = 434 zone years. Robust standard errors are reported in parentheses.Significance levels: ∗∗∗ p < 0.001, ∗∗ p < 0.01, ∗ p < 0.05, †p < .10 (two-tailed tests).a The value of community institutional origin and that of provincial capital city do not vary for a zone in this study period. Further,the value of municipality city only changed for one zone (the one located in Chongqing that was upgraded from a subprovincial cityto a municipality city in the study period). These three dummy variables thus are not included in the zone fixed-effects models.b The nonsignificant results related to regional community density and domain overlap are likely due to the fact that these variablesdid not vary substantially over time in this study. Thus, the effects of these variables are not distinguishable from the zone level fixedeffects.


Date post:	23-May-2020
Category:	Documents
Upload:	others
View:	2 times
Download:	0 times

INTERCOMMUNITY RELATIONSHIPS AND COMMUNITY …haiyang/SMJ zone 2.pdf1986, which formalized China’s...

Documents